+2019-10-11 Paul Eggert <eggert@cs.ucla.edu>
+
+ Simplify and regularize regex use of ‘assert’
+ Also, tell GCC about the asserts even when compiling without
+ debugging, to give it further optimization opportunities.
+ * lib/regex_internal.h (DEBUG_ASSERT): New macro.
+ * lib/regcomp.c (link_nfa_nodes, calc_eclosure)
+ (parse_expression, parse_bracket_exp):
+ * lib/regex_internal.c (build_wcs_buffer)
+ (build_wcs_upper_buffer, re_string_reconstruct)
+ (re_string_context_at):
+ * lib/regexec.c (re_search_stub, re_copy_regs)
+ (re_search_internal, prune_impossible_nodes, check_matching)
+ (check_halt_state_context, set_regs, sift_states_backward)
+ (build_sifted_states, transit_state_mb, transit_state_bkref)
+ (check_arrival_add_next_nodes, check_arrival_expand_ecl)
+ (match_ctx_add_subtop):
+ Use it instead of plain ‘assert’.
+
2019-10-09 Paul Eggert <eggert@cs.ucla.edu>
regex: omit debug assignment when not debugging
break;
case END_OF_RE:
- assert (node->next == NULL);
+ DEBUG_ASSERT (node->next == NULL);
break;
case OP_DUP_ASTERISK:
right = node->right->first->node_idx;
else
right = node->next->node_idx;
- assert (left > -1);
- assert (right > -1);
+ DEBUG_ASSERT (left > -1);
+ DEBUG_ASSERT (right > -1);
err = re_node_set_init_2 (dfa->edests + idx, left, right);
}
break;
break;
default:
- assert (!IS_EPSILON_NODE (node->token.type));
+ DEBUG_ASSERT (!IS_EPSILON_NODE (node->token.type));
dfa->nexts[idx] = node->next->node_idx;
break;
}
{
Idx node_idx;
bool incomplete;
-#ifdef DEBUG
- assert (dfa->nodes_len > 0);
-#endif
+ DEBUG_ASSERT (dfa->nodes_len > 0);
incomplete = false;
/* For each nodes, calculate epsilon closure. */
for (node_idx = 0; ; ++node_idx)
node_idx = 0;
}
-#ifdef DEBUG
- assert (dfa->eclosures[node_idx].nelem != -1);
-#endif
+ DEBUG_ASSERT (dfa->eclosures[node_idx].nelem != -1);
/* If we have already calculated, skip it. */
if (dfa->eclosures[node_idx].nelem != 0)
default:
/* Must not happen? */
-#ifdef DEBUG
- assert (0);
-#endif
+ DEBUG_ASSERT (false);
return NULL;
}
fetch_token (token, regexp, syntax);
goto parse_bracket_exp_free_return;
break;
default:
- assert (0);
+ DEBUG_ASSERT (false);
break;
}
}
{
if (ret_len)
{
- assert (pmatch[0].rm_so == start);
+ DEBUG_ASSERT (pmatch[0].rm_so == start);
rval = pmatch[0].rm_eo - start;
}
else
}
else
{
- assert (regs_allocated == REGS_FIXED);
+ DEBUG_ASSERT (regs_allocated == REGS_FIXED);
/* This function may not be called with REGS_FIXED and nregs too big. */
- assert (regs->num_regs >= nregs);
+ DEBUG_ASSERT (nregs <= regs->num_regs);
rval = REGS_FIXED;
}
|| dfa->init_state_begbuf == NULL))
return REG_NOMATCH;
-#ifdef DEBUG
/* We assume front-end functions already check them. */
- assert (0 <= last_start && last_start <= length);
-#endif
+ DEBUG_ASSERT (0 <= last_start && last_start <= length);
/* If initial states with non-begbuf contexts have no elements,
the regex must be anchored. If preg->newline_anchor is set,
break;
if (__glibc_unlikely (err != REG_NOMATCH))
goto free_return;
-#ifdef DEBUG
match_last = -1;
-#endif
}
else
break; /* We found a match. */
match_ctx_clean (&mctx);
}
-#ifdef DEBUG
- assert (match_last != -1);
- assert (err == REG_NOERROR);
-#endif
+ DEBUG_ASSERT (match_last != -1);
+ DEBUG_ASSERT (err == REG_NOERROR);
/* Set pmatch[] if we need. */
if (nmatch > 0)
: mctx.input.offsets[pmatch[reg_idx].rm_eo]);
}
#else
- assert (mctx.input.offsets_needed == 0);
+ DEBUG_ASSERT (mctx.input.offsets_needed == 0);
#endif
pmatch[reg_idx].rm_so += match_first;
pmatch[reg_idx].rm_eo += match_first;
re_dfastate_t **sifted_states;
re_dfastate_t **lim_states = NULL;
re_sift_context_t sctx;
-#ifdef DEBUG
- assert (mctx->state_log != NULL);
-#endif
+ DEBUG_ASSERT (mctx->state_log != NULL);
match_last = mctx->match_last;
halt_node = mctx->last_node;
/* An initial state must not be NULL (invalid). */
if (__glibc_unlikely (cur_state == NULL))
{
- assert (err == REG_ESPACE);
+ DEBUG_ASSERT (err == REG_ESPACE);
return -2;
}
err = extend_buffers (mctx, next_char_idx + 1);
if (__glibc_unlikely (err != REG_NOERROR))
{
- assert (err == REG_ESPACE);
+ DEBUG_ASSERT (err == REG_ESPACE);
return -2;
}
}
{
Idx i;
unsigned int context;
-#ifdef DEBUG
- assert (state->halt);
-#endif
+ DEBUG_ASSERT (state->halt);
context = re_string_context_at (&mctx->input, idx, mctx->eflags);
for (i = 0; i < state->nodes.nelem; ++i)
if (check_halt_node_context (mctx->dfa, state->nodes.elems[i], context))
regmatch_t *regs, re_node_set *eps_via_nodes)
{
Idx num = --fs->num;
- assert (num >= 0);
+ DEBUG_ASSERT (num >= 0);
*pidx = fs->stack[num].idx;
memcpy (regs, fs->stack[num].regs, sizeof (regmatch_t) * nregs);
re_node_set_free (eps_via_nodes);
regmatch_t *prev_idx_match;
bool prev_idx_match_malloced = false;
-#ifdef DEBUG
- assert (nmatch > 1);
- assert (mctx->state_log != NULL);
-#endif
+ DEBUG_ASSERT (nmatch > 1);
+ DEBUG_ASSERT (mctx->state_log != NULL);
if (fl_backtrack)
{
fs = &fs_body;
Idx str_idx = sctx->last_str_idx;
re_node_set cur_dest;
-#ifdef DEBUG
- assert (mctx->state_log != NULL && mctx->state_log[str_idx] != NULL);
-#endif
+ DEBUG_ASSERT (mctx->state_log != NULL && mctx->state_log[str_idx] != NULL);
/* Build sifted state_log[str_idx]. It has the nodes which can epsilon
transit to the last_node and the last_node itself. */
Idx prev_node = cur_src->elems[i];
int naccepted = 0;
bool ok;
+ DEBUG_ASSERT (!IS_EPSILON_NODE (dfa->nodes[prev_node].type));
-#ifdef DEBUG
- re_token_type_t type = dfa->nodes[prev_node].type;
- assert (!IS_EPSILON_NODE (type));
-#endif
#ifdef RE_ENABLE_I18N
/* If the node may accept "multi byte". */
if (dfa->nodes[prev_node].accept_mb)
err = clean_state_log_if_needed (mctx, dest_idx);
if (__glibc_unlikely (err != REG_NOERROR))
return err;
-#ifdef DEBUG
- assert (dfa->nexts[cur_node_idx] != -1);
-#endif
+ DEBUG_ASSERT (dfa->nexts[cur_node_idx] != -1);
new_nodes = dfa->eclosures + dfa->nexts[cur_node_idx];
dest_state = mctx->state_log[dest_idx];
/* And add the epsilon closures (which is 'new_dest_nodes') of
the backreference to appropriate state_log. */
-#ifdef DEBUG
- assert (dfa->nexts[node_idx] != -1);
-#endif
+ DEBUG_ASSERT (dfa->nexts[node_idx] != -1);
for (; bkc_idx < mctx->nbkref_ents; ++bkc_idx)
{
Idx subexp_len;
{
int naccepted = 0;
Idx cur_node = cur_nodes->elems[cur_idx];
-#ifdef DEBUG
- re_token_type_t type = dfa->nodes[cur_node].type;
- assert (!IS_EPSILON_NODE (type));
-#endif
+ DEBUG_ASSERT (!IS_EPSILON_NODE (dfa->nodes[cur_node].type));
+
#ifdef RE_ENABLE_I18N
/* If the node may accept "multi byte". */
if (dfa->nodes[cur_node].accept_mb)
reg_errcode_t err;
Idx idx, outside_node;
re_node_set new_nodes;
-#ifdef DEBUG
- assert (cur_nodes->nelem);
-#endif
+ DEBUG_ASSERT (cur_nodes->nelem);
err = re_node_set_alloc (&new_nodes, cur_nodes->nelem);
if (__glibc_unlikely (err != REG_NOERROR))
return err;
__attribute_warn_unused_result__
match_ctx_add_subtop (re_match_context_t *mctx, Idx node, Idx str_idx)
{
-#ifdef DEBUG
- assert (mctx->sub_tops != NULL);
- assert (mctx->asub_tops > 0);
-#endif
+ DEBUG_ASSERT (mctx->sub_tops != NULL);
+ DEBUG_ASSERT (mctx->asub_tops > 0);
if (__glibc_unlikely (mctx->nsub_tops == mctx->asub_tops))
{
Idx new_asub_tops = mctx->asub_tops * 2;