Merge branch 'f/main' into sv/add-cdata-codegen-with-eager-outputs

Addressed merge conflicts.

fuzz/target.c

@@ -422,6 +422,11 @@ fsm_eager_output_dump(FILE *f, const struct fsm *fsm);
 static int
 fuzz_eager_output(const uint8_t *data, size_t size)
 {
+	if (size > 0) {
+		const unsigned seed = data[0];
+		srand(seed);
+	}
+
 	struct feo_env env = {
 		.ok = true,
 		.pattern_count = 0,
@@ -446,9 +451,6 @@ fuzz_eager_output(const uint8_t *data, size_t size)
 
 	size_t max_pattern_length = 0;
 
-	const unsigned seed = size == 0 ? 0 : data[0];
-	srand(seed);
-
 	/* chop data into a series of patterns */
 	{
 		size_t prev = 0;
@@ -506,13 +508,14 @@ fuzz_eager_output(const uint8_t *data, size_t size)
 		}
 	}
 
-	struct re_anchoring_info anchorage[MAX_PATTERNS] = {0};
+	enum re_is_anchored_res anchorage[MAX_PATTERNS] = {0};
 
 	/* for each pattern, attempt to compile to a DFA */
 	for (size_t p_i = 0; p_i < env.pattern_count; p_i++) {
 		const char *p = env.patterns[p_i];
 
-		if (!re_is_anchored(RE_PCRE, fsm_sgetc, &p, 0, NULL, &anchorage[p_i])) {
+		enum re_is_anchored_res a = re_is_anchored(RE_PCRE, fsm_sgetc, &p, 0, NULL);
+		if (a == RE_IS_ANCHORED_ERROR) {
 			continue; /* unsupported regex */
 		}
 
@@ -602,8 +605,8 @@ fuzz_eager_output(const uint8_t *data, size_t size)
 			}
 
 			entries[used].fsm = cp;
-			entries[used].anchored_start = anchorage[i].start;
-			entries[used].anchored_end = anchorage[i].end;
+			entries[used].anchored_start = anchorage[i] & RE_IS_ANCHORED_START;
+			entries[used].anchored_end = anchorage[i] & RE_IS_ANCHORED_END;
 			used++;
 		}
 
@@ -648,7 +651,7 @@ fuzz_eager_output(const uint8_t *data, size_t size)
 	 * Use the combined DFA to generate matches, check that the
 	 * match behavior agrees with the individual DFA copies. */
 	env.current_pattern = (size_t)-1;
-	if (!fsm_generate_matches(env.combined, max_pattern_length, seed, gen_combined_check_individual_cb, &env)) {
+	if (!fsm_generate_matches(env.combined, max_pattern_length, 1, gen_combined_check_individual_cb, &env)) {
 		goto cleanup;
 	}
 
@@ -658,7 +661,7 @@ fuzz_eager_output(const uint8_t *data, size_t size)
 	/* check behavior against the combined DFA. */
 	for (size_t i = 0; i < env.pattern_count; i++) {
 		env.current_pattern = i;
-		if (!fsm_generate_matches(env.combined, max_pattern_length, seed, gen_individual_check_combined_cb, &env)) {
+		if (!fsm_generate_matches(env.combined, max_pattern_length, 1, gen_individual_check_combined_cb, &env)) {
 			goto cleanup;
 		}
 	}

include/adt/common.h

@@ -36,7 +36,7 @@
 /* If non-zero, expand the timer macros defined below, otherwise
  * they compile away. */
 #ifndef TRACK_TIMES
-#define TRACK_TIMES 0
+#define TRACK_TIMES (0 && !BUILD_FOR_FUZZER)
 #endif
 
 #if EXPENSIVE_CHECKS && TRACK_TIMES

include/fsm/fsm.h

@@ -432,6 +432,21 @@ fsm_remove_epsilons(struct fsm *fsm);
 int
 fsm_determinise(struct fsm *fsm);
 
+/* Determinise, with a passed in configuration
+ * and a distinct return value for reaching
+ * the state limit. */
+struct fsm_determinise_config {
+	size_t state_limit;	/* 0: no limit */
+};
+enum fsm_determinise_with_config_res {
+	FSM_DETERMINISE_WITH_CONFIG_OK,
+	FSM_DETERMINISE_WITH_CONFIG_STATE_LIMIT_REACHED,
+	FSM_DETERMINISE_WITH_CONFIG_ERRNO,
+};
+enum fsm_determinise_with_config_res
+fsm_determinise_with_config(struct fsm *fsm,
+	const struct fsm_determinise_config *config);
+
 /*
  * Make a DFA complete, as per fsm_iscomplete.
  */

include/fsm/walk.h

@@ -90,9 +90,10 @@ fsm_walk_edges(const struct fsm *fsm, void *opaque,
  * functionally equivalent cases makes testing dramatically faster,
  * but exploring every edge could be added later.
  *
- * If seed is zero then it will generate the first label in the label
- * set, otherwise a label from the set will be chosen using rand()
- * (favoring printable characters).
+ * If randomized is zero then it will generate the first label in the
+ * label set, otherwise a label from the set will be chosen using rand()
+ * (favoring printable characters). The caller can use srand()
+ * beforehand to set a PRNG seed.
  *
  * Note: fsm is non-const because it calls fsm_trim on the FSM
  * internally. This records the shortest distance from each state to an
@@ -118,7 +119,7 @@ fsm_generate_matches_cb(const struct fsm *fsm,
     const char *input, size_t input_length,
     fsm_state_t end_state, void *opaque);
 int
-fsm_generate_matches(struct fsm *fsm, size_t max_length, unsigned seed,
+fsm_generate_matches(struct fsm *fsm, size_t max_length, int randomized,
     fsm_generate_matches_cb *cb, void *opaque);
 
 /* Callback provided for the most basic use case for

include/re/re.h

@@ -136,20 +136,21 @@ re_comp(enum re_dialect dialect,
 	const struct fsm_alloc *alloc,
 	enum re_flags flags, struct re_err *err);
 
-struct re_anchoring_info {
-	int start;
-	int end;
-	/* FIXME: this could also check for AST_FLAG_NULLABLE, AST_FLAG_UNSATISFIABLE,
-	 * AST_FLAG_ALWAYS_CONSUMES, AST_FLAG_CAN_CONSUME */
-};
-
 /* Parse and analyze the regex enough to determine whether it is
- * anchored at the start and/or end. Returns 0 if the regex is not
- * supported, otherwise returns 1 and writes anchoring flags into *info. */
-int
+ * anchored at the start and/or end.
+ *
+ * As long as the result is checked for RE_IS_ANCHORED_ERROR first,
+ * the result can be used like a bitset. */
+enum re_is_anchored_res {
+	RE_IS_ANCHORED_NONE = 0x00,
+	RE_IS_ANCHORED_START = 0x01,
+	RE_IS_ANCHORED_END = 0x02,
+	RE_IS_ANCHORED_BOTH = 0x03,
+	RE_IS_ANCHORED_ERROR = 0xFFFF,
+};
+enum re_is_anchored_res
 re_is_anchored(enum re_dialect dialect, re_getchar_fun *f, void *opaque,
-	enum re_flags flags, struct re_err *err,
-	struct re_anchoring_info *info);
+	enum re_flags flags, struct re_err *err);
 
 /*
  * Return a human-readable string describing a given error code. The string

src/libfsm/determinise.c

@@ -20,16 +20,20 @@ dump_labels(FILE *f, const uint64_t labels[4])
 	}
 }
 
-int
-fsm_determinise(struct fsm *nfa)
+enum fsm_determinise_with_config_res
+fsm_determinise_with_config(struct fsm *nfa,
+	const struct fsm_determinise_config *config)
 {
-	int res = 0;
+	enum fsm_determinise_with_config_res res = FSM_DETERMINISE_WITH_CONFIG_ERRNO;
 	struct mappingstack *stack = NULL;
 
 	struct interned_state_set_pool *issp = NULL;
 	struct map map = { NULL, 0, 0, NULL };
 	struct mapping *curr = NULL;
 	size_t dfacount = 0;
+	const size_t state_limit = config == NULL
+	    ? 0
+	    : config->state_limit;
 
 	struct analyze_closures_env ac_env = { 0 };
 
@@ -43,7 +47,7 @@ fsm_determinise(struct fsm *nfa)
 	 */
 	if (fsm_has(nfa, fsm_hasepsilons)) {
 		if (!fsm_remove_epsilons(nfa)) {
-			return 0;
+			return FSM_DETERMINISE_WITH_CONFIG_ERRNO;
 		}
 	}
 
@@ -55,7 +59,12 @@ fsm_determinise(struct fsm *nfa)
 
 	issp = interned_state_set_pool_alloc(nfa->alloc);
 	if (issp == NULL) {
-		return 0;
+		return FSM_DETERMINISE_WITH_CONFIG_ERRNO;
+	}
+
+	if (state_limit != 0 && fsm_countstates(nfa) > state_limit) {
+		res = FSM_DETERMINISE_WITH_CONFIG_STATE_LIMIT_REACHED;
+		goto cleanup;
 	}
 
 	{
@@ -77,7 +86,7 @@ fsm_determinise(struct fsm *nfa)
 		 */
 
 		if (!fsm_getstart(nfa, &start)) {
-			res = 1;
+			res = FSM_DETERMINISE_WITH_CONFIG_OK;
 			goto cleanup;
 		}
 
@@ -153,6 +162,11 @@ fsm_determinise(struct fsm *nfa)
 				assert(m->dfastate < dfacount);
 			} else {
 				/* not found -- add a new one and push it to the stack for processing */
+
+				if (state_limit != 0 && dfacount > state_limit) {
+					res = FSM_DETERMINISE_WITH_CONFIG_STATE_LIMIT_REACHED;
+					goto cleanup;
+				}
 				if (!map_add(&map, dfacount, iss, &m)) {
 					goto cleanup;
 				}
@@ -174,8 +188,6 @@ fsm_determinise(struct fsm *nfa)
 		}
 
 		ac_env.output_count = 0;
-
-		/* All elements in sclosures[] are interned, so they will be freed later. */
 	} while ((curr = stack_pop(stack)));
 
 	{
@@ -267,7 +279,7 @@ fsm_determinise(struct fsm *nfa)
 	assert(fsm_all(nfa, fsm_isdfa));
 #endif
 
-	res = 1;
+	res = FSM_DETERMINISE_WITH_CONFIG_OK;
 
 cleanup:
 	map_free(&map);
@@ -318,6 +330,22 @@ fsm_determinise(struct fsm *nfa)
 	return res;
 }
 
+int
+fsm_determinise(struct fsm *nfa)
+{
+	enum fsm_determinise_with_config_res res = fsm_determinise_with_config(nfa, NULL);
+	switch (res) {
+	case FSM_DETERMINISE_WITH_CONFIG_OK:
+		return 1;
+	case FSM_DETERMINISE_WITH_CONFIG_STATE_LIMIT_REACHED:
+		/* unreachable */
+		return 0;
+	case FSM_DETERMINISE_WITH_CONFIG_ERRNO:
+	default:
+		return 0;
+	}
+}
+
 /* Add DFA_state to the list for NFA_state. */
 static int
 add_reverse_mapping(const struct fsm_alloc *alloc,

src/libfsm/gen.c

@@ -107,7 +107,7 @@ struct gen_ctx {
 static bool
 gen_init_outer(struct fsm *fsm, size_t max_length,
     fsm_generate_matches_cb *cb, void *opaque,
-    bool randomized, unsigned seed);
+    bool randomized);
 
 static bool
 gen_init(struct gen_ctx *ctx, struct fsm *fsm);
@@ -140,7 +140,7 @@ static bool
 grow_stack(struct gen_ctx *ctx);
 
 int
-fsm_generate_matches(struct fsm *fsm, size_t max_length, unsigned seed,
+fsm_generate_matches(struct fsm *fsm, size_t max_length, int randomized,
     fsm_generate_matches_cb *cb, void *opaque)
 {
 	if (max_length == 0) {
@@ -154,7 +154,7 @@ fsm_generate_matches(struct fsm *fsm, size_t max_length, unsigned seed,
 
 	INIT_TIMERS();
 	TIME(&pre);
-	int res = gen_init_outer(fsm, max_length, cb, opaque, seed != 0, seed);
+	int res = gen_init_outer(fsm, max_length, cb, opaque, randomized != 0);
 	TIME(&post);
 
 	DIFF_MSEC("fsm_generate_matches", pre, post, NULL);
@@ -204,7 +204,7 @@ fsm_generate_cb_printf(const struct fsm *fsm,
 static bool
 gen_init_outer(struct fsm *fsm, size_t max_length,
     fsm_generate_matches_cb *cb, void *opaque,
-    bool randomized, unsigned seed)
+    bool randomized)
 {
 	int res = false;
 	if (fsm == NULL || cb == NULL || max_length == 0) {
@@ -213,10 +213,6 @@ gen_init_outer(struct fsm *fsm, size_t max_length,
 
 	assert(fsm_all(fsm, fsm_isdfa)); /* DFA-only */
 
-	if (randomized) {
-		srand(seed);
-	}
-
 #if LOG_GEN > 1
 	fprintf(stderr, "%s: %u states\n", __func__, fsm_countstates(fsm));
 #endif

src/libfsm/libfsm.syms

@@ -114,6 +114,7 @@ fsm_countstates
 fsm_trim
 fsm_reverse
 fsm_determinise
+fsm_determinise_with_config
 fsm_remove_epsilons
 fsm_complete
 fsm_minimise

src/libfsm/trim.c

@@ -462,6 +462,10 @@ integrity_check(const char *descr, const struct fsm *fsm)
 	return;
 #endif
 
+#if !EXPENSIVE_CHECKS
+	return;
+#endif
+
 	if (LOG_TRIM > 1) {
 		fprintf(stderr, "integrity check: %s...\n", descr);
 	}

src/libre/re.c

@@ -335,37 +335,40 @@ re_is_literal(enum re_dialect dialect, int (*getc)(void *opaque), void *opaque,
 	return -1;
 }
 
-/* FIXME: placeholder interface */
-int
+enum re_is_anchored_res
 re_is_anchored(enum re_dialect dialect, re_getchar_fun *getc, void *opaque,
-	enum re_flags flags, struct re_err *err,
-	struct re_anchoring_info *info)
+	enum re_flags flags, struct re_err *err)
 {
-	/* FIXME: copy/pasted from above, factor out common */
+	/* FIXME: copy/pasted from above, factor out common code later. */
 
 	struct ast *ast;
 	const struct dialect *m;
 	int unsatisfiable;
 
 	assert(getc != NULL);
-	assert(info != NULL);
 
 	m = re_dialect(dialect);
 	if (m == NULL) {
 		if (err != NULL) { err->e = RE_EBADDIALECT; }
-		return 0;
+		return RE_IS_ANCHORED_ERROR;
 	}
 
 	flags |= m->flags;
 
 	ast = re_parse(dialect, getc, opaque, flags, err, &unsatisfiable);
 	if (ast == NULL) {
-		return 0;
+		return RE_IS_ANCHORED_ERROR;
 	}
 
-	info->start = (ast->expr->flags & AST_FLAG_ANCHORED_START) != 0;
-	info->end = (ast->expr->flags & AST_FLAG_ANCHORED_END) != 0;
+	/* Copy anchoring flags, ending up with NONE, START, END, or BOTH. */
+	enum re_is_anchored_res res = RE_IS_ANCHORED_NONE;
+	if (ast->expr->flags & AST_FLAG_ANCHORED_START) {
+		res |= RE_IS_ANCHORED_START;
+	}
+	if (ast->expr->flags & AST_FLAG_ANCHORED_END) {
+		res |= RE_IS_ANCHORED_END;
+	}
 
 	ast_free(ast);
-	return 1;
+	return res;
 }