summaryrefslogtreecommitdiffstats
path: root/js/src/regexp/regexp-compiler.cc
diff options
context:
space:
mode:
Diffstat (limited to 'js/src/regexp/regexp-compiler.cc')
-rw-r--r--js/src/regexp/regexp-compiler.cc115
1 files changed, 56 insertions, 59 deletions
diff --git a/js/src/regexp/regexp-compiler.cc b/js/src/regexp/regexp-compiler.cc
index 9a2aa30dc..c0070061f 100644
--- a/js/src/regexp/regexp-compiler.cc
+++ b/js/src/regexp/regexp-compiler.cc
@@ -5,7 +5,9 @@
#include "regexp/regexp-compiler.h"
#include "regexp/regexp-macro-assembler-arch.h"
-#include "regexp/regexp-macro-assembler-tracer.h"
+#ifdef V8_INTL_SUPPORT
+#include "regexp/special-case.h"
+#endif // V8_INTL_SUPPORT
#ifdef V8_INTL_SUPPORT
#include "unicode/locid.h"
@@ -237,20 +239,15 @@ RegExpCompiler::RegExpCompiler(Isolate* isolate, Zone* zone, int capture_count,
RegExpCompiler::CompilationResult RegExpCompiler::Assemble(
Isolate* isolate, RegExpMacroAssembler* macro_assembler, RegExpNode* start,
int capture_count, Handle<String> pattern) {
-#ifdef DEBUG
- if (FLAG_trace_regexp_assembler)
- macro_assembler_ = new RegExpMacroAssemblerTracer(isolate, macro_assembler);
- else
-#endif
- macro_assembler_ = macro_assembler;
+ macro_assembler_ = macro_assembler;
- std::vector<RegExpNode*> work_list;
+ ZoneVector<RegExpNode*> work_list(zone());
work_list_ = &work_list;
Label fail;
macro_assembler_->PushBacktrack(&fail);
Trace new_trace;
start->Emit(this, &new_trace);
- macro_assembler_->Bind(&fail);
+ macro_assembler_->BindJumpTarget(&fail);
macro_assembler_->Fail();
while (!work_list.empty()) {
RegExpNode* node = work_list.back();
@@ -264,14 +261,9 @@ RegExpCompiler::CompilationResult RegExpCompiler::Assemble(
}
Handle<HeapObject> code = macro_assembler_->GetCode(pattern);
- isolate->IncreaseTotalRegexpCodeGenerated(code->Size());
+ isolate->IncreaseTotalRegexpCodeGenerated(code);
work_list_ = nullptr;
-#ifdef DEBUG
- if (FLAG_trace_regexp_assembler) {
- delete macro_assembler_;
- }
-#endif
return {*code, next_register_};
}
@@ -557,7 +549,7 @@ void Trace::Flush(RegExpCompiler* compiler, RegExpNode* successor) {
}
// On backtrack we need to restore state.
- assembler->Bind(&undo);
+ assembler->BindJumpTarget(&undo);
RestoreAffectedRegisters(assembler, max_register, registers_to_pop,
registers_to_clear);
if (backtrack() == nullptr) {
@@ -720,32 +712,34 @@ static int GetCaseIndependentLetters(Isolate* isolate, uc16 character,
unibrow::uchar* letters,
int letter_length) {
#ifdef V8_INTL_SUPPORT
- // Special case for U+017F which has upper case in ASCII range.
- if (character == 0x017f) {
+ if (RegExpCaseFolding::IgnoreSet().contains(character)) {
letters[0] = character;
return 1;
}
+ bool in_special_add_set =
+ RegExpCaseFolding::SpecialAddSet().contains(character);
+
icu::UnicodeSet set;
set.add(character);
set = set.closeOver(USET_CASE_INSENSITIVE);
+
+ UChar32 canon = 0;
+ if (in_special_add_set) {
+ canon = RegExpCaseFolding::Canonicalize(character);
+ }
+
int32_t range_count = set.getRangeCount();
int items = 0;
for (int32_t i = 0; i < range_count; i++) {
UChar32 start = set.getRangeStart(i);
UChar32 end = set.getRangeEnd(i);
CHECK(end - start + items <= letter_length);
- // Only add to the output if character is not in ASCII range
- // or the case equivalent character is in ASCII range.
- // #sec-runtime-semantics-canonicalize-ch
- // 3.g If the numeric value of ch ≥ 128 and the numeric value of cu < 128,
- // return ch.
- if (!((start >= 128) && (character < 128))) {
- // No range have start and end span across code point 128.
- DCHECK((start >= 128) == (end >= 128));
- for (UChar32 cu = start; cu <= end; cu++) {
- if (one_byte_subject && cu > String::kMaxOneByteCharCode) break;
- letters[items++] = (unibrow::uchar)(cu);
+ for (UChar32 cu = start; cu <= end; cu++) {
+ if (one_byte_subject && cu > String::kMaxOneByteCharCode) break;
+ if (in_special_add_set && RegExpCaseFolding::Canonicalize(cu) != canon) {
+ continue;
}
+ letters[items++] = (unibrow::uchar)(cu);
}
}
return items;
@@ -852,10 +846,6 @@ static bool ShortCutEmitCharacterPair(RegExpMacroAssembler* macro_assembler,
return false;
}
-using EmitCharacterFunction = bool(Isolate* isolate, RegExpCompiler* compiler,
- uc16 c, Label* on_failure, int cp_offset,
- bool check, bool preloaded);
-
// Only emits letters (things that have case). Only used for case independent
// matches.
static inline bool EmitAtomLetter(Isolate* isolate, RegExpCompiler* compiler,
@@ -1843,13 +1833,13 @@ RegExpNode* TextNode::FilterOneByte(int depth) {
if (elm.text_type() == TextElement::ATOM) {
Vector<const uc16> quarks = elm.atom()->data();
for (int j = 0; j < quarks.length(); j++) {
- uint16_t c = quarks[j];
+ uc16 c = quarks[j];
if (elm.atom()->ignore_case()) {
c = unibrow::Latin1::TryConvertToLatin1(c);
}
if (c > unibrow::Latin1::kMaxChar) return set_replacement(nullptr);
// Replace quark in case we converted to Latin-1.
- uint16_t* writable_quarks = const_cast<uint16_t*>(quarks.begin());
+ uc16* writable_quarks = const_cast<uc16*>(quarks.begin());
writable_quarks[j] = c;
}
} else {
@@ -2304,7 +2294,6 @@ void TextNode::TextEmitPass(RegExpCompiler* compiler, TextEmitPassType pass,
for (int j = preloaded ? 0 : quarks.length() - 1; j >= 0; j--) {
if (first_element_checked && i == 0 && j == 0) continue;
if (DeterminedAlready(quick_check, elm.cp_offset() + j)) continue;
- EmitCharacterFunction* emit_function = nullptr;
uc16 quark = quarks[j];
if (elm.atom()->ignore_case()) {
// Everywhere else we assume that a non-Latin-1 character cannot match
@@ -2312,6 +2301,9 @@ void TextNode::TextEmitPass(RegExpCompiler* compiler, TextEmitPassType pass,
// invalid by using the Latin1 equivalent instead.
quark = unibrow::Latin1::TryConvertToLatin1(quark);
}
+ bool needs_bounds_check =
+ *checked_up_to < cp_offset + j || read_backward();
+ bool bounds_checked = false;
switch (pass) {
case NON_LATIN1_MATCH:
DCHECK(one_byte);
@@ -2321,24 +2313,24 @@ void TextNode::TextEmitPass(RegExpCompiler* compiler, TextEmitPassType pass,
}
break;
case NON_LETTER_CHARACTER_MATCH:
- emit_function = &EmitAtomNonLetter;
+ bounds_checked =
+ EmitAtomNonLetter(isolate, compiler, quark, backtrack,
+ cp_offset + j, needs_bounds_check, preloaded);
break;
case SIMPLE_CHARACTER_MATCH:
- emit_function = &EmitSimpleCharacter;
+ bounds_checked = EmitSimpleCharacter(isolate, compiler, quark,
+ backtrack, cp_offset + j,
+ needs_bounds_check, preloaded);
break;
case CASE_CHARACTER_MATCH:
- emit_function = &EmitAtomLetter;
+ bounds_checked =
+ EmitAtomLetter(isolate, compiler, quark, backtrack,
+ cp_offset + j, needs_bounds_check, preloaded);
break;
default:
break;
}
- if (emit_function != nullptr) {
- bool bounds_check = *checked_up_to < cp_offset + j || read_backward();
- bool bound_checked =
- emit_function(isolate, compiler, quark, backtrack, cp_offset + j,
- bounds_check, preloaded);
- if (bound_checked) UpdateBoundsCheck(cp_offset + j, checked_up_to);
- }
+ if (bounds_checked) UpdateBoundsCheck(cp_offset + j, checked_up_to);
}
} else {
DCHECK_EQ(TextElement::CHAR_CLASS, elm.text_type());
@@ -3424,8 +3416,8 @@ void BackReferenceNode::Emit(RegExpCompiler* compiler, Trace* trace) {
DCHECK_EQ(start_reg_ + 1, end_reg_);
if (IgnoreCase(flags_)) {
- assembler->CheckNotBackReferenceIgnoreCase(
- start_reg_, read_backward(), IsUnicode(flags_), trace->backtrack());
+ assembler->CheckNotBackReferenceIgnoreCase(start_reg_, read_backward(),
+ trace->backtrack());
} else {
assembler->CheckNotBackReference(start_reg_, read_backward(),
trace->backtrack());
@@ -3597,12 +3589,17 @@ template <typename... Propagators>
class Analysis : public NodeVisitor {
public:
Analysis(Isolate* isolate, bool is_one_byte)
- : isolate_(isolate), is_one_byte_(is_one_byte), error_message_(nullptr) {}
+ : isolate_(isolate),
+ is_one_byte_(is_one_byte),
+ error_(RegExpError::kNone) {}
void EnsureAnalyzed(RegExpNode* that) {
StackLimitCheck check(isolate());
if (check.HasOverflowed()) {
- fail("Stack overflow");
+ if (FLAG_correctness_fuzzer_suppressions) {
+ FATAL("Analysis: Aborting on stack overflow");
+ }
+ fail(RegExpError::kAnalysisStackOverflow);
return;
}
if (that->info()->been_analyzed || that->info()->being_analyzed) return;
@@ -3612,12 +3609,12 @@ class Analysis : public NodeVisitor {
that->info()->been_analyzed = true;
}
- bool has_failed() { return error_message_ != nullptr; }
- const char* error_message() {
- DCHECK(error_message_ != nullptr);
- return error_message_;
+ bool has_failed() { return error_ != RegExpError::kNone; }
+ RegExpError error() {
+ DCHECK(error_ != RegExpError::kNone);
+ return error_;
}
- void fail(const char* error_message) { error_message_ = error_message; }
+ void fail(RegExpError error) { error_ = error; }
Isolate* isolate() const { return isolate_; }
@@ -3702,19 +3699,19 @@ class Analysis : public NodeVisitor {
private:
Isolate* isolate_;
bool is_one_byte_;
- const char* error_message_;
+ RegExpError error_;
DISALLOW_IMPLICIT_CONSTRUCTORS(Analysis);
};
-const char* AnalyzeRegExp(Isolate* isolate, bool is_one_byte,
+RegExpError AnalyzeRegExp(Isolate* isolate, bool is_one_byte,
RegExpNode* node) {
Analysis<AssertionPropagator, EatsAtLeastPropagator> analysis(isolate,
is_one_byte);
DCHECK_EQ(node->info()->been_analyzed, false);
analysis.EnsureAnalyzed(node);
- DCHECK_IMPLIES(analysis.has_failed(), analysis.error_message() != nullptr);
- return analysis.has_failed() ? analysis.error_message() : nullptr;
+ DCHECK_IMPLIES(analysis.has_failed(), analysis.error() != RegExpError::kNone);
+ return analysis.has_failed() ? analysis.error() : RegExpError::kNone;
}
void BackReferenceNode::FillInBMInfo(Isolate* isolate, int offset, int budget,
generated by cgit v1.2.3 (git 2.25.1) at 2024-07-28 23:28:13 +0000