[ckermit.git] / ckc299.txt

C-KERMIT CHANGE LOG (Changes since 8.0.207 / K95 2.1.3 January 2003)

  Chronological order.
  Go to the bottom to find the newest edits.

  F. da Cruz, The Kermit Project, Columbia University, NYC.

FTP USER, FTP ACCOUNT, plus the various prompts and switches for FTP username,
password, and account all neglected to strip quotes, and in most cases quotes
are necessary to specify a username that contains spaces.  ckcftp.c,
15 Jan 2003.

FTP MPUT f1 f2 f3... gets a parse error if any of the fn's do not match an
existing file.  This is bad for scripts.  In doftpput(), cmfdb() looks for
keywords (switches) or CMIFI.  When it hits CMIFI, it exits from the initial
parse loop and then does additional cmifi()s in a loop until done.  The most
obvious fix is to parse each field with cmfdb(CMIFI,CMFLD), i.e. fall back to
CMFLD if CMIFI doesn't match anything.  Then if CMFLD was used, we don't add
the filespec to the list.  This is a rather big change but it seems to work.
No error messages or failures happen for non-matching fields, but an error
message is printed (and the MPUT command fails) if none of the fields match
any files.  This fix got in too late for 2.1.3; workaround: use C-Shell
like wildcard list (ftp mput "{*.abc,foo.*}").  ckcftp.c, 16 Jan 2003.

GREP did not pass its pattern through the expander, thus variables could
not be used for patterns.  This must have been an oversight -- I can't find
anything in my notes about it.  Fixed in dogrep(): ckuus6.c, 24 Jan 2003.

New makefile target for HP-UX 11.xx with OpenSSL from Tapani Tarvainen.
makefile, 31 Jan 2003.

From Jeff:
 . Avoid core dump when dereferencing tnc_get_signature(): ckuus4.c.
 . Bump version numbers to 8.0.208, 2.1.4: ckcmai.c.

Added /NOLOGIN to FTP [OPEN].  ckcftp.c, 10 Feb 2003.

Don't dump core if FTP DEBUG is ON and FTP OPEN does not include a service.
openftp(): ckcftp.c, 10 Feb 2003.

HELP PATTERN text incorrectly identified commands and functions with
floating and anchored patterns.  The corrected lists are:
Floating: GREP, TYPE /MATCH:, /EXCEPT: patterns, \farraylook(),
Anchored: IF MATCH, file-matching wildcards, \fsearch(), \frsearch()
ckuus2.c, 10 Feb 2003.     

INPUT n \fpattern(xxx) did not work for case-independent comparisons.
Fixed in doinput(): ckuus4.c, 10 Feb 2003.

It seems \fpattern() didn't work with MINPUT at all.  There was no code to
handle \fpattern() in the MINPUT parse loop, so it never worked.  The code
had to be totally rewritten to use cmfld() in a loop, rather than cmtxt()
and then cksplit().  Furthermore, whenever any of the fields was an
\fjoin(), this had to be split.  ckuusr.c, 10 Feb 2003.

Macro replacement via \m() and \fdefinition() does not work as advertised
(i.e. case sensitively) for associative array elements; e.g. \m(xxx<abc>) is
treated the same as \m(xxx<ABC>), contrary to section 7.10.10 of the C-Kermit
7.0 update notes, and to the fact that the two really do exist separately.
Fixed by adding a static function isaarray(s) which succeeds if s is an
associative array reference and fails otherwise, and then having \m()
and \fdef() call mxxlook() (case-sensitive lookup) if isaarray(), otherwise
(as before) mxlook()).  ckuus4.c, 11 Feb 2003.

Fixed FTP OPEN to allow the /USER switch to override SET FTP AUTOLOGIN OFF,
just as /NOLOGIN overrides SET FTP AUTOLOGIN ON.  ckcftp.c, 11 Feb 2003.

In K95, "set key \1234 \27H" (any SET KEY command in which the first char of
the definition was backslash, and the ONLY character after the backslash
quantity was an uppercase letter, that letter would be lowercased).  Diagnosis:
xlookup() poking its argument (see notes from July 2000).  Jeff sent a fix.
ckucmd.c, 15 Feb 2003.

Ran my S-Expression torture test to make sure Sexps still worked.  They do,
except the bitwise & and | operators were broken, e.g. (& 7 2) and (| 1 2 4)
get "Invalid operand" errors.  Jeff's code had added an early failure return
from the lookup loop when when a single-byte keyword matched a keyword that
started with the same byte but was more than one byte long.  So "&" would hit
"&&" and fail instead of continuing its search (xlookup tables aren't sorted
so there can be no early return).  Fixed in xlookup(): ckucmd.c, 16 Feb 2003.

Got rid of "krbmit" target from makefile.  It's still there, but we don't
use it any more.  All secure targets now use "xermit", and produce a binary
called wermit, just like the regular ones do (except the old ckucon.c ones).
Non-secure targets, since they don't define any of the security symbols,
wind up compiling and linking to (mostly) empty security modules.  makefile,
15 Feb 2003.

Added \fcvtdate(xxx,3) to format its result in MDTM format (yyyymmddhhmmss,
all numeric, no spaces or punctuation).  Of course these numeric strings
are too big to be 32-bit numbers and are useless for arithmetic, but they're
useful for lexical comparison, etc.  ckuus[24].c, 16 Feb 2003.

The following FTP commands did not set FAILURE when they failed: RMDIR,
CD, CDUP, Fixed in the corresponding doftpblah() routines.  ckcftp.c,
16 Feb 2003.

RENAME would sometimes not print an error message when it failed, e.g. in K95
when the destination file already existed.  ckuus6.c, 17 Feb 2003.

Fixed COPY error messages, which did not come out in standard format when
/LIST was not included.  ckuus6.c, 17 Feb 2003.

Fixed #ifdefs in ck_crp.c to allow nonsecure builds on old platforms like
System V/68 R3.  19 Feb 2003.

Similar treatment for ck_ssl.c.  20 Feb 2003.

From Jeff, 21 Feb 2003:
 . AIX53 and AIX52 symbols for ckcdeb.h, makefile.
 . New gcc targets for various AIX 4.x/5.x versions: makefile.
 . Copyright date updates: ck_crp.c, ck_ssl.c.
 . ENABLE/DISABLE QUERY broken because keyword table out of order: ckuusr.c.
 . Fixed the use of HTTP proxies for HTTP [RE]OPEN for Unix: ckcnet.c.

Also for K95 only: Allow file transfer when K95 is invoked on the remote end
of a connection to a Pragma Systems Terminal Server connection; automatically
SET EXIT HANGUP OFF when invoked with open port handle ("k95 -l nnnn").

"cd a*" failed even when "a*" matched only one directory.  Fixed in cmifi():
ckucmd.c, 21 Feb 2003.

In the Unix version, replace "extern int errno;" with "#include <errno.h>"
if __GLIBC__ is defined, since glibc now defines a thread-specific errno.
ckcdeb.h, 26 Feb 2003.

Added #ifdefs to skip compilation of ckuath.c in nonsecure builds.  Tested
by building both secure and regular versions in Linux.  ckuath.c, 26 Feb 2003.

Ran the build-in-84-different-configurations script on Linux to make sure it
still builds with all different combinations of feature selection options.
All OK.  26 Feb 2003.

Built on VMS.  Needed to add a prototype for mxxlook*() to ckuusr.h; built
OK otherwise.  26 Feb 2003.

From Jeff: More #ifdef shuffling for nonsecure builds: ckuath.c, ck_ssl.c,
27 Feb 2003.

Added code to ensure \v(download) ends in a directory separator in Unix,
Windows, and OS/2.  ckuus7.c, 27 Feb 2003.

Added code to K95 zfnqfp() to tack on directory separator when returning
a directory name.  ckofio.c, 27 Feb 2003.

Somehow an old copy of ckuath.c popped to replace the new one.  Put the new
one back.  28 Feb 2003.

From Jeff: Fix typo in my K95 zfnqfp() code from yesterday; fixes for handling
UNCs uniformly, no matter which way their slashes are leaning.  ckofio.c,
28 Feb 2003.

At Jeff Mezei's suggestion, separate text and binary mode open sequences
for VMS session log.  ckvfio.c, 28 Feb 2003.

Added freebsd48 target for FreeBSD 4.8.  makefile, 1 Mar 2003.

Changed Mac OS X entries to include -DUSE_STRERROR.  makefile, 2 Mar 2003.

Fixed GETOK /GUI to evaluate its text argument.  ckuus6.c, 3 Mar 2003.

Jeff fixed the K95 Dialer QUICK dialog to (a) allow templates, and (b) have
a Save-As option.  3 Mar 2003.

Jeff fixed a problem with the Xmodem-CRC checksum being crunched whenever
there was a retransmission.  7 Mar 2003.

Added target/banner for Tru64 5.1B.  makefile, ckuver.h, 5 Mar 2003.

In Unix, the zcopy() routine (used by the COPY command) reset the user's umask
to 0 for the remainder of the Kermit process lifetime.  The bug was in
ckufio.c 8.0.194, 24 Oct 2002, and is fixed in ckufio.c 8.0.195, 6 Mar 2003.
Of course this happened after building 155 C-Kermit 8.0.208 binaries.  (But
before officially releasing 8.0.208.)

In the VMS version, changed:

        while ((n--) && xx_inc(2) > -1) ;
to:
        while ((n--) && xx_inc(2) >= 0) ;

to suppress the "...is being compared with a relational operator to a constant
whose value is not greater than zero" warning.  ckvtio.c, 7 Mar 2002.

Added a debug call to dologend in hopes of catching overzealous Locus
switching, which seems to happen only in K95.  ckuus3.c, 7 Mar 2002.

Rebuilt binaries for some of the more current Unix releases: AIX 4.3.3-5.1,
Solaris 7-9 , Red Hat 7.0-8.0, Slackware 8.1, Freebsd 4.7-4.8, NetBSD 1.6,
OpenBSD 3.2, Unixware 7.1.3, Open Unix 8, OSR5.0.6a, etc.  A Unix binary with
COPY umask fix shows a 6 Mar 2003 date for "UNIX File support" in SHOW
VERSIONS; a binary without the fix shows 24 Oct 2002.

C-Kermit 8.0.208 dated 14 March 2003 released on 10 March 2003.

---8.0.208---

From Jeff 13 Mar 2003:
 . Updated SSL module allows importation of tickets from host.
 . freebsd50+openssl target: makefile.
 . FTP PUT /PERMISSIONS error message for K95: ckcftp.c.
 
Fixed MINPUT to strip quotes or braces from around targets (this was broken
on Feb 10th).  Thanks to Jason Heskett for discovering and reporting this
(killer) bug.  ckuusr.c, 14 Mar 2003.

Changed version number to 209 Dev.00.  ckcmai.c, 14 Mar 2003.

While debugging the alphapage script, I found that the command "minput 8 \6\13
\21\13 \13\27\4\13 \30\13" gets "?Not confirmed" in 8.0.208 and 8.0.209, but
not in 206 and earlier.  This problem too was introduced on Feb 10th by
changing MINPUT parsing from cmtxt() followed by cksplit() to cmfld() in a
loop.  cmfld() uses setatm() to return its result and of course setatm()
breaks on \13.  Changing setatm() not to do this would break everything else.
But cmfld() has no arguments that let us tell it to do anything different in
this case.  Changing the API would be a disaster.  The only solution is to add
an "MINPUT ACTIVE" (minputactive) global variable that tells cmfld() to tell
setatm() not to break on CR.  Now MINPUT with braced targets containing CR
and/or LF works in 209, 206, and 201 (but not 208).  ckucmd.c, ckuusr.c,
ckuus5.c, 15 Mar 2003.

MINPUT n \fjoin(&a) works OK if all the members of \&a[] are text strings, but
if they are strings of control chars (as above), they don't get separated by
the spaces.  For example in:

  dcl \&a[] = "\4\5" "\6\7" xxx
  minput 10 \fjoin(&a)

MINPUT gets two targets: "aaa" and "\4\5 \6\7 xxx".  The bug was in the
cksplit() call in the \fjoin() case of MINPUT: it needed to specify an
include set consisting of all the control characters except NUL.  ckuusr.c,
16 Mar 2003.

But there's still a problem:

  dcl \&a[] = "\4\5\13\10" "\6\7" "xxx"

creates an array whose first member is "^D^E (one doublequote included).  But
if braces are used instead, there's no problem.  Same deal as MINPUT: cmfld()
breaks on CR or LF, thus the end quote is lost.  If I set minputactive for
DECLARE initializers too, that fixes it.  Is there any reason not to do this?
Can't think of any (famous last words)...  ckuusr.c, 16 Mar 2003.

Since it has multiple applications, changed the flag's name from minputactive
to keepallchars.  ckucmd.c, ckuus[r5].c, 16 Mar 2003.

\v(exedir) wasn't being set correctly (it included the program name as well
as the directory).  Fixed in getexedir(): ckuus4.c, 16 Mar 2003.

SET CARRIER-WATCH <Esc> "auto matic" (spurious space in supplied keyword).
Cosmetic only; it still worked.  Fixed in setdcd(): ckuus3.c, 16 Mar 2003.

"directory a b c" listed too many files -- all files whose names END WITH a,
b, or c, rather than the files whose names WERE a, b, or c.  Diagnosis: The
filespec is changed into a pattern: {a,b,c}, which is the correct form.  It is
passed to nzxpand(), which goes through the directory getting filenames and
sending each one to ckmatch() with the given pattern.  ckmatch() receives the
correct pattern but then prepends a "*" -- that's not right.  It's not just
in filename matching either.  The following succeeds when it shouldn't:

  if match xxxxc {{a,b,c}} <command>

Changing ckmatch() to not prepend the "*" to each segment fixes the command
above but breaks lots of others.  Running through the "match" torture-test
script shows the problem occurs only when the {a,b,c} list is the entire
pattern, and not embedded within a larger pattern.  Testing for this case
fixed the problem.  ckmatch(): ckclib.c, 16 Mar 2003.

Fixed FTP MODTIME to not print anything if QUIET ON.  ckcftp.c, 16 Mar 2003.

Picked up a new ckuath.c from Jeff, not sure what the changes are. 16 Mar 2003.

Did a few regular and secure builds to make sure I didn't wreck anything.

Changed version number to 209 (final).  ckcmai.c, 16 Mar 2003.

Jason Heskett found another bug: if you define a macro FOO inside the
definition of another macro BAR, and FOO's definition includes an odd number
of doublequotes (such as 1), FOO's definition absorbs the rest of BAR's
definition.  Example:

  def TEST {
    .foo = {X"}
    sho mac foo
  }
  do test
  sho mac foo

Results in:

  foo = {X"}, sho mac foo

Diagnosis: the TEST definition becomes:

  def TEST .foo = {X"}, sho mac foo

and the macro reader is erroneously treating the doublequote as an open
quote, and then automatically closes the quote at the end of the definition.
The error is that a doublequote should be significant only at the beginning of
a field.  But the macro reader isn't a command parser; it doesn't know what
a field is -- it's just looking for commas and skipping over quoted ones.
First we have to fix an oversight: SET COMMAND DOUBLEQUOTING OFF should have
worked here, but it wasn't tested in this case.  Fixed in getncm(): ckuus5.c,
17 Mar 2003.

There are only certain cases where it makes sense to treat doublequotes as
signicant:

 . An open quote must be at the beginning or preceded by a space.
 . A close quote is only at the end or else followed by a space.

This too was fixed in getncm(): ckuus5.c, 17 Mar 2003.

A fix from Jeff SSL/TLS FTP data decoding.  ckcftp.c, 18 Mar 2003.

Tried building C-Kermit on a Cray Y-MP with UNICOS 9.0.  "int suspend",
declared in ckcmai.c and used in many modules, conflicts with:

  unistd.h:extern int suspend __((int _Category, int _Id));

The "=Dsuspend=xsuspend" trick doesn't work for this; there is no way around
the conflict other than to rename the variable: ckcmai.c, ckutio.c,
ckuus[35xy].c.  26 Mar 2003.  VMS and K95 not affected.

OK that gets us past ckcmai.c...  Then in ckutio.c I had to add a new #ifdef
around the LFDEVNO setting, because the Cray didn't have mkdev.h.  Could not
find a Cray-specific manifest symbol, so I made a new makefile target (cray9)
that sets this symbol.  Having done this I have no idea what kind of lockfile
would be created, but I also doubt if anybody dials out from a Cray.  The
binary should run a C90, J90, or Y-MP.  makefile, 26 Mar 2003.

Added a target for SCO OSR5.0.7.  makefile, ckuver.h, 30 Mar 2003.

Changed since 208:
makefile ckuver.h ckcmai.c ckclib.c ckcftp.c ckucmd.c ckuus*.c ckutio.c.

---8.0.209---

From Mark Sapiro, a fix for the March 17th doubleqote fix, getncm(): ckuus5.c,
4 Apr 2003.

From Jeff, 29 Apr 2003:
 . Corrected target for HP-UX 11.00 + OpenSSL: makefile, 
 . Do not allow WILL AUTH before WONT START_TLS: ckctel.h ckctel.c
 . Add hooks for SFTP and SET/SHOW SFTP: ckcdeb.h ckuusr.h ckuusr.c ckuus3.c
 . Add SKERMIT ckuusr.h ckuusr.c
 . Add ADM-5 terminal emulation: ckuus7.c, ckuus5.c
 . Uncomment and update HELP SET SSH V2 AUTO-REKEY: ckuus2.c
 . Enable IF TERMINAL-MACRO and IF STARTED-FROM-DIALER for C-Kermit: ckuus6.c
 . Fix conflicting NOSCROLL keyword definition: ckuusr.h
 . Set ttname when I_AM_SSH: ckuusy.c
 . Add extended arg parsing for SSH, Rlogin, Telnet: ckuusy.c, ckuus4.c
 . Security updates: ckuath.c, ck_ssl.c
 . Change K95 version number to 2.2.0: ckcmai.c
 . Save K95 term i/o state before executing keyboard macro: ckuus4.c
 . Add tests for SSH Subsystem active during INPUT/OUTPUT/CONNECT: ckuus[45].c
 . Enable K95 SET SSH V2 AUTO-REKEY: ckuus3.c

SFTP and SET SFTP subcommands are implemented up to the case statements.

Files of mine that Jeff hadn't picked up:
  ckuver.h ckcftp.c ckutio.c ckuusx.c (just minor changes for last build-all)

On 4 Jan 2003, SET RECEIVE MOVE-TO was changed to convert is argument to an
absolute path, which made it impossible to specify a relative path, then
move to different directories and have it apply relatively to each directory.
Changed this as follows:

 . Parser uses cmtxt() rather than cmdir() so it won't fail at parse time.
 . If path is absolute, we fail at parse time if directory doesn't exist.
 . In reof() we run the the path through xxstring (again, in case deferred
   evaluation of variables is desired) and then, if not null, use it.
 . If the directory doesn't exist, rename() fails and reof() returns -4,
   resulting in a protocol error (this is not a change).  We do NOT create
   the directory on the fly. 

I also fixed SET SEND/RECEIVE RENAME-TO to parse with cmtxt() rather than
cmdir(), since it's parsing a text template, not a directory name, e.g.
"set receive rename-to file-\v(time)-v(date)-\v(pid)".  This was totally
broken, since when I don't know.  We don't call xxstring() in this parse, so
evaluation is always deferred -- I'd better not change this.  ckuus7.c,
ckcfns.c, 1 May 2003.

From Jeff, Sat May  3 14:15:23 2003:
 . Pick up the right isascii definition for K95: ckctel.c
 . malloc...  ckuath.c (new safe malloc routines for K95)
 . Add author listing: ckuus5.c
 . SSH Heartbeat support (K95 only): ckuus[23].c
 . Prescan --height and --width to avoid window resizing at startup: ckuusy.c
 . Add checks for fatal() or doexit() called from sysinit(): ckuusx.c
 . Move some K95-specific definitions to ckoker.h: ckcdeb.h
 . Add support for ON_CD macro in zchdir(): ckufio.c
 . Add a command to let FTP client authenticate with SSLv2: ckcftp.c
 . Fix parsing of FTP file facts like "UNIX.mode": ckcftp.c

ON_CD will need some explaining (to be done).  It's implemented for Unix,
VMS, WIndows, and OS/2.

The FTP file facts fix came from first exposure to the new OpenBSD FTP
server: ftp://ftp7.usa.openbsd.org/pub/os/OpenBSD/3.3/i386/
The period in "UNIX.mode" caused an erroneous word break, adding junk to
the filename.

About the malloc changes, Jeff says "K95 is not behaving well in low memory
environments.  I'm not sure that C-Kermit does much better.  The program does
not crash but it certainly does not behave the way the user expects it to.
I'm beginning to think that any malloc() error should be treated as fatal."

Not visible in these changes because it's in K95-specific modules: Jeff made
SET ATTRIBUTES OFF and SET ATTRIBUTES DATE OFF apply to XYZMODEM transfers.

From Jeff, 11 May 2003:
 . Add support for SSH Keepalive to relevant SET command (K95): ckuus3.c
 . Reduce max overlapped i/o requests from 30 to 7 (K95): ckuus7.c
 . Don't call sysinit() in fatal(): ckuusx.c.
 . Some new conditionalizations for SSL module: ck_ssl.c

The doublequote-parsing fixes from March and April broke the SWITCH statement,
which is implemented by internally defining, then executing, a macro.  If I
drop back to the old dumb handling of doublequotes, everything is fixed except
the problem of March 17th.  But can we really expect getncm() to pre-guess
what the parser is going to do?  getncm()'s only job is to find command
boundaries, which are represented by commas.  Commas, however, is needed IN
commands too.  We take a comma literally if it is quoted with \, or is inside
a matched pair of braces, parens, or doublequotes.  It is not unreasonable to
require a doublequote in a macro definition to be prefixed by \ when it is to
be taken literally.  The proper response to Jason Heskett's complaint of March
17th should have been to leave the code alone and recommand an appropriate
form of quoting:

  def TEST {
      .foo = {X\"}
      sho mac foo
  }

And this is what I have done.  Another reason for sticking with the old method
is that it's explainable.  The "improved" method, even if it worked, would be
be impossible to explain.  Btw, in testing this I noticed that the switch-test
script made 8.0.201 dump core.  Today's version is fine.  The problem with
quoted strings inside of IF {...} clauses and FOR and WHILE loops is fixed
too.  Perhaps "unbroken" would be a better word.  ckuus5.c, 11 May 2003.

Vace discovered that FTP MGET /EXCEPT:{... (with an unterminated /EXCEPT list)
could crash Kermit.  Fixed in ckcftp.c, 11 May 2003.

CONTINUE should not affect SUCCESS/FAILURE status.  ckuusr.c, 11 May 2003.

Fixed an oversight that goes back 15 years.  While \{123} is allowed for
decimal codes, \x{12} and \o{123} were never handled.  ckucmd.c, 11 May 2003.

Added support for Red Hat <baudboy.h> and /usr/sbin/lockdev.  Supposedly this
allows Kermit to be installed without setuid or setgid bits and still be able
to lock and use the serial device.  Compiles and starts, but not tested.
ckcdeb.h, makefile, ckutio.c, ckuus5.c, 16 May 2003.

From Jeff: FTP ASCII send data to host when FTP /SSL was in use was broken.
ftp_dpl is set to Clear when FTP /SSL is in use.  This was causing the data to
be written to the socket with send() instead of the OpenSSL routines.
ckcftp.c, ckuath.c, 21 May 2003.

From Jeff: Stuff for Kerberos 524: ckcdeb.h.  Fixes for FTP; "FTP ASCII send
data did not properly compute the end of line translations.  On Unix (and
similar platforms) the end of line was correct for no character sets but
incorrect when character sets were specified.  On Windows/OS2, the end of line
was correct when character sets were specified and incorrect when they were
not.  On MAC, both were broken.  Also, FTP Send Byte counts were incorrect
when character sets were specified."  ckcftp.c.  17 Jun 2003.

From Jeff: fixes to HTTP /AGENT: and /USER: switch action: ckcnet.c ckuus3.c
ck_crp.c ckcftp.c ckuus2.c ckuusy.c ckuusr.c ckcnet.h, 21 Jun 2003.

From Jeff: Fix SET DIALER BACKSPACE so it can override a previous SET KEY
(e.g. from INI file): ckuus7.c.  Some SSL/TLS updates: ck_ssl.c.  HTTP support
for VMS and other VMS improvements (e.g. a way to not have to hardwire the
C-Kermit version number into the build script) from Martin Vorlaender:
ckcnet.h, ckuus[r3].c, ckcdeb.h, ckvtio.c, ckcnet.c, ckvker.com.  Built on
Solaris (gcc/ansi) and SunOS (cc/k&r).  The new VMS script tests the VMS
version and includes HTTP support only for VMS 6.2 or later.  2 Jul 2003.

Tried to build on our last VMS system but it seems to be dead.  Looks like a
head crash (makes really loud noises, boot says DKA0 not recognized) (fooey, I
just paid good money to renew the VMS license).  Tried building at another
site with:

  Process Software MultiNet V4.3 Rev A-X,
  Compaq AlphaServer ES40, OpenVMS AXP V7.3
  Compaq C V6.4-008 on OpenVMS Alpha V7.3

Had to make a few corrections to ckvker.com.  But still, compilation of
ckcnet.c bombs, indicating that the SELECT definition somehow got lost
somewhere since the 209 release (i.e. no SELECT type is defined so it falls
thru to "SELECT is required for this code").  But I don't see anything in
ckcdeb.h or ckcnet.[ch] that would explain this.  Not ckvker.com either
(putting the old one back gives the same result).  OK, I give up, maybe it's
just that I haven't tried building it on MultiNet recently.  What about UCX?
Aha, builds fine there except for warnings about mlook, dodo, and parser in
ckvfio.c (because of ON_CD) -- I suppose I have #include <ckucmd.h>... (done)
Anyhow it builds OK and the HTTP code is active and almost works (HTTP OPEN
works; HTTP GET seems to succeed but creates an empty file every time).  Tried
building under MultiNet at another installation; same bad result.

OK so why won't it build for MultiNet?  Comparing ckcnet.c with the 209
version, not a single #ifdef or #include is changed.  Tried building with
p3="NOHTTP" -- builds OK, aha.  Where's the problem?  Not ckcnet.h...
Not ckcdeb.h...  OK I give up, will revisit this next time I get time to
do anything with the code.

Later Jeff said "Martin did not implement VMS networking for the HTTP code.
All he did was activate the #define HTTP which happens to work because his
connections are using SSL/TLS connections.  http_inc(), http_tol(), etc have
no support for VMS networking regardless of whether it is UCX or MULTINET.
The vast majority of HTTP connections are not secured by SSL/TLS.  It makes no
sense to support HTTP on VMS until someone is willing to either do the work or
pay have the work done to implement VMS networking in that code base."  So the
fix is to not enable HTTP for VMS after all.  Removed the CKHTTP definition
for VMS from ckcdeb.h, 6 Jul 2003.

Fixed ckvfio.c to #include <ckuusr.h> (instead of <ckucmd.h>) to pick up 
missing prototypes.  6 Jul 2003.

From Arthur Marsh: solaris2xg+openssl+zlib+srp+pam+shadow and the corresponding
Solaris 7 target.  makefile, 6 Jul 2003.

Remove duplicate #includes for <sys/stat.h>, <errno.h>, and <ctype.h> from
ckcftp.c.  6 Jul 2003.

Add -DUSE_MEMCPY to Motorola SV/68 targets because of shuffled #includes in 
ckcftp.c.  8 Jul 2003.

From Jeff: Fix problems mixing SSL and SRP without Kerberos.  Plus a few minor
#define comment changes and a reshuffling of #defines in ckcdeb.h to allow me
to build on X86 Windows without Kerberos.  ckcdeb.h, ck_crp.c, ckuath.c,
10 Jul 2003.

From Jeff: updated ckuat2.h and ckuath.c, 29 Jul 2003.

Mats Peterson noticed that a very small Latin-1 file would be incorrectly
identified as UCS-2 by scanfile().  Fixed in ckuusx.c, 29 Jul 2003.

Fixed ACCESS macro definition to account for the fact that FIND is now a
built-in command.  ckermit.ini, 30 Jul 2003.

From Jeff: Fix for typo in urlparse() (svc/hos): ckuusy.c, 18 Aug 2003.

From Jeff: Redhat9 makefile targets (needed for for OpenSSL 0.9.7):
makefile, 19 Aug 2003.

GREP /NOLIST and /COUNT did too much magic, with some undesirable fallout:
"GREP /NOLIST /COUNT:x args" printed "file:count" for each file.  "GREP
/COUNT:x /NOLIST args" did not print "file:count", but neither did it set the
count variable.  Removed the magic.  Also one of the GREP switches,
/LINENUMBERS, was out of order.  Fixed in ckuus6.c, 20 Aug 2003.

From Jeff: "Reorganizing code to enable building with different subsets of
options; a few typos corrected as well."  ckcdeb.h, ckuver.h (for RH9),
ckcnet.c, ckuus7.c, ckuus3.c: 24 Aug 2003.

Scanfile misidentified a big PDF file as text because the first 800K of it
*was* text (most other PDF files were correctly tagged as binary).  Fixed
by adding a check for the PDF signature at the beginning of the file.
scanfile(): ckuusx.c, 25 Aug 2003.

Ditto for PostScript files, but conservatively.  Signature at beginning of
file must begin with "%!PS-Ado".  If it's just "%!" (or something nonstandard
like "%%Creator: Windows PSCRIPT") we do a regular scan.  Also added "*.ps"
to all binary filename patterns.  ckuusx.c, 4 Sep 2003.

Ditto (but within #ifndef NOPCLSCAN) for PCL (<ESC>E) and PJL (<ESC>%) files,
but no binpatterns (note: ".PCL" is the extension for TOPS-20 EXEC scripts).
ckuusx.c, 4 Sep 2003.

Added comments about OpenSSL 0.9.7 to all linux+openssl targets.
makefile, 4 Sep 2003.

From Jeff: Added - #define ALLOW_KRB_3DES_ENCRYPT.  When this symbol is defined
at compilation Kermit will allow non-DES session keys to be used during Telnet
Auth.  These session keys can then be used for Telnet Encrypt.  The reason
this is not compiled on by default is that the MIT Kerberos Telnet does not
follow the RFC for constructing keys for ENCRYPT DES when the keys are longer
than 8 bytes in length.  ckuath.c, ckuus5.c, 4 Sep 2003.

"ftp mget a b c" succeeded if one or more of the files did not exist, even
with "set ftp error-action proceed".  This is because the server's NLST file
list does not include any files that don't exist, so the client never even
tries to get them.  Fortunately, the way the code is structured, this one was
easy to fix.  ckcftp.c, 14 Sep 2003.

From Jeff: Corrected code in ckcnet.c to ensure that Reverse DNS Lookups are
not performed if tcp_rdns is OFF.  Fixed ck_krb5_getrealm() to actually return
the realm of the credentials cache and not the default realm specified in the
krb5.conf file.  Previously krb5_cc_get_principal() was not being called.
Fixed ck_krb5_is_tgt_valid() to test the TGT in the current ccache and not the
TGT constructed from the default realm.  ckcnet.c, ckuath.c, 14 Sep 2003.

Marco Bernardi noticed that IF DIRECTORY could produce a false positive if
the argument directory had previously been referenced but then removed.  This
is because of the clever isdir() cache that was added to speed up recursion
through big directory trees.  Changed IF DIRECTORY to make a second check
(definitive but more expensive) if isdir() succeeds, and changed the
directory-deleting routine, ckmkdir(), to flush the directory cache (UNIX
only -- this also should be done in K95 but it's not critical).  This was
done by adding a routine, clrdircache() to ckufio.c, which sets prevstat
to -1 and prevpath[0] to NUL.  ckcfn3.c, ckuus6.c, ckufio.c, 18 Sep 2003.

Marco reported the second fix still didn't work for him (even though it did
for me).  Rather than try to figure out why, I concluded that the directory
cache is just not safe: a directory found a second ago might have been deleted
or renamed not only by Kermit but by some other process.  Why did I add this
in the first place?  The log says:

  Some debug logs showed that isdir() is often called twice in a row on the
  same file.  Rather than try to sort out clients, I added a 1-element cache
  to Unix isdir().  ckufio.c, 24 Apr 2000.

Experimentation with DIR and DIR /RECURSIVE does not show this happening at
all.  So I #ifdef'd out the directory cache (see #ifdef ISDIRCACHE in ckufio.c;
ISDIRCACHE is not defined) and backed off the previous changes: ckufio.c,
ckcfn3.c, ckuus6.c, 28 Sep 2003.

From Jeff: Replace the compile time ALLOW_KRB_3DES_ENCRYPT with a run-time
command SET TELNET BUG AUTH-KRB5-DES which defaults to ON: ckctel.[ch],
ckuus[234].c, ck_crp.c, ckuath.c.  4 Oct 2003.

Allow DIAL RETRIES to be any positive number, and catch negative ones.
Also added code to check for atoi() errors (e.g. truncation).  At least on
some platforms (e.g. Solaris) atoi() is supposed to set errno, but it
doesn't.  ckuus3.c, ckucmd.c, 4 Oct 2003.

Added /DEFAULT: to ASK-class commands (ASK, ASKQ, GETOK):

 . For popups: no way to send defaults to popup_readtext() or popup_readpass().
 . For GUI ASK[Q], pass default to gui_txt_dialog().
 . For GUI GETOK, convert "yes" "ok" or "no" default to number for uq_ok().
 . For Text GETOK, add default to cmkey().
 . For Text ASK[Q], add default to cmtxt().
 . For GETC, GETKEY, and READ: no changes.

GETOK, ASK, and ASKQ with /TIMEOUT: no longer fail when the timer goes off
if a /DEFAULT was supplied.  The GUI functions (uq_blah) don't seem to
support timeouts.  Only the text version has been tested.  ckuus[26].c,
4 Oct 2003.

From Jeff: add /DEFAULT: for popups.  ckuus6.c. 6 Oct 2003.

Change SET DIAL INTERVAL to be like SET DIAL RETRIES.  ckuus[34].c, 6 Oct 2003.

Added target for HP-UX 10/11 + OpenSSL built with gcc, from Chris Cheney.
Makefile, 12 Oct 2003.

From Jeff, 6 Nov 2003:
 . #ifdef adjustments: ckcftp.c, ckcdeb.h
 . Fix spurious consumption of first byte(s) on Telnet connection: ckctel.c
 . Another HP PJL test for scanfile: ckuusx.c.
 . K95: Recognize DG4xx protected fields in DG2xx emulation: ckuus7.c.
 . Add SSLeay version display to SHOW AUTH command: ckuus7.c
 . Improved SET MOUSE CLEAR help text: ckuus2.c.
 . Improved Kverbs help text: ckuus2.c (+ new IBM-3151 Kverbs).
 . Some changes to ck_ssl.c, ckuath.c.

From PeterE, 10 Nov 2003:
 . Improved HP-UX 10/11 makefile targets for OpenSSL.
 . #ifdef fix for OpenSSL on HP-UX: ck_ssl.c.

Another new makefile from PeterE with improved and integrated HP-UX targets.
12 Nov 2003.

A couple fixes to the solaris9g+krb5+krb4+openssl+shadow+pam+zlib target
from Jeff.  Added a solaris9g+openssl+shadow+pam+zlib target.  makefile,
21 Nov 2003.

From Jeff, 30 Nov 2003:
 . Fix SEND /MOVE-TO: ckuusr.c.
 . Fix K95 SET TITLE to allow quotes/braces around text: ckuus7.c.
 . Improved "set term autodownload ?" response: ckuus5.c.
 . Fix SHOW FEATURES to specify the protocol for encryption: ckuus5.c
 . Make {SEND, RECEIVE} {MOVE-TO, RENAME-TO} work for XYZMODEM (K95 only).

From Jeff: 7 Jan 2004:
 . At one point Frank started to add a timer parameter to the
   uq_txt() function but he only did it for the non-ANSI
   compilers.  I added it for the ANSI compilers, fixed the
   prototypes and provided a default value easily changed
   DEFAULT_UQ_TIMEOUT: ckcker.h, ckuus[36].c, ck_ssl.c, ckcftp.c, ckuath.c.
 . Fixed SET TERMINAL DEBUG ON (typo in variable name): ckuus7.c.
 . Fixed BEEP INFORMATION; previously it made no sound, now uses
   MB_ICONQUESTION.  ckuusx.c.

From Ian Beckwith <ian@nessie.mcc.ac.uk> (Debianization), 7 Jan 2004:
 . Search dir/ckermit for docs, as well as dir/kermit in cmdini(): ckuus5.c.
 . New linux+krb5+krb4+openssl+shadow+pam target (kitchen sink minus SRP,
   which Debian does not distribute): makefile.
 ? Mangles the DESTDIR support in makefile to install into a staging area:
   makefile (I didn't take this one yet).

Updated copyright notices for 2004, all modules.  7 Jan 2004.

Added INPUT /NOMATCH, allowing INPUT to be used for a fixed amount of time
without attempting to match any text or patterns, so it's no longer
necessary to "input 600 STRING_THAT_WILL_NEVER_COME".  If /NOMATCH is
included, INPUT succeeds if the timeout expires, with \v(instatus) = 1
(meaning "timed out"); fails upon interruption or i/o error.  ckuusr.h,
ckuus[r24].c, 7 Jan 2004.

Added SET INPUT SCALE-FACTOR <float>.  This scales all INPUT timeouts by the
given factor, allowing time-sensitive scripts to be adjusted to changing
conditions such as congested networks or different-speed modems without
having to change each INPUT-class command.  This affects only those timeouts
that are given in seconds, not as wall-clock times.  Although the scale
factor can have a fractional part, the INPUT timeout is still an integer.
Added this to SHOW INPUT, and added a \v(inscale) variable for it.
ckuusr.h, ckuus[r257].c, 7 Jan 2004.

undef \%a, \fverify(abc,\%a) returns 0, which makes it look as if \%a is a
string composed of a's, b's, and/or c's, when in fact it contains nothing.
Changed \fverify() to return -1 in this case.  ckuus4.c, 12 Jan 2004.

\fcode(xxx) returned an empty string if its argument string was empty.  This
makes it unsafe to use in arithmetic or boolean expressions.  Changed it to
return 0 if its argument was missing, null, or empty.  ckuus4.c, 12 Jan 2004.

Updated \verify() and \fcode() help text.  ckuus2.c, 12 Jan 2004.

While setting up IKSD, Ian Beckwith noticed that including the --initfile:
option caused Kermit to start parsing its own Copyright string as if it were
the command line, and eventually crash.  I couldn't reproduce on Solaris /
Sparc but I could in Linux / i386 (what Ian is using) -- a change from Jeff
on 28 Apr 2003 set the command-line arg pointer to a literal empty string in
prescan() about line 1740 of of ckuus4.c; the pointer is incremented next
time thru the loop, resulting in random memory being referenced.  Fixed by
setting the pointer to NULL instead of "".  ckuus4.c, 12 Jan 2004.

declare \&a[999999999999999] would dump core on some platforms.  atoi()
or whatever would truncate the dimension to maxint.  When we add 1 to the
result, we get a negative number, which is used as an index, loop test, etc.
Fixed both dodcl() and dclarray() to check for (n+1 < 0).  ckuus[r5].c,
12 Jan 2004.

Unix zchki() would fail on /dev/tty, which is unreasonable.  This prevented
FOPEN /READ from reading from the terminal.  zchki() already allowed for
/dev/null, so I added /dev/tty to the list of specials.  Ditto for FOPEN
/WRITE and zchko().  ckufio.c 13 Jan 2004.

Added untabify() routine to ckclib.[ch], 13 Jan 2004.
Added FREAD /TRIM and /UNTABIFY.  ckuus[27].c, 13 Jan 2004.
Added \funtabify().  ckuusr.h, ckuus[24].c, 13 Jan 2004.

Dat Nguyen noticed that (setq u 'p') followed by (u) dumped core.  This was
caused by an over-clever optimization that skipped mallocs for short
literals, but then went on later to try to free one that hadn't been
malloc'd.  Fixed in dosexp(): ckuus3.c, 14 Jan 2004.

Catch another copyright date.  ckuus5.c, 14 Jan 2004.

Fixed SWITCH to work even when SET COMMAND DOUBLEQUOTE OFF (from Mark
Sapiro).  ckuus5.c, 15 Jan 2004.

Changed version to 8.0.211 so scripts can test for recently added features.
ckcmai.c, 15 Jan 2004.

Fixed a glitch in K95 "help set port".  ckuus2.c, 20 Jan 2004.

Fix from Jeff: Connections to a TLS-aware protocol which require a reconnect
upon certificate verification failure could not reconnect if the connection
was initiated from the command line or via a URL.  ckctel.c ckcmai.c
ckuusr.c ckuus7.c ckuusy.c, 20 Jan 2004.

From Alex Lewin: makefile target and #ifdef for Mac OS X 10.3 (Panther):
makefile, ckcnet.c, 7 Feb 2004.

Added KFLAGS to sco32v507 targets to make PTY and SSH commands work.  The
same flags could probably also be added to earlier OSR5 targets but they
have not been tested there.  makefile, 7 Feb 2004.

Checked a complaint that "LOCAL &a" did not make array \&a[] local.  Indeed
it did not, and can not.  You have to use the full syntax in the LOCAL
command, "LOCAL \&a[]", or else it doesn't know it's not a macro named &a.
7 Feb 2004.

Fixed some confusion in creating IKSD database file and temp-file names.
I was calling zfnqfp() without remembering that the path member of the
returned struct included the filename, so to get just the directory name,
I needed to strip the filename from the right.  ckuusy.c, 2 Mar 2004.

New ckuath.c, ck_ssl.c from Jeff.  2 Mar 2004.

Updated Jeff's affiliation in VERSION command text.  ckuusr.c, 2 Mar 2004.

Designation changed from Dev.00 to Beta.01.  ckcmai.c, 2 Mar 2004.

Fixed zrename() syslogging -- it had success and failure reversed.
Beta.02: ckufio.c, 4 Mar 2004.

Problem: when accessing IKSD via a kermit:// or iksd:// URL, and a user ID
is given but no password, doxarg() set the password to "" instead of leaving
it NULL, but all the tests in dourl() are for NULL.  Fixed in doxarg():
ckuusy.c, 5 Mar 2004.

The logic in dourl() about which macro to construct (login and connect,
login and get directory listing, or login and fetch a file) was a bit off,
so all three cases were not handled.  ckcmai.c, 5 Mar 2004.

Trial Beta builds:
 . HP-UX B.11.11 PA-RISC
 . HP-UX B.11.23 IA64
 . Tru64 4.0G Alpha
 . Tru64 5.1B Alpha
 . Debian 3.0 i386
 . Red Hat ES 2.1 i386
 . Slackware 9.1 i386
 . VMS 7.3-1 Alpha + UCX 5.3
 . VMS 7.3-1 Alpha no TCP/IP
 . VMS 7.3 Alpha MultiNet 4.3 A-X
 . SCO UnixWare 7.1.4 i386
 . SCO OSR5.0.7 i386
 . Solaris 9 Sparc

Fixed compiler warning in doxarg() caused by typo (NULL instead of NUL) in
the 5 March doxarg() edit.  ckuusy.c, 9 Mar 2004.

IKSD (kermit://) command-line URLs did not work right if the client had
already preauthenticated with Kerberos or somesuch because they tried to log
in again with REMOTE LOGIN.  The macros constructed in doxarg() needed to
check \v(authstate) before attempting REMOTE LOGIN.  ckcmai.c, 10 Mar 2004.

Added ckuker.nr to x.sh (ckdaily upload) and updated ckuker.nr with current
version number and dates.  10 Mar 2004.

Replaced hardwired references to /usr/local in makefile with $(prefix)
(which defaults to /usr/local, but can be overridden on the command line),
suggested by Nelson Beebe for use with Configure.  10 Mar 2004.

From Nelson Beebe: In the Kermit makefile in the install target commands,
line 981 reads:

        cp $(BINARY) $(DESTDIR)$(BINDIR)/kermit || exit 1;\

Could you please add this line before it:

        rm -f $(DESTDIR)$(BINDIR)/kermit;\

Some sites (mine included) keep multiple versions of software around,
with hard links between $(prefix)/progname and $(prefix)/progname-x.y.z.
Failure to remove the $(prefix)/progname at "make install" time then
replaces the old $(prefix)/progname-x.y.z with the new one, destroying
an old version that the site wanted to be preserved.  makefile, 10 Mar 2004.

Minor syntax and typo fixes (mostly prototypes): ckcdeb.h, ckcfns.c,
ckclib.c, ckufio.c, ckuusr.h, ckuusx.c, 10 Mar 2004.  (I still have a few
more to do.)

Added CC=$(CC) CC2=$(CC2) to many (but not all) makefile targets that
reference other makefile targets.  On some platforms (notably AIX, Solaris,
SunOS) there are specific targets for different compilers, so I skipped
those.  makefile, 10 Mar 2004.

Added error checking to kermit:// URL macros, so they don't plow ahead
after the connection is closed.  ckcmai.c, 11 Mar 2004.

Added FreeBSD 4.9 and 5.1 targets (only the herald is affected).
makefile, ckuver.h, 11 Mar 2004.

Added "LIBS=-lcrypt" to bsd44 targets since nowadays crypt is almost always
unbundled from libc.  Also added explanatory notes.  makefile, 11 Mar 2004.

Changed MANDIR to default to $(manroot)/man/man1, and manroot to default
to $(prefix).  More adding of CC=$(CC) clauses: {Free,Net,Open}BSD, 4.4BSD.
makefile, 11 Mar 2004.

Miscellaneous cleanups: ckuusx.c, ckcnet.c, ckufio.c, 11 Mar 2004.

Corrected the check in the linux target to see if /usr/include/crypt.h
exists, and if so to define HAVE_CRYPT_H, which is used in ckcdeb.h to
#include <crypt.h> to get the prototype for crypt() and prevent bogus
conversions on its return type on 64-bit platforms (the previous test wasn't
quite right and the resulting symbol wasn't spelled right).  makefile,
12 Mar 2004.

From Jeff, 14 Mar 2004:
 . Initialize localuidbuf[] in tn_snenv(): ckctel.c.
 . Remove remote-mode checks in hupok() for K95G only (why?): ckuus3.c.
 . Add help text for new K95-only TYPE /GUI switches: ckuus2.c.
 . TYPE /GUI parsing, ...: ckuusr.c.
 . TYPE /GUI action, dotype(): ckuus6.c
 . Change Jeff's affiliation: most modules.

20 Mar 2004: Looked into adding long file support, i.e. handling files more
than 2GB (or 4GB) long.  Discovered very quickly this would be a major
project.  Each platform has a different API, or environment, or transition
plan, or whatever -- a nightmare to handle in portable code.  At the very
least we'll need to convert a lot of Kermit variables from long or unsigned
long to some new Kermit type, which in turn is #defined or typedef'd
appropriately for each platform (to off_t or size_t or whatever).  Then we
have to worry about the details of open() vs fopen(); printf() formats (%lld
vs %Ld vs %"PRId64"...), platforms like HP-UX where you might have to use
different APIs for different file systems on the same computer, etc.  We'll
need to confront this soon, but let's get a good stable 8.0.211 release out
first!  Meanwhile, for future reference, here are a few articles:

General: http://freshmeat.net/articles/view/709/
Linux:   http://www.ece.utexas.edu/~luo/linux_lfs.html
HP-UX:   http://devrsrc1.external.hp.com/STK/partner/lg_files.pdf
Solaris: http://wwws.sun.com/software/whitepapers/wp-largefiles/largefiles.pdf

Looked into FTP timeouts.  It appears I can just call empty() (which is
nothing more than a front end for select()) with the desired timeout before
any kind of network read.  If it returns <= 0, we have a timeout.  This is
not quite the same as using alarm() / signal() around a recv() (which could
get stuck) but alarm() / signal() are not not used in the FTP module and are
not naturally portable to Windows, but select() is already in use in the FTP
module for both Unix and Windows.  This form of timeout could be used
portably for both command response and data reads.  What about writes to the
command or data socket?  They can get stuck for hours and hours without
returning too, but the select() approach won't help here -- we need the
actual send() or recv() to time out, or be wrapped in an alarm()/signal()
kind of mechanism.  But if we can do that for sends, we can also do it for
receives.  Better check with Jeff before I start programming anything.
20 Mar 2004.

Later: Decided to postpone the above two projects (ditto IPv6) until after
8.0.211 is released because both will have major impacts on portability.
Grumble: all i/o APIs should have been designed from the beginning with a
timeout parameter.  To this day, hardly any have this feature.

3-4 Apr 2004: More 8.0.211 Beta.02+ test builds:

 . FreeBSD 3.3
 . FreeBSD 4.4
 . Linux Debian 2.1
 . Linux RH 6.1
 . Linux RH 7.1
 . Linux RH 7.2
 . Linux RH 9 (with 84 different combinations of feature selection)
 . Linux SuSE 6.4
 . Linux SuSE 7.0
 . NetBSD 1.4.1
 . NetBSD 1.5.2
 . OpenBSD 2.5
 . OpenBSD 3.0
 . QNX 4.25
 . SCO UnixWare 2.1.3
 . SCO UnixWare 7.1.4
 . SCO OpenServer 5.0.7
 . SCO XENIX 2.3.4 (no TCP)

Changes needed: None.

Problem: SCO XENIX 2.3.4 network build failed in the FTP module with
header-file syntax and conflicting-definitions trouble.  I'm not going to
try to fix it; 8.0.209 built OK with FTP, so we'll just keep that one
available.

Got access to VMS 8.1 on IA64.  Building the nonet version of C-Kermit
required minor modifications to ckvvms.h, ckv[ft]io.c, and ckvcon.c, to
account for a third architecture.  Also to SHOW FEATURES in ckuus5.c.  Once
that was done, the UCX 5.5 version built OK too.  Starts OK, makes Telnet
connection OK, sends files.  Has some obvious glitches though -- "stat"
after a file transfer reports 0 elapsed time (in fact it was 00:09:48) and
1219174400 cps (when in fact it was 10364).  This doesn't happen on the
Alpha.  Btw, the IA64 binary is twice as big as the Alpha one.  Changed
to Beta.03.  5 Apr 2004.

Fixed the ckdaily script to include the makefile and man page in the Zip
file (they were not included because the Zip file was intended mainly for
VMS users, but some Unix users prefer Zip to tar.gz).  6 Apr 2004.

Traced problems in VMS/IA64 statistics report to rftimer()/gftimer() in
ckvtio.c, which use sys$ and lib$ calls to figure elapsed time.  These work
on VAX and Alpha but not IA64.  Sent a report to the chief engineer of the
IA64 VMS port; he says it's probably a bug in VMS 8.1 (which is not a real
release); he'll make sure it's fixed in 8.2.  As an experiment, tried
swapping in the Unix versions of these routines (which call gettimeofday()
etc).  They seem work just fine (it hung a couple times but I think that's
because the underlying system hung too; trying it later on a new connection,
it was fine; however I noticed a BIG discrepancy in throughput between
sending and receiving).  Moved definitions for VMS64BIT and VMSI64 to
ckcdeb.h so all modules can use them and added them to the SHOW FEATURES
display.  Added VMSV80 definition to build procedure.  Beta.03+.  ckcdeb.h,
ckcuus5.c, ckcvvms.h, ckvtio.c, ckvker.com, 6 Apr 2004.

While doing the build-all, I noticed the VMS version did not build with
Multinet or older UCX versions, always with the same errors -- undeclared
variables, undefined symbols, all TCP/IP related.  This didn't happen a
couple weeks ago...  Somehow the order of #includes was messed up --
ckuusr.h depended on symbols that are defined in ckcnet.h, but ckcnet.h
was being included after ckuusr.h...  this was compounded by two missing
commas in ckvker.com.  11 Apr 2004.

Removed Beta designation, released as 8.0.211, 10 Apr 2004.

I had somehow lost the edit to ckutio.c that changed the UUCP lockfile for
Mac OS X from /var/spool/uucp to /var/spool/lock.  So I slipped it in and
re-uploaded version 8.0.211.  You can tell the difference because SHOW
VERSIONS has 17 Apr 2004 for the Communications I/O module.  Also the 10.3
executable now has a designer banner: "Mac OS X 10.3".  makefile, ckuver.h,
ckutio.c, ckuus[45].c, 17 Apr 2004.

---8.0.211---

Removed "wermit" from "make clean" (how did it get there?).  makefile.

From Jeff, applied 10 May 2004.
 . Rearrange #ifdefs that define OS/2-only features. ckcdeb.h.
 . Fix two strncat()s that should have been ckstrncat()s.  ckuus7.c.
 . Fix two strncat()s that should have been ckstrncat()s.  ckuus4.c.
 . Fix one strncat(). ckcfns.c.
 . SET FTP CHAR ON used backwards byte order when output to screen.  ckcfns.c.
 . Fix two strncat()s.  ckuus3.c.
 . Add SET NETWORK TYPE NAMED-PIPE for K95.  ckuus3.c.
 . Add "No active connections" message to hupok().  ckuus3.c.
 . Fix many strncat()s.  ckcnet.c.
 . Fix some strncat()s.  ckcftp.c
 . Make FTP port unsigned short for 16383 < port < 65536.  ckcftp.c.
 . Improvements to FTP USER command.  ckcftp.c.
 . Fix FEAT parsing to allow for various forms of whitespace.  ckcftp.c.

S-Expression (AND FOO BAR) would not short-circuit if FOO's value was 0,
even though short-circuiting code has been there since Day 1.  Similarly for
(OR BAR FOO).  Turns out the first operand was a special case that bypassed
the short-circuit check.  Fixed in dosexp(): ckuus3.c, 10 May 2004.

Red Hat 7.3 (and maybe others) <baudboy.h> referenced open() without first
ensuring it was declared.  The declaration is in <fcntl.h>, which is after
<baudboy.h> in ckutio.c series of #includes.  Made a special case for this.
ckutio.c (see comments), 10 May 2004.

If the local Kermit's parity is set to SPACE and then a file arrives via
autodownload, automatic parity detection improperly switches it to NONE.
Fixed in rpack() by switching parity automatically only if parchk() returns
> 0 (rather than > -1), since NONE and SPACE are indistinguishable.  A
bigger problem still remains: autodownload does not work at all if the
sender is using actual parity bits (even, odd, or mark) and the receiver's
parity is NONE.  ckcfn2.c, 10 May 2004.

When a DIAL MACRO is defined and the phone number is comprised of more than
one "word" (i.e. contains spaces), the dial macro loses the second and
subsequent words after the first call. Fixed in xdial() by inserting quotes
around phone number before passing it to xdial(). ckuus6.c, 10 May 2004.

DIAL MACRO fix was not right; the quotes were kept as part of the phone
number and sent to the modem.  dodo() pokes its argument to separate the
macro argument string into its component arguments.  xdial() is called
repeatedly on the same string, so after the first time, a NUL has been
deposited after the first word of the telephone number.  The fix is to have
xdial() create a pokeable copy of its argument string before calling
dodo(dial-macro,args...).  It might seem odd that dodo pokes its argument,
but making copies would be would be prohibitive in space and time.
ckuus6.c, 23 May 2004.

FTP CD did not strip braces or quotes from around its argument.  Fixed in
doftprmt(): ckcftp.c, 23 May 2004.

Added client side of REMOTE MESSAGE/RMESSAGE/RMSG: ckuus[r27].c, 23 May 2004.

Server side of REMOTE MESSAGE: ckcpro.w, 23 May 2004.

From Dave Sneddon: an updated CKVKER.COM containing a fix where the
COMPAQ_SSL symbol was not defined but later referenced which generated an
undefined symbol error.  ckvker.com, 5 Jan 2005.

From Andy Tanenbaum (28 May 2005):
 . Fix an errant prototype in ckcker.h and ckucmd.h - () instead of (void).
 . Add support for MINIX 3.0.  makefile, ckutio.c, ckufio.c, ckuver.h.

Fixed messed-up sndhlp() call which apparently had been jiggered to
compensate for the bad prototype which has now been fixed, ckcpro.w,
12 Jun 2005.

From Jeff (12 June 2005):
 . Security updates.  ck_ssl.c, ck_crp.c, ckuath.c.
 . Fix bug in K95 SET PRINTER CHARACTER-SET. ckuus3.c.
 . Add printer character-set to K95 SHOW PRINTER display. ckuus5,c
 . Add SET MSKERMIT FILE-RENAMING to K95. ckuus7.c, ckuusr.h.
 . Add help for K95 SET MSKERMIT.  ckuus2.c.
 . Add SET GUI CLOSE to K95.  ckuusr.h, ckuus2.c, ckuus3.c
 . Add help text for K95 SET GUI MENUBAR and TOOLBAR.  ckuus2.c.
 . Add --noclose command-line option for K95.  ckuusy.c
 . Add PAM support for Mac OS X.  ckufio.c.
 . Add GSSAPI support for Mac OS X.  ckcftp.c.
 . Pick up more URL options.  ckcker.h, ckuusy.c.
 . Fix bug in delta-time calculation across year boundary.  ckucmd.c.
 . Add Secure Endpoints to copyright notices.  ckcmai.c.
 . Fix FTP HELP to override unverbose setting.  ckcftp.c.
 . Fix assorted minor typos.

From Matthias Kurz: automatic herald generation for NetBSD 2.0 and later,
"make netbsd2".  ckuver.h, makefile, 12 Jun 2005.

Added SET TERMINAL LF-DISPLAY, like CR-DISPLAY but for linefeed rather than
carriage return.  ckuusr.h, ckuus[257x].c, 12 Jun 2005.

Made a command-line option --unbuffered to do what the -DNONOSETBUF
compile-time option does, i.e. force unbuffered console i/o.  Unix only.
ckuusr.h, ckuusy.c, ckutio.c, 12 Jun 2005.

Fixed getiact() (which displays TERM IDLE-ACTION setting) to display
space as \{32}.  ckuus7.c, 12 Jun 2005.

Added LMV as a synonym for LRENAME, which is itself a synonym for LOCAL
RENAME.  ckuusr.c, 12 Jun 2005.

Put HELP SET TERMINAL DG-UNIX-MODE text where it belonged.  ckuus2.c,
12 Jun 2005.

Added IF LINK (Unix only) to test if a filename is a symlink.  Uses the most
simpleminded possible method, calls readlink() to see if it succeeds or fails.
No other method is dependable across different Unixes.  This code should be
portable because I already use readlink() elsewhere within exactly the same
#ifdefs.  ckufio.c, ckuus2.c, ckuus6.c, 12 Jun 2005.

Fixed a bug in which \fdir() wouldn't work when its argument was the nonwild
name of a directory file.  zxpand(): ckufio.c, 12 Jun 2005.

Made \fdirectory() a synonym for \fdirectories().  Made \fdir() an
acceptable abbreviation for these, even though it clashes with \fdirname(),
which still works as before.  ckuus4.c, 12 Jun 2005.

Added the long-needed \flopx() function, to return rightmost pieces of
strings, such as file extensions.  \fstripx() and \flopx() are the
orthogonal functions we need to pick filenames apart from the right:
\stripx(foo.tar.gz) = foo.tar; flopx(foo.tar.gz) = gz.  ckuusr.h, ckuusr.c,
ckuus2.c, 12 Jun 2005.

Removed reference to defunct fax number, ckcmai.c, 12 Jun 2005.

Added -DHAVE_PTMX to linux+krb5+openssl+zlib+shadow+pam.  From Timothy Folks.
makefile, 12 Jun 2005.

Built on Solaris 9 and NetBSD 2.0.

From Jeff: New build target for Mac OS X 10.3 with Kerberos 5 and SSL.
makefile, 14 Jun 2005.

Fixed error in ckuver.h NetBSD #ifdefs.  15 Jun 2005.

Fixed SET TERMINAL IDLE-ACTION OUTPUT to work as documented, namely if the
output string is empty, to send a NUL.  Previously there was no way to make
it send a NUL.  ckuus7.c, 15 Jun 2005.

Suppose (in Unix, for example) a filename contains wildcard characters, such
as {abc}.txt.  When referring to such a file (e.g. in a SEND command), these
characters can be quoted, e.g. \{abc\}.txt.  But if the file list has been
obtained programmatically, e.g. stored in an array, there is no way, short
of tedious, complicated, and error-prone string processing, to reference the
file.  For this we need a way to disable wildcard processing.  I added { ON,
OFF } choices for the SET WILD and SHOW FILE commands: ckuusr.h, ckuus[234].c.
{ ON, OFF } turns wildcarding off and on without affecting the { KERMIT,
SHELL } agent choice; it does this by setting a new and separate global
variable, wildena.  Added semantics to ckufio.c.  Crude but effective.  It
might have been more Unixlike to add Yet Another form of quoting but we
have enough of that already (later maybe I'll add a \function() for this).
Needs to be propogated to Windows and VMS.  15 Jun 2005.

Improved and fixed typos in HELP WILDCARD and HELP PATTERN.  ckuus2.c,
15 Jun 2005.

The GREP command, and probably anything else that uses ckmatch() for pattern
matching, failed on patterns like */[0-3]*.html.  The [a-b] handler, when
failing to match at the current position, neglected to back up the pattern
and try again on the remainder of the string.  I also fixed another case, in
which matching a literal string a*b?c against the pattern a[*?]*[?*]c caused
ckmatch() to recurse until it blew up.  ckclib.c, 16 Jun 2005.

Added builds and designer banner for Solaris 10.  makefile, ckuver.h,
27 Jun 2005.

Defined CKHTTP for NetBSD, the HTTP code builds and works fine there.
ckcdeb.h, 2 Jul 2005.

Added #ifndef OSF40..#endif around definition of inet_aton() in ck_ssl()
to allow building in Tru64.  Added tru64-51b+openssl to makefile.
15 Jul 2005.

HTTP GET would fail if the URL contained any metacharacters, no matter how
much you quoted them.  Although it uses cmfld() to parse the (partial) URL,
it then uses cmofi() to get the output filename, which by default is the
"filename" from the URL, which might be something like "rankem.asp?id=1639".
cmofi() refuses to accept unquoted metacharacters in "filenames" and that's
what happens in this case if the output filename is not specified.  Worked
around this by disabling wildcard processing around HTTP GET using the new
"wildena" variable from June 15th.  ckuusr.c, 18 Jul 2005.

Fixed the June 16th fix to the pattern matcher.  I fixed a real problem, but
I made an unrelated optimization that introduced new ones.  ckclib.c,
18 Jul 2005.

Added missing help text for \fb64encode() and \fb64decode().  ckuus2.c,
18 Jul 2005.

Changed SET WILD OFF help text to warn that this setting prevents the
creation of backup files (later I'll have to see if something more useful
can be done about this).  ckuus2.c, 18 Jul 2005.

Built OK on Mac OS X 10.4.2 using macosx103 target (but with some
"signedness" warnings in ckcnet.c and ckcftp.c).  Built on Unixware 7.1.4
with uw7 target.  27-28 Jul 2005. 

Added -DCKHTTP to Mac OS X 10.3-.4 KFLAGS.  Makefile, 4 Aug 2005.

Built on BSDI 4.3.1.  Added -DCKHTTP.  

Compact substring notation extended to accept not only start:length but also
start-end notation.  Thus \s(foo[12:18]) means the substring of foo starting
at position 12 of length 18, and tne new \s(foo[12-18]) means the substring
of foo starting at position 12 and ending with position 18.  Ditto for
\:(\%a), etc.  ckuus4.c, 9 Aug 2005.

See correspondence with Mark Sapiro, Nov 2003 and Sep 2004, about certain
variations on IF syntax having been broken by the introduction of "immediate
macros" circa 1999.  It seems the problem -- variables not being expanded --
always occurs in the ELSE part when (a) the IF condition is false; (b) the
ELSE command is "standalone", i.e. expressed as a separate command after the
IF command (original C-Kermit 5A syntax), and (c) its command list is a block.
This would suggest the problem is in the XXELS parser.

Going back to 1999, I find this:
  Fixed a problem Jim Whitby noticed with quoting in ELSE statements.  This
  problem was introduced when I unified IF and XIF, and occurs only when
  ELSE begins on a line, followed by a { command list } rather than a single
  command.  The solution (gross) was to make a special version of pushcmd()
  (called pushqcmd()) for this situation, which doubles backslashes while
  copying, BUT ONLY IF it's a command list (i.e. starts with "{"); otherwise
  we break lots of other stuff.  Result passes Jim's test and still passes
  ckedemo.ksc and iftest.ksc.  ckucmd.c, ckuus6.c, 27 Sep 99.

I undid this change and it made no difference to all the other IF
constructions (in fact, it fixed an urelated one that was broken, so now
iftest scores 54 out of 54, instead of 53).  However, it does not fix the
ELSE problem; in fact it pushes it all the way in the other direction:

  The opposite occurs any time you try to execute an immediate macro inside a
  macro or any other { block }: not only is the variable evaluated, it is
  evaluated into nothing.  It looks like this happens only in immediate
  macros, i.e. *commands* that start with '{'.  So maybe we really have two
  isolated problems, that can each be fixed.

The situation is illustrated by this simple script:

  def xx {
      if false { echo \%1, echo \%2 }
      else { echo \%3, echo \%4 }
  }
  xx one two three four

With pushqcmd() it echoes the variable names literally; with pushcmd() it 
echoes empty lines.  Since ELSE, when its argument is a block, dispatches
to the immediate-macro handler, it seems we have unified the two problems,
so fixing one should fix the other.

The problem is that we define a new temporary macro and then call dodo() to
execute it.  But if the definition contains macro arguments, we have added a
new level of macro invocation, thus wiping out the current level of args.
The cure is to expand the variables in the immediate macro in the current
context, before executing it.  This means simply changing the cmtxt() call
that reads the immediate macro to specify xxsting as its processing
function, rather than NULL, which is used for real macros to defer their
argument evaluation until after the macro entered.  ckuusr.c, 11 Aug 2005.

Added a new makefile target, macosx10.4, for Mac OS X 10.4.  This one uses
an undocumented trick to get the otherwise unavailable-except-by-clicking
Mac OS X version number (in this case 10.4.2) and stuff it into the HERALD
string.  makefile, 11 Aug 2005.

Built OK on Solaris 9, Solaris 10 (with a few implicit declaration warnings
in ckuusx.c), Mac OS X 10.4.2 (with some warnings in ckcnet.c and ckcftp.c),
Mac OS X 10.3.9 (also using the macos10.4 entry, which gets the right
version number, and gets no warnings at all), RH Enterprise Linux AS4 on AMD
x86_64, Tru64 Unix 4.0F, SCO UnixWare 7.1.4

For docs and/or scriptlib:  Unix C-Kermit can be a stdin/out filter.  The
trick is to use the ASK, ASKQ, or GETC command for input, specifying no
prompt, and ECHO or XECHO for output, e.g.:

while true {
    ask line
    if fail exit 0
    echo \freverse(\m(line))
}
exit 0

FOPEN didn't do anything with the channel number if the open failed, so any
subsequent command that tried to reference it would get a parse error it was
undefined or non-numeric, not very helpful.  Changed FOPEN to set the
channel number to -1 if the file can't be opened.  Now subsequent operations
on the channel fail with "Channel -1: File not open".  I also added two
magic channel numbers: -8 means that any FILE command (besides OPEN and
STATUS) on that channel is a noop that succeeds silently; -9 is a noop that
fails silently.  So now it's possible to simply set a channel number to one
of these values to disable i/o to certain file without getting lots of error
messages.  dofile(): ckuus7.c, 12 Aug 2005.

Added automatic herald construction for UnixWare 7.  makefile, 12 Aug 2005.

Unix isdir() never allowed for arguments that started with tilde, so gave
incorrect results for ~/tmp/ or ~fdc.  The problem was mainly invisible
since most commands that parsed file or directory names used cmifi(), cmdir(),
etc, which did the conversions themselves.  But IF DIRECTORY was an exception,
since its operand had to be treated as just text, and then tested after it
was parsed.  ckufio.c, 13 Aug 2005.

Fixed the following:
"ckuusx.c", line 8959: warning: implicit function declaration: ckgetpeer
"ckufio.c", line 1869: warning: implicit function declaration: ttwait
"ckufio.c", line 2941: warning: implicit function declaration: mlook
"ckufio.c", line 2943: warning: implicit function declaration: dodo
"ckufio.c", line 2944: warning: implicit function declaration: parser
"ckcftp.c", line 2625: warning: implicit function declaration: delta2sec
"ckcftp.c", line 4071: warning: no explicit type given for parameter: prm
"ckcftp.c", line 8389: warning: no explicit type given for parameter: brief
ckuusx.c, ckufio.c, ckcftp.c, ckucmd.h.  13 Aug 2005.

Unbuffered stdout code has never worked because the setbuf(stdout,NULL) call
has to occur before the stdout has been used.  The reason it's needed is
that some Kermit code writes to stderr (which is unbuffered) and other code
writes to stdout, and therefore typescripts can come out jumbled.  Robert
Simmons <robertls@nortel.com> provided the needed clue when he insisted it
worked only when executed at the very beginning of main().  So I moved the
code to that spot.  But since now we also want to make unbuffered a runtime
(command-line) option, I had to do a clunky by-hand pre-prescan inline in
main() to look thru argv[], even before prescan() was called.  ckcmai.c,
ckutio.c, ckuusy.c, 13 Aug 2005.  (Now that this works, it might be a good
idea to remove all use of stderr from Kermit.)

Managed, after some finagling, to build a 64-bit version on Solaris 10 at
Utah Math with Sun cc.  (Can't make any gcc builds at all, 32- or 64-bit,
they all blow up in <sys/siginfo.h>.)  New target: solaris10_64.  makefile,
15 Aug 2005.

The 64-bit Solaris 10 version compiles and links OK and transfers files in
remote mode.  It can make FTP connections and use them, but Telnet connections
always fail with "network unreachable".  This is with all default libs and
include files.  Nelson has a separate set in /usr/local, which he references
explicitly in all his 64-bit builds, but using these makes no difference.
Some data type is wrong in ckcnet.c.  But telnet works fine in 64-bit Linux
and Tru64 builds.  Debug logs trace the difference to netopen() (of course),
the spot where we test the results of inet_addr(), which is already marked
suspicious for 64-bit builds.  It seems that inet_addr() is of type in_addr_t,
which in turn is u_int32, i.e. an unsigned 32-bit int.  Yet the man page says
that failure is indicated by returning -1.  I guess this doesn't matter in
32-bit builds, but in the 64-bit world, the test for failure didn't work
right.  I made a Solaris-specific workaround, and checked that it works in
both 32-bit and 64-builds.  I really hate typedefs.  ckcnet.c, 15 Aug 2005.

Changed the plain-text version (as opposed to the popup or GUI version - the
GUI version, at least, already does this) of ASKQ to echo keystrokes
asterisks rather than simply not echo anything, so it's easier to see what
you're doing, the effects of editing, etc.  Experimental; for now, there's
no way to disable this.  Not sure if there needs to be.  Anyway, to get this
working required a fair amount of cleaning up of gtword(), which was echoing
different ways in different places.  ckuus6.c, ckucmd.c, 15 Aug 2005.

Added a solaris9_64 target for building a 64-bit version on Solaris 9 with
Sun cc.  Verified, using the DIR command and \fsize() function on a 4.4GB
file, that the Solaris 64-bit version of Kermit gets the size correctly, and
that it can copy such a file (thus its fopen/fread/fwrite/fclose interface
works right).  Initiated a large-file transfer between here and Utah over
SSH and verified that it puts the correct file size in the A packet when
sending; the right quantites are shown on the file transfer display (file
size CPS, percent done, etc).  But even at 5Mb/sec, it takes a good while to
transfer 4.4GB, more than 2 hours (not streaming; 30 window slots, 4K
packets, maybe it would go faster with streaming)...  After an hour or so,
it filled up the partition and gave up (gracefully) before it reached the
2GB frontier (drained its pending packets, closed the partial file).
Restarted at 12:54, this time with streaming and 8K packets (the speed
wasn't significantly different).  This time it transferred 95% of the file
(4187660288 bytes) before failing because the disk filled up.  Went to Utah
and started a transfer between two Solaris 10/Sparc hosts; this goes about 8
times faster.  The transfer completed successfully after 17m41s.  All fields
in the f.t. display looked right the whole time.  Then I verified various
other 64-bit combinations transferring the same 4.4GB file:

        To................
  From  Sol  Amd  i64  Tru      
  Sol   OK   OK   OK   OK      Sol = Solaris 10 / Sparc
  Amd   OK                     Amd = AMD x86_64 RH Enterprise Linux AS4
  i64   OK                     i64 = Intel IA64, RH 2.1AS
  Tru                          Tru = Tru64 Unix 4.0F Alpha

(The other combinations are difficult to test for logistical reasons.)

Tried sending the same long file with Kermit's FTP client.  It chugged along
for a while until I stopped it; it would have taken hours to complete.
There is no indication that it wouldn't have worked, assuming the FTP server
could also handle long files, which who knows.  Anyway, Kermit showed all
the right data on the display screen.  17 Aug 2005.

On AMD x86_64 and IA64 native 64-bit Linux builds, the pty routines did not
work at all.  ptsname() dumped core.  If I commented out ptsname(), then the
next thing dumped core.  The same code works on the other 64-bit builds.
Poking around, I see that this version of Linux has an openpty() function,
which I could try using instead of the current API -- grantpty(), etc.  Then
I see that openpty() is already coded into Kermit's pty module,
conditionalized under HAVE_OPENPTY, which has never before been defined for
any build.  I added a test to the makefile linux target (look for the
openpty() prototype in <pty.h>, if found define HAVE_OPENPTY as a CFLAG and
also add -lutil to LNKFLAGS).  Works fine on the problem builds, and also
on previously working 32-bit builds.  makefile, 17 Aug 2005.

Fixed a bug in the ASKQ echo asterisks code, which made the VMS version of
C-Kermit always echo asterisks.  Turns out that some code in the main parse
loop to reset command-specific flags was in the wrong place, which had other
effects too, for example ASKQ temporarily turns off debug logging as a
security measure, but the code to turn it back on was skipped in most cases.
Some other side effects related to the DIRECTORY and CD commands might have
been possible but I haven't seen them.  ckuus[56].c, 23 Aug 2005.

Problem reported when sending a file to VMS when the name in the F packet
starts with a device specification and does not include a directory field,
and PATHNAMES are RELATIVE.  Example: dsk:foo.bar becomes f_oo.bar.  The
code assumes that if there is a device field, it is followed by a directory
field, and it inserts a dot after the '[', which in this case is not there.
Later the dot becomes '_' because of the only-one-dot rule.  Solution: only
insert the dot if there really is an opening bracket.  nzrtol(): ckvfio.c,
23 Aug 2005.

A report on the newsgroup complains that C-Kermit and K95 servers were
sending REMOTE DIR listings with only #J line terminators, rather than #M#J.
Yet all the other REMOTE xxx responses arrived with #M#J.  snddir() was
neglecting to switch to text mode.  ckcfns.c, 26 Aug 2005.

Back to long files.  What happens if 32-bit Kermit is sent a long file?
It gets an A-packet that looks like this:

  ^A_"A."U1""B8#120050815 18:28:03!'42920641*4395073536,#775-!7@ )CP

The 32-bit receiver reacts like so:

  gattr length[4395073536]=100106240

the first number being the string from the A-packet, the second being the
value of the long int it was converted to by atol().  Clearly not equal in
this case.  When this happens Kermit should reject the file instead of
accepting it and then getting a horrible error a long time later.  Added
code to gattr() to convert the result of atol() back to a string and compare
it with the original string; if they're not equal, reject the file on the
assumption that the only reason this could happen is overflow.  Also some
other code in case the sender sends the only LENGTHK attribute.  Now files
whose lengths are too big for a long int are rejected right away, provided
the sender sends the length in an A packet ahead of the file itself.  If
this new code should ever cause a problem, it can be bypassed with SET
ATTRIBUTE LENGTH OFF.  ckcfn3.c, 26 Aug 2005.

As I recall from when I was testing this a few weeks ago, when the too-big
length is not caught at A-packet time, the transfer fails more or less
gracefully when the first attempt is made to write past the limit.  I went
to doublecheck this by sending a big file from the 64-bit Solaris10 version
to a 32-bit Mac OS X version that does not have today's code.  The Mac
thinks the incoming file is 2GB long when it's really 4GB+.  But in this
case, something new happens!  Although the percent done and transfer rate go
negative, the file keeps coming.  It would seem that Mac OS X lets us create
long files without using any special APIs.  The transfer runs to completion.
Mac OS X Kermit says SUCCESS (but gets the byte count and cps wrong, of
course).  But then a STATUS command says FAILURE.  The file was, however,
transferred successfully; it is exactly the same length and compares byte
for byte with the original.  This tells me that in the Mac OS X version --
and how many others like it??? -- today's rejection code should not be
enabled.  Meanwhile I put today's new code in #ifndef NOCHECKOVERFLOW..#endif,
and defined this symbol in the Mac OS X 10.4 target.  Over time, I'll have
to find out what other platforms have this characteristic.  And of course
I'll also have to do something about file-transfer display, statistics, and
status.  makefile, ckcfn3.c, 26 Aug 2005.

From now on I'm going to bump the Dev.xx number each time I upload a new
ckdaily.  This one will be Dev.02.  ckckmai.c, 26 Aug 2005.

Got rid of all the extraneous FreeBSD 4 and 5 build targets.  Now there's
one (freebsd) for all FreeBSD 4.1 and later.  makefile, 27 Aug 2005.

Mac OS X 10.4 (Tiger) is a 64-bit OS.  Building C-Kermit 0n 10.4.2 without
any special switches stilll gives a 32-bit executable.  Ditto building with
-mpowerpc64.  Further investigation turned up a tip sheet on MySQL that says
you have to include all of these: -mpowerpc64 -mcpu=G5 -mtune=G5 -arch
ppc64.  That did the trick.  New makefile target: macosx10.4_64.  But the
10.4.2 system I tried did not have 64-bit [n]curses or resolv libs, so this
build has no -DNOCURSES -DNO_DNS_SRV.  makefile, 27 Aug 2005.

Created a symbol CK_64BIT to indicate true 64-bit builds at compile time.
Added 64-bit announcement to the startup herald and the VERSION text.
ckcdeb.h, ckuus[r5].c, 27 Aug 2005.

Added a built-in variable \v(bits) to indicate the size of the build
(16, 32, 64, or whatever else sizeof() might report).  ckuusr.h, ckuus4.c,
27 Aug 2005.

Got rid of all the warnings in 64-bit Mac OS X about args to getsockopt(),
getsockname(), and getpeername(), and the comparisons on the return value
of inet_addr().  ckcnet.[ch], 27 Aug 2005.

Now to check the effects on other builds...
  Linux on AMD64: ok.
  Linux on IA64: ok.
  Linux on i386: ok.
  Mac OS X 10.3.9 32-bit: ok.
  Solaris 10 64-bit: ok.
  Solaris 9 32-bit: ok.
  Tru64 4.0F: ok.
  FreeBSD 4.11: ok.
  FreeBSD 5.4 ia64 (64-bit): ok.
  FreeBSD 5.4 i386 (32-bit): ok.

The Tru64 5.1B build totally blew up because they have their own unique
sockopt/etc length-argument data type (int!), so I had to roll back on using
socklen_t for this in all 64-bit builds.  Checked to make sure it still
builds on Tru64 4.0F after this change (it does).  ckcnet.h, 27 Aug 2005.

The HP-UX 11i/ia64 build comes out to be 32-bit but thinks it's 64-bit.
CK_64BIT is set because __ia64 is defined.  So how do I actually make a
64-bit HP-UX build?  I tried adding +DD64 to CFLAGS, and this generates
64-bit object files but linking fails to find the needed 64-bit libs
(e.g. -lm).  For now I added an exception for HPUX to the CK_64BIT
definition section.  ckcdeb.h, 27 Aug 2005.

Took the time to verify my recollection about the "graceful failure" on a
regular Pentium Linux system when receiving a too-big file...  OK, it's not
exactly graceful.  It gets a "File size limit exceeded" error; the message
is printed in the middle of the file-transfer display, apparently not by
Kermit, and Kermit exits immediately.  Looks like a trap...  Yup.  "File
size limit exceeded" is SIGXFSZ (25).  What happens if we set it to SIG_IGN?
Just the right thing: The receiver gets "Error writing data" at 2147483647
bytes, sends E-packet to sender with this message, and recovers with total
grace (drains packet buffers, returns to prompt).  ckutio.c, 27 Aug 2005.

Backed off from rejecting a file because its announced size overflows a
long.  Now instead, I set the file size to -2 (a negative size means the
size is unknown, but we have always used -1 for this; -2 means "unknown and
probably too big").  In this case, the f-t display says:

  File Size: POSSIBLY EXCEEDS LOCAL FILE SIZE LIMIT

then the user can interrupt it with X or whatever, or can let it run and
see if maybe (as in the case of Mac OS X) it will be accepted anyway.  This
way, we skip all the bogus calculations of percent done, time remaining, etc.
ckcfn3.c, ckuusx.c, 27 Aug 2005.

Discovered that VMS C-Kermit on Alpha and IA64 is a 32-bit application;
sizeof(long) == sizeof(char *) == 4.  Tried adding /POINTER_SIZE=64 to VMS
DECC builds on Alpha and IA64, but the results aren't great.  Tons of
warnings about pointer size mismatches between Kermit pointers and RMS ones,
and the executable doesn't run.  It appears that access to long files
would require a lot of hacking, similar to what's needed for 32-bit Linux.

--- Dev.02: 27 Aug 2005 ---

From Jeff, 28 Aug 2005.
 . Fix SSH GLOBAL-KNOWN-HOSTS-FILE / USER-KNOWN-HOSTS-FILE parsing, ckuus3.c.
 . Pick up K95STARTFLAGS from environment, ckuus4.c.
 . Fix some typos in command-line processing (-q), ckuus4.c.
 . Be sure to suppress herald if started with -q, ckuus7.c.
 . Fix ssh command-line switches, ckuusy.c.

Eric Smutz complained that HTTP POST was adding an extraneous blank line,
which prevented his application from successfully posting.  RFC 2616 states
(in Section 4.1):

   In the interest of robustness, servers SHOULD ignore any empty
   line(s) received where a Request-Line is expected. In other words, if
   the server is reading the protocol stream at the beginning of a
   message and receives a CRLF first, it should ignore the CRLF.

   Certain buggy HTTP/1.0 client implementations generate extra CRLF's
   after a POST request. To restate what is explicitly forbidden by the
   BNF, an HTTP/1.1 client MUST NOT preface or follow a request with an
   extra CRLF.

This seems pretty clear.  One section of code in http_post() (just above the
postopen: label) was appending a CRLF to a buffer whose last already was
terminated by CRLF, and then appended a second CRLF; thus two empty lines.
I removed the second one.  ckcnet.c, 28 Aug 2005.

I looked into the 64-bitness of NetBSD, it seems to be like Linux and
FreeBSD on 64-bit hardware, i.e. you just build it there and it works, at
least on Alpha and AMD64, going back to NetBSD 1.4 or 1.5.  But I don't have
access to any of these for verification and documentation on the Web is
scanty.

Checked PeterE's complaint again of warnings in ckutio.c about parameter
list of get[ug]id() and gete[ug]id().  When I "make hpux1100o" on HP-UX
11.11 (PA-RISC), there are definitely no warnings.  He says the same thing
happens on 10.xx, but I don't have access to that any more.  I also did
"make hpux1100o" on HP-UX 11.23 (11i v2) (PA-RISC), also no warnings.
(Except in both cases, a warning about a comment within a comment in
/usr/include/sys/ptyio.h).  On HP-UX 11i v2 on Itanium, however, there are
TONS of warnings, mostly of the "variable set but never used" kind.  Also
"dollar sign used in identifier".  Tracking this last one down, I see it's
complaining about code that's in #ifdefs for other platforms, such as
Apollo Aegis.  Is "aegis" defined in HP-UX 11i v2/IA64?  No!  (It would show
up in SHOW FEATURES if it was.)  Some phase of the compiler is complaining
about code that it should be skipping (and that, in fact, it *is* skipping
it because the build is successful).  It's as if cc is running lint for me
but not telling lint which macros are defined and which are not.

Verified that 64-bit linking fails in the same way for HP-UX 11i v2 on both
IA64 and PA-RISC.  Sent a query to HP.

Compiling ckcnet.c and ckcftp.c got the familiar sockopt-related warnings on
HP-UX 11i v2; turns out it is just like Tru64 Unix in using an int for the
length argument.  Added another special case and the warnings went away.
ckcnet.h, 28 Aug 2005.

Added some stuff to SHOW FEATURES to see what kinds of macros are exposed
(e.g. INT_MAX, LONG_MAX, LLONG_MAX, etc) and also show sizeof(long long) and
sizeof(off_t).  Building this code all over the place will give me an idea
of how widespread these data types are, and to what extent I can tell
whether they are available from clues in the header files.  (At first
glance, it appears that I'm not picking up <limits.h>, but adding an
#include for it is just asking for trouble.)  No complaints about long long
or off_t from Solaris 9 or recent Linuxes.  ckuus5.c, 28 Aug 2005.

Fixed a warning in HP-UX 10 and 11 stemming from some old-style prototypes
in ckutio.c for get[re][gu]id().  ckutio.c, 29 Aug 2005.

Updated minix3 target from Andy Tanenbaum.  makefile, 29 Aug 2005.

PeterE confirms that "long long" and off_t are available in all HP-UX 10 and
11, and in HP-UX 9 on PA-RISC but not Motorola.  30 Aug 2005.

Got 64-bit builds to work on HP-UX.  According to my notes, John Bigg of HP
said (in 1999) that HP-UX 10.30 and later require PA-RISC 1.1, and do not
work on PA-RISC 1.0.  But is PA 1.0 64-bit or what?  Today, Alex McKale of
HP said "The 64-bit binaries will work on all machines that have the same or
later release of HP-UX (excluding PA-RISC 1.1 machines)".  Still need
clarification...  Maybe it's that all IA64 builds can be 64-bit but I need
dual builds for PA-RISC.  Meanwhile I started transfer of a 4GB+ file from
Solaris to HP-UX 11i but it exceeded some quota on the HP long before it
approached the 2G point.  It failed cleanly and up until then it was working
fine (numbers, stats, etc).  30 Aug 2005.

Support of large files in 32-bit builds began in 10.20.  64-bit application
support began in 11.00, but not all machines that run 11.00 support 64 bits.
About long files, see HP /usr/share/doc/lg_files.txt.

PeterE found that certain patterns can still make Kermit loop; example:

  if match T01011-00856-21-632-073 *[abc] { echo GOOD } else { echo BAD }
  if match T01011-00856-21-632-073 *[a-z] { echo GOOD } else { echo BAD }

The minimum offending pattern is * followed immediately by an [xxx]
construction, followed by anything else, including nothing.  Previous
versions of Kermit handled this one correctly, without looping (but failed
certain matches that should have succeeded).  The new section of code I
added on 15 June, upon failure to match, advances the string pointer and
backs up the pattern to the previous pattern, and starts again
(recursively).  However, there needed to be a corresponding check at entry
for an empty target string.  ckmatch(): ckclib.c, 12 Sep 2005.

PeterE discovered that "kermit -y filethatdoesnotexit" gives an erroneous
error message that names the user's customization, rather than the name
given on the command line.  doinit(): ckuus5.c, 12 Sep 2005.

FREAD does not get an error if it tries to read a record or file or piece of
file that is too big for its buffer.  In particular, FREAD /SIZE:xxx seems
to succeed even if less than xxx was read.  It should fail unless, perhaps,
it successfully read up to the end of the file.  Furthermore, if xxx is
bigger than the file buffer size, it should complain.  The buffer is
line[LINBUFSIZ], 32K.  The lack of failure was due to code in dofile() that
adjusted the given size silently if it was greater than the buffer size,
which I removed, and also added a check when parsing the /SIZE: switch.
dofile(): ckuus7.c, 12 Sep 2005.

That still didn't help with FREAD /SIZE:n returning less than n bytes, even
when they were available.  That's because the underlying routine, z_in(),
didn't check fread()'s return code, which is the number of bytes read.
If fread() has smaller buffers, it needs to be called in a loop.  z_in():
ckuus7.c, 12 Sep 2005.

Flen() fails on strings of length 8192 or more.  The limitation is in the
callers of zzstring, which seem to be specifying an 8K buffer, in this case
fneval().  The operable symbols are FNVALL (max length of value returned by
a function) and MAXARGLEN (maximum length of an argument to a function).  I
changed both of these for BIGBUFOK builds to be CMDBL.  Buffers can never be
infinite, there has to be a limit.  It's important to make everything work
consistently within that limit, and to make something useful happen when the
limit is exceeded.  At this point, I can probably also increase the limits
for modern 32-bit systems, and certainly for 64-bit ones.  Also there's no
point in worrying about 16-bit platforms any more; earlier C-Kermit versions
can still be used on them if necessary. ckuusr.h, 12 Sep 2005.

Special #ifdefs for finding resolv.h and nameser.h in MINIX3 from Andy
Tanenbaum.  ckcnet.c, 20 Sep 2005.

PeterE noticed that ckmatch(), even though it works pretty well now, does a
lot of extra and unnecessary recursion after determining the string and
pattern do not match, at least when the pattern is of the form *[abc].
After several false starts I was able reduce this effect to a minor level
(but not eliminate it all together) by changing a while loop into a do loop.
ckmatch(): ckclib.c, 15 Oct 2005.

Added -DNOLONGLONG to HP-UX 8.00 and earlier builds, and to Motorola-based
HP-UX 9.00 builds.  This is simply to inhibit the test for whether "long
long" is supported by the compiler, since when it isn't, the module
containing the test won't compile.  makefile, ckuus5.c, 16 Oct 2005.

Making ASKQ always echo askterisks is a bad idea, because when it doesn't
echo, it's the perfect way to read silently from stdin, e.g. in a CGI script
(INPUT can also be used for this but it's not as straightforward).  So I put
the default for ASKQ back to no echoing, then gave ASKQ its own switch
table, which is the same as for ASK with the addition of an /ECHO:x switch,
which tells what character to echo.  ckucmd.c, ckuus[26].c, 17 Oct 2005.

Fixed a bug in FTP GET /COMMAND filename commandname; it always dumped core
dereferencing a null string (the nonexistent local asname).  ckcftp.c,
17 Oct 2005.

For docs: if you don't like the funny business that happens when you type
an IF command at the prompt, use XIF instead and it won't happen.  Also note
that commands like "if xxx { echo blah } else { echo blah blah }" don't
work when typed at the prompt; you have to use XIF for this. 

Back to ckmatch()...  Under certain conditions (e.g. patterns like *[abc])
failure to match would not stop the recursion because the string and pattern
arguments are on the stack, as they must be, so there was no way for level
n-1 to know that level n had detected a definitive nonmatch and that no
further attempts at matching were required.  The right way to handle this is
to recode the whole thing as coroutines, the cheap way out is with a global
static flag.  Works perfectly, in the sense that the match.ksc test results
are identical to what they were before and the extra backing up and
recursion are eliminated.  (The Oct 15th fix wasn't really a fix, it broke
a couple of cases.)  ckclib.c, 20 Oct 2005.

ckuus7.c(2987): warning #267: the format string requires additional arguments
(in PURGE command); fixed 20 Oct 2005.

From Andy Tanenbaum, final changes for MINIX3: #ifdef out the inline
definitions for gettimeofday() and readlink().  ckutio.c, 23 Oct 2005.

From Jeff: struct gss_trials initializers changed from gss_mech_krb5 to
ck_gss_mech_krb5.  ckcftp.c, 23 Oct 2005.

From Jeff: some improvements to K95 GUI SHOW TERMINAL.  ckuus5.c, 23 Oct 2005.

Found and corrected some misplaced #ifdefs in shofeat(), ckuus5.c, 23 Oct 2005.

--- Dev.03 ---

Fixed a compiler warning in a debug() statement in zzstring() by adding
parens.  ckuus4.c, 24 Oct 2005.

Added -DNOLONGLONG to sv68r3v6 target, makefile, 25 Oct 2005.

New makefile targets for HP-UX from PeterE to handle the 'long long'
situation.   26 Oct 2005.

From Jeff: changes to support OpenSSL 0.9.8, ck_ssl.h.  ckcasc.h has had
short names defined for ASCII control characters for 20-some years but now
they are causing conflicts, so EM becomes XEM (also for OpenSSL 0.9.8).
Changed K95's default terminal type from VT320 to VT220 because VT320
termcaps/terminfos are disappearing from Unix hosts: ckuus7.c.  Reorganize
the data-types section of SHOW FEATURES to add more macro tests for integral
sizes and to provide for the proper printf formatting in order to allow the
sizes to be output ("You are going to need to be careful because %llx is not
supported on all platforms.  On Windows, it is the same as %lx, 32 bits"):
ckuus5.c, 26 Oct 2005.

Defined NOLONGLONG ckcdeb.h for various old platforms where we know we are
never going to need 64-bit ints (even if they support a long long datatype,
chances are pretty slim they supported 64-bit file sizes).  ckcdeb.h,
26 Oct 2005.

PeterE noticed that GOTO targets can only be 50 characters long.  This was
by design, a long time ago, on the assumption that nobody would make longer
labels.  But in SWITCH statements, case labels can be variables that expand
to anything at all.  If we chop them off at 50, we might execute the wrong
case.  Changed the maximum label size to be 8K, and added code to dogoto()
to check when a label or target is too long and fail, to prevent spurious
GOTO or SWITCH results.  ckuusr.h, ckuus[r6].c, 26 Oct 2005.

Testing revealed there was still a problem with SWITCH case labels that were
variables that expanded into long strings.  Turns out that I was being
too clever when I decided that, if the SWITCH macro was n1 characters long
and the case-label search target was n2 characters long, I only had to
search the first n1-n2+1 characters of the macro definition.  That was true
before I allowed case labels to be variables, but not any more!  Fixed in
dogoto(): ckuus5.c, 26 Oct 2005.

--- Dev.04 ---

Dev.04 didn't actually contain Jeff's data-type changes to shofeat(),
I think I saved the wrong buffer in EMACS...  Fixed now.  27 Oct 2005.

PeterE corrected a typo in the HP-UX 7.00 makefile target.  27 Oct 2005.

PeterE had been reporting problems stress-testing the new SWITCH code, but
only on HP-UX 9, primarily stack overrun.  Turns out to be the HP-UX 9
optimizing compiler's fault.  No optimization, no problems.

PeterE found that even when dogoto() detects a string that is too long
and fails, this does not stop SWITCH from producing a result, which can not
possibly be trusted.  Changed the part of dogoto() that handles this to
not just fail, but also to exit the script immediately and return to top
level.  ckuus6.c, 28 Oct 2005.

An idea popped into my head after having typed too many commands like "dir
ck[cuw]*.[cwh]" to check the list of matching files, and then having to
retype the same filespec in a SEND command: Why not unleash some unused
control character such as Ctrl-K to spit out the most recently entered input
filespec?  It was easy, just a few lines in cmifi2() and gtword(), plus a
couple declarations.  To see all the changes, search for "lastfile" (all the
new code is protected by #ifndef NOLASTFILE).  ckucmd.c, 28 Oct 2005.

I added a new variable \v(lastfilespec) that expands to the same last
filespec, for use in scripts.  ckuusr.h, ckuus4.c, 28 Oct 2005.

The Unix version of C-Kermit failed to put anything in the session log if
SET TERMINAL DEBUG ON.  Rearranged the pertinent clause so logging happens
independent of TERMINAL DEBUG.  For now, since the user who noticed this
wanted debug format to go into the session log, that's what I do.  The
alternative would be to just log the raw incoming stream as usual, or to add
Yet Another SET Command to choose.  ckucns.c, 11 Nov 2005.

Fixed HELP INTRO text.  ckuus2.c, 11 Nov 2005.

Added NOLONGLONG for SV68.  ckcdeb.h, 11 Nov 2005.

--- Dev.05 ---

Added a debug() statement in FTP secure_getbyte() to see what's going on
with Muhamad Taufiq Tajuddin's 205-byte-per-second FTP/SSL downloads.

--- Dev.06 ---

Result: nothing, SSL_get_error() does not report any errors.  Suggested
testing SSL_read()'s return code, if 0 don't update the screen.

Created a new data type CK_OFF_T in ckcdeb.h that will eventually resolve
to whatever each platform uses for file sizes and offsets.  ckcdeb.h,
17 Nov 2005.

Made a new library routine ckfstoa() that converts a file size or offset to
a string.  This is to solve the problem with having to use different
printf() formats for different representations of file size (int, long, long
long, off_t, signed, unsigned, etc).  Replaced a few printf("%l",size) with
printf("%s",ckfstoa(size)) with the expected results.  This is just a start,
the definitions will need adjustment for many platforms, variables need to
be redeclared, and all the offending printf's (and printw's) will have to
hunted down and converted.  ckclib.[ch], ckuus4.c, 17 Nov 2005.

Built a minimal version on Linux with:
make linux "KFLAGS=-DNOLOCAL -DNOICP -DNOCSETS -DNODEBUG"
Worked fine, result was 260K on i686.  21 Nov 2005.

Discovered that Kermit's date parser, contrary to the documentation, failed
to handle strings like "Wed, 13 Feb 2002 17:43:02 -0800 (PST)", which are
commonly found in email.  This was because of an overzealous and misguided
check in the code; once removed, all was well.  ckucmd.c, 26 Nov 2005.

Added a new format code 4 to \fcvtdate() to emit asctime() format, used in
BSD-format email message envelopes (i.e. the "From " line).  shuffledate(),
ckucmd.c, ckuus[24].c, 26 Nov 2005.

Added a new function \femailaddress().  Given a From: or Sender: header line
from an RFC2822-format email address, extracts and returns the actual email
address, such as kermit@columbia.edu.  ckuusr.h, ckuus[42].c, 26 Nov 2005.

Using the new functions, I wrote a script to fetch mail from a POP3 server
over a TLS connection.  But the line-at-a-time input (needed for changing
line terminators and byte-stuffing text lines that start with "From ") is
slow, 17 sec to read 29 messages totaling 175K.

Added INPUT /CLEAR so INPUT can be started with a clean buffer without
requiring a sepearate CLEAR INPUT command.  ckuusr.h, ckuus[r24].c,
27 Nov 2005.

One thing that INPUT was never able to do well was read and save the
complete incoming data stream.  That's because, while waiting for its
target, the buffer might overflow wrap around.  Yet there was never a way to
tell it to stop when its buffer fills up and let me save it.  I added a
/NOWRAP switch that does this.  If the buffer fills up before any other
completion criterion is met, INPUT returns failure, but with \v(instatus)
set to 6 (the next available instatus value).  Thus a program that wants to
read and save (say) an email message from a POP server, which could be any
length at all, and which terminates with <CRLF>.<CRLF> could do this:

  set flag off
  while open connection {
      input /nowrap 10 \13\10.\13\10 # Wait for <CRLF>.<CRLF>
      if success {
          frwrite /string \%o {\freplace(\v(input),\13\10.\13\10,\13\10)}
          set flag on
          break
      } else if ( == \v(instatus) 6 || == \v(instatus) 1 ) {
          frwrite /string \%o {\v(input)}
          continue
      }
      break
  }
  if flag (handle success)

Note carefully the braces around the FWRITE text; without them, trailing
spaces would be lost.

Previously the only way to INPUT an entire data stream without losing
anything (assuming it was ordinary lines of text that were not "too long"),
was line-by-line:

  while open connection {
      input /clear 10 \13\10
      if fail break
      if eq "\v(input)" "$ \13\10" break
      fwrite /string \%o {\freplace(\v(input),\13\10,\10)}
  }

The new code is 3 times faster using the default INPUT buffer length of 4K.
Raising it to 16K makes it 3.6 times faster (not worth it).  Changing the
POP3 script to use INPUT /NOWRAP makes it about twice as fast (it does more;
it has to do all the byte-stuffing and unstuffing).  27 Nov 2005.

Changed ssl_display_xxx() to just return if SET QUIET ON.  Otherwise there
is no way to suppress the messages.  Also protected a previously unprotected
printf("[SSL - OK]\r\n"); by if ( ssl_verbose_flag ).  ck_ssl.c,
28 Nov 2005.

Discovered that FOPEN /APPEND doesn't work if the file doesn't exist.  It
uses cmiofi() which is a super-hokey front end to cmifi2().  I had code to
call it but for some reason it was commented out, with a note to the effect
it didn't work.  I uncommented it but that didn't help much.  So I wrote an
entirely new cmiofi() that works exactly as it should, using chained FDBs,
_CMIFI to _CMOFI (I think the original cmiofi() predated chained FDBs).
ckuus7.c, ckucmd.c, 29 Nov 2005.

Getting rid of the awful hacks required to call cmiofi() meant I also had to
change the EDIT command, which is the only other place where it's used.
Unfortunately now it's no longer possible to give EDIT without a filename
(to just start an empty editor) but I doubt anyone will notice.  ckuusr.c,
29 Nov 2005.

IF KERBANG didn't always work right.  If a kerbang script TAKEs another
kerbang script, the second one should have IF KERBANG false, but it didn't.
Added a check for \v(cmdlevel) == 1.  Now you can write a wrapper that runs
a kerbang script in a loop, and the latter can use IF KERBANG to know
whether to EXIT (if called at top level) or END (if called by another
script, thus allowing -- in this case -- the loop to continue).  ckuus6.c,
29 Nov 2005.

Changed \flop() and flopx() functions to take a third argument, a number
signifying at which occurrence of the break character to lop, so:

  \flopx(sesame.cc.columbia.edu) = edu
  \flopx(sesame.cc.columbia.edu,,2) = columbia.edu

ckuus[24].c, 1 Dec 2005.

Built OK on VMS 7.2-1 with MultiNet 4.4.  Built with and without OpenSSL on
Linux OK, ditto Solaris 9.  Built OK on RH Linux AS4 on X86_64 (64-bit);
"show var fsize" (using new ckfstoa()) works OK there.  Also Mac OS X 10.3.9
(32-bit), Tru64 UNIX 4.0F (64-bit), HP-UX 11iv2 (64-bit) (picky new compiler
spews out tons of useless warnings), FreeBSD 6.0 on ia64 (64-bit).

--- Dev.07 ---

Changed "make netbsd" to be a synonym for "make netbsd2" because the
original netbsd target was ancient.  Renamed it to netbsd-old.  makefile,
3 Dec 2005.

Updated INPUT and MINPUT help text.  ckuus2.c, 3 Dec 2005.

Discovered that on a SET PORT /SSL connection, Kermit treats incoming
0xff data bytes (e.g. sent from the POP server) as IACs and goes into Telnet
negotiations.  Jeff says "You will need to implement NP_SSLRAW and NP_TLSRAW
that do the same as NP_TCPRAW but negotiate SSL or TLS as appropriate."
This was not as easy as it sounded, because apparently a lot of the Telnet
code is used by SSL and TLS even when Telnet protocol is not being executed.
I wound up doing this as follows: I added /SSL-RAW and /TLS-RAW to the
switch table.  Rather than disable Telnet, they do exactly what the /SSL and
/TLS switches do, but also set a special flag.  This flag is checked in only
two place: netclos() (to prevent Kermit from sending TELNET LOGOUT when
closing the connection), and tn_doop() (to prevent Kermit from reacting to
incoming IACs; it makes tn_doop() return(3), which means "quoted IAC", which
causes the caller to keep the IAC as data).  ckcnet.h, ckctel.h, ckctel.c,
ckuus7.c, 4 Dec 2005.

The INPUT command did not account for tn_doop() returning 3.  Fixed in
doinput(), ckuus4.c, 4 Dec 2005.

Added another debug() statement in FTP secure_getbyte() to see what's going on
with Muhamad Taufiq Tajuddin's 205-byte-per-second FTP/SSL downloads, plus
new code to test SSL_read()'s return code (byte count); if 0 don't update
the screen.  ckcftp.c, 4 Dec 2005.

--- Dev.08 ---

Fixed a typo in the non-ANSIC definition of ckfstoa().  ckclib.c, 7 Dec 2005.

Our Ctrl-C trap (the ON_CTRLC macro) wasn't working for kerbang files.
Rearranged some code to make it work.  ckcmai.c, 8 Dec 2005.

Started converting code to use CK_OFF_T for file sizes and offsets, and
all [s]printf's to replace "%ld" or whatever with "%s", and the size
variable with a call to ckfstoa().  Since I haven't actually changed the
definition of CK_OFF_T from what all the size variables were to begin
with (i.e. long), it shouldn't do any harm.  So far just ckcfn3.c
10 Dec 2005.

An updated HP-UX 9.xx makefile target from PeterE to fix a core dump that
happens on that platform due to insufficient resources.  14 Dec 2005.

Added debug() statements to http_blah() routines to tell whether the
connection is "chunked".  There seems to be a bad performance problem.
ckcnet.c, 14 Dec 2005.

PeterE complained about ugly DIRECTORY error message, ?No files match -
"{blah}".  The braces are used internally in case the user typed more than
one filespec.  I changed the error message to remove them.  Ditto DELETE.
ckuus6.c, 15 Dec 2005.

The problem with HTTP downloads is that Kermit always does single-character
read() or socket_read() calls (or the SSL equivalent); see http_inc().  I
added buffering code for non-SSL connections only but it's gross because it
has to swap ttyfd and httpfd before calling nettchk().  I tried making a
nettchk() clone that accepts a file descriptor as an argument but it didn't
work because too many other routines that are invoked directly or implicitly
by nettchk() (such as in_chk()) are still hardwired to use ttyfd.  HTTP GETs
are now 20 times faster on the local network (the improvement is less
dramatic over a clogged Internet).  ckcnet.[ch], 15 Dec 2005.

--- Dev.09 ---

HTTP file-descriptor swapping is not thread safe.  Doing it right, of
course, is a big deal, so for now I just don't define HTTP_BUFFERING for
Windows.  ckcnet.c, 15 Dec 2005.

Noticed that HTTP not included in FreeBSD and OpenBSD builds.  Fixed in
ckcdeb.h, 22 Dec 2005.

Fleshed out 32/64-bit data type definitions and changed struct zattr
(file attribute structure) members length and lengthk to have the new
CK_OFF_T type.  Changed final arguments of debug() and tlog() to be the new
LONGLONG type.  ckcdeb.h, 22 Dec 2005.

Changed ckfstoa() to return a signed number in string form, rather than an
unsigned one.  That's because off_t is signed (thank goodness).  Added the
inverse function, ckatofs() so we can convert file sizes and offsets back
and forth between binary number and string.  ckclib.c, 22 Dec 2005.

Changed Attribute Packet reader to convert incoming file size attribute
with ckatofs() rather than atol().  ckcfn3.c, 22 Dec 2005.

Converted debug(), tlog(), ckscreen(), etc, to handle potentially "long long"
arguments by making their "n" argument CK_OFF_T.  ckuusx.c, ckcdeb.h,
22 Dec 2005.

Converted the rest of the source files to use CK_OFF_T for all file size
and offset and byte-count related variables, and converted all references to
these variables in printfs to go through ckfstoa().  Then I built it on
Linux/i386 with:

  make linux "KFLAGS=-D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64"

which makes off_t be 64 bits and magically makes all the regular file APIs
use 64-bit sizes and offsets without changing the API calls in the source
code.  It's going to be a lot of work to get through all the kinks but I was
able to send a long file, do directory listings of long files, do
\fsize(longfile), etc.  When it sends a file, the length is shown correctly
in the A packet.  If the receiver does not support big numbers, it receives
the file OK anyway, without showing the size, the thermometer, or percent
done (and then will get an error when the file keeps coming after the 2G
mark).  Kermit 95 actually refuses long files for "Size", but only if the
announced is less than 2^63 bytes.  When today's Linux version receives a
file, it shows the length correctly in the file-transfer display, as well as
percent done, thermometer, etc.  Also built this version on true 64-bit
Linux, and it worked fine.  Many files changed, 22 Dec 2005.

For the record, this API is specified in X/Open's Single UNIX Specification
Version 2, which is branded as UNIX 98.  It is called Large File Support, or
LFS, and was developed at the Large File Summit.

It looks like the operative feature-test macro in glibc for transitional
large file support is __USE_LARGEFILE64.  So if this is defined, we can also
supply _LARGEFILE_SOURCE and _FILE_OFFSET_BITS=64 automatically for 32-bit
Linux builds.  But there's a Catch-22, you don't know if this is defined
until you read the header files, but you have to define _LARGEFILE_SOURCE
and _FILE_OFFSET_BITS before you read the header files.  Maybe it's good
enough to grep through <features.h> for __USE_LARGEFILE64.  makefile,
23 Dec 2005.

Checked this on true 64-bit Linux.  The same symbols are defined in CFLAGS,
but they do no harm; it builds without complaint and works fine.  24 Dec 2005.

Built it on Red Hat Linux 6.1 from 1999.  This picked up the long file
support too.  Guess 6.1 isn't old enough to not have it!  Kermit seems to
work OK on regular files but I don't have enough disk space to create a long
file, and my bigfile.c program (which creates a long file containing only 1
byte) doesn't work ("fseeko: invalid argument").  It looks like parts of
this API were visible in Linux before they were actually working.
24 Dec 2005.

Converted all fseek() and ftell() to macros that expand to fseek() and ftell()
or fseeko() and ftello() depending on whether _LARGEFILE_SOURCE is defined.
ckufio.c, ckuus7.c, ckuusx.c, 24 Dec 2005.

Made a CK_OFF_T version of cmnum().  It would be a very big deal to just
change cmnum() to return a new type, so another idea is to rename cmnum() to
something else, cmnumw(), change its result argument to CK_OFF_T, and then
make a stub cmnum() to call it to get an int, then call cmnumw() explicitly
any time we need a big number.  ckucmd.c, 24 Dec 2005.

Calling cmnumw() directly requires changes to each routine that uses it.
The INCREMENT and DECREMENT commands, for example, required changes to
doincr(), varval(), and incvar(), and all references to them.  ckuusr.[ch],
ckuus[56].c, 24 Dec 2005.

Calling cmnumw() in chained FDBs required defining a new function code,
_CMNUW, adding a new member to the OFDB struct for returning wide results,
and adding a new case to cmfdb().  ckucmd.[ch], 24 Dec 2005.

Changed FSEEK and FCOUNT to use the new chained FDB interface, now we can
seek and look past 2GB.  ckuus7.c, 24 Dec 2005.

Next come switches, which store their results in a struct stringint.  This
struct was defined in each module where it was used (ckuus[r367].c, ckcftp.c).
I moved the definition to ckuusr.h and added a wval member, which can be
referenced by any switch-parsing code that calls cmnumw().  24 Dec 2005.

Changed SEND /CALIBRATE:n to allow big values of n.  This makes it possible
to test the protocol aspects of long-file transfer without actually having a
long file handy.  ckuusr.c, 24 Dec 2005.

SEND /SMALLER-THAN:n, SEND /LARGER-THAN:n, and and SEND /START:n also now
allow large values of n.  ckuusr.c, 24 Dec 2005.

Changed the algebraic expression evaluator to use wide values.
ckuus5.c, 24 Dec 2005.

Fixed ckfstoa() to handle the case when n is negative and (0 - n) is also
negative, which happens for numbers 2^(n-1) or greater, where n is the
number of bits in the word size we're dealing with, e.g. 64, in which case
2^63 has its sign bit set so seems to be negative.  In such cases, ckfstoa()
returns "OVERFLOW" instead of a numeric string.  We'll have to see how this
plays out but I think it's better to cause a parse error and stop things
dead than to return a spurious number.  ckclib.c, 24 Dec 2005.

Converted the S-Expression handler to use wide integers. ckuus3.c, 24 Dec 2005.

Took all the LONGLONG stuff out of ckcdeb.h, we don't need it.

All of these changes result in 64-bit arithmetic (more or less) on 32-bit
Linux, as well as on true 64-bit platforms.

Rebuilt today's code on Solaris 9 in the 32-bit and 64-bit worlds, on Red
Hat 6.1, Red Hat AS4.2.  I haven't bothered trying a 32/64 hybrid build for
Solaris, since I can build a pure 64-bit version there.  Quick tests show
the large-number arithmetic works OK in all cases except, of course, on pure
32-bit builds (unfortunately I can't find a running Linux system old enough
to verify this for Linux, but it's true for other 32-bit platforms).
24 Dec 2005.

Tried building a hybrid version on Solaris 9 after all since the LFS API is
ostensibly the same as for Linux:

 make solaris9 "KFLAGS=-D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64"

It built smoothly and the resulting binary is 2.5MB compared to 3.4MB for
the 100% 64-bit version.  Looks like a keeper.  For now, added solaris9lfs
and solaris10lfs entries to the makefile but if these work on PCs we can
make these the regular entries for Solaris 9 and 10.  27 Dec 2005.

Built on Mac OS X 10.4 with the regular target.  It seems that in that case,
off_t is 64 bits anyway.  Noticed that a lot of stuff didn't work, like
exponentiation in S-Expressions.  Tried building it as above, which worked,
and now CK_OFF_T is 64 bits instead of 32, but (^ 2 30) is still 2.0.  In
fact 2-to-the-any-power is 2.0.  It seems that the Mac OS X version did not
have FNFLOAT defined.  It also seems that every test in dosexp() like:

  if (result != fpresult) fpflag++;

should have been protected by #ifdef FNFLOAT..#endif /* FNFLOAT */ -- a
double-ended break, as they say in the nuclear power industry.  ckuus3.c,
27 Dec 2005.

Added GREP /EXCEPT:pattern.  ckuus[26].c, 27 Dec 2005.

Fixed a problem with uninitialized pv[].wval (switch-parsing parameter-value)
members that showed up on certain platforms or with certain compilers.  Now
the Mac OS X 10.4 version works.  ckuus[r367].c, ckcftp.c, 28 Dec 2005.

Built on Unixware 7.1.1, a pure 32-bit build, seems fine.  Rebuilt on Red
Hat AS 4.2 just to make sure I didn't break anything, it's OK.  No testing
on HP-UX, etc, because HP testdrive file sytem is full, can't upload anything.
29 Dec 2005.

Commented out the SHOW FEATURES section that displays constants like
INT_MAX, CHAR_MAX, etc, because printing each value in the appropriate
format is too tricky, and we don't need them anyway.  ckuus5.c, 29 Dec 2005.

Updated ckvfio.c to use CK_OFF_T for the relevant variables.  Built and
tested on VMS/Alpha 7.2: file transfer in remote mode; making a Telnet
connection and then local-mode file transfer; S-Expressions, all OK.  Also
built a no-net version OK.  29 Dec 2005.

Built and tested on Red Hat AS4 AMD X86_64, used it to upload new sources to
FreeBSD 4.11.  Built on FreeBSD 4.11/i386.  Here's another one where off_t
is 64 bits, even though long is 32 bits.  But it seems to work ok, not sure
why, when CK_OFF_T is 32 bits.  There is no _LARGEFILE_SOURCE stuff in the
header files.  29 Dec 2005.

Built on Mac OS X 10.3.9 using the new macosx10.4 target to pick up LFS.
Works fine.

Built on Red Hat Linux 4WS on IA64 (64-bit).  Now this one is odd, stat()
fails on big files.  It happens also if I use the "linuxnolfs" target, which
does not define _USE_LARGEFILE or _FILE_OFFSET_BITS=64.  DIRECTORY BIGFILE
shows the size as -1, but if "log debug", it says "no files match", i.e.
different behavior, observer effect.  I hate when that happens.

Let's see if that's an anomoly...  Built on Tru64 Unix 4.0F (64-bit Alpha).
It sees long files just fine.  Rebuilt and checked on x86_64 again... fine.
OK, let's not worry about IA64 yet.

Another small fix to the HP-UX 9.0 target from PeterE.  makefile, 29 Dec 2005.

---Dev.10---

Code adjustments from Jeff, mainly to the SSL and TLS Raw mode code from
several weeks ago, plus changing some data types in the security code to
CK_OFF_T, plus a different data type for CK_OFF_T for K95 because Windows
size_T isn't signed.  This presumably will allow large-number arithmetic but
it will not give large file access because that will require replacing all C
library file i/o calls (esp. in ckofio.c) with native Windows APIs.  Build
on Solaris 9 with and without SSL and on Linux RH AS4.2 with and without
SSL.  ck_crp.c, ck_ssl.c, ck_ssl.h, ckcdeb.h, ckcftp.c, ckcmai.c, ckcnet.c,
ckcnet.h, ckctel.c, ckuat2.h, ckuus4.c, ckuus7.c, ckuusr.c, 30 Dec 2005.

It was reported that WRITE SESSION always returned a failure status, even
when it succeeded.  The problem was that Unix versions of zsout() and
zsoutl(), for the session log only, were using write() and returning
write()'s return code, which is different from what zsout() and zsoutl() are
documented to return.  Also plugged a couple potential holes in zsoutx()
that I noticed while I was in the neighborhood.  ckufio.c, 30 Dec 2005.

Added FSEEK /FIND:pattern.  This form of FSEEK accepts all the other
switches and arguments and performs the desired seek.  Then, if the seek was
successful, it starts from that point and reads through the file, line by
line, searching for the first line that contains the given string or matches
the given (unanchored) pattern and, if found, sets the file pointer to the
beginning of that line.  Useful, e.g., for very long timestamped logs, where
you want to start processing at a certain date or time; searching for a
particular string is much faster than doing date comparisons on each line.
ckuus[27].c, 30 Dec 2005.

It was annoying me that FILE STATUS (FSTATUS) required a channel number to
be given even if only one file was open, so I supplied the correct default
in that case.  ckuus7.c, 30 Dec 2005.

INPUT /NOWRAP, added recently, is used for efficiently copying the INPUT
stream intact, but it's not good for matching because if the INPUT target is
broken between the end of the previous buffer and the beginning of the next
one, the context is lost and the match does not occur.  I thought of several
ways around this, but they all involve saving a huge amount of context --
old input buffers, the arrays of target strings and corresponding match
positions, etc.  The alternative is fairly simple but it's not transparent
to the user.  Here's what I did in a POP script:

    .eom := "\13\10.\13\10"
    set flag off                           # FLAG ON = success
    while ( open connection && not flag ) {
        .oldinput := \fright(\v(input),8)  # Save tail of previous INPUT buffer
        input /clear /nowrap 4 \m(eom)     # Get new INPUT buffer
        if success {                       # INPUT matched - good
            .s := {\freplace(\v(input),\m(eom),\13\10)}
            set flag on
        } else {                           # No match
            .s := \v(input)                # Check if target crossed the border
            .oldinput := \m(oldinput)\fsubstr(\v(input),1,8)
            if \findex(\m(eom),\m(oldinput)) set flag on
        }
        ...
    }

I think this will be easier to explain than any dangerous and grotesque
magic I might put into doinput() itself.  For now, added a few words about
this to HELP INPUT.  ckuus2.c, 30 Dec 2005.

Back to the pattern matcher.  Noticed that "IF MATCH index.html [a-hj-z]*"
succeeded when it should have failed.  In ckmatch(), the clist section
needed one more clause: it can't float the pattern if an asterisk does not
occur in the pattern before the clist.  This change fixes the problem
without breaking any other cases that weren't already broken, most of which
involve slists, i.e. {string,string,string,...}.  ckclib.c, 30 Dec 2005.

Tried FSEEK /FIND: on a largish file (over 100,000 lines), using it to seek
to a line near the end.  It took 0.756 seconds, compared with Unix grep,
which did the same thing in 0.151 sec.  That's because C-Kermit is using
ckmatch().  But if the search target is not a pattern, it should be a bit
faster to use ckindex().  Yup, 0.554 sec, a 36% improvement.  Can't expect
to compete with grep, though; it's highly tuned for its single purpose.
ckclib.[ch], ckuus7.c, 1 Jan 2006.

Updated visible copyright dates to 2006: ckcmai.c, ckuus2.c, ckuus5.c,
1 Jan 2006.

Noticed that NetBSD 2.0.3 has 64-bit off_t, and that _LARGEFILE_SOURCE is
mentioned in <stdio.h>.  Tried building Kermit with _LARGEFILE_SOURCE added
to CFLAGS, it's good.  Added it to the netbsd target.  makefile, 1 Jan 2006.

Fixed typo, #ifdef CK_NOLONGLONG in ckuus5.c should have been #ifndef
CK_LONGLONG (which, it turns out, we don't use anyway).  2 Jan 2005.

Observed that FreeBSD 4.x has a 64-bit off_t, but does not use the
_LARGEFILE_SOURCE convention.  Reasoning that all versions of FreeBSD have
off_t (I was able to check back to FreeBSD 3.3), I simply #define CK_OFF_T
to be off_t in ckcdeb.h within #ifdef __FreeBSD__ .. #endif.  Another one
down.  This can be done for any platform that is guaranteed to have off_t.
Turns out FreeBSD 3.3 has 64-bit off_t too.  2 Jan 2005.

OpenBSD, same as FreeBSD.  Also, added OS-version-getting thing to makefile
target for the program herald, as in the other BSDs.  Built on OpenBSD 2.5
from 1998, it has 64-bit off_t too.  ckcdeb.h, makefile, 2 Jan 2005.

Dumping the command stack every time there's an error is really too much.
I added SET COMMAND ERROR-DISPLAY {0,1,2,3} to set the verbosity level of
error messages.  Only level 3 dumps the stack.  ckuus[235].c, 2 Jan 2005.

Built on HP-UX 11.11 with _LARGEFILE_SOURCE and _FILE_OFFSET_BITS=64.  The
result works fine as far as I can tell.  It sees big files, it can open
them, seek to positions past the 2^31 boundary.  It can send large files.
It can do large-number arithmetic (^ 2 62).  The only problem is that during
compilation, every single modules warns:

  cc: "/usr/include/sys/socket.h", line 504: warning 562: Redeclaration of
  "sendfile" with a different storage class specifier: "sendfile" will have
  internal linkage.
  cc: "/usr/include/sys/socket.h", line 505: warning 562: Redeclaration of
  "sendpath" with a different storage class specifier: "sendpath" will have
  internal linkage.

These warnings should be perfectly harmless since they are not coming from
C-Kermit code, nor does C-Kermit use either one of those functions.  These
warnings don't come out in HP-UX 11i v2, but on that one we get tons and tons
of picky compiler warnings (variables set but not used, defined but not
referenced, etc).  A couple, however, turned out to be valid; one case of
"expression has no effect", and two of "string format incompatible with
data type" (I missed a couple file-size printfs).

There were also numerous warnings about signedness mismatch or sign
conversion of constants like IAC (0xff).  Does the HP-UX Optimizing Compiler
have a compiler flag to make all chars unsigned?  Yes, +uc, but the man page
says "Be careful when using this option.  Your application may have problems
interfacing with HP-UX system libraries and other libraries that do not use
this option".  Sigh, better not use it.

After reviewing "HP-UX Large Files White Paper Version 1.4" and HP's
"Writing Portable Code" documents, I added -D_LARGEFILE_SOURCE
-D_FILE_OFFSET_BITS=64 to the hpux1000 target, which is the basis for all
HP-UX 10.00 and later builds.  Large files are available in HP-UX 10.20 and
later.  10.00 and 10.10 were not real releases, and anyway these flags
should be harmless there unless the large-file implementation was only
partly done.  Built OK on both PA-RISC and IA64, optimized and plain.
makefile, 4 Jan 2006.

Built on FreeBSD 6.0 on IA64.  All OK except I got a warning about the
argument passed to time() in logwtmp() in ckufio.c.  This section had
already been partially fixed; thus I put the improved version into
#ifdef CK_64BIT, which is our newly available symbol that should be
automatically defined for any true 64-bit build.  ckufio.c, 4 Jan 2006.

Finally got around to testing Jeff's changes to SSL/TLS RAW mode from
December 30th against our POP server.  It didn't work, couldn't log in.
Tried backing off the ckctel.c changes first; that allowed login and
communication, but it did not suppress activation of Telnet protocol
whenever a 0xff byte arrived.  Backed off the rest of the changes and now
all is OK again.  ckctel.c, ckcnet.c, ckuus7.c, 9 Jan 2006.

Built on NetBSD 1.4.1 (1999), found that it did not like the large file
assumption -- fseeko() and ftello() do not exist; added a clause to the
netbsd target to check for fseeko and not define _LARGEFILE_SOURCE if not
found.  Oddly enough, off_t is 64 bits anyway, but it doesn't look like the
APIs are half-done.  For example, stat() uses off_t (64 bits) for the file
length, but fseek() uses long (32 bits) and there is no 64-bit analog.
Anyway the new netbsd target works on both 1.4.1 and 1.5.2 (no large files)
and on 2.0.3 (large files).  makefile, 9 Jan 2006.

Built on QNX-32 4.25, which has no large file support.  Got a few strange
compiler (WatCom) warnings, but it built and runs OK.  Noticed that file
transfers into QNX over a Telnet connection can't use streaming, but that's
nothing new to this version; same thing happens with C-Kermit 7.0.  9 Jan 2006.

Built on IRIX 6.5.  I didn't bother with large files there because it does
not support the _LARGEFILE_SOURCE interface; you have to change all the APIs
at the source level from blah() to blah64().  Seems to work fine as a 32-bit
app even though its off_t is 64 bits.  Tried a pure 64-bit IRIX 6.5 build
but it dies in ckcnet.c when it hits SOCKOPT_T and GSOCKNAME_T with "The
identifier 'socklen_t' is undefined".

Looks like I no longer have access to SCO OSR5.

Made a pure 32-bit build on SCO UnixWare 7.1.4, all OK.  Found that this
version also supports LFS, added it to the uw7 target.  makefile, 9 Jan 2006.

--- Dev.11 ---

Evidently the HP-UX bundled (non-ANSI non-optimizing) compiler doesn't like
long integers in switch expressions.  Changed three examples of these in the
S-expression code.  ckuus3.c, 10 Jan 2006.

A section of tstats() where GFTIMER isn't defined (e.g. on Motorola
sv68r3v6) was garbled.  Fixed in ckcfn2.c, 10 Jan 2006.

A fix for setting 921600 bps on Linux from Paul Fulghum, Microgate Systems Ltd.
ttgspd(): ckutio.c, 11 Jan 2006.

Noticed that when I changed the compact substring notation code back on
August 9th, I broke the ability to use arithmetic expressions within the
brackets, which explains some rather odd behavior I saw with some of my
scripts.  Looking more deeply into this, I also see that all the parsers I
have been using up to now for this, as well as for array bounds pairs, have
been inadequate because they never allowed for nested constructions, such as
a member of a bounds pair that itself was an array element, possibly with
another array element as a subscript.  I wrote a new routine for this,
called boundspair(), which is like arraybounds() except it accepts an extra
argument, an array of characters that can serve as bounds-pair delimiters,
and it returns the pair separator that was encountered in another new
argument.  For the alternative substring notation for [startpos-endpos] I
had to change the delimiter from '-' to '_' because '-' can be used in
arithmetic but '_' is not a recognized operator.  This is so I can parse,
e.g. [a:b] or [a_b] in the same context, and then find out which form was
used, e.g. \s(line[9:12]) or \s(line[9_12]); the first string is 4 bytes
long, the second is 12.  Everything seems to be OK now.  \s(line[10]) gives
everything starting at 10, but \s(line[10:0]) gives the null string.  Bad
syntax in the bounds pairs results in a null string; missing pieces of the
bounds pair result in defaults that should be compatible with previous
behavior.  ckuus[45].c, ckuusr.h, 13 Jan 2005.

Changed arraybounds() to call boundsdpair().  This was a rather drastic
change, not strictly necessary, but I think I got all the kinks out.
ckuus5.c, 13 Jan 2005.

Changes from PeterE to the makefile for HP-UX 6 and 7, to accommodate bigger
symbol tables, etc.  19 Jan 2005.

Determined that SCO OSR5.0.6 (and earlier) do(es) not support large files.
Don't know about 5.0.7.  30 Jan 2005.

Created a new build target for SCO OSR6.0.0.  Gets the exact 6.x.x version
dynamically.  Supports large files and big-number arithmetic via CK_OFF_T.
The sockopt() family of functions changed the data types of some of their
arguments since OSR5.  It was already possible to define SOCKOPT_T and
GSOCKNAME_T from the command line but I had to add code to also allow this
for GPEERNAME_T too.  ckcnet.c, makefile, 30 Jan 2005.

Apparently, ever since C-Kermit 7.0 was released, it has never been possible
to use a variable for the as-name in a RECEIVE command in Kermit 95.  This
is because evaluation of the as-name field was deferred until after we could
check whether it might be a directory name (which, in Windows, could start
with a backslash).   This little bit of magic was not a good idea, magic
hardly ever is.  I changed the code to evaluate both as-name fields in the
normal way.  If they want to receive to a directory called "\%1", they'll
just have to spell it differently.  The workaround is to turn the whole
command into a macro and evaluate it before executing it, e.g.:

  assign xx receive /as-name:\%1
  do xx

ckuus6.c, 1 Feb 2006.

Built OK on FreeBSD 6.1 on AMD64.  Adjusted some copyrights and date stamps.
ckcmai.c, makefile, 8 Feb 2006.

--- Dev.12 ---

Fixed a signed/unsigned char warning in the new boundspair() calling code
in the compact substring notation handler.  ckuus4.c, 9 Feb 2006.

Removed a spurious extra linux+openssl label from the makefile, added
solaris10g_64 synonym.  9 Feb 2006.

Satisfied myself that LFS is OK on Solaris 10 i386, and I'm going to assume
it's also OK on Solaris 9.  Made LFS standard for all Solaris 9 and 10
builds (including the secure ones) except the explicitly 64-bit ones, and
made the provisional solarisXXlfs targets into synonyms.  makefile, 9 Feb 2006.

--- Dev.13 ---

Further attempts at SSL/TLS message suppression when QUIET is ON.
ck_ssl.c, 16 Feb 2006.

From J.Scott Kasten: (quote...) I just uploaded a patch to /kermit/incoming.
The file name is "jsk-patch-for-cku211.diff".  I have also included the
patch as ASCII text in this email below.  This patch may be applied to the
cku211.tar.gz source code via:
  cd cku211, patch -p1 <../jsk-patch-for-cku211.diff
The patch adds 4 new build targets:
  netbsdwoc - a stripped no curses target for iksd used.
  netbsdse  - security enhanced target with srp, ssl, and zlib.
  irix65gcc - build on SGI Irix 6.5 platform using gcc.
  irix65se  - security enhanced target with srp, ssl, and zlib.
The patch fixes one build target:
  irix64gcc - The "-s" option is not supported by gcc under Irix.
I thank all of you in the Kermit Project for such a fine utility.  I
recently had to get a 16 MB file overseas across a spotty communications
link to repair a computer remotely.  Kermit was the only thing that could do
the job, so I wanted to contribute these patches back to the mainstream to
say thanks.  This digitally signed email is a binding contract that
officially assigns the rights to the source code patch (shown below) that I
developed to the Kermit Project at Columbia University. (...end quote)
ck_ssl.c, makefile, 23 Feb 2006.

Changed the new NetBSD target names to be consistent with the conventions
used in most other targets:

  netbsdwoc -> netbsdnc
  netbsdse  -> netbsd+ssl+srp+zlib
  irix65se  -> irix65+ssl+srp+zlib

and removed old, now superfluous, NetBSD targets (old-netbsd, netbst15,
netbst16), leaving synonym labels in their place.  Also updated (crudely)
the Linux target variations (curses instead of nocurses, no curses at all)
to be (appropriately modified) copies of the current linux target.  It would
be nicer to combine them, but this gets the job done.  makefile, 23 Feb 2006.

--- Dev.14 ---

Fixed the HELP command when used with tokens like @, ^, #, and ;.  The first
two had been omitted from the table.  The second two required a new path
into the guts of the parser, since comments are normally stripped at a very
low level.  ckuus[r2].c, ckucmd.c, 24 Feb 2006.

Built on AIX 5.1 ("make aix51") without incident.  Then I tried:

 make aix51 "KFLAGS=-D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64"

This had no effect.  I found the relevant document ath the IBM website.  It
says to use -D_LARGE_FILES instead.  I added this to the AIX 4.2 target
since (a) IBM says large files are supported by AIX 4.2 and later, and (b)
all Kermit AIX targets past 4.2 use the 4.2 one.  Plus a clause to make
sure CK_OFF_T is defined appropriately.  ckcdeb.h, makefile, 6 Mar 2006.

Added a 32-bit aix51+openssl target.  Builds OK, works fine (tested against
our SSL POP server).  Tried I tried adding -D_LARGE_FILES.  It seems to work
fine, so we'll keep it.  Cleaned up the other aix5blah entries a bit also.
makefile, 6 Mar 2006.

Fixes from J. Scott Kasten to the IRIX 6.4 and 6.5 makefile targets.  They
were badly wrong.  makefile, 6 Mar 2006.

The reason Kermit was looping on directories in IRIX was a classic
"double-ended break".  The makefile targets failed to define DIRENT so
Kermit was open/read on directories rather than opendir()/readdir().  But
then it was also failing to account for the fact that read() would return -1
on error.  The makefile fix adds -DDIRENT, and the read() case in traverse()
now properly terminates its loop on error.  ckufio.c, 6 Mar 2006.

--- Dev.14 ---

In response to a complaint that C-Kermit would not build on HP-UX 11 with
OpenSSL, I tried it myself on both 11.11/PA-RISC and 11i v2/Itanium.  It built
OK on both but I had to add a new target (hpux1000o+openssl-nozlib) for no
Zlib since these boxes did not have it installed.  makefile, 9 Mar 2006.

Added OpenSSL version number display to SHOW FEATURES.  ckuus5.c, 9 Mar 2006.

Gavin Graham noticed that FTP [M]GET /DELETE /MOVE-TO: was rejected with
"?Sorry, /DELETE conflicts with /MOVE or /RENAME".  This check belongs in the
PUT code but not in the GET code.  Commented it out and tested the result.
The combination is now accepted but then Kermit refuses the incoming file as
if it had been given a /SMALLER-THAN: or /LARGER-THAN: switch, which it didn't
happen.  Turns out there was one more place where I wasn't initializing the
new "wide int" member of the switch-parsing pv[] struct.  Once this was fixed,
the /MOVE-TO part still didn't work.  Turned out the /DELETE case was part of
a long if-else-if-else- chain, which effectively made /DELETE and /MOVE-TO: or
/RENAME-TO: mutually exclusive.  Fixed this, now it works fine.  ckcftp.c,
13 Mar 2006.

Got access to AIX 5.3, built there, all OK, including large files. 13 Mar 2006.

--- Dev.16 ---

Patches from Mark Sapiro to suppress getsockopt() and getsockname() warnings
in Mac OS X.  ckcnet.[ch], 18 Mar 2006.

In response to a complaint from Clarence Dold, tried "make redhat9" (which
is the rather dated target that tried to include all forms of security) on
RH Linux AS4.3, it failed miserably.  I made a new makefile target, removing
Kerberos IV and got a lot farther.  But then in ckcftp.c, the following
struct definition:

  struct {
      CONST gss_OID_desc * CONST * mech_type;
      char *service_name;
  } gss_trials[] = {
      { &ck_gss_mech_krb5, "ftp" },
      { &ck_gss_mech_krb5, "host" },
  };

refers to a variable, ck_gss_mech_krb5, that is not defined anywhere.  Up
above, however, is a static definition for gss_mech_krb5, so I changed the
struct definition to match.  Next, in ckuath.c, the compiler could not find
the com_err.h file.  Turns out in Linux this is in a subdirectory, et, so we
have to add a -I clause to the makefile target for this.  Made a target for
Linux+SSL only.  Made a target for Linux+Krb5 only; this required moving an
#ifdef in ckuus7.c to prevent an unguarded reference to SSLEAY_VERSION.
New targets: linux+krb5+ssl, linux+krb5, linux+krb5.  ckcftp.c, ckuus7.c,
makefile, 27 Mar 2006.  

New targets of HP-UX 10/11 with OpenSSL from PeterE.  makefile, 27 Mar 2006.

Added large file/integer support to SHOW FEATURES.  ckuus5.c, 27 Mar 2006.

Built OK on Solaris 9 and 10 with gcc (someone was complaining that this
didn't work, but that was 8.0.211).

Started build on a Sun 3/80 mc68030 with NetBSD 2.0 and gcc 3.3.3.  But it
died with an assembler error in ckcfn2.c (compiler bug).  27 Mar 2006.

--- Dev.17 ---

NebBSD 2.0 build completed by turning off optimization on ckcfn2.c
("KFLAGS=-O0").  Result supports 64-bit ints and, presumably, large files.
uname -p = "m68k", -m = "sun3". 29 Mar 2006.

Corrected an omission in applying PeterE's updates to the HP-UX targets.
makefile, 28 Mar 2006.

solaris2xg+krb5+krb4+openssl+shadow:

Tried resurrecting the solaris2xg+krb5+krb4+openssl+shadow target.  It asks
to link with libdes but there is no libdes.  Removed -ldes from the target,
now at least it builds and runs wart.  The compilation blows up in ckcftp.c
for missing header files:

  ckcftp.c:462: kerberosIV/krb.h: No such file or directory
  ckcftp.c:500: gssapi/gssapi_generic.h: No such file or directory
  ckcftp.c:501: gssapi/gssapi_krb5.h: No such file or directory

Got a bit farther by adding appropriate -I's and -L's to KFLAGS but it still
dies compiling (or linking?) ckcftp.c, but it doesn't say exactly why.  OK,
deferred.

Added SET SEXPRESSION TRUNCATE-ALL-RESULTS { ON, OFF }.  This can be used
for force integer arithmetic in any kind of calculation that requires it,
such as date calculations.  This is a global setting, not on any kind of
stack.  Also, updated SHOW SEXP and added HELP SET SEXP which wasn't there
before.  ckuus[23].c, 30 Mar 2006.

To make the RENAME command a bit more useful, need to add some switches.
But it shares a switch table, qvswtab[], with some other commands.  Broke
this off into its own switch table.  ckuus6.c, 17 Apr 2006.

Added RENAME switch values that can be used in the same table with the DELETE
switch values, which are shared by many commands.  ckuusr.h, 17 Apr 2006.

Discovered that the RENAME command could be entered without any arguments
and it would still succeed.  Fixed in dorenam(): ckuus6.c, 17 Apr 2006.

Added parsing for RENAME /UPPER:option (to uppercase the file name(s)),
/LOWER:option (to lowercase), and /REPLACE:{{s1}{s2}} (to do string
replacement on the filename(s)), but not the semantics.  When any of these
switches is given, the target ("to") name is not parsed; they act on the
source name.  The /LOWER: switch takes keyword args to specify whether it
should act only only files that have all UPPER case latters, or on ALL files
(i.e., including files with mixed-case names); similarly for the /UPPER:
switch.  There is some creative parsing allowing these to be given with or
without a colon and keyword argument, which works fine except if you include
the colon but no argument, execute the command (which works fine), and then
recall the command.  I haven't yet decided about the interaction among these
switches.  Clearly if /UPPER is given after /LOWER, it overrides.  But if
/UPPER (or /LOWER) is given with /REPLACE, what should happen?  ckuus6.c,
17 Apr 2006.

Filled in actions for RENAME /UPPER: and /LOWER: for the single file case,
and tested all combinations of switch values and filename configurations.
Once that was OK, moved the code out into a separate routine, renameone(),
and then called it from both the single-file case and the multifile case.
ckuus6.c, 19 Apr 2006.

Added RENAME /SIMULATE.  Filled in the code for string replacement, needs
testing.  ckuus6.c, 20 Apr 2006.

Changed /REPLACE options to allow a negative number to specify an occurrence
from the right, so -1 means the last occurrence, -2 means the next-to-last,
etc.  ckuus6.c, 24 Apr 2006.

Added RENAME /COLLISION:{OVERWRITE,PROCEED,FAIL}.  This is implemented but
not tested.  ckuus6.c, 24 Apr 2006.

Worked on RENAME /COLLISION:FAIL.  I decided it was less than useful to ...

Added SET RENAME { COLLISION, LIST } to let user change default collision
and listing actions.  ckuusr.[ch], ckuus[36].c, 25 Apr 2006.

Experimented with parsing for /CONVERT:cset1:cset2.  The problem here is
that there is no straightforward way for a switch to have multiple
arguments.  Or is there...?  If I parse cset1 with cmswi() rather than
cmkey(), it almost works; the only problem is that the character-set
keywords don't have CM_ARG set, so they don't know to stop on, and ignore, a
colon.  If I make a copy of the table and set CM_ARG in the flags field for
each keyword, it works fine: if I Tab in the first name, it fills itself
out, supplies a colon, and waits for the second name.  So in the code, the
first time that RENAME /CONVERT is invoked, I put code to copy fcstab[] and
set CM_ARG in each flags field.  Works fine, and now we know how to make a
switch that takes multiple arguments.  ckuus6.c, 24 Apr 2006.

I thought I had a function to convert the character set of a string but I
don't, so actually implementing /CONVERT: will be difficult.

Actually the parsing wasn't that easy either.  It works OK interactively,
but not in a TAKE file.  To make a long story short, I had to change
gtword() and cmkey2() to not require "/" at the beginning of a switch, and
then to parse arguments-that-are-followed-by-other-arguments as if they were
switches, so that they can end with colon rather than space.  This might
seem dangerous, but switches always have "/" at the beginning, so the check
is superfluous.  ckucmd.c, 26 Apr 2006.

Back to /CONVERT...  Once I was able to get the code to call cvtstring() I
was able to debug it (at first it was skipping every second character).  And
now we have a general-purpose string-translating function we can call from
anywhere.  Requires that C-Kermit be built with Unicode support.
ckuus6.c, 26 Apr 2006.

Added SHOW RENAME.  ckuusr.h, ckuus[r5].c, 26 Apr 2006.

Conditionalized some Unix/Windows assumptions in renameone() so the code
could work in VMS.  ckuus6.c, 2 May 2006.

Added RENAME /FIXSPACES to change all spaces in the filename(s) to
underscore or any other character or string that is given.  This is just a
special case of RENAME /REPLACE:{{ }{x}} with easier syntax.
ckuusr.h, ckuus6.c, 2 May 2006.

Added an "all-but" control to the /REPLACE options:
/REPLACE:{{.}{_}{~1}} means replace all but the first (this one works);
/REPLACE:{{.}{_}{~-1}} means replace all but the last (this one not yet).
ckuus6.c, 2 May 2006.

Filled in the second one ("all but" the given occurrence).  The algorithm is
simply to reverse the three strings and then use the same code as we use in
the left-right-case, and then unreverse the result.  At first I used
yystring() for this but yikes, what a bad design!  So I made a better
string-reversal routine, gnirts(), for this (luckily yystring() is only used
in one place, for which its design is appropriate). ckuus6.c, 3-4 May 2006.

Added code to handle the case where the file being renamed includes a path
specification.  In this case we separate the path, apply the renaming
functions to the filename only, and then at the end rejoin the original
filename with the path, and join the new name with same path or, if a
destination directory was given, with that.  ckuus6.c, 4 May 2006.

Added HELP SET RENAME and updated HELP RENAME.  ckuus2.c, 4 May 2006.

"Tom Violin" (Tom Hansen) noticed that the first time you FOPEN a file,
Kermit's memory consumption goes way up.  In fact there's a warning to that
effect in the code, where, upon first open, a potentially big array of
potentially big structs is allocated.  I rewrote the code to allocate each
array member (struct ckz_file) as needed, i.e. when a file is opened, and to
free it when the file is closed (or the open fails).  This was actually
quite a lot of work, which is why I didn't do it the first time around:
every single "."  had to be changed to "->".  Every check for a valid
channel first had to check if the channel's struct was allocated and every
other reference to z_file[i]->anything had to be prechecked that z_file[i]
was not a NULL pointer.  Also I made some improvements to FILE STATUS, and I
fixed FILE CLOSE to default the channel number if only one channel was open,
as I did for FILE STATUS a while back.  ckuus7.c, Cinco de mayo 2006.

Ran my old BUILDS script that builds C-Kermit with about 100 different
combinations of feature-selection switches.  Fixed a few small glitches so
now they all build OK (except can't do NOANSI builds any more on recent
Linuxes because of varargs()).  ckuus3.c, ckuus5.c, ckuus6.c, ckuus7.c,
ckucmd.c, ckcfns.c, 6 May 2006.

Fixed RENAME /LOWER and /UPPER, when given with no colon or agrument, to
default to ALL.  ckuus6.c, 13 May 2006.

Built on VMS 7.2-1, tested new RENAME command there; seems to be OK.
13 May 2006.

--- Dev.18 ---

I wanted to test large files against RESEND but I don't have access to any
system that can run C-Kermit and that also has enough space for a large
file.  I created a "fake" large file on Linux (3G hole plus 1 byte), and
sent it over a localhost connection, and interrupted it repeatedly and then
initiated a RESEND at the sender.  In each case, it picked up where it left
off.  But before the 2G boundary was crossed the disk filled up.
Inconclusive.  14 May 2006.

PeterE got a warning in the new FILE OPEN code when building in HP-UX 9.
I added a cast, built on HP-UX 11, no more complaint.  However there
are warnings about internal vs external bindings of sendpath and sendfile
in every module.  Too bad, these are not Kermit tokens, it's a conflict in
HP's header files.  Marc Sapiro doesn't see them; probably it's something
on the HP testdrive site.  ckuus7.c, 17 May 2006.

Fixed the tru64-51b+openssl target -- the terminating doublequote of KFLAGS
was missing -- and also the osf target, which failed to import the LIBS
definition from whatever other target invoked it.  Now the SSL build goes OK
on Tru64 5.1B.  Replaced x.tar.z in the download areas without declaring a
new Dev number.  The new one has a makefile with today's date.  Software
engineering at its best!  makefile, 18 May 2006.

Scott Kasten noted that the estimated-time-remaining calculation would go
bonkers on LFS systems when RESENDing a large file.  It looks like the
shocps() and shoetl() functions escaped the CK_OFF_T conversion.  I made
what seemed to be the right adjustments, and then was lucky enough to find a
computer that had enough free disk space for me to send a large file,
interrupt it several times, resend it, all seems to be OK.  28 May 2006.
Later Scott verified these changes independently for Linux, but the problems
in IRIX remain.

Patches from Scott Kasten for large files on IRIX 6.5: ckcdeb.h, makefile,
12 Jun 2006.

--- Dev.19 ---

Added a new function for dealing with JPGs and GIFs:

\fpicture(filename,&a)
  returns 0 if file not recognized or can't be opened;
  returns 1 if landscape, 2 if portrait or square.
  If array given, element 1 is width, element 2 is height.

ckuusr.h, ckuus4.c, 19 Jun 2006.

Scott Kasten reports that the FTP client can transfer large files OK, at
least in Linux, but has trouble with recovery:

 . Kermit takes a very long time to start the transfer, sometimes over
   30 minutes.  Suspect the ftp server is counting the bytes in a long file?
   Or maybe it's a text-mode transfer and it's counting the lines?  Probably
   in response to Kermit's SIZE command.

 . The size shown in the FT display is wrong by a random amount.  And of
   course so are the progress bar, percent done, and time remaining.

 . The file, however, is transferred correctly.  REGET works correctly too.

I tried setting up a test scenario locally but our Solaris FTP server does
not support large files:

  FTP SENT [SIZE BIGFILE]
  FTP RCVD [550 BIGFILE: not a plain file.]
  FTP SENT [PASV]
  FTP RCVD [227 Entering Passive Mode (128,59,48,24,246,37)]
  FTP SENT [RETR BIGFILE]
  FTP RCVD [550 BIGFILE: Value too large for defined data type.]

Created the same 3GB on a Tru64 Unix system that allows FTP access.  Made
the connection from C-Kermit on Solaris (32-bit with LFS):

  16:46:12.908 FTP SENT [SIZE BIGFILE]
  16:46:12.947 FTP RCVD [213 3000000001]

Note that it takes less than half a second to get the reply.  Now I start
the download and then interrupt it at about 2%:

  16:46:12.979 FTP SENT [TYPE I]
  16:46:13.174 FTP RCVD [200 Type set to I.]
  16:46:13.226 FTP SENT [PASV]
  16:46:13.262 FTP RCVD [227 Entering Passive Mode (15,170,178,171,11,37)]
  16:46:13.299 FTP SENT [RETR BIGFILE]
  16:46:13.337 FTP RCVD [150 Opening BINARY mode data connection for BIGFILE..]
  16:47:24.895 FTP RCVD [426 Transfer aborted. Data connection closed.]
  16:47:24.934 FTP RCVD [226 Abort successful]
  16:47:24.991 FTP SENT [MDTM BIGFILE]
  16:47:25.028 FTP RCVD [213 20060706204458]

Now I do a REGET:

  16:51:55.321 FTP SENT [PASV]
  16:51:55.357 FTP RCVD [227 Entering Passive Mode (15,170,178,171,11,43)]
  16:51:55.394 FTP SENT [REST 122736640]
  16:51:55.430 FTP RCVD [350 Restarting at 122736640. Send STORE or RETRIEVE..]
  16:51:55.431 FTP SENT [RETR BIGFILE]
  16:51:55.469 FTP RCVD [150 Opening BINARY mode data connection for BIGFILE..]

This worked perfectly, as far as I can tell; the FT display picked up in the
right place; the thermometer, percent done, and estimated time remaining
were the same as when we left off last time.  I did the same thing several
more times, everything was OK.  It would have taken a really long time to
let this run to completion, but I think this demonstrates that Scott's
symptoms are server-dependent.  No changes.  6 July 2006.

Checked current code on VMS 8.2-1 on IA64 / UCX 5.5, builds fine.
No changes.  Updated listing at HP.  6 July 2006.

Checked FTP GET of large file in ASCII mode against Tru64 FTP server.  It
was fine, and there was no delay in the server's response to our SIZE command
(as there would be if it were scanning the entire file to count how many
bytes would be required to send it in text mode).  7 Jul 2006.

Tested FTP PUT big file against Tru64, OK.  Ditto FTP RESEND big file:

  C-Kermit>resend BIGFILE
   PUT BIGFILE (binary) (3000000001 bytes)---> PASV
  227 Entering Passive Mode (15,170,178,171,13,186)
  ---> SIZE BIGFILE
  213 343211280
  ---> MDTM BIGFILE
  213 20060707141243
  ---> APPE BIGFILE
  150 Opening BINARY mode data connection for BIGFILE (128.59.59.56,45470).

Made REPUT a synonym for RESEND.  ckuusr.c, 7 Jul 2006.

Added FTP REPUT and FTP RESEND since previously there was no FTP-prefixed
command for recovering uploads, only the regular RESEND command, which might
not have been obvious to people.  ckcftp.c, 7 Jul 2006.

Added help text for FTP RESEND and REPUT and amended RESEND help text.
ckcftp.c, ckuus2.c, 7 Jul 2006.

Changed name of \fpicture() to \fpictureinfo() and added help text.  By the
way, ImageMagick can do this too: identify -format "%w %h" dscf0520.jpg.
The advantage of having it in Kermit is that not everybody has ImageMagick.
ckuus[24].c, 7 Jul 2006.

Changed the numeric comparisons = < > <= >= != to allow long integers by
changing the data type to CK_OFF_T, etc.  ckuus6.c, 7 Jul 2006.

Noticed that \fkeywordvalue(foo=this is a string) only kept the first word.
Fixed it to keep the whole definition.  Also added \fkwvalue() as a briefer
synomym.  ckuus4.c, 7 Jul 2006

Sometimes we want to check if a file's status before we've FOPEN'd it, in
which case the channel variable is likely to be empty and \f_status(\%c)
would get an error.  Making the obvious change didn't fix this, however.  It
turns out that the function evaluator failed to adjust argn (argument count)
when trailing arguments were empty, and argn was being used in this case,
and probably others, to test whether an argument existed.  I added code to
adjust argn to reflect the number of aruments up to and including the
rightmost non-empty one.  ckuus4.c, 7 Jul 2006.

Fixed \fstripb() to not dump core if second argument is missing.
ckuus4.c, 7 Jul 2006.

Discovered that it was not obvious what pattern to use to match strings
enclosed in square brackets.  "if match [abc] \[*\]" didn't work.  Neither
did various other tricks like NCRs for the brackets.  However, "if match
[abc] \\[*\\]" does work.  Trying to fix this would no doubt break 100 other
things, so let's call it a feature.  7 Jul 2006.

Added \fgetpidinfo(n) to return info about a process ID; for now it simply
returns 1 if the process is alive and 0 if not (or -1 if the argument is
bad or missing or on any kind of error).  ckuusr.h, ckuus[24].c, 7 Jul 2006.

The "where-did-my-file-go" message seemed to be ending with a LF rather
than CRLF, probably because the terminal modes had not yet been restored,
leaving the next prompt hanging below it, rather than on the left margin,
if C-Kermit exited immediately after the transfer.  Fixed by changing
all \n's to \r\n's in wheremsg(): ckcpro.w, 8 Jul 2006.

Added \v(lastkwval) so we can retrieve programmatically the keyword most
recently processed by \fkeywordval().  ckuusr.h, ckuus4.c, 9 Jul 2006.

--- Dev.20 ---

Added #ifdef SV68, #include <unistd.h>, #endif because Unix System V/68 on
Motorola choked on the SEEK_CUR reference without it.  ckuus4.c, 10 Jul 2006.

Make \fkeywordval(xxx) undefine xxx (i.e. when a keyword is given with no
value).  This way command-line keywords will always override preexisting
default definitions, whether they have a value or not, which makes it easier
to parse command lines like "foo=bar blah xx=yy".  ckuus[24].c, 12 Jul 2006.

On 29 Nov 2005 I changed IF KERBANG to solve a problem (see entry for that
date), but introduced a new one; namely that you can't have (e.g.)  a FATAL
macro that uses IF KERBANG to decide whether to EXIT all the way or STOP
back to the prompt.  Changed it again, this time to require not that the
command level be 1, but that the command *file* level be 0 (i.e. that we are
in the top-level command file, irrespective of the command or macro level,
but not in a subfile).  ckuus6.c, 12 Jul 2006.

It is unhelpful when Kermit gets a syntax error in the middle of a big
compound statement block (e.g. FOR or WHILE loop) and dumps out the whole
thing in an error message.  I changed the two places where this can happen
to call a new routine that, instead of dumping out the entire cmdbuf,
checks its length first and if it's more than a line long, truncates it
and adds an ellipsis.  ckuus6.c, 12 Jul 2006.

The new RENAME command didn't give very good error messages, e.g. if the
filespec didn't match any files.  Fixed in dorenam(): ckuus6.c, 12 Jul 2006.

Fixed DIR /TOP to work if the /TOP:n argument was omitted, defaulting
to 10.  domydir(): ckuus6.c, 12 Jul 2006.

Added DIR /COUNT:v to count the number of files that match the given
criteria and store result in the variable v.  ckuusr.h, ckuus[r26].c,
24 Aug 2006.

Added HDIRECTORY as an invisible synonym for DIR /SORT:SIZE /REVERSE.
Can be used with other switches, of course, so (e.g.) HD /TOP shows the
ten biggest files.  ckuusr.h, ckuus[r26].c, 24 Aug 2006.

DIR /FOLLOWLINKS and /NOFOLLOWLINKS always did the same thing; the switch
was ignored, a symlink is always followed.  Fixed in ckuus6.c, 24 Aug 2006.

Added DIR /NOLINKS, which means don't show or count symlinks at all.
ckuusr.h, ckuus[r26].c, 24 Aug 2006.

Build on Solaris 9 and NetBSD 3.0, 24 Aug 2006.

Added a missing definition for LOCK_DIR in the Linux HAVE_BAUDBOY case,
suggested by Gerry Belanger.  ckutio.c, 6 Oct 2006.

Suggested by Jim Crapuchettes: \v(dialmessage) is the text string
corresponding to \v(dialstatus).  ckuusr.h, ckuus4.c, 6 Oct 2006.

Soewono Effendi sent code for exit sequence to leave DTR on; this amounted
to unsetting HPUCL in c_cflag.  I did it a simpler way, hopefully portable
to all Unixes, but who knows at this late date.  The code is inside
#ifndef CK_NOHUPCL..#endif in case it causes trouble.  It is executed if
SET EXIT HANGUP is OFF and a serial port was open at the time Kermit exits
(or closes it explicitly).  ttclos(): ckutio.c, 6 Oct 2006.

Built on Solaris9/Sparc; FreeBSD 6.2/AMD64; NetBSD 3.0/i386; HP-UX 11i v2;
SCO OSR6.00.

--- Dev.21 ---

Added netbsd+openssl target to makefile.  Built OK (NetBSD 3.0, OpenSSL
0.9.7d) except with some warnings in ck_crp.c.  Connects and logs in OK to a
secure site.  10 Oct 2006.

Added a debug statement to ftp_hookup() to record the TCP port that was used.
ckcftp.c, 11 Oct 2006.

Built with OpenSSL 0.9.7l on Solaris 9.  Built with OpenSSL 0.9.8d on
Solaris 9; connects and logs in to a secure site.  11 Oct 2006.

The new RENAME command didn't work if both the source and destination names
included directory segments, e.g. "rename /tmp/foo ~/bar" (see notes of
4 May 2006).  This was fixed in renameone() by a special case in which
the second argument is given but it is a filename, not a directory name.
ckuus6.c, 11 Oct 2006.

Fixed unguarded reference to dialmsg[] for \fdialmessage(), noticed by
Gerry Belanger.  ckuus4.c, 12 Oct 2006.

Added a TOUCH command that does what UNIX touch does: creates the file if it
does not exist, updates the timestamp if it does.  If a wildcard is given,
it operates only on existing files.  It shares the DIRECTORY command parser,
so all the same file selection switches can be given.  ckuusr.[ch],
ckuus[26].c, 12 Oct 2006.

PeterE noticed that if you FOPEN a file, do some seeks or reads, then FCLOSE
it, then FOPEN it again (or open a different one), some of the old
information is still there (e.g. current line number).  This is an artifact
of the changes of May 4th.  Now the file closing and opening routines are a
bit more careful about scrubbing and initializing the file info struct.
ckuus7.c, 12 Oct 2006.

--- Dev.22 ---

Built OK on Red Hat Linux AS4 with both "make linux" and "make linuxnc". 
15 Oct 2006.

DIRECTORY /BRIEF ignored file selection switches and always listed all
files.  This was because of how I cleverly called filhelp() (the routine
that lists matching files when ? is typed in a filename field) and, of
course, filhelp() doesn't know anything about the DIRECTORY command's file
selections.  Changed filhelp() to accept all the args needed for passing
along to fileselect(), renamed it to xfilhelp(), and made a filhelp() stub
that chains to xfilhelp() with null selections.  ckcker.h, ckucmd.[ch],
ckuus6.c, 29 Nov 2006.

SHOW CONNECTION for an SSH connection said the connection type was "NET"
rather than "SSH".  Fixed in dolognet(): ckuus3.c, 29 Nov 2006.

SHOW CONNECTION didn't show the TCP port number.  This command works by
parsing the current connection log entry string, which doesn't have a field
for this, but which sometimes shows the port number as part of the hostname
(but more often not).  Added code to dolognet() to log the TCP port number,
if known.  This involved adding a gettcpport() function to ckcnet.c.
ckcnet.[ch], ckuus3.c, 29 Nov 2006.

This was impossible: def \%1 upper, echo \f\%1(abc) -- i.e. to "compose" a
function name.  Fixed in zzstring().  But note that it's still not possible
to do this: def \%1 \fupper, echo \%1(abc) -- because at the point where
"\fupper" is encountered, which is automatically fed to fneval(), the
argument list hasn't been read yet.  ckuus4.c, 29 Nov 2006.

The meaningless Lisp command (=) would cause Kermit to hang.  Due to some
idiosyncracy in the parser, it would see this as ((=) and would go into
"wait for the closing paren" mode.  There was already a hack in the code to
compensate for this, but it didn't work.  I fixed the hack but I don't
understand the real problem.  Anyway, comparing Kermit with real (Franz)
Lisp I discovered that comparison operators do not require two arguments, as
Kermit has been doing, although they do require at least one.  I changed
Kermit to not require two, so now all the comparison predicates behave
exactly like Franz Lisp, including getting an error if there are no args).
ckuus[r3].c, 29 Nov 2006.

From to-do list: Make a way to inhibit pattern matching in SWITCH labels.
It's already there; just quote the wildcard characters; the only trick is
that for some reason (such as that SWITCH is really an internally defined
macro), a double quote is needed:

  switch \%1 {
    :a\\*z, echo literally "a*z", break
    :abcxyz, echo literally "abcxyz", break
    :a*z, echo a...z, break
    :default, echo NO MATCH
  }

In first case, the asterisk is taken literally; in the third it's a
metacharacter and the label matches any string that starts with 'a' and
ends with 'z'.

Array initializion would quit early if any initializers were undefined,
e.g. "decl \&a[] = \%a \%b \%c" would stop at the first element if \%b
was not defined, even though \%c might be defined.  Fixed in dodcl():
ckuusr.c, 30 Nov 2006.

DIR /ARRAY:a filespec, when the filespec does not match any files,
terminates with the array undeclared.  It would be better to return a
declared but empty array (\&a[0] = 0).  The code is already there to do
that, but isn't working.  And yet "declare \&a[0]" does indeed create a
0-element array ("show array" shows a dimension of 0).  Turns out there were
two problems; one was the careless recycling of a local variable ("array"),
resulting in failure to create \&a[] (but not any other array).  Fixed in
domydir(): ckuus6.c, 30 Nov 2006.

The other problem was that dclarray(), when called with an array name and a
dimension of zero, does two different things depending on whether the array
already existed.  There is still a fair amount of confusion about whether a
dimension of 0 indicates an array with 1 element (as it should) or a
nonexistent array.  We call dclarray() with a size of 0 to undeclare an
array but we also need to able able to declare an array with only element 0.
I changed dclarray() to treat a negative dimension as a command to destroy
the array, and 0 or positive as a command to create the array with the given
dimension.  ckuus[r56].c, 30 Nov 2006.

Next problem: when chkarray() returns 0, this should not be interpreted to
mean the array does not exist.  Looks like the only place this happened was
in \fcontents(); fixed in ckuus4.c, 30 Nov 2006.

If we include file selectors with DIR /ARRAY:&a and some of the files that
match the given filespec but don't fit the selectors, the array's dimension
is bigger than its number of elements.  Added code at the end of domydir()
to resize the array so \fdim() returns the number of filenames in the array,
and also made sure that element 0 contains that number too.  ckuus6.c,
30 Nov 2006.

This would be a nice elegant way to loop over a bunch of files, if it worked:

  for \%i 1 \ffiles(*) 1 { rename \fnextfile() xxx_\flpad(\%i,3,0) }

But in this loop, Kermit skips every other file (beginning with the first)
and then runs out of files halfway through the loop.  Why?  Because in
commands like RENAME and DELETE, the filename parser is in a chained FDB
with the switch parser.  First the switch parser, cmswi(), gets its hands on
\fnextfile(), passing it through the evaluator and thus getting the first
filename, which it then sees is not a switch, so now the field is parsed by
the next parser in the chain, cmifi(), which causes \fnextfile() to be 
executed again.  In fact, the FOR loop has nothing to do with; the same
thing happens like this:

  void \ffiles(*)
  delete \fnextfile()

This deletes not the first file, but the second one.  Obviously users can be
told not to refer to \fnextfile() in chained-fdb fields:

  for \%i 1 \ffiles(*) 1 { .f := \fnextfile(), delete \m(f) }

but this is hardly intuitive.  I had some clever ideas of how to make
\fnextfile() work as expected in this context but it's way too much magic.
Better to simply document that \fnextfile() is "deprecated" and the array
format should be used:

  for \%i 1 \ffiles(*,&a) 1 { delete \&a[\%i] }

The difference is, an array element doesn't change every time it's referred to!

Added a /PRESERVE switch to the COPY command to preserve the timestamp and
permissions of the file.  I did this using the Kermit APIs so it should work
for any version of C-Kermit or K95.  ckuus[26].c, 30 Nov 2006.

Added COPY /OVERWRITE:{ALWAYS,NEVER,OLDER,NEWER} to control name collisions
when copying across directories.  ckuus[26].c, 1 Dec 2006.

--- Dev.23 ---

Fixed a bug in SET TELNET PROMPT-FOR-USERID, SET AUTH KERBEROS[45] PROMPT,
and SET AUTH SRP PROMPT in which the user's string was compared with a
literal (s == ""), reported by Pavol Rusnak.  Worse, empty strings (if the
test succeeded) were turned into null pointers, and then fed to strlen().
Fixed in ckuus3.c, 5 Dec 2006.

Added an optional 4th argument to \findex(), \frindex(), \fsearch(), and
\frsearch(): the desired occurrence number of the searched-for string.
\frsearch() was a bit tricky.  ckuus[24].c, 7 Dec 2006.

Added \fcount(s1,s2) to tell the number of occurrences of s1 in s2.
ckuus[24].c, 8 Dec 2006.

Added \ffunction(s1) to tell if a given built-in function is available.
ckuus[24].c, 8 Dec 2006.

Changed RENAME /COLLISION:PROCEED to be /COLLISION:SKIP, which is clearer.
ckuus[26].c, 8 Dec 2006.

For communication protocols: INPUT /COUNT:n to read exactly n characters
without any matching.  Can be used, for example, with CONTENT_LENGTH in CGI
scripts; NUL characters are counted but not collected.  ckuusr.[ch],
ckuus4.c, 8 Dec 2006.

There was a bad bug in the date-parsing routines; it's been there for years.
If a date string includes a timezone, e.g. "Sat, 9 Dec 2006 19:26:23 EST",
and converting to GMT changes the date, the variables for day, month, and
year (which are used later) were not updated, and the final result was a day
off.  Fixed in cmcvtdate(): ckucmd.c, 10 Dec 2006.

Built OK with SSL/TLS.  Tested with the POP script, found that I broke INPUT
when adding the /COUNT feature; there was a path through the code that could
leave the "anychar" variable unset and therefore random.  Fixed in
doinput().  The POP script, which does not use /COUNT, works again and so
does a new CGI script, which does use /COUNT.  ckuus4.c, 10 Dec 2006.

Supplied a missing comma in the help-text array for HELP SET TERMINAL, which
resulted in bad formatting in K95 around SET SNI-FIRMWARE-VERSIONS.
ckuus2.c, 10 Dec 2006.

Made "help locus" a synonym for "help set locus".  ckuusr.[ch], ckuus2.c,
11 Dec 2006.

This morning the Columbia FTP server was malfunctioning in a perfect way
for me to implement and test an FTP timeout mechanism.  The server would
close the data connection after sending the file, but the client never saw
the close and was stuck forever in a recv().  I added code to do a select()
on the data connection prior to entering the recv(), with a timeout on the
select() that the user can establish with SET FTP TIMEOUT.  Built and tested
on Solaris 9, clear-text FTP.  Also built cleanly for FTPS and tested
against a server that does not hang; I don't have access to an FTPS server
that would tickle the timeout code.  ckcftp.c, 11 Dec 2006.

--- Dev.24 ---

Fixed a bug in the INPUT /COUNT: parser: the array of search strings was
never initialized, which didn't matter before, but with /COUNT:, if the
first element was not a NULL pointer, we'd treat it as a search string, and
then if it happened to match something in the input stream, the operation
would stop before the count was exhausted.  Fixed by (a) initializing the
array, and (b) ignoring any search strings if /COUNT: was given.  ckuusr.c,
13 Dec 2006.

Removed a debug() statement from zsattr() that suddenly started making some
version of gcc complain, reported by Gerry Belanger.  ckufio.c, 13 Dec 2006.

--- Dev.25 ---

Some casts for the 3 interior args of the new select() call in ckcftp.c
for HP-UX 9.  14 Dec 2006.

Changed \fkeywordvalue() to accept a string rather than a single word
as its second argument, so that more than one separator could be specified,
and to return -1 on error, 0 if it found nothing, 1 if given a kewyord but
no value, and 2 if there was a keyword and a value.  dokwval(): ckuus[24].c,
14 Dec 2006.

Checked FTP timeout on command channel with FTP DIRECTORY of a big directory
using a path into our ftp server that preserves the hanging behavior.  The
timeout was actually working, but the failure condition wasn't propogating
back to the user, and there was no error message.  Fixed in doftprecv2() and
failftprecv2(): ckcftp.c, 15 Dec 2006.

Added the obvious timeout checks for FTP uploads, but I have no way to test
the code since our misbehaving FTP server does not hang when receiving
files, only when sending them.  But uploads work both with and without a
timeout set, so at least no harm is done.  ckcftp.c, 17 Dec 2006.

When downloading with FILE DESTINATION NOWHERE (= /CALIBRATE), Kermit still
checked the size of the incoming file and refused it if there wasn't enough
free disk space, on platforms (such as VMS) where zchkspa()) actually works;
reported by Bob Gezelter.  ckcfn3.c, 18 Dec 2006.

Built on Mac OS X 10.4.8 and NetBSD 3.1_RC3, all OK.  19 Dec 2006.

--- Dev.26 ---

Built on VMS 7.3-2/Alpha.  Had to squelch a couple compiler warnings by
changing some ints in the new \fpictureinfo() code from unsigned to signed,
and fix a typo in the prototype for the new gettcpport() function.
ckcnet.h, ckuus4.c, 22 Dec 2006.

--- Dev.27 ---

Parameterized pty routines and all references to them for file descriptor,
rather than to use global ttyfd, thus allowing ptys to be created for
different purposes.  Tested on Solaris 9 and Mac OS X 10.4.8, with "set host
/connect /pty emacs" (fine in both cases), and (more to the point) "set host
/connect /pty kermit" -- here we make a connection from one Kermit process
to another and transfer a file; works fine and wasn't especially slow either;
a good sign.  ckcdeb.h, ckutio.c, ckupty.c, 22 Dec 2006.

Created a new version of ttruncmd() called ttyptycmd(), which works by
calling do_pty() to get a pty to run the command on, and then in a loop,
reads from the pty and writes to the net and reads from the net and writes
to the pty, using select() to which of those it should do on each pass.
First cut just uses single-byte reads and writes.  Tested using Kermit
itself as an external protocol.  Works but slowly: 6000cps.  Zmodem doesn't
work at all.  ckutio.c, 24 Dec 2006.

Changed single-character read() and write() to buffered reads and writes,
with ttxin() and ttol() used for network i/o.  Using Kermit as the external
protocol, this gives 450Kcps (about 1/3 normal on this connection).  

But now there's a problem: the loop doesn't know when to stop.  How does it
know when the process that is running on the pty has exited?  With single
character read()'s that are executed unconditionally when select() says the
pty has data waiting, as in the first pass, I get EIO if there actually
isn't any, and can exit the loop.  But now, to avoid blocking, I call
in_chk() to see how much data is waiting, and I don't try to read anything
if it says nothing is waiting.  If the process associated with the pty file
descriptor has terminated, in_chk() would presumably get some kind of error,
but it doesn't.  I changed do_pty to return the pid of the fork where it
execs its command so we can check the pid with kill(pid,0) when in_chk() of
the pty says 0, but this doesn't help either; it seems like the process is
not exiting, but of course it is.

I could not find any legitimate way to test when the pty fork terminated.
Select() always says the pty file descriptor was ready, no matter what.
Select() never reports an exception on the pty file descriptor;
in_chk(ptyfd) returns 0 and not an error.  read(ptyfd,...) gets 0 but not an
error.  fcntl(ptyfd,...) doesn't get an error.  Finally I tried
write(ptyfd,c,0) and this indeed gets EIO (i/o error).  With this, using
Kermit as the external protocol works fine in Solaris but I tend to think
this trick will not be very portable (it isn't).  24 Dec 2006.

Made ttptycmd() use a more intelligent buffering scheme, fixed a few things
about how I was setting up the select() call that should address some of
yesterday's problems.  Still doesn't work but it's progress.  A: 25 Dec 2006.

Debugging yesterday's code...  Still, the error conditions are never set,
we never detect when the pty closes.  In Solaris, if select() says ptyfd is
ready to read but in_chk() says there are no characters there, we can treat
this as a loop-exit condition.  But in NetBSD, in_chk() always says 0 when
used on a pty (but works OK on a serial or net connection).

Realized I could not use in_chk() on the pty because there is too much
baggage with the communication path -- myread(), etc etc) -- so I replaced
this with a simple ioctl(ptyfd,FIONREAD,&n).  This works fine in Solaris but
always returns 0 in NetBSD, despite what the man page says (i.e. that this
function can be used on any file descriptor).

OK, let's see.... select() does not return useful results.  It says
characters are waiting on ptyfd when they are not, and it never detects the
closure of the pty.....  Well of course not, because we are the ones who
have to close it.  Just because the process has stopped doesn't mean the pty
is closed.  So we're back to square one, how do we know when to close it?
ckupty.c seems to keep the process ID in a global variable, pty_fork_pid
(which is not the same as the pid now returned by do_pty(), which is
useless, but I don't understand why).  But it doesn't matter because when we
kill(pty_fork_pid,0), we still get no error of any kind, even after we know
the process has exited.  I am completely flummoxed.  select() lies, and even
if it didn't, there is simply no completion criterion.  In the loop,
select() always says that the pty is ready to read.  To be continued.
26 Dec 2006.

Back to Square One, single-byte reads and writes.

 . This works for both ripple and Kermit.
 . Doesn't work for Zmodem but we'll deal with that later.
 . In this case FD_ISSET(ptyfd) is still true after pty closes.

But the ensuing read() gets EIO so we know the pty is gone.  That means the
same thing should happen in the buffered version, no?  Yes; I went back to
the buffered version and replaced all the other nonworking tests by a
blocking read of 1 byte on the pty and this detects the termination.  Now:

 . ripple works perfectly (of course it's only one-way).
 . Kermit fails

Let's call the remote, forked, redirected, external Kermit A and its
local partner B.  A sends its S-packet, B receives it OK and Acks.
A apparently does not receive the ACK in time, so sends the S again, but OK.
followed immediately by the F.  B Acks the F.  A sends the A, B Acks it.
But now A sends a piece of the previous F packet and the the first piece
of a D packet.

Clearly the buffering is messed up.  Sure enough, there was an extraneous
statement incrementing a read pointer in a write section.  Removing that
cleared up the problems with Kermit, now we can send and receive substantial
files efficiently in remote mode.  Zmodem seems to work too, except that at
the beginning a bunch of "**B0800000000022d"'s are stuffed into Kermit's
command buffer, so after the transfer we get some error messages.

In local mode, over a Telnet connection, Kermit works fine.  Zmodem works
OK too except it doesn't finish right, so at the very end rz on the far end
is still waiting for something; if I cancel out of it with ^X^X^X^X^X, it
deletes the file.  So there still is something wrong with the termination
test.

Also you don't see anything on your screen when running Kermit or Zmodem
this way.  That's to be expected, since they are using stdio for the
transfer, so they can't also be displaying progress or other messages.

Built this on NetBSD again...  Seems to work this time, but has trouble
finishing, like Zmodem.  Hmmm, on closer examination, it turns out that
since in_chk() always returns 0 on the ptyfd, we fall into our new
single-byte read code, so it's really slow, like 10K cps on a connection
where 1M is the norm.  27 Dec 2006.

Switched the pty from buffer peeking (FIONREAD) and blocking reads to to
nonblocking reads (O_NONBLOCK / O_NDELAY).  Works just fine on NetBSD except
now we no longer get EIO at the end when trying to read from the pty process
that has exited.  In fact, we're back to square one again.  not ioctl(), not
fcntl(), not select(), not even read() gets an i/o error after the pty
process exits.  But in NetBSD, we have to use nonblocking reads because ...
Hmmmm, maybe switch the fd between blocking and nonblocking for the test...
Nope, NetBSD seems to be hopeless (later, Ed Ravin confirmed that similar
problems have been observed with other applications that try to do this).

Switching to Linux, I see that yesterday's Solaris code (blocking reads)
works exactly the same way on Linux.

Tried today's O_NDELAY method on Solaris.  It works perfectly.  And then I
moved this one to Linux and it works perfectly there too.  Except in both
cases we have the wierd thing with Zmodem at the end, but I think that's
because rz/sz don't use standard i/o.  On NetBSD, it still hangs at the end.

Turns out that testing the pid works in NetBSD, even though it didn't in
Solaris.  Turns out read() gets an i/o error in Solaris and Linux but not
in NetBSD.  So checking the read result first, and then checking the pid
if read() got zero bytes catches all three.  28 Dec 2006.

Now the question of return code.  In the original ttruncmd() function, we do
a fork() and a wait().  When the external protocol program finishes, wait()
gives us its return code and we can pass it on through \v(pexitstat) as well
ttruncmd's own return code.  But ttptycmd() has to interact with the pty
continuously, so it can't just sit back and wait() for it.  Instead we have
to detect when the process has exited and then call waitpid() on the fork
pid, before shutting down the pty.  Tested on Solaris using Kermit as the
external protocol and then inducing failure, or letting it run to
completion.  FAILURE and SUCCESS set appropriately in each case.  Tested
with Zmodem too, works OK except for the aforementioned cosmetic glitch at
the end.  Tested on NetBSD, all OK.

To make K5 connection to Panix from Spam:

 set telnet debug on
 authenticate K5 init /realm:PANIX.COM /password:xxxxx
 set host shell.panix.com 23 /k5login

Good...  Now I try to send a file from Spam to Panix over the K5 connection
using Kermit itself as the external protocol.  It fails.  Inspection of the
debug log on the far side shows that the S-Packet was received correctly,
good!  This means we are reading the clear-text S-Packet from the external
Kermit program, and that ttol() is encrypting appropriately.

The remote Kermit sends the Ack and goes to read the next packet: ttinl()
calls myfillbuf() and:

  SVORPOSIX myfillbuf calling read()
  SVORPOSIX myfillbuf=0                <-- read returns 0
  SVORPOSIX myfillbuf ttcarr=2
  SVORPOSIX myfillbuf errno=0          <-- and reports no error
  HEXDUMP: mygetbuf read (-3 bytes)
  mygetbuf errno=0
  ttinl myread failure, n=-3
  ttinl myread errno=0
  ttinl non-EINTR -3[closing]

This happens because myfillbuf() deliberately returns -3 when read() gets 0
bytes.  I don't understand why this happens but the real problem is yet to
come.  The local Kermit (the one that has made the secure connection and is
running the external protocol through ttptycmd()) eventually figures out
that the transfer failed and when we reconnect, we get total garbage -- the
encryption either stopped happening, or got out of sync.

Looking at the local debug log, ttol() is doing its job, converting the
initial "kermit -r\13" from plaintext to cyphertext, as shown by the
hexdumps.  Then it enters ttptycmd()...  Hmmmm, wait, how can it send the
"kermit -r" before it starts the external protocol?  Never mind, worry about
that later...  Anyway, ttptycmd() says:

  ttptycmd loop top have_pty=1
  ttptycmd loop top have_net=1
  ttptycmd FD_SET ptyfd in
  ttptycmd FD_SET ttyfd in
  ttptycmd nfds=5
  ttptycmd select=1
  ttptycmd FD_ISSET ttyfd in
  ...
  ttptycmd in_chk(ttyfd) n=11
  ttptycmd ttxin n=11

ttxin() asks for 11 bytes, myfillbuf() gets 11 bytes, and hexdump() shows
the cyphertext, there doesn't seem to be any decrypting going on.  Hmmm, it
looks like the regular code calls ttinc() in a loop, rather than ttxin().
Maybe ttxin() doesn't have decryption hooks.  No, that's not it, the code is
there, but the Kermit packet reader does not use ttxin(), it uses ttinl().
But of course we can't use that for external protocols because it's designed
only to read Kermit packets.  Substituting a loop of ttinc()s for the ttxin()
call fixes things (and stangely enough, it seems to be faster).  And now we
have our first external protocol transfer over a secure connection (external
Kermit program, Linux over Kerberos 5 to NetBSD).  Zmodem worked too for a
short file but "something happens" with longer ones.  29 Dec 2006.

New makefile target for Linux with Kerberos 5, linux+krb5, that doesn't
include anything extra from SSL or other security methods (but apparently it
is still necessary to include -DOPENSSL_097 in order to get the right names
for the DES routines?).  Ditto netbsd+krb5 for NetBSD, except in this case
-DOPENSSL_097 is not necessary.  makefile, 30 Dec 2006.

Note to myself: On Panix:

  export LD_LIBRARY_PATH=/usr/local/kerblib
  make netbsd+krb5 "K5LIB=-L/usr/local/kerblib" "K5INC=-I/usr/local/include"

Can't telnet-k5 from newly built Kermit on NetBSD; partway through the
negotiations, just after "TELNET RCVD SB ENCRYPTION SUPPORT DES_CFB64
DES_OFB64 IAC SE" it dumps core.  The last two lines in debug.log after
this are:

  tn_sb[len]=5
  encrypt_support[cnt]=2

Rebuilding with -DOPENSSL_097 doesn't change anything.  Ed Ravin said they
have two different Kerberos installations, Heimdahl and MIT; maybe some
mixup between the two explains the problem (Jeff concurs). The core dump
occurs in ck_crp: encrypt_support():

   debug(F100,"XXX ep not NULL","",0);
   type = ep->start ? (*ep->start)(DIR_ENCRYPT, 0) : 0; <-- Here
   debug(F101,"XXX new type","",type);

Anyway, I can log in with Kerberos 5 to Panix OK from Columbia (sesame)
using 8.0.201.  So let's try to resurrect the Solaris version with everything:

  solaris9g+krb5+krb4+openssl+shadow+pam+zlib

I hunted around to find where the current library and header file
directories were...  Last time I tried this (March 2006) it bombed, not
finding libdes.  Instead we have /opt/kerberos5125/lib/libdes425.a.  Made a
new cu-specific target that includes this; now we get farther; it blows up
in ckcftp.c with tons of errors and warnings, which we can worry about
later.  Building again with -DNOFTP, it gets to ckuath.c (the first security
module) and:

  ckuath.c:151:18: error: krb5.h: No such file or directory
  ckuath.c:152:21: error: profile.h: No such file or directory
  ckuath.c:153:21: error: com_err.h: No such file or directory
  ckuath.c:176:28: error: kerberosIV/krb.h: No such file or directory
  In file included from /opt/openssl-0.9.8d/include/openssl/des.h:101,
                   from ckuath.c:219:

Found krb5.h in /opt/kerberos5125/include/krb5.h, added a -I for this
directory ...  Now we get lots of warnings in ckuath.c, but it completes OK,
then we wind up bombing out in ck_crp.c; I don't know why -- there are all
the same warnings (related to argument passing to DES functions), but no
errors.  I have no clue.

Tried to resurrect the solaris2x+krb4 target; this required changing -lkrb
to -lkrb4 and -ldes to -ldes425.  Lots of warnings in ckutio.c, ckcnet.c,
ckctel.c, then it bombs out in ckcftp.c because it can't find krb.h.  I
found it, adjusted the -I flags, but now it bombs because krb.h itself
#includes <kerberosIV/des.h>, which of course it can't find because the
brackets mean it's looking in /usr/include/kerberosIV/, which, of course,
the sys folks have removed.  Giving up on Solaris again.  Later, Jeff said
"Solaris does not publicly export the krb5 libraries.  You need to build
the MIT Kerberos libraries separately and link to them."  30 December 2006.

Changed copyright date to 2007.  ckcmai.c, 1 Jan 2007.

With Ed Ravin's help, successfully built C-Kermit with Kerberos 5 and
OpenSSL (netbsd+krb5+openssl+zlib), but it does not make K5 connections; it
gets hung up in the Telnet negotiations.  3 Jan 2007.

Downloaded MIT Kerberos 5 v1.4.4 to Solaris 9, 54MB worth.  This is just so
I can build a Kerberized C-Kermit for testing ttyptycmd().  Ran the
configure program, got a few warnings but it didn't fail (should it?)  Did
"make install", specifying a private directory but it failed immediately
with "cannot stat libkrb5support.so.0.0: No such file or directory".
OK, I tried.  3 Jan 2007.

Made a new makefile target for Mac OS X, macosx10.4+krb5+ssl, ran it on Mac
OS X 10.4.8.  It bombs out in ckcftp.c with: ckcftp.c:551: error: static
declaration of 'gss_mech_krb5' follows non-static declaration
/usr/include/gssapi/gssapi_krb5.h:76: error: previous declaration of
'gss_mech_krb5' was here".  Ditto for gss_mech_krb5_old, gss_nt_krb5_name,
and gss_nt_krb5_principal.  Tried again with -DNOFTP.  We get lots of
warnings in the network modules, but they complete.  But ck_ssl.c bombed
with a conflict between its own declarations of encrypt_output and
decrypt_input and the ones in ckuat2.h; removed the prototypes from the
latter (as Jeff advised) it built OK and it works OK too.  Built with FTP
too, but with link-time warnings about the aformentioned gss_* symbols.
#ifdef'd them out (gss_mech_krb5, gss_mech_krb5_old, gss_mech_name, and
gss_mech_principal) for MACOSX, where these symbols are exported by the
library.  Now it all compiles and links OK, and runs OK too.  3 Jan 2007.

Spent a day hunting around for a version of Zmodem that would build and
execute on Mac OS X, finally found one.  Now at last I could try a Zmodem
external-protocol transfer over a secure connection.  But phooey, C-Kermit's
pty support didn't work on this box.  Kermit finds master /dev/ptypa OK,
then in ptyint_void_association() tries to open /dev/tty but gets ERRNO=6
"device not configured" (which is apparently OK, because the same thing
happens on other platforms where this works), then tries to open slave
/dev/ttypa and gets ERRNO=13 "permission denied" because, indeed, I don't
have r/w permission on the device.  Left a message.  4 Jan 2007.

Changed TRANSMIT /BINARY output buffer size from 252 to 508 to avoid
TCP fragmentation.  Need to add a SET command for this later.
ckuus4.c, 5 Jan 2007.

Found another Mac where the ptys weren't protected against me, make a K5
connection and transferred a largish file with Zmodem with zero glitches,
except it was kind of slow, 84K cps.  Well, we're doing single-character
reads on the net (ttinc()'s instead of ttxin()).  Hmmm, but then I did it
again and got 2.2Mcps.  Success was reported, but it actually didn't work;
it only sent the first quarter of the file....  Oh well, at least now we
have a testbed.  5 Jan 2007.

Tried again, saw that the file is actually transferred instantly but then
we're not picking up the protocol at the end.  Theory: after the transfer
finishes, we come back to the prompt on the remote host, which means we have
something to read from the net and write to the pty, but the pty has already
exited.  AFTER THE PTY IS GONE, WE DO NOT WANT TO READ FROM THE NET ANY
MORE.  Adding this test makes Kermit succeed right away when sending the
same largish file, with a transfer rate of 4M cps, that's better.  But the
rz program on the far end is evidently not receiving the goodbye handshake
from the receiver, because it sits there foreever in its *B09002402009418
mode until I ^X^X^X^X^X out of it, at which point it deletes the file it
already received, not very helpful.  In the code, I read from the pty if the
pty is open and there is room in the buffer.  This means that when we get to
the end, either there is no room in the buffer (unlikely) or the last bit
sent by sz before exiting was cut off when the fork closed.  Why do we get
in this fix only with Zmodem and not with Kermit?

In Mac OS X, after sz exits, we get ERRNO=5 if we try to write to the pty,
but we still get no errors after that if we try to read from it.  Still,
prior to this we did more than 20 unproductive nonblocking reads from the
pty (no error, no bytes) without incident; there did not seem to be anything
waiting.  In fact, the last thing we read from the pty were the text
messages that are issued at the end of the transfer: "rz 3.73 1-30-03
finished."  After which it pauses a second and spits out a message about
UNREGISTERED COPY.

Figured out how to build lrzsz, in hopes that the previous problems were
with rzsz and crzsz's fiddling with file descriptors, but I get the same
behavior.  Which is good, I guess, because if I can fix one, I fix them all.
Or not...  Testing lrz by itself (not under C-Kermit), I see that it doesn't
work at all with Kermit's own Zmodem implementation.

OK, here's one problem: at the end of the transfer, the Omen Zmodems print
stuff like "Please read the license agreement", Kermit dutifully reads this
from the pty and sends it to the host, the host shell says "Please: command
not found", issues its prompt again, which Kermit reads, feeds to the pty,
and apparently the pty echoes it, so we send it back to the host, and there
ensues an infinite loop of getty babble until the pty closes.  Now, there
ought to be a way to make the external protocol shut up, like Kermit's
-q(uiet) flag, but these are unregistered versions so you can't shut up the
messages.  In fact, the transfer works, but the getty babble at the end
ruins the experience.  Now I'm beginning to wonder how any of these programs
ever worked as external protocols.  Hmmm, now that I try it, I see the
same thing happens the old way, when using ttruncmd() rather than ttptycmd().

Reading the crzsz documentation I see it says that messages come out on
stderr.  OK, that's progress.  In ckupty.c I try redirecting 2 to /dev/null.
Well good, this filters out the messages from csz, but we still get getty
babble on the prompt.  In the debug log, we read the last bunch of stuff
from net, 618 bytes of Zmodem stuff...   Now what happens?

Zmodem on the remote exits, the host prints its prompt.  Kermit, of course,
reads the prompt from the net, now come to the bottom of the loop and we
have 7 bytes to write to the pty, and no error condition, so we continue the
loop.  select() says that the pty is ready for writing.  We write the 7
bytes and and get no error.  Loop again, this time select() says the pty has
data waiting.  Sure enough we get the prompt back, and send it to the net,
and thus begins the getty babble.  There are two causes for this:

  1. crzsz does not exit immediately; it sleeps for 10 seconds after
     printing its nag message.

  2. During this interval the pty seems to be echoing what is sent to it.
     csz is not echoing; I checked.  Anyway, removing the pause doesn't
     seem to make a difference.

ttptycmd() needs to:

 . TELL the pty module to redirect stderr to /dev/null
 . SET PTY TO NOECHO (master or slave?)

Tried setting the pty to noecho:

  termbuf.c_lflag &= ~(ECHO|ECHOE|ECHOK);

and this seemed to stop the getty babble.  After the file transfer, I read
back the prompt from the host shell, I write the prompt bytes to the pty;
there is no error.  And now select() simply hangs forever (or times out if
a timeout is set).  The question here is: why didn't writing to the pty
produce an error?  And, because we never detect the pty has exited, we can't
set a good return code.  5 Jan 2007.

Moved pty fork testing to a separate routine, pty_get_status(), and 
added a call to it from the place where we time out, in case the fork
terminated; then we can get and return its status.  6 Jan 2007.

Added calls to pty_get_status() to every place where we suspect a pty error,
tried again with lrzsz, crzsz, and regular rzsz.  All three work, but in
each case waitpid() indicates that the sz program gave exit code 1 (failure).
ckutio.c, 7 Jan 2007.

Changing the subject...  On my test system, every time I execute ttptycmd(),
I get "permission denied" on /dev/ttyp3.  Then I run it again and get to
ttyp4 which is OK.  I wanted to skip past any pty for which I lack
permission and try the next without raising an error.  Added debugging code:

  16:25:23.524 pty_getpty() pty master open error[/dev/ptyp0]=5
  16:25:23.524 pty_getpty() pty master open error[/dev/ptyp1]=5
  16:25:23.524 pty_getpty() pty master open error[/dev/ptyp2]=5
  16:25:23.524 pty_getpty() found pty master[/dev/ptyp3]
  16:25:23.524 pty_getpty() slavebuf [2][/dev/ttyp3]

So it already was skipping past open errors; ttyp3 was opened successfully.
The problem is that ptyp3 is rw-rw-rw-, but the corresponding master,
ttyp3, is rw--r----.  It seems the code assumes that if the master can be
opened, then so can the corresponding slave.  Unfortunately, the code is
not structured to allow us to skip ahead to the next master if the slave
can't be opened.  7 Jan 2007.

Spent a couple hours trying to rearrange the code in the pty module to skip
past inaccessible slaves but it was a rabbit hole, not worth it, backed off.
8 Jan 2008.

Tried an upload over a secure connection using lsz.  Unexpectedly, this time
it worked; not only was the file (about 0.5MB) transferred correctly, but
Kermit detected the fork's termination and got the pid's exit status, and,
for the first time, correctly reported a successful transfer.  I have no
idea why this works today and not yesterday.  More tests; it works most of
the time.  It works with csz and with regular sz too.

(days later...)

ckucns.c seems to do the right thing; it recognize the ZSTART string,
activates the Zmodem-Receive APC, and returns.  doconect() sees the APC and
begins to execute it.  The RECEIVE command results in a call to the GET
command parser, doxget() (IS THAT RIGHT?), then comes a ttflui(), which
throws away a bunch of stuff.  Finally we get to ttptycmd(), we get a pty
and run lrz in it, select() says stuff is waiting from the pty, but read
returns 0, errno 0.  Skipping the ttflui() in doxget() if the protocol was
not Kermit didn't seem to make difference.  ckuus6.c, 8 Jan 2007.

The problem is that in this case, reads from the pty never get anything (no
data, no error), write always gets an error.  It's as if the pty was not
being set up right, or we're using the wrong file descriptor.  And if we
skip the autodownload?  Same thing.

OK, putting downloads aside for a moment, let's get uploads working as well
as possible.  At this point we have the odd situation (at least in this
configuration) that the upload succeeds, but now for some reason we are
unable to read the exit status from the process, even though this was
working before, so ttptycmd() returns 0 (failure), yet Kermit reports
success.

Well, it turns out that kill(pty_fork_pid,0) was gumming up the works.
If we use only waitpid() all is well, I think.  waitpid() with WNOHANG
returns -1 with status -1 errno 0 if the pid has not exited, and it returns
the pid and status > -1 if the process has exited.  Fixed pty_get_status()
to do it this way.  ckutio.c, 7 Jan 2007.

Let's move this from Mac OS to NetBSD and see how it works.  Well, the file
transfer was just fine, but then I used some sexps to calculate the elapsed
time and transfer rate, and Kermit hung in dosexp().  Fine, ignoring that...
The debug log shows that ttptycmd() gets the pty OK, master and slave, the
i/o goes smoothly, and waitpid() does its job perfectly.  Solaris, same
deal; ttruncmd() goes smoothly, but then the sexps afterward get "Arithmetic
exception".  Turns out there was a BAD bug in dosexp() that allowed an
integer division by 0 to occur under certain circumstances; it's always been
there.  Fixed in dosexp(): ckuus3.c, 8 Jan 2007.

After noticing a few problems running the pop.ksc script in production over
the past year, rewrote \femailaddress() to be more reliable and a lot
simpler.  ckuus4.c, 9 Jan 2007.

Back to ttptycmd()...  When we left off, we could send but not receive.  Set
up a test case using Kermit as the external protocol for receiving a short
file.  If I SET STREAMING OFF and use short packets, it actually does work,
so it's not a complete failure to function, but apparently a lack of flow
control for the pty.  Began by completing the parameterization of the pty
module, so it can be called for interactive use (fc 0) or for running
protocols (1).  Confirmed that everything works at least as well as before
(e.g. "set host /pty emacs" vs external protocols).  ckcdeb.h, ckutio.c,
ckupty.c, 9 Jan 2007.

Found in HP-UX "man 7 pty" a description of ioctl(fd,TIOCTTY,fc) which is
exactly what we want: fc 0 turns off all termio processing and guarantees an
ininterrupted, unmolested, flow-controlled stream of bytes in both
directions.  This function also exists in Linux, but not in Solaris, NetBSD,
or Mac OS X (TIOCNOTTY is not what we want, it does something else entirely).

Another possibility is TIOCREMOTE, which "causes input to the pseudoterminal
to be flow controlled and not input edited, regardless of the terminal
mode".  This one exists in at least HPUX, NetBSD, Solaris, and Mac OS X.

Solaris: builds OK, but at runtime we get ENOTTY ("Inappropriate ioctl for
device").  By the time this happens, it's hard to tell from the code whether
the fd we're using is for the master or the slave; TIOCREMOTE can be used
only on the master.  Close inspection shows that I am indeed doing that;
ptyfd as seen by ttptycmd() is truly the master, i.e. the /dev/ptyXX device,
not the /dev/ttyXX device (the slave fd can't be seen at all, as it exists
only in a separate fork).  OK, so now we know that TIOCREMOTE can't be used
on Solaris.

NetBSD: Somehow, whether as a result of today's fiddling or the phase of the
moon, the code in pty_open_slave() that tries to open /dev/tty started
failing on NetBSD ("Device not configured").  Changing it to be run only if
fc == 0 (which doesn't seem to hurt anything), once again I get ENOTTY on
the TIOCREMOTE ioctl.  Zmodem works but Kermit totally fails (the fork exits
immediately with an exit code of 0, even though it didn't do anything).

Mac OS X: Exactly the same sequence and results as NetBSD.

Linux:  It did not execute the new ioctl at all; apparently the TIOC symbols
are hidden or not exported or something.

Where we stand:
 . Downloads don't work
 . Uploads got slow again
 . Kermit doesn't work at all as an external protocol

Actually if I take the debugging out it goes fast, but it doesn't finish.

All today's work on ttptycmd() looks like a dead end.  To roll back to
yesterday:

  cp ckutio.c-20070108 ckutio.c
  cp ckupty.c-20070108 ckupty.c
  cp ckupty.h-20070108 ckupty.h

or to continue with today's:

  cp ckutio.c-20070109 ckutio.c
  cp ckupty.c-20070109 ckupty.c
  cp ckupty.h-20070109 ckupty.h

Comparing Monday's and Tuesday's pty-related code, the differences are:
 1. Passing of function code to and among pty modules.
 2. Skipping the TIOCSCTTY ioctl and the open("/dev/tty") test.
 3. Attempting to put pty in TIOCTTY or TIOCREMOTE mode.

Commenting out 2 and 3 should put us back where we were on Monday if the
parameterization was done right.  And with this, on Solaris, downloading
with Kermit external protocol works but slowly, 8K cps, with or without
debugging.  Debug log does not show any obvious bottlenecks; select() takes
anywhere between no time at all and 0.1 seconds to return.  If I increase
the pty-net buffer size from 1K to 4K, the rate goes up to 55K cps.  If I
make it 8K I get 136K cps.  With 16K I get 346K cps.  32K: 395K cps -- this
last one isn't worth the doubling.  But at 24K I get 490K cps, sometimes
twice that.  Let's stick with 24K for now.  Downloading with Zmodem (rzsz)
works at the same rate, but now we're back to seeing the getty babble
(Several "**B0800000000022d") at the end. 10 Jan 2007.

Moving to Mac OS X, everything works the same as on Solaris, except I don't
get the Zmodem getty babble there, not even with Omen rzsz.  Tested sends
in both remote and local mode, the latter over a secure Kerberos 5 Telnet
connection, using C-Kermit, rzsz, lrzsz, and crzsz, all good.  10 Jan 2007.

Now we're back where we were yesterday morning, but with better throughput.
The big issue then was receiving files.  But yikes, now it works!  Not only
that, I got a transfer rate of 2.1M cps.  That's using Kermit protocol,
streaming, and big (4K) packets.  Which didn't work before.  Not a fluke
either, I uploaded bigger and bigger files up to 6MB, they all went
smoothly, at rates between 1 and 2 MBps.  10 Jan 2007.

Not so great in Zmodem land, however.  If I start the external-protocol
receiver on the far end, escape back and start a Zmodem send... nothing.
If I leave the remote C-Kermit at its prompt (where it supposed to recognize
the Zmodem start string), still nothing.  On the other hand, if I do it
with a script instead of by hand:

  def xx output take blah\13, send /proto:zmodem \%1
  
it works, at least intermittently.  But that's in remote mode.  We won't be
using this in remote mode.  In local mode, where we have a secure connection
to another computer, it seems we can read from the pty and write to the net,
but we time out waiting to read from the net; nothing arrives.  Well, we
know that i/o works both ways, so there is some kind of screwup with the
Zmodem protocol start itself.  Increasing the (still hardwired timeout) from
5 to 22sec and driving the whole process with a script so as to avoid
autodownload as well as manual dexterity effects...  It just sits there
forever, way longer than 22 sec.  ^C'ing out, I see that sz was indeed
started on the far end and the protocol was executing.  But it looks like
the receiver (the one running under ttptycmd()) is getting trashed packets,
because (a) it seems to be sending the same thing over and over again, and
(b) sometimes it waits as long as 10 seconds before anything arrives from
the remote.  Maybe I was too impatient; I interrupted it after 4 minutes but
it seems to have been making some progress.  Whenever there was data
available to read from the net, it was always 65 bytes, and it was not
actually the same data over and over.  This is using lrz as the external
protocol.  crz gets a bit farther.  In this case we read up to 24K at a
gulp, but the amount varies a lot.  It looks like we took in about 1.2MB of
Zmodem protocol data, but were only able to output the first 20K of the
file.  Clearly there were lots of errors.  In the end, the crz exits with
status 1 (failure).

Anyway it looks like we're back at needing to find a way to accomplish
something like TIOCREMOTE on the pty, which is where we came in.  10 Jan 2007.

Without any way to make the pty transparent and flow controlled, it would
seem to make sense to write to the pty in smaller chunks than we do to the
net.  I left the read-from-pty-write-to-net buffer at 24K and changed the
read-from-net-write-to-pty buffer to 48 bytes.

Upload using lsz worked but took about 3 minutes.  Actually it didn't work.
On the local end it seemed to work, but the file did not appear on the
remote end.  Tried this several times, each time with different results,
adding more debugging each time.  The problem this time was that the pty
read could get EWOULDBLOCK.  Changed the code to not treat this as an error,
now Zmodem uploads are solid again except I never got EWOULDBLOCK again
either, even though I repeated the same upload about 1000 times (with
throughput of over 2MBps even with debugging on), so the test for it has
not been exercised.

OK, uploads still work.  Back to downloading...  The very first pty read
gets 0 bytes, followed by the fork test that shows that it exited with
exit status 2. 

Next we try starting sz with some different options on the far end:

 -q: quiet (no messages):
     for some reason this gets totally stuck.
     it looks as if this option is misdocumented;
     sz seems to be sending the letter C (as in Xmodem 1K or whatever)

 -e: escape (all control chars):
     first attempt to read pty finds the process gone with exit status 2.

 -k: send 1k blocks:
     this one didn't stop immediately.  It reads 48 bytes from net, writes
     48 to the pty with no error.  Then reads 21 bytes from the pty, writes
     them to the net OK.  Then reads 48 bytes from net, writes them to pty OK,
     reads 21 from pty, writes to net OK, etc etc...  It appears to have
     worked but (final read from pty returned 0, fork test showed lrz exited
     with status 0), but only 754 bytes were received from the net when the
     file is 420K...

Well this only goes to show that the faster we shove stuff into the pty, the
worse it gets.  Zmodem downloads won't work unless we can make the pty
transparent and flow-controlled.  So to summarize today's developments:

 . separated in/out buffer sizes
 . handled EWOULDBLOCK
 . found out that sz options don't help much

11 Jan 2007.

Next day.  This has got to be the most delicate code ever, it's like
Whack-A-Mole, fix A and B pops up.  Even without touching it, something that
worked perfectly a 2:00 doesn't work at all an hour later.  Maybe I could
have used pipes instead of ptys, but pipes have problems of their own.
There has to be a way to do this.  The telnet server, the SSH server, etc --
they all run on ptys, and we can upload files to them with Kermit.  Why?
Because Kermit puts its terminal into all the right modes using the
time-honored methods of ttpkt() and ttvt().  Perhaps all we need is a copy
of ttpkt() that operates on the pty.

On that theory, let's go back to Kermit as the external protocol.
It's important to suppress all messages and displays.  With that,
uploads work fine, no hitches.

Downloads:  We fail right away.  The debug log shows the Kermit program that
we are starting in the pty says:

  "" - Invalid command-line option, type "kermit -h" for help.

But of course we are not giving it an invalid command-line option.
Switching to gkermit for the external protocol, now we see that no matter
command-line options we use, we read 0d 0d 0a from the pty and then the
next time we go to read from the pty we get 0 bytes and waitpid() says the
program has exited with status 1.

Why should downloading be different from uploading?  ttptycmd has no idea,
it does everything the same.  The only difference would seem to be which
side sends first, but even that tends to get washed out by each program's
startup messages.

Downloading with Kermit worked 2 days ago, what's different now?  The buffer
sizes.  Putting the net-to-pty back up to 24K (from 48 bytes)...  Now it
works again.

Conclusion: Kermit conditions the pty correctly, Zmodem does not.  Therefore
ttruncmd() must duplicate what ttpkt() does.

Or not.  Because rz works fine on ssh/telnet ptys too.  But not on our pty.
lrz exits immediately with status code 2 = 01000 but there are no clues in
the lrz.c source code, I don't even see this exit status set anywhere.
Unredirecting stderr, I see that the error is "lrz: garbage on command line".

Why do both Kermit and Zmodem sometimes think they are receiving an invalid
command line?  If I could capture the garbage...

Side trip #1: ("pty.log",O_WRONLY) gives "no such file or directory".
Changed this to ("pty.log",O_CREAT,0644) and now it doesn't get an error,
and it creates the file, but not with 0644 permissions, and with nothing
written in it.  How come nothing works?

Fine, the debug log shows that ttptycmd() receives the correct string
(e.g. "lrz -v").  It passes it to do_pty() correctly, and do_pty() passes it
to exec_cmd(), which runs cksplit() on it, coming up (in this case) with
"lrz" and "-v", which is right, and then:

    args = q->a_head + 1;
    execvp(args[0],args);

execvp() wants the args array to have a null element at the end.  cksplit()
does indeed do that, or at least the code is there.  Added code to exec_cmd()
to verify the argument list and that it is null-terminated.  So far it is.

Anyway, we have traffic between the Zmodem partners, but no joy.
Commenting out the bit that redirects stderr, now I can see it on my screen
in real time:

  lrz waiting to receive.Retry 0: Bad CRC
  Retry 0: Got ERROR
  Retry 0: TIMEOUT
  Retry 0: TIMEOUT
  Retry 0: TIMEOUT
  Retry 0: TIMEOUT

etc etc, forever.  Trying sz -e on the far end, I get:

  Retry 0: Bad CRC
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  ...
  Retry 0: Got ERROR
  Retry 0: Bad CRC
  Retry 0: Got ERROR
  Retry 0: Got ERROR
  lrz: xxufio.c removed.

So apparently it's not a matter of escaping.  Trying some other stuff, I
caught the command-line problem in the act:

  lrz: garbage on commandline
  Try `lrz --help' for more information.

Debug log shows:

  cksplit result[lrz]=1
  cksplit result[-v]=2
  exec_cmd arg[lrz]=0
  exec_cmd arg[-v]=1
  exec_cmd arg[]=2

An empty string at the end instead of a null pointer.  I really do not see
any way that could happen, but rather than dig into cksplit() again after
all these years I added a test for this in exec_cmd(), which, of course
after adding it, never encountered this behavior again.

Fiddled with pty buffer size again.  Made it 512 bytes instead of 24K.
Zmodem downloads are the same (Rety 0: TIMEOUT, over and over).  But I don't
see what the problem is -- every time we receive n bytes from the net, we
write n bytes successfully to the pty and there are no errors.  But it also
looks like the remote sender is sending the file header over and over
because it's not receiving an acknowledgment.  If we're not losing data,
then maybe it's a transparency problem.

Tried uncommenting the TIOCblah stuff I commented out before.  Now instead
of only timeouts I get:

  lrz waiting to receive.Retry 0: Bad CRC
  Retry 0: Got ERROR
  Retry 0: Bad CRC
  Retry 0: Got ERROR
  Retry 0: Bad CRC
  Retry 0: Got ERROR
  Retry 0: TIMEOUT

which is odd because the TIOCREMOTE ioctl failed with errno 14, EFAULT,
bad address, which should indicate it had no effect.  We're still receiving
data from the remote in tiny chunks (from 12 to 65 bytes), apparently the
same stuff (file header), and writing them to the pty successfully but
nothing...

Looked at cloning ttpkt() for the pty, but these stupid routines use global
tty mode structs so it's not going to be easy.

Well, we got exactly nowhere today, but I think I'll leave stderr as it is
so users will see some feedback; no reason not to.

WHY DO KERMIT DOWNLOADS WORK AND ZMODEM NOT?

Is it 8-bit transparency?  Up til now I've been testing with text files.
If I try to download a binary what happens?  Fails after 99 seconds.  Packet
log from the far end shows that as soon as the first packet containing 8-bit
data is sent, everything stops.  At least I got one of these:

  17:23:56.475 exec_cmd arg[gkermit]=0
  17:23:56.475 exec_cmd arg[-qr]=1
  17:23:56.475 exec_cmd arg[]=2
  17:23:56.475 exec_cmd SUBSTITUTING NULL=2  <-- the code I just added

Doing this again shows the same thing on the near end.  All the 7-bit-only
packets are sent and acknowledged OK.  Three 8-bit data packets arrive and
nothing else happens after that.  This is with G-Kermit.

The same thing happens with C-Kermit receiving.  But if I change C-Kermit's
.kermrc to turn off streaming and use a short packet length:

The transfer works, even though it's sending 8-bit bytes.  So the problem is
not 8-bit data after all, per se.  Facts:

 . Kermit can receive streaming transfers of 7-bit files.
 . Kermit can not receive streaming transfers of 8-bit files.
 . Kermit can receive nonstreaming transfers of 8-bit files with short packets.
 . Kermit can receive nonstreaming transfers of 8-bit files with 1K packets.
 . Kermit can receive nonstreaming transfers of 8-bit files with 4K packets.

So it's the combination of streaming and 8-bit data?  12 Jan 2007.

As a test I made a new routine pty_make_raw() that does cfmakeraw() (a
nonportable "POSIX-like" function known to be used on ptys in applications
that do approximately what we're attempting).  Results:

 Solaris: errno 25 - inappropriate ioctl for device.

This happens even when we try to get the terminal modes with tcgetattr(),
which is completely nuts.  We pass it the file descriptor of the pty master,
which is supposed to work.  But in Mac OS X, there are no errors.  But
downloads still don't work; lots of errors but the pattern is different.
Using a very small buffer:

  Retry 0: Bad CRC
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Got TIMEOUT
  Retry 0: TIMEOUT
  Retry 0: Bad CRC
  Retry 0: Bad CRC
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: TIMEOUT
  Retry 0: Got ERROR
  Retry 0: TIMEOUT
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Bad CRC

Using a bigger buffer:

  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  (several screensful)

Various other combinations... Nothing seems to work.

Insight: telnetd does exactly what we want to do, sort of.
But it uses TIOCPKT, so every time it reads from pty, it receives
one control byte and then the data bytes, which would complicate our
buffering scheme considerably.  Anyway the TIOCPKT ioctl() fails on
Mac OS X with 14 "Bad address".

Also see: snoopserver.c (found in Google).  It seems to do things in a
slightly different way -- it sets stdout to raw and then dups it to the
slave side of the pty?

Maybe it's a mistake to use the ckupty.c routines.  They are designed for
creating and accessing an interactive session.  Maybe just copy one of the
other programs.

18 Jan 2007.  Tried going back to blocking rather than nonblocking reads
to see if it would make a difference, after all the other changes.  Nope.
OK, let's look at some of these other programs...

snoopserver.c.  I don't know exactly what this is or where it's from or what
platform it runs on and there are no comments to speak of, but it does
approximately what ttptycmd() does.  To get a pty it uses openpty():

  if (openpty(&pty, &tty, NULL, NULL, NULL) == -1)

then creates a fork.  In the fork, it closes the pty (master) and
manipulates the modes of the tty (slave), dups tty to be stdio, and then
doex execv() on the command.  Meanwhile the upper fork closes the tty
(slave), gets the attributes of stdin, using atexit() to have them
automatically restored on exit.  Then it sets stdin to raw mode and enters
the select() loop on stdin, the pty master, and the net.  It uses regular
blocking reads.  It does not use TIOCPKT or anything like it.

openpty() is supported on: Linux, Mac OS X, NetBSD, FreeBSD, ...
openpty() is NOT supported on: Solaris, HP-UX, ...

 1. Try copying the pty code, but keep everything else the same.

I did this; it compiles and starts OK, upper fork (ttptycmd) debug log shows
no errors, but nothing happens.  Logs show that the Kermit program that is
started in the subfork seems to die as soon as it reaches eof on its init
file.  The good news, at least, is that select() doesn't report report that
the pty is ready to be read.  Clearly the file descriptors aren't being
assigned as expected, or as before.
  
In ckupty.c getptyslave() dup2's the slave fd to 0 and 1.  The new code
does exactly the same thing.  Debug log makes it look like the forked kermit
is not receiving its command line.  But now I'm not even sure that the
forked kermit started at all.  ps from another terminal doesn't show it.

19 Jan 2007: Noticed that in snoopserver, the select() calls use standard
input and output file descriptors, rather than the pty master.  Made that
change...  In doing that I had to look at every file descriptor in every
line of code and discovered a couple mistakes, fixed them, put back the
original code but with the fixes, tried it, but no change; can upload OK but
still can't download with Zmodem without lots of errors and ultimate
failure.  Going back to the alternative version and trying to get the the
file descriptors sorted out, now it appears that the external Kermit program
never even starts in the lower fork.  After a bit more fiddling I sort that
out, but now when the lower Kermit program goes to open "/dev/tty" it gets
errno 6 "Device not configured".  Forcing it to use stdio with "-l 0", it
gets past this and actually sends its first packet.  But the Kermit on top
reads nothing from the pty.

Next, I change the pty fd from STDIN_FILENO and STDOUT_FILENO to slavefd.
No difference.  Next I comment out the dup2() calls.  This time I get some
action.  The transfer starts, but only one packet comes.  Log shows that
the lower Kermit sends its S packet.  The upper Kermit receives the ACK
but the lower Kermit never gets it.  The write to the pty succeeds, no
error.  Different combinations give different results.  If write to master
and read from the slave, I get packets in both directions but tons of
errors....  This happens only if I comment out the dup2()'s.

25 Jan 2007: After leaving it sit for a while, and realizing that what I'm
trying to do has to be possible because so much other software does the same
thing (e.g. Telnet servers), I put things back to how they were originally
-- the upper fork (Kermit) uses the master and the lower fork the slave.
The upper fork puts the master in raw mode, the lower fork puts the slave in
raw mode.  The lower fork dup2's the slave fd to stdin/out.  Send file in
remote mode using external Kermit: works OK but select() times out at the
end.  This means that the self-contained pty code in ttptycmd() is sorted
out -- all the file descriptors go to the right place, etc, and now we can
use this routine as a testbed, rather than the original ckupty.c-based one.

But send with lsz, csz, and regular rz: Nothing happens, times out after 0
bytes of i/o.  Once again, Kermit works, Zmodem doesn't.  The reason for
running Zmodem in a pty is so its i/o will work as it does on a terminal,
no matter how it may fiddle the file descriptors.  So why don't we see a
single byte come out?

Commenting out pty_make_raw(), I get a successful Zmodem send using lsz.
csz manages to get the filename across, but then gets stuck.  regular sz, on
the other hand, works perfectly.  Testing csz by itself (not under Kermit),
I see it fails in exactly the same way ("Got phony ZEOF", etc).  OK, forget
crzsz.

OK, let's move to local mode over a Kerberized Telnet connection...
Uploading (sending) with external Kermit protocol... works.
Downloading (receiving) with external Kermit protocol... works.
Uploading with sz... works.
Downloading with lrz...  Gets tons of errors and fails.

Running pty_make_raw() on the slave but not on the master: no difference.
Running pty_make_raw() on the master but not on the slave: no difference.

Back where we started...  Either:

 . Zmodem is overdriving the pty, no matter what modes we put it in.  
 . It's a transparency problem.

Theoretically we should be able to test these by using different sz switches:
  -q:    quiet (should always use this)
  -e:    escape all control characters
  -B n:  Buffer n bytes (rather than whole file)
  -L n:  Packet length
  -l n:  Frame length (>= packet length)
  -w n:  Window size
  -4:    4K blocksize (doesn't help)

-q by itself doesn't help.
-q -e, this one worked but still got about 100 errors and was very slow.
-q -e -l 200 -L 100, failed fast and bad.
-q -e -w 1.  Failed quickly.
-q -e -w 1 -B 100.  Eventually failed.
-q -w 1, Eventually failed.
-q -l 1024, this gets much more errors, definitely need -e.
-q -e -l 1024, got pretty far before failing.
-q -e -w 1 -l 1024, also got pretty far before failing.
-q -e, this one got farthest of all, about 48K, before getting errors.

In the latter combinations that work somewhat better, we always get up to
16K, or 32K, or 48K, before the errors start coming out and piling up.
Sometimes the errors are recoverable and we receive as much as 300K
successfully before giving up.

Now that we have data flowing pretty well (but not well enough), tried
reinstating pty_make_raw(), but it hurt more than helped.

As a sanity check, I tried transferring from the same host over the same
kind of connection (Kerberized Telnet) directly to K95's built-in Zmodem
protocol, and that worked fine.  So the problem is definitely in the pty.
Or more precisely, where Kermit writes incoming net data to the pty master.

26 Jan 2007: Tried changing the size of the net-to-pty buffer from 24K to
1K.  Result: total failure.  Set both buffers to 1K.  Still total failure.
Set both to 4K: now we get about 45K of data, then failure.  Put them both
back to 24K, still fails totally -- the same code that worked pretty well
yesterday.  Actually, no downloads work, not even Kermit, not even of
text files.

27 Jan 2007: Since I have not been able to find a way to make ptys work for
this, I made a third copy of this routine, this time using pipes instead of
ptys.  The disadvantage here is that if the external protocol does not use
stdio, the pipes won't work, but one thing a time...

Inferior Kermit starts in lower fork, but when it tries to send its first
packet it gets errno=9 EBADF, Bad File Descriptor.  Substituting G-Kermit as
the external protocol, which is simpler, reveals that the problem is that
the external protocol gets errors when it tries to manipulate the its stdio
file descriptors with ioctls, etc; these are not valid for a pipe.  The pipe
mechanism itself works.  If I take out the test for ttpkt() failing in
gkermit, the file transfer works OK.  Trying Zmodem... Sending works OK;
receiving works a lot better than with ptys (it got 360K into the file
before failing).  Making the buffers smaller, doesn't help.

I'm starting to wonder if the problem might be in my buffering code, rather
than in the pty or pipe interface...  Try making a version that does
single-character reads and writes.

This one reads the first packet from the lower Kermit and sends it.  It is
recognized by the other Kermit, which sends an ACK.  We see the ^A of the
ACK, but then select() times out on the next character -- OF COURSE: because
at a lower level, it has already been read.  We have to check the myread
buffer, and then call select() only if it's empty.  Making this change:

 . SEND with G-Kermit works (but very slowly).
 . SEND with lsz works but gets a lot of errors, eventually succeeds.

Let's work our way back...  With the same changes to the buffered pipe version:

 . SEND with G-Kermit/streaming works (fast).
 . SEND with lsz works too (fast), but we get gubbish at the end.
 . RECEIVE with Kermit fails because "/dev/tty is not a terminal device".
 . RECEIVE with rsz... lots of errors ("garbage count exceeded") but succeeded.

But maybe now we're seeing pipe artifacts, so going back one more step to
the version that gets its own pty and starts its own fork:

 . SEND with G-Kermit/Streaming works (fast) but select() times out at the end.

Another breakthrough: Moved the write pieces to below the read pieces.  This
is what was preventing the buffer reset code from working -- with the writes
done before the reads, we never catch up and can never reset the buffers.

 . SEND with G-Kermit/streaming works (fast) (but there's a pause at the end)
 . SEND with lsz works (fast) (but there's a pause at the end)
 . RECEIVE with rsz... lots of errors ("garbage count exceeded") and fails.
 . RECEIVE with Kermit -- nothing happens (it thinks it succeeded), then we
     reconnect, terminal sees S packet and goes into autodownload

From the log it looks like ttpkt() fails in the lower Kermit.  Switching
this with the hacked G-Kermit...  it gets "transmission error on reliable
link".  Tried again with real Kermit below, this time with "-l 0" and not
streaming.  This was actually working, but slowly, I don't see any NAKs in
the packet log, but then select() timed out.

28 Jan 2007: Restored both the calls to pty_make_raw():

 . SEND with C-Kermit streaming works, but slow (54Kcps)
 . Ditto, but with debugging off -- hangs forever.
 . Ditto, but using G-Kermit instead of C-Kermit -- also hangs forever.

Backed off on calling pty_make_raw().  Same thing.
Reduced size of net-to-pty buffer.  Same thing.

15 Feb 2007...  Decided to give up on this and publish it as is, in hopes
that somebody with more experience with ptys can make it work, because I'm
just going in circles.  So today I just have to get the code into shape so
people could choose among the three alternative routines.  The second one,
yttyptycmd(), is the one that uses openpty(), which is not portable, so it
can be enabled only for Mac OS X, NetBSD, FreeBSD, and Linux, or by also
defining HAVE_OPENTPY at compile time.  Anyway, if you build Kermit in the
normal way, you get the regular behavior -- ttruncmd() is used to execute
external protocols.  If you build it with -DXTTPTYCMD, you get the first
version of ttptycmd(); with -DYTTPTYCMD the second, and with -DZTTPTYCMD the
third.

(Then some interruptions, then...)

From Jeff, fix hostname comparison in X.509 certificate checking to work
right in the case of names that contain no periods.  dNSName_cmp(): cl_ssl.c,
21 Feb 2007.

John Dunlap noticed some strange behavior when transferring files between
home base and the EM-APEX oceanographic floats via satellite... long story,
but every so often the transfer would get stuck for a long time, and it
happened only when C-Kermit was sending a file and received two or more
packets (Ack or Nak) back to back from the float.  Years ago I added some
lookahead code to ttinl() to clear the input buffer of any interpacket junk
so that, in the windowing case, we wouldn't be tricked next time around into
thinking a packet was waiting to be read when there wasn't.  The code, which
has been there for a while, was a bit fractured; luckily, it would be
executed only when the debug log was active so it didn't have much effect.
The problem was that if the SOP came immediately after the EOP, it could be
missed because the loop read the next character before checking the current
one.  Fixed by rearranging the loop.  Also I changed it so it would execute
in all cases, not only when the debug log was active.  Also, cleaned up a
bunch of confusing #ifdefs and removed some chunks that had been commented
out for years, decades maybe.  ttinl(): ckutio.c, 21-22 Feb 2007.

Added NOW keyword info to HELP DATE, plus a tip about how to convert to UTC;
suggested by Arthur Marsh.  ckuus2.c, 22 Feb 2007.

When an FTP client sends NLST to the server and no matching files are found,
the server is supposed to respond with an error message on the control
channel and nothing on the data channel.  However it seems that at least one
server sends the error message back on the data channel, as if it were a
filename ("/bin/ls: blah: No such file or directory"), and on the control
channel there is no error indication ("226 ASCII Transfer complete").  At
this point remote_files() has a listfile and, if a match pattern was given,
it looks through list to see if any of the lines match the given filename,
e.g. "blah".  This makes FTP CHECK give false positives.  The problem
(diagnosed by Jeff) is that the match pattern was not given in this case, so
it takes some random default action, resulting in the spurious success
return.  Fixed by using the user's string as the pattern.  Not tested,
however, since I don't have access to a server that behaves this way.
ckcftp.c, 22 Feb 2007.

If an external-protocol file transfer fails, don't print Kermit-specific
hints.  ckuus5.c, 22 Feb 2007.

One more time with ttinl().  Got rid of the "csave" junk, which never could
have worked (which is no doubt why it was in a debugging section).  The
problem was that saving the beginning of the next packet locally did not
synchronize with the buffer clearing (ttflui()) done at a higher level,
between calls to ttinl().  So now, the lookahead code, if it finds the
beginning an as-yet unread packet, puts it back at the head of the input
queue.  This way, if the protocol engine clears the input buffer, it will
get the whole packet, not just the part after the SOH.  ckutio.c, 24 Feb 2007.

From Steven M Schweda, Saint Paul, MN: adaptation of large file support to
VMS (it was already possible to transfer large files in VMS C-Kermit but the
file-transfer display and statistics were wrong).  And a minimal adaptation
of the FTP client to VMS -- no RMS, no special VMS file stuff, Stream_LF and
binary files only, developed and tested only with UCX.  SSL/TLS is
supported.  The source-code changes are minimal; most have nothing to do
with VMS, but with header files, prototypes, and data types (e.g. ftp_port
int rather than short, various signed/unsigned conflicts) to shut up
compiler warnings.  Some of these could be dangerous in terms of
portability; I've marked them with /* SMS 2007/02/15 */.  ckcfns.c,
ckcnet.h, ck_ssl.h, ckuus3.c, ckuus4.c, ckvfio.c, ckcftp.c, ckvker.mms
(which was rewritten to actually reflect the source module dependencies),
ckvker.com (also heavily modified).  ckvker.com (the "makefile" for VMS
C-Kermit) now indludes "F" and "I" option flags for the large File and
Internal ftp features, plus better handling of Vax/Alpha/IA64 distinction.
26 Feb 2007.

Changed NetBSD targets to include -DHAVE_OPENPTY and -lutil, so they
can use openpty().  makefile, 26 Feb 2007.

Built on Solaris without and with SSL OK.
Built on NetBSD with Kerberos 5, OK.
Built on Mac OS X 10.4, regular version, OK.
Built on Mac OS X 10.4 with SSL and Kerberos 5, OK.

On VMS 7.2-1/Alpha with MultiNet 4.4A-X...

'CC' 'CCOPT' KSP:ckuus3
%DCL-W-TKNOVF, command element is too long - shorten
 \CKUUS4.OBJ "'CC' 'CCOPT' KSP:ckuus4" "KSP:ckuus4.c KSP:ckcsym.h KSP:ckcdeb.h
 KSP:ckclib.h" "KSP:ckcasc.h KSP:ckcker.h KSP:ckcnet.h KSP:ckvioc.h"
"KSP:ckctel.h KSP:ckuusr.h KSP:ckucmd.h KSP:ckuver.h" "KSP:ckcxla.h
 KSP:ckuxla.h KSP:ckcuni.h KSP:ckuath.h"

The new rule for ckuus4.c was too long.  I removed one file from the
dependency list (ckcxla.h, which will probably never change again) and that
made it OK.  Built Nonet and Net versions OK, but this is without the new
stuff.

"make f" (large-file support) on VMS 7.2-1...
'CC' 'CCOPT' KSP:ckuus4
                    if (CKFSEEK(fp,(CK_OFF_T)j,SEEK_CUR) != 0) {
........................^
%CC-I-IMPLICITFUNC, In this statement, the identifier "fseeko" is implicitly
declared as a function.

Ditto for ftello and fseeko in various other places, and then fseeko and
ftello come up up undefined at link time.

The rule for ckcftp in "make i" (Internal FTP support) had the same problem.
I removed ckcxla.h from its dependency list too, but "utime" comes up
undeclared at compile time and undefined at link time.

Verdict: neither one of the two new features can be used in VMS 7.2 or
earlier, but the code still builds OK if you don't ask for them.

VMS 8.3 on IA64...  Can't build anything:
%MMS-F-BADTARG, Specified target (WERMIT) does not exist in description file

27 Feb 2007: Changed CKVKER.COM to keep all its dependencies but use a
shorter logical name (Steven M Schweda).  The problem on VMS 8.3 is that MMS
now supports case-sensitive file systems, and so it can't find anything.
Workaround: bypass MMS (include "m" in P1).  With this, "@ckvker.com ifm"
builds OK on HP Testdrive, but I can't test the new features since outbound
connections are not allowed there.  As for fseeko(), ftello(), and utime(),
they simply are not availble prior to VMS 7.3.  It would probably be a good
idea to test for this in CKVKER.COM, but actually it is possible to install
newer C's and CRTLs on older VMS versions, so don't stand in their way.

28 Feb 2007: With additional chages from SMS, and then some further
adjustments, I was able to build the FTP version on VMS 7.2-1.  First I
tested it with GET of a binary file, but it transferred it in text mode.
After a few more attempts with PUT and GET, it crashed with "floating/decimal
divide by zero" in ckscreen, ckuusx.c line 27859.  Of course, that's the
listing line, not the source line, and I don't have a listing.

To get a listing, I deleted CKUUSX.OBJ and then did:

  $ make i "" "" "/LIST"

Surprisingly, it recompiled everything.

Anyway, the divide by zero happened in a section of code where the divisor
was not checked, but it was a section of code we should not have been
executing at all, since the file-transfer display was fullscreen, and this
was in the "brief" section.  Anyway, I added the needed check.  Again, it
recompiles everything.  Maybe there's no MMS on grumpy -- right, there isn't.

ANYWAY... Try to GET a binary file like this:

  binary
  ---> TYPE I
  200 Type set to I.
  get gkermit
  ---> TYPE A
  200 Type set to A.
  ---> SIZE gkermit
  550 gkermit: file too large for SIZE.
  GET gkermit (text) (-1 bytes)---> TYPE A

Anyway... "get /binary gkermit" downloads it, seemingly correctly (the byte
count is right).

But "put /binary gkermit.;1" results in a 0-length GKERMIT file being sent.
Here's the debug log:

FTP PUT gnfile[DISK$MSA4:[C.FDC.NET]gkermit.;1]=1
ftp putfile flg[DISK$MSA4:[C.FDC.NET]gkermit.;1]=0
zltor fncnv[DISK$MSA4:[C.FDC.NET]gkermit.;1]=-1
FTP PUT nzltor[GKERMIT]
zfnqfp 1[DISK$MSA4:[C.FDC.NET]gkermit.;1]=675
zfnqfp 2[DISK$MSA4:[C.FDC.NET]GKERMIT.;1]=31
zfnqfp 3[DISK$MSA4:[C.FDC.NET]GKERMIT.;1]=31
zrelnam result 2[gkermit.;1]
ftp sendrequest restart[DISK$MSA4:[C.FDC.NET]gkermit.;1]=0
openi name[DISK$MSA4:[C.FDC.NET]gkermit.;1]
openi sndsrc=-1
openi file number=2
zopeni[DISK$MSA4:[C.FDC.NET]gkermit.;1]=2
zopeni fp=0
chkfn=2
chkfn return=0
zopeni fixed file format - using blk I/O
zopeni binary flag at open=1
zopeni ifile_bmode=1
zopeni binary=1
zopeni RMS operations completed ok
openi zopeni 1[DISK$MSA4:[C.FDC.NET]gkermit.;1]=1
ftpcmd cmd[PASV]
FTP SENT [PASV]
FTP RCVD [227 Entering Passive Mode (166,84,1,2,233,216)]
initconn connect ok
FTP SENT [STOR GKERMIT]
FTP RCVD [150 Opening BINARY mode data connection for 'GKERMIT'.]
doftpsend2 ftpcode[STOR]=150

  Here is where the file is supposed to be read and sent but there is nothing
  in the log between the "doftpsend2 ftpcode" line and the following line.

rftimer status=1
gftimer status 1=1
gftimer status 2=1409025
gftimer status 3=1409025
gftimer s[0.000000]
zclose n=2
chkfn=2
chkfn return=1
zclose ZIFILE RMS operations completed ok
ftp getreply lcs=0
ftp getreply rcs=-1
ftp getreply fc=0
FTP RCVD [226 Transfer complete.]
ftp getreply[226 Transfer complete.]=2
doftpsend2 ok=0

Everything is OK up until we go to send the file, then it behaves as if we
got EOF immediately and so closes the data connection, and reports success;
an empty copy of the file is left on the far end.

Starting over with a text file....  PUT LOGIN.COM gets another divide by
zero.  But it happened in the code I just fixed, which is impossible.  Swell.
I recompiled everything and this time the upload worked, and downloading it
again worked too.

But a binary file still can't be uploaded.  Trying to upload a text file
after doing this seems to succeed (reports the right number of bytes sent)
but nothing appears on the far side.

SUMMARY:

  To download a text file: GET /ASCII blah.txt    (/ASCII is optional)
  To download a binary file: GET /BINARY blah.bin (/BINARY is required)
  To upload a text file: PUT blah.txt             (/ASCII switch not needed)
  To upload a binary file: PUT /BINARY blah.bin   (doesn't work)

Problems:
  . Why doesn't BINARY "stick"?
  . Why don't binary uploads work?

The culprit seems to be the VMS version of zxin().  In the FTP module,
zxin() is called only when sending binary files.  In VMS, zxin() is just
a front end for C-Library fread().  It probably needs to do just do
zminchar() in a loop, like binary mode does, but calling zzout instead
of xxout.  Or something like that.  FINISH THIS TOMORROW (debug on grumpy).

2 Mar 2007: New logs from John Dunlap.

ema-1636-log-0175.dbg: C-Kermit uploads a short file. It receives an Ack for
the Z packet it just sent, tailgated by the beginning of a Nak for the next
packet.  When the second SOH is encountered, it is put back in the myread
queue.  Then the protocol engine, to which we return the Ack, says, "I have
the packet I wanted so I'm clearing the buffer", and away go the first two
bytes of the Nak from the myread buffer.  Then, having just received the Ack
of our Z packet, we send our B, and go to read the reply.  in_chk finds 0 in
the myread buffer (which we just cleared) and 6 waiting to be read from the
comms channel, which it does, obtaining the remaining 6 bytes of the Nak,
which it properly discards.  (The reason this is proper is that, having
already received the Ack for the last packet it sent, no Ack or Nak that
arrives subsequently -- in the non-windowing case -- could possibly affect
what it does next.)  Since it hasn't yet found a good packet, it keeps
reading, and now it finds the Ack to the B, as soon as it showed up.  This
is how it's supposed to work.  No time was lost because of anything that
C-Kermit did.

ema-1636-log-0174.dbg: C-Kermit uploads a short file. It sends Data packet
#3 and receives the Ack followed immediately by the first 3 bytes of a Nak
for packet 4.  When it gets to the SOH of the second packet, it pushes it
back in the queue.  Again, input() flushes the input buffer (myread queue
and device buffer).  C-Kermit detects EOF on the file it is sending, and
sends the Z packet.  Then it reads the remaining bytes of the Nak,
which it discards, and then it finds the Ack for Z which comes in 23 seconds
later, sends the B, gets a Nak for the B, sends the B again, gets the Ack
for the B 4 seconds later, and done.  Again, it's working right and losing
no time.

The question remains: what would happen if the protocol engine did not clear
the buffer?  Would ttinl() retrieve all packets in sequence even when they
come back to back?  To test this, I had C-Kermit send a file using 30 window
slots and observed the stream of Acks in the reverse direction:

  HEXDUMP: mygetbuf read (16 bytes)
  01 25 23 59 2f 52 39 0d | 01 25 24 59 2b 26 31 0d  .%#Y/R9. .%$Y+&1.
  ttinl lookahead my_count=9
  ttinl lookahead removed=^M
  ttinl lookahead pushback SOP=^A
  HEXDUMP: ttinl got (7 bytes)
  01 25 23 59 2f 52 39    |                          .%#Y/R9
  RECEIVE BUFFERS:
  buffer inuse address length data type seq flag retries
     0     1     29212 9667     0   Y    3     0
  [\ 1%#Y]
  ...
  in_chk my_count=8
  ...
  ttinl lookahead my_count=1
  ttinl lookahead removed=^M
  HEXDUMP: ttinl got (7 bytes)
  01 25 24 59 2b 26 31    |                          .%$Y+&1
  RECEIVE BUFFERS:
  buffer inuse address length data type seq flag retries
     0     1     29212 9667     0   Y    4     0
  [\ 1%$Y]

Here we can see that the pushed-back SOH was properly retrieved next time
around, and the tailgating Ack was not lost.  This scenario repeats itself
212 times in the log, and there are no screwups.

Back to VMS FTP...  The problem with sending binary files is that zxin()
uses C-Library fopen()/fread() instead of RMS, so it can't access the input
file, which was opened by zopeni(), which is totally RMS-ified in VMS
C-Kermit.  For VMS only, I replaced the zxin() loop by a zminchar() loop
like the one used in text mode, except without the character set or
record-format conversion.  Tested by PUT /BINARY of some binary files, which
worked fine.  ckcftp.c, 2 Mar 2007.

Next problem...  VMS C-Kermit ftp client sending binary files in text mode.
Variation 1: We just send the file.  zopeni() is supposed to detect that
it's a binary file and automatically set the mode.  And it does:

  zopeni fixed file format - using blk I/O
  zopeni binary flag at open=0
  zopeni ifile_bmode=1
  zopeni binary=0
  zopeni autoswitch from TEXT to BINARY
  zopeni RMS operations completed ok

but then in gnfile():

  if (!server || (server && ((whatru & WMI_FLAG) == 0)))
    binary = gnf_binary;        /* Restore prevailing transfer mode */

Well, since VMS sets text/binary mode automatically when sending files,
this code can (and should) be skipped in VMS.  gnfile(): ckcfns.c, 2 Mar 2007.

Variation 2: BINARY or SET FILE TYPE BINARY doesn't force binary mode.  But
SET FTP TYPE BINARY does.  But BINARY does indeed call doftptyp() so what's
the problem?  We do indeed set ftp_typ to 1 but it gets reset somewhere
before we call zopeni().  But then zopeni() puts it back to 1.  Tracing
through a transfer, it looks like all of this works right, it's only that
the file transfer display says TEXT when the transfer is really in binary
mode.  This is because screen() is called before openi().  I wonder if we
can call scrft() from the ftp module...  No, that would be too easy.  OK,
sendrequest calls openi() and sets the file mode; putfile() calls
screen(SCR_FN), which prints the transfer mode.  But putfile calls
sendrequest() after it puts up the screen that says the file type.  So it
looks like sendrequest() has to call screen(SCR_FN) again if it changes the
file type.  OK, that did it.  ckuusx.c, ckcftp.c, 2 Mar 2007.

The BINARY and TEXT (ASCII) commands do not inhibit automatic type switching
in VMS.  They don't in Unix either.  They never have.  Should they?  I think
so, otherwise what good are they?  Plus we want the Kermit FTP client to
behave like the others.  I added code for this but it doesn't work, due to
the layers and layers of text/binary detection and switching and
if-this-but-then-if-that...  Anyway, no harm done.  The normal rule is:
when you PUT a file, Kermit figures out on a per-file basis whether to use
text or binary mode unless you include a /TEXT (/ASCII) or /BINARY switch
in the PUT (or MPUT) command.  ckuus[r3].c, ckcftp.c, 2 Mar 2007.

Wed Mar 7 16:21:13 2007 WROTE SHORT TEST PROGRAM for ttruncmd (the openpty
version) on Mac OS X.  On dulce: ~/kermit/ttpty.c / ttpty.sh.  It starts the
external protocol in the lower fork.  The command to run is a command-line
argument.  Sending and receiving files with Kermit works OK.  But again, the
standalone program totally fails when I use sz or lsz as the external
protocol.  So it looks like we can rule out any environmental effects of
running the code inside C-Kermit.

Mon Mar 12 16:52:20 2007: Put some effort into making ttpty.c more useful;
added a debug log.  Found that for some reason, at least on Mac OS X,
select() always timed out at the the end.  I added a SIGCHLD alarm and that
seems to handle the fork exit condition very nicely.  Now we can send (say)
a 3MB file at good speed on Ethernet (1Mcps) considering the debugging, etc,
and it terminates instantly.  But when sending a file into ttptycmd (with
"gkermit -r"), things go wrong at the end -- the Z packet is never
acknowledged.  This is reproducible.  Maybe this is a good lead....  The log
shows that select() timed out, even though the gkermit fork had not yet
exited (or finished).  It looks like gkermit sent the Ack, ttpty.c read it
from the pty and sent it out the net:

  0003: read pty=8                <-- read Ack from pty
  0003: loop top have_pty=1
  0003: loop top have_net=1
  0003: FD_SET pty_in
  0003: FD_SET ttyfd in
  0003: FD_SET ttyfd out=8
  0003: nfds=5
  0003: select=1
  0003: FD_ISSET ttyfd out
  0003: write net=8               <-- send ack to net
  0003: loop top have_pty=1
  0003: loop top have_net=1
  0003: FD_SET pty_in
  0003: FD_SET ttyfd in
  0003: nfds=5
  0009: select=0
  0009: select timeout - have_pty=1

But Ack never arrived.  This is a streaming transfer.  But nope, streaming
is not the problem.  If I disable streaming ("gkermit -Sr"), we hang in in
the middle of sending the data.  If I use small packets, we don't hang:
1000 is OK, 2000 is not.  In fact, the cutoff is 1024.  OK, TBC...

Wed 14 Mar 2007: Receiving a file thru ttpty "gkermit -e 1200 -Srd"
produces a debug log that shows that gkermit gets a lot of EAGAIN errors
when it tries to read from its stdin.  In fact, it takes 6 tries (read()
calls) to read the S packet (27 bytes).  Then when the first data packet
arrives (1200 bytes), read() never returns even one single byte.  The
timeout interval is 15 seconds and it times out repeatedly.  Added a
primitive hex dump to the ttpty debug log for each read/write (showing only
the first 24 characters and the last character, so it fits on one line).
Tried uploading a file.  The S, F, and A packets (short) are received and
Ack'd OK, but then ttpty select() times out, never receiving even one byte
from the D packet.  Clearly, when the pty driver receives a burst of > 1K
bytes, stops working.  As before, if I limit the packets to < 1K, it works
fine.

Can I send an 8-bit binary file?  Nope.  ttpty reads the binary data just
fine from the net and writes it exactly as it was received to the pty, but
the first time we write an 8-bit byte, we never hear back from the PTY
again.  But the log shows that when the initial 7-bit packets from the pty,
it looks like the PTY is not in rawmode, because these packets end with ^J
rather than ^M.  Calling pty_make_raw() on the masterfd and slavefd
explicitly, however, doesn't change anything.  It doesn't matter if I do
this in the lower fork or the upper fork.  So maybe it's the actual
semantics of pty_make_raw() that are wrong.

Thu 15 Mar 2007: Went thru all the terminal mode flags in Mac OS X; didn't
help.  Changed hex dump to show whole packet.  Put hex dump routine in a
private copy of G-Kermit.  Tried to transfer an 8-bit file, logging both
ttpty and gkermit.  Compared what ttpty received on stdin with what it sent
to the pty (same) and what was received by G-Kermit (same).  Then I realized
that my little test program was not putting its controlling terminal into
raw mode; when I did that, I could upload binary files (streaming, 2MB/sec).
And with Zmodem too (with rz; lrz doesn't work for some reason).  Looking
back at the original in ckutio.c, I see that ttptycmd() never called
ttpkt().  Maybe that was the trouble all along.  (Yup, but maybe not the
whole trouble.)

Moving back to C-Kermit and the original ttptycmd() routine, adding the call
to ttpkt(), and stripping out a lot of cruft, and moving the pty_make_raw()
code to ckupty.c, Kermit uploads and downloads (streaming) work fine in
Solaris.  Zmodem sends a file, but then the transfer hangs at the very end,
as if the signoff protocol were lost.  This happens on Solaris.  If I move
back to Mac OS X, everything works just fine.  Then, making a Kerberized
connection from the Mac to NetBSD, I can send files from the Mac with both
Zmodem and Kermit.  Receiving...  Kermit OK.  Zmodem...  Nope.  "rz:
Persistent CRC or other ERROR" (and created a 265MB debug.log!)

Fri 16 Mar 2007: ttptycmd() was for sending files with Zmodem across
encrypted connections.  But it occurred to me that it's necessary for
clear-text connections too; e.g. Telnet, where 0xff has to be doubled.  Of
course Zmodem doesn't do that itself, so there's no way Zmodem external
protocol could work when executed over a Telnet connection, and in fact
it doesn't.  I wonder why I ever thought it did.

Wed 21 Mar 2007: Back to where we left off a week ago.  Trying C-Kermit's
ttptycmd() on the Mac again, in remote mode:

 . G-Kermit send txt (kst): OK  832Kcps  
 . G-Kermit recv txt (kr):  OK  425Kcps  
 . G-Kermit send bin (ksb): OK 1000Kcps  
 . G-Kermit recv bin (kr):  OK  188Kcps  

And Zmodem:

 . sz txt (zst): OK  563Kcps  
 . sz bin (zsb): OK  714Kcps  
 . rz txt (zr):  OK  863Kcps  
 . rz bin (zr):  OK  198Kcps  

So in remote mode, everything works.  Now let's try a clear-text Telnet
connection...

 . G-Kermit send txt (kst): OK  841Kcps
 . G-Kermit recv txt (krt): OK  391Kcps
 . G-Kermit send bin (ksb): OK  811Kcps
 . G-Kermit recv bin (krb): OK  171Kcps

And Zmodem over the same clear-text telnet connection:

 . sz txt (zst): OK  91Kcps (*)

Kermit is sending sz messages like "sz 3.73 1-30-03 finished." to the
host, which tries to execute them, after the transfer is finished.
Of course "sz" is a command, but:

  sz: cannot open 3.73: No such file or directory
  sz: cannot open 1-30-03: No such file or directory
  sz: cannot open finished.: No such file or directory

Did I lose that code that dis-redirects stderr when I went back to using the
pty code from the ckupty module?  No, it's there and it's being executed.
Apparently the copy of sz I have is writing its "finished" message to stdout
because "sz blah 2> /dev/null" does not suppress it.  Starting again with
lsz instead of sz:

 . sz txt (lzst): OK  413Kcps
 . sz bin (lzsb): OK  FAILED (*)
 . rz txt (lzrt): OK  
 . rz bin (lzrb): OK

(*) Sigh.  Using lsz, we get "garbage count exceeded" errors and eventual
failure.  But using regular sz, we get the extraneous message that starts
sz on the far tend, and the resulting getty babble.

But even without changing the code, it will work one minute, and then fail
consistently the next.  For example, I was able to send files with sz
successfully over and over, but with the getty babble at the end.  Then,
after trying lsz and then going back to sz, every attempt at sending a file
quits with "Got ZCAN".  The difference has to be that Kermit always does at
least some minimal encoding of C0/C1 control characters such NUL and DEL and
IAC, and I doubt that Zmodem does.

http://zssh.sourceforge.net/ says:

  If file transfer is initiated but never completes (ie a line like :

     Bytes Sent:      0/    513   BPS:0        ETA 00:00  Retry 0: Got ZCAN

   can be seen, but transfer never completes), chances are the pty/tty on one
   of the systems are not 8-bit clean.  (Linux is 8-bit clean, NetBSD is not).
   Using the -e (escape) option of rz should solve this problem.

It doesn't, at least not with lrz.  And yes, the receiving end happens to be
NetBSD.  But it looks like the zssh people have been down this road too.

But with rz and sz, it worked.  Once.  Twice.  Three times.  But of course,
with the getty babble at the end.  This can be taken care of by doing:

  rz -eq ; cat > foo

which puts "sz 3.73 1-30-03 finished" and any other messages in foo (but you
have to type ^D to finish the cat).  Using this method I was also able to
send an 8K binary file that contained a test pattern of all 256 possible byte
values.  Then I tried a 3MB binary executable.  All OK.  So here we go again:

 . sz txt (zst): OK
 . sz bin (zsb): OK
 . rz txt (zrt): 
 . rz bin (zrb): 

Downloading fails about halfway through a fairly large file.  I tried an
even bigger file, guaranteed to be 100% ASCII; same thing -- halfway
through: "rz: Persistent CRC or other ERROR".  But it worked with a smaller
version of the same file (82K versus 2MB).  Tried again with the bigger
version, it failed in exactly the same way at exactly the same spot: byte
number 1048320.  But this is just ASCII text so it can't be a transparency
problem.  Substituting another plain ASCII file of the same size but totally
different contents, it doesn't fail (2.36MB).  Back to the previous file, it
fails again, but in a different spot (832960).  So it's not totally
deterministic.

To round things out, I tried downloading the binary test-pattern file; it's
only 8K.  This failed.

  -4, --try-4k                go up to 4K blocksize
  -B, --bufsize N             buffer N bytes (N==auto: buffer whole file)
  -e, --escape                escape all control characters (Z)
  -E, --rename                force receiver to rename files it already has
  -L, --packetlen N           limit subpacket length to N bytes (Z)
  -l, --framelen N            limit frame length to N bytes (l>=L) (Z)

Tried again with "sz -L 256 -B 256 -4aeq".  Doesn't change anything.

NOTE: Mac OS X rz 3.73 1-30-03 does not support -e.
NetBSD rz 0.12.20 does support -e.

Thu 22 Mar 2007: It occurs to me that ttpkt() might still be a problem;
maybe it's the network connection and not the pty that is not transparent
enough.  To test this theory I did "stty raw ; stty -a" and then copied all
of the flag values into ttpkt in the BSD44ORPOSIX section:

 . rz txt (zrt): OK (2.36MB file, worked 2 out of 3 times)
 . rz bin (zrb): "rz: Persistent CRC or other ERROR"

A little more fiddling with the flags and I got the 8K binary test pattern
to SEEM to download OK (in the sense that rz gave a 0 return code) but the
file itself was truncated, always at 224.  If I changed the test pattern
file to not include any bytes with value 224 (0xe0) or 255 (0xff), the
download worked.  So we have a transparency problem somewhere.  The debug
log shows that all byte values are being received from the network correctly
so the problem has to occur when we try to feed them to the pty.

But no amount of twiddling with the termios flags seems to let these
characters pass through.  Of course, since they are not in the C0 or C1
control range, "sz -e" doesn't quote them (which it does by prefixing with
Ctrl-X and then adding 0x40 to the byte value so (e.g.) NUL becomes ^X@.
Note that 255 does not cause problems because it coincides with the IAC
character; the remote Telnet server doubles outbound IACs, and Kermit's
ttinc() undoubles them automatically (as the log shows).

Trying to send a different file (a C-Kermit binary) shows that 255 is the
real killer; the file is truncated where the first one appears (at about
6K), even though some 224's precede it.  Going back to the remote-mode test,
I see the same thing happens with the binary test-pattern file, if I send it
from K95 direct to rz-under-C-Kermit-in-remote-mode.  So it has nothing to
do with C-Kermit having a network connection.  Yet if I send the same file
direct from K95 to rz, it goes OK and the result is not truncated, so it's
not Zmodem either.  The data arrives to C-Kermit intact, so the failure is
definitely in writing it to the rz process through the slave and master ptys.

BUT if I send the same file from K95 to rz-under-ttpty, that works.  What's
the difference?  Suppose I just transplant ttpty literally into C-Kermit...
It makes no difference.  When receiving the test-pattern, it truncates it
in exactly the same place.

Well, all this is on Mac OS X.  What if I move it to a different platform?
OK, building on Solaris and following the exact same procedure, ttptycmd()
doesn't even use the network connection.  I think that's because rzsz on
Solaris is hardwired to use the controlling terminal and can't be
redirected, even in a pty?

Moved to NetBSD.

 . sz txt (zst): Failed ("Got ZCAN")
 . sz bin (zsb): 
 . rz txt (zrt): OK
 . rz bin (zrb): 

Well, this is a big mess.  Sending doesn't work (or sometimes it does but
reports that it didn't).  Receiving...  well, actually it's the same thing;
the file is completely transferred but then the final protocol handshake is
lost.  The local C-Kermit returns to its prompt, but rz is still running:

  Retry 0: Got TIMEOUT
  Retry 0: TIMEOUT
  Retry 0: Got TIMEOUT
  Retry 0: TIMEOUT
  Retry 0: Got TIMEOUT
  Retry 0: TIMEOUT
  Retry 0: Got TIMEOUT
  Retry 0: TIMEOUT
  Retry 0: Got TIMEOUT
  Retry 0: TIMEOUT
  Retry 0: Got TIMEOUT

I don't see how that is even possible.  Even after I exit from Kermit the
messages keep coming, even though ps doesn't show the rz process anywhere.
Looking at the code, I see a place where end_pty() was still commented out
from the ttpty.c episode, I uncommented it.  But still:

 . sz txt (zst): Fails ("Got ZCAN")
 . sz bin (zsb): Fails instantly (but with no diagnostic)
 . rz txt (zrt): OK
 . rz bin (zrb): Fails with tons of "Bad CRC", "Gargage Count exceeded"

Conclusion for the day: I think this is hopeless.  Even if I can get it to
work somewhere, the results depend on the exact Zmodem software, how it uses
stdin/out vs stderr versus getting its own nonredirectable file descriptor,
versus the Zmodem version on the other end and which options are available
on each, versus the pty and select() quirks on each platform, and on and on.
It will be so hard to explain and to set up that nobody would ever use it.
It would be better to just implement Zmodem internally.

Fri 23 Mar 2007: Went back to the small test program, ttpty.c.  Tried
setting both the master and the slave pty to rawmode, even though I have
never seen any other software that did this.  I had it receive the binary
test pattern file; it worked.  I made a bigger test-pattern file, 3MB,
containing single, double, and triple copies of each byte in byte order and
in random order, this one was accepted too.

So it would seem that the ckupty.c module is something to avoid after all.
It's full of stuff I don't understand and probably should not undo.  So
changing C-Kermit's ttptycmd() to manage its own pty again, using openpty()
(which is not portable), I got it all to work in remote mode: Kermit
text/binary up/down and Zmodem text/binary up/down.  But in local mode on
the client side of a Telnet connection...

  zst: OK, but we still get the getty babble at the end that starts sz.
  zsb: OK, ditto.  This is with the 3MB test-pattern file.
  zrt: Not OK -- "Persistent CRC or other ERROR"
  zrb: Not OK -- got the cutoff at 224 again "Persistent CRC or other ERROR"

It's close.  But actually this was still with USE_CKUPTY_C defined.  When I
undefined it, it was back to being totally broken.  Start over.  (Check the
new cfmakeraw() code.)

Tue 27 Mar 2007: Starting over.  Back to ttpty.c.  Let's verify, VERY
CAREFULLY, that it really does work, using the most stressful of the four
tests: sending the big (3.2768MB) binary test pattern from K95 into rz
through ttpty, logging everything.  ttpty definitely receives the big file
smoothly with no errors or hiccups when I have it set to use the master side
of the pty for i/o.  The application program (Zmodem in this case) runs on
the slave, and the network and/or control program communicates with the
master.  This implies that Zmodem controls the terminal modes of the slave,
and ttpty should be concerned with those of the master.  Doing it this way
in ttpty confirms this.

Fine.  But if I tell ttpty to SEND a file with sz, nothing happens.  Ditto
with lsz.  Select times out waiting for input from the pty.  But if I
manually tell K95 to RECEIVE /PROTOCOL:ZMODEM it works OK.  Somehow sz's
initial B000000 string is being swallowed somewhere, and it's waiting for
a reply from the receiver.  sigh...  But "ttpty gkermit -s filename" works
fine.  What's the difference?  It has nothing to do with stdout vs stderr;
sz is not writing to stderr at all.  Is it some timing thing between the
forks?  Aha.  It's that I change the modes of the pty master in one fork
while sz is already starting in the other fork.

OK, good, now for the first time we have Kermit and Zmodem both able to
upload and download a large worst-case binary test-pattern file... in
remote mode.  Now taking today's lessons and fitting them back into
C-Kermit so I can try it local mode...

Using G-Kermit as the external protocol, first in remote mode...  All good:
text/binary up/down.  The "halting problem" is solved by SIGCHLD, which
catches fork termination instantly and lets ttptycmd() know there is no more
pty.  Zmodem:

  zst: OK
  zsb: OK
  zrt: OK
  zrb: OK

That's a first.  Next, repeat in local mode, in which C-Kermit is the client
and has made a Telnet connection to another host over a secure (Kerberos V)
connection:

  kst: OK     zst: ...
  ksb: OK
  krt: OK  
  krb: OK
  
It seems we can never end a day on a high note.  Somehow I seem to have
broken regular internal Kermit protocol transfers over encrypted connections
-- the en/decryption engine loses sync.  But they still work OK over a
clear-text Telnet connection.

Today's code in ~/80/dulce.tar (27 Mar 2007).

Added makefile target solaris10g+openssl.  Gathered all the standard CFLAGS
for Solaris into cdcdeb.h so they don't have to be included in every single
makefile target for Solaris.  On local Solaris 10 host OpenSSL is in
/opt/openssl-0.9.8e/.  Tried the new makefile target, works OK.  Also made
solaris10+openssl for Sun CC, but couldn't test it because I can't find any
Solaris 10 host that has Sun CC.  Built with gcc at another site that has
OpenSSL 0.9.8f-dev, all OK.  ckcdeb.h, makefile, 24 Jun 2007.

It occurs to me that Kermit transfers on secure connections might have been
broken by the changes I made back in February to ttinl() for John Dunlap.
Here, for the first time, we invoke myunrd() to push a byte back into the
input queue, and there is also some funny business with "csave", which
changed, and which an old comment notes that it has to be treated specially
when encrypting.  So it could be that the broken Kermit transfer has nothing
to do with the work on external protocols, and that putting back the
previous ttinl() will fix it.  But now I can't seem to make a Kerberized
connection from Panix to Panix, even though I can make one from Columbia to
Panix.  This means I have to build a Kerberized binary from the current
source code on either Solaris or Mac OS X.  Trying Solaris
first... [~/solaris9k5/mk5.sh] This didn't work the first time due to
undefined krb5_init_ets, which is referenced if MIT_CURRENT is not defined
(it should be for Kerberos 5 1.05 and later and we have 1.42 here), tried
again with -DMIT_CURRENT=1...  Nope, that one totally blew up in ck_crp.c.
Later, Jeff says krb5_init_ets is a no-op in Kerberos 1.4.x and later,
so I added an #ifdef (NO_KRB5_INIT_ETS) for skipping it; now it builds and
runs OK.  ckuath.c, makefile, 9 Jul 2007.

Meanwhile, using C-Kermit on Mac OS X, which makes the Kerberized connection
just fine, but still has the problem transferring files over it.  Packet log
shows:

  s-00-01-^A9 Sz/ @-#Y3~Z! z0___F"U1@A^M 
  r-00-01-^A9 Y~/ @-#Y3~^>J)0___J"U1@I
  s-01-01-^A(!Fx.x)(V^M
  r-xx-08-<timeout>
  S-01-08-^A(!Fx.x)(V^M
  r-xx-08-<timeout>
  S-01-08-^A(!Fx.x)(V^M
  r-xx-16-<timeout>

Note that S packet is sent, received, and Ack'd OK.  The F packet is sent but
is never Ack'd.  Tried this several times and noticed that it's just
receiving that is screwed up, not sending.  After ^C'ing out of the
transfer, I can still type commands, and they are executed on the far end,
but the results coming back are gibberish.  Mon Jul 9 16:08:22 2007 (come
back to this later... substitute Dev.27 ttinl for current one and see if
the problem goes away, and if so, conditionalize the new code for clear-text
connections).

Built C-Kermit with Kerberos 5 on Solaris with a version of ckutio.c that
uses the old ttinl() and transferred a file OK over a Kerberized connection.
So now it's just a matter of reconciling the old and new ttinl.  The easiest
way to do this is to have new ttinl() chain to old ttinl() if the connection
is encrypted, which is what I did and it works fine.  At some point the two
versions of ttinl() should be reconciled.  ckutio.c, 12 Aug 2007.

There was a function, islink(), used in only one place (ckuus6.c) that had
the same name as a commonly used scalar variable, and it was missing a
prototype.  Changed its name to isalink() and added the prototype (Unix
only), ckuus6.c, ckufio.c, ckcdeb.h. 12 Aug 2007.

Revisiting the ASCII and BINARY top-level commands, which are supposed to
be like in other FTP clients, but don't seem to have any effect.  I added a
new routine to the FTP module, doftpglobaltype(), that sets the global,
sticky, permanent transfer mode (ASCII or BINARY) (TENEX could be added to
if anybody asks).  These commands (now that they work) are different from
SET FTP TYPE { ASCII, BINARY }, which set the *default* transfer mode when
automatic switching fails for a given file.  ckuusr.c, ckcftp.c, 12 Aug 2007.
 (notify: Matt <mlist@cmcflex.com>)

Even though the code hasn't changed, suddenly we're getting:

  "ckuusx.c", line 5682: warning: implicit function declaration: tgetent
  "ckuusx.c", line 6183: warning: implicit function declaration: tgetstr
  "ckuusx.c", line 6262: warning: implicit function declaration: tputs
  "ckuusx.c", line 6266: warning: implicit function declaration: tgoto

in ckuusx.c on Solaris 9.  <curses.h> is still in /usr/include, dated 2002.
A quick search shows the missing functions are hiding in <term.h>, which
until now was included only in Linux.  Added a USE_TERM_H clause.  No, that
doesn't help, the prototypes are not selected at compile time; there are
#ifdefs in that file that skip over these prototypes.  I had to put them in
the code under #ifdef BUG999..#endif (I could have used a longer name like
#ifdef ADD_PROTOTYPES_FOR_CURSES_FUNCTIONS, but that would not be portable).
ckuusx.c, 12 Aug 2007.

Also:

  "ckuusx.c", line 9232: warning: implicit function declaration: creat

This is called in the IKSD dababase code, used for getting a lockfile.
creat() is a Unixism in code that is supposed to be portable.  But IKSD only
runs on Unix and Windows, so I assume the Windows C library has a creat()
function.  Anyway, suddenly the Solaris header files seem to have blocked
whatever path previously existed to the creat() prototype (which is in
<fcntl.h>), so I added an #include in the appropriate spot.  ckuusx.c,
12 Aug 2007.

Kermit functions for converting the number base -- \fradix(), \fhexify(),
\unfhexify() -- did not work with big numbers; ckradix() was missed in the
CK_OFF_T conversion.  Fixed in ckclib.c, 12 Aug 2007.

Updated the help text for ASCII, BINARY, and SET FTP TYPE to clarify the
semantics.  ckuus2.c, ckcftp.c, 12 Aug 2007.

Error messages were printed upon failure to open any of the four log file,
even with SET QUIET ON.  Fixed in ckuus4.c, 12 Aug 2007.

Built OK on NetBSD 1.3_RC3.  Tried to build secure version but the libraries
had disappeared.  13 Aug 2007.

Built OK on Mac OS X 10.4.9.  Tried the secure version, macosx10.4+krb5+ssl.
Here we get the usual pile of "pointer targets in passing argument 1 of
(function name) differ in signedness", regarding security functions, but it
built OK.  13 Aug 2007.

Reconciling the two ttinl's...  On encrypted connections myread() returns
encrypted bytes; ttinl() has to decrypt them; it wasn't doing this in the
lookahead section so I fixed it.  The new code works on both encrypted and
clear-text connections.  I removed the chaining to oldttinl(), and
oldttinl() itself.  ckutio.c, 13 Aug 2007.

  (Wouldn't it make more sense and be more efficient and less confusing
  for myfillbuf() to do the decrypting?)

When C-Kermit uses Zmodem as an external protocol, it doesn't seem to scan
files before sending them to set text or binary mode appropriately.  It's
that external protocols bypass Kermit's whole "get next file" mechanism; the
(possibly wild) filespec is simply passed to the external protocol program.
Changing this would be a very big deal.  But if only one file is being sent
(the filespec is not wild) it's easy enough to check.  I added this to the
external protocols section of the protocol module.  It can be overridden in
any of the regular ways (/TEXT or /BINARY switch on SEND command, SET
PATTERNS OFF, SET TRANSFER MODE MANUAL, etc).  ckcpro.w, 13 Aug 2007.

[FTP SEND /RECURSIVE]
Peter Crowley reported a problem with FTP recursive uploads getting the
directory tree wrong when the previous pathname was a left substring of the
new pathname (e.g. foo/bar/ and foo/bar2/).  The logic did not handle this
case and created the bar2 directory as a subdirectory of bar, rather than as
a parallel directory.  Fixed in syncdir() and tested with various edge cases.
ckcftp.c 14 Aug 2007.

  notify <peter.crowley@alumni.utexas.net>

Added CD messages to FTP BRIEF display to track the ups and downs of
recursive uploads.  ckcftp.c, 14 Aug 2007.

The OUTPUT command gave a misleading error message ("Connection to xxx not
open") when used on a serial port that was, indeed, open but was not
presenting the Carrier signal, when CARRIER-WATCH was not OFF.  Added a new
message for this, and some others.  ckuus5.c, 14 Aug 2007.

Sending from the command line, e.g. kermit -s foo, did not give an
informative error message if the file could not be found or opened.  Fixed
in ckuusy.c, 14 Aug 2007.

OK, back to ttptycmd....  It seems that back on March 27th, I got everything
working but I thought that there was still something wrong with it because
an unrelated problem so I put it aside.  The version of ttpty.c from that
date worked OK, and it looks like I updated ckutio.c from it, but that
version of ckutio.c was put aside.  Since then I have been working on the
ckutio.c version that was NOT put aside and so now I have to reconcile the
two:

  ~/80/ttypty/20070327/ckutio.c
  ~/80/ckutio.c

As a first cut I did this simply by replacing the contents of the #ifdef
CK_REDIR section of the latter with that of the former.  Of course in
Solaris this comes up with openty() implicitly declared at compile time and
unresolved at link time.  So the first task is to get HAVE_OPENPTY defined
for platforms that have it and have the others use the ttruncmd().  For
starters I put an #ifdef block in ckcdeb.h that defines HAVE_OPENPTY for
Linux, FreeBSD, NetBSD, OpenBSD, and Mac OS X.  Ones that don't have
openpty() include AIX, HP-UX, and Solaris.  Others like SCO I don't know but
I doubt it.  The real solution is to get the ckupty.c module to work but one
thing at a time...  This version is supposed work with secure builds on the
openpty() platforms, and on the others like Solaris, if an external protocol
is attempted on a secure (encrypted) connection, an error message is
printed and the command fails.  ckutio.c, 14 Aug 2007.

How to test?  Apparently I did all my testing on Panix before, and that's
where all my Zmodem builds are, but now when I build a Kerberized version
(which works if I do it on the right pool host), it won't make a local
connection, and there is no other place I can connect to that has a
Kerberized Telnet server.  I can, however, connect to Panix from here, using
the same code, but on Mac OS X...

Slight detour: Got access to AIX again (5.3.0.0).  Picky compiler, some
things needed fixing....  Also it says "1506-507 (W) No licenses available.
Contact your program supplier to add additional users. Compilation will
proceed shortly" and of course it goes kind of slow.  For some reason, I
can't do streaming transfers into AIX over a local network (to its SSH
server), but windowed transfers are OK.  Anyway, noting that we've been
using the same basic makefile target since AIX 4.2, changing nothing but the
version herald, I made a new target, simply "aix", that picks up the AIX
version automatically and sets the herald from it.  Ditto for aix+openssl,
but on this host requires setting SSLINC and SSLLIB to /opt/ssl/include and
/opt/ssl/lib.  Also the make program here was extremely sensitive to spacing
so I had to make some minor edits to get the link step to work for the SSL
version.  ckuusy.c, makefile, 14-15 Aug 2007.

Got rid of the special Panix secure NetBSD target, replaced it with a
regular one, which is invoked in the normal way by defining K5INC and K5LIB
to point to to where the stuff is hidden.  Cleaned up and modernized the
comments in the makefile a bit.  makefile 15 Aug 2007.

Changed some data types and added some casts to ckctel.c to do away with
tons of "pointer targets in passing argument 1 of 'xxx' differ in signedness"
warnings.  15 Aug 2007.

Set up Mac OS X as the testbed for ttptycmd(), with Panix as the remote
partner over a Kerberos 5 connection.  The first test is to send a 300K
text file with gkermit as the external protocol.  It worked fine, and the
debug log showed all the right components were active (namely encryption and
ttptycmd) [kermit/zmodem send/receive text/binary]:

  Kermit    Zmodem
  kst OK    zst OK
  ksb OK    zsb OK
  krt OK    zrt OK
  krb OK    zrb Failed "rz: Persistent CRC or other ERROR"

We've seen this before.  The problem is 0xff, Telnet IAC, as I proved to
myself by constructing a 3MB file that contained every byte but 0xff in every
mixture and order and transferring it successfully over the same connection.
Presumably the Telnet server is doubling IACs, whereas of course rz is not
undoubling, thus the CRC error.  This is progress.  15 Aug 2007.

Log shows that indeed every IAC in the source file arrives doubled.  Adding
code to remove the first IAC of every adjacent pair, a small test file with
different-length runs of IACs transfers OK.  The 3MB all.bin file does not.

Starting over...  I can receive a big text file with Zmodem OK.  The 3.2MB
binary test pattern that contains no IACs failed after 1.8MB, but the part
that it transferred was OK.  A second try, almost the whole thing arrived,
it stopped just 584 bytes short of the end.  Could be that file size is a
separate problem.  Making a new copy exactly 1MB long...  Well, that's
interesting, this one too stopped just short of the end.  And again, the
same thing.  When connecting back to the host, the last Zmodem packet can
be seen on the screen; it's as if the local Zmodem exited before reading
the last packet...  But OK, if I change the options on the remote sz
sender to use small blocks, etc, then it works.

Now, changing from the 1MB no-IAC-binary test pattern, to the 1MB all-values
test pattern, we fail after 81K.  But the part that was transferred is
correct.  Second try, same thing, but 57K.  Third: 40K.  Each time, upon
connecting back, the session is completely dead.

IF I HAVE TO undouble IACs for incoming files, don't I have to double them
going out?  To send a block to net we just call ttol(), but ttol() doesn't
do any doubling (because Kermit protocol always quotes 0xff).  To see what
happens, I changed the ttol() call to ttoc() in a loop that doubles IACs.  I
tested this by sending the full 3.2MB test pattern, which worked fine.

For receiving, it's slow but it works OK with files that don't contain IACs
(my concern was that IACs might appear in outbound files or in Zmodem
protocol messages).  It receives the 1MB no-IAC test pattern, so there are
no problems with protocol or timing.  But the full test pattern always gets
cut off, but at different points, as before, with the remote session dead.
Changing the Zmodem receiver from rz to lrz on the local end (since the
sender on the remote end is lsz) does not change the behavior.

Anyway, I went back and replaced the byte loop with something more
efficient, and it goes about 20 times faster.  But this doesn't help either,
it only makes it fail faster.  But aha, what if a doubled IAC is broken
across successive pty reads -- we have to make the "previous character"
memory persistent.  Well, that was a good insight, but it still didn't fix
it.  The log shows the IAC handling code is working fine.

What does sz say?  Capturing its stderr to a file... "Retry 1: Got ZCAN".
Next time: "Retry 1: Got TIMEOUT".  Next time: Got ZCAN.

Trying different Zmodem options...  apparently I don't need to use short
blocks.  But I do need to use -e, probably because of Telnet NVT treatment
of carriage return; without -e, there is a "persistent CRC error".  -O
disables timeouts, but this makes no difference.

OK, we still have two Big Problems:

 1. When a long file has no IACs, the final < 1K of the file is not received.
 2. When a long file has IACs, the transfer generally stops very early.

Problem 1: the transfer consistently fails less than 1K from the end of the
file.  Upon CONNECT back to the host, a big Zmodem packet is sitting there
waiting to be read, which means ttptycmd()'s copy of rz is terminating
early.  Can we catch it in the debug log?  Doing this takes forever and
writes a GB to the disk...  And then the problem doesn't happen.  Also, I
can receive a HUGE text file almost instantly with no errors at all.

Switching to lrz on the receiving end, now I see the error messages, about
300 lines like this:

  Retry 0: Garbage count exceeded
  Bytes received:  872352/1000000   BPS:85464  ETA 00:01  Retry 0: Bad CRC
  Bytes received:  892448/1000000   BPS:86690  ETA 00:01  Retry 0: Bad CRC
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Retry 0: Got ERROR
  Bytes received:  898336/1000000   BPS:84293  ETA 00:01  Retry 0: Bad CRC
  Retry 0: Garbage count exceeded
  Retry 0: Garbage count exceeded
  Bytes received:  900384/1000000   BPS:83751  ETA 00:01  Bad escape sequence
  2fRe
  try 0: Bad data subpacket
  Bytes received:  941472/1000000   BPS:86191  ETA 00:00  Retry 0: Bad CRC
  Retry 0: Garbage count exceeded

Even when it succeeds, it gets these.  But if I receive a text file, no
matter how big, no errors or retries or timeouts at all.  So it appears that
there is only one problem: a big-time lack of transparency regarding 8-bit
and/or control characters.  The odd thing is, it's not that the characters
can't get through -- they all can -- but they seem to cause transitory
blockages.  16 Aug 2007.

Cleaned up the remaining pointer signedness warnings in ckutio.c, but this
was a mistake, it broke Kerberos connections completely.  Undid the changes.
ckutio.c, 17 Aug 2007.

Changed all return() in the fork()==0 section of ttptycmd() to exit().
ckutio.c, 17 Aug 2007.

Tried explicitly setting the slave pty to rawmode.  Makes no difference.
Tried using the Mac OS X (curses) raw() function, and also system("stty
raw"); still no difference.  Tried doing all of these in different
combinations and orders.  I found one combination that cuts the errors about
in half, and the transfer of the no-IAC test pattern almost always succeeds
(but it's slow).  Anyway, it doesn't help much with the test pattern that
contains IACs.  Well, the code is more solid than it was before but
functionally we have not advanced much if we can't download a binary file
with Zmodem!  On the other hand, we can upload them, and we can transfer
text files in both directions, which is an improvement over the previous
situation, in which the entire session would hang due to loss of
synchronization of the encryption stream.

Tried adding -funsigned-char to CFLAGS of Mac OS X target.  It does not
make the "signedness" warnings go away and it doesn't change the runtime
symptoms.

I tried a simpler version of pty_make_raw(), the one from Serg Iakovlev, but
it was a total failure.  That's encouraging though, because it indicates
that pty_make_raw() is the right place to be working.

Then I made pty_make_raw() set or unset every single terminal flag
explicitly.  This made no difference, but didn't hurt anything either.

Then I made pty_make_raw() explicitly set all the c_cc[] characters to 0
(but left c_cc[VMIN] as 1).  This made no difference either.

I checked pty_make_raw() against ttpkt() and the only difference I found in
the terminal flags is that ttpkt() sets IGNPAR thinking it means "ignore
parity errors" when really it means "discard any character that has a parity
error" (at least according to Iakovlev) -- exactly the opposite.  But I
tried it both ways, no difference.  17 Aug 2007.

I noticed that even Zmodem text receives can fail.  They don't get any
errors, they just get cut off shortly before the end.  (But usually they
succeed, and fast too, like 500K cps).

What if I don't call pty_make_raw() at all on the slave pty?

zrt: EESSSSSSSS: 80% good (E = stopped just before end but no other errors)

zrb no-IAC test pattern, short blocks:
 1. S/5 (success with 5 screens of errors.
 2. S/7
 3. S/7
 4. S/6
 5. E/7 (failed just before end)
 6. S/7
 7. S/6
 8. S/6
 9. S/6
10. S/4

So, lots of errors, but it recovered 90% of the time.
Next, same thing, but without requesting short blocks:

 1. E/5
 2. S/5
 3. E/4
 4. S/5
 5. S/5
 6. S/5
 7. X/0 (hard failure right away: "Got ZCAN"
 8. S/5
 9. S/5
10. S/5

So it doesn't look like short blocks make that much difference.  Now what if
I turn off prefixing?  Bad CRC, fails immediately every time.  Putting back
pty_make_raw(slave), it still fails hard.

Tried a new strategy with pty_make_raw(): rather than modify existing flags,
I set all flags to 0, and then turn on only those few that we need like CS8.
Now we get only 2.5 screens of errors instead 4-7 and the transfer rate is
higher for binary files (all of the previous ones were under 100K CPS, while
for text files it was 400-500K CPS):

 1. S/2 195669 CPS
 2. S/2 194720
 3. E/3
 4. S/2 192550
 5. S/3 192325
 6. S/3 145066
 7. S/2 200689
 8. S/3 188948
 9. S/2 209461
10. S/3 181991

I noticed that there was no TIOCSTTY ioctl in the pty/fork setup sequence,
which is recommended somewhere, so I tried that and it was a disaster; the
entire session hung.  I took it back out.  18 Aug 2007.

Tried some transfers over a clear-text (not encrypted) connection with the
same results: smooth, fast transfer of a big text file (400K cps); rocky but
successful transfer of the no-IAC binary pattern file (135K cps).  Switching
back to ttruncmd(), the same binary file is received at 1.5M cps, and the
no-IAC binary file totally fails after too many "Bad CRC"s; and we already
know that any file that contains IACs will fail.  One might say that
ttptycmd() is better in every respect than ttruncmd() except in speed
(when it works).

Let's see if ttyptycmd still works in remote mode (to local K95):
 . sz / text works, but slowly.
 . lsz / text works but some wierd errors are reported.
 . lsz / binary / no IAC doesn't work at all (CRC-32 mismatch for a header;
        Unexpected control character ignored: 13, etc).
 . sz / binary / no IAC works OK but slow.
 . sz / binary / full test pattern with IAC works OK but slow.
 . Sending text into rz fails completely.

What about ttruncmd() in remote mode?
 . send /text works, fast.
 . send /binary works, fast. 
 . receive /text works, not so fast but not bad.
 . receive /binary works, not so fast but not bad.

So we use ttruncmd() for remote mode, and we use it for local mode
serial-port and modem connections, and we use ttptycmd() on network
connections because (a) they might be encrypted, and (b) even if they are
not, they use some protocol that we have to handle, e.g. Telnet, Rlogin.
19 Aug 2007.

Discovered that Sending binary files no longer works.  Text is OK, binary
transfers don't even start.  This happens on both encrypted and clear-text
connections.  ttptycmd() is being used in both cases.  But oddly enough,
receiving binary still works as before.  What did I break, and when?
Oh, it was just the script, when I changed it from using sz to lsz.  Putting
it back to sz makes it work, even with the full 3.2MB binary pattern with
IACs.

I backed off the changes I made to ckctel.c to suppress some warnings, in
view of the fact that similar changes to ckutio.c broke things so badly.
19 Aug 2007.

If sz is not given the -e flag, it sends control characters bare, except ^P,
^Q, ^S, and ^X.  ^X is the control prefix, so ^A is sent ^X followed by A.
With -e, all C0 control chars are prefixed, but with ^X, which is, of
course, a control character.  Interestingly, the C1 analogs of ^P, ^Q, ^S
(but not ^X and, unfortunately, not IAC) are also prefixed.  -e makes no
difference for 8-bit characters.

If we have a Telnet connection and the server is in ASCII (NVT) mode, CR is
always followed by LF or NUL.  Well, it seems the server is putting us
(Kermit) in binary mode in this case, but staying in ASCII mode itself.
Added code to handle NVT byte stuffing and unstuffing in each direction
independently, according to the TRANSMIT_BINARY state in that direction.  I
made a file containing just the bytes 0-31 and 127 and 128-159 and 255 (66
bytes all together) and sending it from the host to C-Kermit, the local log
shows that every control character was received correctly and all TELNET
conversions were done right -- NUL removed after CR (and only after CR); IAC
removed after IAC (and only after an IAC meant as a quote).  For the first
time, I can receive the 1MB all-values test pattern, but there are still
tons of (correctable) CRC errors, so the transfer rate is really awful, like
about 5% of what we get with a text file (25Kcps instead of 500).

Further experimentation shows that the fundamental transparency problem is
fixed; we can receive short files (say, 1K or less) containing absolutely
any byte values in any combination with no errors at all.  But once the file
size reaches (say) 10K, we get CRC errors, like one every 2 or 3K of data.
These are not deterministic.  In successive transfers of the same file, they
come in different spots.  It's tempting to blame pty buffer overruns, but
then text files would show the same behavior.  When a binary file size
exceeds, say, 1MB, the chances of successful completion go way down,
independent of whether my external protocol is rz or lrz.  I like lrz better
because the error reports come out on the screen as the transfer is going
on.  Trying to download a real-world binary file -- a 2.2MB C-Kermit
executable -- I get 4500 error messages but the transfer evenually succeeds,
with an effective throughput of 21Kcps.

Actually it turns out that "sz -a somebigtextfile" (2.2MB) also gets a lot
of CRC errors.  The -e flag (escape all control characters) makes the same
big text file transfer with few or no errors.  It's not sure-fire.
Sometimes no errors, sometimes one or two, and sometimes a fatal error that
kills the transfer.

With binary files... a 32K binary file seems to make it every time.  40K
fails about 50% of the time.  48K fails 60% and every time it fails, it has
created a partial file of exactly 32K (32768 bytes).  96K fails 9 out of 10
times, when it fails, the partial file is always 0 bytes, or 32768, or
65536, but that just means that rz's file output buffer is 32K.

Why, then, do binary files cause trouble if it is not a solid transparency
problem?  If a certain file can get through once, why can't it get through
every time?  When a character arrives at the pty, the pty driver probably
takes a different path through its code, checking the terminal flags that
would affect that character.  I tried making Kermit's network read buffers
very small but, surprisingly, this made things worse.  I also tried making
them very much bigger, which didn't help either.  24K still seems to be the
right size.

So, is it that some characters take longer to process than others?  So long
that data is lost due to lack of flow control between TCP and the pty?  One
way to test this theory is to slow Zmodem down.  I tried "-l 32" which,
according to the man page, tells sz to "wait for the receiver to acknowledge
correct data every N (32 <= N <= 1024) characters.  This may be used to
avoid network over-run when XOFF flow control is lacking."  Makes no
difference.  I also tried the -w (Window) switch, ditto.  In fact there are
all sorts of options to set the "window size", "packet length", "block
size", and "frame length", but with no explanation of what these mean or how
they are related.  If I crank everything down to minimum value:

  lsz q -L 32 -l 32 -w 1

I get 50% success with the 96K file instead of 10%.  Adding -e, oddly
enough, made it worse.  I also tried setting the environment variable
ZNULLS to different numbers like 512, no help there either.

I tried making the read-from-net-write-to-pty buffer small (1K) but leaving
the pty-to-net one big.  This improves chances of success, but it's
intolerably slow (3Kcps when the connection is capable of 500K).

I also changed the write-to-pty operation from a single write() call of
possibly many K characters to a byte loop, one write() per byte.  Same
result: success (but still about 300 recoverable errors), throughput 3Kcps.
20 Aug 2007.

With ttptycmd() configured to write to the pty in a byte loop, it is
possible to delay each write.  Adding a 10msec delay per character results
in a transfer that runs at about 20 cps and (for the 96K test file) would
take about 80 minutes to complete.  And yet it still gets just as many
errors.  So it's not a matter of timing either.  The errors come, on
average, every file 388 bytes, but not at regular intervals.

I tried the TIOCREMOTE ioctl on the pty master, as discussed somewhat
obliquely in the Mac OS X "man pty" page; "This mode causes input to the
pseudo terminal to be flow controlled and not input edited (regardless of
the terminal mode)" -- sounds like just the ticket but it made no
difference.  Actually, looking at a man page on another OS (Solaris), it
says this is only for lines of text, EOLs are supplied, so that would mess
up the protocol.  So remember: don't use this.

Tried without O_NDELAY; the behavior was the same but the speed was much
slower.

Tried switching back to the ckupty.c routines on Mac OS X and found that it
works now the same as with openpty(), except that I seem to get more getty
babble at the end.  But this means I can run some tests on Solaris.  I moved
the entire test environment from Mac OS X 10.4.9 to Solaris 9.  But it
doesn't work at all.

Trying to figure out the ckupty.c modules again.
 . do_pty() calls pty_getpty() which returns in arg1 the fd of the pty master.
 . Then it creates a pipe as a way to tell when the child dies
 . Then it creates a fork:
    - The parent does a blocking read from the pipe
    - The child calls getptyslave() to get the pty slave
      and writes one byte to the pipe
      and then execs the command it's supposed to run
Note that the file descriptor of the slave is known only to the lower fork.
Therefore the lower fork is the one that has to set all the tty modes, etc.
I took care of all that but the ckupty.c method doesn't work at all on
Solaris.  But it works "fine" on Mac OS X (the 32K all-bytes test file
transfers instantly with no errors, but the 96K one errors out).

The problem on Solaris is that pty_make_raw() fails on the masterfd (but not
on the slavefd) with errno 25 "ioctl inappropriate for device".  It doesn't
matter whether I do it in ckupty.c or ckutio.c.  I found a web page on
kde.org that says Solaris does not allow tcget/setattr() on a pty master.
But the Sun "knowledge base" is not open to the public.  Well, presumably
changes made to the slave are reflected in the master (comments in Solaris
telnetd seem to confirm this...)  Let's come back to Solaris later.

Moving to a Linux with lrzsz installed...  Built a Kerberos 5 version with
USE_CKUPTY_C.  Like on Mac OS X, it transfers short files OK and chokes on
longer ones.  Switched to openpty(), it behaves the same.  So the problems
on Mac OS X are evidently not OS-specific, which is good I guess, since that
means finding the way around them will apply to more than one platform.
21 Aug 2007.

Look into TIOCSCTTY again.  On System V based OS's, opening a pty acquires a
controlling terminal automatically.  On BSD-based OS's, no; you have to use
the TIOCSCTTY on the slave file descriptor to give it one.  I'm not sure why
a controlling terminal would be needed, except that without one, the virtual
device "/dev/tty" does not exist for the process that runs on the pty, and
maybe the application that runs there (e.g. rzsz) checks for it.  On the
downside, having a controlling terminal opens the process up to terminal
interrupts like SIGINT and SIGQUIT.  Until now I have not been using this
ioctl().  Results (in Linux):

  With TIOCSCTTY: 96K all-bytes test: 11 screens of errors, then success
  Without TIOCSCTTY: exactly the same.

Tried the same thing with TIOCNOTTY instead of TIOCSCTTY, with exactly the
same results (no effect whatsoever).

There has to be a way to make this work, because Zmodem works through
telnetd, which basically the same thing as ttptycmd(): a relay between the
network and a pty.  ttptycmd() is like telnetd backwards.  Modern telnetds
are not much help; they don't access ptys or the network directly, they go
through "mux" devices so I can't see what they're doing to get transparency
and flow control.  An old BSD telnetd uses packet mode but that would be a
big deal...

I tried ignoring various signals like SIGTTOU and SITSTP, since some Telnet
clients do this.  No effect, no difference.  Anyway, in Linux the transfers
almost always finish OK despite the many errors.  There is just some trick
I'm missing to make the pty accept a stream of arbitrary bytes without
hiccuping.

What about Solaris, which uses ckupty.c?  In streams-based OS's, where line
disciplines and whatnot are pushed on top of the pty, it looks like the pty
module saves the file descriptor of the "bare" slave pty (as 'spty') before
pushing things onto it, and then later uses spty rather than the regular
slave pty file descriptor when getting/setting terminal modes.  I'm not sure
what this is all about but it's definitely SysVish...  It happens if
STREAMSPTY is defined, but I noticed that STREAMSPTY is never defined
anywhere.  I tried defining it so we take an entirely different path through
the code.  It made absolutely no difference.

Then I noticed that HAVE_STREAMS is not defined for Solaris either.  Tried
defining it, but the session didn't work at all, no i/o.  Removing the
HAVE_STREAMS definition but keeping the STREAMSPTY defined, I rebuilt and
tried "set host /connect /pty emacs".  I got an EMACS screen but could not
type anything into it, which means that STREAMSPTY should not be defined
either.  Removed the definition and "set host /pty" works again.  So what's
the problem with ttptycmd()?

In fact, ttptycmd() works on Solaris with Kermit as the external protocol,
but not with Zmodem, not even with text files.  So again, there is no
fundamental problem with the code or the logic, it's Just A Matter Of
Transparency to control and/or 8-bit characters -- some trick I don't know
about.

Looking at the Solaris debug log...  I see that ckupty.c is calling
init_termbuf() to set the tty modes of the master, not the slave, and
set_termbuf() to set them, but you can't do that in Solaris, error 25.  This
is in getptyslave().  Shouldn't getptyslave() be setting the tty modes of
the slave, not the master?  I changed it to do this, but like all other
changes, it made no difference.  I checked to make sure that after the change,
"set host /pty /connect emacs" still worked and it did.

And then what...  I had some code to redirect stderr in ckupty.c that was
not being executing due to a typo.  When I fixed the typo, poof, Zmodem
binary transfers started working, or working as well as they work in Linux
and Mac OS X.  It turns out that if I don't redirect stderr, sz and rz
just don't work.  But lsz and lrz do.  But if I do redirect it, I don't see
the progress messages from lsz/lrz.  22 Aug 2007.

Built on HP-UX 11i v3 (B.11.31 U ia64) with optimizing compiler, got tons of
picky warnings, but it finished and linked and runs OK.  Many of the
warnings were like this:

  "ckucns.c", line 1606: warning #2068-D: integer conversion resulted in a
  change of sign:   tnopt[0] = (CHAR) IAC;

IAC is defined as 255 in ckctel.h.  If I define it as 0xff, I don't get the
warnings.  I changed the definitions of all the Telnet commands to be in hex
notation rather than decimal.  If cuts way down on the HP-UX warnings and
doesn't seem to cause problems elsewhere.  ckctel.h, 23 Aug 2007.

Now it looks like Solaris is working but then it hangs at the end.  It
appears as if the ckupty.c module is blocking SIGCHLD.  Debug log shows that
when the transfer is complete, we received IAC DM (Telnet Data Mark) after
sz's last gasp and before the shell prompt is printed.  But calling
tn_doop() in this case is a mistake because we are reading the number of
bytes that we know are available in a counted loop, but tn_doop() would
consume an unknown number of bytes and we would never know when to exit the
loop.  Anyway, C-Kermit doesn't do anything with DM.  Skipping over
tn_doop() (and not writing out the Telnet command bytes) fixes the hanging
condition at the end, even though SIGCHLD is never raised.  ckutio.c,
23 Aug 2007.

Some tests, Solaris to NetBSD over K5.
zst sends ascii.txt, a 2.36MB ascii text file (Kcps / Errors).
zrt receives the same file:

  zst 587/0 526/0 542/0 434/0 423/0
  zrt 827/0 800/0 847/0 FAIL  610/0

So text is good.  Binary not so good.  Here we transfer the 1MB all-bytes
pattern file.  zrb receives it successfully, but with 1248 errors, at only
15Kcps.  Sending the same file out always fails:

  Begin 20070823 16:32:07: SEND BINARY all2.bin [sz]
  Sending: all2.bin
  Bytes Sent:   5600/1000000   BPS:12446    ETA 01:19   FAILURE
  End 20070823 16:32:13
  Elapsed time: 6.617992999999842
  cps = 151103.2121067556
  lsz: caught signal 1; exiting

Decided to move to Linux but found that something is screwed up in Linux
C-Kermit with tilde expansion:

  send ~/testfiles/all.bin

doesn't expand at all (but it did yesterday!).  The problem was in the
ancient, ancient realuid/setuid handling code; real_uid() no longer works in
Linux.  I worked around this in whoami() by setting ruid to getuid() if
real_uid() returned a negative number.  Maybe dangerous, worry about it
later.  ckufio.c, 23 Aug 2007.

ANYWAY... after fixing that, I tested zsb on Linux, and it's broken there
too, using openpty(), so it's nothing to do with ckupty.c.  After sending
the first Zmodem data packet, it just hangs, nothing comes back.  In text
mode it gets farther, but then the same thing happens.  Captured stderr from
rz on the far end:

  Bytes received:     608/1000000   BPS:21137  ETA 00:47  Retry 0: Bad CRC
  Bytes received:     864/1000000   BPS:23540  ETA 00:42  Retry 0: Bad CRC
  Bytes received:    1120/1000000   BPS:25003  ETA 00:39  Retry 0: Bad CRC
  Bytes received:    5696/1000000   BPS:56988  ETA 00:17  Retry 0: Bad CRC
  Bytes received:    9120/1000000   BPS:62227  ETA 00:15  Retry 0: Bad CRC
  Bytes received:    9376/1000000   BPS:60766  ETA 00:16  Retry 0: Bad CRC
  Bytes received:    9632/1000000   BPS:60361  ETA 00:16  Retry 0: Got TIMEOUT
  Retry 0: Sender Canceled
  Retry 0: Got ZCAN

The local sz, however, doesn't give any error message.  ZCAN means: "other
end canceled session by sending 5 ^X's" (or user typed them).  What actually
happens is that ttptycmd()'s select() times out waiting for something from
the Zmodem partner and ttptycmd() itself kills the sz fork with SIGHUP.
When lsz receives SIGHUP it sends the ZCAN.  So the real problem is that
after some point we're not receiving anything.

I changed the timeout from 4 seconds to 30 seconds and now I see it just
stops for long periods of time and then resumes.  The lrz log on the
receiving end shows tons of timouts, CRC errors, and other errors.  The
local log shows that lsz wound up sending ZCAN (2 x (10 x ^H, 10 x ^X)).

Moving on to another problem...  Turns out Ctrl-C (SIGINT) is working right
after all.  Since I'm using my test scripts like kerbang scripts, Ctrl-C
exits through trap(), as it should, closing the connection and cleaning up.
If I start Kermit and tell it to TAKE the script, then Ctrl-C brings me back
to the prompt with the connection still open (as it should).  However, until
now I haven't done anything about the fork or the ptys.  Added code to
trap() to kill the fork and close the master pty.  ckuusx.c, 24 Aug 2007.

Added code to try to break the deadlock.  If select() times out, but we have
stuff to write either to the pty or the net, try to do it anyway, even
though select() did not say we could.  But this doesn't help because when
select() times out we don't have anything to write.  The problem is that
after receiving that last packet from the remote rz, the local lsz doesn't
seem to do anything, as if the lower fork wasn't running (and to confirm
this hypothesis, sometimes I noticed that when I Ctrl-C'd out of this, the
transfer would take off again).

Backing up and testing with gkermit rather than zmodem:

 kst ripple.txt [824K] OK
 kst ascii.txt [1359K] OK
 krt ripple.txt -- FAILED

It seems that we can't handle streaming.  If I set up krt to disable
streaming on receipt, it works OK.

 krt ripple.txt [824K] OK
 krb all2.bin  [1000K] OK

So here we have no trouble sending but big trouble receiving unless we
disable streaming.  Whereas with Zmodem we have trouble receiving.

But this wasn't happening before, what changed?  Using C-Kermit on the far
end to receive the file with debug log on, I see that it is sending 4K data
packet after 4K data packet, with the local gkermit silent, as expected.
About midway through the transfer, the local Kermit sends an error packet
"Transmission error on reliable link".  Looking at G-Kermit's debug log...
It receives the first five 4K data packets OK, but gets a CRC error on the
fifth one, and sends the Error packet.  So it has received a stream of
20-some thousand bytes OK and then messes up.  That number sounds a lot like
ttptycmd()'s buffer size.  I changed the buffer sizes to be different:

  Read from pty and write to net: 4K  
  Read from net and write to pty: 1K

This time it received the first 4K packet and failed on the second one.
Then I increased the buffers to 98K each, expecting to receive lots more
packets successfully but it bombed out on the 5th one.  But that's good, it
confirms there's no logic error in the buffer management.  Just to make
sure, though, let's set the buffer size smaller than the packet size and
disable streaming.  In this case we get 4 good data packets and a CRC error
on the 5th one and so we request retransmission, and the next 8 times it
arrives it gets a different CRC error, but the 9th copy is OK.  Then the
next packet comes and it gets a CRC error every time.  And this is nothing
but plain ASCII text.

Switching to remote mode:

  REMOTE=1 kk kst

(after tricking myself because it was using ttruncmd() for this...) I see
that nothing works at all.  What did I break?  24 Aug 2007.

Fixed ttptycmd() to restore console modes after a remote-mode transfer.
ckutio.c, 25 Aug 2007.

Noticed that error codes like ESRCH are not available in all modules.
That's because of some complicated in #ifdefs in ckcdeb.h that wind up not
always #including <errno.h>.  But I notice that ckutio.c includes it
unconditionally with no ill effects, and so does ckvfio.c.  Does any version
of Unix at all not have <errno.h>?  Added a catch-all clause to ckcdeb.h to
#include <errno.h> (in UNIX only) if, after the other clauses, ESRCH was
still not defined.  ckcdeb.h, 25 Aug 2007.

Now back to debugging ttptycmd()...  Remote-mode transfers with ttptycmd()
were broken in two places, maybe as long as 2 weeks ago (this would have
affected non-network transfers too, which I can't test any more).
The logic was missing in a couple places for the non-network and/or
non-Telnet and/or non-encrypting connections (if statements with no else
parts).  Fixed in ckutio.c, 25 Aug 2007.

Testing remote mode:

 kst OK   zst OK
 ksb OK   zsb OK
 krt OK   zrt OK
 krb OK   zrb OK

Functionally it all works but there are hitches with Zmodem as always.
When sending to K95:

 . If I send with lsz, there are hundreds of "Subpacket too long" errors,
   and the transfer is very slow, but it succeeds.

 . If I send with the 1994 Omen version of sz, transmission is instantaneous
   and without errors, but then it hangs at the end.

 . If I bypass C-Kermit and send direct from lsz or sz, both work fine.

So clearly the ptys are getting in the way.  The hanging at the end would be
caused by the sz process closing before its last output reached the master
pty.  It would need to do some form of flushing and/or pausing at the end
but there's nothing I can do about that; these programs were not designed to
be used in this way.  Anyway, it only seems to happen with files longer than
100K.

For local mode, testing in Solaris over our Kerberos 5 connection again:

 gkermit  lrzsz
 kst OK   zst FAIL
 ksb OK   zsb FAIL
 krt OK   zrt OK but with errors
 krb OK   zrb FAIL

If I use Omen rzsz as the external protocol (e.g. with zst), it blocks
redirection and it sends the file to my terminal, rather than over the
connection.  This would probably be because it finds out the device name of
the job's controlling terminal and opens it, to prevent redirection.  This
is hard to prevent in Solaris because there is no TIOCSTTY ioctl().
Supposedly the same thing is accomplished by closing and reopening the slave
pty after doing setsid().  I added code to do this, but it made no
difference.  (If I use lsz instead of sz, it is indeed redirected, but jams
up after about 15K.)  ckupty.c, 27 Aug 2007.

On Mac OS X with sz 3.73 1-30-03, however, the redirection works, so I
assume it would also work in Linux, FreeBSD, NetBSD, etc, too.  Doing the
full test suite on Mac OS X:

 gkermit   lrzsz          rzsz
  kst OK   zst FAIL (1)    OK
  ksb OK   zsb FAIL (2)    OK
  krt OK   zrt OK   (3)    OK for 100K file, fails for longer.
  krb OK   zrb FAIL (4)    OK (1MB all-bytes test pattern)

(1) 64K file OK every time; 100K file fails every time.
(2) 10K file fails every time.
(3) Succeeds with 800K file but gets a few recoverable errors.
(4) Succeeds with 48K binary file with some errors, fails with longer ones.

So actually it looks pretty good, it's just that lrzsz messes up.  When
sending with lsz if I include -L 512 it sends the 100K test file with no
errors, but still chokes on longer ones.

Testing on Mac OS X again, but this time over a clear-text Telnet connection:

 gkermit        lrzsz    rzsz
  kst OK   zst  FAIL(1)   OK
  ksb OK   zsb  FAIL(2)   OK
  krt OK   zrt  OK(3)     OK
  krb OK   zrb  FAIL(4)   OK 

(1) Almost worked, finished 777K out of 824K without errors.
(2) Got tons of errors, failed in first 30K out of 1000K.
(3) OK for 100K file but fails for larger.
(4) OK for 48K binary fail but fails for larger.

Maybe see if we can do without the OPENPTY part.

TOMORROW -- just clean up the code, add some SET / SHOW / HELP commands,
document it, and move on.

Note: In K95, SET WINDOW sets the Zmodem packet length, 32 - 1024, multiple
of 64.

SEE ~/80/external.txt

Changed ftp port from int to unsigned int.  ckcftp.c, 30 Aug 2007.

Tried again to build KRB4/KRB5/SSL/TLS version for Solaris 9.  Had to update
the build procedure again, of course, because of new file and directory
names, but ran into problems anyway because the
cu-solaris9g+krb5+krb4+openssl+shadow+pam+zlib target was calling another
target that did not know about the hardwired pathnames.  Integrated the two
targets and tried building again.  It actually compiled ok (but with lots of
warnings from the security modules), but failed at link time with
krb5_init_ets not found; fixed that with an #ifdef NO_KRB5_INIT_ETS, now it
builds OK but without the ftp client.  Tried building it WITH the FTP and
that was OK too, no changes needed except to the build procedure.  12 Feb
2008, that is: C-Kermit 8.0.212 : 20080212.

Tried to build with -DCK_SRP and -lsrp but:

  hash_supported                      ckcftp.o
  hash_getdescbyname                  ckcftp.o
  hash_getdescbyid                    ckcftp.o
  cipher_getdescbyname                ckcftp.o
  krypto_delete                       ckcftp.o
  krypto_new                          ckcftp.o
  cipher_supported                    ckcftp.o
  krypto_msg_priv                     ckcftp.o
  krypto_msg_safe                     ckcftp.o
  hash_getlist                        ckcftp.o
  cipher_getlist                      ckcftp.o
  cipher_getdescbyid                  ckcftp.o

Sent mail to Tom Wu and backed off for now.  makefile, 14 Feb 2008.
(Tom Wu never answered; seems like SRP is defunct.)

The ".blah = xxx" form of variable assignment only worked for variables
names of length 22 or less, noticed and fixed by Wolfram Sang.  ckucmd.c,
5 Mar 2008.

In "set host /pty ssh ..." connections, the INPUT command suddenly stopped
working.  This is in Solaris 9.  It happens with all 8.0.* versions of
C-Kermit, so it's nothing to do with ttptycmd().  Added some debug()
statements but they don't show anything.  Turns out there wasn't a problem
after all.  Wed Mar 26 16:04:53 2008

Changed cmifi() to not print "?No files match" (or whatever) if SET QUIET ON.
ckucmd.c, 26 Mar 2008.

Added \v(remoteip) for the IP address of the host we're connected to,
and \v(inmessage) for INPUT status messages corresponding to \v(instatus).
ckuusr.h, ckcmai.c, ckuus[24].c, 26 Mar 2008.

Made \fkeywordval() strip braces/quotes from the right-hand side so we can
handle things like:

  password="stringwithspaceatend "

ckuus4.c, 6 Aug 2008.

Added invisible PUTENV command for UNIX only.  Value should not be enclosed
in doublequotes.  Requires lge \v(buildid) 20080826.  ckuusr.[ch], 26 Aug 2008.

Added SET VARIABLE-EVALUATION { RECURSIVE, SIMPLE }.  This is highly
experimental, but also highly desirable if it works out.  SIMPLE inhibits
the default recursive method of evaluating \%x and \&x[] variables, which
is, quite frankly, nuts and makes programming in Kermit at best
counterintuitive.  I made an exception in the case of array subscripts,
because changing how they are evaluated could break a lot of scripts, and
anyway there should never be any harm in evaluating them recursively because
their final value is always (or should be) numeric, not some string that
might contain backslashes.  The SET VAR setting is on the stack, just like
SET QUIET (it follows the quiet/xquiet code in ckuus[356].c), so macros or
command files that change it can't break the script that invokes them.
Added \frecurse() to force recursive evaluation of a \%x or \&x[] variable
regardless of the VARIABLE-EVALUATION setting.  Added \v(vareval) to allow
programmatic setting to current setting.  Tested on Solaris 9 but should be
totally portable.  ckuusr.[ch], ckuus[356].c, 11 Sep 2008.

From Günter Knauf: 64-bit builds were failing on SuSE Linux because
libresolv and libcrypt were in lib64 rather than lib; updated the tests in
the linux makefile target to find them.  makefile, 12 Jan 2009.

Tried building on Red Hat Enterprise Linux Server release 5.3 64-bit.
There is no curses or ncurses.  "make linuxnc" compiled OK but collapsed at
link time looking for crypt(), res_search(), and dn_expand().  Turned out
the linuxnc (and linuxc) targets needed the same treatment as the Linux one
for 64-bit Linuxes.  makefile, 3 Mar 2009.

Consolidated the linux targets so we no longer need three separate ones for
curses, ncurses, and no curses.  "make linux" works ok on computers with and
without (n)curses.  "make linux+ssl", ditto.  "linux+krb5+ssl builds OK but
needs -DNO_KRB5_INIT_ETS".  Makefile, 3 Mar 2009.

Fixed copyright date announced in herald, ckuus5.c, 3 Mar 2009.

Patch from Seth Therault to avoid deprecation warning for utmp references
in ckufio.c in Mac OS X 10.5 (later, this became a consolidated makefile
target that works automatically for at least Mac OS X 10.3.9 through
10.5.6).  makefile, ckufio.c, 28 April 2009.

zshcmd() (the function used by RUN and ! to run external commands) was not
falling back as expected in Linux RHEL4/5 if SHELL was not defined in the
environment.  Also in all Unix versions, there was no indication if a RUN/!
command failed (other than the return code) because the specified shell
didn't exist or was not executable (e.g. the SHELL environment variable was
misdefined).  Now it prints the name of the offending shell and the reason
it couldn't be executed (Not found, Permission denied, etc).  ckufio.c,
28 April 2009.

There is no easy way to get the last field of string; for example, the
extension from a filename, which might have any number of fields.  In
general we want to be able to get "word number n" counting from the right;
\fword() lacks this ability.  Now if you give it a negative word number,
that says to count from the right; for example \fword(one two three four
five, -2) returns "four".  ckclib.c, ckuusr.c, 14 May 2009.

Fixed a typo in the aix51+openssl (SSLLIBS should have been SSLLIB).
From Jason Lehr.  makefile, 27 May 2009.

Updated the linux+openssl+zlib+shadow+pam target to chain to the new main
Linux target.  A bunch of other ones remain un-updated. makefile, 12 Jun 2009.

Updates to the new Mac OS X 10.5 target from Seth Therault (which is
supposed to work on all Mac OS 10-point-anything) to avoid warnings
that came up on on Mac OS 10.4.11/Intel.  Once this one is proven we should
be able to remove/consolidate lots of other ones.  makefile, 12 Jun 2009.

C-Kermit disables SSL with the message "?OpenSSL libraries do not match
required version." if the version of OpenSSL that Kermit was built with is
not exactly the same as the version that is loaded dynamically at runtime.
This is actually the proper behavior, since APIs are not guaranteed not to
change between OpenSSL versions prior to 1.0.0.  Made the error message more
informative.  ck_ssl.c, 26 Aug 2009, and again 28 Aug 2009.

AIX 6.1 is out, it is really just a new name for AIX 5.4.  Added makefile
targets, plus for the first I made AIX 4.2 and later figure out its version
number in the makefile target so we don't have to keep adding new -DAIXnn
sections to the code, and also get its hardware name (e.g. "powerpc") from
uname at make time, rather than hardwiring "rs6000" as I did before.
Consolidated all AIX 4.2 and later targets so now just "make aix" or "make
aix+ssl" can be used.  Except not the gcc ones as they have some quirks so
I'd rather not disturb them.  Tested this on AIX 5.3.
makefile, 28 Aug 2009.

From Kinjal Shah, a correction to the Linux makefile entry that allows it
find the 64-bit curses or ncurses library.  makefile, 29 Aug 2009.

Renamed aix4[23]: to oldaix4[23]: in makefile to fix the warning messages
I didn't notice before.  I didn't want to remove them because they have
some special things that might still be needed, if anybody still has these 
AIX versions.  makefile, 29 Aug 2009.

Built on RHEL 5.3 64-bit, regular and with OpenSSL 0.9.8e.  31 Aug 2009.

Built on NetBSD 5.0.1/i386, regular and with OpenSSL 0.9.9-dev, 1 Sep 2009.

Changed SSL message to mention LD_LIBRARY_PATH (Solaris), SHLIB_PATH (HP-UX),
LIBPATH (AIX), or LD_LIBRARY_PATH (Linux).  ck_ssl.c, 3 Sep 2009

Noticed that "make linux+openssl" fails to include -lutil a link time, which
it needs for openpty().  That's because this target is obsolete.  I renamed
it to be oldlinux+openssl and added linux+openssl as a synonym for
linux+ssl.  makefile, 3 Sep 2009.

Tested linux+openssl+zlib+shadow+pam, it's OK.  Also linux+krb5.  Also
linux+krb5+ssl.  makefile, 3 Sep 2009.

Tried building on Solaris 9 with OpenSSL 0.9.8k with
solaris9g+openssl+shadow+pam+zlib, it failed like so:

  ck_ssl.c:2875: error: conflicting types for 'inet_aton'
  /usr/include/arpa/inet.h:52: previous declaration of 'inet_aton' was here
  make[2]: [ck_ssl.o] Error 1
  make[2]: Leaving directory hmt/sirius1/prv0/kd/fdc/solaris9ssl'
  make[1]: [solaris2xg+openssl+zlib+pam+shadow] Error 2
  make[1]: Leaving directory hmt/sirius1/prv0/kd/fdc/solaris9ssl'
  make: [solaris9g+openssl+shadow+pam+zlib] Error 2

The problem was caused by including an inet_aton() function ck_ssl.c for
the benefit of platforms that don't have one in their libraries.  This is
defeated by including NO_DCL_INET_ATON in KFLAGS.  I added this, but then
I thought it would be a good idea to automatically sense the OpenSSL
version so we can automatically set OPENSSL_097 or OPENSSL_098 rather than
bombing out, so I added code to do that too, and also to set the Solaris
version number: 9, 10, or 11.  The new entry is solaris9g+openssl.
ckcdeb.h, makefile, 3 Sep 2009.

Fixed a complaint in ckufio.c about implicit declaration of initgroups.
ckufio.c, 4 Sep 2009.

Built on Solaris 10 with gcc and Sun CC using new solaris{9,10,11} target
that is like the new solaris{9,10,11}g one but without the gccisms.
makefile, 4 Sep 2009.

Changed solaris{9,10,11}g+ssl target to set only the SSL-specific things and
then chain to the main solaris{9,10,11}g target.  Tested OK on Solaris 9 and
10.  makefile, 4 Sep 2009.

Created solaris{9,10,11}+ssl target that is exactly like the
solaris{9,10,11}g+ssl except it chains to the solaris{9,10,11} target
instead of the solaris{9,10,11}g one.  That is, it builds an SSL version of
C-Kermit using Sun CC rather than gcc.  makefile, 4 Sep 2009.

Tried building on HP-UX 10.20, bundled (non-ANSI) compiler ("make
hpux1000").  This failed until I:

 . Moved a struct inititialization out of setextern(), ckuus3.c.
 . Removed an ANSIism from the declaration of sigchld_handler() in ckutio.c
 . Added a cast to strcmp() in zvuser(), ckufio.c.

Builds OK now.  Built OK with "hpux1000o" (the ANSI compiler) too.
And with "hpux1000gcc".  Couldn't test "hpux1000o+openssl".  21 Sep 2009.

The Sony Playstation 2 and 3 are 64-bit PowerPC platforms that can run Linux
if it is installed as an "other OS" on its hard disk; and the Linux kernel
since 2.6.21 supports the PS3 without any patching required.  Pawel Rogocz
reported that "make linuxppc" (one of the old targets that has not yet been
integrated into the main "linux" target) compiles OK on 2.6.29-ydl61.3
(Yellow Dog Linux release 6.2 'Pyxis'), but fails at link time because
'openpty' isn't found, because -lutil was not included, because that part
was added only to the main linux target.  I asked him to try "make linux"
and he sent back a transcript in which there were thousands of errors from
the curses code ckuusx.c.  Later I tried it myself and it built without a
hitch.  My theory is that between then and now, a missing piece of the
ncurses library (/usr/include/ncursesw) was installed.  21 Sep 2009.

HP-UX 9.05 on PA-RISC 9000/712 building with hpux0900 (bundled compiler):
 . ckutio.c compilation failed with PENDIN and FLUSHO not defined in
   pty_make_raw().  I dummied definitions for them to handle this situation
   on this or any other platform where it might crop up.
   ckutio.c, 24 Sep 2009.
 . Ditto for the PTY module, + IMAXBEL.  ckupty.c, 24 Sep 2009.
 . References to endusershell() were fatal in the bundled compiler.  Changed
   the hpux0900 target to define NODCLENDUSERSHELL, and put a special case
   in ckufio.c to not put a cast in front of the call if NODCLENDUSERSHELL
   is defined.  Now it builds and links OK.  makefile, ckufio.c, 24 Sep 2009.

HP-UX 9.05 on PA-RISC 9000/712 building with hpux0900o (optimizing compiler):
 . Warnings in ckutio.c at line 14860 about arguments to select (pointers
   are not assignment-compatible).  "man select" says arguments are ints.
   Defining INTSELECT fixes these warnings but results in fatal errors later
   around line 14881 and others in the area involving FD_SET.  This was too
   involved so I put it back as it was.  24 Sep 2009.

Built OK on Solaris 10 with Sun CC.  A couple warnings about implicit
function declarations for curses routines because apparently they aren't
declared in curses.h.  Tuff.  25 Sep 2009.

Tried building on Solaris 10 with Sun CC and OpenSSL 0.9.8k, and this
uncovered various loose ends in the solaris9+openssl target, which I fixed.
makefile, 25 Sep 2005.

Fixed four typos in printfs in ck_ssl.c, \% instead of just %.  25 Sep 2009.

Squelched 20-some complaints about a character array being referred to
directly instead of by a pointer, plus several other similar nits to get rid
of all the compilation warnings on Solaris 10 with Sun C 5.8 Patch 121015-06
2007/10/03.  ckctel.c, ckctel.h, 25 Sep 2009.

Built the result on the same Solaris 10 system with gcc 4.2.4 using the
new solari10g+openssl target, working out a few kinks here too.
makefile, 25 Sep 2009.

Made consolidated Solaris 9/10/11 64-bit targets for gcc, solaris9g64,
solaris10g64, solaris11g64, tested on Solaris 10 Sparc. makefile, 25 Sep 2009.

Made consolidated Solaris 9/10/11 64-bit targets for Sun cc: solaris9_64,
solaris10_64, solaris11_64.  These simply set a couple flags and chain to
the main solaris9 target.  makefile, 25 Sep 2009.

Removed a bunch of old superfluous Solaris 9 and 10 targets: oldsolaris9,
oldsolaris9lfs, solaris9g64 solaris9g_64, oldsolaris10 old solaris10lfs,
oldsolaris10+openssl, oldsolaris10g+openssl, solaris10_64, oldsolaris10g,
solaris10g_64, solaris10g64.  There are still plenty more to prune but it's
a start.  makefile, 25 Sep 2009.

Added or fixed some missing prototypes in ckctel.h:
fwdx_send_xauth_to_xserver(), fwdx_parse_displayname.  25 Sep 2009.

Improved the instructions for building secure versions in the makefile,
using this example:

  make solaris9+openssl "SSLINC=-I/opt/openssl-0.9.8k/include" \
   "SSLLIB=-L/opt/openssl-0.9.8k/lib"

makefile, http://kermit.columbia.edu/security.html, 25 Sep 2009.

Built on HP-UX 11.11, 26 Sep 2009:
 . make hpux1100 (ok)
 . make hpux1100gcc (ok)
 . make hpux1100o (gets a lot of warnings about sendpath and sendfile,
    because they are also declared in <sys/socket.h>, but builds OK)
 . make hpux1000gcc+openssl \
    SSLINC=-I/opt/openssl/include SSLLIB=-L/opt/openssl/lib

Note: sendpath and sendfile are not Kermit symbols.  The warnings are coming
from socket.h: 'Redeclaration of "sendfile" with a different storage class
specifier'.  This is nothing new; see notes of 2-4 Jan 2005.

From Peter Eichhorn:
 . Update to makefile to make current code build OK on HP-UX 8.00.
 . Changes to format of some hints to make them more copy-and-pastable.
makefile, ckuu5.c, 28 Sep 2009.

From Peter Eichhorn: Changes to HP-UX 7.0 target to increase the switch table
stack size, which was overflowing.  makefile, 30 Sep 2009

HP-UX 6.5 (1989), "make hpux0650tcpc"... (8:19...)  Needed to not include
arpa/inet.h (which doesn't exist) and not use host address lists (add
-DNOHADDRLIST), which gets us past ckcnet.c, but in ckcftp.c we bomb out on
FD_SETSIZE undefined.  Somehow we worked around this in ckcnet.c.  Patched
in a definition in ckcftp.c, and also added -DINTSELECT to compiler flags.
Compiles ok, bombs at link time on bcopy, bzero, FD_ZERO, FD_SET, FD_ISSET.
Now it compiles and links OK but dumps core when started.  Added
-DNOCKGETFQHOST, rebuilt from scratch (takes 35 minutes).  It starts OK, but
it dumps core when given a "telnet xxx" command, where xxx is a hostname.
However, it works OK if an IP address is used: "telnet 123.45.6.78".  It
took all day to track this down, but now it's fixed (see the #ifdef HPUX6
sections of ckcnet.c).  So now (for the first time, I think) we have both
telnet and ftp in HP-UX 6.x, if anyone cares.  ckcnet.[ch], ckcftp.c,
makefile, 2 Oct 2009.

Changed default SET TERMINAL TYPE type for K95 from vt320 to vt220.  This is
because Unix OS's such as Solaris have dropped vt320 as a terminal type.
settrmtyp(), ckuus7.c, 5 Oct 2009.

I moved the PUTENV command code, which was inline, to a function, doputenv().
ckuus[r7].c, ckuusr.h, 5 Oct 2009.

Changed the UNIX version of SET TERMINAL TYPE to take a value and then do
the equivalent of "export TERM=value" by calling doputenv().  This sets
\$(TERM) correctly and passes its value along to inferior processes.
However, to make this take effect within Kermit itself (for the fullscreen
file transfer display and for the SCREEN command, Ctrl-L, etc) I also had to
reinitialize the curses database, which is tricky because normally if you
feed it an unknown terminal name, it just exits.  ckuus7.c, 5 Oct 2009.

Changed the little-known and little-used RESET command (which closes all
open files) to also put command echoing back to normal in case it got
messed up somehow (as in HP-UX 6.5, upon returning from PUSH).
ckuusx.c, 5 Oct 2009.

For Unix, increased string buffer sizes for wildcard expansion for all
platforms that have BIGBUFOK defined from 500000 (0.5M) to 10000000 (10M)
bytes, and for 64-bit builds to 2000000000 (2G) bytes.  No point making
it bigger than that because malloc's argument is a size_t, which is an int.
ckufio.c, 5 Oct 2009.

Built on Mac OS X 10.4.11, required one minor adjustment to the makefile
(-DNODCLINITGROUPS).  This was using the macosx10.5 target, which is
supposed to be universal like the linux and netbsd targets, but not yet
proven.  Also built a 64-bit version (-mpowerpc64 -mcpu=G5 -mtune=G5
-arch ppc64); it compiles and links OK but won't start: "Bad CPU Type
in executable".  Fix later...  makefile, 5 Oct 2009.

Changes from Seth Theriault to suppress signed vs unsigned char warnings in
Mac OS 10.5.8 from gcc4, and a new makefile target for Mac OS X (presumably
10.3.9 or later) + Kerberos 5 and OpenSSL.  ckutio.c, ckuath.c, ckctel.c,
ckcnet.c, ckcftp.c, ck_crp.c, makefile, 6 Oct 2009.

  Later I had to back out of these, because although it made for a
  clean build, in the resulting executable SSL connections didn't work.

Tue Oct  6 17:23:27 2009
FTP address resolution is broken, but ftp_hookup() hasn't changed.
So... (see the #ifdef HPUX6 sections of ckcnet.c)  (I did, and I rolled
back some of the changes from the other day, but it made no difference.)
Putting back the ckcftp.c from a few weeks ago makes no difference.
Putting back the ckcnet.c from a few weeks ago makes no difference.

Added patches from Seth Theriault so macosx10.5+krb5+openssl would build
on Mac OS X 10.3.9.  makefile, ckcftp.c, 7 Oct 2009.

Built today's code on Linux RHEL4, NetBSD 5.0.1, Solaris 9, and Mac OS X
10.4.11, both with and without SSL.  The NetBSD system has OpenSSL 0.9.9-dev.
7 Oct 2009.

In Mac OS X 10.6, the following symbols are unresolved at link time:
_des_key_sched, _des_new_random_key, _des_ecb_encrypt,
_des_init_random_number_generator, _des_fixup_key_parity.  This is
with OpenSSL 0.9.8k.  But it doesn't happen on other platforms that
have 0.9.8k.

Added SET SESSION-LOG NULL-TERMINATED-TEXT.  This is for the benefit of a
speech synthesizer that will speak a line of text only after receiving a
NUL character.  A more general solution would be to define a filter or
whatever, but who has time.  ckuus[23x].c, 7 Oct 2009.

Consolidated Mac OS X targets, and removed experimental 64-bit ones, because
they never could work in 10.5 and earlier because 64-bit libs are missing,
and 10.6 and later are 64-bit automatically.  makefile, 8 Oct 2009.

Built on Mac OS X 10.6.1.  It came out automatically as a 64-bit build
because __LP64__ is defined somewhere that I can't find.  But this explains
why the 0.9.8k on 10.6 comes up with missing symbols when the 0.9.8k lib
10.5 (or on Solaris or on Linux) does not: it's a different library: "Mach-O
64-bit dynamically linked shared library x86_64", rather than "Mach-O
dynamically linked shared library ppc".  Probably the 64-bit version has
some things #ifdef'd out.  Added -m32 to the CFLAGS and LNKFLAGS for the
macosx+krb5+openssl targets, and it built OK one time.  But then the errors
came back.  makefile, 8 Oct 2009.

Updated C-Kermit installation for Mac OS X in ckuwr.html on the website.
8 Oct 2009.

Tried some things to get around the problem with OpenSSL in Mac OS X 10.6,
to no avail.  Asked Jeff.  He said, "MacOS X no longer includes DES anywhere
on the system.  Not for SSL, not for Kerberos, not for anything.  This will
increasingly become the situation on new operating systems.  Windows 7 and
2008 R2 will also ship with no DES."  Sure enough, the Mac OS X Server
Upgrading and Migrating document for 10.6 says, "Mac OS X Server v10.6 does
not support single DES encryption. It supports AES 128 and 256 encryption
types. However, during a migration or upgrade from v10.4 to v10.6, servers
that were Kerberized by the v10.5 Open Directory server will not use the AES
128 or 256 encryption types. To use the AES 128 or 256 encryption types you
must re-Kerberize all servers."  12 Oct 2009.

DES and 3DES encryption can be excluding removing the -DCK_DES flag.  I
removed this one and -DLIBDES (and -m32) and this makes a working 64-bit
version.  Then I added code to the macosx+krb5+openssl target to use these
flags if the Mac OS X version was 10.5 or less and leave them out for 10.6
or later.  Tested on 10.4.11 and 10.6.1.  A better way to do it might have
been "nm -gj libssl.dylib | grep des_", but that gives the same results on
10.4 and 10.6.  Also, 10.6 still has /usr/include/ssl/des.h.
makefile, 13 Oct 2009.

Next issue:
  In file included from ckutio.c:15674:
  /usr/lib/gcc/i386-redhat-linux/3.4.6/include/varargs.h:4:2: #error "GCC no
  longer implements <varargs.h>."
  /usr/lib/gcc/i386-redhat-linux/3.4.6/include/varargs.h:5:2: #error "Revise
  your code to use <stdarg.h>."

The problem occurs when trying to force a non-ANSIC build with GCC.
Changing the source file to include <stdarg.h> instead of <varargs.h>
doesn't help because evidently <stdarg.h> requires an ANSI C compiler.
Nothing can be done about this.  13 Oct 2009.

Next issue: Can't compile ckcftp.c with -DNOCSETS or -DNOSPL; some
#ifdef/#endif doesn't match up.  Sigh, this is the hardest kind of thing to
debug.  There's 17,622 lines of code in this module and no tool that I know
of.... Wait, I wrote one.  But it shows all the #if/#ifdef/#ifndef's and
#endifs matching up just fine.  Backing off to ckcftp.c of a few days ago
(before char / unsigned char casts were added), I see that it builds OK, so
I backed off to that one, but put back the special case #ifdef for MACOSX103
declaring CONST gss_OID_desc, and it builds OK (the other stuff was purely
cosmetic, when will I learn?).  ckcftp.c, 13 Oct 2009.

Protected cvtstring() and related functions with #ifdef NOCSETS..#endif,
and ditto for the character-set conversion code in dorename().
ckuus6.c, 13 Oct 2009.

Fixed an #endif /* TNCODE */ that was a line too low in ttptycmd(),
causing -DNONET builds to fail.  ckutio.c, 13 Oct 2009.

There was a reference to doputenv() that wasn't guarded by #ifndef NOPUTENV,
fixed in ckuus7.c, 13 Oct 2009.

Moved doputenv() and settermtyp() out of an #ifdef NOLOCAL section because
these are useful even when not making connections.  ckuus7.c, 13 Oct 2009.

Moved havelfs declaration outside of #ifdef NOXFER because it was also used
for other things.  ckcmai.c, 13 Oct 2009.

COPY /PRESERVE depended on code from the Kermit protocol module, which
is omitted in -DNOXFER builds.  Disabled COPY /PRESERVE in -DNOXFER
builds.  ckuus6.c, 14 Oct 2009.

SHOW PROTOCOL code for external protocols had to be #ifdef'd out for
-DNOPUSH builds.  ckuus4.c, 14 Oct 2009.

There was some confusion between "No XYZMODEM" and "No extermal protocols";
cleared up in ckuus3.c, 14 Oct 2009.

After all that, 86 different combinations of feature selections built OK on
Linux.  And the Kerberized version (K5) works OK on Linux for Telnet and FTP.
14 Oct 2009.

Changed version number to 9.0.  All modules, 16 Oct 2009.

Need to make LOG SESSION log to a tty.  Right now "log session
/dev/ttyKeySerial1" says "Write permission denied" even though the device is
crw-rw-rw-.  This happens in zchko(), which is called by cmofi().  The
problem is that /dev/ is not writeable.  I added a Unix-only clause that
attempts to open the file for write access using open(), in order to get a
file descriptor, which then can be passed to isatty() to check if it's a
tty, and if so, to allow access.  And then close it.  I tested this on Mac
OS X as follows:

  log session /dev/ttyKeySerial1
  telnet somehost

The Mac's serial port was connected to the serial port of another computer
where Kermit displayed the incoming characters in CONNECT mode.  Glitches:

 1. The port has to be set up as desired in advance, outside of Kermit.
 2. log session /dev/ttyKeySerial1 will hang if any required modem signals
    are not present when the port is opened.
 3. Bypasses lockfile mechanism - so we do this only if -DNOUUCP.

For (2), I tried setting O_NDELAY / O_NONBLOCK, and this allowed zchko() to
continue, but then it freezes in the subsequent fopen().  So I changed
zopeno() to also check if the device is a serial port, and if so, to open()
it with O_NDELAY / O_NONBLOCK, and then convert the file descriptor into a
file pointer with fdopen().

Now for the speaking device that needs lines to be terminated by NUL...

  set session-log binary       <-- need to put these in SHOW LOG
  set session-log null-padded      (and in HELP SET LOG)
  set line /dev/ttyKeySerial1

This part works.

This feature is enabled only for -DNOUUCP builds because serial ports aren't
like other Unix files; we would have to create a lockfile, but we can't do
that...  actually, ttlock() takes a name as an argument, but ttunlck() does
not, so there would be no way to remove the lock.  Anyway, there is only one
API for configuring the port (speed, flow control, etc) and it only works
with the SET LINE device, not any random file.  To fix this would require
massive redesign and changes.  ckuus[23].c, ckufio.c, 19-20 Oct 2009.

I made -DNOUUCP the default for Mac OS X, since everybody winds up building
it that way anyhow.  To undo this, do "make macosx KFLAGS=-UNOUUCP".
makefile, 21 Oct 2009.

Changed SET SESSION-LOG TEXT to strip out ANSI escape sequences; 
previously there wasn't that much difference between TEXT and BINARY logs.
It's still not perfect; for example it doesn't delete characters that the
user erased.  (Made sure this still builds with -DNOESCSEQ.)
ckucns.c, 22 Oct 2009.

Changed SHOW LOG to show the SET SESSION-LOG settings, as well as
SET DEBUG, which was not shown before.  ckuus5.c, 22 Oct 2009.

If a series of PUTENV commands is given, each new one undoes the previous
one, so only the last definition is seen by the new fork (or by Kermit
itself).  Turns out you can't feed automatic variables to putenv(); they
have to be static, so to allow for multiple PUTENV commands Kermit has to
maintain an array of static strings.  ckuus7.c, 6 Nov 2009.

From Seth Theriault, a better way for the makefile to determine the
Mac OS X version number; there's a program for this, sw_ver.  makefile,
6 Nov 2009.

Peter Eichhorn reported that file-transfer failure hints were not coming
out since Dev.27.  The only change I made since then was to skip them if
the file-transfer protocol was not Kermit.  I was using the wrong variable
in the tests, 'proto' instead of 'protocol'.  ckuus5.c, 6 Nov 2009.

Changed Mac OS X targets to correctly extract the Mac OS major version
from uname -r in order to choose correctly between utmp and utmpx; this
wasn't working in 10.6.1.  makefile, 6 Nov 2009.

Fix from Seth T. for an oversight in the previous edit.  Also add
MACOSX103 to "show features" display.  makefile, ckuus5.c, 10 Nov 2009.

Added REJECT as a synonym for DISCARD in SET FILE COLLISION; it's more
intuitive and more accurate.  ckuus[27].c, 15 Nov 2009.

\fsplit() and \fword() always break on 8-bit characters unless you explicitly
put every single 8-bit value into the include set, e.g. (for a TSV file):

  undef include
  for \%i 128 255 1 {
      if == \%i 9 continue
      .include := \m(include)\fchar(\%i)
  }
  .\%n := \fsplit(\m(line),&a,\9,\m(include))

I changed cksplit() to treat all 8-bit bytes 128-255 as non-break characters
by default.  It might have made more sense to do this for 160-255 (since
128-159 are traditionaly C1 control characters) but thanks to Microsoft
tradition is out the window.  To treat one or more 8-bit characters as break
characters, put them in the break set.  This might break some scripts, but I
doubt it because this flaw was so awful that if anyone had come up against
they would have let me know.  ckclib.c, 16 Nov 2009.

Changed the netbsd target to set -funsigned-char, since cc on NetBSD is
actually gcc.  makefile, 16 Nov 2009.

Changed macosx targets to get the CPU type from the HOSTTYPE environment
variable.  Also added getenv("HOSTTYPE") as a last-resort method to set the
\v(cpu) variable at runtime (maybe it should be the first resort?)...
ckuus4.c, makefile, 16 Nov 2009.

Made sure the solaris9_64 and solaris10 targets still work.  16 Nov 2009.

Made sure the current source package builds OK on HP-UX 10.20...  Got a lot
of "warning 6062: Optdriver: Exceeding compiler resource limits in xxx; some
optimizations skipped. Use +Onolimit if override desired" but it builds OK.
Tested long file transfer; works OK.  17 Nov 2009.

Built on FreeBSD 7.2 with and without OpenSSL, all OK.  17 Nov 2009.

Built on NetBSD 5.0.1 with and without OpenSSL, all OK, but netbsd+krb5
fails with "can't find -lgssapi_krb5"; worked around this with
"K5LIB=-L/usr/local/kerblib" (where the lib actually is on this host) but
then it failed with "ckcftp.c:13868: error: 'gss_nt_service_name' undeclared".
17 Nov 2009.

I found a VMS 6.2 system... Takes a loooong time to build there.  In
ckuusy.c, DEC C didn't like the prototypes and declarations of dorlgarg()
and dotnarg() as static so I made them not static.  But that didn't help,
now it fails at the very end, saying the final #ifdef is an invalid
statement.  It looks like an #ifdef mismatch that affects only VMS.  I ran
my #ifdef matcher, it turned up nothing.  I substituted a copy of ckuusy.c
from 2007, it comes up with the same errors.  Then I substituted the copy
from 8.0.211 from 2004, and this one compiled OK and, miraculously, the
whole mess even linked OK and runs OK.  The Alpha binary is 2.84MB.  Now I
have 4500 lines of code to compare....  I went through the two files line by
line and I can't see a single thing wrong.  I gave up and tried building the
TCP/IP version.  It builds fine except for ckuusy.c, with the utterly
useless error message:

  #endif /* NOCMDL */
  ...................^
  %CC-E-BADSTMT, Invalid statement.

Indicating the last line in the file.  Just for the heck of it, I put
another line after that one:

  /* This is a test */

and got:

  /* This is a test */
  ....................^
  %CC-E-BADSTMT, Invalid statement.

So it is not objecting to anything in the file.  Trying the old LISP trick,
I put an extraneous closing bracket after that.  Success!  Honestly, I don't
see anything wrong with file.  It's DEC C V5.3-006.  I suspect a C bug.
I'll leave it like this for now until I get access to some other VMS
versions.  Another clue is that when building the network version I get a
horrible warning I never saw before from a module that hasn't been touched
in a very long time (ckvrtl.c).  Also, in the network version, I note that
the FTP code is not compiled in.  We have to try this again with some
command-line switches, but it'll do for now.  ckuusy.c, 18 Nov 2009.

---C-Kermit 9.0 Alpha.01---

From Steven Schweda (SMS), the real solution for the VMS closing brace
problem, it wasn't a DECC bug, it was a me bug.  ckuusy.c, 20 Nov 2009.

Rediscovered the new VMS build options: f for Long Files, i for Internal
FTP.  "make mnf" doesn't work on VMS 6.2, it looks like the VMS definition
for CK_OFF_T got lost.  Same thing with "make mfi".  Come back to this later.

From Gerry Belanger, a fix to INPUT /COUNT:n.  ckuus4.c, 26 Nov 2009.

Added \fsqueeze(s), returns string s with leading and trailing whitespace
removed, Tabs converted to Spaces, and multiple spaces converted to single
spaces.  For now, ASCII only, no options.  ckuusr.h, ckuus[24].c, 27 Nov 2009.

I wrote a Kermit script to read a big file of addresses on Solaris 9,
\fsqueeze()ing each line.  After about 14000 lines, there was a malloc
failure in getnct() (the command-file reader).  There's nothing wrong with
\fsqueeze(), the failure is on a deeper level, because the same thing
happens if I use \fupper() (which is structurally identical to \fsqueeze())
in the same script.  The problem is not in getnct() either, because every
malloc() is freed (I checked).  On the other hand, the same script (with
\fupper() instead of \fsqueeze() completes OK in C-Kermit 8.0.201.  If I
remove the function call (\fsqueeze() or \fupper()) from the script, it also
runs OK in 9.0.  This seems to point the finger at fnevel(), which contains
countless malloc's and free's.  But comparing fneval() between 8.0.211 and
9.0, I don't see any difference that would explain this behavior -- nothing
at all that involves malloc(), makstr(), or free().  Nor any pertinent
change in the caller (zzstring) of fneval().  27 Nov 3009.

Another problem is that when this happens, the error is not caught (e.g. by
the IF FAIL statement after the command that contains the function call);
instead, C-Kermit returns immediately to its prompt.  27 Nov 2009.

It could simply be that some of the buffers we allocate are much bigger now.
But again, I don't see much difference between 8.0.211 and 9.0; we were
already allocating 32K command-related buffers (malloc() takes a size_t, and
size_t is an int almost everywere).  I built the same source on NetBSD and
ran the same script (with \fqueeze()), and it worked fine.  Let's worry
about this later, if it comes up.  27 Nov 2009.

Built OK on Silicon Graphics IRIX 6.5 R10000; regular build OK, SSL and
Kerberos builds failed.  30 Nov 3009.

Tried to build on Digital Unix 4.0F but it blew up in ckutio.c, apparently
not recognizing any of the terminal struct symbols from termios.h.  Tried
again with gcc, same thing.  Tried explicitly #including <sys/termios.h>
within #ifdef TRU64, same thing.  What could have changed?  30 Nov 2009.

Built OK on Linux RHEL5.4/Itanium-2, make linux.  The secure build
required "FLAGS=-DNO_KRB5_INIT_ETS" and built OK.  30 Nov 2009.

Built OK on Digital Unix 4.0F using "make osf" instead of "make tru64-40f".
I don't know why the specific target doesn't work, but it's not worth
chasing down.  2 Dec 2009.

Built OK on MirBSD 10, despite a lot of gratuitous compiler warnings.  Built
OK on MirBSD 10, OpenBSD 4.5, and Fedora 10.  3 Dec 2009.

(Various other successful Unix builds in these weeks...)

Built on VMS 7.2 and 8.3 with and without TCP/IP, no problems.  11 Jan 2010.

Built on VMS 8.3 with "make fi" to include the FTP client and long-file
support (mid Jan 2010).

Built on VMS 8.3 with UXC 5.6 and HP SSL 1.3, which is OpenSSL 0.9.7e.
It compiled and linked OK but when I tried to make an FTP SSL connection
it crashed in SSL$LIBSSL_SHR, which is called from ssl_auth(), after having
had TLS accepted as an authentication type, but before actually
authenticating.  In Unix:

 19. ftp open ftp.somecompany.com /user:pge.com/test_quota /password:xxxxxx
Connected to ftp.somecompany.com.
220-Somecompany FTP v6.0 for WinSock ready...
220 Welcome to the online storage FTP server.  Please check the main web
site for system announcements and AUP. (O)
---> AUTH TLS
234 AUTH command OK. Initializing SSL connection.
TLS accepted as authentication type
SSL DEBUG ACTIVE
=>START SSL/TLS connect on COMMAND

In VMS:

 19. ftp open ftp.somecompany.com /user:pge.com/test_quota /password:xxxxxx
Connected to ftp.somecompany.com.
220 Somecompany FTP v6.0 for WinSock ready...
---> AUTH TLS
234 AUTH command OK. Initializing SSL connection.
TLS accepted as authentication type
SSL DEBUG ACTIVE
%SYSTEM-F-ACCVIO, access violation, reason mask=04, virtual
address=FFFFFFFF8001A120, PC=000000000068B118, PS=0000001B

Note: The Unix version received the second 220 response, the VMS version did
not.  That's odd, it's the same code...  25 Jan 2010.

Added some essential details to the HELP FSEEK text.  ckuus2.c, 25 Jan 2010.

Discovered that the result returned by \fsearch() is totally unreliable.
This is probably too hard to fix.

FSEEK did not pay attention to SET CASE, searches were always case sensitive.
Fixed in ckuus7.c, 26 Jan 2010.

FSEEK failed to find anything if the search pattern was matched in the first
line of the file.  Fixed in ckuus7.c, 26 Jan 2010.

\fword() and \fsplit()....  Another change, but not backwards-incompatible.
One may now put the word ALL (just like that, all uppercase) as the include
set (4th argument) to indicate that there will be no break characters other
than those explicitly given in the break set, e.g. \fsplit(\m(xx),&a,:,ALL)
breaks a line only on a colon (:), nothing else.  The original rules for
cksplit() were more than a little counterintuitive: the default break set is
all non alphanums, and the default include set is all alphanums, so if you
wanted to parse (say) a CSV file, breaking only on comma, you had to think
of all the characters you wanted to keep.  This way you just say ALL.
ckclib.c, 26 Jan 2010.

Speaking of CSV files... How can you put comma as a function argument when
comma is the function-argument separator?  Use one of these forms:

 \fsplit(\m(xx),&a,",",ALL)
 \fsplit(\m(xx),&a,{,},ALL)
 \fsplit(\m(xx),&a,\44,ALL)
 \fsplit(\m(xx),&a,\fchar(44),ALL)

From John Dunlap, U. of Washington Applied Physics Lab: 'When "stty -a <
/dev/ttyS0 | grep crtscts" shows "crtscts" (not "-crtscts") and when using a
three wire serial interface and when asking kermit to not use flow control
(set flow none) then "ckutio.c1" (see attachments) fails while "ckutio.c"
works.  The result of "diff -u ckutio.c1 ckutio.c" is attached as "diffs"'.
ckutio.c, 26 Jan 2010.

Changed the year from 2009 to 2010 in the modules I worked on today and in
the heralds, etc.  ckckmai.c, ckuus5.c, ckutio.c, ckclib.c, ckuus7.c,
26 Jan 2010.

Built on Linux Fedora Core 3, regular and with OpenSSL 0.9.7a.  Built on
Ubuntu 9.4 OK, but SSL and Kerberos builds failed due to not finding libs
and/or header files.  I'm sure this could be fixed...  27 Jan 2010.

Added SSL, KRB4, and KRB5 to the startup herald for versions that were
built with SSL, Kerberos 4, or Kerberos 5.  Built OK on Fedora 3 with
linux+krb5+ssl and new banner shows correctly.  ckuus5.c, 27 Jan 2010.

Set NO_KRB5_INIT_ETS by default in ckuath.h since krb5_init_ets() is a no-op
in Kerberos 1.4.x and later and in some installations it can't be found,
which clobbers the build.  ckuath.h, 27 Jan 2010.

Adapted to MINIX 3 1.5, the first version that has virtual memory according
to Andy T, who should know.  On earlier versions (e.g. MINIX 3 1.2) any
attempt to build C-Kermit causes the compiler to crash.  Now the compiler
doesn't crash but it spews out countless warnings about old-fashioned
function declarations that I don't get anywhere else.  The real problems
came in ckutio.c where numerous symbols were undefined at compile time and
the POSIX function tcgetpgrp() was not found at link time, even though there
is a prototype for it in the MINIX header files, and there is no alternative
(since POSIX doesn't let us use ioctl()).  Also note that there is some
confusion over the compile-time symbols MINIX, MINIX2, MINIX3, and MINIX315.
You would expect MINIX to mean "any version of MINIX" but in some parts of
ckutio.c it means MINIX 1.0.  I sincerely doubt that C-Kermit 9.0 can be
built on any version of Minix before 3.1.5 so I removed the confusion and
made MINIX mean "any Minix".  It builds on 3.1.5 OK now, except for the FTP
client.  This can probably be fixed but...  Modules changed: ckcdeb.h,
ckuver.h, ckcmai.c, ckuus5.c, ckutio.c, 1 Feb 2010.

Later.. Andy says MINIX does not support job control, so no program is ever
in the background.  That settles that!  1 Feb 2010.

Built OK on Minix, Linux, Mac OS X, Solaris 9, NetBSD 5.0.1...  1 Feb 2010.

---C-Kermit 9.0 Alpha.02---

From Christian Corti at Uni-Stuttgart.de: fixes to allow building on SunOS
4.1, which once was my main development platform but which is long-gone from
here.  ckupty.c, ckutio.c, 9 Feb 2010.  (He says it is also necessary to
comment out the "struct winsize" and "struct ttysize" in sys/ioctl.h;
otherwise there will be a conflict with sys/ttycom.h (included by termios.h)
which also declares these structs. But you need both includes.')

From John Dunlap, a fix for Kermit protocol fixed packet-timeout interval
going to a unexpected value (missing else clause in two places).
ckcfn2.c, 9 Feb 2010.

Added an aixg target to build on AIX with gcc when gcc is not installed as
cc, and also added CC=$(CC) CC2=$(CC) clauses to the aix and aix+ssl
targets.  Wow, AIX really loses bigtime when receiving files through its ssh
server.  Streaming can't be used, sliding windows recover from errors but
there are tons of them using the default 4K packets; 500 works much better.
Built with IBM cc and gcc, and also tested (successfully) the new aix+ibmssl
target, in which the OpenSSL headers and libs are in a standard place.
makefile, 9 Feb 2010.

In ckupty.h, make the #include <sys/ioctl.h> be #ifndef SUNOS41.
From Christian Corti.  10 Feb 2010.

Built on VMS E8.4.  12 Feb 2010.

Tried to build on a real VAX-11/785 but the machine seems to be seriously
wedged.  12-15 Feb 2010.

Added note to CKVKER.COM to the effect the the 'f' option has no effect
on VAX architecture.  15 Feb 2010.

Moved the #include "ckvrtl.h" in the FTP module to below the include for
utime.h, because building the VMS version with the 'i' option (meaning
"include internal ftp client") results in "struct utimbuf tp" erroring out
because struct utimbuf is not defined yet (at least in some version of VMS
with some version of C).  From Rob Brown, ckcftp.c, 20 Feb 2010.

From Martin Vorlaender: new code in VMS C-Kermit build procedure to detect
OpenSSL version automatically.  ckvker.com, 22 Feb 2010.

Added code to INPUT command to strip ANSI escape sequences.  It's activated
by SET SESSION-LOG TEXT.  ckuusr.h: added prototype for chkaes();
ckucon.c, ckucns.c: made inesc[] and oldesc[] global instead of static; 
ckuus4.c: doinput() code for skipping escape sequences.  1 Mar 2010.

Peter Eichhorn complained that if you make an ssh connection with Kermit,
then log out from the ssh host, and then use a "connect" command to
make a new connection to the same host (which you can do with Telnet),
Kermit says (e.g.):

 DNS Lookup... Can't get address for ssh -e none somehostname
 Sorry, can't open ssh -e none somehostname: Error 0

I added code to detect and handle this case and it seems to work OK, even
though it's kind of a hack.  ckuusr.[ch], ckuus7.c, 1 Mar 2010.

There has never been a clean way to put debugging messages (ECHO commands)
in a script which are executed only if debugging is desired and ignored
otherwise.  You'd have to set a random variable and test it, or define a
macro or whatever.  To make this more straightforward, I added SET DEBUG
MESSAGE ON/OFF/STDERR, and added a new MESSAGE (syn: MSG) command for printing
debugging messages to stdout if SET DEBUG MESSAGE is ON or to stderr if SET
DEBUG MESSAGE is STDERR.  ckcmai.c, ckuus[r23].c, 12 Mar 2010.

Also for debugging and error messages, I added \v(lastcommmand) so that
the command that failed can be included in an IF FAIL or DEBUG error message.
This works even for commands that have syntax errors.
ckuusr.h, ckuus5.c, ckucmd.c, 12 Mar 2010.

From SMS for VMS: 'Added/documented P3 options INTSELECT, OLDFIB, OLDIP.
Disabled (commented out) automatic definition of NOSETTIME for VMS before
V7.2 (vms_ver .lts. "VMS_V72").'  ckcdeb.h, ckcftp.c, ckcnet.c, ckuus[2567].c,
ckvfio.c, ckvker.com, ckvrtl.[ch], 15 Mar 2010.

Exposed inesc[] and oldesc[] for VMS, so new INPUT command escape-sequence
stripping can work (really, chkaes() and related global variables should be
moved out of ck[uvd]con.c/ckucns.c and into a common module; do that later).
ckuusr.h, ckvcon.c, 15 Mar 2010.

Built OK on Solaris9, Mac OS X 10.4.11, RHEL4 (32-bit), RHEL5 (64-bit),
AIX 5.3, SCO OpenServr 6.0.0...  15 Mar 2010.

Not so good on VMS, turns out I made a typo in one of the VMS updates
(#ifndef OLDIP instead of #ifdef...).  ckcnet.c, 16 Mar 2010.

More from SMS for VMS, 16 Mar 2010:
 . Set MAXPATH correctly for VMS, ckcdeb.h.
 . NAM -> NAML, QIO replaces system( "SET PROTECTION"), bugfixes in
   cvtdir() and nzltor(), ...  (See comments): ckvfio.c, new ckvrms.h.
   (The RMS code in ckvfio.c was almost totally rewritten)
 . Moved "NAMX$*" (and related) macros to ckvrms.h, and renamed to
   "NAMX_*" (and similar "$" -> "_"), moved "FIB_*" macros from ckvrtl.c.

These changes are mainly to accommodate the ODS5 file system, which has
longer and mixed-case filenames, and also to execute certain commands
(e.g. for setting file protection, deleting directories) directly instead
of using a system() command.

Built OK on VMS 8.3 (with and without network support).  16 Mar 2010.

Failed to build on VMS 6.2.  16 Mar 2010.

FreeBSD 8.0 <libutil.h> has a hexdump() prototype that conflicts with the
hexdump macro defined in ckcdeb.h.  Since the same thing is likely to happen
elsewhere, I changed the Kermit macro to ckhexdump as well all references to
it: ckcdeb.h, ckcftp.c, ckcnet.c, ckctel.c, ckuath.c, ckutio.c, 16 Mar 2010.

Built OK on Digital Unix Tru-64 4.0E using "make osf", 16 Mar 2010.

Tried again to build Digital Unix Tru64 4.0E using "make tru64-40e", but
something prevents it from picking up the termios symbols and it blows up in
ckutio.c, whereas this used to work in earlier C-Kermit versions.  This is
the only Tru64 system I still have access to, so I can't tell if it's a
local peculiarity or what.  Note that POSIX is not defined for this build.
But if I define it, I get into trouble with "struct timeval".  Tried again
with "KFLAGS=-DPOSIX -DNOTIMEVAL" but that doesn't help.  Tried "make
dec-osf" and that worked OK but oddly enough it makes a Kermit with less
features than "make osf".  16 Mar 2010.

To go with MESSAGE and SET DEBUG MESSAGE, I added IF DEBUG, which is true
if SET DEBUG MESSAGE is not OFF and false otherwise.  ckuusr.h, ckuus6.c,
16 Mar 2010.

From SMS: Corrections to my merging of SMS's changes, ckcftp.c, ckvrtl.h.
Builds OK on VMS 6.2 now.  Also did an SSL build on VMS 8.3 with OpenSSL
m0.9.7e and "OPENSSL_DISABLE_OLD_DES_SUPPORT" was included in P3
automatically by Martin V's addition to ckvker.com.  17 Mar 2010.

From SMS: #include <types.h> earlier for VMS in ckcdeb.h to pick up off_t
before it is referenced.  This allows C-Kermit to compile on VMS/Alpha 6.2
but linking fails on fseeko() and ftello() (and yet, a functional executable
is created, and FSEEK works right).  Builds the same way with no problems at
all on VMS 8.3 / Alpha.  In this case we get the full 64-bit arithmetic...
Well, 62 bits:

  ATLAS::C-Kermit>( ^ 2 63)
   9223372036854775000.0
  ATLAS::C-Kermit>( ^ 2 62)
   4611686018427387904

whereas on VMS 6.2 we get integers only up to (^ 2 30).  17 Mar 2010.

Changed the VMS build procedure to enable large file support automatically
for non-VAX and VMS 7.3 or greater.  No reason not to include this feature.
Changed the sense of the F option to DISABLE large file support in the
unlikely case that C-Kermit is being built on a suitable platform but the
C library is older than VMS73_ACRTL-V0200, in which case fseeko() and
ftello() will come up missing at link time.  ckvker.com, 18 Mar 2010.

Changed VMS build procedure to include the FTP client in any network build
by default.  Changed the sense of the I option to exclude the FTP client,
in case anybody would want to do that.  ckvker.com, 18 Mar 2010.

From SMS: updated dependencies in CKVKER.COM, fix the "don't reinclude me"
clause in CKVRTL.H.  19 Mar 2010.

Built OK on VMS 6.2 and 8.3 with and without networking.  Large file support
included automatically in VMS 8.3  FTP client included automatically in both
network builds.  19 Mar 2010.

Changed hexdump() to ckhexdump() in ck_crp.c, which I missed before.
19 Mar 2010.

---C-Kermit 9.0 Alpha.03---

In HP-UX with the bundled-non ANSI compiler, we get warnings about functions
such as endusershell(), which are declared void in the header files.  But in
non-ANSI builds we defind VOID to be int rather than void, so our prototypes
are wrong.  I checked that HP-UX 9, 10, and 11 all have void datatype and
changed the definition of VOID to void in those cases.  ckcdeb.h, 29 Mar 2010.

Fixed a typo in a debug() statement in cksplit() that was causing some
warnings.  ckclib.c, 29 Mar 2010.

Ditto in tls_load_certs().  ck_ssl.c, 29 Mar 2010.

"make hpux1000o+ssl" files with:
/usr/ccs/bin/ld: Unsatisfied symbols:
   __umoddi3 (code)
   __udivdi3 (code)
   __eprintf (code)

It appears that OpenSSL (0.9.7c in this case) requires -lgcc.
And indeed hpux1000gcc+ssl builds fine.  29 Mar 2010.

There are various warnings in the SSL code in ckutio.c, ckcftp.c, and
ckcnet.c about pointers not being assignment compatible, but I have learned
from experience not to try to fix these (see notes from 6 Oct 2009).
29 Mar 2010.

connect(s, (struct sockaddr *)&hisctladdr, sizeof (hisctladdr)): In FTP,
this doesn't work on RHEL5 / Mac OX X 6.1/2 64-bit.  But the connect() in
Telnet works.  On Mac OS X 6.2 I tried changing the socket() call to be like
the one in ckcnet.c for Telnet, but it made no difference.  On a RHEL5.4
system on i386, FTP works fine, so it's not the Red Hat version.  On Digital
Unix 4.0E 64-bit, same thing:

  11:23:10.722 ftp_hookup[kermit.columbia.edu]=21
  11:23:10.722 ftp hookup A[kermit.columbia.edu]
  11:23:10.722 ftp hookup C[kermit.columbia.edu]
  11:23:10.722 ftp hookup socket=4
  11:23:10.722 ftp hookup HADDRLIST
  11:23:10.723 ftp hookup connect failed=13
  11:23:10.723 ftp hookup bad

13 = Permission denied:

  [EACCESS] Search permission is denied for a component of the path prefix;
    or write access to the named socket is denied.

On Gentoo Linux, also on Alpha, the errno is 51: Network is unreachable.
Clearly some data type in the sockets structs is out of whack.

The third connect() argument is "address length".  The address is a
struct sockaddr.  About the third argument, RHEL5 "man connect" says:

  The third argument of connect() is in reality an int (and this is what 
  4.x BSD and libc4 and libc5 have).  Some POSIX confusion resulted in 
  the present socklen_t, also used by glibc.  See also accept(2).

Building on RHEL5 on x86_64, where size_t is 8 and socklen_t is 4, I get a
warning:

  ckcftp.c: In function 'ftp_hookup':
  ckcftp.c:14667: warning:
   comparison is always true due to limited range of data

Referring to:

  if (hisctladdr.sin_addr.s_addr != (unsigned long) -1)

This seems to be the problem; if I remove the (unsigned long) cast (in two
places), the problem goes away.  Actually what I should be comparing it with
is INADDR_NONE, which is defined appropriately in some header file, e.g. as
0xffffffff.  Also I define it explicitly as -1 if it is not defined in any
header file (as is the case in Solaris 9).  Tested OK on 64-bit RHEL5,
32-bit RHEL5, Digital Unix 4.0E 64-bit, Solaris 9 32-bit, Mac OS X 10.4.11
32-bit, Mac OS X 10.6.3 64-bit, AIX 5.3, Gentoo Linux 2.6.31 on Alpha
64-bit, NetBSD 5.0.1 32-bit....  ckcftp.c, 29 Mar 2010.

---C-Kermit 9.0 Alpha.04---

Yesterday's VOID redefinition caused problems for HP-UX in ckuusx.c, in the
curses section where VOID is undef'd and not used to avoid a conflict with
curses.h.  As a workaround I defined a new macro CKVOID with the same
definition as VOID and used it in the offending section of ckuusx.  The real
solution is to replace all references to VOID with CKVOID (since VOID is
increasingly likely to cause conflicts), but a mass search and replace is
not without risks.  ckcdeb.h, ckuusx.c, 30 Mar 2010.

Changed VOID and CKVOID definition to be 'void' for all HP-UX (verified by
PeterE back to HP-UX 6.5, 1989).  Still need to check this on HP-UX 5.21;
if that's an exception it can be done in the makefile.  ckcdeb.h, 30 Mar 2010.

The change I made to allow CONNECT to reestablish a previous SSH connection
prevented a new SSH connection to a different host to be made.  Fixed in
ckuus7.c, 30 Mar 2010.

Fixed mistaken extern declarations of krb4_errno and krb5_errno as strings
in nvlook(); they are ints. Built OK on Mac OS X 10.6.3. ckuus4.c, 30 Mar 2010.

A fix to Trusted HP-UX makefile target from PeterE, to account for the
equivalence of +openssl and +ssl as target suffixes.  30 Mar 2010.

Added a new function \fcvtcsets(string,cset1,cset1) that converts a string
from one character set to another.  The csets are File Character-Set names.
ckuus4.c, 31 Mar 2010.

Added a new function \fdecodehex(string,prefix) that decodes a string
containing prefixed hex bytes.  Default prefix is %%, but any prefix of
one of two chars (such as % or 0x) can be specified.  ckuusr.h, ckclib.h,
ckclib.c, ckuusr.c, 31 Mar 2010.

Richard Nolde reports that Kermit can't find -lpam on Fedora 12 because it's
in /lib rather than /usr/lib.  RHEL5 has symlinks, FC12 should too.  Added a
note to the makefile.  1 Apr 2010.

Build on Solaris 11 for the first time.  Had to adjust ckuver.h to get the
version herald right.  This was on a box that reported its architecture as
i86pc.  1 Apr 2010.

Added MIME character-set names as invisible synonyms in the file and
terminal character-set tables, fcstab[] and tcstab[].  Note that not all the
character sets known to Kermit are registered in MIME.  But at least now
MIME-registered character sets can be referred to by their MIME names, e.g.
ISO-8859-1, ISO646-ES, IBM437, WINDOWS-1252.  These are not listed if you
type ? in a field that is parsing them, unless you type a letter first,
e.g. "i?" lists ISO- and IBM set names.  Later maybe I'll make parallel
tables, or keyword attribute bit that says whether a name is MIME or not.
The real benefit of this change is that now Kermit can take its
character-set names from external sources like email headers or web logs.
ckuxla.c, 1 Apr 2010.

Changed the IF command to accept a bare macro name its condition.  This will
parse and execute correctly if the macro is defined and if it has a numeric
value, or if it is not defined, in which case it evaluates to 0 (FALSE).  If
it is defined but has a non-numeric value, a parse error occurs.  ckuus6.c,
2 Apr 2010.

Added \fstringtype() function.  Given a string argument, it tells whether
the string is 7bit, 8bit, utf8, binary, etc.  ckuusr.h, ckuus[4x].c,
2 Apr 2010.

Did a few builds to make sure there were no booboos.  Solaris 9, NetBSD
5.01, Linux RHEL4, HP-UX 10.20 (non-ANSI compiler and ANSI optimizing
compiler), Mac OS X 10.4.11, SCO OSR 6.00.  5 Apr 2010.

---C-Kermit 9.0 Alpha.05---

Increased maximum variable name length from 4K to 16K. Verified that
too-long names are caught and recovered from correctly.  ckuusr.h, 6 Apr 2010.

Implemented a new \fsplit() option for parsing CSV files, which turns out to
be a little complicated, because the separator is not just a comma, but a
comma and all its surrounding spaces.  Also there are special quoting rules
for fields with embedded commas and fields with embedded quotes.  ckclib.c,
7 Apr 2010.

---C-Kermit 9.0 Alpha.06---

VMS changes from SMS.  They build OK, Kermit file transfers are still OK,
but FTP text-mode GETs always hang on the 10th 8K network read.  Couldn't
get a debug log this time.  ckcmai.c, ckvfio.c, ckvrms.h, ckvker.com.
8 Apr 2010.

Changing VNAML from 4K to 16K broke the build on HP-UX 9.  Put it back to
4K.  9 Apr 2010.

John Dunlap, running days-long stress tests between E-Kermit and C-Kermit,
found a bug in the packet-reading and -decoding code: If a NAK packet
arrives with its length field corrupted to indicate a bigger size, and there
are enough bytes following in the pipeline, ttinl() will return a too-long
packet (if there are not enough bytes waiting to be read, then ttinl() will
properly time out).  In the bad case rpack() trusts the packet length, uses
it as the basis for computation of the block-check length, which is then
used to access memory that might not be there, causing (at least on John's
Linux system) a segmentation fault.  John added the normal clause to check
the result of the block-check calculation, and I changed ttinl() to always
break on the eol character (normally carriage return), since this can never
appear in a packet, even if we "set control unprefix all".  Also added a
check to ttinl() to protect against length fields corrupted into illegal
values.  ckcfn2.c, ckutio.c, 13 Apr 2010.

From Lewis McCarthy:
  Based on code inspection, C-Kermit appears to have an SSL-related security
  vulnerability analogous to that identified as CVE-2009-3767 (see e.g.
  http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2009-3767).

  I'm attaching a patch for this issue relative to the revision of ck_ssl.c
  obtained from a copy of http://www.columbia.edu/kermit/ftp/test/tar/x.zip
  downloaded on 2010/07/30, which I believe is the latest.

  When this flaw was first widely publicized at last year's Black Hat
  conference, it was claimed that some public certificate authorities had
  indeed issued certificates that could be used to exploit this class of
  vulnerability. As far as I know they have not revealed specifically which
  public CA(s) had been found issuing such certificates.
  Some references: http://www.mseclab.com/?p=180
  http://www.theregister.co.uk/2009/07/30/universal_ssl_certificate/

Patches added to ck_ssl.c, 4 Aug 2010.

Peter Eichhorn reported that "RENAME ../x ." didn't work.  This is a side
effect of the changes of 2006 to the RENAME command, there was a little
confusion in the renameone() routine; fixed in ckuus6.c, 4 Aug 2010.

If only one file is FOPEN'd, FCLOSE given with no arguments would close it.
Turns out to be a bad idea. Example: program with an input and output file,
try to close the output file before it is opened by just typing FCLOSE; this
can mess up the input file.   For safety FCLOSE has to require a channel
number or ALL.  ckuus7.c, 4 Aug 2010.

Added \fstrcmp(s1,s2,case,start,length), which has the advantage over IF
EQU,LGT,LLT that case senstivity can be specified as a function arg, and
also substrings can be specified.  ckuusr.h, ckuus[24].c, 5 Aug 2010.

The CSV feature of Alpha.06 had a subtle flaw, namely that if the last item
in a comma separated list was enclosed within doublequotes with a trailing
space after the closing doublequote, a spurious empty final element would be
created in the result array.  Fixed in cksplit(), ckclib.c, 5 Aug 2010.

---Alpha.07---

The CSV feature of \fsplit() splits a comma-separated list into an array.
To turn the array back into a comma separated list, \fjoin(&a,\44,1) almost
works, except for elements contain literal doublequotes, such as:

  Mohammad "The Greatest" Ali

This calls for making a symbolic CSV argument for \fjoin() like the one that
was made for \fsplit(): \fjoin(&a,CSV).  Also \fjoin(&a,TSV) for
Tab-separated list.  Thus if Kermit reads a record in CSV format, splits it
into an array, and then joins the array back into a CSV record, the result
will be equivalent to the original, according to the CSV definition.  It
might not be identical, because if the result had extraneous spaces before
or after the separating commas, these are discarded, but that does not
affect the elements themselves.  Furthermore it is now possible to convert
a comma-separated list into a tab-separated list, and vice versa (which is
not a simple matter of changing commas to tabs or vice versa).  ckuus4.c,
12 Aug 2010.

From Joop Boonen 26 Juli 2010: "Added HAVE_LOCKDEV as openSuSE >= 11.3 uses
lockdev but not baudboy.  They use ttylock directly.  The program code has
been added so the the program works without a problem."  makefile, ckcdeb.h,
ckutio.c, ckuus5.c, 23 Aug 2010.

---Alpha.08---

From Gary Mills at the U of Manitoba: convert Solaris version from BSD ptys
to streams ptys because there are only 48 BSD-style ptys and he was running
out.  No code changes needed, the only change necessary was to add the
following flags to the makefile target:

  -DHAVE_STREAMS -DHAVE_GRANTPT -DHAVE_PTSNAME
  -DPUSH_PTEM -DPUSH_LDTERM -DPUSH_TTCOMPAT

makefile, ckcmai.c, 21 Sep 2010.

Testing this in Solaris 9 I see that the DES library disappeared.  Added
code to the solaris9 targets (also used by Solaris 10 and 11) to check for
this.  makefile, 21 Sep 2010.

The Solaris target checked the OpenSSL version automatically to set the
right flag, the Linux target didn't.  Put the OpenSSL-version testing code
in the Linux target too.  makefile, 21 Sep 2010.

A couple minor changes to the tru64-51b makefile targets from Steven Schweda
but there still are some problems with the Tru64 Unix builds.
makefile, 21 Sep 2010.

---Alpha.09---

\fcontents(\&a[3]) got an error if the array was declared but its dimension
was less than 3, which is bad when dealing with (say) an array created
dynamically by \fsplit(), which might or might not have a third element.
In case it doesn't -- i.e. in case we are referring to an out of range
element of any array that is declared -- we should just return a null
string, as we do with other types of variables that are not defined.
For that matter, ditto even if the array is not declared; what useful
purpose is served by throwing an error in this case?
ckuus4.c, 30 Dec 2010.

cksplit() treats \ as a quoting character.  If the source string contains
backslashes, they are swallowed (or, if doubled, one is kept).  That's not
good for parsing external data, such as lines read from files, where there
are no quoting rules.  This came up when parsing CSV files; as a workaround,
I made \fsplit() treat backslash as an ordinary character for CSV and TSV
splitting (a better solution might be yet another argument that specifies
a quote character).  ckclib.c, 30 Dec 2010.

Began converting C-Kermit to Open Source with the Simplified 3-Clause BSD
license.  Updated the text for the INTRO, LICENSE, NEWS, and SUPPORT
commands.  Fixed things so the copyright year to be displayed is defined in
one place (ck_cryear in ckcmai.c), rather than hardwired into text strings
all over the place.  COPYING.TXT, ckcmai.c, ckuus[256].c, 2 Jun 2011.

When I added MIME synonyms for Kermit character-set names, I left a bogus
entry in the tables ("windows-1251") that was in the wrong place
alphabetically, thus preventing most references to file character-set names
from working right.  Removed the bogus entry.  ckuxla.c, 2 Jun 2011.

Most combinations work OK, but not translating Cyrillic text from UTF-8
to Latin/Cyrillic, and probably the same would be true for any case of
converting from UTF-8 or UCS-2 to anything else.  The problem was in
xgnbyte(), which converts the input stream from the specified character to
UCS2; it needed to make a special case for when the input file was already
Unicode.  Believe it or not, this problem occurred at least as far back as
8.0.201 (9.5 years ago) and nobody noticed.  So if the fix isn't perfect
probably nobody will notice that either.  ckcfns.c, 3 Jun 2011.

The SET BLOCK CHECK command did not parse all the items in its keyword
list.  Fixed in ckuus3.c, 3 Jun 2011.

For EM-APEX ocean floats project, where buoys in stormy waters have to
transmit data through an earth satellite using non-error-correcting modems,
John Dunlap ran exhaustive stress tests of Kermit protocol transfers through
a simulated connection that injected errors and delays and identified a
weakness in Kermit protocol when it is used under extremely bad conditions:
If a data byte of the S packet (or its Ack) is corrupted and the 1-byte
checksum is also corrupted in such a way that that the checksum matches the
corrupted data, the two Kermit programs will disagree as to the negotiated
parameters.  For example, if file Sender's RPT field is changed from '~' to
'^', the receiver will decode the packet incorrectly.  Ditto for most of the
other parameters.  The result is that a corrupted file is received but
reported correct.  John suggested a new mode of operation in which the Type
3 block check is used for all packets.  Such a mode can not be negotiated
because the negotiation packet itself is assumed by all Kermit programs to
have a 1-byte checksum.  Added SET BLOCK-CHECK 5 to the parser (with
invisible synonym FORCE-3".  ckuus3.c, 3 Jun 2011.

Added supporting code for SET BLOCK 5: ckcfn[23].c, ckcpro.w, ckcmai.c,
ckuus3.c, 3 Jun 2011.

Added code to skip the heuristic that S and I packets always have block
check type 1.  File transfer OK between two C-Kermits with SET BLOCK 5.
rpack(): ckcfn2.c, 5 Jun 2011.

Made the file receiver put "5" in the block-check-type in its ACK to the
S-Packet.  spar(): ckcfns.c, 5 Jun 2011.

Now the question is: Can we make the file receiver automatically and safely
recognize a three-byte block check on an incoming S or I packet?  It's
tricky because the block check field is not self-identified, it's just the
last "n" characters of string indicated by the length field, so correct
decoding of the packet depends on stateful knowledge of "n".  How about this:
rpack() already knows what type of packet it is, so if it's an S or I packet
and the 8th byte of the data field is "5" and last 3 bytes, when interpreted
as the CRC, match the packet contents, then we accept the packet and switch
to BLOCK 5 mode.  

On the other hand, if the "5" was put there by corruption, the CRC should
catch the error.  In that case we NAK the packet and presumabely get a
different version back.  There would be no reason to try to re-read the same
packet with a different block check, because the "5" could not possibly be
there legitimately unless it had a 3-byte CRC.  To be clear, this is
cheating.  We read the packet contents before we know the packet is correct,
then we check that it *is* correct.  I made the 4-line change to rpack()
and it works OK in the absense of transmission errors.  ckcfn2.c, 3 Jun 2011.

So the various combinations should work as desired:

 . Sender and receiver both support and are told to SET BLOCK 5 ("SB5").
 . Sender SB5, but receiver doesn't support it (errors out).
 . Sender SB5, receiver supports it but wasn't told (auto-recognizes it).
 . Receiver SB5 but sender no (errors out).

Note in the last case, the receiver should NOT automatically fall back to
standard behavior because if the user said SET BLOCK 5 that means every
packet MUST be protected by CRC to prevent the I/S packets from being
corrupted.

Installed new HELP SET BLOCK-CHECK text.  ckuus2.c, 5 Jun 2011.

Autodownload didn't work when the S or I packet had a 3-byte block check
because kstart() checked it for a 1-byte checksum.  Fixed in kstart(),
ckcfn2.c, 6 Jun 2011.  However, older Kermit versions and programs that
claim to do "autodownload" will never recognize this type of packet.  No
big deal since even if they did, the transfer would fail anyway.

Added 'FORCE 3' to E-Kermit, called it EK 1.7.  The option is "-b 5".  Works
OK for sending and receiving, both with and without the new option.  Also
works with "-b 5" if you send an S packet to it with '5' in the BCT field.
Changes were minimal, I have them all in ek17.diff.

I could probably also make a new G-Kermit in about 10 minutes, but who cares
about G-Kermit...  We already have two useful Kermit programs that
interoperate with the new protocol.  6 Jun 2011.

Replaced the very inadequate help texts for functions \fword() and
\fsplit() with new ones.  ckuus2.c, 6 Jun 2011.

There were a couple reports of file corruption that I was saving for later.
Now that now is later I dug up the messages, files, and logs and it turns
out that nobody had reported a reproducible case of Kermit corrupting a
file.  There have been non-reproducible cases though, almost certainly due
to corruption of the S or I packet or its ACK, which is why we now have SET
BLOCK 5.  Even with BLOCK CHECK 5, there is no guarantee that the same thing
won't happen, it is just far less likely.  Even if we added a 32-bit CRC or
even 64-bit one, there would still be a small chance it could happen.

7 Jun 2011:

Corrected various #ifdefs (or lack of them) when building C-Kermit with
different combinations of feature-selection options such as NOCSETS, NOICP,
NOLOCAL, NOSPL, NOUNICODE, etc.  ckcfns.c ckcmai.c ckcxla.h ckuus2.c
ckuus4.c ckuus5.c ckuus6.c ckuusr.c, 7 Jun 2011.  After running the script
that does all these builds (84 of them) I ran it again to make sure that
none of the changes broke builds that succeeded before the changes were made.

Built OK on Solaris9 ("make solaris9")
Ditto with Krb5 and OpenSSL 0.9.8q  ("make solaris9g+openssl+shadow+pam+zlib")

Built OK on Mac OS X 10.4.11 ("make macosx").
Also "make macosx+krb5+openssl.

Built OK on Linux RHEL4 ("make linux").
Built OK on Linux RHEL4 with OpenSSL 0.9.7a ("make linux+ssl").
Built OK on Linux RHEL5 ("make linux").

"make linux+ssl" fails on RHEL5 because of DES, even though the target
tests for the presence or absence of the DES libraries.  In this case the
libraries are there but they lack the functions des_ecb3_encrypt,
des_random_seed, and des_set_odd_parity.  The build succeeds as:

  make linux+ssl KFLAGS=-UCK_SSL

Since DES is now considered harmful, Jeff Altman suggests that all OpenSSL
builds, even for old versions, should omit it ("If you are building with
openssl and no kerberos or srp, just disable DES.  Disabling DES will impact
telnet and rlogin but it won't matter if you have no ability to negotiate a
session key").

From Ian Beckwith, patches for Debian Linux:
 . Change all '-' to '\(hy' in man page (new pedantry): ckuker.nr.
 . Make IKSD authentication (using PAM) ask for a password when an invalid
   username has been given, to avoid disclosing which account names are valid:
   ckufio.c, ckuus7.c.
 . Fix spelling errors: ckcftp.c, ckuus2.c, ckuker.nr, ckcpro.w, ckuusr.h.
 . Patch makefile to support install to a staging area with DESTDIR.
 . Some other patches (mainly for typos) were for plain-text documentation
   files that were generated from Web pages; I updated the web pages.

A big corporate C-Kermit user has an application where a local C-Kermit
makes a connection to a remote one, uploads some files, and then if the
server has any new patch files for the local, it sends the patches and
does a REMOTE HOST command to run the patch program.  This stopped working
in C-Kermit 6.0 or 7.0 when I put a check to prevent it, because "it makes
no sense to send REMOTE commands to the local end, because the results are
sent back to the remote to be displayed on its screen but it has no screen".
That may be true, but if the user needs to control the local from the
remote, they should be able to.  I removed the checks.  This doesn't solve
the problem of where the output goes; ideally it would go to the local
screen but I don't see any elegant and simple way to make that change.
However the output redirectors can still be used with the REMOTE command
so the results can be captured to a remote file, which could then be sent.
ckuus7.c, 7 Jun 2011.

Changed SET VARIABLE-EVALUATION to SET COMMAND VARIABLE-EVALUATION, but left
the former version available.  ckuusr.c, 9 Jun 2011.

Documented the SET COMMAND VARIABLE-EVALUATION command, which I added in
2008.  ck90.html, 9 Jun 2011.

Renamed all old Mac OS X makefile targets to have the prefix "old" to avoid
confusing them with the current targets, and made macosx10 a synonym for
macosx, so those who used previous makefiles will get a current target
without having to know the new name.  makefile, 9 Jun 2011.

Added XMESSAGE, which is to MESSAGE as XECHO is ECHO: prints the text
without a line terminator, so it can be continued by subsequent [X]MESSAGE
commands.  ckuusr.[ch], 9 Jun 2011.

Back to "make linux+ssl" on RHEL5...  I took the coward's way out and added
code to the makefile target to check whether the build worked (somebody let
me know if there is a better way to check), and if not to give a message
suggesting they "make clean ; make linux+ssl KFLAGS=-UCK_DES".  makefile,
9 Jun 2011.

Noticed that \frecurse() would dump core if called with no arguments.
Fixed in ckuus4.c, 9 Jun 2011.

Added \q() as an alternative to the more verbose \fliteral() for quoting
strings that contain characters (like \) that would otherwise be significant
to Kermit.  It's more efficient because it isn't a function call, and 'q'
is an intuitive letter to mean 'quote'.  It also works better than
\fliteral() because functions treat commas and braces specially.  ckuus4.c,
10 Jun 2011.

Built OK on VMS 8.3 on Alpha, no net.  DEC C caught a couple glitches in the
new code that gcc didn't catch, which I fixed.  ckuus[25].c, 10 Jun 2011.

Built OK on VMS 8.3 on Alpha with Multinet 5.3.  The SSL build failed but
I'm not going to worry about it.  10 Jun 2011.

Built OK on NetBSD 5.1.  10 Jun 2011.

Tried to resurrect my old "build-all" machine, an IBM Netfinity 3500 from
1997 with 20-some mountable bootable hard disks with lots of 1990s OS's on
them.  No dice.  I can see the BIOS but not the hard disks.  The
configuration is still correct because it tries to boot from the mountable
hard disk, but it fails (I tried six different ones).

Tried to resurrect my old Siemens Nixdorf RM 200 MIPS machine.  Booted OK,
headless even, but makes a hellish high-pitched whine, like a dentist drill.
It's pretty slow too.  "make sinix542" (for SINIX 5.4.2) bombed at link
time on no rdchk().  Fixed by #including <sys/filio.h>.  ckutio.c, 10 Jun 2011.

Tried to resurrect my old SCO Xenix 2.3.4 machine, also headless.  Amazingly
it still works; it can't use a monitor but I can Telnet to it.  Had to tweak
some #ifdefs but I got a no-net version built successfully.  According to my
notes, it hasn't been possible to build with TCP/IP since C-Kermit 8.0,
but how many people ever had SCO Xenix 2.3.4 with TCP/IP anyway?  Anyway we
still have the binaries for C-Kermit 7.0.  ckuus4.c, 10 Jun 2011.

Built OK on AIX 5.3.  Built OK on Solaris 10.  11 Jun 2011.

Tried harder to revive the build-all machine, now it sort of works, but not
all of the bootable OS's work.  Built C-Kermit 9.0 OK on OpenBSD 3.0.  Built
OK on QNX 4.25 but had to #ifdef references to IXANY in ckutio.c and ckupty.
Built OK on NetBSD 1.5.1 (2000).  Tried "make netbsd+ssl" on this one, it's
OpenSSL 0.9.5a 1 Apr 2000, but it bombs out in ckuath.c, no big deal.
Another problem in NetBSD 1.5.2 is that even though off_t is 8, CK_OFF_T
is 4.  Worth noting but not worth fixing unless someone else notices.
13 Jun 2011.

SuSE 7.0... boots OK but telnet server doesn't work.  Can telnet out but
it's too flaky, connection drops if I try to transfer a file.

OpenBSD 2.5 [1999] OK.  Red Hat 7.1 OK.  Red Hat 7.1 with OpenSSL 0.9.6
not OK, same error as with 0.9.5a:

ckuath.c
In file included from ck_ssl.h:48,
                 from ckuath.c:225:
/usr/include/openssl/des.h:77: warning: redefinition of `Block'
ckuat2.h:86: warning: `Block' previously declared here
/usr/include/openssl/des.h:83: redefinition of `struct des_ks_struct'
/usr/include/openssl/des.h:91: warning: redefinition of `Schedule'
ckuat2.h:90: warning: `Schedule' previously declared here

So it appears that OpenSSL support is broken for pre-0.9.7.  Tried
building it again with -UCK_SSL (since the errors are originating from
from des.h)...  But it still failed exactly the same way.  I found
#includes for des.h in ckuath.c and and ck_ssl.h and #ifdef'd them out,
but it still fails:

In file included from /usr/include/openssl/evp.h:89,
                 from /usr/include/openssl/x509.h:67,
                 from /usr/include/openssl/ssl.h:69,
                 from ck_ssl.h:51,
                 from ckuath.c:227:
/usr/include/openssl/des.h:77: warning: redefinition of `Block'
ckuat2.h:86: warning: `Block' previously declared here
/usr/include/openssl/des.h:83: redefinition of `struct des_ks_struct'
/usr/include/openssl/des.h:91: warning: redefinition of `Schedule'
ckuat2.h:90: warning: `Schedule' previously declared here

Built OK on Debian 2.1.  13 Jun 2011.

On FreeBSD 4.4, it blows up with:
ckufio.c: In function vpass':
ckufio.c:8201: conflicting types for 'initgroups'
/usr/include/unistd.h:154: previous declaration of 'initgroups'
ckufio.c:8201: warning: extern declaration of 'initgroups' doesn't match global
one.  Fixed by defining NODCLINITGROUPS for FreeBSD in ckufio.c. It might not 
be the right fix, but I don't have a lot of other FreeBSD versions to
compare with.  Anyway now it builds OK on 4.4, and also on FreeBSD 3.3.
ckufio.c, 13 Jun 2011.

Tried to build on SCO Open Server 5.0.7 but it fails at link time because
it can't find rdchk().  But it's supposed to be there!  Come back to this
later...

Red Hat 6.1             i386   32/64   linux          2332545
Red Hat 7.1             i386   32/64   linux          2368528
Red Hat EL4             i386   32/74   linux          2363067
Red Hat EL5.6           i386    64     linux          2371279
Solaris9                sparc  32/64   solaris9       2849896
Solaris9+ssl            sparc  32/64   solaris9       5021764
Solaris10               sparc  32/64   solaris10      2855776
QNX                     i386    32     qnx32          2012323 
NetBSD 1.5.1            i386   32/64   netbsd         2198055
NetBSD 5.1              i386   32/64   netbsd         2159863
OpenBSD 2.5             i386   32/64   openbsd        2236036
Mac OS X 10.6.7         x86_64  64     macosx         2.7M
Mac OS X 10.4.11        ppc    32/64   macosx         2496304
Debian 2.1              i386   32/64   linux          2213221
FreeBSD 4.4             i386   32/64   freebsd        2291333
FreeBSD 3.3             i386   32/64   freebsd        2147370
SINIX 5.42              mips    32     sinix542       3319325 (1995)
SCO Unixware 2.1.3      i386    32     uw213          2242176
SCO OSR6.0.0            i386   32/64   sco_osr600     2368300

More builds, 14 June 2011:

VMS 6.2                 alpha   32     make mn        2556928 No TCP/IP
VMS 6.2                 alpha   32     make m         3112960 UCX 4.0
Solaris 11              i386   32/64   solaris11      2823860
Solaris 11              i386   32/64   solaris11+ssl  2993660 OpenSSL 0.9.8l
NetBSD 5.1              i386   32/64   netbsd+krb5    2307855 Kerberos 5
Linux Slackware 12.1.0  i386   32/65   linux          2175754
Linux Fedora 14         i386   32/64   linux          2256514
Linux Fedora 14         i386   32/64   linux+ssl      ....... OpenSSL 1.0.0d
Linux Fedora 14         i386   32/64   linux+krb      2449614 (*)

(*) make linux+krb5 "LIBS=$LIBS /lib/libk5crypto.so.3 /lib/libcom_err.so.2"

Noticed that netbsd+ssl build on NetBSD 5.1 said "NetBSD 1.5" in its banner.
Fixed by replacing the old hardwired target with the new "subroutinized"
target a'la linux+ssl and adapting it to NetBSD.  makefile, 15 Jun 2011.

Same deal for Kerberos 5, make a new netbsd+krb5 target and it builds ok,
at least once one figures out where the Kerberos headers and libs are.
makefile, 15 Jun 2011.

Same deal for the netbsdnc target, now it simply defined NOCURSES and
chains to the main netbsd target.  makefile, 15 Jun 2011.

Tried to build with Kerberos 5 on Solaris, fails because the DES library
no longer exists.  This one is beyond me, sorry.

Made new targets for MirBSD, mirbsd and mirbsd+ssl, makefile 15 Jun 2011.

In OpenSUSE 11.2 with OpenSSL 0.9.8r we bomb on undefined references from
various DES library routines.  Builds OK without DES.

Various linux+krb5 builds fail because can't find -lgssapi_krb5

SSL builds with OpenSSL < 0.9.7 fail even though there is code to support
the older SSL.

Fixed some printf %ld vs int instances in the sizeofs section of SHOW FEATURES.
ckuus5.c, 15 Jun 2011.

Fixed the new linux+ssl target to actually use the SSLINC and SSLLIBS
definitions, oops.  makefile, 15 Jun 2011.

15 June 2011 builds (Beta.01):

AIX 5.3                 ppc    32/64   aix+ssl        3283846 OpenSSL 0.9.8m
NetBSD 5.1              i386   32/64   netbsd         2159863
NetBSD 5.1              i386   32/64   netbsd+ssl     2350274 OpenSSL 0.9.9-dev
NetBSD 5.1              i386   32/64   netbsd+krb5    2349627 MIT Krb5 1.6.3
FreeBSD 8.2             i386   32/64   freebsd        2298414
FreeBSD 8.2             i386   32/64   freebsd+ssl    2448961 OpenSSL 0.9.8q
OpenBSD 4.7             i386   32/64   openbsd        2266132
OpenBSD 4.7             i386   32/64   openbsd+ssl    2409263 OpenSSL 0.9.8k
MirBSD 10               i386   32/64   mirbsd         2216601
MirBSD 10               i386   32/64   mirbsd+ssl     2358318 OpenSSL 0.9.8r
OpenSuse 11.2           x86_64  64     linux          2348468
OpenSuse 11.2           x86_64  64     linux+ssl (*)  2546540 OpenSSL 0.9.8r
RHEL 5.6                ia64    64     linux          4390687
RHEL 5.6                ia64    64     linux+ssl (*)  4775007 OpenSSL 0.9.8e
Ubuntu 9.10             i386   32/64   linux          2275523
Ubuntu 9.10             i386   32/64   linux+ssl      2466708 OpenSSL 0.9.8r
Gentoo 1.12.13          ppc    32/64   linux          2386597
Gentoo 1.12.13          ppc64   64     linux          2749015
Gentoo 1.12.13          ppc64   64     linux+ssl      3002150 OpenSSL 0.9.8r
Gentoo 1.12.13          sparc  32/64   linux          2478382
Gentoo 1.12.13          sparc  32/64   linux+ssl      2690499 OpenSSL 0.9.8r
Solaris 9               sparc  32/64   solaris9       2849896
Solaris 10              i386   32/64   solaris10      2837620
IRIX 6.5                R10000 32/64   irix65         2869704

* and KFLAGS=-UCK_DES

Tried building on NetBSD 5.1 with Heimdal Kerberos using:

make netbsd+krb5 \
 "KFLAGS=-DHEIMDAL" \
 "K5INC=-I/usr/include" \
 "K5LIB=-L/usr/lib"

It found all its headers OK, but it blew up in ckuath.c.  Small wonder,
ckccfg.html says:

HEIMDAL
    Should be defined if Kerberos V support is provided by HEIMDAL. Support
    for this option is not complete in C-Kermit 8.0. Anyone interested in
    working on this should contact kermit-support. 

'krb5-config --version' gives the MIT Kerberos 5 version number.

Make a new netbsd+krb5+ssl target based on the combination of the new
netbsd+ssl and netbsd+krb5 targets.  There were lots of warnings in the
compilation but no errors, but it produced an executable that starts and
does normal things but I have no idea if the SSL or Kerberos functions work.
makefile, 16 Jun 2011.

Changed the cu-solaris9-krb5 target to test for the presence of DES because
DES isn't there, to see if this would allow a Kerberos build to proceed.
And it worked, amazing.  At least the build completed, I have no way to test
the Kerberos part.  makefile, 16 Jun 2011.

Updated the solaris9+ssl target to do the DES testing.  makefile, 16 Jun 2011.

Updated cu-solaris+krb5 target to test whether the GSSAPI library is called
libgassapi or libgassapi_krb5.  makefile, 16 Jun 2011.

Added lots of tests to the Linux Kerberos 5 entries, linux+krb5 and
linux+krb5+ssl, because some have libk5crypto and some don't; some have
libcom_err and some don't; and some have libgssapi_krb5 (e.g. RHEL5,
OpenSuse 11.2) whereas others have libgssapi (Gentoo).

16 June 2011 builds (Beta.01):

NetBSD 5.1 i386   32/64  netbsd+krb5+ssl   2451757 OpenSSL 0.9.9 MIT Krb5 1.6.3
Solaris 9  sparc  32/64  solaris9+krb5     2543036 MIT Kerberos 5 1.7.1
Solaris 9  sparc  32/64  solaris9+ssl      5021544 OpenSSL 0.9.8q (gcc)
Gentoo...  ppc    32/64  linux             2386597
Gentoo...  ppc    32/64  linux+ssl         2593561 OpenSSL 0.9.8r 
Gentoo...  ppc64   64    linux             2749015
Gentoo...  ppc64   64    linux+ssl         3002150 OpenSSL 0.9.8r 
RHEL5      x86_64  64    linux+krb5 (*)    2563878 MIT Kerberos 5 1.6.1
RHEL5      x86_64  64    linux+krb5+ssl(*) 2563878 MIT Kerberos 5 1.6.1
Fedora 14  i386   32/64  linux+krb5+ssl    2539891 MIT Krb5 + OpenSSL 0.9.8r

* KFLAGS=-UCK_DES

--- C-Kermit 9.0.299 Beta.01 ---

sizeof() can return a long or an int, so neither printf("%d",sizeof(blah));
or printf("%ld",sizeof(blah)); can be used everywhere.  Changed the
"sizeofs" section of SHOW FEATURES in the dumbest (and therefore most
portable) way to squelch the warnings.  ckuus5.c, 17 Jun 2011.

From John Dunlap: "Watching the server screen led me to offer a cosmetic
patch for ckuusx.c.  I noticed that the server screen said it was
"RESENDING" when it really wasn't.  The attached patch emits blanks to
insure that old labels are completely erased."  ckuusx.c, 17 Jun 2011.

Nelson Beebe found two places where I had SSLLIBS in the makefile instead of
SSLLIB.  makefile, 18 Jun 2011.

More important he knew how to force gcc to load the right header files for
OpenSSL 1.0.0d (by using '-isystem' rather than '-I').  Previously it was
using the 0.9.8r header files but linking with the 1.0.0d libraries.  This
is not in the sources, it's done in the 'make' command, e.g.:

  export PATH=/usr/bin:$PATH
  export SSLINC=-isystem/usr/include
  export "SSLLIB=-L/usr/lib -Wl,-rpath,/usr/lib"
  make linux+ssl

Folded the previous linux+openssl+zlib+shadow+pam and linux+openssl+shadow
targets into linux+ssl.  Checked the linuxso (scripting only) target, builds
OK, 600K.  Made new subroutinized linux+krb5+krb4 target but can't find
anyplace to test it.   Made new subroutinized linux+shadow+pam target, works
fine on RHEL4.  Revised comments and lists again.  makefile, 18 Jun 2011.

For the pluggable-disk OS's that boot OK but lack a working network, I
rigged up a serial connection using a DB9-FF null modem cable, and then a
DB9-MF modem cable to make it reach.  I don't see any modem signals on
either end, but the data goes through OK.  COM1 on the desktop PC,
/dev/ttyS1 or whatever on Lab.  Since there are no modem signals, can't use
RTS/CTS.  At 57600bps with Xon/Xoff, 500-byte packets and sliding windows,
transfers work OK at about 5000cps using 5 window slots; takes 8 minutes to
transfer the gzipped C-Kermit tarball.  Kermit to the rescue.  19 Jun 2011.

Transferred the tarball over serial ports to SCO OSR5.0.5 at 38.4Kbps, the
highest speed supported, 12 minutes, no errors, 3300cps.  Unpack, make
sco32v505udk, OK.  Also built the TCP/IP version and it almost made an
outbound connection, but only once.  19 Jun 2011.

Ditto for Solaris 2.6/i386, except 57.6Kbps, 4K-byte packets, no problem.
Solaris 8/i386, ditto.  19 Jun 2011.

SCO OpenServer 5.0.5  i386   32   sco32v505udk     1940964  No TCP/IP
SCO OpenServer 5.0.5  i386   32   sco32v505udknet  2314668  With TCP/IP
Sun Solaris 2.6       i386   32   solaris26g       4661368
Sun Solaris 8         i386   32   solaris8g        4675432

When using compact substring notation, \s(xx[4]) returns the whole string
xx starting at position 4, but \s(xx[4:]) returns an empty string.  Fixed
the latter to be like the former.  ckuus5.c, 20 Jun 2010.

Really it would have been nicer if \s(xx[4]) returned a single character,
the 4th character of xx, but it's too late now.  Added another "separator"
character '.' (period) for that: \s(xx[4.]) is the 4th character of xx.
ckuus4.c, 20 Jun 2010.

Back to SCO OSR5.0.7... This failed before because 'rdchk' came up unknown
at link time, unlike all previous OSR5's, that used rdchk() in place of the
FIONREAD ioctl.  Added #ifdefs to make a special case for 5.0.7.  I'm not
sure this is the best way, but this is the minimal change to get it to work.
If anybody cares, maybe the same can be done for previous OSR5 releases.
ckutio.c, 20 Jun 2010 (search for SCO_OSR507).

SCO OpenServer 5.0.7  i386   32   sco32v507        1895724  No TCP/IP
SCO OpenServer 5.0.7  i386   32   sco32v507net     2246792  With TCP/IP

Checked current code on RHEL4, found that my GSSAPI-lib finding makefile
target didn't look in enough places; added some more.  makefile, 21 Jun 2011.

Got reports back on HPUX from Peter Eichhorn, almost all good on HP-UX 8, 9,
10, and 11.  21 Jun 2011.

Got access to Debian 5.0 and 7-to-be ("Wheezy/Sid").  Regular 'make linux' is
OK in Debian 5, but in 7 can't find crypt, res_search, or dn_expand; had
to add more library search clauses to 'make linux'.  makefile, 21 Jun 2011.

In Debian 7.0, libk5crypto could not be found without adding another clause
to 'make linux+krb5'.  That done, the SSL build (1.0.0d) was OK, as well as
the krb5+ssl one.  makefile, 21 Jun 2011.

I found a Linux box that had both Kerberos 4 and 5 installed and tried 'make
linux+krb5+krb4', which failed because of missing DES functions.  Tried
'make linux+krb5+krb4 KFLAGS=-UCK_DES', but that fails too, even though it
doesn't fail for Kerberos 5 alone, so probably some Krb4 code is making
unguarded calls to the DES routines.  What is really needed is a way to
completely strip all DES references from any given build.  21 Jun 2011.

-----------------------------------
To check:

after logging out from "ssh jezebel":
(/home/fdc/) C-Kermit>ssh jezebel
 Closing connection
Sorry, network type not supported
(/home/fdc/) C-Kermit> # This happens in Linux but not Solaris

---------------------------------
***************************