gnuregex.c

-

			    return ret;
		    } else if (p[0] == '-' && p[1] != ']') {	/* This handles ranges made up of characters only.  */
			reg_errcode_t ret;

			/* Move past the `-'.  */
			PATFETCH(c1);

			ret = compile_range(&p, pend, translate, syntax, b);
			if (ret != REG_NOERROR)
			    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)
			    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))
				return REG_ECTYPE;

			    /* Throw away the ] at the end of the character
			     * class.  */
			    PATFETCH(c);

			    if (p == pend)
				return REG_EBRACK;

			    for (ch = 0; ch < 1 << BYTEWIDTH; ch++) {
				if ((is_alnum && ISALNUM(ch))
				    || (is_alpha && ISALPHA(ch))
				    || (is_blank && ISBLANK(ch))
				    || (is_cntrl && ISCNTRL(ch))
				    || (is_digit && ISDIGIT(ch))
				    || (is_graph && ISGRAPH(ch))
				    || (is_lower && ISLOWER(ch))
				    || (is_print && ISPRINT(ch))
				    || (is_punct && ISPUNCT(ch))
				    || (is_space && ISSPACE(ch))
				    || (is_upper && ISUPPER(ch))
				    || (is_xdigit && ISXDIGIT(ch)))
				    SET_LIST_BIT(ch);
			    }
			    had_char_class = true;
			} else {
			    c1++;
			    while (c1--)
				PATUNFETCH;
			    SET_LIST_BIT('[');
			    SET_LIST_BIT(':');
			    had_char_class = false;
			}
		    } else {
			had_char_class = false;
			SET_LIST_BIT(c);
		    }
		}

		/* Discard any (non)matching list bytes that are all 0 at the
		 * end of the map.  Decrease the map-length byte too.  */
		while ((int) b[-1] > 0 && b[b[-1] - 1] == 0)
		    b[-1]--;
		b += b[-1];
	    }
	    break;


	case '(':
	    if (syntax & RE_NO_BK_PARENS)
		goto handle_open;
	    else
		goto normal_char;


	case ')':
	    if (syntax & RE_NO_BK_PARENS)
		goto handle_close;
	    else
		goto normal_char;


	case '\n':
	    if (syntax & RE_NEWLINE_ALT)
		goto handle_alt;
	    else
		goto normal_char;


	case '|':
	    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)
		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 we've reached MAX_REGNUM groups, then this open
		 * won't actually generate any code, so we'll have to
		 * clear pending_exact explicitly.  */
		pending_exact = 0;
		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
			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
			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;
		    /* If we've reached MAX_REGNUM groups, then this open
		     * won't actually generate any code, so we'll have to
		     * clear pending_exact explicitly.  */
		    pending_exact = 0;

		    /* 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
			    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
			    return REG_BADBR;
		    }
		    if (!(syntax & RE_NO_BK_BRACES)) {
			if (c != '\\')
			    return REG_EBRACE;

			PATFETCH(c);
		    }
		    if (c != '}') {
			if (syntax & RE_NO_BK_BRACES)
			    goto unfetch_interval;
			else
			    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)
			    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> <succed_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;

#ifdef emacs
		/* There is no way to specify the before_dot and after_dot
		 * operators.  rms says this is ok.  --karl  */