-
-
-/* Read the ending character of a range (in a bracket expression) from the
- uncompiled pattern *P_PTR (which ends at PEND). We assume the
- starting character is in `P[-2]'. (`P[-1]' is the character `-'.)
- Then we set the translation of all bits between the starting and
- ending characters (inclusive) in the compiled pattern B.
-
- Return an error code.
-
- We use these short variable names so we can use the same macros as
- `regex_compile' itself. */
-
-static reg_errcode_t
-compile_range (p_ptr, pend, translate, syntax, b)
- const char **p_ptr, *pend;
- char *translate;
- reg_syntax_t syntax;
- unsigned char *b;
-{
- unsigned this_char;
-
- const char *p = *p_ptr;
-
- /* Even though the pattern is a signed `char *', we need to fetch into
- `unsigned char's. Reason: if the high bit of the pattern character
- is set, the range endpoints will be negative if we fetch into a
- signed `char *'. */
- unsigned char range_end;
- unsigned char range_start = p[-2];
-
- if (p == pend)
- return REG_ERANGE;
-
- PATFETCH (range_end);
-
- /* Have to increment the pointer into the pattern string, so the
- caller isn't still at the ending character. */
- (*p_ptr)++;
-
- /* If the start is after the end, the range is empty. */
- if (range_start > range_end)
- return syntax & RE_NO_EMPTY_RANGES ? REG_ERANGE : REG_NOERROR;
-
- /* Here we see why `this_char' has to be larger than an `unsigned
- char' -- the range is inclusive, so if `range_end' == 0xff
- (assuming 8-bit characters), we would otherwise go into an infinite
- loop, since all characters <= 0xff. */
- for (this_char = range_start; this_char <= range_end; this_char++)
- {
- SET_LIST_BIT (TRANSLATE (this_char));
- }
-
- return REG_NOERROR;
-}
-\f
-/* Failure stack declarations and macros; both re_compile_fastmap and
- re_match_2 use a failure stack. These have to be macros because of
- REGEX_ALLOCATE. */
-
-
-/* Number of failure points for which to initially allocate space
- when matching. If this number is exceeded, we allocate more
- space, so it is not a hard limit. */
-#ifndef INIT_FAILURE_ALLOC
-#define INIT_FAILURE_ALLOC 5
-#endif
-
-/* Roughly the maximum number of failure points on the stack. Would be
- exactly that if always used MAX_FAILURE_SPACE each time we failed.
- This is a variable only so users of regex can assign to it; we never
- change it ourselves. */
-int re_max_failures = 2000;
-
-typedef const unsigned char *fail_stack_elt_t;
-
-typedef struct
-{
- fail_stack_elt_t *stack;
- unsigned size;
- unsigned avail; /* Offset of next open position. */
-} fail_stack_type;
-
-#define FAIL_STACK_EMPTY() (fail_stack.avail == 0)
-#define FAIL_STACK_PTR_EMPTY() (fail_stack_ptr->avail == 0)
-#define FAIL_STACK_FULL() (fail_stack.avail == fail_stack.size)
-#define FAIL_STACK_TOP() (fail_stack.stack[fail_stack.avail])
-
-
-/* Initialize `fail_stack'. Do `return -2' if the alloc fails. */
-
-#define INIT_FAIL_STACK() \
- do { \
- fail_stack.stack = (fail_stack_elt_t *) \
- REGEX_ALLOCATE (INIT_FAILURE_ALLOC * sizeof (fail_stack_elt_t)); \
- \
- if (fail_stack.stack == NULL) \
- return -2; \
- \
- fail_stack.size = INIT_FAILURE_ALLOC; \
- fail_stack.avail = 0; \
- } while (0)
-
-
-/* Double the size of FAIL_STACK, up to approximately `re_max_failures' items.
-
- Return 1 if succeeds, and 0 if either ran out of memory
- allocating space for it or it was already too large.
-
- REGEX_REALLOCATE requires `destination' be declared. */
-
-#define DOUBLE_FAIL_STACK(fail_stack) \
- ((fail_stack).size > re_max_failures * MAX_FAILURE_ITEMS \
- ? 0 \
- : ((fail_stack).stack = (fail_stack_elt_t *) \
- REGEX_REALLOCATE ((fail_stack).stack, \
- (fail_stack).size * sizeof (fail_stack_elt_t), \
- ((fail_stack).size << 1) * sizeof (fail_stack_elt_t)), \
- \
- (fail_stack).stack == NULL \
- ? 0 \
- : ((fail_stack).size <<= 1, \
- 1)))
-
-
-/* Push PATTERN_OP on FAIL_STACK.
-
- Return 1 if was able to do so and 0 if ran out of memory allocating
- space to do so. */
-#define PUSH_PATTERN_OP(pattern_op, fail_stack) \
- ((FAIL_STACK_FULL () \
- && !DOUBLE_FAIL_STACK (fail_stack)) \
- ? 0 \
- : ((fail_stack).stack[(fail_stack).avail++] = pattern_op, \
- 1))
-
-/* This pushes an item onto the failure stack. Must be a four-byte
- value. Assumes the variable `fail_stack'. Probably should only
- be called from within `PUSH_FAILURE_POINT'. */
-#define PUSH_FAILURE_ITEM(item) \
- fail_stack.stack[fail_stack.avail++] = (fail_stack_elt_t) item
-
-/* The complement operation. Assumes `fail_stack' is nonempty. */
-#define POP_FAILURE_ITEM() fail_stack.stack[--fail_stack.avail]
-
-/* Used to omit pushing failure point id's when we're not debugging. */
-#ifdef DEBUG
-#define DEBUG_PUSH PUSH_FAILURE_ITEM
-#define DEBUG_POP(item_addr) *(item_addr) = POP_FAILURE_ITEM ()
-#else
-#define DEBUG_PUSH(item)
-#define DEBUG_POP(item_addr)
-#endif
-
-
-/* Push the information about the state we will need
- if we ever fail back to it.
-
- Requires variables fail_stack, regstart, regend, reg_info, and
- num_regs be declared. DOUBLE_FAIL_STACK requires `destination' be
- declared.
-
- Does `return FAILURE_CODE' if runs out of memory. */
-
-#define PUSH_FAILURE_POINT(pattern_place, string_place, failure_code) \
- do { \
- char *destination; \
- /* Must be int, so when we don't save any registers, the arithmetic \
- of 0 + -1 isn't done as unsigned. */ \
- int this_reg; \
- \
- DEBUG_STATEMENT (failure_id++); \
- DEBUG_STATEMENT (nfailure_points_pushed++); \
- DEBUG_PRINT2 ("\nPUSH_FAILURE_POINT #%u:\n", failure_id); \
- DEBUG_PRINT2 (" Before push, next avail: %d\n", (fail_stack).avail);\
- DEBUG_PRINT2 (" size: %d\n", (fail_stack).size);\
- \
- DEBUG_PRINT2 (" slots needed: %d\n", NUM_FAILURE_ITEMS); \
- DEBUG_PRINT2 (" available: %d\n", REMAINING_AVAIL_SLOTS); \
- \
- /* Ensure we have enough space allocated for what we will push. */ \
- while (REMAINING_AVAIL_SLOTS < NUM_FAILURE_ITEMS) \
- { \
- if (!DOUBLE_FAIL_STACK (fail_stack)) \
- return failure_code; \
- \
- DEBUG_PRINT2 ("\n Doubled stack; size now: %d\n", \
- (fail_stack).size); \
- DEBUG_PRINT2 (" slots available: %d\n", REMAINING_AVAIL_SLOTS);\
- } \
- \
- /* Push the info, starting with the registers. */ \
- DEBUG_PRINT1 ("\n"); \
- \
- for (this_reg = lowest_active_reg; this_reg <= highest_active_reg; \
- this_reg++) \
- { \
- DEBUG_PRINT2 (" Pushing reg: %d\n", this_reg); \
- DEBUG_STATEMENT (num_regs_pushed++); \
- \
- DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \
- PUSH_FAILURE_ITEM (regstart[this_reg]); \
- \
- DEBUG_PRINT2 (" end: 0x%x\n", regend[this_reg]); \
- PUSH_FAILURE_ITEM (regend[this_reg]); \
- \
- DEBUG_PRINT2 (" info: 0x%x\n ", reg_info[this_reg]); \
- DEBUG_PRINT2 (" match_null=%d", \
- REG_MATCH_NULL_STRING_P (reg_info[this_reg])); \
- DEBUG_PRINT2 (" active=%d", IS_ACTIVE (reg_info[this_reg])); \
- DEBUG_PRINT2 (" matched_something=%d", \
- MATCHED_SOMETHING (reg_info[this_reg])); \
- DEBUG_PRINT2 (" ever_matched=%d", \
- EVER_MATCHED_SOMETHING (reg_info[this_reg])); \
- DEBUG_PRINT1 ("\n"); \
- PUSH_FAILURE_ITEM (reg_info[this_reg].word); \
- } \
- \
- DEBUG_PRINT2 (" Pushing low active reg: %d\n", lowest_active_reg);\
- PUSH_FAILURE_ITEM (lowest_active_reg); \
- \
- DEBUG_PRINT2 (" Pushing high active reg: %d\n", highest_active_reg);\
- PUSH_FAILURE_ITEM (highest_active_reg); \
- \
- DEBUG_PRINT2 (" Pushing pattern 0x%x: ", pattern_place); \
- DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern_place, pend); \
- PUSH_FAILURE_ITEM (pattern_place); \
- \
- DEBUG_PRINT2 (" Pushing string 0x%x: `", string_place); \
- DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, \
- size2); \
- DEBUG_PRINT1 ("'\n"); \
- PUSH_FAILURE_ITEM (string_place); \
- \
- DEBUG_PRINT2 (" Pushing failure id: %u\n", failure_id); \
- DEBUG_PUSH (failure_id); \
- } while (0)
-
-/* This is the number of items that are pushed and popped on the stack
- for each register. */
-#define NUM_REG_ITEMS 3
-
-/* Individual items aside from the registers. */
-#ifdef DEBUG
-#define NUM_NONREG_ITEMS 5 /* Includes failure point id. */
-#else
-#define NUM_NONREG_ITEMS 4
-#endif
-
-/* We push at most this many items on the stack. */
-#define MAX_FAILURE_ITEMS ((num_regs - 1) * NUM_REG_ITEMS + NUM_NONREG_ITEMS)
-
-/* We actually push this many items. */
-#define NUM_FAILURE_ITEMS \
- ((highest_active_reg - lowest_active_reg + 1) * NUM_REG_ITEMS \
- + NUM_NONREG_ITEMS)
-
-/* How many items can still be added to the stack without overflowing it. */
-#define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail)
-
-
-/* Pops what PUSH_FAIL_STACK pushes.
-
- We restore into the parameters, all of which should be lvalues:
- STR -- the saved data position.
- PAT -- the saved pattern position.
- LOW_REG, HIGH_REG -- the highest and lowest active registers.
- REGSTART, REGEND -- arrays of string positions.
- REG_INFO -- array of information about each subexpression.
-
- Also assumes the variables `fail_stack' and (if debugging), `bufp',
- `pend', `string1', `size1', `string2', and `size2'. */
-
-#define POP_FAILURE_POINT(str, pat, low_reg, high_reg, regstart, regend, reg_info)\
-{ \
- DEBUG_STATEMENT (fail_stack_elt_t failure_id;) \
- int this_reg; \
- const unsigned char *string_temp; \
- \
- assert (!FAIL_STACK_EMPTY ()); \
- \
- /* Remove failure points and point to how many regs pushed. */ \
- DEBUG_PRINT1 ("POP_FAILURE_POINT:\n"); \
- DEBUG_PRINT2 (" Before pop, next avail: %d\n", fail_stack.avail); \
- DEBUG_PRINT2 (" size: %d\n", fail_stack.size); \
- \
- assert (fail_stack.avail >= NUM_NONREG_ITEMS); \
- \
- DEBUG_POP (&failure_id); \
- DEBUG_PRINT2 (" Popping failure id: %u\n", failure_id); \
- \
- /* If the saved string location is NULL, it came from an \
- on_failure_keep_string_jump opcode, and we want to throw away the \
- saved NULL, thus retaining our current position in the string. */ \
- string_temp = POP_FAILURE_ITEM (); \
- if (string_temp != NULL) \
- str = (const char *) string_temp; \
- \
- DEBUG_PRINT2 (" Popping string 0x%x: `", str); \
- DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \
- DEBUG_PRINT1 ("'\n"); \
- \
- pat = (unsigned char *) POP_FAILURE_ITEM (); \
- DEBUG_PRINT2 (" Popping pattern 0x%x: ", pat); \
- DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \
- \
- /* Restore register info. */ \
- high_reg = (unsigned) POP_FAILURE_ITEM (); \
- DEBUG_PRINT2 (" Popping high active reg: %d\n", high_reg); \
- \
- low_reg = (unsigned) POP_FAILURE_ITEM (); \
- DEBUG_PRINT2 (" Popping low active reg: %d\n", low_reg); \
- \
- for (this_reg = high_reg; this_reg >= low_reg; this_reg--) \
- { \
- DEBUG_PRINT2 (" Popping reg: %d\n", this_reg); \
- \
- reg_info[this_reg].word = POP_FAILURE_ITEM (); \
- DEBUG_PRINT2 (" info: 0x%x\n", reg_info[this_reg]); \
- \
- regend[this_reg] = (const char *) POP_FAILURE_ITEM (); \
- DEBUG_PRINT2 (" end: 0x%x\n", regend[this_reg]); \
- \
- regstart[this_reg] = (const char *) POP_FAILURE_ITEM (); \
- DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \
- } \
- \
- DEBUG_STATEMENT (nfailure_points_popped++); \
-} /* POP_FAILURE_POINT */