- { "Success", /* REG_NOERROR */
- "No match", /* REG_NOMATCH */
- "Invalid regular expression", /* REG_BADPAT */
- "Invalid collation character", /* REG_ECOLLATE */
- "Invalid character class name", /* REG_ECTYPE */
- "Trailing backslash", /* REG_EESCAPE */
- "Invalid back reference", /* REG_ESUBREG */
- "Unmatched [ or [^", /* REG_EBRACK */
- "Unmatched ( or \\(", /* REG_EPAREN */
- "Unmatched \\{", /* REG_EBRACE */
- "Invalid content of \\{\\}", /* REG_BADBR */
- "Invalid range end", /* REG_ERANGE */
- "Memory exhausted", /* REG_ESPACE */
- "Invalid preceding regular expression", /* REG_BADRPT */
- "Premature end of regular expression", /* REG_EEND */
- "Regular expression too big", /* REG_ESIZE */
- "Unmatched ) or \\)", /* REG_ERPAREN */
+ {
+ gettext_noop ("Success"), /* REG_NOERROR */
+ gettext_noop ("No match"), /* REG_NOMATCH */
+ gettext_noop ("Invalid regular expression"), /* REG_BADPAT */
+ gettext_noop ("Invalid collation character"), /* REG_ECOLLATE */
+ gettext_noop ("Invalid character class name"), /* REG_ECTYPE */
+ gettext_noop ("Trailing backslash"), /* REG_EESCAPE */
+ gettext_noop ("Invalid back reference"), /* REG_ESUBREG */
+ gettext_noop ("Unmatched [ or [^"), /* REG_EBRACK */
+ gettext_noop ("Unmatched ( or \\("), /* REG_EPAREN */
+ gettext_noop ("Unmatched \\{"), /* REG_EBRACE */
+ gettext_noop ("Invalid content of \\{\\}"), /* REG_BADBR */
+ gettext_noop ("Invalid range end"), /* REG_ERANGE */
+ gettext_noop ("Memory exhausted"), /* REG_ESPACE */
+ gettext_noop ("Invalid preceding regular expression"), /* REG_BADRPT */
+ gettext_noop ("Premature end of regular expression"), /* REG_EEND */
+ gettext_noop ("Regular expression too big"), /* REG_ESIZE */
+ gettext_noop ("Unmatched ) or \\)"), /* REG_ERPAREN */
- case '?':
- if ((syntax & RE_BK_PLUS_QM)
- || (syntax & RE_LIMITED_OPS))
- goto normal_char;
- handle_plus:
- case '*':
- /* If there is no previous pattern... */
- if (!laststart)
- {
- if (syntax & RE_CONTEXT_INVALID_OPS)
- FREE_STACK_RETURN (REG_BADRPT);
- else if (!(syntax & RE_CONTEXT_INDEP_OPS))
- goto normal_char;
- }
-
- {
- /* Are we optimizing this jump? */
- boolean keep_string_p = false;
-
- /* 1 means zero (many) matches is allowed. */
- char zero_times_ok = 0, many_times_ok = 0;
-
- /* If there is a sequence of repetition chars, collapse it
- down to just one (the right one). We can't combine
- interval operators with these because of, e.g., `a{2}*',
- which should only match an even number of `a's. */
-
- for (;;)
- {
- zero_times_ok |= c != '+';
- many_times_ok |= c != '?';
-
- if (p == pend)
- break;
-
- PATFETCH (c);
-
- if (c == '*'
- || (!(syntax & RE_BK_PLUS_QM) && (c == '+' || c == '?')))
- ;
-
- else if (syntax & RE_BK_PLUS_QM && c == '\\')
- {
- if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
-
- PATFETCH (c1);
- if (!(c1 == '+' || c1 == '?'))
- {
- PATUNFETCH;
- PATUNFETCH;
- break;
- }
-
- c = c1;
- }
- else
- {
- PATUNFETCH;
- break;
- }
-
- /* If we get here, we found another repeat character. */
- }
-
- /* Star, etc. applied to an empty pattern is equivalent
- to an empty pattern. */
- if (!laststart)
- break;
-
- /* Now we know whether or not zero matches is allowed
- and also whether or not two or more matches is allowed. */
- if (many_times_ok)
- { /* More than one repetition is allowed, so put in at the
- end a backward relative jump from `b' to before the next
- jump we're going to put in below (which jumps from
- laststart to after this jump).
-
- But if we are at the `*' in the exact sequence `.*\n',
- insert an unconditional jump backwards to the .,
- instead of the beginning of the loop. This way we only
- push a failure point once, instead of every time
- through the loop. */
- assert (p - 1 > pattern);
-
- /* Allocate the space for the jump. */
- GET_BUFFER_SPACE (3);
-
- /* We know we are not at the first character of the pattern,
- because laststart was nonzero. And we've already
- incremented `p', by the way, to be the character after
- the `*'. Do we have to do something analogous here
- for null bytes, because of RE_DOT_NOT_NULL? */
- if (TRANSLATE (*(p - 2)) == TRANSLATE ('.')
+ case '?':
+ if ((syntax & RE_BK_PLUS_QM)
+ || (syntax & RE_LIMITED_OPS))
+ goto normal_char;
+ handle_plus:
+ case '*':
+ /* If there is no previous pattern... */
+ if (!laststart)
+ {
+ if (syntax & RE_CONTEXT_INVALID_OPS)
+ FREE_STACK_RETURN (REG_BADRPT);
+ else if (!(syntax & RE_CONTEXT_INDEP_OPS))
+ goto normal_char;
+ }
+
+ {
+ /* Are we optimizing this jump? */
+ boolean keep_string_p = false;
+
+ /* 1 means zero (many) matches is allowed. */
+ char zero_times_ok = 0, many_times_ok = 0;
+
+ /* If there is a sequence of repetition chars, collapse it
+ down to just one (the right one). We can't combine
+ interval operators with these because of, e.g., `a{2}*',
+ which should only match an even number of `a's. */
+
+ for (;;)
+ {
+ zero_times_ok |= c != '+';
+ many_times_ok |= c != '?';
+
+ if (p == pend)
+ break;
+
+ PATFETCH (c);
+
+ if (c == '*'
+ || (!(syntax & RE_BK_PLUS_QM) && (c == '+' || c == '?')))
+ ;
+
+ else if (syntax & RE_BK_PLUS_QM && c == '\\')
+ {
+ if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
+
+ PATFETCH (c1);
+ if (!(c1 == '+' || c1 == '?'))
+ {
+ PATUNFETCH;
+ PATUNFETCH;
+ break;
+ }
+
+ c = c1;
+ }
+ else
+ {
+ PATUNFETCH;
+ break;
+ }
+
+ /* If we get here, we found another repeat character. */
+ }
+
+ /* Star, etc. applied to an empty pattern is equivalent
+ to an empty pattern. */
+ if (!laststart)
+ break;
+
+ /* Now we know whether or not zero matches is allowed
+ and also whether or not two or more matches is allowed. */
+ if (many_times_ok)
+ { /* More than one repetition is allowed, so put in at the
+ end a backward relative jump from `b' to before the next
+ jump we're going to put in below (which jumps from
+ laststart to after this jump).
+
+ But if we are at the `*' in the exact sequence `.*\n',
+ insert an unconditional jump backwards to the .,
+ instead of the beginning of the loop. This way we only
+ push a failure point once, instead of every time
+ through the loop. */
+ assert (p - 1 > pattern);
+
+ /* Allocate the space for the jump. */
+ GET_BUFFER_SPACE (3);
+
+ /* We know we are not at the first character of the pattern,
+ because laststart was nonzero. And we've already
+ incremented `p', by the way, to be the character after
+ the `*'. Do we have to do something analogous here
+ for null bytes, because of RE_DOT_NOT_NULL? */
+ if (TRANSLATE (*(p - 2)) == TRANSLATE ('.')
- laststart = b;
-
- /* We test `*p == '^' twice, instead of using an if
- statement, so we only need one BUF_PUSH. */
- BUF_PUSH (*p == '^' ? charset_not : charset);
- if (*p == '^')
- p++;
-
- /* Remember the first position in the bracket expression. */
- p1 = p;
-
- /* Push the number of bytes in the bitmap. */
- BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH);
-
- /* Clear the whole map. */
- bzero (b, (1 << BYTEWIDTH) / BYTEWIDTH);
-
- /* charset_not matches newline according to a syntax bit. */
- if ((re_opcode_t) b[-2] == charset_not
- && (syntax & RE_HAT_LISTS_NOT_NEWLINE))
- SET_LIST_BIT ('\n');
-
- /* Read in characters and ranges, setting map bits. */
- for (;;)
- {
- if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
-
- PATFETCH (c);
-
- /* \ might escape characters inside [...] and [^...]. */
- if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\')
- {
- if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
-
- PATFETCH (c1);
- SET_LIST_BIT (c1);
- continue;
- }
-
- /* Could be the end of the bracket expression. If it's
- not (i.e., when the bracket expression is `[]' so
- far), the ']' character bit gets set way below. */
- if (c == ']' && p != p1 + 1)
- break;
-
- /* Look ahead to see if it's a range when the last thing
- was a character class. */
- if (had_char_class && c == '-' && *p != ']')
- FREE_STACK_RETURN (REG_ERANGE);
-
- /* Look ahead to see if it's a range when the last thing
- was a character: if this is a hyphen not at the
- beginning or the end of a list, then it's the range
- operator. */
- if (c == '-'
- && !(p - 2 >= pattern && p[-2] == '[')
- && !(p - 3 >= pattern && p[-3] == '[' && p[-2] == '^')
- && *p != ']')
- {
- reg_errcode_t ret
- = compile_range (&p, pend, translate, syntax, b);
- if (ret != REG_NOERROR) FREE_STACK_RETURN (ret);
- }
-
- else if (p[0] == '-' && p[1] != ']')
- { /* This handles ranges made up of characters only. */
- reg_errcode_t ret;
+ laststart = b;
+
+ /* We test `*p == '^' twice, instead of using an if
+ statement, so we only need one BUF_PUSH. */
+ BUF_PUSH (*p == '^' ? charset_not : charset);
+ if (*p == '^')
+ p++;
+
+ /* Remember the first position in the bracket expression. */
+ p1 = p;
+
+ /* Push the number of bytes in the bitmap. */
+ BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH);
+
+ /* Clear the whole map. */
+ bzero (b, (1 << BYTEWIDTH) / BYTEWIDTH);
+
+ /* charset_not matches newline according to a syntax bit. */
+ if ((re_opcode_t) b[-2] == charset_not
+ && (syntax & RE_HAT_LISTS_NOT_NEWLINE))
+ SET_LIST_BIT ('\n');
+
+ /* Read in characters and ranges, setting map bits. */
+ for (;;)
+ {
+ if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
+
+ PATFETCH (c);
+
+ /* \ might escape characters inside [...] and [^...]. */
+ if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\')
+ {
+ if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
+
+ PATFETCH (c1);
+ SET_LIST_BIT (c1);
+ continue;
+ }
+
+ /* Could be the end of the bracket expression. If it's
+ not (i.e., when the bracket expression is `[]' so
+ far), the ']' character bit gets set way below. */
+ if (c == ']' && p != p1 + 1)
+ break;
+
+ /* Look ahead to see if it's a range when the last thing
+ was a character class. */
+ if (had_char_class && c == '-' && *p != ']')
+ FREE_STACK_RETURN (REG_ERANGE);
+
+ /* Look ahead to see if it's a range when the last thing
+ was a character: if this is a hyphen not at the
+ beginning or the end of a list, then it's the range
+ operator. */
+ if (c == '-'
+ && !(p - 2 >= pattern && p[-2] == '[')
+ && !(p - 3 >= pattern && p[-3] == '[' && p[-2] == '^')
+ && *p != ']')
+ {
+ reg_errcode_t ret
+ = compile_range (&p, pend, translate, syntax, b);
+ if (ret != REG_NOERROR) FREE_STACK_RETURN (ret);
+ }
+
+ else if (p[0] == '-' && p[1] != ']')
+ { /* This handles ranges made up of characters only. */
+ reg_errcode_t ret;
- PATFETCH (c1);
-
- ret = compile_range (&p, pend, translate, syntax, b);
- if (ret != REG_NOERROR) FREE_STACK_RETURN (ret);
- }
-
- /* See if we're at the beginning of a possible character
- class. */
-
- else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == ':')
- { /* Leave room for the null. */
- char str[CHAR_CLASS_MAX_LENGTH + 1];
-
- PATFETCH (c);
- c1 = 0;
-
- /* If pattern is `[[:'. */
- if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
-
- for (;;)
- {
- PATFETCH (c);
- if (c == ':' || c == ']' || p == pend
- || c1 == CHAR_CLASS_MAX_LENGTH)
- break;
- str[c1++] = c;
- }
- str[c1] = '\0';
-
- /* If isn't a word bracketed by `[:' and:`]':
- undo the ending character, the letters, and leave
- the leading `:' and `[' (but set bits for them). */
- if (c == ':' && *p == ']')
- {
- int ch;
- boolean is_alnum = STREQ (str, "alnum");
- boolean is_alpha = STREQ (str, "alpha");
- boolean is_blank = STREQ (str, "blank");
- boolean is_cntrl = STREQ (str, "cntrl");
- boolean is_digit = STREQ (str, "digit");
- boolean is_graph = STREQ (str, "graph");
- boolean is_lower = STREQ (str, "lower");
- boolean is_print = STREQ (str, "print");
- boolean is_punct = STREQ (str, "punct");
- boolean is_space = STREQ (str, "space");
- boolean is_upper = STREQ (str, "upper");
- boolean is_xdigit = STREQ (str, "xdigit");
-
- if (!IS_CHAR_CLASS (str))
+ PATFETCH (c1);
+
+ ret = compile_range (&p, pend, translate, syntax, b);
+ if (ret != REG_NOERROR) FREE_STACK_RETURN (ret);
+ }
+
+ /* See if we're at the beginning of a possible character
+ class. */
+
+ else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == ':')
+ { /* Leave room for the null. */
+ char str[CHAR_CLASS_MAX_LENGTH + 1];
+
+ PATFETCH (c);
+ c1 = 0;
+
+ /* If pattern is `[[:'. */
+ if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
+
+ for (;;)
+ {
+ PATFETCH (c);
+ if (c == ':' || c == ']' || p == pend
+ || c1 == CHAR_CLASS_MAX_LENGTH)
+ break;
+ str[c1++] = c;
+ }
+ str[c1] = '\0';
+
+ /* If isn't a word bracketed by `[:' and:`]':
+ undo the ending character, the letters, and leave
+ the leading `:' and `[' (but set bits for them). */
+ if (c == ':' && *p == ']')
+ {
+ int ch;
+ boolean is_alnum = STREQ (str, "alnum");
+ boolean is_alpha = STREQ (str, "alpha");
+ boolean is_blank = STREQ (str, "blank");
+ boolean is_cntrl = STREQ (str, "cntrl");
+ boolean is_digit = STREQ (str, "digit");
+ boolean is_graph = STREQ (str, "graph");
+ boolean is_lower = STREQ (str, "lower");
+ boolean is_print = STREQ (str, "print");
+ boolean is_punct = STREQ (str, "punct");
+ boolean is_space = STREQ (str, "space");
+ boolean is_upper = STREQ (str, "upper");
+ boolean is_xdigit = STREQ (str, "xdigit");
+
+ if (!IS_CHAR_CLASS (str))
- if (syntax & RE_NO_BK_VBAR)
- goto handle_alt;
- else
- goto normal_char;
-
-
- case '{':
- if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES)
- goto handle_interval;
- else
- goto normal_char;
-
-
- case '\\':
- if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
-
- /* Do not translate the character after the \, so that we can
- distinguish, e.g., \B from \b, even if we normally would
- translate, e.g., B to b. */
- PATFETCH_RAW (c);
-
- switch (c)
- {
- case '(':
- if (syntax & RE_NO_BK_PARENS)
- goto normal_backslash;
-
- handle_open:
- bufp->re_nsub++;
- regnum++;
-
- if (COMPILE_STACK_FULL)
- {
- RETALLOC (compile_stack.stack, compile_stack.size << 1,
- compile_stack_elt_t);
- if (compile_stack.stack == NULL) return REG_ESPACE;
-
- compile_stack.size <<= 1;
- }
-
- /* These are the values to restore when we hit end of this
- group. They are all relative offsets, so that if the
- whole pattern moves because of realloc, they will still
- be valid. */
- COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer;
- COMPILE_STACK_TOP.fixup_alt_jump
- = fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0;
- COMPILE_STACK_TOP.laststart_offset = b - bufp->buffer;
- COMPILE_STACK_TOP.regnum = regnum;
-
- /* We will eventually replace the 0 with the number of
- groups inner to this one. But do not push a
- start_memory for groups beyond the last one we can
- represent in the compiled pattern. */
- if (regnum <= MAX_REGNUM)
- {
- COMPILE_STACK_TOP.inner_group_offset = b - bufp->buffer + 2;
- BUF_PUSH_3 (start_memory, regnum, 0);
- }
-
- compile_stack.avail++;
-
- fixup_alt_jump = 0;
- laststart = 0;
- begalt = b;
+ if (syntax & RE_NO_BK_VBAR)
+ goto handle_alt;
+ else
+ goto normal_char;
+
+
+ case '{':
+ if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES)
+ goto handle_interval;
+ else
+ goto normal_char;
+
+
+ case '\\':
+ if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
+
+ /* Do not translate the character after the \, so that we can
+ distinguish, e.g., \B from \b, even if we normally would
+ translate, e.g., B to b. */
+ PATFETCH_RAW (c);
+
+ switch (c)
+ {
+ case '(':
+ if (syntax & RE_NO_BK_PARENS)
+ goto normal_backslash;
+
+ handle_open:
+ bufp->re_nsub++;
+ regnum++;
+
+ if (COMPILE_STACK_FULL)
+ {
+ RETALLOC (compile_stack.stack, compile_stack.size << 1,
+ compile_stack_elt_t);
+ if (compile_stack.stack == NULL) return REG_ESPACE;
+
+ compile_stack.size <<= 1;
+ }
+
+ /* These are the values to restore when we hit end of this
+ group. They are all relative offsets, so that if the
+ whole pattern moves because of realloc, they will still
+ be valid. */
+ COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer;
+ COMPILE_STACK_TOP.fixup_alt_jump
+ = fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0;
+ COMPILE_STACK_TOP.laststart_offset = b - bufp->buffer;
+ COMPILE_STACK_TOP.regnum = regnum;
+
+ /* We will eventually replace the 0 with the number of
+ groups inner to this one. But do not push a
+ start_memory for groups beyond the last one we can
+ represent in the compiled pattern. */
+ if (regnum <= MAX_REGNUM)
+ {
+ COMPILE_STACK_TOP.inner_group_offset = b - bufp->buffer + 2;
+ BUF_PUSH_3 (start_memory, regnum, 0);
+ }
+
+ compile_stack.avail++;
+
+ fixup_alt_jump = 0;
+ laststart = 0;
+ begalt = b;
- break;
-
-
- case ')':
- if (syntax & RE_NO_BK_PARENS) goto normal_backslash;
-
- if (COMPILE_STACK_EMPTY)
- if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD)
- goto normal_backslash;
- else
- FREE_STACK_RETURN (REG_ERPAREN);
-
- handle_close:
- if (fixup_alt_jump)
- { /* Push a dummy failure point at the end of the
- alternative for a possible future
- `pop_failure_jump' to pop. See comments at
- `push_dummy_failure' in `re_match_2'. */
- BUF_PUSH (push_dummy_failure);
-
- /* We allocated space for this jump when we assigned
- to `fixup_alt_jump', in the `handle_alt' case below. */
- STORE_JUMP (jump_past_alt, fixup_alt_jump, b - 1);
- }
-
- /* See similar code for backslashed left paren above. */
- if (COMPILE_STACK_EMPTY)
- if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD)
- goto normal_char;
- else
- FREE_STACK_RETURN (REG_ERPAREN);
-
- /* Since we just checked for an empty stack above, this
- ``can't happen''. */
- assert (compile_stack.avail != 0);
- {
- /* We don't just want to restore into `regnum', because
- later groups should continue to be numbered higher,
- as in `(ab)c(de)' -- the second group is #2. */
- regnum_t this_group_regnum;
-
- compile_stack.avail--;
- begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset;
- fixup_alt_jump
- = COMPILE_STACK_TOP.fixup_alt_jump
- ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1
- : 0;
- laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset;
- this_group_regnum = COMPILE_STACK_TOP.regnum;
+ break;
+
+
+ case ')':
+ if (syntax & RE_NO_BK_PARENS) goto normal_backslash;
+
+ if (COMPILE_STACK_EMPTY)
+ if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD)
+ goto normal_backslash;
+ else
+ FREE_STACK_RETURN (REG_ERPAREN);
+
+ handle_close:
+ if (fixup_alt_jump)
+ { /* Push a dummy failure point at the end of the
+ alternative for a possible future
+ `pop_failure_jump' to pop. See comments at
+ `push_dummy_failure' in `re_match_2'. */
+ BUF_PUSH (push_dummy_failure);
+
+ /* We allocated space for this jump when we assigned
+ to `fixup_alt_jump', in the `handle_alt' case below. */
+ STORE_JUMP (jump_past_alt, fixup_alt_jump, b - 1);
+ }
+
+ /* See similar code for backslashed left paren above. */
+ if (COMPILE_STACK_EMPTY)
+ if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD)
+ goto normal_char;
+ else
+ FREE_STACK_RETURN (REG_ERPAREN);
+
+ /* Since we just checked for an empty stack above, this
+ ``can't happen''. */
+ assert (compile_stack.avail != 0);
+ {
+ /* We don't just want to restore into `regnum', because
+ later groups should continue to be numbered higher,
+ as in `(ab)c(de)' -- the second group is #2. */
+ regnum_t this_group_regnum;
+
+ compile_stack.avail--;
+ begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset;
+ fixup_alt_jump
+ = COMPILE_STACK_TOP.fixup_alt_jump
+ ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1
+ : 0;
+ laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset;
+ this_group_regnum = COMPILE_STACK_TOP.regnum;
- /* We're at the end of the group, so now we know how many
- groups were inside this one. */
- if (this_group_regnum <= MAX_REGNUM)
- {
- unsigned char *inner_group_loc
- = bufp->buffer + COMPILE_STACK_TOP.inner_group_offset;
-
- *inner_group_loc = regnum - this_group_regnum;
- BUF_PUSH_3 (stop_memory, this_group_regnum,
- regnum - this_group_regnum);
- }
- }
- break;
-
-
- case '|': /* `\|'. */
- if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR)
- goto normal_backslash;
- handle_alt:
- if (syntax & RE_LIMITED_OPS)
- goto normal_char;
-
- /* Insert before the previous alternative a jump which
- jumps to this alternative if the former fails. */
- GET_BUFFER_SPACE (3);
- INSERT_JUMP (on_failure_jump, begalt, b + 6);
- pending_exact = 0;
- b += 3;
-
- /* The alternative before this one has a jump after it
- which gets executed if it gets matched. Adjust that
- jump so it will jump to this alternative's analogous
- jump (put in below, which in turn will jump to the next
- (if any) alternative's such jump, etc.). The last such
- jump jumps to the correct final destination. A picture:
- _____ _____
- | | | |
- | v | v
- a | b | c
-
- If we are at `b', then fixup_alt_jump right now points to a
- three-byte space after `a'. We'll put in the jump, set
- fixup_alt_jump to right after `b', and leave behind three
- bytes which we'll fill in when we get to after `c'. */
-
- if (fixup_alt_jump)
- STORE_JUMP (jump_past_alt, fixup_alt_jump, b);
-
- /* Mark and leave space for a jump after this alternative,
- to be filled in later either by next alternative or
- when know we're at the end of a series of alternatives. */
- fixup_alt_jump = b;
- GET_BUFFER_SPACE (3);
- b += 3;
-
- laststart = 0;
- begalt = b;
- break;
-
-
- case '{':
- /* If \{ is a literal. */
- if (!(syntax & RE_INTERVALS)
- /* If we're at `\{' and it's not the open-interval
- operator. */
- || ((syntax & RE_INTERVALS) && (syntax & RE_NO_BK_BRACES))
- || (p - 2 == pattern && p == pend))
- goto normal_backslash;
-
- handle_interval:
- {
- /* If got here, then the syntax allows intervals. */
-
- /* At least (most) this many matches must be made. */
- int lower_bound = -1, upper_bound = -1;
-
- beg_interval = p - 1;
-
- if (p == pend)
- {
- if (syntax & RE_NO_BK_BRACES)
- goto unfetch_interval;
- else
- FREE_STACK_RETURN (REG_EBRACE);
- }
-
- GET_UNSIGNED_NUMBER (lower_bound);
-
- if (c == ',')
- {
- GET_UNSIGNED_NUMBER (upper_bound);
- if (upper_bound < 0) upper_bound = RE_DUP_MAX;
- }
- else
- /* Interval such as `{1}' => match exactly once. */
- upper_bound = lower_bound;
-
- if (lower_bound < 0 || upper_bound > RE_DUP_MAX
- || lower_bound > upper_bound)
- {
- if (syntax & RE_NO_BK_BRACES)
- goto unfetch_interval;
- else
- FREE_STACK_RETURN (REG_BADBR);
- }
-
- if (!(syntax & RE_NO_BK_BRACES))
- {
- if (c != '\\') FREE_STACK_RETURN (REG_EBRACE);
-
- PATFETCH (c);
- }
-
- if (c != '}')
- {
- if (syntax & RE_NO_BK_BRACES)
- goto unfetch_interval;
- else
- FREE_STACK_RETURN (REG_BADBR);
- }
-
- /* We just parsed a valid interval. */
-
- /* If it's invalid to have no preceding re. */
- if (!laststart)
- {
- if (syntax & RE_CONTEXT_INVALID_OPS)
- FREE_STACK_RETURN (REG_BADRPT);
- else if (syntax & RE_CONTEXT_INDEP_OPS)
- laststart = b;
- else
- goto unfetch_interval;
- }
-
- /* If the upper bound is zero, don't want to succeed at
- all; jump from `laststart' to `b + 3', which will be
- the end of the buffer after we insert the jump. */
- if (upper_bound == 0)
- {
- GET_BUFFER_SPACE (3);
- INSERT_JUMP (jump, laststart, b + 3);
- b += 3;
- }
-
- /* Otherwise, we have a nontrivial interval. When
- we're all done, the pattern will look like:
- set_number_at <jump count> <upper bound>
- set_number_at <succeed_n count> <lower bound>
- succeed_n <after jump addr> <succeed_n count>
- <body of loop>
- jump_n <succeed_n addr> <jump count>
- (The upper bound and `jump_n' are omitted if
- `upper_bound' is 1, though.) */
- else
- { /* If the upper bound is > 1, we need to insert
- more at the end of the loop. */
- unsigned nbytes = 10 + (upper_bound > 1) * 10;
-
- GET_BUFFER_SPACE (nbytes);
-
- /* Initialize lower bound of the `succeed_n', even
- though it will be set during matching by its
- attendant `set_number_at' (inserted next),
- because `re_compile_fastmap' needs to know.
- Jump to the `jump_n' we might insert below. */
- INSERT_JUMP2 (succeed_n, laststart,
- b + 5 + (upper_bound > 1) * 5,
- lower_bound);
- b += 5;
-
- /* Code to initialize the lower bound. Insert
- before the `succeed_n'. The `5' is the last two
- bytes of this `set_number_at', plus 3 bytes of
- the following `succeed_n'. */
- insert_op2 (set_number_at, laststart, 5, lower_bound, b);
- b += 5;
-
- if (upper_bound > 1)
- { /* More than one repetition is allowed, so
- append a backward jump to the `succeed_n'
- that starts this interval.
-
- When we've reached this during matching,
- we'll have matched the interval once, so
- jump back only `upper_bound - 1' times. */
- STORE_JUMP2 (jump_n, b, laststart + 5,
- upper_bound - 1);
- b += 5;
-
- /* The location we want to set is the second
- parameter of the `jump_n'; that is `b-2' as
- an absolute address. `laststart' will be
- the `set_number_at' we're about to insert;
- `laststart+3' the number to set, the source
- for the relative address. But we are
- inserting into the middle of the pattern --
- so everything is getting moved up by 5.
- Conclusion: (b - 2) - (laststart + 3) + 5,
- i.e., b - laststart.
-
- We insert this at the beginning of the loop
- so that if we fail during matching, we'll
- reinitialize the bounds. */
- insert_op2 (set_number_at, laststart, b - laststart,
- upper_bound - 1, b);
- b += 5;
- }
- }
- pending_exact = 0;
- beg_interval = NULL;
- }
- break;
-
- unfetch_interval:
- /* If an invalid interval, match the characters as literals. */
- assert (beg_interval);
- p = beg_interval;
- beg_interval = NULL;
-
- /* normal_char and normal_backslash need `c'. */
- PATFETCH (c);
-
- if (!(syntax & RE_NO_BK_BRACES))
- {
- if (p > pattern && p[-1] == '\\')
- goto normal_backslash;
- }
- goto normal_char;
+ /* We're at the end of the group, so now we know how many
+ groups were inside this one. */
+ if (this_group_regnum <= MAX_REGNUM)
+ {
+ unsigned char *inner_group_loc
+ = bufp->buffer + COMPILE_STACK_TOP.inner_group_offset;
+
+ *inner_group_loc = regnum - this_group_regnum;
+ BUF_PUSH_3 (stop_memory, this_group_regnum,
+ regnum - this_group_regnum);
+ }
+ }
+ break;
+
+
+ case '|': /* `\|'. */
+ if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR)
+ goto normal_backslash;
+ handle_alt:
+ if (syntax & RE_LIMITED_OPS)
+ goto normal_char;
+
+ /* Insert before the previous alternative a jump which
+ jumps to this alternative if the former fails. */
+ GET_BUFFER_SPACE (3);
+ INSERT_JUMP (on_failure_jump, begalt, b + 6);
+ pending_exact = 0;
+ b += 3;
+
+ /* The alternative before this one has a jump after it
+ which gets executed if it gets matched. Adjust that
+ jump so it will jump to this alternative's analogous
+ jump (put in below, which in turn will jump to the next
+ (if any) alternative's such jump, etc.). The last such
+ jump jumps to the correct final destination. A picture:
+ _____ _____
+ | | | |
+ | v | v
+ a | b | c
+
+ If we are at `b', then fixup_alt_jump right now points to a
+ three-byte space after `a'. We'll put in the jump, set
+ fixup_alt_jump to right after `b', and leave behind three
+ bytes which we'll fill in when we get to after `c'. */
+
+ if (fixup_alt_jump)
+ STORE_JUMP (jump_past_alt, fixup_alt_jump, b);
+
+ /* Mark and leave space for a jump after this alternative,
+ to be filled in later either by next alternative or
+ when know we're at the end of a series of alternatives. */
+ fixup_alt_jump = b;
+ GET_BUFFER_SPACE (3);
+ b += 3;
+
+ laststart = 0;
+ begalt = b;
+ break;
+
+
+ case '{':
+ /* If \{ is a literal. */
+ if (!(syntax & RE_INTERVALS)
+ /* If we're at `\{' and it's not the open-interval
+ operator. */
+ || ((syntax & RE_INTERVALS) && (syntax & RE_NO_BK_BRACES))
+ || (p - 2 == pattern && p == pend))
+ goto normal_backslash;
+
+ handle_interval:
+ {
+ /* If got here, then the syntax allows intervals. */
+
+ /* At least (most) this many matches must be made. */
+ int lower_bound = -1, upper_bound = -1;
+
+ beg_interval = p - 1;
+
+ if (p == pend)
+ {
+ if (syntax & RE_NO_BK_BRACES)
+ goto unfetch_interval;
+ else
+ FREE_STACK_RETURN (REG_EBRACE);
+ }
+
+ GET_UNSIGNED_NUMBER (lower_bound);
+
+ if (c == ',')
+ {
+ GET_UNSIGNED_NUMBER (upper_bound);
+ if (upper_bound < 0) upper_bound = RE_DUP_MAX;
+ }
+ else
+ /* Interval such as `{1}' => match exactly once. */
+ upper_bound = lower_bound;
+
+ if (lower_bound < 0 || upper_bound > RE_DUP_MAX
+ || lower_bound > upper_bound)
+ {
+ if (syntax & RE_NO_BK_BRACES)
+ goto unfetch_interval;
+ else
+ FREE_STACK_RETURN (REG_BADBR);
+ }
+
+ if (!(syntax & RE_NO_BK_BRACES))
+ {
+ if (c != '\\') FREE_STACK_RETURN (REG_EBRACE);
+
+ PATFETCH (c);
+ }
+
+ if (c != '}')
+ {
+ if (syntax & RE_NO_BK_BRACES)
+ goto unfetch_interval;
+ else
+ FREE_STACK_RETURN (REG_BADBR);
+ }
+
+ /* We just parsed a valid interval. */
+
+ /* If it's invalid to have no preceding re. */
+ if (!laststart)
+ {
+ if (syntax & RE_CONTEXT_INVALID_OPS)
+ FREE_STACK_RETURN (REG_BADRPT);
+ else if (syntax & RE_CONTEXT_INDEP_OPS)
+ laststart = b;
+ else
+ goto unfetch_interval;
+ }
+
+ /* If the upper bound is zero, don't want to succeed at
+ all; jump from `laststart' to `b + 3', which will be
+ the end of the buffer after we insert the jump. */
+ if (upper_bound == 0)
+ {
+ GET_BUFFER_SPACE (3);
+ INSERT_JUMP (jump, laststart, b + 3);
+ b += 3;
+ }
+
+ /* Otherwise, we have a nontrivial interval. When
+ we're all done, the pattern will look like:
+ set_number_at <jump count> <upper bound>
+ set_number_at <succeed_n count> <lower bound>
+ succeed_n <after jump addr> <succeed_n count>
+ <body of loop>
+ jump_n <succeed_n addr> <jump count>
+ (The upper bound and `jump_n' are omitted if
+ `upper_bound' is 1, though.) */
+ else
+ { /* If the upper bound is > 1, we need to insert
+ more at the end of the loop. */
+ unsigned nbytes = 10 + (upper_bound > 1) * 10;
+
+ GET_BUFFER_SPACE (nbytes);
+
+ /* Initialize lower bound of the `succeed_n', even
+ though it will be set during matching by its
+ attendant `set_number_at' (inserted next),
+ because `re_compile_fastmap' needs to know.
+ Jump to the `jump_n' we might insert below. */
+ INSERT_JUMP2 (succeed_n, laststart,
+ b + 5 + (upper_bound > 1) * 5,
+ lower_bound);
+ b += 5;
+
+ /* Code to initialize the lower bound. Insert
+ before the `succeed_n'. The `5' is the last two
+ bytes of this `set_number_at', plus 3 bytes of
+ the following `succeed_n'. */
+ insert_op2 (set_number_at, laststart, 5, lower_bound, b);
+ b += 5;
+
+ if (upper_bound > 1)
+ { /* More than one repetition is allowed, so
+ append a backward jump to the `succeed_n'
+ that starts this interval.
+
+ When we've reached this during matching,
+ we'll have matched the interval once, so
+ jump back only `upper_bound - 1' times. */
+ STORE_JUMP2 (jump_n, b, laststart + 5,
+ upper_bound - 1);
+ b += 5;
+
+ /* The location we want to set is the second
+ parameter of the `jump_n'; that is `b-2' as
+ an absolute address. `laststart' will be
+ the `set_number_at' we're about to insert;
+ `laststart+3' the number to set, the source
+ for the relative address. But we are
+ inserting into the middle of the pattern --
+ so everything is getting moved up by 5.
+ Conclusion: (b - 2) - (laststart + 3) + 5,
+ i.e., b - laststart.
+
+ We insert this at the beginning of the loop
+ so that if we fail during matching, we'll
+ reinitialize the bounds. */
+ insert_op2 (set_number_at, laststart, b - laststart,
+ upper_bound - 1, b);
+ b += 5;
+ }
+ }
+ pending_exact = 0;
+ beg_interval = NULL;
+ }
+ break;
+
+ unfetch_interval:
+ /* If an invalid interval, match the characters as literals. */
+ assert (beg_interval);
+ p = beg_interval;
+ beg_interval = NULL;
+
+ /* normal_char and normal_backslash need `c'. */
+ PATFETCH (c);
+
+ if (!(syntax & RE_NO_BK_BRACES))
+ {
+ if (p > pattern && p[-1] == '\\')
+ goto normal_backslash;
+ }
+ goto normal_char;
- goto fail;
-
-
- /* on_failure_keep_string_jump is used to optimize `.*\n'. It
- pushes NULL as the value for the string on the stack. Then
- `pop_failure_point' will keep the current value for the
- string, instead of restoring it. To see why, consider
- matching `foo\nbar' against `.*\n'. The .* matches the foo;
- then the . fails against the \n. But the next thing we want
- to do is match the \n against the \n; if we restored the
- string value, we would be back at the foo.
-
- Because this is used only in specific cases, we don't need to
- check all the things that `on_failure_jump' does, to make
- sure the right things get saved on the stack. Hence we don't
- share its code. The only reason to push anything on the
- stack at all is that otherwise we would have to change
- `anychar's code to do something besides goto fail in this
- case; that seems worse than this. */
- case on_failure_keep_string_jump:
- DEBUG_PRINT1 ("EXECUTING on_failure_keep_string_jump");
-
- EXTRACT_NUMBER_AND_INCR (mcnt, p);
- DEBUG_PRINT3 (" %d (to 0x%x):\n", mcnt, p + mcnt);
-
- PUSH_FAILURE_POINT (p + mcnt, NULL, -2);
- break;
+ goto fail;
+
+
+ /* on_failure_keep_string_jump is used to optimize `.*\n'. It
+ pushes NULL as the value for the string on the stack. Then
+ `pop_failure_point' will keep the current value for the
+ string, instead of restoring it. To see why, consider
+ matching `foo\nbar' against `.*\n'. The .* matches the foo;
+ then the . fails against the \n. But the next thing we want
+ to do is match the \n against the \n; if we restored the
+ string value, we would be back at the foo.
+
+ Because this is used only in specific cases, we don't need to
+ check all the things that `on_failure_jump' does, to make
+ sure the right things get saved on the stack. Hence we don't
+ share its code. The only reason to push anything on the
+ stack at all is that otherwise we would have to change
+ `anychar's code to do something besides goto fail in this
+ case; that seems worse than this. */
+ case on_failure_keep_string_jump:
+ DEBUG_PRINT1 ("EXECUTING on_failure_keep_string_jump");
+
+ EXTRACT_NUMBER_AND_INCR (mcnt, p);
+ DEBUG_PRINT3 (" %d (to 0x%x):\n", mcnt, p + mcnt);
+
+ PUSH_FAILURE_POINT (p + mcnt, NULL, -2);
+ break;
- on_failure:
- DEBUG_PRINT1 ("EXECUTING on_failure_jump");
-
- EXTRACT_NUMBER_AND_INCR (mcnt, p);
- DEBUG_PRINT3 (" %d (to 0x%x)", mcnt, p + mcnt);
-
- /* If this on_failure_jump comes right before a group (i.e.,
- the original * applied to a group), save the information
- for that group and all inner ones, so that if we fail back
- to this point, the group's information will be correct.
- For example, in \(a*\)*\1, we need the preceding group,
- and in \(zz\(a*\)b*\)\2, we need the inner group. */
-
- /* We can't use `p' to check ahead because we push
- a failure point to `p + mcnt' after we do this. */
- p1 = p;
-
- /* We need to skip no_op's before we look for the
- start_memory in case this on_failure_jump is happening as
- the result of a completed succeed_n, as in \(a\)\{1,3\}b\1
- against aba. */
- while (p1 < pend && (re_opcode_t) *p1 == no_op)
- p1++;
-
- if (p1 < pend && (re_opcode_t) *p1 == start_memory)
- {
- /* We have a new highest active register now. This will
- get reset at the start_memory we are about to get to,
- but we will have saved all the registers relevant to
- this repetition op, as described above. */
- highest_active_reg = *(p1 + 1) + *(p1 + 2);
- if (lowest_active_reg == NO_LOWEST_ACTIVE_REG)
- lowest_active_reg = *(p1 + 1);
- }
-
- DEBUG_PRINT1 (":\n");
- PUSH_FAILURE_POINT (p + mcnt, d, -2);
- break;
-
-
- /* A smart repeat ends with `maybe_pop_jump'.
- We change it to either `pop_failure_jump' or `jump'. */
- case maybe_pop_jump:
- EXTRACT_NUMBER_AND_INCR (mcnt, p);
- DEBUG_PRINT2 ("EXECUTING maybe_pop_jump %d.\n", mcnt);
- {
+ on_failure:
+ DEBUG_PRINT1 ("EXECUTING on_failure_jump");
+
+ EXTRACT_NUMBER_AND_INCR (mcnt, p);
+ DEBUG_PRINT3 (" %d (to 0x%x)", mcnt, p + mcnt);
+
+ /* If this on_failure_jump comes right before a group (i.e.,
+ the original * applied to a group), save the information
+ for that group and all inner ones, so that if we fail back
+ to this point, the group's information will be correct.
+ For example, in \(a*\)*\1, we need the preceding group,
+ and in \(zz\(a*\)b*\)\2, we need the inner group. */
+
+ /* We can't use `p' to check ahead because we push
+ a failure point to `p + mcnt' after we do this. */
+ p1 = p;
+
+ /* We need to skip no_op's before we look for the
+ start_memory in case this on_failure_jump is happening as
+ the result of a completed succeed_n, as in \(a\)\{1,3\}b\1
+ against aba. */
+ while (p1 < pend && (re_opcode_t) *p1 == no_op)
+ p1++;
+
+ if (p1 < pend && (re_opcode_t) *p1 == start_memory)
+ {
+ /* We have a new highest active register now. This will
+ get reset at the start_memory we are about to get to,
+ but we will have saved all the registers relevant to
+ this repetition op, as described above. */
+ highest_active_reg = *(p1 + 1) + *(p1 + 2);
+ if (lowest_active_reg == NO_LOWEST_ACTIVE_REG)
+ lowest_active_reg = *(p1 + 1);
+ }
+
+ DEBUG_PRINT1 (":\n");
+ PUSH_FAILURE_POINT (p + mcnt, d, -2);
+ break;
+
+
+ /* A smart repeat ends with `maybe_pop_jump'.
+ We change it to either `pop_failure_jump' or `jump'. */
+ case maybe_pop_jump:
+ EXTRACT_NUMBER_AND_INCR (mcnt, p);
+ DEBUG_PRINT2 ("EXECUTING maybe_pop_jump %d.\n", mcnt);
+ {