#if HAVE_ICONV
# include <iconv.h>
-# include "utf8-ucs4-safe.h"
-# include "ucs4-utf8.h"
# include "unistr.h"
#endif
-#include "strdup.h"
#include "c-strcase.h"
+#include "c-strcaseeq.h"
#ifndef SIZE_MAX
# define SIZE_MAX ((size_t) -1)
iconv (cd, (ICONV_CONST char **) (inbuf), inbytesleft, outbuf, outbytesleft))
# endif
+/* iconv_carefully_1 is like iconv_carefully, except that it stops after
+ converting one character. */
+static size_t
+iconv_carefully_1 (iconv_t cd,
+ const char **inbuf, size_t *inbytesleft,
+ char **outbuf, size_t *outbytesleft,
+ bool *incremented)
+{
+ const char *inptr = *inbuf;
+ const char *inptr_end = inptr + *inbytesleft;
+ char *outptr = *outbuf;
+ size_t outsize = *outbytesleft;
+ const char *inptr_before = inptr;
+ size_t res = (size_t)(-1);
+ size_t insize;
+
+ for (insize = 1; inptr + insize <= inptr_end; insize++)
+ {
+ res = iconv (cd,
+ (ICONV_CONST char **) &inptr, &insize,
+ &outptr, &outsize);
+ if (!(res == (size_t)(-1) && errno == EINVAL))
+ break;
+ /* We expect that no input bytes have been consumed so far. */
+ if (inptr != inptr_before)
+ abort ();
+ }
+
+ *inbuf = inptr;
+ *inbytesleft = inptr_end - inptr;
+# if !defined _LIBICONV_VERSION && !defined __GLIBC__
+ /* Irix iconv() inserts a NUL byte if it cannot convert.
+ NetBSD iconv() inserts a question mark if it cannot convert.
+ Only GNU libiconv and GNU libc are known to prefer to fail rather
+ than doing a lossy conversion. */
+ if (res != (size_t)(-1) && res > 0)
+ {
+ /* iconv() has already incremented INPTR. We cannot go back to a
+ previous INPTR, otherwise the state inside CD would become invalid,
+ if FROM_CODESET is a stateful encoding. So, tell the caller that
+ *INBUF has already been incremented. */
+ *incremented = (inptr > inptr_before);
+ errno = EILSEQ;
+ return (size_t)(-1);
+ }
+# endif
+
+ if (res != (size_t)(-1))
+ {
+ *outbuf = outptr;
+ *outbytesleft = outsize;
+ }
+ *incremented = false;
+ return res;
+}
+
+/* utf8conv_carefully is like iconv, except that
+ - it converts from UTF-8 to UTF-8,
+ - it stops as soon as it encounters a conversion error, and it returns
+ in *INCREMENTED a boolean telling whether it has incremented the input
+ pointers past the error location,
+ - if one_character_only is true, it stops after converting one
+ character. */
+static size_t
+utf8conv_carefully (bool one_character_only,
+ const char **inbuf, size_t *inbytesleft,
+ char **outbuf, size_t *outbytesleft,
+ bool *incremented)
+{
+ const char *inptr = *inbuf;
+ size_t insize = *inbytesleft;
+ char *outptr = *outbuf;
+ size_t outsize = *outbytesleft;
+ size_t res;
+
+ res = 0;
+ do
+ {
+ ucs4_t uc;
+ int n;
+ int m;
+
+ n = u8_mbtoucr (&uc, (const uint8_t *) inptr, insize);
+ if (n < 0)
+ {
+ errno = (n == -2 ? EINVAL : EILSEQ);
+ n = u8_mbtouc (&uc, (const uint8_t *) inptr, insize);
+ inptr += n;
+ insize -= n;
+ res = (size_t)(-1);
+ *incremented = true;
+ break;
+ }
+ if (outsize == 0)
+ {
+ errno = E2BIG;
+ res = (size_t)(-1);
+ *incremented = false;
+ break;
+ }
+ m = u8_uctomb ((uint8_t *) outptr, uc, outsize);
+ if (m == -2)
+ {
+ errno = E2BIG;
+ res = (size_t)(-1);
+ *incremented = false;
+ break;
+ }
+ inptr += n;
+ insize -= n;
+ if (m == -1)
+ {
+ errno = EILSEQ;
+ res = (size_t)(-1);
+ *incremented = true;
+ break;
+ }
+ outptr += m;
+ outsize -= m;
+ }
+ while (!one_character_only && insize > 0);
+
+ *inbuf = inptr;
+ *inbytesleft = insize;
+ *outbuf = outptr;
+ *outbytesleft = outsize;
+ return res;
+}
+
static int
mem_cd_iconveh_internal (const char *src, size_t srclen,
iconv_t cd, iconv_t cd1, iconv_t cd2,
enum iconv_ilseq_handler handler,
size_t extra_alloc,
+ size_t *offsets,
char **resultp, size_t *lengthp)
{
/* When a conversion error occurs, we cannot start using CD1 and CD2 at
char *result;
size_t allocated;
size_t length;
+ size_t last_length = (size_t)(-1); /* only needed if offsets != NULL */
- if (*lengthp >= sizeof (tmpbuf))
+ if (*resultp != NULL && *lengthp >= sizeof (tmpbuf))
{
initial_result = *resultp;
allocated = *lengthp;
allocated = sizeof (tmpbuf);
}
result = initial_result;
+
+ /* Test whether a direct conversion is possible at all. */
+ if (cd == (iconv_t)(-1))
+ goto indirectly;
+
+ if (offsets != NULL)
+ {
+ size_t i;
+
+ for (i = 0; i < srclen; i++)
+ offsets[i] = (size_t)(-1);
+
+ last_length = (size_t)(-1);
+ }
length = 0;
/* First, try a direct conversion, and see whether a conversion error
size_t res;
bool grow;
- /* Use iconv_carefully instead of iconv here, because:
- - If TO_CODESET is UTF-8, we can do the error handling in this loop,
- no need for a second loop,
- - With iconv() implementations other than GNU libiconv and GNU libc,
- if we use iconv() in a big swoop, checking for an E2BIG return,
- we lose the number of irreversible conversions. */
- res = iconv_carefully (cd,
- &inptr, &insize,
- &outptr, &outsize,
- &incremented);
+ if (offsets != NULL)
+ {
+ if (length != last_length) /* ensure that offset[] be increasing */
+ {
+ offsets[inptr - src] = length;
+ last_length = length;
+ }
+ res = iconv_carefully_1 (cd,
+ &inptr, &insize,
+ &outptr, &outsize,
+ &incremented);
+ }
+ else
+ /* Use iconv_carefully instead of iconv here, because:
+ - If TO_CODESET is UTF-8, we can do the error handling in this
+ loop, no need for a second loop,
+ - With iconv() implementations other than GNU libiconv and GNU
+ libc, if we use iconv() in a big swoop, checking for an E2BIG
+ return, we lose the number of irreversible conversions. */
+ res = iconv_carefully (cd,
+ &inptr, &insize,
+ &outptr, &outsize,
+ &incremented);
length = outptr - result;
grow = (length + extra_alloc > allocated / 2);
goto done;
indirectly:
- /* The direct conversion failed, handler != iconveh_error,
- and cd2 != (iconv_t)(-1).
+ /* The direct conversion failed.
Use a conversion through UTF-8. */
+ if (offsets != NULL)
+ {
+ size_t i;
+
+ for (i = 0; i < srclen; i++)
+ offsets[i] = (size_t)(-1);
+
+ last_length = (size_t)(-1);
+ }
length = 0;
{
+ const bool slowly = (offsets != NULL || handler == iconveh_error);
# define utf8bufsize 4096 /* may also be smaller or larger than tmpbufsize */
char utf8buf[utf8bufsize + 1];
size_t utf8len = 0;
/* Set to the initial state. */
if (cd1 != (iconv_t)(-1))
iconv (cd1, NULL, NULL, NULL, NULL);
- iconv (cd2, NULL, NULL, NULL, NULL);
+ if (cd2 != (iconv_t)(-1))
+ iconv (cd2, NULL, NULL, NULL, NULL);
# endif
while (in1size > 0 || do_final_flush1 || utf8len > 0 || do_final_flush2)
/* Conversion step 1: from FROM_CODESET to UTF-8. */
if (in1size > 0)
{
+ if (offsets != NULL
+ && length != last_length) /* ensure that offset[] be increasing */
+ {
+ offsets[in1ptr - src] = length;
+ last_length = length;
+ }
if (cd1 != (iconv_t)(-1))
- res1 = iconv_carefully (cd1,
- (ICONV_CONST char **) &in1ptr, &in1size,
- &out1ptr, &out1size,
- &incremented1);
+ {
+ if (slowly)
+ res1 = iconv_carefully_1 (cd1,
+ &in1ptr, &in1size,
+ &out1ptr, &out1size,
+ &incremented1);
+ else
+ res1 = iconv_carefully (cd1,
+ &in1ptr, &in1size,
+ &out1ptr, &out1size,
+ &incremented1);
+ }
else
{
/* FROM_CODESET is UTF-8. */
- res1 = 0;
- do
- {
- ucs4_t uc;
- int n;
- int m;
-
- n = u8_mbtouc_safe (&uc, (const uint8_t *) in1ptr, in1size);
- if (uc == 0xfffd
- && !(n >= 3
- && (uint8_t)in1ptr[0] == 0xEF
- && (uint8_t)in1ptr[1] == 0xBF
- && (uint8_t)in1ptr[2] == 0xBD))
- {
- in1ptr += n;
- in1size -= n;
- errno = EILSEQ;
- res1 = (size_t)(-1);
- incremented1 = true;
- break;
- }
- if (out1size == 0)
- {
- errno = E2BIG;
- res1 = (size_t)(-1);
- incremented1 = false;
- break;
- }
- m = u8_uctomb ((uint8_t *) out1ptr, uc, out1size);
- if (m == -2)
- {
- errno = E2BIG;
- res1 = (size_t)(-1);
- incremented1 = false;
- break;
- }
- in1ptr += n;
- in1size -= n;
- if (m == -1)
- {
- errno = EILSEQ;
- res1 = (size_t)(-1);
- incremented1 = true;
- break;
- }
- out1ptr += m;
- out1size -= m;
- }
- while (in1size > 0);
+ res1 = utf8conv_carefully (slowly,
+ &in1ptr, &in1size,
+ &out1ptr, &out1size,
+ &incremented1);
}
}
else if (do_final_flush1)
errno1 = errno;
utf8len = out1ptr - utf8buf;
- if (in1size == 0
+ if (offsets != NULL
+ || in1size == 0
|| utf8len > utf8bufsize / 2
|| (res1 == (size_t)(-1) && errno1 == E2BIG))
{
bool grow;
if (in2size > 0)
- res2 = iconv_carefully (cd2,
- &in2ptr, &in2size,
- &out2ptr, &out2size,
- &incremented2);
+ {
+ if (cd2 != (iconv_t)(-1))
+ res2 = iconv_carefully (cd2,
+ &in2ptr, &in2size,
+ &out2ptr, &out2size,
+ &incremented2);
+ else
+ /* TO_CODESET is UTF-8. */
+ res2 = utf8conv_carefully (false,
+ &in2ptr, &in2size,
+ &out2ptr, &out2size,
+ &incremented2);
+ }
else /* in1size == 0 && !do_final_flush1
&& in2size == 0 && do_final_flush2 */
{
state. But avoid glibc-2.1 bug and Solaris 2.7 bug. */
# if defined _LIBICONV_VERSION \
|| !((__GLIBC__ == 2 && __GLIBC_MINOR__ <= 1) || defined __sun)
- res2 = iconv (cd2, NULL, NULL, &out2ptr, &out2size);
-# else
- res2 = 0;
+ if (cd2 != (iconv_t)(-1))
+ res2 = iconv (cd2, NULL, NULL, &out2ptr, &out2size);
+ else
# endif
+ res2 = 0;
do_final_flush2 = false;
incremented2 = true;
}
int n;
if (in2size == 0)
abort ();
- n = u8_mbtouc (&uc, (const uint8_t *) in2ptr,
- in2size);
+ n = u8_mbtouc_unsafe (&uc, (const uint8_t *) in2ptr,
+ in2size);
in2ptr += n;
in2size -= n;
}
inptr = scratchbuf;
insize = scratchlen;
- res = iconv (cd2,
- (ICONV_CONST char **) &inptr, &insize,
- &out2ptr, &out2size);
+ if (cd2 != (iconv_t)(-1))
+ res = iconv (cd2,
+ (ICONV_CONST char **) &inptr, &insize,
+ &out2ptr, &out2size);
+ else
+ {
+ /* TO_CODESET is UTF-8. */
+ if (out2size >= insize)
+ {
+ memcpy (out2ptr, inptr, insize);
+ out2ptr += insize;
+ out2size -= insize;
+ inptr += insize;
+ insize = 0;
+ res = 0;
+ }
+ else
+ {
+ errno = E2BIG;
+ res = (size_t)(-1);
+ }
+ }
length = out2ptr - result;
if (res == (size_t)(-1) && errno == E2BIG)
{
out2ptr = result + length;
out2size = allocated - extra_alloc - length;
- res = iconv (cd2,
- (ICONV_CONST char **) &inptr, &insize,
- &out2ptr, &out2size);
+ if (cd2 != (iconv_t)(-1))
+ res = iconv (cd2,
+ (ICONV_CONST char **) &inptr,
+ &insize,
+ &out2ptr, &out2size);
+ else
+ {
+ /* TO_CODESET is UTF-8. */
+ if (!(out2size >= insize))
+ abort ();
+ memcpy (out2ptr, inptr, insize);
+ out2ptr += insize;
+ out2size -= insize;
+ inptr += insize;
+ insize = 0;
+ res = 0;
+ }
length = out2ptr - result;
}
# if !defined _LIBICONV_VERSION && !defined __GLIBC__
mem_cd_iconveh (const char *src, size_t srclen,
iconv_t cd, iconv_t cd1, iconv_t cd2,
enum iconv_ilseq_handler handler,
+ size_t *offsets,
char **resultp, size_t *lengthp)
{
return mem_cd_iconveh_internal (src, srclen, cd, cd1, cd2, handler, 0,
- resultp, lengthp);
+ offsets, resultp, lengthp);
}
char *
char *result = NULL;
size_t length = 0;
int retval = mem_cd_iconveh_internal (src, strlen (src),
- cd, cd1, cd2, handler, 1,
+ cd, cd1, cd2, handler, 1, NULL,
&result, &length);
if (retval < 0)
mem_iconveh (const char *src, size_t srclen,
const char *from_codeset, const char *to_codeset,
enum iconv_ilseq_handler handler,
+ size_t *offsets,
char **resultp, size_t *lengthp)
{
if (srclen == 0)
*lengthp = 0;
return 0;
}
- else if (c_strcasecmp (from_codeset, to_codeset) == 0)
+ else if (offsets == NULL && c_strcasecmp (from_codeset, to_codeset) == 0)
{
char *result;
# endif
cd = iconv_open (to_codeset, from_codeset);
- if (cd == (iconv_t)(-1))
- return -1;
- if (c_strcasecmp (from_codeset, "UTF-8") == 0)
+ if (STRCASEEQ (from_codeset, "UTF-8", 'U','T','F','-','8',0,0,0,0))
cd1 = (iconv_t)(-1);
else
{
if (cd1 == (iconv_t)(-1))
{
int saved_errno = errno;
- iconv_close (cd);
+ if (cd != (iconv_t)(-1))
+ iconv_close (cd);
errno = saved_errno;
return -1;
}
}
- if (c_strcasecmp (to_codeset, "UTF-8") == 0)
+ if (STRCASEEQ (to_codeset, "UTF-8", 'U','T','F','-','8',0,0,0,0))
cd2 = (iconv_t)(-1);
else
{
int saved_errno = errno;
if (cd1 != (iconv_t)(-1))
iconv_close (cd1);
- iconv_close (cd);
+ if (cd != (iconv_t)(-1))
+ iconv_close (cd);
errno = saved_errno;
return -1;
}
result = *resultp;
length = *lengthp;
- retval =
- mem_cd_iconveh (src, srclen, cd, cd1, cd2, handler, &result, &length);
+ retval = mem_cd_iconveh (src, srclen, cd, cd1, cd2, handler, offsets,
+ &result, &length);
if (retval < 0)
{
iconv_close (cd2);
if (cd1 != (iconv_t)(-1))
iconv_close (cd1);
- iconv_close (cd);
+ if (cd != (iconv_t)(-1))
+ iconv_close (cd);
errno = saved_errno;
}
else
int saved_errno = errno;
if (cd1 != (iconv_t)(-1))
iconv_close (cd1);
- iconv_close (cd);
+ if (cd != (iconv_t)(-1))
+ iconv_close (cd);
if (result != *resultp && result != NULL)
free (result);
errno = saved_errno;
/* Return -1, but free the allocated memory, and while doing
that, preserve the errno from iconv_close. */
int saved_errno = errno;
- iconv_close (cd);
+ if (cd != (iconv_t)(-1))
+ iconv_close (cd);
if (result != *resultp && result != NULL)
free (result);
errno = saved_errno;
return -1;
}
- if (iconv_close (cd) < 0)
+ if (cd != (iconv_t)(-1) && iconv_close (cd) < 0)
{
/* Return -1, but free the allocated memory, and while doing
that, preserve the errno from iconv_close. */
# endif
cd = iconv_open (to_codeset, from_codeset);
- if (cd == (iconv_t)(-1))
- return NULL;
- if (c_strcasecmp (from_codeset, "UTF-8") == 0)
+ if (STRCASEEQ (from_codeset, "UTF-8", 'U','T','F','-','8',0,0,0,0))
cd1 = (iconv_t)(-1);
else
{
if (cd1 == (iconv_t)(-1))
{
int saved_errno = errno;
- iconv_close (cd);
+ if (cd != (iconv_t)(-1))
+ iconv_close (cd);
errno = saved_errno;
return NULL;
}
}
- if (c_strcasecmp (to_codeset, "UTF-8") == 0)
+ if (STRCASEEQ (to_codeset, "UTF-8", 'U','T','F','-','8',0,0,0,0))
cd2 = (iconv_t)(-1);
else
{
int saved_errno = errno;
if (cd1 != (iconv_t)(-1))
iconv_close (cd1);
- iconv_close (cd);
+ if (cd != (iconv_t)(-1))
+ iconv_close (cd);
errno = saved_errno;
return NULL;
}
iconv_close (cd2);
if (cd1 != (iconv_t)(-1))
iconv_close (cd1);
- iconv_close (cd);
+ if (cd != (iconv_t)(-1))
+ iconv_close (cd);
errno = saved_errno;
}
else
int saved_errno = errno;
if (cd1 != (iconv_t)(-1))
iconv_close (cd1);
- iconv_close (cd);
+ if (cd != (iconv_t)(-1))
+ iconv_close (cd);
free (result);
errno = saved_errno;
return NULL;
/* Return NULL, but free the allocated memory, and while doing
that, preserve the errno from iconv_close. */
int saved_errno = errno;
- iconv_close (cd);
+ if (cd != (iconv_t)(-1))
+ iconv_close (cd);
free (result);
errno = saved_errno;
return NULL;
}
- if (iconv_close (cd) < 0)
+ if (cd != (iconv_t)(-1) && iconv_close (cd) < 0)
{
/* Return NULL, but free the allocated memory, and while doing
that, preserve the errno from iconv_close. */
projects / gnulib.git / blobdiff