| File: | var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc |
| Warning: | line 1466, column 38 The result of left shift is undefined because the right operand '32' is not smaller than 32, the capacity of 'uint32_t' |
Press '?' to see keyboard shortcuts
Keyboard shortcuts:
| 1 | // Copyright 2019 the V8 project authors. All rights reserved. | |||
| 2 | // Use of this source code is governed by a BSD-style license that can be | |||
| 3 | // found in the LICENSE file. | |||
| 4 | ||||
| 5 | #include "irregexp/imported/regexp-compiler.h" | |||
| 6 | ||||
| 7 | #include "irregexp/imported/regexp-macro-assembler-arch.h" | |||
| 8 | ||||
| 9 | #ifdef V8_INTL_SUPPORT1 | |||
| 10 | #include "irregexp/imported/special-case.h" | |||
| 11 | #include "unicode/locid.h" | |||
| 12 | #include "unicode/uniset.h" | |||
| 13 | #include "unicode/utypes.h" | |||
| 14 | #endif // V8_INTL_SUPPORT | |||
| 15 | ||||
| 16 | namespace v8 { | |||
| 17 | namespace internal { | |||
| 18 | ||||
| 19 | using namespace regexp_compiler_constants; // NOLINT(build/namespaces) | |||
| 20 | ||||
| 21 | // ------------------------------------------------------------------- | |||
| 22 | // Implementation of the Irregexp regular expression engine. | |||
| 23 | // | |||
| 24 | // The Irregexp regular expression engine is intended to be a complete | |||
| 25 | // implementation of ECMAScript regular expressions. It generates either | |||
| 26 | // bytecodes or native code. | |||
| 27 | ||||
| 28 | // The Irregexp regexp engine is structured in three steps. | |||
| 29 | // 1) The parser generates an abstract syntax tree. See ast.cc. | |||
| 30 | // 2) From the AST a node network is created. The nodes are all | |||
| 31 | // subclasses of RegExpNode. The nodes represent states when | |||
| 32 | // executing a regular expression. Several optimizations are | |||
| 33 | // performed on the node network. | |||
| 34 | // 3) From the nodes we generate either byte codes or native code | |||
| 35 | // that can actually execute the regular expression (perform | |||
| 36 | // the search). The code generation step is described in more | |||
| 37 | // detail below. | |||
| 38 | ||||
| 39 | // Code generation. | |||
| 40 | // | |||
| 41 | // The nodes are divided into four main categories. | |||
| 42 | // * Choice nodes | |||
| 43 | // These represent places where the regular expression can | |||
| 44 | // match in more than one way. For example on entry to an | |||
| 45 | // alternation (foo|bar) or a repetition (*, +, ? or {}). | |||
| 46 | // * Action nodes | |||
| 47 | // These represent places where some action should be | |||
| 48 | // performed. Examples include recording the current position | |||
| 49 | // in the input string to a register (in order to implement | |||
| 50 | // captures) or other actions on register for example in order | |||
| 51 | // to implement the counters needed for {} repetitions. | |||
| 52 | // * Matching nodes | |||
| 53 | // These attempt to match some element part of the input string. | |||
| 54 | // Examples of elements include character classes, plain strings | |||
| 55 | // or back references. | |||
| 56 | // * End nodes | |||
| 57 | // These are used to implement the actions required on finding | |||
| 58 | // a successful match or failing to find a match. | |||
| 59 | // | |||
| 60 | // The code generated (whether as byte codes or native code) maintains | |||
| 61 | // some state as it runs. This consists of the following elements: | |||
| 62 | // | |||
| 63 | // * The capture registers. Used for string captures. | |||
| 64 | // * Other registers. Used for counters etc. | |||
| 65 | // * The current position. | |||
| 66 | // * The stack of backtracking information. Used when a matching node | |||
| 67 | // fails to find a match and needs to try an alternative. | |||
| 68 | // | |||
| 69 | // Conceptual regular expression execution model: | |||
| 70 | // | |||
| 71 | // There is a simple conceptual model of regular expression execution | |||
| 72 | // which will be presented first. The actual code generated is a more | |||
| 73 | // efficient simulation of the simple conceptual model: | |||
| 74 | // | |||
| 75 | // * Choice nodes are implemented as follows: | |||
| 76 | // For each choice except the last { | |||
| 77 | // push current position | |||
| 78 | // push backtrack code location | |||
| 79 | // <generate code to test for choice> | |||
| 80 | // backtrack code location: | |||
| 81 | // pop current position | |||
| 82 | // } | |||
| 83 | // <generate code to test for last choice> | |||
| 84 | // | |||
| 85 | // * Actions nodes are generated as follows | |||
| 86 | // <push affected registers on backtrack stack> | |||
| 87 | // <generate code to perform action> | |||
| 88 | // push backtrack code location | |||
| 89 | // <generate code to test for following nodes> | |||
| 90 | // backtrack code location: | |||
| 91 | // <pop affected registers to restore their state> | |||
| 92 | // <pop backtrack location from stack and go to it> | |||
| 93 | // | |||
| 94 | // * Matching nodes are generated as follows: | |||
| 95 | // if input string matches at current position | |||
| 96 | // update current position | |||
| 97 | // <generate code to test for following nodes> | |||
| 98 | // else | |||
| 99 | // <pop backtrack location from stack and go to it> | |||
| 100 | // | |||
| 101 | // Thus it can be seen that the current position is saved and restored | |||
| 102 | // by the choice nodes, whereas the registers are saved and restored by | |||
| 103 | // by the action nodes that manipulate them. | |||
| 104 | // | |||
| 105 | // The other interesting aspect of this model is that nodes are generated | |||
| 106 | // at the point where they are needed by a recursive call to Emit(). If | |||
| 107 | // the node has already been code generated then the Emit() call will | |||
| 108 | // generate a jump to the previously generated code instead. In order to | |||
| 109 | // limit recursion it is possible for the Emit() function to put the node | |||
| 110 | // on a work list for later generation and instead generate a jump. The | |||
| 111 | // destination of the jump is resolved later when the code is generated. | |||
| 112 | // | |||
| 113 | // Actual regular expression code generation. | |||
| 114 | // | |||
| 115 | // Code generation is actually more complicated than the above. In order to | |||
| 116 | // improve the efficiency of the generated code some optimizations are | |||
| 117 | // performed | |||
| 118 | // | |||
| 119 | // * Choice nodes have 1-character lookahead. | |||
| 120 | // A choice node looks at the following character and eliminates some of | |||
| 121 | // the choices immediately based on that character. This is not yet | |||
| 122 | // implemented. | |||
| 123 | // * Simple greedy loops store reduced backtracking information. | |||
| 124 | // A quantifier like /.*foo/m will greedily match the whole input. It will | |||
| 125 | // then need to backtrack to a point where it can match "foo". The naive | |||
| 126 | // implementation of this would push each character position onto the | |||
| 127 | // backtracking stack, then pop them off one by one. This would use space | |||
| 128 | // proportional to the length of the input string. However since the "." | |||
| 129 | // can only match in one way and always has a constant length (in this case | |||
| 130 | // of 1) it suffices to store the current position on the top of the stack | |||
| 131 | // once. Matching now becomes merely incrementing the current position and | |||
| 132 | // backtracking becomes decrementing the current position and checking the | |||
| 133 | // result against the stored current position. This is faster and saves | |||
| 134 | // space. | |||
| 135 | // * The current state is virtualized. | |||
| 136 | // This is used to defer expensive operations until it is clear that they | |||
| 137 | // are needed and to generate code for a node more than once, allowing | |||
| 138 | // specialized an efficient versions of the code to be created. This is | |||
| 139 | // explained in the section below. | |||
| 140 | // | |||
| 141 | // Execution state virtualization. | |||
| 142 | // | |||
| 143 | // Instead of emitting code, nodes that manipulate the state can record their | |||
| 144 | // manipulation in an object called the Trace. The Trace object can record a | |||
| 145 | // current position offset, an optional backtrack code location on the top of | |||
| 146 | // the virtualized backtrack stack and some register changes. When a node is | |||
| 147 | // to be emitted it can flush the Trace or update it. Flushing the Trace | |||
| 148 | // will emit code to bring the actual state into line with the virtual state. | |||
| 149 | // Avoiding flushing the state can postpone some work (e.g. updates of capture | |||
| 150 | // registers). Postponing work can save time when executing the regular | |||
| 151 | // expression since it may be found that the work never has to be done as a | |||
| 152 | // failure to match can occur. In addition it is much faster to jump to a | |||
| 153 | // known backtrack code location than it is to pop an unknown backtrack | |||
| 154 | // location from the stack and jump there. | |||
| 155 | // | |||
| 156 | // The virtual state found in the Trace affects code generation. For example | |||
| 157 | // the virtual state contains the difference between the actual current | |||
| 158 | // position and the virtual current position, and matching code needs to use | |||
| 159 | // this offset to attempt a match in the correct location of the input | |||
| 160 | // string. Therefore code generated for a non-trivial trace is specialized | |||
| 161 | // to that trace. The code generator therefore has the ability to generate | |||
| 162 | // code for each node several times. In order to limit the size of the | |||
| 163 | // generated code there is an arbitrary limit on how many specialized sets of | |||
| 164 | // code may be generated for a given node. If the limit is reached, the | |||
| 165 | // trace is flushed and a generic version of the code for a node is emitted. | |||
| 166 | // This is subsequently used for that node. The code emitted for non-generic | |||
| 167 | // trace is not recorded in the node and so it cannot currently be reused in | |||
| 168 | // the event that code generation is requested for an identical trace. | |||
| 169 | ||||
| 170 | namespace { | |||
| 171 | ||||
| 172 | constexpr base::uc32 MaxCodeUnit(const bool one_byte) { | |||
| 173 | static_assert(String::kMaxOneByteCharCodeU <= | |||
| 174 | std::numeric_limits<uint16_t>::max()); | |||
| 175 | static_assert(String::kMaxUtf16CodeUnitU <= | |||
| 176 | std::numeric_limits<uint16_t>::max()); | |||
| 177 | return one_byte ? String::kMaxOneByteCharCodeU : String::kMaxUtf16CodeUnitU; | |||
| 178 | } | |||
| 179 | ||||
| 180 | constexpr uint32_t CharMask(const bool one_byte) { | |||
| 181 | static_assert(base::bits::IsPowerOfTwo(String::kMaxOneByteCharCodeU + 1)); | |||
| 182 | static_assert(base::bits::IsPowerOfTwo(String::kMaxUtf16CodeUnitU + 1)); | |||
| 183 | return MaxCodeUnit(one_byte); | |||
| 184 | } | |||
| 185 | ||||
| 186 | } // namespace | |||
| 187 | ||||
| 188 | void RegExpTree::AppendToText(RegExpText* text, Zone* zone) { UNREACHABLE()do { do { } while (false); MOZ_ReportCrash("" "unreachable code" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 188); AnnotateMozCrashReason("MOZ_CRASH(" "unreachable code" ")"); do { *((volatile int*)__null) = 188; __attribute__((nomerge )) ::abort(); } while (false); } while (false); } | |||
| 189 | ||||
| 190 | void RegExpAtom::AppendToText(RegExpText* text, Zone* zone) { | |||
| 191 | text->AddElement(TextElement::Atom(this), zone); | |||
| 192 | } | |||
| 193 | ||||
| 194 | void RegExpClassRanges::AppendToText(RegExpText* text, Zone* zone) { | |||
| 195 | text->AddElement(TextElement::ClassRanges(this), zone); | |||
| 196 | } | |||
| 197 | ||||
| 198 | void RegExpText::AppendToText(RegExpText* text, Zone* zone) { | |||
| 199 | for (int i = 0; i < elements()->length(); i++) | |||
| 200 | text->AddElement(elements()->at(i), zone); | |||
| 201 | } | |||
| 202 | ||||
| 203 | TextElement TextElement::Atom(RegExpAtom* atom) { | |||
| 204 | return TextElement(ATOM, atom); | |||
| 205 | } | |||
| 206 | ||||
| 207 | TextElement TextElement::ClassRanges(RegExpClassRanges* class_ranges) { | |||
| 208 | return TextElement(CLASS_RANGES, class_ranges); | |||
| 209 | } | |||
| 210 | ||||
| 211 | int TextElement::length() const { | |||
| 212 | switch (text_type()) { | |||
| 213 | case ATOM: | |||
| 214 | return atom()->length(); | |||
| 215 | ||||
| 216 | case CLASS_RANGES: | |||
| 217 | return 1; | |||
| 218 | } | |||
| 219 | UNREACHABLE()do { do { } while (false); MOZ_ReportCrash("" "unreachable code" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 219); AnnotateMozCrashReason("MOZ_CRASH(" "unreachable code" ")"); do { *((volatile int*)__null) = 219; __attribute__((nomerge )) ::abort(); } while (false); } while (false); | |||
| 220 | } | |||
| 221 | ||||
| 222 | class RecursionCheck { | |||
| 223 | public: | |||
| 224 | explicit RecursionCheck(RegExpCompiler* compiler) : compiler_(compiler) { | |||
| 225 | compiler->IncrementRecursionDepth(); | |||
| 226 | } | |||
| 227 | ~RecursionCheck() { compiler_->DecrementRecursionDepth(); } | |||
| 228 | ||||
| 229 | private: | |||
| 230 | RegExpCompiler* compiler_; | |||
| 231 | }; | |||
| 232 | ||||
| 233 | // Attempts to compile the regexp using an Irregexp code generator. Returns | |||
| 234 | // a fixed array or a null handle depending on whether it succeeded. | |||
| 235 | RegExpCompiler::RegExpCompiler(Isolate* isolate, Zone* zone, int capture_count, | |||
| 236 | RegExpFlags flags, bool one_byte) | |||
| 237 | : next_register_(JSRegExp::RegistersForCaptureCount(capture_count)), | |||
| 238 | unicode_lookaround_stack_register_(kNoRegister), | |||
| 239 | unicode_lookaround_position_register_(kNoRegister), | |||
| 240 | work_list_(nullptr), | |||
| 241 | recursion_depth_(0), | |||
| 242 | flags_(flags), | |||
| 243 | one_byte_(one_byte), | |||
| 244 | reg_exp_too_big_(false), | |||
| 245 | limiting_recursion_(false), | |||
| 246 | optimize_(v8_flagsjs::jit::JitOptions.regexp_optimization), | |||
| 247 | read_backward_(false), | |||
| 248 | current_expansion_factor_(1), | |||
| 249 | frequency_collator_(), | |||
| 250 | isolate_(isolate), | |||
| 251 | zone_(zone) { | |||
| 252 | accept_ = zone->New<EndNode>(EndNode::ACCEPT, zone); | |||
| 253 | DCHECK_GE(RegExpMacroAssembler::kMaxRegister, next_register_ - 1)do { static_assert( mozilla::detail::AssertionConditionType< decltype((RegExpMacroAssembler::kMaxRegister) >= (next_register_ - 1))>::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!((RegExpMacroAssembler::kMaxRegister) >= (next_register_ - 1)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(RegExpMacroAssembler::kMaxRegister) >= (next_register_ - 1)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 253); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(RegExpMacroAssembler::kMaxRegister) >= (next_register_ - 1)" ")"); do { *((volatile int*)__null) = 253; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 254 | } | |||
| 255 | ||||
| 256 | RegExpCompiler::CompilationResult RegExpCompiler::Assemble( | |||
| 257 | Isolate* isolate, RegExpMacroAssembler* macro_assembler, RegExpNode* start, | |||
| 258 | int capture_count, Handle<String> pattern) { | |||
| 259 | macro_assembler_ = macro_assembler; | |||
| 260 | ||||
| 261 | ZoneVector<RegExpNode*> work_list(zone()); | |||
| 262 | work_list_ = &work_list; | |||
| 263 | Label fail; | |||
| 264 | macro_assembler_->PushBacktrack(&fail); | |||
| 265 | Trace new_trace; | |||
| 266 | start->Emit(this, &new_trace); | |||
| 267 | macro_assembler_->BindJumpTarget(&fail); | |||
| 268 | macro_assembler_->Fail(); | |||
| 269 | while (!work_list.empty()) { | |||
| 270 | RegExpNode* node = work_list.back(); | |||
| 271 | work_list.pop_back(); | |||
| 272 | node->set_on_work_list(false); | |||
| 273 | if (!node->label()->is_bound()) node->Emit(this, &new_trace); | |||
| 274 | } | |||
| 275 | if (reg_exp_too_big_) { | |||
| 276 | if (v8_flagsjs::jit::JitOptions.correctness_fuzzer_suppressions) { | |||
| 277 | FATAL("Aborting on excess zone allocation")do { do { } while (false); MOZ_ReportCrash("" "Aborting on excess zone allocation" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 277); AnnotateMozCrashReason("MOZ_CRASH(" "Aborting on excess zone allocation" ")"); do { *((volatile int*)__null) = 277; __attribute__((nomerge )) ::abort(); } while (false); } while (false); | |||
| 278 | } | |||
| 279 | macro_assembler_->AbortedCodeGeneration(); | |||
| 280 | return CompilationResult::RegExpTooBig(); | |||
| 281 | } | |||
| 282 | ||||
| 283 | Handle<HeapObject> code = macro_assembler_->GetCode(pattern); | |||
| 284 | isolate->IncreaseTotalRegexpCodeGenerated(code); | |||
| 285 | work_list_ = nullptr; | |||
| 286 | ||||
| 287 | return {code, next_register_}; | |||
| 288 | } | |||
| 289 | ||||
| 290 | bool Trace::DeferredAction::Mentions(int that) { | |||
| 291 | if (action_type() == ActionNode::CLEAR_CAPTURES) { | |||
| 292 | Interval range = static_cast<DeferredClearCaptures*>(this)->range(); | |||
| 293 | return range.Contains(that); | |||
| 294 | } else { | |||
| 295 | return reg() == that; | |||
| 296 | } | |||
| 297 | } | |||
| 298 | ||||
| 299 | bool Trace::mentions_reg(int reg) { | |||
| 300 | for (DeferredAction* action = actions_; action != nullptr; | |||
| 301 | action = action->next()) { | |||
| 302 | if (action->Mentions(reg)) return true; | |||
| 303 | } | |||
| 304 | return false; | |||
| 305 | } | |||
| 306 | ||||
| 307 | bool Trace::GetStoredPosition(int reg, int* cp_offset) { | |||
| 308 | DCHECK_EQ(0, *cp_offset)do { static_assert( mozilla::detail::AssertionConditionType< decltype((0) == (*cp_offset))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((0) == (*cp_offset)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(0) == (*cp_offset)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 308); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(0) == (*cp_offset)" ")"); do { *((volatile int*)__null) = 308; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 309 | for (DeferredAction* action = actions_; action != nullptr; | |||
| 310 | action = action->next()) { | |||
| 311 | if (action->Mentions(reg)) { | |||
| 312 | if (action->action_type() == ActionNode::STORE_POSITION) { | |||
| 313 | *cp_offset = static_cast<DeferredCapture*>(action)->cp_offset(); | |||
| 314 | return true; | |||
| 315 | } else { | |||
| 316 | return false; | |||
| 317 | } | |||
| 318 | } | |||
| 319 | } | |||
| 320 | return false; | |||
| 321 | } | |||
| 322 | ||||
| 323 | // A (dynamically-sized) set of unsigned integers that behaves especially well | |||
| 324 | // on small integers (< kFirstLimit). May do zone-allocation. | |||
| 325 | class DynamicBitSet : public ZoneObject { | |||
| 326 | public: | |||
| 327 | V8_EXPORT_PRIVATE bool Get(unsigned value) const { | |||
| 328 | if (value < kFirstLimit) { | |||
| 329 | return (first_ & (1 << value)) != 0; | |||
| 330 | } else if (remaining_ == nullptr) { | |||
| 331 | return false; | |||
| 332 | } else { | |||
| 333 | return remaining_->Contains(value); | |||
| 334 | } | |||
| 335 | } | |||
| 336 | ||||
| 337 | // Destructively set a value in this set. | |||
| 338 | void Set(unsigned value, Zone* zone) { | |||
| 339 | if (value < kFirstLimit) { | |||
| 340 | first_ |= (1 << value); | |||
| 341 | } else { | |||
| 342 | if (remaining_ == nullptr) | |||
| 343 | remaining_ = zone->New<ZoneList<unsigned>>(1, zone); | |||
| 344 | if (remaining_->is_empty() || !remaining_->Contains(value)) | |||
| 345 | remaining_->Add(value, zone); | |||
| 346 | } | |||
| 347 | } | |||
| 348 | ||||
| 349 | private: | |||
| 350 | static constexpr unsigned kFirstLimit = 32; | |||
| 351 | ||||
| 352 | uint32_t first_ = 0; | |||
| 353 | ZoneList<unsigned>* remaining_ = nullptr; | |||
| 354 | }; | |||
| 355 | ||||
| 356 | int Trace::FindAffectedRegisters(DynamicBitSet* affected_registers, | |||
| 357 | Zone* zone) { | |||
| 358 | int max_register = RegExpCompiler::kNoRegister; | |||
| 359 | for (DeferredAction* action = actions_; action != nullptr; | |||
| 360 | action = action->next()) { | |||
| 361 | if (action->action_type() == ActionNode::CLEAR_CAPTURES) { | |||
| 362 | Interval range = static_cast<DeferredClearCaptures*>(action)->range(); | |||
| 363 | for (int i = range.from(); i <= range.to(); i++) | |||
| 364 | affected_registers->Set(i, zone); | |||
| 365 | if (range.to() > max_register) max_register = range.to(); | |||
| 366 | } else { | |||
| 367 | affected_registers->Set(action->reg(), zone); | |||
| 368 | if (action->reg() > max_register) max_register = action->reg(); | |||
| 369 | } | |||
| 370 | } | |||
| 371 | return max_register; | |||
| 372 | } | |||
| 373 | ||||
| 374 | void Trace::RestoreAffectedRegisters(RegExpMacroAssembler* assembler, | |||
| 375 | int max_register, | |||
| 376 | const DynamicBitSet& registers_to_pop, | |||
| 377 | const DynamicBitSet& registers_to_clear) { | |||
| 378 | for (int reg = max_register; reg >= 0; reg--) { | |||
| 379 | if (registers_to_pop.Get(reg)) { | |||
| 380 | assembler->PopRegister(reg); | |||
| 381 | } else if (registers_to_clear.Get(reg)) { | |||
| 382 | int clear_to = reg; | |||
| 383 | while (reg > 0 && registers_to_clear.Get(reg - 1)) { | |||
| 384 | reg--; | |||
| 385 | } | |||
| 386 | assembler->ClearRegisters(reg, clear_to); | |||
| 387 | } | |||
| 388 | } | |||
| 389 | } | |||
| 390 | ||||
| 391 | void Trace::PerformDeferredActions(RegExpMacroAssembler* assembler, | |||
| 392 | int max_register, | |||
| 393 | const DynamicBitSet& affected_registers, | |||
| 394 | DynamicBitSet* registers_to_pop, | |||
| 395 | DynamicBitSet* registers_to_clear, | |||
| 396 | Zone* zone) { | |||
| 397 | // The "+1" is to avoid a push_limit of zero if stack_limit_slack() is 1. | |||
| 398 | const int push_limit = (assembler->stack_limit_slack() + 1) / 2; | |||
| 399 | ||||
| 400 | // Count pushes performed to force a stack limit check occasionally. | |||
| 401 | int pushes = 0; | |||
| 402 | ||||
| 403 | for (int reg = 0; reg <= max_register; reg++) { | |||
| 404 | if (!affected_registers.Get(reg)) continue; | |||
| 405 | ||||
| 406 | // The chronologically first deferred action in the trace | |||
| 407 | // is used to infer the action needed to restore a register | |||
| 408 | // to its previous state (or not, if it's safe to ignore it). | |||
| 409 | enum DeferredActionUndoType { IGNORE, RESTORE, CLEAR }; | |||
| 410 | DeferredActionUndoType undo_action = IGNORE; | |||
| 411 | ||||
| 412 | int value = 0; | |||
| 413 | bool absolute = false; | |||
| 414 | bool clear = false; | |||
| 415 | static const int kNoStore = kMinInt((int32_t)0x80000000); | |||
| 416 | int store_position = kNoStore; | |||
| 417 | // This is a little tricky because we are scanning the actions in reverse | |||
| 418 | // historical order (newest first). | |||
| 419 | for (DeferredAction* action = actions_; action != nullptr; | |||
| 420 | action = action->next()) { | |||
| 421 | if (action->Mentions(reg)) { | |||
| 422 | switch (action->action_type()) { | |||
| 423 | case ActionNode::SET_REGISTER_FOR_LOOP: { | |||
| 424 | Trace::DeferredSetRegisterForLoop* psr = | |||
| 425 | static_cast<Trace::DeferredSetRegisterForLoop*>(action); | |||
| 426 | if (!absolute) { | |||
| 427 | value += psr->value(); | |||
| 428 | absolute = true; | |||
| 429 | } | |||
| 430 | // SET_REGISTER_FOR_LOOP is only used for newly introduced loop | |||
| 431 | // counters. They can have a significant previous value if they | |||
| 432 | // occur in a loop. TODO(lrn): Propagate this information, so | |||
| 433 | // we can set undo_action to IGNORE if we know there is no value to | |||
| 434 | // restore. | |||
| 435 | undo_action = RESTORE; | |||
| 436 | DCHECK_EQ(store_position, kNoStore)do { static_assert( mozilla::detail::AssertionConditionType< decltype((store_position) == (kNoStore))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((store_position) == (kNoStore )))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(store_position) == (kNoStore)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 436); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(store_position) == (kNoStore)" ")"); do { *((volatile int*)__null) = 436; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 437 | DCHECK(!clear)do { static_assert( mozilla::detail::AssertionConditionType< decltype(!clear)>::isValid, "invalid assertion condition") ; if ((__builtin_expect(!!(!(!!(!clear))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("!clear", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 437); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!clear" ")") ; do { *((volatile int*)__null) = 437; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 438 | break; | |||
| 439 | } | |||
| 440 | case ActionNode::INCREMENT_REGISTER: | |||
| 441 | if (!absolute) { | |||
| 442 | value++; | |||
| 443 | } | |||
| 444 | DCHECK_EQ(store_position, kNoStore)do { static_assert( mozilla::detail::AssertionConditionType< decltype((store_position) == (kNoStore))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((store_position) == (kNoStore )))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(store_position) == (kNoStore)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 444); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(store_position) == (kNoStore)" ")"); do { *((volatile int*)__null) = 444; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 445 | DCHECK(!clear)do { static_assert( mozilla::detail::AssertionConditionType< decltype(!clear)>::isValid, "invalid assertion condition") ; if ((__builtin_expect(!!(!(!!(!clear))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("!clear", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 445); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!clear" ")") ; do { *((volatile int*)__null) = 445; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 446 | undo_action = RESTORE; | |||
| 447 | break; | |||
| 448 | case ActionNode::STORE_POSITION: { | |||
| 449 | Trace::DeferredCapture* pc = | |||
| 450 | static_cast<Trace::DeferredCapture*>(action); | |||
| 451 | if (!clear && store_position == kNoStore) { | |||
| 452 | store_position = pc->cp_offset(); | |||
| 453 | } | |||
| 454 | ||||
| 455 | // For captures we know that stores and clears alternate. | |||
| 456 | // Other register, are never cleared, and if the occur | |||
| 457 | // inside a loop, they might be assigned more than once. | |||
| 458 | if (reg <= 1) { | |||
| 459 | // Registers zero and one, aka "capture zero", is | |||
| 460 | // always set correctly if we succeed. There is no | |||
| 461 | // need to undo a setting on backtrack, because we | |||
| 462 | // will set it again or fail. | |||
| 463 | undo_action = IGNORE; | |||
| 464 | } else { | |||
| 465 | undo_action = pc->is_capture() ? CLEAR : RESTORE; | |||
| 466 | } | |||
| 467 | DCHECK(!absolute)do { static_assert( mozilla::detail::AssertionConditionType< decltype(!absolute)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(!absolute))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("!absolute", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 467); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!absolute" ")" ); do { *((volatile int*)__null) = 467; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 468 | DCHECK_EQ(value, 0)do { static_assert( mozilla::detail::AssertionConditionType< decltype((value) == (0))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((value) == (0)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(value) == (0)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 468); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(value) == (0)" ")"); do { *((volatile int*)__null) = 468; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 469 | break; | |||
| 470 | } | |||
| 471 | case ActionNode::CLEAR_CAPTURES: { | |||
| 472 | // Since we're scanning in reverse order, if we've already | |||
| 473 | // set the position we have to ignore historically earlier | |||
| 474 | // clearing operations. | |||
| 475 | if (store_position == kNoStore) { | |||
| 476 | clear = true; | |||
| 477 | } | |||
| 478 | undo_action = RESTORE; | |||
| 479 | DCHECK(!absolute)do { static_assert( mozilla::detail::AssertionConditionType< decltype(!absolute)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(!absolute))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("!absolute", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 479); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!absolute" ")" ); do { *((volatile int*)__null) = 479; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 480 | DCHECK_EQ(value, 0)do { static_assert( mozilla::detail::AssertionConditionType< decltype((value) == (0))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((value) == (0)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(value) == (0)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 480); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(value) == (0)" ")"); do { *((volatile int*)__null) = 480; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 481 | break; | |||
| 482 | } | |||
| 483 | default: | |||
| 484 | UNREACHABLE()do { do { } while (false); MOZ_ReportCrash("" "unreachable code" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 484); AnnotateMozCrashReason("MOZ_CRASH(" "unreachable code" ")"); do { *((volatile int*)__null) = 484; __attribute__((nomerge )) ::abort(); } while (false); } while (false); | |||
| 485 | } | |||
| 486 | } | |||
| 487 | } | |||
| 488 | // Prepare for the undo-action (e.g., push if it's going to be popped). | |||
| 489 | if (undo_action == RESTORE) { | |||
| 490 | pushes++; | |||
| 491 | RegExpMacroAssembler::StackCheckFlag stack_check = | |||
| 492 | RegExpMacroAssembler::kNoStackLimitCheck; | |||
| 493 | if (pushes == push_limit) { | |||
| 494 | stack_check = RegExpMacroAssembler::kCheckStackLimit; | |||
| 495 | pushes = 0; | |||
| 496 | } | |||
| 497 | ||||
| 498 | assembler->PushRegister(reg, stack_check); | |||
| 499 | registers_to_pop->Set(reg, zone); | |||
| 500 | } else if (undo_action == CLEAR) { | |||
| 501 | registers_to_clear->Set(reg, zone); | |||
| 502 | } | |||
| 503 | // Perform the chronologically last action (or accumulated increment) | |||
| 504 | // for the register. | |||
| 505 | if (store_position != kNoStore) { | |||
| 506 | assembler->WriteCurrentPositionToRegister(reg, store_position); | |||
| 507 | } else if (clear) { | |||
| 508 | assembler->ClearRegisters(reg, reg); | |||
| 509 | } else if (absolute) { | |||
| 510 | assembler->SetRegister(reg, value); | |||
| 511 | } else if (value != 0) { | |||
| 512 | assembler->AdvanceRegister(reg, value); | |||
| 513 | } | |||
| 514 | } | |||
| 515 | } | |||
| 516 | ||||
| 517 | // This is called as we come into a loop choice node and some other tricky | |||
| 518 | // nodes. It normalizes the state of the code generator to ensure we can | |||
| 519 | // generate generic code. | |||
| 520 | void Trace::Flush(RegExpCompiler* compiler, RegExpNode* successor) { | |||
| 521 | RegExpMacroAssembler* assembler = compiler->macro_assembler(); | |||
| 522 | ||||
| 523 | DCHECK(!is_trivial())do { static_assert( mozilla::detail::AssertionConditionType< decltype(!is_trivial())>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(!is_trivial()))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("!is_trivial()", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 523); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!is_trivial()" ")"); do { *((volatile int*)__null) = 523; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 524 | ||||
| 525 | if (actions_ == nullptr && backtrack() == nullptr) { | |||
| 526 | // Here we just have some deferred cp advances to fix and we are back to | |||
| 527 | // a normal situation. We may also have to forget some information gained | |||
| 528 | // through a quick check that was already performed. | |||
| 529 | if (cp_offset_ != 0) assembler->AdvanceCurrentPosition(cp_offset_); | |||
| 530 | // Create a new trivial state and generate the node with that. | |||
| 531 | Trace new_state; | |||
| 532 | successor->Emit(compiler, &new_state); | |||
| 533 | return; | |||
| 534 | } | |||
| 535 | ||||
| 536 | // Generate deferred actions here along with code to undo them again. | |||
| 537 | DynamicBitSet affected_registers; | |||
| 538 | ||||
| 539 | if (backtrack() != nullptr) { | |||
| 540 | // Here we have a concrete backtrack location. These are set up by choice | |||
| 541 | // nodes and so they indicate that we have a deferred save of the current | |||
| 542 | // position which we may need to emit here. | |||
| 543 | assembler->PushCurrentPosition(); | |||
| 544 | } | |||
| 545 | ||||
| 546 | int max_register = | |||
| 547 | FindAffectedRegisters(&affected_registers, compiler->zone()); | |||
| 548 | DynamicBitSet registers_to_pop; | |||
| 549 | DynamicBitSet registers_to_clear; | |||
| 550 | PerformDeferredActions(assembler, max_register, affected_registers, | |||
| 551 | ®isters_to_pop, ®isters_to_clear, | |||
| 552 | compiler->zone()); | |||
| 553 | if (cp_offset_ != 0) { | |||
| 554 | assembler->AdvanceCurrentPosition(cp_offset_); | |||
| 555 | } | |||
| 556 | ||||
| 557 | // Create a new trivial state and generate the node with that. | |||
| 558 | Label undo; | |||
| 559 | assembler->PushBacktrack(&undo); | |||
| 560 | if (successor->KeepRecursing(compiler)) { | |||
| 561 | Trace new_state; | |||
| 562 | successor->Emit(compiler, &new_state); | |||
| 563 | } else { | |||
| 564 | compiler->AddWork(successor); | |||
| 565 | assembler->GoTo(successor->label()); | |||
| 566 | } | |||
| 567 | ||||
| 568 | // On backtrack we need to restore state. | |||
| 569 | assembler->BindJumpTarget(&undo); | |||
| 570 | RestoreAffectedRegisters(assembler, max_register, registers_to_pop, | |||
| 571 | registers_to_clear); | |||
| 572 | if (backtrack() == nullptr) { | |||
| 573 | assembler->Backtrack(); | |||
| 574 | } else { | |||
| 575 | assembler->PopCurrentPosition(); | |||
| 576 | assembler->GoTo(backtrack()); | |||
| 577 | } | |||
| 578 | } | |||
| 579 | ||||
| 580 | void NegativeSubmatchSuccess::Emit(RegExpCompiler* compiler, Trace* trace) { | |||
| 581 | RegExpMacroAssembler* assembler = compiler->macro_assembler(); | |||
| 582 | ||||
| 583 | // Omit flushing the trace. We discard the entire stack frame anyway. | |||
| 584 | ||||
| 585 | if (!label()->is_bound()) { | |||
| 586 | // We are completely independent of the trace, since we ignore it, | |||
| 587 | // so this code can be used as the generic version. | |||
| 588 | assembler->Bind(label()); | |||
| 589 | } | |||
| 590 | ||||
| 591 | // Throw away everything on the backtrack stack since the start | |||
| 592 | // of the negative submatch and restore the character position. | |||
| 593 | assembler->ReadCurrentPositionFromRegister(current_position_register_); | |||
| 594 | assembler->ReadStackPointerFromRegister(stack_pointer_register_); | |||
| 595 | if (clear_capture_count_ > 0) { | |||
| 596 | // Clear any captures that might have been performed during the success | |||
| 597 | // of the body of the negative look-ahead. | |||
| 598 | int clear_capture_end = clear_capture_start_ + clear_capture_count_ - 1; | |||
| 599 | assembler->ClearRegisters(clear_capture_start_, clear_capture_end); | |||
| 600 | } | |||
| 601 | // Now that we have unwound the stack we find at the top of the stack the | |||
| 602 | // backtrack that the BeginNegativeSubmatch node got. | |||
| 603 | assembler->Backtrack(); | |||
| 604 | } | |||
| 605 | ||||
| 606 | void EndNode::Emit(RegExpCompiler* compiler, Trace* trace) { | |||
| 607 | if (!trace->is_trivial()) { | |||
| 608 | trace->Flush(compiler, this); | |||
| 609 | return; | |||
| 610 | } | |||
| 611 | RegExpMacroAssembler* assembler = compiler->macro_assembler(); | |||
| 612 | if (!label()->is_bound()) { | |||
| 613 | assembler->Bind(label()); | |||
| 614 | } | |||
| 615 | switch (action_) { | |||
| 616 | case ACCEPT: | |||
| 617 | assembler->Succeed(); | |||
| 618 | return; | |||
| 619 | case BACKTRACK: | |||
| 620 | assembler->GoTo(trace->backtrack()); | |||
| 621 | return; | |||
| 622 | case NEGATIVE_SUBMATCH_SUCCESS: | |||
| 623 | // This case is handled in a different virtual method. | |||
| 624 | UNREACHABLE()do { do { } while (false); MOZ_ReportCrash("" "unreachable code" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 624); AnnotateMozCrashReason("MOZ_CRASH(" "unreachable code" ")"); do { *((volatile int*)__null) = 624; __attribute__((nomerge )) ::abort(); } while (false); } while (false); | |||
| 625 | } | |||
| 626 | UNIMPLEMENTED()do { do { } while (false); MOZ_ReportCrash("" "unimplemented code" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 626); AnnotateMozCrashReason("MOZ_CRASH(" "unimplemented code" ")"); do { *((volatile int*)__null) = 626; __attribute__((nomerge )) ::abort(); } while (false); } while (false); | |||
| 627 | } | |||
| 628 | ||||
| 629 | void GuardedAlternative::AddGuard(Guard* guard, Zone* zone) { | |||
| 630 | if (guards_ == nullptr) guards_ = zone->New<ZoneList<Guard*>>(1, zone); | |||
| 631 | guards_->Add(guard, zone); | |||
| 632 | } | |||
| 633 | ||||
| 634 | ActionNode* ActionNode::SetRegisterForLoop(int reg, int val, | |||
| 635 | RegExpNode* on_success) { | |||
| 636 | ActionNode* result = | |||
| 637 | on_success->zone()->New<ActionNode>(SET_REGISTER_FOR_LOOP, on_success); | |||
| 638 | result->data_.u_store_register.reg = reg; | |||
| 639 | result->data_.u_store_register.value = val; | |||
| 640 | return result; | |||
| 641 | } | |||
| 642 | ||||
| 643 | ActionNode* ActionNode::IncrementRegister(int reg, RegExpNode* on_success) { | |||
| 644 | ActionNode* result = | |||
| 645 | on_success->zone()->New<ActionNode>(INCREMENT_REGISTER, on_success); | |||
| 646 | result->data_.u_increment_register.reg = reg; | |||
| 647 | return result; | |||
| 648 | } | |||
| 649 | ||||
| 650 | ActionNode* ActionNode::StorePosition(int reg, bool is_capture, | |||
| 651 | RegExpNode* on_success) { | |||
| 652 | ActionNode* result = | |||
| 653 | on_success->zone()->New<ActionNode>(STORE_POSITION, on_success); | |||
| 654 | result->data_.u_position_register.reg = reg; | |||
| 655 | result->data_.u_position_register.is_capture = is_capture; | |||
| 656 | return result; | |||
| 657 | } | |||
| 658 | ||||
| 659 | ActionNode* ActionNode::ClearCaptures(Interval range, RegExpNode* on_success) { | |||
| 660 | ActionNode* result = | |||
| 661 | on_success->zone()->New<ActionNode>(CLEAR_CAPTURES, on_success); | |||
| 662 | result->data_.u_clear_captures.range_from = range.from(); | |||
| 663 | result->data_.u_clear_captures.range_to = range.to(); | |||
| 664 | return result; | |||
| 665 | } | |||
| 666 | ||||
| 667 | ActionNode* ActionNode::BeginPositiveSubmatch(int stack_reg, int position_reg, | |||
| 668 | RegExpNode* on_success) { | |||
| 669 | ActionNode* result = | |||
| 670 | on_success->zone()->New<ActionNode>(BEGIN_POSITIVE_SUBMATCH, on_success); | |||
| 671 | result->data_.u_submatch.stack_pointer_register = stack_reg; | |||
| 672 | result->data_.u_submatch.current_position_register = position_reg; | |||
| 673 | return result; | |||
| 674 | } | |||
| 675 | ||||
| 676 | ActionNode* ActionNode::BeginNegativeSubmatch(int stack_reg, int position_reg, | |||
| 677 | RegExpNode* on_success) { | |||
| 678 | ActionNode* result = | |||
| 679 | on_success->zone()->New<ActionNode>(BEGIN_NEGATIVE_SUBMATCH, on_success); | |||
| 680 | result->data_.u_submatch.stack_pointer_register = stack_reg; | |||
| 681 | result->data_.u_submatch.current_position_register = position_reg; | |||
| 682 | return result; | |||
| 683 | } | |||
| 684 | ||||
| 685 | ActionNode* ActionNode::PositiveSubmatchSuccess(int stack_reg, int position_reg, | |||
| 686 | int clear_register_count, | |||
| 687 | int clear_register_from, | |||
| 688 | RegExpNode* on_success) { | |||
| 689 | ActionNode* result = on_success->zone()->New<ActionNode>( | |||
| 690 | POSITIVE_SUBMATCH_SUCCESS, on_success); | |||
| 691 | result->data_.u_submatch.stack_pointer_register = stack_reg; | |||
| 692 | result->data_.u_submatch.current_position_register = position_reg; | |||
| 693 | result->data_.u_submatch.clear_register_count = clear_register_count; | |||
| 694 | result->data_.u_submatch.clear_register_from = clear_register_from; | |||
| 695 | return result; | |||
| 696 | } | |||
| 697 | ||||
| 698 | ActionNode* ActionNode::EmptyMatchCheck(int start_register, | |||
| 699 | int repetition_register, | |||
| 700 | int repetition_limit, | |||
| 701 | RegExpNode* on_success) { | |||
| 702 | ActionNode* result = | |||
| 703 | on_success->zone()->New<ActionNode>(EMPTY_MATCH_CHECK, on_success); | |||
| 704 | result->data_.u_empty_match_check.start_register = start_register; | |||
| 705 | result->data_.u_empty_match_check.repetition_register = repetition_register; | |||
| 706 | result->data_.u_empty_match_check.repetition_limit = repetition_limit; | |||
| 707 | return result; | |||
| 708 | } | |||
| 709 | ||||
| 710 | ActionNode* ActionNode::ModifyFlags(RegExpFlags flags, RegExpNode* on_success) { | |||
| 711 | ActionNode* result = | |||
| 712 | on_success->zone()->New<ActionNode>(MODIFY_FLAGS, on_success); | |||
| 713 | result->data_.u_modify_flags.flags = flags; | |||
| 714 | return result; | |||
| 715 | } | |||
| 716 | ||||
| 717 | #define DEFINE_ACCEPT(Type) \ | |||
| 718 | void Type##Node::Accept(NodeVisitor* visitor) { visitor->Visit##Type(this); } | |||
| 719 | FOR_EACH_NODE_TYPE(DEFINE_ACCEPT)DEFINE_ACCEPT(End) DEFINE_ACCEPT(Action) DEFINE_ACCEPT(Choice ) DEFINE_ACCEPT(LoopChoice) DEFINE_ACCEPT(NegativeLookaroundChoice ) DEFINE_ACCEPT(BackReference) DEFINE_ACCEPT(Assertion) DEFINE_ACCEPT (Text) | |||
| 720 | #undef DEFINE_ACCEPT | |||
| 721 | ||||
| 722 | // ------------------------------------------------------------------- | |||
| 723 | // Emit code. | |||
| 724 | ||||
| 725 | void ChoiceNode::GenerateGuard(RegExpMacroAssembler* macro_assembler, | |||
| 726 | Guard* guard, Trace* trace) { | |||
| 727 | switch (guard->op()) { | |||
| 728 | case Guard::LT: | |||
| 729 | DCHECK(!trace->mentions_reg(guard->reg()))do { static_assert( mozilla::detail::AssertionConditionType< decltype(!trace->mentions_reg(guard->reg()))>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!(!trace->mentions_reg(guard->reg())))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("!trace->mentions_reg(guard->reg())" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 729); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!trace->mentions_reg(guard->reg())" ")"); do { *((volatile int*)__null) = 729; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 730 | macro_assembler->IfRegisterGE(guard->reg(), guard->value(), | |||
| 731 | trace->backtrack()); | |||
| 732 | break; | |||
| 733 | case Guard::GEQ: | |||
| 734 | DCHECK(!trace->mentions_reg(guard->reg()))do { static_assert( mozilla::detail::AssertionConditionType< decltype(!trace->mentions_reg(guard->reg()))>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!(!trace->mentions_reg(guard->reg())))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("!trace->mentions_reg(guard->reg())" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 734); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!trace->mentions_reg(guard->reg())" ")"); do { *((volatile int*)__null) = 734; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 735 | macro_assembler->IfRegisterLT(guard->reg(), guard->value(), | |||
| 736 | trace->backtrack()); | |||
| 737 | break; | |||
| 738 | } | |||
| 739 | } | |||
| 740 | ||||
| 741 | namespace { | |||
| 742 | ||||
| 743 | #ifdef DEBUG1 | |||
| 744 | bool ContainsOnlyUtf16CodeUnits(unibrow::uchar* chars, int length) { | |||
| 745 | static_assert(sizeof(unibrow::uchar) == 4); | |||
| 746 | for (int i = 0; i < length; i++) { | |||
| 747 | if (chars[i] > String::kMaxUtf16CodeUnit) return false; | |||
| 748 | } | |||
| 749 | return true; | |||
| 750 | } | |||
| 751 | #endif // DEBUG | |||
| 752 | ||||
| 753 | // Returns the number of characters in the equivalence class, omitting those | |||
| 754 | // that cannot occur in the source string because it is Latin1. This is called | |||
| 755 | // both for unicode modes /ui and /vi, and also for legacy case independent | |||
| 756 | // mode /i. In the case of Unicode modes we handled surrogate pair expansions | |||
| 757 | // earlier so at this point it's all about single-code-unit expansions. | |||
| 758 | int GetCaseIndependentLetters(Isolate* isolate, base::uc16 character, | |||
| 759 | RegExpCompiler* compiler, unibrow::uchar* letters, | |||
| 760 | int letter_length) { | |||
| 761 | bool one_byte_subject = compiler->one_byte(); | |||
| 762 | bool unicode = IsEitherUnicode(compiler->flags()); | |||
| 763 | static const base::uc16 kMaxAscii = 0x7f; | |||
| 764 | if (!unicode && character <= kMaxAscii) { | |||
| 765 | // Fast case for common characters. | |||
| 766 | base::uc16 upper = character & ~0x20; | |||
| 767 | if ('A' <= upper && upper <= 'Z') { | |||
| 768 | letters[0] = upper; | |||
| 769 | letters[1] = upper | 0x20; | |||
| 770 | return 2; | |||
| 771 | } | |||
| 772 | letters[0] = character; | |||
| 773 | return 1; | |||
| 774 | } | |||
| 775 | #ifdef V8_INTL_SUPPORT1 | |||
| 776 | ||||
| 777 | if (!unicode && RegExpCaseFolding::IgnoreSet().contains(character)) { | |||
| 778 | letters[0] = character; | |||
| 779 | DCHECK(ContainsOnlyUtf16CodeUnits(letters, 1))do { static_assert( mozilla::detail::AssertionConditionType< decltype(ContainsOnlyUtf16CodeUnits(letters, 1))>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!(ContainsOnlyUtf16CodeUnits(letters, 1)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("ContainsOnlyUtf16CodeUnits(letters, 1)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 779); AnnotateMozCrashReason("MOZ_ASSERT" "(" "ContainsOnlyUtf16CodeUnits(letters, 1)" ")"); do { *((volatile int*)__null) = 779; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 780 | return 1; | |||
| 781 | } | |||
| 782 | bool in_special_add_set = | |||
| 783 | RegExpCaseFolding::SpecialAddSet().contains(character); | |||
| 784 | ||||
| 785 | icu::UnicodeSet set; | |||
| 786 | set.add(character); | |||
| 787 | set = set.closeOver(unicode ? USET_SIMPLE_CASE_INSENSITIVE | |||
| 788 | : USET_CASE_INSENSITIVE); | |||
| 789 | ||||
| 790 | UChar32 canon = 0; | |||
| 791 | if (in_special_add_set && !unicode) { | |||
| 792 | canon = RegExpCaseFolding::Canonicalize(character); | |||
| 793 | } | |||
| 794 | ||||
| 795 | int32_t range_count = set.getRangeCount(); | |||
| 796 | int items = 0; | |||
| 797 | for (int32_t i = 0; i < range_count; i++) { | |||
| 798 | UChar32 start = set.getRangeStart(i); | |||
| 799 | UChar32 end = set.getRangeEnd(i); | |||
| 800 | CHECK(end - start + items <= letter_length)do { static_assert( mozilla::detail::AssertionConditionType< decltype(end - start + items <= letter_length)>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!(end - start + items <= letter_length))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("end - start + items <= letter_length" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 800); AnnotateMozCrashReason("MOZ_RELEASE_ASSERT" "(" "end - start + items <= letter_length" ")"); do { *((volatile int*)__null) = 800; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 801 | for (UChar32 cu = start; cu <= end; cu++) { | |||
| 802 | if (one_byte_subject && cu > String::kMaxOneByteCharCode) continue; | |||
| 803 | if (!unicode && in_special_add_set && | |||
| 804 | RegExpCaseFolding::Canonicalize(cu) != canon) { | |||
| 805 | continue; | |||
| 806 | } | |||
| 807 | letters[items++] = static_cast<unibrow::uchar>(cu); | |||
| 808 | } | |||
| 809 | } | |||
| 810 | DCHECK(ContainsOnlyUtf16CodeUnits(letters, items))do { static_assert( mozilla::detail::AssertionConditionType< decltype(ContainsOnlyUtf16CodeUnits(letters, items))>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!(ContainsOnlyUtf16CodeUnits(letters, items)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("ContainsOnlyUtf16CodeUnits(letters, items)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 810); AnnotateMozCrashReason("MOZ_ASSERT" "(" "ContainsOnlyUtf16CodeUnits(letters, items)" ")"); do { *((volatile int*)__null) = 810; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 811 | return items; | |||
| 812 | #else | |||
| 813 | int length = | |||
| 814 | isolate->jsregexp_uncanonicalize()->get(character, '\0', letters); | |||
| 815 | // Unibrow returns 0 or 1 for characters where case independence is | |||
| 816 | // trivial. | |||
| 817 | if (length == 0) { | |||
| 818 | letters[0] = character; | |||
| 819 | length = 1; | |||
| 820 | } | |||
| 821 | ||||
| 822 | if (one_byte_subject) { | |||
| 823 | int new_length = 0; | |||
| 824 | for (int i = 0; i < length; i++) { | |||
| 825 | if (letters[i] <= String::kMaxOneByteCharCode) { | |||
| 826 | letters[new_length++] = letters[i]; | |||
| 827 | } | |||
| 828 | } | |||
| 829 | length = new_length; | |||
| 830 | } | |||
| 831 | ||||
| 832 | DCHECK(ContainsOnlyUtf16CodeUnits(letters, length))do { static_assert( mozilla::detail::AssertionConditionType< decltype(ContainsOnlyUtf16CodeUnits(letters, length))>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!(ContainsOnlyUtf16CodeUnits(letters, length)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("ContainsOnlyUtf16CodeUnits(letters, length)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 832); AnnotateMozCrashReason("MOZ_ASSERT" "(" "ContainsOnlyUtf16CodeUnits(letters, length)" ")"); do { *((volatile int*)__null) = 832; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 833 | return length; | |||
| 834 | #endif // V8_INTL_SUPPORT | |||
| 835 | } | |||
| 836 | ||||
| 837 | inline bool EmitSimpleCharacter(Isolate* isolate, RegExpCompiler* compiler, | |||
| 838 | base::uc16 c, Label* on_failure, int cp_offset, | |||
| 839 | bool check, bool preloaded) { | |||
| 840 | RegExpMacroAssembler* assembler = compiler->macro_assembler(); | |||
| 841 | bool bound_checked = false; | |||
| 842 | if (!preloaded) { | |||
| 843 | assembler->LoadCurrentCharacter(cp_offset, on_failure, check); | |||
| 844 | bound_checked = true; | |||
| 845 | } | |||
| 846 | assembler->CheckNotCharacter(c, on_failure); | |||
| 847 | return bound_checked; | |||
| 848 | } | |||
| 849 | ||||
| 850 | // Only emits non-letters (things that don't have case). Only used for case | |||
| 851 | // independent matches. | |||
| 852 | inline bool EmitAtomNonLetter(Isolate* isolate, RegExpCompiler* compiler, | |||
| 853 | base::uc16 c, Label* on_failure, int cp_offset, | |||
| 854 | bool check, bool preloaded) { | |||
| 855 | RegExpMacroAssembler* macro_assembler = compiler->macro_assembler(); | |||
| 856 | bool one_byte = compiler->one_byte(); | |||
| 857 | unibrow::uchar chars[4]; | |||
| 858 | int length = GetCaseIndependentLetters(isolate, c, compiler, chars, 4); | |||
| 859 | if (length < 1) { | |||
| 860 | // This can't match. Must be an one-byte subject and a non-one-byte | |||
| 861 | // character. We do not need to do anything since the one-byte pass | |||
| 862 | // already handled this. | |||
| 863 | return false; // Bounds not checked. | |||
| 864 | } | |||
| 865 | bool checked = false; | |||
| 866 | // We handle the length > 1 case in a later pass. | |||
| 867 | if (length == 1) { | |||
| 868 | if (one_byte && chars[0] > String::kMaxOneByteCharCodeU) { | |||
| 869 | // Can't match - see above. | |||
| 870 | return false; // Bounds not checked. | |||
| 871 | } | |||
| 872 | if (!preloaded) { | |||
| 873 | macro_assembler->LoadCurrentCharacter(cp_offset, on_failure, check); | |||
| 874 | checked = check; | |||
| 875 | } | |||
| 876 | macro_assembler->CheckNotCharacter(chars[0], on_failure); | |||
| 877 | } | |||
| 878 | return checked; | |||
| 879 | } | |||
| 880 | ||||
| 881 | bool ShortCutEmitCharacterPair(RegExpMacroAssembler* macro_assembler, | |||
| 882 | bool one_byte, base::uc16 c1, base::uc16 c2, | |||
| 883 | Label* on_failure) { | |||
| 884 | const uint32_t char_mask = CharMask(one_byte); | |||
| 885 | base::uc16 exor = c1 ^ c2; | |||
| 886 | // Check whether exor has only one bit set. | |||
| 887 | if (((exor - 1) & exor) == 0) { | |||
| 888 | // If c1 and c2 differ only by one bit. | |||
| 889 | // Ecma262UnCanonicalize always gives the highest number last. | |||
| 890 | DCHECK(c2 > c1)do { static_assert( mozilla::detail::AssertionConditionType< decltype(c2 > c1)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(c2 > c1))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("c2 > c1", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 890); AnnotateMozCrashReason("MOZ_ASSERT" "(" "c2 > c1" ")" ); do { *((volatile int*)__null) = 890; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 891 | base::uc16 mask = char_mask ^ exor; | |||
| 892 | macro_assembler->CheckNotCharacterAfterAnd(c1, mask, on_failure); | |||
| 893 | return true; | |||
| 894 | } | |||
| 895 | DCHECK(c2 > c1)do { static_assert( mozilla::detail::AssertionConditionType< decltype(c2 > c1)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(c2 > c1))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("c2 > c1", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 895); AnnotateMozCrashReason("MOZ_ASSERT" "(" "c2 > c1" ")" ); do { *((volatile int*)__null) = 895; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 896 | base::uc16 diff = c2 - c1; | |||
| 897 | if (((diff - 1) & diff) == 0 && c1 >= diff) { | |||
| 898 | // If the characters differ by 2^n but don't differ by one bit then | |||
| 899 | // subtract the difference from the found character, then do the or | |||
| 900 | // trick. We avoid the theoretical case where negative numbers are | |||
| 901 | // involved in order to simplify code generation. | |||
| 902 | base::uc16 mask = char_mask ^ diff; | |||
| 903 | macro_assembler->CheckNotCharacterAfterMinusAnd(c1 - diff, diff, mask, | |||
| 904 | on_failure); | |||
| 905 | return true; | |||
| 906 | } | |||
| 907 | return false; | |||
| 908 | } | |||
| 909 | ||||
| 910 | // Only emits letters (things that have case). Only used for case independent | |||
| 911 | // matches. | |||
| 912 | inline bool EmitAtomLetter(Isolate* isolate, RegExpCompiler* compiler, | |||
| 913 | base::uc16 c, Label* on_failure, int cp_offset, | |||
| 914 | bool check, bool preloaded) { | |||
| 915 | RegExpMacroAssembler* macro_assembler = compiler->macro_assembler(); | |||
| 916 | bool one_byte = compiler->one_byte(); | |||
| 917 | unibrow::uchar chars[4]; | |||
| 918 | int length = GetCaseIndependentLetters(isolate, c, compiler, chars, 4); | |||
| 919 | if (length <= 1) return false; | |||
| 920 | // We may not need to check against the end of the input string | |||
| 921 | // if this character lies before a character that matched. | |||
| 922 | if (!preloaded) { | |||
| 923 | macro_assembler->LoadCurrentCharacter(cp_offset, on_failure, check); | |||
| 924 | } | |||
| 925 | Label ok; | |||
| 926 | switch (length) { | |||
| 927 | case 2: { | |||
| 928 | if (ShortCutEmitCharacterPair(macro_assembler, one_byte, chars[0], | |||
| 929 | chars[1], on_failure)) { | |||
| 930 | } else { | |||
| 931 | macro_assembler->CheckCharacter(chars[0], &ok); | |||
| 932 | macro_assembler->CheckNotCharacter(chars[1], on_failure); | |||
| 933 | macro_assembler->Bind(&ok); | |||
| 934 | } | |||
| 935 | break; | |||
| 936 | } | |||
| 937 | case 4: | |||
| 938 | macro_assembler->CheckCharacter(chars[3], &ok); | |||
| 939 | [[fallthrough]]; | |||
| 940 | case 3: | |||
| 941 | macro_assembler->CheckCharacter(chars[0], &ok); | |||
| 942 | macro_assembler->CheckCharacter(chars[1], &ok); | |||
| 943 | macro_assembler->CheckNotCharacter(chars[2], on_failure); | |||
| 944 | macro_assembler->Bind(&ok); | |||
| 945 | break; | |||
| 946 | default: | |||
| 947 | UNREACHABLE()do { do { } while (false); MOZ_ReportCrash("" "unreachable code" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 947); AnnotateMozCrashReason("MOZ_CRASH(" "unreachable code" ")"); do { *((volatile int*)__null) = 947; __attribute__((nomerge )) ::abort(); } while (false); } while (false); | |||
| 948 | } | |||
| 949 | return true; | |||
| 950 | } | |||
| 951 | ||||
| 952 | void EmitBoundaryTest(RegExpMacroAssembler* masm, int border, | |||
| 953 | Label* fall_through, Label* above_or_equal, | |||
| 954 | Label* below) { | |||
| 955 | if (below != fall_through) { | |||
| 956 | masm->CheckCharacterLT(border, below); | |||
| 957 | if (above_or_equal != fall_through) masm->GoTo(above_or_equal); | |||
| 958 | } else { | |||
| 959 | masm->CheckCharacterGT(border - 1, above_or_equal); | |||
| 960 | } | |||
| 961 | } | |||
| 962 | ||||
| 963 | void EmitDoubleBoundaryTest(RegExpMacroAssembler* masm, int first, int last, | |||
| 964 | Label* fall_through, Label* in_range, | |||
| 965 | Label* out_of_range) { | |||
| 966 | if (in_range == fall_through) { | |||
| 967 | if (first == last) { | |||
| 968 | masm->CheckNotCharacter(first, out_of_range); | |||
| 969 | } else { | |||
| 970 | masm->CheckCharacterNotInRange(first, last, out_of_range); | |||
| 971 | } | |||
| 972 | } else { | |||
| 973 | if (first == last) { | |||
| 974 | masm->CheckCharacter(first, in_range); | |||
| 975 | } else { | |||
| 976 | masm->CheckCharacterInRange(first, last, in_range); | |||
| 977 | } | |||
| 978 | if (out_of_range != fall_through) masm->GoTo(out_of_range); | |||
| 979 | } | |||
| 980 | } | |||
| 981 | ||||
| 982 | // even_label is for ranges[i] to ranges[i + 1] where i - start_index is even. | |||
| 983 | // odd_label is for ranges[i] to ranges[i + 1] where i - start_index is odd. | |||
| 984 | void EmitUseLookupTable(RegExpMacroAssembler* masm, | |||
| 985 | ZoneList<base::uc32>* ranges, uint32_t start_index, | |||
| 986 | uint32_t end_index, base::uc32 min_char, | |||
| 987 | Label* fall_through, Label* even_label, | |||
| 988 | Label* odd_label) { | |||
| 989 | static const uint32_t kSize = RegExpMacroAssembler::kTableSize; | |||
| 990 | static const uint32_t kMask = RegExpMacroAssembler::kTableMask; | |||
| 991 | ||||
| 992 | base::uc32 base = (min_char & ~kMask); | |||
| 993 | USE(base)do { ::v8::base::Use unused_tmp_array_for_use_macro[]{base}; ( void)unused_tmp_array_for_use_macro; } while (false); | |||
| 994 | ||||
| 995 | // Assert that everything is on one kTableSize page. | |||
| 996 | for (uint32_t i = start_index; i <= end_index; i++) { | |||
| 997 | DCHECK_EQ(ranges->at(i) & ~kMask, base)do { static_assert( mozilla::detail::AssertionConditionType< decltype((ranges->at(i) & ~kMask) == (base))>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!((ranges->at(i) & ~kMask) == (base)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(ranges->at(i) & ~kMask) == (base)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 997); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(ranges->at(i) & ~kMask) == (base)" ")"); do { *((volatile int*)__null) = 997; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 998 | } | |||
| 999 | DCHECK(start_index == 0 || (ranges->at(start_index - 1) & ~kMask) <= base)do { static_assert( mozilla::detail::AssertionConditionType< decltype(start_index == 0 || (ranges->at(start_index - 1) & ~kMask) <= base)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(start_index == 0 || (ranges-> at(start_index - 1) & ~kMask) <= base))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("start_index == 0 || (ranges->at(start_index - 1) & ~kMask) <= base" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 999); AnnotateMozCrashReason("MOZ_ASSERT" "(" "start_index == 0 || (ranges->at(start_index - 1) & ~kMask) <= base" ")"); do { *((volatile int*)__null) = 999; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1000 | ||||
| 1001 | char templ[kSize]; | |||
| 1002 | Label* on_bit_set; | |||
| 1003 | Label* on_bit_clear; | |||
| 1004 | int bit; | |||
| 1005 | if (even_label == fall_through) { | |||
| 1006 | on_bit_set = odd_label; | |||
| 1007 | on_bit_clear = even_label; | |||
| 1008 | bit = 1; | |||
| 1009 | } else { | |||
| 1010 | on_bit_set = even_label; | |||
| 1011 | on_bit_clear = odd_label; | |||
| 1012 | bit = 0; | |||
| 1013 | } | |||
| 1014 | for (uint32_t i = 0; i < (ranges->at(start_index) & kMask) && i < kSize; | |||
| 1015 | i++) { | |||
| 1016 | templ[i] = bit; | |||
| 1017 | } | |||
| 1018 | uint32_t j = 0; | |||
| 1019 | bit ^= 1; | |||
| 1020 | for (uint32_t i = start_index; i < end_index; i++) { | |||
| 1021 | for (j = (ranges->at(i) & kMask); j < (ranges->at(i + 1) & kMask); j++) { | |||
| 1022 | templ[j] = bit; | |||
| 1023 | } | |||
| 1024 | bit ^= 1; | |||
| 1025 | } | |||
| 1026 | for (uint32_t i = j; i < kSize; i++) { | |||
| 1027 | templ[i] = bit; | |||
| 1028 | } | |||
| 1029 | Factory* factory = masm->isolate()->factory(); | |||
| 1030 | // TODO(erikcorry): Cache these. | |||
| 1031 | Handle<ByteArray> ba = factory->NewByteArray(kSize, AllocationType::kOld); | |||
| 1032 | for (uint32_t i = 0; i < kSize; i++) { | |||
| 1033 | ba->set(i, templ[i]); | |||
| 1034 | } | |||
| 1035 | masm->CheckBitInTable(ba, on_bit_set); | |||
| 1036 | if (on_bit_clear != fall_through) masm->GoTo(on_bit_clear); | |||
| 1037 | } | |||
| 1038 | ||||
| 1039 | void CutOutRange(RegExpMacroAssembler* masm, ZoneList<base::uc32>* ranges, | |||
| 1040 | uint32_t start_index, uint32_t end_index, uint32_t cut_index, | |||
| 1041 | Label* even_label, Label* odd_label) { | |||
| 1042 | bool odd = (((cut_index - start_index) & 1) == 1); | |||
| 1043 | Label* in_range_label = odd ? odd_label : even_label; | |||
| 1044 | Label dummy; | |||
| 1045 | EmitDoubleBoundaryTest(masm, ranges->at(cut_index), | |||
| 1046 | ranges->at(cut_index + 1) - 1, &dummy, in_range_label, | |||
| 1047 | &dummy); | |||
| 1048 | DCHECK(!dummy.is_linked())do { static_assert( mozilla::detail::AssertionConditionType< decltype(!dummy.is_linked())>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(!dummy.is_linked()))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("!dummy.is_linked()" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1048); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!dummy.is_linked()" ")"); do { *((volatile int*)__null) = 1048; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1049 | // Cut out the single range by rewriting the array. This creates a new | |||
| 1050 | // range that is a merger of the two ranges on either side of the one we | |||
| 1051 | // are cutting out. The oddity of the labels is preserved. | |||
| 1052 | for (uint32_t j = cut_index; j > start_index; j--) { | |||
| 1053 | ranges->at(j) = ranges->at(j - 1); | |||
| 1054 | } | |||
| 1055 | for (uint32_t j = cut_index + 1; j < end_index; j++) { | |||
| 1056 | ranges->at(j) = ranges->at(j + 1); | |||
| 1057 | } | |||
| 1058 | } | |||
| 1059 | ||||
| 1060 | // Unicode case. Split the search space into kSize spaces that are handled | |||
| 1061 | // with recursion. | |||
| 1062 | void SplitSearchSpace(ZoneList<base::uc32>* ranges, uint32_t start_index, | |||
| 1063 | uint32_t end_index, uint32_t* new_start_index, | |||
| 1064 | uint32_t* new_end_index, base::uc32* border) { | |||
| 1065 | static const uint32_t kSize = RegExpMacroAssembler::kTableSize; | |||
| 1066 | static const uint32_t kMask = RegExpMacroAssembler::kTableMask; | |||
| 1067 | ||||
| 1068 | base::uc32 first = ranges->at(start_index); | |||
| 1069 | base::uc32 last = ranges->at(end_index) - 1; | |||
| 1070 | ||||
| 1071 | *new_start_index = start_index; | |||
| 1072 | *border = (ranges->at(start_index) & ~kMask) + kSize; | |||
| 1073 | while (*new_start_index < end_index) { | |||
| 1074 | if (ranges->at(*new_start_index) > *border) break; | |||
| 1075 | (*new_start_index)++; | |||
| 1076 | } | |||
| 1077 | // new_start_index is the index of the first edge that is beyond the | |||
| 1078 | // current kSize space. | |||
| 1079 | ||||
| 1080 | // For very large search spaces we do a binary chop search of the non-Latin1 | |||
| 1081 | // space instead of just going to the end of the current kSize space. The | |||
| 1082 | // heuristics are complicated a little by the fact that any 128-character | |||
| 1083 | // encoding space can be quickly tested with a table lookup, so we don't | |||
| 1084 | // wish to do binary chop search at a smaller granularity than that. A | |||
| 1085 | // 128-character space can take up a lot of space in the ranges array if, | |||
| 1086 | // for example, we only want to match every second character (eg. the lower | |||
| 1087 | // case characters on some Unicode pages). | |||
| 1088 | uint32_t binary_chop_index = (end_index + start_index) / 2; | |||
| 1089 | // The first test ensures that we get to the code that handles the Latin1 | |||
| 1090 | // range with a single not-taken branch, speeding up this important | |||
| 1091 | // character range (even non-Latin1 charset-based text has spaces and | |||
| 1092 | // punctuation). | |||
| 1093 | if (*border - 1 > String::kMaxOneByteCharCode && // Latin1 case. | |||
| 1094 | end_index - start_index > (*new_start_index - start_index) * 2 && | |||
| 1095 | last - first > kSize * 2 && binary_chop_index > *new_start_index && | |||
| 1096 | ranges->at(binary_chop_index) >= first + 2 * kSize) { | |||
| 1097 | uint32_t scan_forward_for_section_border = binary_chop_index; | |||
| 1098 | uint32_t new_border = (ranges->at(binary_chop_index) | kMask) + 1; | |||
| 1099 | ||||
| 1100 | while (scan_forward_for_section_border < end_index) { | |||
| 1101 | if (ranges->at(scan_forward_for_section_border) > new_border) { | |||
| 1102 | *new_start_index = scan_forward_for_section_border; | |||
| 1103 | *border = new_border; | |||
| 1104 | break; | |||
| 1105 | } | |||
| 1106 | scan_forward_for_section_border++; | |||
| 1107 | } | |||
| 1108 | } | |||
| 1109 | ||||
| 1110 | DCHECK(*new_start_index > start_index)do { static_assert( mozilla::detail::AssertionConditionType< decltype(*new_start_index > start_index)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(*new_start_index > start_index ))), 0))) { do { } while (false); MOZ_ReportAssertionFailure( "*new_start_index > start_index", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1110); AnnotateMozCrashReason("MOZ_ASSERT" "(" "*new_start_index > start_index" ")"); do { *((volatile int*)__null) = 1110; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1111 | *new_end_index = *new_start_index - 1; | |||
| 1112 | if (ranges->at(*new_end_index) == *border) { | |||
| 1113 | (*new_end_index)--; | |||
| 1114 | } | |||
| 1115 | if (*border >= ranges->at(end_index)) { | |||
| 1116 | *border = ranges->at(end_index); | |||
| 1117 | *new_start_index = end_index; // Won't be used. | |||
| 1118 | *new_end_index = end_index - 1; | |||
| 1119 | } | |||
| 1120 | } | |||
| 1121 | ||||
| 1122 | // Gets a series of segment boundaries representing a character class. If the | |||
| 1123 | // character is in the range between an even and an odd boundary (counting from | |||
| 1124 | // start_index) then go to even_label, otherwise go to odd_label. We already | |||
| 1125 | // know that the character is in the range of min_char to max_char inclusive. | |||
| 1126 | // Either label can be nullptr indicating backtracking. Either label can also | |||
| 1127 | // be equal to the fall_through label. | |||
| 1128 | void GenerateBranches(RegExpMacroAssembler* masm, ZoneList<base::uc32>* ranges, | |||
| 1129 | uint32_t start_index, uint32_t end_index, | |||
| 1130 | base::uc32 min_char, base::uc32 max_char, | |||
| 1131 | Label* fall_through, Label* even_label, | |||
| 1132 | Label* odd_label) { | |||
| 1133 | DCHECK_LE(min_char, String::kMaxUtf16CodeUnit)do { static_assert( mozilla::detail::AssertionConditionType< decltype((min_char) <= (String::kMaxUtf16CodeUnit))>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!((min_char) <= (String::kMaxUtf16CodeUnit)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(min_char) <= (String::kMaxUtf16CodeUnit)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1133); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(min_char) <= (String::kMaxUtf16CodeUnit)" ")"); do { *((volatile int*)__null) = 1133; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1134 | DCHECK_LE(max_char, String::kMaxUtf16CodeUnit)do { static_assert( mozilla::detail::AssertionConditionType< decltype((max_char) <= (String::kMaxUtf16CodeUnit))>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!((max_char) <= (String::kMaxUtf16CodeUnit)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(max_char) <= (String::kMaxUtf16CodeUnit)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1134); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(max_char) <= (String::kMaxUtf16CodeUnit)" ")"); do { *((volatile int*)__null) = 1134; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1135 | ||||
| 1136 | base::uc32 first = ranges->at(start_index); | |||
| 1137 | base::uc32 last = ranges->at(end_index) - 1; | |||
| 1138 | ||||
| 1139 | DCHECK_LT(min_char, first)do { static_assert( mozilla::detail::AssertionConditionType< decltype((min_char) < (first))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((min_char) < (first)))), 0 ))) { do { } while (false); MOZ_ReportAssertionFailure("(min_char) < (first)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1139); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(min_char) < (first)" ")"); do { *((volatile int*)__null) = 1139; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1140 | ||||
| 1141 | // Just need to test if the character is before or on-or-after | |||
| 1142 | // a particular character. | |||
| 1143 | if (start_index == end_index) { | |||
| 1144 | EmitBoundaryTest(masm, first, fall_through, even_label, odd_label); | |||
| 1145 | return; | |||
| 1146 | } | |||
| 1147 | ||||
| 1148 | // Another almost trivial case: There is one interval in the middle that is | |||
| 1149 | // different from the end intervals. | |||
| 1150 | if (start_index + 1 == end_index) { | |||
| 1151 | EmitDoubleBoundaryTest(masm, first, last, fall_through, even_label, | |||
| 1152 | odd_label); | |||
| 1153 | return; | |||
| 1154 | } | |||
| 1155 | ||||
| 1156 | // It's not worth using table lookup if there are very few intervals in the | |||
| 1157 | // character class. | |||
| 1158 | if (end_index - start_index <= 6) { | |||
| 1159 | // It is faster to test for individual characters, so we look for those | |||
| 1160 | // first, then try arbitrary ranges in the second round. | |||
| 1161 | static uint32_t kNoCutIndex = -1; | |||
| 1162 | uint32_t cut = kNoCutIndex; | |||
| 1163 | for (uint32_t i = start_index; i < end_index; i++) { | |||
| 1164 | if (ranges->at(i) == ranges->at(i + 1) - 1) { | |||
| 1165 | cut = i; | |||
| 1166 | break; | |||
| 1167 | } | |||
| 1168 | } | |||
| 1169 | if (cut == kNoCutIndex) cut = start_index; | |||
| 1170 | CutOutRange(masm, ranges, start_index, end_index, cut, even_label, | |||
| 1171 | odd_label); | |||
| 1172 | DCHECK_GE(end_index - start_index, 2)do { static_assert( mozilla::detail::AssertionConditionType< decltype((end_index - start_index) >= (2))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((end_index - start_index) >= (2)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(end_index - start_index) >= (2)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1172); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(end_index - start_index) >= (2)" ")"); do { *((volatile int*)__null) = 1172; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1173 | GenerateBranches(masm, ranges, start_index + 1, end_index - 1, min_char, | |||
| 1174 | max_char, fall_through, even_label, odd_label); | |||
| 1175 | return; | |||
| 1176 | } | |||
| 1177 | ||||
| 1178 | // If there are a lot of intervals in the regexp, then we will use tables to | |||
| 1179 | // determine whether the character is inside or outside the character class. | |||
| 1180 | static const int kBits = RegExpMacroAssembler::kTableSizeBits; | |||
| 1181 | ||||
| 1182 | if ((max_char >> kBits) == (min_char >> kBits)) { | |||
| 1183 | EmitUseLookupTable(masm, ranges, start_index, end_index, min_char, | |||
| 1184 | fall_through, even_label, odd_label); | |||
| 1185 | return; | |||
| 1186 | } | |||
| 1187 | ||||
| 1188 | if ((min_char >> kBits) != first >> kBits) { | |||
| 1189 | masm->CheckCharacterLT(first, odd_label); | |||
| 1190 | GenerateBranches(masm, ranges, start_index + 1, end_index, first, max_char, | |||
| 1191 | fall_through, odd_label, even_label); | |||
| 1192 | return; | |||
| 1193 | } | |||
| 1194 | ||||
| 1195 | uint32_t new_start_index = 0; | |||
| 1196 | uint32_t new_end_index = 0; | |||
| 1197 | base::uc32 border = 0; | |||
| 1198 | ||||
| 1199 | SplitSearchSpace(ranges, start_index, end_index, &new_start_index, | |||
| 1200 | &new_end_index, &border); | |||
| 1201 | ||||
| 1202 | Label handle_rest; | |||
| 1203 | Label* above = &handle_rest; | |||
| 1204 | if (border == last + 1) { | |||
| 1205 | // We didn't find any section that started after the limit, so everything | |||
| 1206 | // above the border is one of the terminal labels. | |||
| 1207 | above = (end_index & 1) != (start_index & 1) ? odd_label : even_label; | |||
| 1208 | DCHECK(new_end_index == end_index - 1)do { static_assert( mozilla::detail::AssertionConditionType< decltype(new_end_index == end_index - 1)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(new_end_index == end_index - 1))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("new_end_index == end_index - 1", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1208); AnnotateMozCrashReason("MOZ_ASSERT" "(" "new_end_index == end_index - 1" ")"); do { *((volatile int*)__null) = 1208; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1209 | } | |||
| 1210 | ||||
| 1211 | DCHECK_LE(start_index, new_end_index)do { static_assert( mozilla::detail::AssertionConditionType< decltype((start_index) <= (new_end_index))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((start_index) <= (new_end_index )))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(start_index) <= (new_end_index)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1211); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(start_index) <= (new_end_index)" ")"); do { *((volatile int*)__null) = 1211; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1212 | DCHECK_LE(new_start_index, end_index)do { static_assert( mozilla::detail::AssertionConditionType< decltype((new_start_index) <= (end_index))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((new_start_index) <= (end_index )))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(new_start_index) <= (end_index)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1212); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(new_start_index) <= (end_index)" ")"); do { *((volatile int*)__null) = 1212; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1213 | DCHECK_LT(start_index, new_start_index)do { static_assert( mozilla::detail::AssertionConditionType< decltype((start_index) < (new_start_index))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((start_index) < (new_start_index )))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(start_index) < (new_start_index)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1213); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(start_index) < (new_start_index)" ")"); do { *((volatile int*)__null) = 1213; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1214 | DCHECK_LT(new_end_index, end_index)do { static_assert( mozilla::detail::AssertionConditionType< decltype((new_end_index) < (end_index))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((new_end_index) < (end_index )))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(new_end_index) < (end_index)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1214); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(new_end_index) < (end_index)" ")"); do { *((volatile int*)__null) = 1214; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1215 | DCHECK(new_end_index + 1 == new_start_index ||do { static_assert( mozilla::detail::AssertionConditionType< decltype(new_end_index + 1 == new_start_index || (new_end_index + 2 == new_start_index && border == ranges->at(new_end_index + 1)))>::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!(new_end_index + 1 == new_start_index || (new_end_index + 2 == new_start_index && border == ranges->at(new_end_index + 1))))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("new_end_index + 1 == new_start_index || (new_end_index + 2 == new_start_index && border == ranges->at(new_end_index + 1))" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1217); AnnotateMozCrashReason("MOZ_ASSERT" "(" "new_end_index + 1 == new_start_index || (new_end_index + 2 == new_start_index && border == ranges->at(new_end_index + 1))" ")"); do { *((volatile int*)__null) = 1217; __attribute__((nomerge )) ::abort(); } while (false); } } while (false) | |||
| 1216 | (new_end_index + 2 == new_start_index &&do { static_assert( mozilla::detail::AssertionConditionType< decltype(new_end_index + 1 == new_start_index || (new_end_index + 2 == new_start_index && border == ranges->at(new_end_index + 1)))>::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!(new_end_index + 1 == new_start_index || (new_end_index + 2 == new_start_index && border == ranges->at(new_end_index + 1))))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("new_end_index + 1 == new_start_index || (new_end_index + 2 == new_start_index && border == ranges->at(new_end_index + 1))" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1217); AnnotateMozCrashReason("MOZ_ASSERT" "(" "new_end_index + 1 == new_start_index || (new_end_index + 2 == new_start_index && border == ranges->at(new_end_index + 1))" ")"); do { *((volatile int*)__null) = 1217; __attribute__((nomerge )) ::abort(); } while (false); } } while (false) | |||
| 1217 | border == ranges->at(new_end_index + 1)))do { static_assert( mozilla::detail::AssertionConditionType< decltype(new_end_index + 1 == new_start_index || (new_end_index + 2 == new_start_index && border == ranges->at(new_end_index + 1)))>::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!(new_end_index + 1 == new_start_index || (new_end_index + 2 == new_start_index && border == ranges->at(new_end_index + 1))))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("new_end_index + 1 == new_start_index || (new_end_index + 2 == new_start_index && border == ranges->at(new_end_index + 1))" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1217); AnnotateMozCrashReason("MOZ_ASSERT" "(" "new_end_index + 1 == new_start_index || (new_end_index + 2 == new_start_index && border == ranges->at(new_end_index + 1))" ")"); do { *((volatile int*)__null) = 1217; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1218 | DCHECK_LT(min_char, border - 1)do { static_assert( mozilla::detail::AssertionConditionType< decltype((min_char) < (border - 1))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((min_char) < (border - 1) ))), 0))) { do { } while (false); MOZ_ReportAssertionFailure( "(min_char) < (border - 1)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1218); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(min_char) < (border - 1)" ")"); do { *((volatile int*)__null) = 1218; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1219 | DCHECK_LT(border, max_char)do { static_assert( mozilla::detail::AssertionConditionType< decltype((border) < (max_char))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((border) < (max_char)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(border) < (max_char)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1219); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(border) < (max_char)" ")"); do { *((volatile int*)__null) = 1219; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1220 | DCHECK_LT(ranges->at(new_end_index), border)do { static_assert( mozilla::detail::AssertionConditionType< decltype((ranges->at(new_end_index)) < (border))>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!((ranges->at(new_end_index)) < (border)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(ranges->at(new_end_index)) < (border)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1220); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(ranges->at(new_end_index)) < (border)" ")"); do { *((volatile int*)__null) = 1220; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1221 | DCHECK(border < ranges->at(new_start_index) ||do { static_assert( mozilla::detail::AssertionConditionType< decltype(border < ranges->at(new_start_index) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(border < ranges->at(new_start_index ) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("border < ranges->at(new_start_index) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1224); AnnotateMozCrashReason("MOZ_ASSERT" "(" "border < ranges->at(new_start_index) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1)" ")"); do { *((volatile int*)__null) = 1224; __attribute__((nomerge )) ::abort(); } while (false); } } while (false) | |||
| 1222 | (border == ranges->at(new_start_index) &&do { static_assert( mozilla::detail::AssertionConditionType< decltype(border < ranges->at(new_start_index) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(border < ranges->at(new_start_index ) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("border < ranges->at(new_start_index) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1224); AnnotateMozCrashReason("MOZ_ASSERT" "(" "border < ranges->at(new_start_index) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1)" ")"); do { *((volatile int*)__null) = 1224; __attribute__((nomerge )) ::abort(); } while (false); } } while (false) | |||
| 1223 | new_start_index == end_index && new_end_index == end_index - 1 &&do { static_assert( mozilla::detail::AssertionConditionType< decltype(border < ranges->at(new_start_index) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(border < ranges->at(new_start_index ) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("border < ranges->at(new_start_index) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1224); AnnotateMozCrashReason("MOZ_ASSERT" "(" "border < ranges->at(new_start_index) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1)" ")"); do { *((volatile int*)__null) = 1224; __attribute__((nomerge )) ::abort(); } while (false); } } while (false) | |||
| 1224 | border == last + 1))do { static_assert( mozilla::detail::AssertionConditionType< decltype(border < ranges->at(new_start_index) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(border < ranges->at(new_start_index ) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("border < ranges->at(new_start_index) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1224); AnnotateMozCrashReason("MOZ_ASSERT" "(" "border < ranges->at(new_start_index) || (border == ranges->at(new_start_index) && new_start_index == end_index && new_end_index == end_index - 1 && border == last + 1)" ")"); do { *((volatile int*)__null) = 1224; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1225 | DCHECK(new_start_index == 0 || border >= ranges->at(new_start_index - 1))do { static_assert( mozilla::detail::AssertionConditionType< decltype(new_start_index == 0 || border >= ranges->at(new_start_index - 1))>::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!(new_start_index == 0 || border >= ranges->at(new_start_index - 1)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("new_start_index == 0 || border >= ranges->at(new_start_index - 1)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1225); AnnotateMozCrashReason("MOZ_ASSERT" "(" "new_start_index == 0 || border >= ranges->at(new_start_index - 1)" ")"); do { *((volatile int*)__null) = 1225; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1226 | ||||
| 1227 | masm->CheckCharacterGT(border - 1, above); | |||
| 1228 | Label dummy; | |||
| 1229 | GenerateBranches(masm, ranges, start_index, new_end_index, min_char, | |||
| 1230 | border - 1, &dummy, even_label, odd_label); | |||
| 1231 | if (handle_rest.is_linked()) { | |||
| 1232 | masm->Bind(&handle_rest); | |||
| 1233 | bool flip = (new_start_index & 1) != (start_index & 1); | |||
| 1234 | GenerateBranches(masm, ranges, new_start_index, end_index, border, max_char, | |||
| 1235 | &dummy, flip ? odd_label : even_label, | |||
| 1236 | flip ? even_label : odd_label); | |||
| 1237 | } | |||
| 1238 | } | |||
| 1239 | ||||
| 1240 | void EmitClassRanges(RegExpMacroAssembler* macro_assembler, | |||
| 1241 | RegExpClassRanges* cr, bool one_byte, Label* on_failure, | |||
| 1242 | int cp_offset, bool check_offset, bool preloaded, | |||
| 1243 | Zone* zone) { | |||
| 1244 | ZoneList<CharacterRange>* ranges = cr->ranges(zone); | |||
| 1245 | CharacterRange::Canonicalize(ranges); | |||
| 1246 | ||||
| 1247 | // Now that all processing (like case-insensitivity) is done, clamp the | |||
| 1248 | // ranges to the set of ranges that may actually occur in the subject string. | |||
| 1249 | if (one_byte) CharacterRange::ClampToOneByte(ranges); | |||
| 1250 | ||||
| 1251 | const int ranges_length = ranges->length(); | |||
| 1252 | if (ranges_length == 0) { | |||
| 1253 | if (!cr->is_negated()) { | |||
| 1254 | macro_assembler->GoTo(on_failure); | |||
| 1255 | } | |||
| 1256 | if (check_offset) { | |||
| 1257 | macro_assembler->CheckPosition(cp_offset, on_failure); | |||
| 1258 | } | |||
| 1259 | return; | |||
| 1260 | } | |||
| 1261 | ||||
| 1262 | const base::uc32 max_char = MaxCodeUnit(one_byte); | |||
| 1263 | if (ranges_length == 1 && ranges->at(0).IsEverything(max_char)) { | |||
| 1264 | if (cr->is_negated()) { | |||
| 1265 | macro_assembler->GoTo(on_failure); | |||
| 1266 | } else { | |||
| 1267 | // This is a common case hit by non-anchored expressions. | |||
| 1268 | if (check_offset) { | |||
| 1269 | macro_assembler->CheckPosition(cp_offset, on_failure); | |||
| 1270 | } | |||
| 1271 | } | |||
| 1272 | return; | |||
| 1273 | } | |||
| 1274 | ||||
| 1275 | if (!preloaded) { | |||
| 1276 | macro_assembler->LoadCurrentCharacter(cp_offset, on_failure, check_offset); | |||
| 1277 | } | |||
| 1278 | ||||
| 1279 | if (cr->is_standard(zone) && macro_assembler->CheckSpecialClassRanges( | |||
| 1280 | cr->standard_type(), on_failure)) { | |||
| 1281 | return; | |||
| 1282 | } | |||
| 1283 | ||||
| 1284 | static constexpr int kMaxRangesForInlineBranchGeneration = 16; | |||
| 1285 | if (ranges_length > kMaxRangesForInlineBranchGeneration) { | |||
| 1286 | // For large range sets, emit a more compact instruction sequence to avoid | |||
| 1287 | // a potentially problematic increase in code size. | |||
| 1288 | // Note the flipped logic below (we check InRange if negated, NotInRange if | |||
| 1289 | // not negated); this is necessary since the method falls through on | |||
| 1290 | // failure whereas we want to fall through on success. | |||
| 1291 | if (cr->is_negated()) { | |||
| 1292 | if (macro_assembler->CheckCharacterInRangeArray(ranges, on_failure)) { | |||
| 1293 | return; | |||
| 1294 | } | |||
| 1295 | } else { | |||
| 1296 | if (macro_assembler->CheckCharacterNotInRangeArray(ranges, on_failure)) { | |||
| 1297 | return; | |||
| 1298 | } | |||
| 1299 | } | |||
| 1300 | } | |||
| 1301 | ||||
| 1302 | // Generate a flat list of range boundaries for consumption by | |||
| 1303 | // GenerateBranches. See the comment on that function for how the list should | |||
| 1304 | // be structured | |||
| 1305 | ZoneList<base::uc32>* range_boundaries = | |||
| 1306 | zone->New<ZoneList<base::uc32>>(ranges_length * 2, zone); | |||
| 1307 | ||||
| 1308 | bool zeroth_entry_is_failure = !cr->is_negated(); | |||
| 1309 | ||||
| 1310 | for (int i = 0; i < ranges_length; i++) { | |||
| 1311 | CharacterRange& range = ranges->at(i); | |||
| 1312 | if (range.from() == 0) { | |||
| 1313 | DCHECK_EQ(i, 0)do { static_assert( mozilla::detail::AssertionConditionType< decltype((i) == (0))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((i) == (0)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(i) == (0)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1313); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(i) == (0)" ")"); do { *((volatile int*)__null) = 1313; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1314 | zeroth_entry_is_failure = !zeroth_entry_is_failure; | |||
| 1315 | } else { | |||
| 1316 | range_boundaries->Add(range.from(), zone); | |||
| 1317 | } | |||
| 1318 | // `+ 1` to convert from inclusive to exclusive `to`. | |||
| 1319 | // [from, to] == [from, to+1[. | |||
| 1320 | range_boundaries->Add(range.to() + 1, zone); | |||
| 1321 | } | |||
| 1322 | int end_index = range_boundaries->length() - 1; | |||
| 1323 | if (range_boundaries->at(end_index) > max_char) { | |||
| 1324 | end_index--; | |||
| 1325 | } | |||
| 1326 | ||||
| 1327 | Label fall_through; | |||
| 1328 | GenerateBranches(macro_assembler, range_boundaries, | |||
| 1329 | 0, // start_index. | |||
| 1330 | end_index, | |||
| 1331 | 0, // min_char. | |||
| 1332 | max_char, &fall_through, | |||
| 1333 | zeroth_entry_is_failure ? &fall_through : on_failure, | |||
| 1334 | zeroth_entry_is_failure ? on_failure : &fall_through); | |||
| 1335 | macro_assembler->Bind(&fall_through); | |||
| 1336 | } | |||
| 1337 | ||||
| 1338 | } // namespace | |||
| 1339 | ||||
| 1340 | RegExpNode::~RegExpNode() = default; | |||
| 1341 | ||||
| 1342 | RegExpNode::LimitResult RegExpNode::LimitVersions(RegExpCompiler* compiler, | |||
| 1343 | Trace* trace) { | |||
| 1344 | // If we are generating a greedy loop then don't stop and don't reuse code. | |||
| 1345 | if (trace->stop_node() != nullptr) { | |||
| 1346 | return CONTINUE; | |||
| 1347 | } | |||
| 1348 | ||||
| 1349 | RegExpMacroAssembler* macro_assembler = compiler->macro_assembler(); | |||
| 1350 | if (trace->is_trivial()) { | |||
| 1351 | if (label_.is_bound() || on_work_list() || !KeepRecursing(compiler)) { | |||
| 1352 | // If a generic version is already scheduled to be generated or we have | |||
| 1353 | // recursed too deeply then just generate a jump to that code. | |||
| 1354 | macro_assembler->GoTo(&label_); | |||
| 1355 | // This will queue it up for generation of a generic version if it hasn't | |||
| 1356 | // already been queued. | |||
| 1357 | compiler->AddWork(this); | |||
| 1358 | return DONE; | |||
| 1359 | } | |||
| 1360 | // Generate generic version of the node and bind the label for later use. | |||
| 1361 | macro_assembler->Bind(&label_); | |||
| 1362 | return CONTINUE; | |||
| 1363 | } | |||
| 1364 | ||||
| 1365 | // We are being asked to make a non-generic version. Keep track of how many | |||
| 1366 | // non-generic versions we generate so as not to overdo it. | |||
| 1367 | trace_count_++; | |||
| 1368 | if (KeepRecursing(compiler) && compiler->optimize() && | |||
| 1369 | trace_count_ < kMaxCopiesCodeGenerated) { | |||
| 1370 | return CONTINUE; | |||
| 1371 | } | |||
| 1372 | ||||
| 1373 | // If we get here code has been generated for this node too many times or | |||
| 1374 | // recursion is too deep. Time to switch to a generic version. The code for | |||
| 1375 | // generic versions above can handle deep recursion properly. | |||
| 1376 | bool was_limiting = compiler->limiting_recursion(); | |||
| 1377 | compiler->set_limiting_recursion(true); | |||
| 1378 | trace->Flush(compiler, this); | |||
| 1379 | compiler->set_limiting_recursion(was_limiting); | |||
| 1380 | return DONE; | |||
| 1381 | } | |||
| 1382 | ||||
| 1383 | bool RegExpNode::KeepRecursing(RegExpCompiler* compiler) { | |||
| 1384 | return !compiler->limiting_recursion() && | |||
| 1385 | compiler->recursion_depth() <= RegExpCompiler::kMaxRecursion; | |||
| 1386 | } | |||
| 1387 | ||||
| 1388 | void ActionNode::FillInBMInfo(Isolate* isolate, int offset, int budget, | |||
| 1389 | BoyerMooreLookahead* bm, bool not_at_start) { | |||
| 1390 | base::Optional<RegExpFlags> old_flags; | |||
| 1391 | if (action_type_ == MODIFY_FLAGS) { | |||
| 1392 | // It is not guaranteed that we hit the resetting modify flags node, due to | |||
| 1393 | // recursion budget limitation for filling in BMInfo. Therefore we reset the | |||
| 1394 | // flags manually to the previous state after recursing. | |||
| 1395 | old_flags = bm->compiler()->flags(); | |||
| 1396 | bm->compiler()->set_flags(flags()); | |||
| 1397 | } | |||
| 1398 | if (action_type_ == POSITIVE_SUBMATCH_SUCCESS) { | |||
| 1399 | // Anything may follow a positive submatch success, thus we need to accept | |||
| 1400 | // all characters from this position onwards. | |||
| 1401 | bm->SetRest(offset); | |||
| 1402 | } else { | |||
| 1403 | on_success()->FillInBMInfo(isolate, offset, budget - 1, bm, not_at_start); | |||
| 1404 | } | |||
| 1405 | SaveBMInfo(bm, not_at_start, offset); | |||
| 1406 | if (old_flags.has_value()) { | |||
| 1407 | bm->compiler()->set_flags(*old_flags); | |||
| 1408 | } | |||
| 1409 | } | |||
| 1410 | ||||
| 1411 | void ActionNode::GetQuickCheckDetails(QuickCheckDetails* details, | |||
| 1412 | RegExpCompiler* compiler, int filled_in, | |||
| 1413 | bool not_at_start) { | |||
| 1414 | if (action_type_ == SET_REGISTER_FOR_LOOP) { | |||
| 1415 | on_success()->GetQuickCheckDetailsFromLoopEntry(details, compiler, | |||
| 1416 | filled_in, not_at_start); | |||
| 1417 | } else { | |||
| 1418 | if (action_type() == MODIFY_FLAGS) { | |||
| 1419 | compiler->set_flags(flags()); | |||
| 1420 | } | |||
| 1421 | on_success()->GetQuickCheckDetails(details, compiler, filled_in, | |||
| 1422 | not_at_start); | |||
| 1423 | } | |||
| 1424 | } | |||
| 1425 | ||||
| 1426 | void AssertionNode::FillInBMInfo(Isolate* isolate, int offset, int budget, | |||
| 1427 | BoyerMooreLookahead* bm, bool not_at_start) { | |||
| 1428 | // Match the behaviour of EatsAtLeast on this node. | |||
| 1429 | if (assertion_type() == AT_START && not_at_start) return; | |||
| 1430 | on_success()->FillInBMInfo(isolate, offset, budget - 1, bm, not_at_start); | |||
| 1431 | SaveBMInfo(bm, not_at_start, offset); | |||
| 1432 | } | |||
| 1433 | ||||
| 1434 | void NegativeLookaroundChoiceNode::GetQuickCheckDetails( | |||
| 1435 | QuickCheckDetails* details, RegExpCompiler* compiler, int filled_in, | |||
| 1436 | bool not_at_start) { | |||
| 1437 | RegExpNode* node = continue_node(); | |||
| 1438 | return node->GetQuickCheckDetails(details, compiler, filled_in, not_at_start); | |||
| 1439 | } | |||
| 1440 | ||||
| 1441 | namespace { | |||
| 1442 | ||||
| 1443 | // Takes the left-most 1-bit and smears it out, setting all bits to its right. | |||
| 1444 | inline uint32_t SmearBitsRight(uint32_t v) { | |||
| 1445 | v |= v >> 1; | |||
| 1446 | v |= v >> 2; | |||
| 1447 | v |= v >> 4; | |||
| 1448 | v |= v >> 8; | |||
| 1449 | v |= v >> 16; | |||
| 1450 | return v; | |||
| 1451 | } | |||
| 1452 | ||||
| 1453 | } // namespace | |||
| 1454 | ||||
| 1455 | bool QuickCheckDetails::Rationalize(bool asc) { | |||
| 1456 | bool found_useful_op = false; | |||
| 1457 | const uint32_t char_mask = CharMask(asc); | |||
| 1458 | mask_ = 0; | |||
| 1459 | value_ = 0; | |||
| 1460 | int char_shift = 0; | |||
| 1461 | for (int i = 0; i < characters_; i++) { | |||
| 1462 | Position* pos = &positions_[i]; | |||
| 1463 | if ((pos->mask & String::kMaxOneByteCharCode) != 0) { | |||
| 1464 | found_useful_op = true; | |||
| 1465 | } | |||
| 1466 | mask_ |= (pos->mask & char_mask) << char_shift; | |||
| ||||
| 1467 | value_ |= (pos->value & char_mask) << char_shift; | |||
| 1468 | char_shift += asc
| |||
| 1469 | } | |||
| 1470 | return found_useful_op; | |||
| 1471 | } | |||
| 1472 | ||||
| 1473 | int RegExpNode::EatsAtLeast(bool not_at_start) { | |||
| 1474 | return not_at_start ? eats_at_least_.eats_at_least_from_not_start | |||
| 1475 | : eats_at_least_.eats_at_least_from_possibly_start; | |||
| 1476 | } | |||
| 1477 | ||||
| 1478 | EatsAtLeastInfo RegExpNode::EatsAtLeastFromLoopEntry() { | |||
| 1479 | // SET_REGISTER_FOR_LOOP is only used to initialize loop counters, and it | |||
| 1480 | // implies that the following node must be a LoopChoiceNode. If we need to | |||
| 1481 | // set registers to constant values for other reasons, we could introduce a | |||
| 1482 | // new action type SET_REGISTER that doesn't imply anything about its | |||
| 1483 | // successor. | |||
| 1484 | UNREACHABLE()do { do { } while (false); MOZ_ReportCrash("" "unreachable code" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1484); AnnotateMozCrashReason("MOZ_CRASH(" "unreachable code" ")"); do { *((volatile int*)__null) = 1484; __attribute__((nomerge )) ::abort(); } while (false); } while (false); | |||
| 1485 | } | |||
| 1486 | ||||
| 1487 | void RegExpNode::GetQuickCheckDetailsFromLoopEntry(QuickCheckDetails* details, | |||
| 1488 | RegExpCompiler* compiler, | |||
| 1489 | int characters_filled_in, | |||
| 1490 | bool not_at_start) { | |||
| 1491 | // See comment in RegExpNode::EatsAtLeastFromLoopEntry. | |||
| 1492 | UNREACHABLE()do { do { } while (false); MOZ_ReportCrash("" "unreachable code" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1492); AnnotateMozCrashReason("MOZ_CRASH(" "unreachable code" ")"); do { *((volatile int*)__null) = 1492; __attribute__((nomerge )) ::abort(); } while (false); } while (false); | |||
| 1493 | } | |||
| 1494 | ||||
| 1495 | EatsAtLeastInfo LoopChoiceNode::EatsAtLeastFromLoopEntry() { | |||
| 1496 | DCHECK_EQ(alternatives_->length(), 2)do { static_assert( mozilla::detail::AssertionConditionType< decltype((alternatives_->length()) == (2))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((alternatives_->length()) == (2)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(alternatives_->length()) == (2)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1496); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(alternatives_->length()) == (2)" ")"); do { *((volatile int*)__null) = 1496; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); // There's just loop and continue. | |||
| 1497 | ||||
| 1498 | if (read_backward()) { | |||
| 1499 | // The eats_at_least value is not used if reading backward. The | |||
| 1500 | // EatsAtLeastPropagator should've zeroed it as well. | |||
| 1501 | DCHECK_EQ(eats_at_least_info()->eats_at_least_from_possibly_start, 0)do { static_assert( mozilla::detail::AssertionConditionType< decltype((eats_at_least_info()->eats_at_least_from_possibly_start ) == (0))>::isValid, "invalid assertion condition"); if (( __builtin_expect(!!(!(!!((eats_at_least_info()->eats_at_least_from_possibly_start ) == (0)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(eats_at_least_info()->eats_at_least_from_possibly_start) == (0)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1501); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(eats_at_least_info()->eats_at_least_from_possibly_start) == (0)" ")"); do { *((volatile int*)__null) = 1501; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1502 | DCHECK_EQ(eats_at_least_info()->eats_at_least_from_not_start, 0)do { static_assert( mozilla::detail::AssertionConditionType< decltype((eats_at_least_info()->eats_at_least_from_not_start ) == (0))>::isValid, "invalid assertion condition"); if (( __builtin_expect(!!(!(!!((eats_at_least_info()->eats_at_least_from_not_start ) == (0)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(eats_at_least_info()->eats_at_least_from_not_start) == (0)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1502); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(eats_at_least_info()->eats_at_least_from_not_start) == (0)" ")"); do { *((volatile int*)__null) = 1502; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1503 | return {}; | |||
| 1504 | } | |||
| 1505 | ||||
| 1506 | // Figure out how much the loop body itself eats, not including anything in | |||
| 1507 | // the continuation case. In general, the nodes in the loop body should report | |||
| 1508 | // that they eat at least the number eaten by the continuation node, since any | |||
| 1509 | // successful match in the loop body must also include the continuation node. | |||
| 1510 | // However, in some cases involving positive lookaround, the loop body under- | |||
| 1511 | // reports its appetite, so use saturated math here to avoid negative numbers. | |||
| 1512 | uint8_t loop_body_from_not_start = base::saturated_cast<uint8_t>( | |||
| 1513 | loop_node_->EatsAtLeast(true) - continue_node_->EatsAtLeast(true)); | |||
| 1514 | uint8_t loop_body_from_possibly_start = base::saturated_cast<uint8_t>( | |||
| 1515 | loop_node_->EatsAtLeast(false) - continue_node_->EatsAtLeast(true)); | |||
| 1516 | ||||
| 1517 | // Limit the number of loop iterations to avoid overflow in subsequent steps. | |||
| 1518 | int loop_iterations = base::saturated_cast<uint8_t>(min_loop_iterations()); | |||
| 1519 | ||||
| 1520 | EatsAtLeastInfo result; | |||
| 1521 | result.eats_at_least_from_not_start = | |||
| 1522 | base::saturated_cast<uint8_t>(loop_iterations * loop_body_from_not_start + | |||
| 1523 | continue_node_->EatsAtLeast(true)); | |||
| 1524 | if (loop_iterations > 0 && loop_body_from_possibly_start > 0) { | |||
| 1525 | // First loop iteration eats at least one, so all subsequent iterations | |||
| 1526 | // and the after-loop chunk are guaranteed to not be at the start. | |||
| 1527 | result.eats_at_least_from_possibly_start = base::saturated_cast<uint8_t>( | |||
| 1528 | loop_body_from_possibly_start + | |||
| 1529 | (loop_iterations - 1) * loop_body_from_not_start + | |||
| 1530 | continue_node_->EatsAtLeast(true)); | |||
| 1531 | } else { | |||
| 1532 | // Loop body might eat nothing, so only continue node contributes. | |||
| 1533 | result.eats_at_least_from_possibly_start = | |||
| 1534 | continue_node_->EatsAtLeast(false); | |||
| 1535 | } | |||
| 1536 | return result; | |||
| 1537 | } | |||
| 1538 | ||||
| 1539 | bool RegExpNode::EmitQuickCheck(RegExpCompiler* compiler, | |||
| 1540 | Trace* bounds_check_trace, Trace* trace, | |||
| 1541 | bool preload_has_checked_bounds, | |||
| 1542 | Label* on_possible_success, | |||
| 1543 | QuickCheckDetails* details, | |||
| 1544 | bool fall_through_on_failure, | |||
| 1545 | ChoiceNode* predecessor) { | |||
| 1546 | DCHECK_NOT_NULL(predecessor)do { static_assert( mozilla::detail::AssertionConditionType< decltype((predecessor) != nullptr)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((predecessor) != nullptr))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(predecessor) != nullptr" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1546); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(predecessor) != nullptr" ")"); do { *((volatile int*)__null) = 1546; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1547 | if (details->characters() == 0) return false; | |||
| 1548 | GetQuickCheckDetails(details, compiler, 0, | |||
| 1549 | trace->at_start() == Trace::FALSE_VALUE); | |||
| 1550 | if (details->cannot_match()) return false; | |||
| 1551 | if (!details->Rationalize(compiler->one_byte())) return false; | |||
| 1552 | DCHECK(details->characters() == 1 ||do { static_assert( mozilla::detail::AssertionConditionType< decltype(details->characters() == 1 || compiler->macro_assembler ()->CanReadUnaligned())>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(details->characters() == 1 || compiler->macro_assembler()->CanReadUnaligned()))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("details->characters() == 1 || compiler->macro_assembler()->CanReadUnaligned()" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1553); AnnotateMozCrashReason("MOZ_ASSERT" "(" "details->characters() == 1 || compiler->macro_assembler()->CanReadUnaligned()" ")"); do { *((volatile int*)__null) = 1553; __attribute__((nomerge )) ::abort(); } while (false); } } while (false) | |||
| 1553 | compiler->macro_assembler()->CanReadUnaligned())do { static_assert( mozilla::detail::AssertionConditionType< decltype(details->characters() == 1 || compiler->macro_assembler ()->CanReadUnaligned())>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(details->characters() == 1 || compiler->macro_assembler()->CanReadUnaligned()))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("details->characters() == 1 || compiler->macro_assembler()->CanReadUnaligned()" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1553); AnnotateMozCrashReason("MOZ_ASSERT" "(" "details->characters() == 1 || compiler->macro_assembler()->CanReadUnaligned()" ")"); do { *((volatile int*)__null) = 1553; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1554 | uint32_t mask = details->mask(); | |||
| 1555 | uint32_t value = details->value(); | |||
| 1556 | ||||
| 1557 | RegExpMacroAssembler* assembler = compiler->macro_assembler(); | |||
| 1558 | ||||
| 1559 | if (trace->characters_preloaded() != details->characters()) { | |||
| 1560 | DCHECK(trace->cp_offset() == bounds_check_trace->cp_offset())do { static_assert( mozilla::detail::AssertionConditionType< decltype(trace->cp_offset() == bounds_check_trace->cp_offset ())>::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!(trace->cp_offset() == bounds_check_trace->cp_offset ()))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("trace->cp_offset() == bounds_check_trace->cp_offset()" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1560); AnnotateMozCrashReason("MOZ_ASSERT" "(" "trace->cp_offset() == bounds_check_trace->cp_offset()" ")"); do { *((volatile int*)__null) = 1560; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1561 | // The bounds check is performed using the minimum number of characters | |||
| 1562 | // any choice would eat, so if the bounds check fails, then none of the | |||
| 1563 | // choices can succeed, so we can just immediately backtrack, rather | |||
| 1564 | // than go to the next choice. The number of characters preloaded may be | |||
| 1565 | // less than the number used for the bounds check. | |||
| 1566 | int eats_at_least = predecessor->EatsAtLeast( | |||
| 1567 | bounds_check_trace->at_start() == Trace::FALSE_VALUE); | |||
| 1568 | DCHECK_GE(eats_at_least, details->characters())do { static_assert( mozilla::detail::AssertionConditionType< decltype((eats_at_least) >= (details->characters()))> ::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!((eats_at_least) >= (details->characters())))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(eats_at_least) >= (details->characters())" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1568); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(eats_at_least) >= (details->characters())" ")"); do { *((volatile int*)__null) = 1568; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1569 | assembler->LoadCurrentCharacter( | |||
| 1570 | trace->cp_offset(), bounds_check_trace->backtrack(), | |||
| 1571 | !preload_has_checked_bounds, details->characters(), eats_at_least); | |||
| 1572 | } | |||
| 1573 | ||||
| 1574 | bool need_mask = true; | |||
| 1575 | ||||
| 1576 | if (details->characters() == 1) { | |||
| 1577 | // If number of characters preloaded is 1 then we used a byte or 16 bit | |||
| 1578 | // load so the value is already masked down. | |||
| 1579 | const uint32_t char_mask = CharMask(compiler->one_byte()); | |||
| 1580 | if ((mask & char_mask) == char_mask) need_mask = false; | |||
| 1581 | mask &= char_mask; | |||
| 1582 | } else { | |||
| 1583 | // For 2-character preloads in one-byte mode or 1-character preloads in | |||
| 1584 | // two-byte mode we also use a 16 bit load with zero extend. | |||
| 1585 | static const uint32_t kTwoByteMask = 0xFFFF; | |||
| 1586 | static const uint32_t kFourByteMask = 0xFFFFFFFF; | |||
| 1587 | if (details->characters() == 2 && compiler->one_byte()) { | |||
| 1588 | if ((mask & kTwoByteMask) == kTwoByteMask) need_mask = false; | |||
| 1589 | } else if (details->characters() == 1 && !compiler->one_byte()) { | |||
| 1590 | if ((mask & kTwoByteMask) == kTwoByteMask) need_mask = false; | |||
| 1591 | } else { | |||
| 1592 | if (mask == kFourByteMask) need_mask = false; | |||
| 1593 | } | |||
| 1594 | } | |||
| 1595 | ||||
| 1596 | if (fall_through_on_failure) { | |||
| 1597 | if (need_mask) { | |||
| 1598 | assembler->CheckCharacterAfterAnd(value, mask, on_possible_success); | |||
| 1599 | } else { | |||
| 1600 | assembler->CheckCharacter(value, on_possible_success); | |||
| 1601 | } | |||
| 1602 | } else { | |||
| 1603 | if (need_mask) { | |||
| 1604 | assembler->CheckNotCharacterAfterAnd(value, mask, trace->backtrack()); | |||
| 1605 | } else { | |||
| 1606 | assembler->CheckNotCharacter(value, trace->backtrack()); | |||
| 1607 | } | |||
| 1608 | } | |||
| 1609 | return true; | |||
| 1610 | } | |||
| 1611 | ||||
| 1612 | // Here is the meat of GetQuickCheckDetails (see also the comment on the | |||
| 1613 | // super-class in the .h file). | |||
| 1614 | // | |||
| 1615 | // We iterate along the text object, building up for each character a | |||
| 1616 | // mask and value that can be used to test for a quick failure to match. | |||
| 1617 | // The masks and values for the positions will be combined into a single | |||
| 1618 | // machine word for the current character width in order to be used in | |||
| 1619 | // generating a quick check. | |||
| 1620 | void TextNode::GetQuickCheckDetails(QuickCheckDetails* details, | |||
| 1621 | RegExpCompiler* compiler, | |||
| 1622 | int characters_filled_in, | |||
| 1623 | bool not_at_start) { | |||
| 1624 | // Do not collect any quick check details if the text node reads backward, | |||
| 1625 | // since it reads in the opposite direction than we use for quick checks. | |||
| 1626 | if (read_backward()) return; | |||
| 1627 | Isolate* isolate = compiler->macro_assembler()->isolate(); | |||
| 1628 | DCHECK(characters_filled_in < details->characters())do { static_assert( mozilla::detail::AssertionConditionType< decltype(characters_filled_in < details->characters())> ::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!(characters_filled_in < details->characters())) ), 0))) { do { } while (false); MOZ_ReportAssertionFailure("characters_filled_in < details->characters()" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1628); AnnotateMozCrashReason("MOZ_ASSERT" "(" "characters_filled_in < details->characters()" ")"); do { *((volatile int*)__null) = 1628; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1629 | int characters = details->characters(); | |||
| 1630 | const uint32_t char_mask = CharMask(compiler->one_byte()); | |||
| 1631 | for (int k = 0; k < elements()->length(); k++) { | |||
| 1632 | TextElement elm = elements()->at(k); | |||
| 1633 | if (elm.text_type() == TextElement::ATOM) { | |||
| 1634 | base::Vector<const base::uc16> quarks = elm.atom()->data(); | |||
| 1635 | for (int i = 0; i < characters && i < quarks.length(); i++) { | |||
| 1636 | QuickCheckDetails::Position* pos = | |||
| 1637 | details->positions(characters_filled_in); | |||
| 1638 | base::uc16 c = quarks[i]; | |||
| 1639 | if (IsIgnoreCase(compiler->flags())) { | |||
| 1640 | unibrow::uchar chars[4]; | |||
| 1641 | int length = | |||
| 1642 | GetCaseIndependentLetters(isolate, c, compiler, chars, 4); | |||
| 1643 | if (length == 0) { | |||
| 1644 | // This can happen because all case variants are non-Latin1, but we | |||
| 1645 | // know the input is Latin1. | |||
| 1646 | details->set_cannot_match(); | |||
| 1647 | pos->determines_perfectly = false; | |||
| 1648 | return; | |||
| 1649 | } | |||
| 1650 | if (length == 1) { | |||
| 1651 | // This letter has no case equivalents, so it's nice and simple | |||
| 1652 | // and the mask-compare will determine definitely whether we have | |||
| 1653 | // a match at this character position. | |||
| 1654 | pos->mask = char_mask; | |||
| 1655 | pos->value = chars[0]; | |||
| 1656 | pos->determines_perfectly = true; | |||
| 1657 | } else { | |||
| 1658 | uint32_t common_bits = char_mask; | |||
| 1659 | uint32_t bits = chars[0]; | |||
| 1660 | for (int j = 1; j < length; j++) { | |||
| 1661 | uint32_t differing_bits = ((chars[j] & common_bits) ^ bits); | |||
| 1662 | common_bits ^= differing_bits; | |||
| 1663 | bits &= common_bits; | |||
| 1664 | } | |||
| 1665 | // If length is 2 and common bits has only one zero in it then | |||
| 1666 | // our mask and compare instruction will determine definitely | |||
| 1667 | // whether we have a match at this character position. Otherwise | |||
| 1668 | // it can only be an approximate check. | |||
| 1669 | uint32_t one_zero = (common_bits | ~char_mask); | |||
| 1670 | if (length == 2 && ((~one_zero) & ((~one_zero) - 1)) == 0) { | |||
| 1671 | pos->determines_perfectly = true; | |||
| 1672 | } | |||
| 1673 | pos->mask = common_bits; | |||
| 1674 | pos->value = bits; | |||
| 1675 | } | |||
| 1676 | } else { | |||
| 1677 | // Don't ignore case. Nice simple case where the mask-compare will | |||
| 1678 | // determine definitely whether we have a match at this character | |||
| 1679 | // position. | |||
| 1680 | if (c > char_mask) { | |||
| 1681 | details->set_cannot_match(); | |||
| 1682 | pos->determines_perfectly = false; | |||
| 1683 | return; | |||
| 1684 | } | |||
| 1685 | pos->mask = char_mask; | |||
| 1686 | pos->value = c; | |||
| 1687 | pos->determines_perfectly = true; | |||
| 1688 | } | |||
| 1689 | characters_filled_in++; | |||
| 1690 | DCHECK(characters_filled_in <= details->characters())do { static_assert( mozilla::detail::AssertionConditionType< decltype(characters_filled_in <= details->characters()) >::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!(characters_filled_in <= details->characters()) )), 0))) { do { } while (false); MOZ_ReportAssertionFailure("characters_filled_in <= details->characters()" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1690); AnnotateMozCrashReason("MOZ_ASSERT" "(" "characters_filled_in <= details->characters()" ")"); do { *((volatile int*)__null) = 1690; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1691 | if (characters_filled_in == details->characters()) { | |||
| 1692 | return; | |||
| 1693 | } | |||
| 1694 | } | |||
| 1695 | } else { | |||
| 1696 | QuickCheckDetails::Position* pos = | |||
| 1697 | details->positions(characters_filled_in); | |||
| 1698 | RegExpClassRanges* tree = elm.class_ranges(); | |||
| 1699 | ZoneList<CharacterRange>* ranges = tree->ranges(zone()); | |||
| 1700 | if (tree->is_negated() || ranges->is_empty()) { | |||
| 1701 | // A quick check uses multi-character mask and compare. There is no | |||
| 1702 | // useful way to incorporate a negative char class into this scheme | |||
| 1703 | // so we just conservatively create a mask and value that will always | |||
| 1704 | // succeed. | |||
| 1705 | // Likewise for empty ranges (empty ranges can occur e.g. when | |||
| 1706 | // compiling for one-byte subjects and impossible (non-one-byte) ranges | |||
| 1707 | // have been removed). | |||
| 1708 | pos->mask = 0; | |||
| 1709 | pos->value = 0; | |||
| 1710 | } else { | |||
| 1711 | int first_range = 0; | |||
| 1712 | while (ranges->at(first_range).from() > char_mask) { | |||
| 1713 | first_range++; | |||
| 1714 | if (first_range == ranges->length()) { | |||
| 1715 | details->set_cannot_match(); | |||
| 1716 | pos->determines_perfectly = false; | |||
| 1717 | return; | |||
| 1718 | } | |||
| 1719 | } | |||
| 1720 | CharacterRange range = ranges->at(first_range); | |||
| 1721 | const base::uc32 first_from = range.from(); | |||
| 1722 | const base::uc32 first_to = | |||
| 1723 | (range.to() > char_mask) ? char_mask : range.to(); | |||
| 1724 | const uint32_t differing_bits = (first_from ^ first_to); | |||
| 1725 | // A mask and compare is only perfect if the differing bits form a | |||
| 1726 | // number like 00011111 with one single block of trailing 1s. | |||
| 1727 | if ((differing_bits & (differing_bits + 1)) == 0 && | |||
| 1728 | first_from + differing_bits == first_to) { | |||
| 1729 | pos->determines_perfectly = true; | |||
| 1730 | } | |||
| 1731 | uint32_t common_bits = ~SmearBitsRight(differing_bits); | |||
| 1732 | uint32_t bits = (first_from & common_bits); | |||
| 1733 | for (int i = first_range + 1; i < ranges->length(); i++) { | |||
| 1734 | range = ranges->at(i); | |||
| 1735 | const base::uc32 from = range.from(); | |||
| 1736 | if (from > char_mask) continue; | |||
| 1737 | const base::uc32 to = | |||
| 1738 | (range.to() > char_mask) ? char_mask : range.to(); | |||
| 1739 | // Here we are combining more ranges into the mask and compare | |||
| 1740 | // value. With each new range the mask becomes more sparse and | |||
| 1741 | // so the chances of a false positive rise. A character class | |||
| 1742 | // with multiple ranges is assumed never to be equivalent to a | |||
| 1743 | // mask and compare operation. | |||
| 1744 | pos->determines_perfectly = false; | |||
| 1745 | uint32_t new_common_bits = (from ^ to); | |||
| 1746 | new_common_bits = ~SmearBitsRight(new_common_bits); | |||
| 1747 | common_bits &= new_common_bits; | |||
| 1748 | bits &= new_common_bits; | |||
| 1749 | uint32_t new_differing_bits = (from & common_bits) ^ bits; | |||
| 1750 | common_bits ^= new_differing_bits; | |||
| 1751 | bits &= common_bits; | |||
| 1752 | } | |||
| 1753 | pos->mask = common_bits; | |||
| 1754 | pos->value = bits; | |||
| 1755 | } | |||
| 1756 | characters_filled_in++; | |||
| 1757 | DCHECK(characters_filled_in <= details->characters())do { static_assert( mozilla::detail::AssertionConditionType< decltype(characters_filled_in <= details->characters()) >::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!(characters_filled_in <= details->characters()) )), 0))) { do { } while (false); MOZ_ReportAssertionFailure("characters_filled_in <= details->characters()" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1757); AnnotateMozCrashReason("MOZ_ASSERT" "(" "characters_filled_in <= details->characters()" ")"); do { *((volatile int*)__null) = 1757; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1758 | if (characters_filled_in == details->characters()) return; | |||
| 1759 | } | |||
| 1760 | } | |||
| 1761 | DCHECK(characters_filled_in != details->characters())do { static_assert( mozilla::detail::AssertionConditionType< decltype(characters_filled_in != details->characters())> ::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!(characters_filled_in != details->characters()))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("characters_filled_in != details->characters()" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1761); AnnotateMozCrashReason("MOZ_ASSERT" "(" "characters_filled_in != details->characters()" ")"); do { *((volatile int*)__null) = 1761; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1762 | if (!details->cannot_match()) { | |||
| 1763 | on_success()->GetQuickCheckDetails(details, compiler, characters_filled_in, | |||
| 1764 | true); | |||
| 1765 | } | |||
| 1766 | } | |||
| 1767 | ||||
| 1768 | void QuickCheckDetails::Clear() { | |||
| 1769 | for (int i = 0; i < characters_; i++) { | |||
| 1770 | positions_[i].mask = 0; | |||
| 1771 | positions_[i].value = 0; | |||
| 1772 | positions_[i].determines_perfectly = false; | |||
| 1773 | } | |||
| 1774 | characters_ = 0; | |||
| 1775 | } | |||
| 1776 | ||||
| 1777 | void QuickCheckDetails::Advance(int by, bool one_byte) { | |||
| 1778 | if (by >= characters_ || by < 0) { | |||
| 1779 | DCHECK_IMPLIES(by < 0, characters_ == 0)do { if (by < 0) { do { static_assert( mozilla::detail::AssertionConditionType <decltype(characters_ == 0)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(characters_ == 0))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("characters_ == 0" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1779); AnnotateMozCrashReason("MOZ_ASSERT" "(" "characters_ == 0" ")"); do { *((volatile int*)__null) = 1779; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); } } while ( false); | |||
| 1780 | Clear(); | |||
| 1781 | return; | |||
| 1782 | } | |||
| 1783 | DCHECK_LE(characters_ - by, 4)do { static_assert( mozilla::detail::AssertionConditionType< decltype((characters_ - by) <= (4))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((characters_ - by) <= (4) ))), 0))) { do { } while (false); MOZ_ReportAssertionFailure( "(characters_ - by) <= (4)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1783); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(characters_ - by) <= (4)" ")"); do { *((volatile int*)__null) = 1783; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1784 | DCHECK_LE(characters_, 4)do { static_assert( mozilla::detail::AssertionConditionType< decltype((characters_) <= (4))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((characters_) <= (4)))), 0 ))) { do { } while (false); MOZ_ReportAssertionFailure("(characters_) <= (4)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1784); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(characters_) <= (4)" ")"); do { *((volatile int*)__null) = 1784; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1785 | for (int i = 0; i < characters_ - by; i++) { | |||
| 1786 | positions_[i] = positions_[by + i]; | |||
| 1787 | } | |||
| 1788 | for (int i = characters_ - by; i < characters_; i++) { | |||
| 1789 | positions_[i].mask = 0; | |||
| 1790 | positions_[i].value = 0; | |||
| 1791 | positions_[i].determines_perfectly = false; | |||
| 1792 | } | |||
| 1793 | characters_ -= by; | |||
| 1794 | // We could change mask_ and value_ here but we would never advance unless | |||
| 1795 | // they had already been used in a check and they won't be used again because | |||
| 1796 | // it would gain us nothing. So there's no point. | |||
| 1797 | } | |||
| 1798 | ||||
| 1799 | void QuickCheckDetails::Merge(QuickCheckDetails* other, int from_index) { | |||
| 1800 | DCHECK(characters_ == other->characters_)do { static_assert( mozilla::detail::AssertionConditionType< decltype(characters_ == other->characters_)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(characters_ == other->characters_ ))), 0))) { do { } while (false); MOZ_ReportAssertionFailure( "characters_ == other->characters_", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1800); AnnotateMozCrashReason("MOZ_ASSERT" "(" "characters_ == other->characters_" ")"); do { *((volatile int*)__null) = 1800; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1801 | if (other->cannot_match_) { | |||
| 1802 | return; | |||
| 1803 | } | |||
| 1804 | if (cannot_match_) { | |||
| 1805 | *this = *other; | |||
| 1806 | return; | |||
| 1807 | } | |||
| 1808 | for (int i = from_index; i < characters_; i++) { | |||
| 1809 | QuickCheckDetails::Position* pos = positions(i); | |||
| 1810 | QuickCheckDetails::Position* other_pos = other->positions(i); | |||
| 1811 | if (pos->mask != other_pos->mask || pos->value != other_pos->value || | |||
| 1812 | !other_pos->determines_perfectly) { | |||
| 1813 | // Our mask-compare operation will be approximate unless we have the | |||
| 1814 | // exact same operation on both sides of the alternation. | |||
| 1815 | pos->determines_perfectly = false; | |||
| 1816 | } | |||
| 1817 | pos->mask &= other_pos->mask; | |||
| 1818 | pos->value &= pos->mask; | |||
| 1819 | other_pos->value &= pos->mask; | |||
| 1820 | uint32_t differing_bits = (pos->value ^ other_pos->value); | |||
| 1821 | pos->mask &= ~differing_bits; | |||
| 1822 | pos->value &= pos->mask; | |||
| 1823 | } | |||
| 1824 | } | |||
| 1825 | ||||
| 1826 | class VisitMarker { | |||
| 1827 | public: | |||
| 1828 | explicit VisitMarker(NodeInfo* info) : info_(info) { | |||
| 1829 | DCHECK(!info->visited)do { static_assert( mozilla::detail::AssertionConditionType< decltype(!info->visited)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(!info->visited))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("!info->visited" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1829); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!info->visited" ")"); do { *((volatile int*)__null) = 1829; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1830 | info->visited = true; | |||
| 1831 | } | |||
| 1832 | ~VisitMarker() { info_->visited = false; } | |||
| 1833 | ||||
| 1834 | private: | |||
| 1835 | NodeInfo* info_; | |||
| 1836 | }; | |||
| 1837 | ||||
| 1838 | // Temporarily sets traversed_loop_initialization_node_. | |||
| 1839 | class LoopInitializationMarker { | |||
| 1840 | public: | |||
| 1841 | explicit LoopInitializationMarker(LoopChoiceNode* node) : node_(node) { | |||
| 1842 | DCHECK(!node_->traversed_loop_initialization_node_)do { static_assert( mozilla::detail::AssertionConditionType< decltype(!node_->traversed_loop_initialization_node_)>:: isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!(!node_->traversed_loop_initialization_node_))), 0 ))) { do { } while (false); MOZ_ReportAssertionFailure("!node_->traversed_loop_initialization_node_" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1842); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!node_->traversed_loop_initialization_node_" ")"); do { *((volatile int*)__null) = 1842; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1843 | node_->traversed_loop_initialization_node_ = true; | |||
| 1844 | } | |||
| 1845 | ~LoopInitializationMarker() { | |||
| 1846 | DCHECK(node_->traversed_loop_initialization_node_)do { static_assert( mozilla::detail::AssertionConditionType< decltype(node_->traversed_loop_initialization_node_)>:: isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!(node_->traversed_loop_initialization_node_))), 0) )) { do { } while (false); MOZ_ReportAssertionFailure("node_->traversed_loop_initialization_node_" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1846); AnnotateMozCrashReason("MOZ_ASSERT" "(" "node_->traversed_loop_initialization_node_" ")"); do { *((volatile int*)__null) = 1846; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1847 | node_->traversed_loop_initialization_node_ = false; | |||
| 1848 | } | |||
| 1849 | LoopInitializationMarker(const LoopInitializationMarker&) = delete; | |||
| 1850 | LoopInitializationMarker& operator=(const LoopInitializationMarker&) = delete; | |||
| 1851 | ||||
| 1852 | private: | |||
| 1853 | LoopChoiceNode* node_; | |||
| 1854 | }; | |||
| 1855 | ||||
| 1856 | // Temporarily decrements min_loop_iterations_. | |||
| 1857 | class IterationDecrementer { | |||
| 1858 | public: | |||
| 1859 | explicit IterationDecrementer(LoopChoiceNode* node) : node_(node) { | |||
| 1860 | DCHECK_GT(node_->min_loop_iterations_, 0)do { static_assert( mozilla::detail::AssertionConditionType< decltype((node_->min_loop_iterations_) > (0))>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!((node_->min_loop_iterations_) > (0)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(node_->min_loop_iterations_) > (0)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1860); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(node_->min_loop_iterations_) > (0)" ")"); do { *((volatile int*)__null) = 1860; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1861 | --node_->min_loop_iterations_; | |||
| 1862 | } | |||
| 1863 | ~IterationDecrementer() { ++node_->min_loop_iterations_; } | |||
| 1864 | IterationDecrementer(const IterationDecrementer&) = delete; | |||
| 1865 | IterationDecrementer& operator=(const IterationDecrementer&) = delete; | |||
| 1866 | ||||
| 1867 | private: | |||
| 1868 | LoopChoiceNode* node_; | |||
| 1869 | }; | |||
| 1870 | ||||
| 1871 | RegExpNode* SeqRegExpNode::FilterOneByte(int depth, RegExpCompiler* compiler) { | |||
| 1872 | if (info()->replacement_calculated) return replacement(); | |||
| 1873 | if (depth < 0) return this; | |||
| 1874 | DCHECK(!info()->visited)do { static_assert( mozilla::detail::AssertionConditionType< decltype(!info()->visited)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(!info()->visited))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("!info()->visited" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1874); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!info()->visited" ")"); do { *((volatile int*)__null) = 1874; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1875 | VisitMarker marker(info()); | |||
| 1876 | return FilterSuccessor(depth - 1, compiler); | |||
| 1877 | } | |||
| 1878 | ||||
| 1879 | RegExpNode* SeqRegExpNode::FilterSuccessor(int depth, | |||
| 1880 | RegExpCompiler* compiler) { | |||
| 1881 | RegExpNode* next = on_success_->FilterOneByte(depth - 1, compiler); | |||
| 1882 | if (next == nullptr) return set_replacement(nullptr); | |||
| 1883 | on_success_ = next; | |||
| 1884 | return set_replacement(this); | |||
| 1885 | } | |||
| 1886 | ||||
| 1887 | // We need to check for the following characters: 0x39C 0x3BC 0x178. | |||
| 1888 | bool RangeContainsLatin1Equivalents(CharacterRange range) { | |||
| 1889 | // TODO(dcarney): this could be a lot more efficient. | |||
| 1890 | return range.Contains(0x039C) || range.Contains(0x03BC) || | |||
| 1891 | range.Contains(0x0178); | |||
| 1892 | } | |||
| 1893 | ||||
| 1894 | namespace { | |||
| 1895 | ||||
| 1896 | bool RangesContainLatin1Equivalents(ZoneList<CharacterRange>* ranges) { | |||
| 1897 | for (int i = 0; i < ranges->length(); i++) { | |||
| 1898 | // TODO(dcarney): this could be a lot more efficient. | |||
| 1899 | if (RangeContainsLatin1Equivalents(ranges->at(i))) return true; | |||
| 1900 | } | |||
| 1901 | return false; | |||
| 1902 | } | |||
| 1903 | ||||
| 1904 | } // namespace | |||
| 1905 | ||||
| 1906 | RegExpNode* TextNode::FilterOneByte(int depth, RegExpCompiler* compiler) { | |||
| 1907 | RegExpFlags flags = compiler->flags(); | |||
| 1908 | if (info()->replacement_calculated) return replacement(); | |||
| 1909 | if (depth < 0) return this; | |||
| 1910 | DCHECK(!info()->visited)do { static_assert( mozilla::detail::AssertionConditionType< decltype(!info()->visited)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(!info()->visited))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("!info()->visited" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1910); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!info()->visited" ")"); do { *((volatile int*)__null) = 1910; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1911 | VisitMarker marker(info()); | |||
| 1912 | int element_count = elements()->length(); | |||
| 1913 | for (int i = 0; i < element_count; i++) { | |||
| 1914 | TextElement elm = elements()->at(i); | |||
| 1915 | if (elm.text_type() == TextElement::ATOM) { | |||
| 1916 | base::Vector<const base::uc16> quarks = elm.atom()->data(); | |||
| 1917 | for (int j = 0; j < quarks.length(); j++) { | |||
| 1918 | base::uc16 c = quarks[j]; | |||
| 1919 | if (!IsIgnoreCase(flags)) { | |||
| 1920 | if (c > String::kMaxOneByteCharCode) return set_replacement(nullptr); | |||
| 1921 | } else { | |||
| 1922 | unibrow::uchar chars[4]; | |||
| 1923 | int length = GetCaseIndependentLetters(compiler->isolate(), c, | |||
| 1924 | compiler, chars, 4); | |||
| 1925 | if (length == 0 || chars[0] > String::kMaxOneByteCharCode) { | |||
| 1926 | return set_replacement(nullptr); | |||
| 1927 | } | |||
| 1928 | } | |||
| 1929 | } | |||
| 1930 | } else { | |||
| 1931 | // A character class can also be impossible to match in one-byte mode. | |||
| 1932 | DCHECK(elm.text_type() == TextElement::CLASS_RANGES)do { static_assert( mozilla::detail::AssertionConditionType< decltype(elm.text_type() == TextElement::CLASS_RANGES)>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!(elm.text_type() == TextElement::CLASS_RANGES))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("elm.text_type() == TextElement::CLASS_RANGES" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 1932); AnnotateMozCrashReason("MOZ_ASSERT" "(" "elm.text_type() == TextElement::CLASS_RANGES" ")"); do { *((volatile int*)__null) = 1932; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 1933 | RegExpClassRanges* cr = elm.class_ranges(); | |||
| 1934 | ZoneList<CharacterRange>* ranges = cr->ranges(zone()); | |||
| 1935 | CharacterRange::Canonicalize(ranges); | |||
| 1936 | // Now they are in order so we only need to look at the first. | |||
| 1937 | // If we are in non-Unicode case independent mode then we need | |||
| 1938 | // to be a bit careful here, because the character classes have | |||
| 1939 | // not been case-desugared yet, but there are characters and ranges | |||
| 1940 | // that can become Latin-1 when case is considered. | |||
| 1941 | int range_count = ranges->length(); | |||
| 1942 | if (cr->is_negated()) { | |||
| 1943 | if (range_count != 0 && ranges->at(0).from() == 0 && | |||
| 1944 | ranges->at(0).to() >= String::kMaxOneByteCharCode) { | |||
| 1945 | bool case_complications = !IsEitherUnicode(flags) && | |||
| 1946 | IsIgnoreCase(flags) && | |||
| 1947 | RangesContainLatin1Equivalents(ranges); | |||
| 1948 | if (!case_complications) { | |||
| 1949 | return set_replacement(nullptr); | |||
| 1950 | } | |||
| 1951 | } | |||
| 1952 | } else { | |||
| 1953 | if (range_count == 0 || | |||
| 1954 | ranges->at(0).from() > String::kMaxOneByteCharCode) { | |||
| 1955 | bool case_complications = !IsEitherUnicode(flags) && | |||
| 1956 | IsIgnoreCase(flags) && | |||
| 1957 | RangesContainLatin1Equivalents(ranges); | |||
| 1958 | if (!case_complications) { | |||
| 1959 | return set_replacement(nullptr); | |||
| 1960 | } | |||
| 1961 | } | |||
| 1962 | } | |||
| 1963 | } | |||
| 1964 | } | |||
| 1965 | return FilterSuccessor(depth - 1, compiler); | |||
| 1966 | } | |||
| 1967 | ||||
| 1968 | RegExpNode* LoopChoiceNode::FilterOneByte(int depth, RegExpCompiler* compiler) { | |||
| 1969 | if (info()->replacement_calculated) return replacement(); | |||
| 1970 | if (depth < 0) return this; | |||
| 1971 | if (info()->visited) return this; | |||
| 1972 | { | |||
| 1973 | VisitMarker marker(info()); | |||
| 1974 | ||||
| 1975 | RegExpNode* continue_replacement = | |||
| 1976 | continue_node_->FilterOneByte(depth - 1, compiler); | |||
| 1977 | // If we can't continue after the loop then there is no sense in doing the | |||
| 1978 | // loop. | |||
| 1979 | if (continue_replacement == nullptr) return set_replacement(nullptr); | |||
| 1980 | } | |||
| 1981 | ||||
| 1982 | return ChoiceNode::FilterOneByte(depth - 1, compiler); | |||
| 1983 | } | |||
| 1984 | ||||
| 1985 | RegExpNode* ChoiceNode::FilterOneByte(int depth, RegExpCompiler* compiler) { | |||
| 1986 | if (info()->replacement_calculated) return replacement(); | |||
| 1987 | if (depth < 0) return this; | |||
| 1988 | if (info()->visited) return this; | |||
| 1989 | VisitMarker marker(info()); | |||
| 1990 | int choice_count = alternatives_->length(); | |||
| 1991 | ||||
| 1992 | for (int i = 0; i < choice_count; i++) { | |||
| 1993 | GuardedAlternative alternative = alternatives_->at(i); | |||
| 1994 | if (alternative.guards() != nullptr && | |||
| 1995 | alternative.guards()->length() != 0) { | |||
| 1996 | set_replacement(this); | |||
| 1997 | return this; | |||
| 1998 | } | |||
| 1999 | } | |||
| 2000 | ||||
| 2001 | int surviving = 0; | |||
| 2002 | RegExpNode* survivor = nullptr; | |||
| 2003 | for (int i = 0; i < choice_count; i++) { | |||
| 2004 | GuardedAlternative alternative = alternatives_->at(i); | |||
| 2005 | RegExpNode* replacement = | |||
| 2006 | alternative.node()->FilterOneByte(depth - 1, compiler); | |||
| 2007 | DCHECK(replacement != this)do { static_assert( mozilla::detail::AssertionConditionType< decltype(replacement != this)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(replacement != this))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("replacement != this" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2007); AnnotateMozCrashReason("MOZ_ASSERT" "(" "replacement != this" ")"); do { *((volatile int*)__null) = 2007; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); // No missing EMPTY_MATCH_CHECK. | |||
| 2008 | if (replacement != nullptr) { | |||
| 2009 | alternatives_->at(i).set_node(replacement); | |||
| 2010 | surviving++; | |||
| 2011 | survivor = replacement; | |||
| 2012 | } | |||
| 2013 | } | |||
| 2014 | if (surviving < 2) return set_replacement(survivor); | |||
| 2015 | ||||
| 2016 | set_replacement(this); | |||
| 2017 | if (surviving == choice_count) { | |||
| 2018 | return this; | |||
| 2019 | } | |||
| 2020 | // Only some of the nodes survived the filtering. We need to rebuild the | |||
| 2021 | // alternatives list. | |||
| 2022 | ZoneList<GuardedAlternative>* new_alternatives = | |||
| 2023 | zone()->New<ZoneList<GuardedAlternative>>(surviving, zone()); | |||
| 2024 | for (int i = 0; i < choice_count; i++) { | |||
| 2025 | RegExpNode* replacement = | |||
| 2026 | alternatives_->at(i).node()->FilterOneByte(depth - 1, compiler); | |||
| 2027 | if (replacement != nullptr) { | |||
| 2028 | alternatives_->at(i).set_node(replacement); | |||
| 2029 | new_alternatives->Add(alternatives_->at(i), zone()); | |||
| 2030 | } | |||
| 2031 | } | |||
| 2032 | alternatives_ = new_alternatives; | |||
| 2033 | return this; | |||
| 2034 | } | |||
| 2035 | ||||
| 2036 | RegExpNode* NegativeLookaroundChoiceNode::FilterOneByte( | |||
| 2037 | int depth, RegExpCompiler* compiler) { | |||
| 2038 | if (info()->replacement_calculated) return replacement(); | |||
| 2039 | if (depth < 0) return this; | |||
| 2040 | if (info()->visited) return this; | |||
| 2041 | VisitMarker marker(info()); | |||
| 2042 | // Alternative 0 is the negative lookahead, alternative 1 is what comes | |||
| 2043 | // afterwards. | |||
| 2044 | RegExpNode* node = continue_node(); | |||
| 2045 | RegExpNode* replacement = node->FilterOneByte(depth - 1, compiler); | |||
| 2046 | if (replacement == nullptr) return set_replacement(nullptr); | |||
| 2047 | alternatives_->at(kContinueIndex).set_node(replacement); | |||
| 2048 | ||||
| 2049 | RegExpNode* neg_node = lookaround_node(); | |||
| 2050 | RegExpNode* neg_replacement = neg_node->FilterOneByte(depth - 1, compiler); | |||
| 2051 | // If the negative lookahead is always going to fail then | |||
| 2052 | // we don't need to check it. | |||
| 2053 | if (neg_replacement == nullptr) return set_replacement(replacement); | |||
| 2054 | alternatives_->at(kLookaroundIndex).set_node(neg_replacement); | |||
| 2055 | return set_replacement(this); | |||
| 2056 | } | |||
| 2057 | ||||
| 2058 | void LoopChoiceNode::GetQuickCheckDetails(QuickCheckDetails* details, | |||
| 2059 | RegExpCompiler* compiler, | |||
| 2060 | int characters_filled_in, | |||
| 2061 | bool not_at_start) { | |||
| 2062 | if (body_can_be_zero_length_ || info()->visited) return; | |||
| 2063 | not_at_start = not_at_start || this->not_at_start(); | |||
| 2064 | DCHECK_EQ(alternatives_->length(), 2)do { static_assert( mozilla::detail::AssertionConditionType< decltype((alternatives_->length()) == (2))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((alternatives_->length()) == (2)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(alternatives_->length()) == (2)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2064); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(alternatives_->length()) == (2)" ")"); do { *((volatile int*)__null) = 2064; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); // There's just loop and continue. | |||
| 2065 | if (traversed_loop_initialization_node_ && min_loop_iterations_ > 0 && | |||
| 2066 | loop_node_->EatsAtLeast(not_at_start) > | |||
| 2067 | continue_node_->EatsAtLeast(true)) { | |||
| 2068 | // Loop body is guaranteed to execute at least once, and consume characters | |||
| 2069 | // when it does, meaning the only possible quick checks from this point | |||
| 2070 | // begin with the loop body. We may recursively visit this LoopChoiceNode, | |||
| 2071 | // but we temporarily decrease its minimum iteration counter so we know when | |||
| 2072 | // to check the continue case. | |||
| 2073 | IterationDecrementer next_iteration(this); | |||
| 2074 | loop_node_->GetQuickCheckDetails(details, compiler, characters_filled_in, | |||
| 2075 | not_at_start); | |||
| 2076 | } else { | |||
| 2077 | // Might not consume anything in the loop body, so treat it like a normal | |||
| 2078 | // ChoiceNode (and don't recursively visit this node again). | |||
| 2079 | VisitMarker marker(info()); | |||
| 2080 | ChoiceNode::GetQuickCheckDetails(details, compiler, characters_filled_in, | |||
| 2081 | not_at_start); | |||
| 2082 | } | |||
| 2083 | } | |||
| 2084 | ||||
| 2085 | void LoopChoiceNode::GetQuickCheckDetailsFromLoopEntry( | |||
| 2086 | QuickCheckDetails* details, RegExpCompiler* compiler, | |||
| 2087 | int characters_filled_in, bool not_at_start) { | |||
| 2088 | if (traversed_loop_initialization_node_) { | |||
| 2089 | // We already entered this loop once, exited via its continuation node, and | |||
| 2090 | // followed an outer loop's back-edge to before the loop entry point. We | |||
| 2091 | // could try to reset the minimum iteration count to its starting value at | |||
| 2092 | // this point, but that seems like more trouble than it's worth. It's safe | |||
| 2093 | // to keep going with the current (possibly reduced) minimum iteration | |||
| 2094 | // count. | |||
| 2095 | GetQuickCheckDetails(details, compiler, characters_filled_in, not_at_start); | |||
| 2096 | } else { | |||
| 2097 | // We are entering a loop via its counter initialization action, meaning we | |||
| 2098 | // are guaranteed to run the loop body at least some minimum number of times | |||
| 2099 | // before running the continuation node. Set a flag so that this node knows | |||
| 2100 | // (now and any times we visit it again recursively) that it was entered | |||
| 2101 | // from the top. | |||
| 2102 | LoopInitializationMarker marker(this); | |||
| 2103 | GetQuickCheckDetails(details, compiler, characters_filled_in, not_at_start); | |||
| 2104 | } | |||
| 2105 | } | |||
| 2106 | ||||
| 2107 | void LoopChoiceNode::FillInBMInfo(Isolate* isolate, int offset, int budget, | |||
| 2108 | BoyerMooreLookahead* bm, bool not_at_start) { | |||
| 2109 | if (body_can_be_zero_length_ || budget <= 0) { | |||
| 2110 | bm->SetRest(offset); | |||
| 2111 | SaveBMInfo(bm, not_at_start, offset); | |||
| 2112 | return; | |||
| 2113 | } | |||
| 2114 | ChoiceNode::FillInBMInfo(isolate, offset, budget - 1, bm, not_at_start); | |||
| 2115 | SaveBMInfo(bm, not_at_start, offset); | |||
| 2116 | } | |||
| 2117 | ||||
| 2118 | void ChoiceNode::GetQuickCheckDetails(QuickCheckDetails* details, | |||
| 2119 | RegExpCompiler* compiler, | |||
| 2120 | int characters_filled_in, | |||
| 2121 | bool not_at_start) { | |||
| 2122 | not_at_start = (not_at_start || not_at_start_); | |||
| 2123 | int choice_count = alternatives_->length(); | |||
| 2124 | DCHECK_LT(0, choice_count)do { static_assert( mozilla::detail::AssertionConditionType< decltype((0) < (choice_count))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((0) < (choice_count)))), 0 ))) { do { } while (false); MOZ_ReportAssertionFailure("(0) < (choice_count)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2124); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(0) < (choice_count)" ")"); do { *((volatile int*)__null) = 2124; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2125 | alternatives_->at(0).node()->GetQuickCheckDetails( | |||
| 2126 | details, compiler, characters_filled_in, not_at_start); | |||
| 2127 | for (int i = 1; i < choice_count; i++) { | |||
| 2128 | QuickCheckDetails new_details(details->characters()); | |||
| 2129 | RegExpNode* node = alternatives_->at(i).node(); | |||
| 2130 | node->GetQuickCheckDetails(&new_details, compiler, characters_filled_in, | |||
| 2131 | not_at_start); | |||
| 2132 | // Here we merge the quick match details of the two branches. | |||
| 2133 | details->Merge(&new_details, characters_filled_in); | |||
| 2134 | } | |||
| 2135 | } | |||
| 2136 | ||||
| 2137 | namespace { | |||
| 2138 | ||||
| 2139 | // Check for [0-9A-Z_a-z]. | |||
| 2140 | void EmitWordCheck(RegExpMacroAssembler* assembler, Label* word, | |||
| 2141 | Label* non_word, bool fall_through_on_word) { | |||
| 2142 | if (assembler->CheckSpecialClassRanges( | |||
| 2143 | fall_through_on_word ? StandardCharacterSet::kWord | |||
| 2144 | : StandardCharacterSet::kNotWord, | |||
| 2145 | fall_through_on_word ? non_word : word)) { | |||
| 2146 | // Optimized implementation available. | |||
| 2147 | return; | |||
| 2148 | } | |||
| 2149 | assembler->CheckCharacterGT('z', non_word); | |||
| 2150 | assembler->CheckCharacterLT('0', non_word); | |||
| 2151 | assembler->CheckCharacterGT('a' - 1, word); | |||
| 2152 | assembler->CheckCharacterLT('9' + 1, word); | |||
| 2153 | assembler->CheckCharacterLT('A', non_word); | |||
| 2154 | assembler->CheckCharacterLT('Z' + 1, word); | |||
| 2155 | if (fall_through_on_word) { | |||
| 2156 | assembler->CheckNotCharacter('_', non_word); | |||
| 2157 | } else { | |||
| 2158 | assembler->CheckCharacter('_', word); | |||
| 2159 | } | |||
| 2160 | } | |||
| 2161 | ||||
| 2162 | // Emit the code to check for a ^ in multiline mode (1-character lookbehind | |||
| 2163 | // that matches newline or the start of input). | |||
| 2164 | void EmitHat(RegExpCompiler* compiler, RegExpNode* on_success, Trace* trace) { | |||
| 2165 | RegExpMacroAssembler* assembler = compiler->macro_assembler(); | |||
| 2166 | ||||
| 2167 | // We will load the previous character into the current character register. | |||
| 2168 | Trace new_trace(*trace); | |||
| 2169 | new_trace.InvalidateCurrentCharacter(); | |||
| 2170 | ||||
| 2171 | // A positive (> 0) cp_offset means we've already successfully matched a | |||
| 2172 | // non-empty-width part of the pattern, and thus cannot be at or before the | |||
| 2173 | // start of the subject string. We can thus skip both at-start and | |||
| 2174 | // bounds-checks when loading the one-character lookbehind. | |||
| 2175 | const bool may_be_at_or_before_subject_string_start = | |||
| 2176 | new_trace.cp_offset() <= 0; | |||
| 2177 | ||||
| 2178 | Label ok; | |||
| 2179 | if (may_be_at_or_before_subject_string_start) { | |||
| 2180 | // The start of input counts as a newline in this context, so skip to ok if | |||
| 2181 | // we are at the start. | |||
| 2182 | assembler->CheckAtStart(new_trace.cp_offset(), &ok); | |||
| 2183 | } | |||
| 2184 | ||||
| 2185 | // If we've already checked that we are not at the start of input, it's okay | |||
| 2186 | // to load the previous character without bounds checks. | |||
| 2187 | const bool can_skip_bounds_check = !may_be_at_or_before_subject_string_start; | |||
| 2188 | assembler->LoadCurrentCharacter(new_trace.cp_offset() - 1, | |||
| 2189 | new_trace.backtrack(), can_skip_bounds_check); | |||
| 2190 | if (!assembler->CheckSpecialClassRanges(StandardCharacterSet::kLineTerminator, | |||
| 2191 | new_trace.backtrack())) { | |||
| 2192 | // Newline means \n, \r, 0x2028 or 0x2029. | |||
| 2193 | if (!compiler->one_byte()) { | |||
| 2194 | assembler->CheckCharacterAfterAnd(0x2028, 0xFFFE, &ok); | |||
| 2195 | } | |||
| 2196 | assembler->CheckCharacter('\n', &ok); | |||
| 2197 | assembler->CheckNotCharacter('\r', new_trace.backtrack()); | |||
| 2198 | } | |||
| 2199 | assembler->Bind(&ok); | |||
| 2200 | on_success->Emit(compiler, &new_trace); | |||
| 2201 | } | |||
| 2202 | ||||
| 2203 | } // namespace | |||
| 2204 | ||||
| 2205 | // Emit the code to handle \b and \B (word-boundary or non-word-boundary). | |||
| 2206 | void AssertionNode::EmitBoundaryCheck(RegExpCompiler* compiler, Trace* trace) { | |||
| 2207 | RegExpMacroAssembler* assembler = compiler->macro_assembler(); | |||
| 2208 | Isolate* isolate = assembler->isolate(); | |||
| 2209 | Trace::TriBool next_is_word_character = Trace::UNKNOWN; | |||
| 2210 | bool not_at_start = (trace->at_start() == Trace::FALSE_VALUE); | |||
| 2211 | BoyerMooreLookahead* lookahead = bm_info(not_at_start); | |||
| 2212 | if (lookahead == nullptr) { | |||
| 2213 | int eats_at_least = | |||
| 2214 | std::min(kMaxLookaheadForBoyerMoore, EatsAtLeast(not_at_start)); | |||
| 2215 | if (eats_at_least >= 1) { | |||
| 2216 | BoyerMooreLookahead* bm = | |||
| 2217 | zone()->New<BoyerMooreLookahead>(eats_at_least, compiler, zone()); | |||
| 2218 | FillInBMInfo(isolate, 0, kRecursionBudget, bm, not_at_start); | |||
| 2219 | if (bm->at(0)->is_non_word()) next_is_word_character = Trace::FALSE_VALUE; | |||
| 2220 | if (bm->at(0)->is_word()) next_is_word_character = Trace::TRUE_VALUE; | |||
| 2221 | } | |||
| 2222 | } else { | |||
| 2223 | if (lookahead->at(0)->is_non_word()) | |||
| 2224 | next_is_word_character = Trace::FALSE_VALUE; | |||
| 2225 | if (lookahead->at(0)->is_word()) next_is_word_character = Trace::TRUE_VALUE; | |||
| 2226 | } | |||
| 2227 | bool at_boundary = (assertion_type_ == AssertionNode::AT_BOUNDARY); | |||
| 2228 | if (next_is_word_character == Trace::UNKNOWN) { | |||
| 2229 | Label before_non_word; | |||
| 2230 | Label before_word; | |||
| 2231 | if (trace->characters_preloaded() != 1) { | |||
| 2232 | assembler->LoadCurrentCharacter(trace->cp_offset(), &before_non_word); | |||
| 2233 | } | |||
| 2234 | // Fall through on non-word. | |||
| 2235 | EmitWordCheck(assembler, &before_word, &before_non_word, false); | |||
| 2236 | // Next character is not a word character. | |||
| 2237 | assembler->Bind(&before_non_word); | |||
| 2238 | Label ok; | |||
| 2239 | BacktrackIfPrevious(compiler, trace, at_boundary ? kIsNonWord : kIsWord); | |||
| 2240 | assembler->GoTo(&ok); | |||
| 2241 | ||||
| 2242 | assembler->Bind(&before_word); | |||
| 2243 | BacktrackIfPrevious(compiler, trace, at_boundary ? kIsWord : kIsNonWord); | |||
| 2244 | assembler->Bind(&ok); | |||
| 2245 | } else if (next_is_word_character == Trace::TRUE_VALUE) { | |||
| 2246 | BacktrackIfPrevious(compiler, trace, at_boundary ? kIsWord : kIsNonWord); | |||
| 2247 | } else { | |||
| 2248 | DCHECK(next_is_word_character == Trace::FALSE_VALUE)do { static_assert( mozilla::detail::AssertionConditionType< decltype(next_is_word_character == Trace::FALSE_VALUE)>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!(next_is_word_character == Trace::FALSE_VALUE))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("next_is_word_character == Trace::FALSE_VALUE" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2248); AnnotateMozCrashReason("MOZ_ASSERT" "(" "next_is_word_character == Trace::FALSE_VALUE" ")"); do { *((volatile int*)__null) = 2248; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2249 | BacktrackIfPrevious(compiler, trace, at_boundary ? kIsNonWord : kIsWord); | |||
| 2250 | } | |||
| 2251 | } | |||
| 2252 | ||||
| 2253 | void AssertionNode::BacktrackIfPrevious( | |||
| 2254 | RegExpCompiler* compiler, Trace* trace, | |||
| 2255 | AssertionNode::IfPrevious backtrack_if_previous) { | |||
| 2256 | RegExpMacroAssembler* assembler = compiler->macro_assembler(); | |||
| 2257 | Trace new_trace(*trace); | |||
| 2258 | new_trace.InvalidateCurrentCharacter(); | |||
| 2259 | ||||
| 2260 | Label fall_through; | |||
| 2261 | Label* non_word = backtrack_if_previous == kIsNonWord ? new_trace.backtrack() | |||
| 2262 | : &fall_through; | |||
| 2263 | Label* word = backtrack_if_previous == kIsNonWord ? &fall_through | |||
| 2264 | : new_trace.backtrack(); | |||
| 2265 | ||||
| 2266 | // A positive (> 0) cp_offset means we've already successfully matched a | |||
| 2267 | // non-empty-width part of the pattern, and thus cannot be at or before the | |||
| 2268 | // start of the subject string. We can thus skip both at-start and | |||
| 2269 | // bounds-checks when loading the one-character lookbehind. | |||
| 2270 | const bool may_be_at_or_before_subject_string_start = | |||
| 2271 | new_trace.cp_offset() <= 0; | |||
| 2272 | ||||
| 2273 | if (may_be_at_or_before_subject_string_start) { | |||
| 2274 | // The start of input counts as a non-word character, so the question is | |||
| 2275 | // decided if we are at the start. | |||
| 2276 | assembler->CheckAtStart(new_trace.cp_offset(), non_word); | |||
| 2277 | } | |||
| 2278 | ||||
| 2279 | // If we've already checked that we are not at the start of input, it's okay | |||
| 2280 | // to load the previous character without bounds checks. | |||
| 2281 | const bool can_skip_bounds_check = !may_be_at_or_before_subject_string_start; | |||
| 2282 | assembler->LoadCurrentCharacter(new_trace.cp_offset() - 1, non_word, | |||
| 2283 | can_skip_bounds_check); | |||
| 2284 | EmitWordCheck(assembler, word, non_word, backtrack_if_previous == kIsNonWord); | |||
| 2285 | ||||
| 2286 | assembler->Bind(&fall_through); | |||
| 2287 | on_success()->Emit(compiler, &new_trace); | |||
| 2288 | } | |||
| 2289 | ||||
| 2290 | void AssertionNode::GetQuickCheckDetails(QuickCheckDetails* details, | |||
| 2291 | RegExpCompiler* compiler, | |||
| 2292 | int filled_in, bool not_at_start) { | |||
| 2293 | if (assertion_type_ == AT_START && not_at_start) { | |||
| 2294 | details->set_cannot_match(); | |||
| 2295 | return; | |||
| 2296 | } | |||
| 2297 | return on_success()->GetQuickCheckDetails(details, compiler, filled_in, | |||
| 2298 | not_at_start); | |||
| 2299 | } | |||
| 2300 | ||||
| 2301 | void AssertionNode::Emit(RegExpCompiler* compiler, Trace* trace) { | |||
| 2302 | RegExpMacroAssembler* assembler = compiler->macro_assembler(); | |||
| 2303 | switch (assertion_type_) { | |||
| 2304 | case AT_END: { | |||
| 2305 | Label ok; | |||
| 2306 | assembler->CheckPosition(trace->cp_offset(), &ok); | |||
| 2307 | assembler->GoTo(trace->backtrack()); | |||
| 2308 | assembler->Bind(&ok); | |||
| 2309 | break; | |||
| 2310 | } | |||
| 2311 | case AT_START: { | |||
| 2312 | if (trace->at_start() == Trace::FALSE_VALUE) { | |||
| 2313 | assembler->GoTo(trace->backtrack()); | |||
| 2314 | return; | |||
| 2315 | } | |||
| 2316 | if (trace->at_start() == Trace::UNKNOWN) { | |||
| 2317 | assembler->CheckNotAtStart(trace->cp_offset(), trace->backtrack()); | |||
| 2318 | Trace at_start_trace = *trace; | |||
| 2319 | at_start_trace.set_at_start(Trace::TRUE_VALUE); | |||
| 2320 | on_success()->Emit(compiler, &at_start_trace); | |||
| 2321 | return; | |||
| 2322 | } | |||
| 2323 | } break; | |||
| 2324 | case AFTER_NEWLINE: | |||
| 2325 | EmitHat(compiler, on_success(), trace); | |||
| 2326 | return; | |||
| 2327 | case AT_BOUNDARY: | |||
| 2328 | case AT_NON_BOUNDARY: { | |||
| 2329 | EmitBoundaryCheck(compiler, trace); | |||
| 2330 | return; | |||
| 2331 | } | |||
| 2332 | } | |||
| 2333 | on_success()->Emit(compiler, trace); | |||
| 2334 | } | |||
| 2335 | ||||
| 2336 | namespace { | |||
| 2337 | ||||
| 2338 | bool DeterminedAlready(QuickCheckDetails* quick_check, int offset) { | |||
| 2339 | if (quick_check == nullptr) return false; | |||
| 2340 | if (offset >= quick_check->characters()) return false; | |||
| 2341 | return quick_check->positions(offset)->determines_perfectly; | |||
| 2342 | } | |||
| 2343 | ||||
| 2344 | void UpdateBoundsCheck(int index, int* checked_up_to) { | |||
| 2345 | if (index > *checked_up_to) { | |||
| 2346 | *checked_up_to = index; | |||
| 2347 | } | |||
| 2348 | } | |||
| 2349 | ||||
| 2350 | } // namespace | |||
| 2351 | ||||
| 2352 | // We call this repeatedly to generate code for each pass over the text node. | |||
| 2353 | // The passes are in increasing order of difficulty because we hope one | |||
| 2354 | // of the first passes will fail in which case we are saved the work of the | |||
| 2355 | // later passes. for example for the case independent regexp /%[asdfghjkl]a/ | |||
| 2356 | // we will check the '%' in the first pass, the case independent 'a' in the | |||
| 2357 | // second pass and the character class in the last pass. | |||
| 2358 | // | |||
| 2359 | // The passes are done from right to left, so for example to test for /bar/ | |||
| 2360 | // we will first test for an 'r' with offset 2, then an 'a' with offset 1 | |||
| 2361 | // and then a 'b' with offset 0. This means we can avoid the end-of-input | |||
| 2362 | // bounds check most of the time. In the example we only need to check for | |||
| 2363 | // end-of-input when loading the putative 'r'. | |||
| 2364 | // | |||
| 2365 | // A slight complication involves the fact that the first character may already | |||
| 2366 | // be fetched into a register by the previous node. In this case we want to | |||
| 2367 | // do the test for that character first. We do this in separate passes. The | |||
| 2368 | // 'preloaded' argument indicates that we are doing such a 'pass'. If such a | |||
| 2369 | // pass has been performed then subsequent passes will have true in | |||
| 2370 | // first_element_checked to indicate that that character does not need to be | |||
| 2371 | // checked again. | |||
| 2372 | // | |||
| 2373 | // In addition to all this we are passed a Trace, which can | |||
| 2374 | // contain an AlternativeGeneration object. In this AlternativeGeneration | |||
| 2375 | // object we can see details of any quick check that was already passed in | |||
| 2376 | // order to get to the code we are now generating. The quick check can involve | |||
| 2377 | // loading characters, which means we do not need to recheck the bounds | |||
| 2378 | // up to the limit the quick check already checked. In addition the quick | |||
| 2379 | // check can have involved a mask and compare operation which may simplify | |||
| 2380 | // or obviate the need for further checks at some character positions. | |||
| 2381 | void TextNode::TextEmitPass(RegExpCompiler* compiler, TextEmitPassType pass, | |||
| 2382 | bool preloaded, Trace* trace, | |||
| 2383 | bool first_element_checked, int* checked_up_to) { | |||
| 2384 | RegExpMacroAssembler* assembler = compiler->macro_assembler(); | |||
| 2385 | Isolate* isolate = assembler->isolate(); | |||
| 2386 | bool one_byte = compiler->one_byte(); | |||
| 2387 | Label* backtrack = trace->backtrack(); | |||
| 2388 | QuickCheckDetails* quick_check = trace->quick_check_performed(); | |||
| 2389 | int element_count = elements()->length(); | |||
| 2390 | int backward_offset = read_backward() ? -Length() : 0; | |||
| 2391 | for (int i = preloaded ? 0 : element_count - 1; i >= 0; i--) { | |||
| 2392 | TextElement elm = elements()->at(i); | |||
| 2393 | int cp_offset = trace->cp_offset() + elm.cp_offset() + backward_offset; | |||
| 2394 | if (elm.text_type() == TextElement::ATOM) { | |||
| 2395 | base::Vector<const base::uc16> quarks = elm.atom()->data(); | |||
| 2396 | for (int j = preloaded ? 0 : quarks.length() - 1; j >= 0; j--) { | |||
| 2397 | if (first_element_checked && i == 0 && j == 0) continue; | |||
| 2398 | if (DeterminedAlready(quick_check, elm.cp_offset() + j)) continue; | |||
| 2399 | base::uc16 quark = quarks[j]; | |||
| 2400 | bool needs_bounds_check = | |||
| 2401 | *checked_up_to < cp_offset + j || read_backward(); | |||
| 2402 | bool bounds_checked = false; | |||
| 2403 | switch (pass) { | |||
| 2404 | case NON_LATIN1_MATCH: { | |||
| 2405 | DCHECK(one_byte)do { static_assert( mozilla::detail::AssertionConditionType< decltype(one_byte)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(one_byte))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("one_byte", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2405); AnnotateMozCrashReason("MOZ_ASSERT" "(" "one_byte" ")" ); do { *((volatile int*)__null) = 2405; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); // This pass is only done in one-byte mode. | |||
| 2406 | if (IsIgnoreCase(compiler->flags())) { | |||
| 2407 | // We are compiling for a one-byte subject, case independent mode. | |||
| 2408 | // We have to check whether any of the case alternatives are in | |||
| 2409 | // the one-byte range. | |||
| 2410 | unibrow::uchar chars[4]; | |||
| 2411 | // Only returns characters that are in the one-byte range. | |||
| 2412 | int length = | |||
| 2413 | GetCaseIndependentLetters(isolate, quark, compiler, chars, 4); | |||
| 2414 | if (length == 0) { | |||
| 2415 | assembler->GoTo(backtrack); | |||
| 2416 | return; | |||
| 2417 | } | |||
| 2418 | } else { | |||
| 2419 | // Case-dependent mode. | |||
| 2420 | if (quark > String::kMaxOneByteCharCode) { | |||
| 2421 | assembler->GoTo(backtrack); | |||
| 2422 | return; | |||
| 2423 | } | |||
| 2424 | } | |||
| 2425 | break; | |||
| 2426 | } | |||
| 2427 | case NON_LETTER_CHARACTER_MATCH: | |||
| 2428 | bounds_checked = | |||
| 2429 | EmitAtomNonLetter(isolate, compiler, quark, backtrack, | |||
| 2430 | cp_offset + j, needs_bounds_check, preloaded); | |||
| 2431 | break; | |||
| 2432 | case SIMPLE_CHARACTER_MATCH: | |||
| 2433 | bounds_checked = EmitSimpleCharacter(isolate, compiler, quark, | |||
| 2434 | backtrack, cp_offset + j, | |||
| 2435 | needs_bounds_check, preloaded); | |||
| 2436 | break; | |||
| 2437 | case CASE_CHARACTER_MATCH: | |||
| 2438 | bounds_checked = | |||
| 2439 | EmitAtomLetter(isolate, compiler, quark, backtrack, | |||
| 2440 | cp_offset + j, needs_bounds_check, preloaded); | |||
| 2441 | break; | |||
| 2442 | default: | |||
| 2443 | break; | |||
| 2444 | } | |||
| 2445 | if (bounds_checked) UpdateBoundsCheck(cp_offset + j, checked_up_to); | |||
| 2446 | } | |||
| 2447 | } else { | |||
| 2448 | DCHECK_EQ(TextElement::CLASS_RANGES, elm.text_type())do { static_assert( mozilla::detail::AssertionConditionType< decltype((TextElement::CLASS_RANGES) == (elm.text_type()))> ::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!((TextElement::CLASS_RANGES) == (elm.text_type())))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(TextElement::CLASS_RANGES) == (elm.text_type())" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2448); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(TextElement::CLASS_RANGES) == (elm.text_type())" ")"); do { *((volatile int*)__null) = 2448; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2449 | if (pass == CHARACTER_CLASS_MATCH) { | |||
| 2450 | if (first_element_checked && i == 0) continue; | |||
| 2451 | if (DeterminedAlready(quick_check, elm.cp_offset())) continue; | |||
| 2452 | RegExpClassRanges* cr = elm.class_ranges(); | |||
| 2453 | bool bounds_check = *checked_up_to < cp_offset || read_backward(); | |||
| 2454 | EmitClassRanges(assembler, cr, one_byte, backtrack, cp_offset, | |||
| 2455 | bounds_check, preloaded, zone()); | |||
| 2456 | UpdateBoundsCheck(cp_offset, checked_up_to); | |||
| 2457 | } | |||
| 2458 | } | |||
| 2459 | } | |||
| 2460 | } | |||
| 2461 | ||||
| 2462 | int TextNode::Length() { | |||
| 2463 | TextElement elm = elements()->last(); | |||
| 2464 | DCHECK_LE(0, elm.cp_offset())do { static_assert( mozilla::detail::AssertionConditionType< decltype((0) <= (elm.cp_offset()))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((0) <= (elm.cp_offset())) )), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(0) <= (elm.cp_offset())" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2464); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(0) <= (elm.cp_offset())" ")"); do { *((volatile int*)__null) = 2464; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2465 | return elm.cp_offset() + elm.length(); | |||
| 2466 | } | |||
| 2467 | ||||
| 2468 | TextNode* TextNode::CreateForCharacterRanges(Zone* zone, | |||
| 2469 | ZoneList<CharacterRange>* ranges, | |||
| 2470 | bool read_backward, | |||
| 2471 | RegExpNode* on_success) { | |||
| 2472 | DCHECK_NOT_NULL(ranges)do { static_assert( mozilla::detail::AssertionConditionType< decltype((ranges) != nullptr)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((ranges) != nullptr))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(ranges) != nullptr" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2472); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(ranges) != nullptr" ")"); do { *((volatile int*)__null) = 2472; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2473 | // TODO(jgruber): There's no fundamental need to create this | |||
| 2474 | // RegExpClassRanges; we could refactor to avoid the allocation. | |||
| 2475 | return zone->New<TextNode>(zone->New<RegExpClassRanges>(zone, ranges), | |||
| 2476 | read_backward, on_success); | |||
| 2477 | } | |||
| 2478 | ||||
| 2479 | TextNode* TextNode::CreateForSurrogatePair( | |||
| 2480 | Zone* zone, CharacterRange lead, ZoneList<CharacterRange>* trail_ranges, | |||
| 2481 | bool read_backward, RegExpNode* on_success) { | |||
| 2482 | ZoneList<TextElement>* elms = zone->New<ZoneList<TextElement>>(2, zone); | |||
| 2483 | if (lead.from() == lead.to()) { | |||
| 2484 | ZoneList<base::uc16> lead_surrogate(1, zone); | |||
| 2485 | lead_surrogate.Add(lead.from(), zone); | |||
| 2486 | RegExpAtom* atom = zone->New<RegExpAtom>(lead_surrogate.ToConstVector()); | |||
| 2487 | elms->Add(TextElement::Atom(atom), zone); | |||
| 2488 | } else { | |||
| 2489 | ZoneList<CharacterRange>* lead_ranges = CharacterRange::List(zone, lead); | |||
| 2490 | elms->Add(TextElement::ClassRanges( | |||
| 2491 | zone->New<RegExpClassRanges>(zone, lead_ranges)), | |||
| 2492 | zone); | |||
| 2493 | } | |||
| 2494 | elms->Add(TextElement::ClassRanges( | |||
| 2495 | zone->New<RegExpClassRanges>(zone, trail_ranges)), | |||
| 2496 | zone); | |||
| 2497 | return zone->New<TextNode>(elms, read_backward, on_success); | |||
| 2498 | } | |||
| 2499 | ||||
| 2500 | TextNode* TextNode::CreateForSurrogatePair( | |||
| 2501 | Zone* zone, ZoneList<CharacterRange>* lead_ranges, CharacterRange trail, | |||
| 2502 | bool read_backward, RegExpNode* on_success) { | |||
| 2503 | ZoneList<CharacterRange>* trail_ranges = CharacterRange::List(zone, trail); | |||
| 2504 | ZoneList<TextElement>* elms = zone->New<ZoneList<TextElement>>(2, zone); | |||
| 2505 | elms->Add( | |||
| 2506 | TextElement::ClassRanges(zone->New<RegExpClassRanges>(zone, lead_ranges)), | |||
| 2507 | zone); | |||
| 2508 | elms->Add(TextElement::ClassRanges( | |||
| 2509 | zone->New<RegExpClassRanges>(zone, trail_ranges)), | |||
| 2510 | zone); | |||
| 2511 | return zone->New<TextNode>(elms, read_backward, on_success); | |||
| 2512 | } | |||
| 2513 | ||||
| 2514 | // This generates the code to match a text node. A text node can contain | |||
| 2515 | // straight character sequences (possibly to be matched in a case-independent | |||
| 2516 | // way) and character classes. For efficiency we do not do this in a single | |||
| 2517 | // pass from left to right. Instead we pass over the text node several times, | |||
| 2518 | // emitting code for some character positions every time. See the comment on | |||
| 2519 | // TextEmitPass for details. | |||
| 2520 | void TextNode::Emit(RegExpCompiler* compiler, Trace* trace) { | |||
| 2521 | LimitResult limit_result = LimitVersions(compiler, trace); | |||
| 2522 | if (limit_result == DONE) return; | |||
| 2523 | DCHECK(limit_result == CONTINUE)do { static_assert( mozilla::detail::AssertionConditionType< decltype(limit_result == CONTINUE)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(limit_result == CONTINUE))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("limit_result == CONTINUE" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2523); AnnotateMozCrashReason("MOZ_ASSERT" "(" "limit_result == CONTINUE" ")"); do { *((volatile int*)__null) = 2523; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2524 | ||||
| 2525 | if (trace->cp_offset() + Length() > RegExpMacroAssembler::kMaxCPOffset) { | |||
| 2526 | compiler->SetRegExpTooBig(); | |||
| 2527 | return; | |||
| 2528 | } | |||
| 2529 | ||||
| 2530 | if (compiler->one_byte()) { | |||
| 2531 | int dummy = 0; | |||
| 2532 | TextEmitPass(compiler, NON_LATIN1_MATCH, false, trace, false, &dummy); | |||
| 2533 | } | |||
| 2534 | ||||
| 2535 | bool first_elt_done = false; | |||
| 2536 | int bound_checked_to = trace->cp_offset() - 1; | |||
| 2537 | bound_checked_to += trace->bound_checked_up_to(); | |||
| 2538 | ||||
| 2539 | // If a character is preloaded into the current character register then | |||
| 2540 | // check that first to save reloading it. | |||
| 2541 | for (int twice = 0; twice < 2; twice++) { | |||
| 2542 | bool is_preloaded_pass = twice == 0; | |||
| 2543 | if (is_preloaded_pass && trace->characters_preloaded() != 1) continue; | |||
| 2544 | if (IsIgnoreCase(compiler->flags())) { | |||
| 2545 | TextEmitPass(compiler, NON_LETTER_CHARACTER_MATCH, is_preloaded_pass, | |||
| 2546 | trace, first_elt_done, &bound_checked_to); | |||
| 2547 | TextEmitPass(compiler, CASE_CHARACTER_MATCH, is_preloaded_pass, trace, | |||
| 2548 | first_elt_done, &bound_checked_to); | |||
| 2549 | } else { | |||
| 2550 | TextEmitPass(compiler, SIMPLE_CHARACTER_MATCH, is_preloaded_pass, trace, | |||
| 2551 | first_elt_done, &bound_checked_to); | |||
| 2552 | } | |||
| 2553 | TextEmitPass(compiler, CHARACTER_CLASS_MATCH, is_preloaded_pass, trace, | |||
| 2554 | first_elt_done, &bound_checked_to); | |||
| 2555 | first_elt_done = true; | |||
| 2556 | } | |||
| 2557 | ||||
| 2558 | Trace successor_trace(*trace); | |||
| 2559 | // If we advance backward, we may end up at the start. | |||
| 2560 | successor_trace.AdvanceCurrentPositionInTrace( | |||
| 2561 | read_backward() ? -Length() : Length(), compiler); | |||
| 2562 | successor_trace.set_at_start(read_backward() ? Trace::UNKNOWN | |||
| 2563 | : Trace::FALSE_VALUE); | |||
| 2564 | RecursionCheck rc(compiler); | |||
| 2565 | on_success()->Emit(compiler, &successor_trace); | |||
| 2566 | } | |||
| 2567 | ||||
| 2568 | void Trace::InvalidateCurrentCharacter() { characters_preloaded_ = 0; } | |||
| 2569 | ||||
| 2570 | void Trace::AdvanceCurrentPositionInTrace(int by, RegExpCompiler* compiler) { | |||
| 2571 | // We don't have an instruction for shifting the current character register | |||
| 2572 | // down or for using a shifted value for anything so lets just forget that | |||
| 2573 | // we preloaded any characters into it. | |||
| 2574 | characters_preloaded_ = 0; | |||
| 2575 | // Adjust the offsets of the quick check performed information. This | |||
| 2576 | // information is used to find out what we already determined about the | |||
| 2577 | // characters by means of mask and compare. | |||
| 2578 | quick_check_performed_.Advance(by, compiler->one_byte()); | |||
| 2579 | cp_offset_ += by; | |||
| 2580 | if (cp_offset_ > RegExpMacroAssembler::kMaxCPOffset) { | |||
| 2581 | compiler->SetRegExpTooBig(); | |||
| 2582 | cp_offset_ = 0; | |||
| 2583 | } | |||
| 2584 | bound_checked_up_to_ = std::max(0, bound_checked_up_to_ - by); | |||
| 2585 | } | |||
| 2586 | ||||
| 2587 | void TextNode::MakeCaseIndependent(Isolate* isolate, bool is_one_byte, | |||
| 2588 | RegExpFlags flags) { | |||
| 2589 | if (!IsIgnoreCase(flags)) return; | |||
| 2590 | #ifdef V8_INTL_SUPPORT1 | |||
| 2591 | // This is done in an earlier step when generating the nodes from the AST | |||
| 2592 | // because we may have to split up into separate nodes. | |||
| 2593 | if (NeedsUnicodeCaseEquivalents(flags)) return; | |||
| 2594 | #endif | |||
| 2595 | ||||
| 2596 | int element_count = elements()->length(); | |||
| 2597 | for (int i = 0; i < element_count; i++) { | |||
| 2598 | TextElement elm = elements()->at(i); | |||
| 2599 | if (elm.text_type() == TextElement::CLASS_RANGES) { | |||
| 2600 | RegExpClassRanges* cr = elm.class_ranges(); | |||
| 2601 | // None of the standard character classes is different in the case | |||
| 2602 | // independent case and it slows us down if we don't know that. | |||
| 2603 | if (cr->is_standard(zone())) continue; | |||
| 2604 | ZoneList<CharacterRange>* ranges = cr->ranges(zone()); | |||
| 2605 | CharacterRange::AddCaseEquivalents(isolate, zone(), ranges, is_one_byte); | |||
| 2606 | } | |||
| 2607 | } | |||
| 2608 | } | |||
| 2609 | ||||
| 2610 | int TextNode::GreedyLoopTextLength() { return Length(); } | |||
| 2611 | ||||
| 2612 | RegExpNode* TextNode::GetSuccessorOfOmnivorousTextNode( | |||
| 2613 | RegExpCompiler* compiler) { | |||
| 2614 | if (read_backward()) return nullptr; | |||
| 2615 | if (elements()->length() != 1) return nullptr; | |||
| 2616 | TextElement elm = elements()->at(0); | |||
| 2617 | if (elm.text_type() != TextElement::CLASS_RANGES) return nullptr; | |||
| 2618 | RegExpClassRanges* node = elm.class_ranges(); | |||
| 2619 | ZoneList<CharacterRange>* ranges = node->ranges(zone()); | |||
| 2620 | CharacterRange::Canonicalize(ranges); | |||
| 2621 | if (node->is_negated()) { | |||
| 2622 | return ranges->length() == 0 ? on_success() : nullptr; | |||
| 2623 | } | |||
| 2624 | if (ranges->length() != 1) return nullptr; | |||
| 2625 | const base::uc32 max_char = MaxCodeUnit(compiler->one_byte()); | |||
| 2626 | return ranges->at(0).IsEverything(max_char) ? on_success() : nullptr; | |||
| 2627 | } | |||
| 2628 | ||||
| 2629 | // Finds the fixed match length of a sequence of nodes that goes from | |||
| 2630 | // this alternative and back to this choice node. If there are variable | |||
| 2631 | // length nodes or other complications in the way then return a sentinel | |||
| 2632 | // value indicating that a greedy loop cannot be constructed. | |||
| 2633 | int ChoiceNode::GreedyLoopTextLengthForAlternative( | |||
| 2634 | GuardedAlternative* alternative) { | |||
| 2635 | int length = 0; | |||
| 2636 | RegExpNode* node = alternative->node(); | |||
| 2637 | // Later we will generate code for all these text nodes using recursion | |||
| 2638 | // so we have to limit the max number. | |||
| 2639 | int recursion_depth = 0; | |||
| 2640 | while (node != this) { | |||
| 2641 | if (recursion_depth++ > RegExpCompiler::kMaxRecursion) { | |||
| 2642 | return kNodeIsTooComplexForGreedyLoops; | |||
| 2643 | } | |||
| 2644 | int node_length = node->GreedyLoopTextLength(); | |||
| 2645 | if (node_length == kNodeIsTooComplexForGreedyLoops) { | |||
| 2646 | return kNodeIsTooComplexForGreedyLoops; | |||
| 2647 | } | |||
| 2648 | length += node_length; | |||
| 2649 | SeqRegExpNode* seq_node = static_cast<SeqRegExpNode*>(node); | |||
| 2650 | node = seq_node->on_success(); | |||
| 2651 | } | |||
| 2652 | if (read_backward()) { | |||
| 2653 | length = -length; | |||
| 2654 | } | |||
| 2655 | // Check that we can jump by the whole text length. If not, return sentinel | |||
| 2656 | // to indicate the we can't construct a greedy loop. | |||
| 2657 | if (length < RegExpMacroAssembler::kMinCPOffset || | |||
| 2658 | length > RegExpMacroAssembler::kMaxCPOffset) { | |||
| 2659 | return kNodeIsTooComplexForGreedyLoops; | |||
| 2660 | } | |||
| 2661 | return length; | |||
| 2662 | } | |||
| 2663 | ||||
| 2664 | void LoopChoiceNode::AddLoopAlternative(GuardedAlternative alt) { | |||
| 2665 | DCHECK_NULL(loop_node_)do { static_assert( mozilla::detail::AssertionConditionType< decltype((loop_node_) == nullptr)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((loop_node_) == nullptr))), 0 ))) { do { } while (false); MOZ_ReportAssertionFailure("(loop_node_) == nullptr" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2665); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(loop_node_) == nullptr" ")"); do { *((volatile int*)__null) = 2665; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2666 | AddAlternative(alt); | |||
| 2667 | loop_node_ = alt.node(); | |||
| 2668 | } | |||
| 2669 | ||||
| 2670 | void LoopChoiceNode::AddContinueAlternative(GuardedAlternative alt) { | |||
| 2671 | DCHECK_NULL(continue_node_)do { static_assert( mozilla::detail::AssertionConditionType< decltype((continue_node_) == nullptr)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((continue_node_) == nullptr) )), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(continue_node_) == nullptr" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2671); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(continue_node_) == nullptr" ")"); do { *((volatile int*)__null) = 2671; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2672 | AddAlternative(alt); | |||
| 2673 | continue_node_ = alt.node(); | |||
| 2674 | } | |||
| 2675 | ||||
| 2676 | void LoopChoiceNode::Emit(RegExpCompiler* compiler, Trace* trace) { | |||
| 2677 | RegExpMacroAssembler* macro_assembler = compiler->macro_assembler(); | |||
| 2678 | if (trace->stop_node() == this) { | |||
| ||||
| 2679 | // Back edge of greedy optimized loop node graph. | |||
| 2680 | int text_length = | |||
| 2681 | GreedyLoopTextLengthForAlternative(&(alternatives_->at(0))); | |||
| 2682 | DCHECK_NE(kNodeIsTooComplexForGreedyLoops, text_length)do { static_assert( mozilla::detail::AssertionConditionType< decltype((kNodeIsTooComplexForGreedyLoops) != (text_length))> ::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!((kNodeIsTooComplexForGreedyLoops) != (text_length))) ), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(kNodeIsTooComplexForGreedyLoops) != (text_length)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2682); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(kNodeIsTooComplexForGreedyLoops) != (text_length)" ")"); do { *((volatile int*)__null) = 2682; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2683 | // Update the counter-based backtracking info on the stack. This is an | |||
| 2684 | // optimization for greedy loops (see below). | |||
| 2685 | DCHECK(trace->cp_offset() == text_length)do { static_assert( mozilla::detail::AssertionConditionType< decltype(trace->cp_offset() == text_length)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(trace->cp_offset() == text_length ))), 0))) { do { } while (false); MOZ_ReportAssertionFailure( "trace->cp_offset() == text_length", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2685); AnnotateMozCrashReason("MOZ_ASSERT" "(" "trace->cp_offset() == text_length" ")"); do { *((volatile int*)__null) = 2685; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2686 | macro_assembler->AdvanceCurrentPosition(text_length); | |||
| 2687 | macro_assembler->GoTo(trace->loop_label()); | |||
| 2688 | return; | |||
| 2689 | } | |||
| 2690 | DCHECK_NULL(trace->stop_node())do { static_assert( mozilla::detail::AssertionConditionType< decltype((trace->stop_node()) == nullptr)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((trace->stop_node()) == nullptr ))), 0))) { do { } while (false); MOZ_ReportAssertionFailure( "(trace->stop_node()) == nullptr", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2690); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(trace->stop_node()) == nullptr" ")"); do { *((volatile int*)__null) = 2690; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2691 | if (!trace->is_trivial()) { | |||
| 2692 | trace->Flush(compiler, this); | |||
| 2693 | return; | |||
| 2694 | } | |||
| 2695 | ChoiceNode::Emit(compiler, trace); | |||
| 2696 | } | |||
| 2697 | ||||
| 2698 | int ChoiceNode::CalculatePreloadCharacters(RegExpCompiler* compiler, | |||
| 2699 | int eats_at_least) { | |||
| 2700 | int preload_characters = std::min(4, eats_at_least); | |||
| 2701 | DCHECK_LE(preload_characters, 4)do { static_assert( mozilla::detail::AssertionConditionType< decltype((preload_characters) <= (4))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((preload_characters) <= ( 4)))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(preload_characters) <= (4)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2701); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(preload_characters) <= (4)" ")"); do { *((volatile int*)__null) = 2701; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2702 | if (compiler->macro_assembler()->CanReadUnaligned()) { | |||
| 2703 | bool one_byte = compiler->one_byte(); | |||
| 2704 | if (one_byte) { | |||
| 2705 | // We can't preload 3 characters because there is no machine instruction | |||
| 2706 | // to do that. We can't just load 4 because we could be reading | |||
| 2707 | // beyond the end of the string, which could cause a memory fault. | |||
| 2708 | if (preload_characters == 3) preload_characters = 2; | |||
| 2709 | } else { | |||
| 2710 | if (preload_characters > 2) preload_characters = 2; | |||
| 2711 | } | |||
| 2712 | } else { | |||
| 2713 | if (preload_characters > 1) preload_characters = 1; | |||
| 2714 | } | |||
| 2715 | return preload_characters; | |||
| 2716 | } | |||
| 2717 | ||||
| 2718 | // This class is used when generating the alternatives in a choice node. It | |||
| 2719 | // records the way the alternative is being code generated. | |||
| 2720 | class AlternativeGeneration : public Malloced { | |||
| 2721 | public: | |||
| 2722 | AlternativeGeneration() | |||
| 2723 | : possible_success(), | |||
| 2724 | expects_preload(false), | |||
| 2725 | after(), | |||
| 2726 | quick_check_details() {} | |||
| 2727 | Label possible_success; | |||
| 2728 | bool expects_preload; | |||
| 2729 | Label after; | |||
| 2730 | QuickCheckDetails quick_check_details; | |||
| 2731 | }; | |||
| 2732 | ||||
| 2733 | // Creates a list of AlternativeGenerations. If the list has a reasonable | |||
| 2734 | // size then it is on the stack, otherwise the excess is on the heap. | |||
| 2735 | class AlternativeGenerationList { | |||
| 2736 | public: | |||
| 2737 | AlternativeGenerationList(int count, Zone* zone) : alt_gens_(count, zone) { | |||
| 2738 | for (int i = 0; i < count && i < kAFew; i++) { | |||
| 2739 | alt_gens_.Add(a_few_alt_gens_ + i, zone); | |||
| 2740 | } | |||
| 2741 | for (int i = kAFew; i < count; i++) { | |||
| 2742 | alt_gens_.Add(new AlternativeGeneration(), zone); | |||
| 2743 | } | |||
| 2744 | } | |||
| 2745 | ~AlternativeGenerationList() { | |||
| 2746 | for (int i = kAFew; i < alt_gens_.length(); i++) { | |||
| 2747 | delete alt_gens_[i]; | |||
| 2748 | alt_gens_[i] = nullptr; | |||
| 2749 | } | |||
| 2750 | } | |||
| 2751 | ||||
| 2752 | AlternativeGeneration* at(int i) { return alt_gens_[i]; } | |||
| 2753 | ||||
| 2754 | private: | |||
| 2755 | static const int kAFew = 10; | |||
| 2756 | ZoneList<AlternativeGeneration*> alt_gens_; | |||
| 2757 | AlternativeGeneration a_few_alt_gens_[kAFew]; | |||
| 2758 | }; | |||
| 2759 | ||||
| 2760 | void BoyerMoorePositionInfo::Set(int character) { | |||
| 2761 | SetInterval(Interval(character, character)); | |||
| 2762 | } | |||
| 2763 | ||||
| 2764 | namespace { | |||
| 2765 | ||||
| 2766 | ContainedInLattice AddRange(ContainedInLattice containment, const int* ranges, | |||
| 2767 | int ranges_length, Interval new_range) { | |||
| 2768 | DCHECK_EQ(1, ranges_length & 1)do { static_assert( mozilla::detail::AssertionConditionType< decltype((1) == (ranges_length & 1))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((1) == (ranges_length & 1 )))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(1) == (ranges_length & 1)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2768); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(1) == (ranges_length & 1)" ")"); do { *((volatile int*)__null) = 2768; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2769 | DCHECK_EQ(String::kMaxCodePoint + 1, ranges[ranges_length - 1])do { static_assert( mozilla::detail::AssertionConditionType< decltype((String::kMaxCodePoint + 1) == (ranges[ranges_length - 1]))>::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!((String::kMaxCodePoint + 1) == (ranges[ranges_length - 1])))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(String::kMaxCodePoint + 1) == (ranges[ranges_length - 1])" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2769); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(String::kMaxCodePoint + 1) == (ranges[ranges_length - 1])" ")"); do { *((volatile int*)__null) = 2769; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2770 | if (containment == kLatticeUnknown) return containment; | |||
| 2771 | bool inside = false; | |||
| 2772 | int last = 0; | |||
| 2773 | for (int i = 0; i < ranges_length; inside = !inside, last = ranges[i], i++) { | |||
| 2774 | // Consider the range from last to ranges[i]. | |||
| 2775 | // We haven't got to the new range yet. | |||
| 2776 | if (ranges[i] <= new_range.from()) continue; | |||
| 2777 | // New range is wholly inside last-ranges[i]. Note that new_range.to() is | |||
| 2778 | // inclusive, but the values in ranges are not. | |||
| 2779 | if (last <= new_range.from() && new_range.to() < ranges[i]) { | |||
| 2780 | return Combine(containment, inside ? kLatticeIn : kLatticeOut); | |||
| 2781 | } | |||
| 2782 | return kLatticeUnknown; | |||
| 2783 | } | |||
| 2784 | return containment; | |||
| 2785 | } | |||
| 2786 | ||||
| 2787 | int BitsetFirstSetBit(BoyerMoorePositionInfo::Bitset bitset) { | |||
| 2788 | static_assert(BoyerMoorePositionInfo::kMapSize == | |||
| 2789 | 2 * kInt64Size * kBitsPerByte); | |||
| 2790 | ||||
| 2791 | // Slight fiddling is needed here, since the bitset is of length 128 while | |||
| 2792 | // CountTrailingZeros requires an integral type and std::bitset can only | |||
| 2793 | // convert to unsigned long long. So we handle the most- and least-significant | |||
| 2794 | // bits separately. | |||
| 2795 | ||||
| 2796 | { | |||
| 2797 | static constexpr BoyerMoorePositionInfo::Bitset mask(~uint64_t{0}); | |||
| 2798 | BoyerMoorePositionInfo::Bitset masked_bitset = bitset & mask; | |||
| 2799 | static_assert(kInt64Size >= sizeof(decltype(masked_bitset.to_ullong()))); | |||
| 2800 | uint64_t lsb = masked_bitset.to_ullong(); | |||
| 2801 | if (lsb != 0) return base::bits::CountTrailingZeros(lsb); | |||
| 2802 | } | |||
| 2803 | ||||
| 2804 | { | |||
| 2805 | BoyerMoorePositionInfo::Bitset masked_bitset = bitset >> 64; | |||
| 2806 | uint64_t msb = masked_bitset.to_ullong(); | |||
| 2807 | if (msb != 0) return 64 + base::bits::CountTrailingZeros(msb); | |||
| 2808 | } | |||
| 2809 | ||||
| 2810 | return -1; | |||
| 2811 | } | |||
| 2812 | ||||
| 2813 | } // namespace | |||
| 2814 | ||||
| 2815 | void BoyerMoorePositionInfo::SetInterval(const Interval& interval) { | |||
| 2816 | w_ = AddRange(w_, kWordRanges, kWordRangeCount, interval); | |||
| 2817 | ||||
| 2818 | if (interval.size() >= kMapSize) { | |||
| 2819 | map_count_ = kMapSize; | |||
| 2820 | map_.set(); | |||
| 2821 | return; | |||
| 2822 | } | |||
| 2823 | ||||
| 2824 | for (int i = interval.from(); i <= interval.to(); i++) { | |||
| 2825 | int mod_character = (i & kMask); | |||
| 2826 | if (!map_[mod_character]) { | |||
| 2827 | map_count_++; | |||
| 2828 | map_.set(mod_character); | |||
| 2829 | } | |||
| 2830 | if (map_count_ == kMapSize) return; | |||
| 2831 | } | |||
| 2832 | } | |||
| 2833 | ||||
| 2834 | void BoyerMoorePositionInfo::SetAll() { | |||
| 2835 | w_ = kLatticeUnknown; | |||
| 2836 | if (map_count_ != kMapSize) { | |||
| 2837 | map_count_ = kMapSize; | |||
| 2838 | map_.set(); | |||
| 2839 | } | |||
| 2840 | } | |||
| 2841 | ||||
| 2842 | BoyerMooreLookahead::BoyerMooreLookahead(int length, RegExpCompiler* compiler, | |||
| 2843 | Zone* zone) | |||
| 2844 | : length_(length), | |||
| 2845 | compiler_(compiler), | |||
| 2846 | max_char_(MaxCodeUnit(compiler->one_byte())) { | |||
| 2847 | bitmaps_ = zone->New<ZoneList<BoyerMoorePositionInfo*>>(length, zone); | |||
| 2848 | for (int i = 0; i < length; i++) { | |||
| 2849 | bitmaps_->Add(zone->New<BoyerMoorePositionInfo>(), zone); | |||
| 2850 | } | |||
| 2851 | } | |||
| 2852 | ||||
| 2853 | // Find the longest range of lookahead that has the fewest number of different | |||
| 2854 | // characters that can occur at a given position. Since we are optimizing two | |||
| 2855 | // different parameters at once this is a tradeoff. | |||
| 2856 | bool BoyerMooreLookahead::FindWorthwhileInterval(int* from, int* to) { | |||
| 2857 | int biggest_points = 0; | |||
| 2858 | // If more than 32 characters out of 128 can occur it is unlikely that we can | |||
| 2859 | // be lucky enough to step forwards much of the time. | |||
| 2860 | const int kMaxMax = 32; | |||
| 2861 | for (int max_number_of_chars = 4; max_number_of_chars < kMaxMax; | |||
| 2862 | max_number_of_chars *= 2) { | |||
| 2863 | biggest_points = | |||
| 2864 | FindBestInterval(max_number_of_chars, biggest_points, from, to); | |||
| 2865 | } | |||
| 2866 | if (biggest_points == 0) return false; | |||
| 2867 | return true; | |||
| 2868 | } | |||
| 2869 | ||||
| 2870 | // Find the highest-points range between 0 and length_ where the character | |||
| 2871 | // information is not too vague. 'Too vague' means that there are more than | |||
| 2872 | // max_number_of_chars that can occur at this position. Calculates the number | |||
| 2873 | // of points as the product of width-of-the-range and | |||
| 2874 | // probability-of-finding-one-of-the-characters, where the probability is | |||
| 2875 | // calculated using the frequency distribution of the sample subject string. | |||
| 2876 | int BoyerMooreLookahead::FindBestInterval(int max_number_of_chars, | |||
| 2877 | int old_biggest_points, int* from, | |||
| 2878 | int* to) { | |||
| 2879 | int biggest_points = old_biggest_points; | |||
| 2880 | static const int kSize = RegExpMacroAssembler::kTableSize; | |||
| 2881 | for (int i = 0; i < length_;) { | |||
| 2882 | while (i < length_ && Count(i) > max_number_of_chars) i++; | |||
| 2883 | if (i == length_) break; | |||
| 2884 | int remembered_from = i; | |||
| 2885 | ||||
| 2886 | BoyerMoorePositionInfo::Bitset union_bitset; | |||
| 2887 | for (; i < length_ && Count(i) <= max_number_of_chars; i++) { | |||
| 2888 | union_bitset |= bitmaps_->at(i)->raw_bitset(); | |||
| 2889 | } | |||
| 2890 | ||||
| 2891 | int frequency = 0; | |||
| 2892 | ||||
| 2893 | // Iterate only over set bits. | |||
| 2894 | int j; | |||
| 2895 | while ((j = BitsetFirstSetBit(union_bitset)) != -1) { | |||
| 2896 | DCHECK(union_bitset[j])do { static_assert( mozilla::detail::AssertionConditionType< decltype(union_bitset[j])>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(union_bitset[j]))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("union_bitset[j]" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2896); AnnotateMozCrashReason("MOZ_ASSERT" "(" "union_bitset[j]" ")"); do { *((volatile int*)__null) = 2896; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); // Sanity check. | |||
| 2897 | // Add 1 to the frequency to give a small per-character boost for | |||
| 2898 | // the cases where our sampling is not good enough and many | |||
| 2899 | // characters have a frequency of zero. This means the frequency | |||
| 2900 | // can theoretically be up to 2*kSize though we treat it mostly as | |||
| 2901 | // a fraction of kSize. | |||
| 2902 | frequency += compiler_->frequency_collator()->Frequency(j) + 1; | |||
| 2903 | union_bitset.reset(j); | |||
| 2904 | } | |||
| 2905 | ||||
| 2906 | // We use the probability of skipping times the distance we are skipping to | |||
| 2907 | // judge the effectiveness of this. Actually we have a cut-off: By | |||
| 2908 | // dividing by 2 we switch off the skipping if the probability of skipping | |||
| 2909 | // is less than 50%. This is because the multibyte mask-and-compare | |||
| 2910 | // skipping in quickcheck is more likely to do well on this case. | |||
| 2911 | bool in_quickcheck_range = | |||
| 2912 | ((i - remembered_from < 4) || | |||
| 2913 | (compiler_->one_byte() ? remembered_from <= 4 : remembered_from <= 2)); | |||
| 2914 | // Called 'probability' but it is only a rough estimate and can actually | |||
| 2915 | // be outside the 0-kSize range. | |||
| 2916 | int probability = (in_quickcheck_range ? kSize / 2 : kSize) - frequency; | |||
| 2917 | int points = (i - remembered_from) * probability; | |||
| 2918 | if (points > biggest_points) { | |||
| 2919 | *from = remembered_from; | |||
| 2920 | *to = i - 1; | |||
| 2921 | biggest_points = points; | |||
| 2922 | } | |||
| 2923 | } | |||
| 2924 | return biggest_points; | |||
| 2925 | } | |||
| 2926 | ||||
| 2927 | // Take all the characters that will not prevent a successful match if they | |||
| 2928 | // occur in the subject string in the range between min_lookahead and | |||
| 2929 | // max_lookahead (inclusive) measured from the current position. If the | |||
| 2930 | // character at max_lookahead offset is not one of these characters, then we | |||
| 2931 | // can safely skip forwards by the number of characters in the range. | |||
| 2932 | int BoyerMooreLookahead::GetSkipTable( | |||
| 2933 | int min_lookahead, int max_lookahead, | |||
| 2934 | DirectHandle<ByteArray> boolean_skip_table) { | |||
| 2935 | const int kSkipArrayEntry = 0; | |||
| 2936 | const int kDontSkipArrayEntry = 1; | |||
| 2937 | ||||
| 2938 | std::memset(boolean_skip_table->begin(), kSkipArrayEntry, | |||
| 2939 | boolean_skip_table->length()); | |||
| 2940 | ||||
| 2941 | for (int i = max_lookahead; i >= min_lookahead; i--) { | |||
| 2942 | BoyerMoorePositionInfo::Bitset bitset = bitmaps_->at(i)->raw_bitset(); | |||
| 2943 | ||||
| 2944 | // Iterate only over set bits. | |||
| 2945 | int j; | |||
| 2946 | while ((j = BitsetFirstSetBit(bitset)) != -1) { | |||
| 2947 | DCHECK(bitset[j])do { static_assert( mozilla::detail::AssertionConditionType< decltype(bitset[j])>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(bitset[j]))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("bitset[j]", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2947); AnnotateMozCrashReason("MOZ_ASSERT" "(" "bitset[j]" ")" ); do { *((volatile int*)__null) = 2947; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); // Sanity check. | |||
| 2948 | boolean_skip_table->set(j, kDontSkipArrayEntry); | |||
| 2949 | bitset.reset(j); | |||
| 2950 | } | |||
| 2951 | } | |||
| 2952 | ||||
| 2953 | const int skip = max_lookahead + 1 - min_lookahead; | |||
| 2954 | return skip; | |||
| 2955 | } | |||
| 2956 | ||||
| 2957 | // See comment above on the implementation of GetSkipTable. | |||
| 2958 | void BoyerMooreLookahead::EmitSkipInstructions(RegExpMacroAssembler* masm) { | |||
| 2959 | const int kSize = RegExpMacroAssembler::kTableSize; | |||
| 2960 | ||||
| 2961 | int min_lookahead = 0; | |||
| 2962 | int max_lookahead = 0; | |||
| 2963 | ||||
| 2964 | if (!FindWorthwhileInterval(&min_lookahead, &max_lookahead)) return; | |||
| 2965 | ||||
| 2966 | // Check if we only have a single non-empty position info, and that info | |||
| 2967 | // contains precisely one character. | |||
| 2968 | bool found_single_character = false; | |||
| 2969 | int single_character = 0; | |||
| 2970 | for (int i = max_lookahead; i >= min_lookahead; i--) { | |||
| 2971 | BoyerMoorePositionInfo* map = bitmaps_->at(i); | |||
| 2972 | if (map->map_count() == 0) continue; | |||
| 2973 | ||||
| 2974 | if (found_single_character || map->map_count() > 1) { | |||
| 2975 | found_single_character = false; | |||
| 2976 | break; | |||
| 2977 | } | |||
| 2978 | ||||
| 2979 | DCHECK(!found_single_character)do { static_assert( mozilla::detail::AssertionConditionType< decltype(!found_single_character)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(!found_single_character))), 0 ))) { do { } while (false); MOZ_ReportAssertionFailure("!found_single_character" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2979); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!found_single_character" ")"); do { *((volatile int*)__null) = 2979; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2980 | DCHECK_EQ(map->map_count(), 1)do { static_assert( mozilla::detail::AssertionConditionType< decltype((map->map_count()) == (1))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((map->map_count()) == (1) ))), 0))) { do { } while (false); MOZ_ReportAssertionFailure( "(map->map_count()) == (1)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2980); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(map->map_count()) == (1)" ")"); do { *((volatile int*)__null) = 2980; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2981 | ||||
| 2982 | found_single_character = true; | |||
| 2983 | single_character = BitsetFirstSetBit(map->raw_bitset()); | |||
| 2984 | ||||
| 2985 | DCHECK_NE(single_character, -1)do { static_assert( mozilla::detail::AssertionConditionType< decltype((single_character) != (-1))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((single_character) != (-1))) ), 0))) { do { } while (false); MOZ_ReportAssertionFailure("(single_character) != (-1)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 2985); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(single_character) != (-1)" ")"); do { *((volatile int*)__null) = 2985; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 2986 | } | |||
| 2987 | ||||
| 2988 | int lookahead_width = max_lookahead + 1 - min_lookahead; | |||
| 2989 | ||||
| 2990 | if (found_single_character && lookahead_width == 1 && max_lookahead < 3) { | |||
| 2991 | // The mask-compare can probably handle this better. | |||
| 2992 | return; | |||
| 2993 | } | |||
| 2994 | ||||
| 2995 | if (found_single_character) { | |||
| 2996 | Label cont, again; | |||
| 2997 | masm->Bind(&again); | |||
| 2998 | masm->LoadCurrentCharacter(max_lookahead, &cont, true); | |||
| 2999 | if (max_char_ > kSize) { | |||
| 3000 | masm->CheckCharacterAfterAnd(single_character, | |||
| 3001 | RegExpMacroAssembler::kTableMask, &cont); | |||
| 3002 | } else { | |||
| 3003 | masm->CheckCharacter(single_character, &cont); | |||
| 3004 | } | |||
| 3005 | masm->AdvanceCurrentPosition(lookahead_width); | |||
| 3006 | masm->GoTo(&again); | |||
| 3007 | masm->Bind(&cont); | |||
| 3008 | return; | |||
| 3009 | } | |||
| 3010 | ||||
| 3011 | Factory* factory = masm->isolate()->factory(); | |||
| 3012 | Handle<ByteArray> boolean_skip_table = | |||
| 3013 | factory->NewByteArray(kSize, AllocationType::kOld); | |||
| 3014 | int skip_distance = | |||
| 3015 | GetSkipTable(min_lookahead, max_lookahead, boolean_skip_table); | |||
| 3016 | DCHECK_NE(0, skip_distance)do { static_assert( mozilla::detail::AssertionConditionType< decltype((0) != (skip_distance))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((0) != (skip_distance)))), 0 ))) { do { } while (false); MOZ_ReportAssertionFailure("(0) != (skip_distance)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3016); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(0) != (skip_distance)" ")"); do { *((volatile int*)__null) = 3016; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3017 | ||||
| 3018 | Label cont, again; | |||
| 3019 | masm->Bind(&again); | |||
| 3020 | masm->LoadCurrentCharacter(max_lookahead, &cont, true); | |||
| 3021 | masm->CheckBitInTable(boolean_skip_table, &cont); | |||
| 3022 | masm->AdvanceCurrentPosition(skip_distance); | |||
| 3023 | masm->GoTo(&again); | |||
| 3024 | masm->Bind(&cont); | |||
| 3025 | } | |||
| 3026 | ||||
| 3027 | /* Code generation for choice nodes. | |||
| 3028 | * | |||
| 3029 | * We generate quick checks that do a mask and compare to eliminate a | |||
| 3030 | * choice. If the quick check succeeds then it jumps to the continuation to | |||
| 3031 | * do slow checks and check subsequent nodes. If it fails (the common case) | |||
| 3032 | * it falls through to the next choice. | |||
| 3033 | * | |||
| 3034 | * Here is the desired flow graph. Nodes directly below each other imply | |||
| 3035 | * fallthrough. Alternatives 1 and 2 have quick checks. Alternative | |||
| 3036 | * 3 doesn't have a quick check so we have to call the slow check. | |||
| 3037 | * Nodes are marked Qn for quick checks and Sn for slow checks. The entire | |||
| 3038 | * regexp continuation is generated directly after the Sn node, up to the | |||
| 3039 | * next GoTo if we decide to reuse some already generated code. Some | |||
| 3040 | * nodes expect preload_characters to be preloaded into the current | |||
| 3041 | * character register. R nodes do this preloading. Vertices are marked | |||
| 3042 | * F for failures and S for success (possible success in the case of quick | |||
| 3043 | * nodes). L, V, < and > are used as arrow heads. | |||
| 3044 | * | |||
| 3045 | * ----------> R | |||
| 3046 | * | | |||
| 3047 | * V | |||
| 3048 | * Q1 -----> S1 | |||
| 3049 | * | S / | |||
| 3050 | * F| / | |||
| 3051 | * | F/ | |||
| 3052 | * | / | |||
| 3053 | * | R | |||
| 3054 | * | / | |||
| 3055 | * V L | |||
| 3056 | * Q2 -----> S2 | |||
| 3057 | * | S / | |||
| 3058 | * F| / | |||
| 3059 | * | F/ | |||
| 3060 | * | / | |||
| 3061 | * | R | |||
| 3062 | * | / | |||
| 3063 | * V L | |||
| 3064 | * S3 | |||
| 3065 | * | | |||
| 3066 | * F| | |||
| 3067 | * | | |||
| 3068 | * R | |||
| 3069 | * | | |||
| 3070 | * backtrack V | |||
| 3071 | * <----------Q4 | |||
| 3072 | * \ F | | |||
| 3073 | * \ |S | |||
| 3074 | * \ F V | |||
| 3075 | * \-----S4 | |||
| 3076 | * | |||
| 3077 | * For greedy loops we push the current position, then generate the code that | |||
| 3078 | * eats the input specially in EmitGreedyLoop. The other choice (the | |||
| 3079 | * continuation) is generated by the normal code in EmitChoices, and steps back | |||
| 3080 | * in the input to the starting position when it fails to match. The loop code | |||
| 3081 | * looks like this (U is the unwind code that steps back in the greedy loop). | |||
| 3082 | * | |||
| 3083 | * _____ | |||
| 3084 | * / \ | |||
| 3085 | * V | | |||
| 3086 | * ----------> S1 | | |||
| 3087 | * /| | | |||
| 3088 | * / |S | | |||
| 3089 | * F/ \_____/ | |||
| 3090 | * / | |||
| 3091 | * |<----- | |||
| 3092 | * | \ | |||
| 3093 | * V |S | |||
| 3094 | * Q2 ---> U----->backtrack | |||
| 3095 | * | F / | |||
| 3096 | * S| / | |||
| 3097 | * V F / | |||
| 3098 | * S2--/ | |||
| 3099 | */ | |||
| 3100 | ||||
| 3101 | GreedyLoopState::GreedyLoopState(bool not_at_start) { | |||
| 3102 | counter_backtrack_trace_.set_backtrack(&label_); | |||
| 3103 | if (not_at_start) counter_backtrack_trace_.set_at_start(Trace::FALSE_VALUE); | |||
| 3104 | } | |||
| 3105 | ||||
| 3106 | void ChoiceNode::AssertGuardsMentionRegisters(Trace* trace) { | |||
| 3107 | #ifdef DEBUG1 | |||
| 3108 | int choice_count = alternatives_->length(); | |||
| 3109 | for (int i = 0; i < choice_count - 1; i++) { | |||
| 3110 | GuardedAlternative alternative = alternatives_->at(i); | |||
| 3111 | ZoneList<Guard*>* guards = alternative.guards(); | |||
| 3112 | int guard_count = (guards == nullptr) ? 0 : guards->length(); | |||
| 3113 | for (int j = 0; j < guard_count; j++) { | |||
| 3114 | DCHECK(!trace->mentions_reg(guards->at(j)->reg()))do { static_assert( mozilla::detail::AssertionConditionType< decltype(!trace->mentions_reg(guards->at(j)->reg())) >::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!(!trace->mentions_reg(guards->at(j)->reg())) )), 0))) { do { } while (false); MOZ_ReportAssertionFailure("!trace->mentions_reg(guards->at(j)->reg())" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3114); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!trace->mentions_reg(guards->at(j)->reg())" ")"); do { *((volatile int*)__null) = 3114; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3115 | } | |||
| 3116 | } | |||
| 3117 | #endif | |||
| 3118 | } | |||
| 3119 | ||||
| 3120 | void ChoiceNode::SetUpPreLoad(RegExpCompiler* compiler, Trace* current_trace, | |||
| 3121 | PreloadState* state) { | |||
| 3122 | if (state->eats_at_least_ == PreloadState::kEatsAtLeastNotYetInitialized) { | |||
| 3123 | // Save some time by looking at most one machine word ahead. | |||
| 3124 | state->eats_at_least_ = | |||
| 3125 | EatsAtLeast(current_trace->at_start() == Trace::FALSE_VALUE); | |||
| 3126 | } | |||
| 3127 | state->preload_characters_ = | |||
| 3128 | CalculatePreloadCharacters(compiler, state->eats_at_least_); | |||
| 3129 | ||||
| 3130 | state->preload_is_current_ = | |||
| 3131 | (current_trace->characters_preloaded() == state->preload_characters_); | |||
| 3132 | state->preload_has_checked_bounds_ = state->preload_is_current_; | |||
| 3133 | } | |||
| 3134 | ||||
| 3135 | void ChoiceNode::Emit(RegExpCompiler* compiler, Trace* trace) { | |||
| 3136 | int choice_count = alternatives_->length(); | |||
| 3137 | ||||
| 3138 | if (choice_count == 1 && alternatives_->at(0).guards() == nullptr) { | |||
| 3139 | alternatives_->at(0).node()->Emit(compiler, trace); | |||
| 3140 | return; | |||
| 3141 | } | |||
| 3142 | ||||
| 3143 | AssertGuardsMentionRegisters(trace); | |||
| 3144 | ||||
| 3145 | LimitResult limit_result = LimitVersions(compiler, trace); | |||
| 3146 | if (limit_result == DONE) return; | |||
| 3147 | DCHECK(limit_result == CONTINUE)do { static_assert( mozilla::detail::AssertionConditionType< decltype(limit_result == CONTINUE)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(limit_result == CONTINUE))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("limit_result == CONTINUE" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3147); AnnotateMozCrashReason("MOZ_ASSERT" "(" "limit_result == CONTINUE" ")"); do { *((volatile int*)__null) = 3147; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3148 | ||||
| 3149 | // For loop nodes we already flushed (see LoopChoiceNode::Emit), but for | |||
| 3150 | // other choice nodes we only flush if we are out of code size budget. | |||
| 3151 | if (trace->flush_budget() == 0 && trace->actions() != nullptr) { | |||
| 3152 | trace->Flush(compiler, this); | |||
| 3153 | return; | |||
| 3154 | } | |||
| 3155 | ||||
| 3156 | RecursionCheck rc(compiler); | |||
| 3157 | ||||
| 3158 | PreloadState preload; | |||
| 3159 | preload.init(); | |||
| 3160 | GreedyLoopState greedy_loop_state(not_at_start()); | |||
| 3161 | ||||
| 3162 | int text_length = GreedyLoopTextLengthForAlternative(&alternatives_->at(0)); | |||
| 3163 | AlternativeGenerationList alt_gens(choice_count, zone()); | |||
| 3164 | ||||
| 3165 | if (choice_count
| |||
| 3166 | trace = EmitGreedyLoop(compiler, trace, &alt_gens, &preload, | |||
| 3167 | &greedy_loop_state, text_length); | |||
| 3168 | } else { | |||
| 3169 | preload.eats_at_least_ = EmitOptimizedUnanchoredSearch(compiler, trace); | |||
| 3170 | ||||
| 3171 | EmitChoices(compiler, &alt_gens, 0, trace, &preload); | |||
| 3172 | } | |||
| 3173 | ||||
| 3174 | // At this point we need to generate slow checks for the alternatives where | |||
| 3175 | // the quick check was inlined. We can recognize these because the associated | |||
| 3176 | // label was bound. | |||
| 3177 | int new_flush_budget = trace->flush_budget() / choice_count; | |||
| 3178 | for (int i = 0; i < choice_count; i++) { | |||
| 3179 | AlternativeGeneration* alt_gen = alt_gens.at(i); | |||
| 3180 | Trace new_trace(*trace); | |||
| 3181 | // If there are actions to be flushed we have to limit how many times | |||
| 3182 | // they are flushed. Take the budget of the parent trace and distribute | |||
| 3183 | // it fairly amongst the children. | |||
| 3184 | if (new_trace.actions() != nullptr) { | |||
| 3185 | new_trace.set_flush_budget(new_flush_budget); | |||
| 3186 | } | |||
| 3187 | bool next_expects_preload = | |||
| 3188 | i == choice_count - 1 ? false : alt_gens.at(i + 1)->expects_preload; | |||
| 3189 | EmitOutOfLineContinuation(compiler, &new_trace, alternatives_->at(i), | |||
| 3190 | alt_gen, preload.preload_characters_, | |||
| 3191 | next_expects_preload); | |||
| 3192 | } | |||
| 3193 | } | |||
| 3194 | ||||
| 3195 | Trace* ChoiceNode::EmitGreedyLoop(RegExpCompiler* compiler, Trace* trace, | |||
| 3196 | AlternativeGenerationList* alt_gens, | |||
| 3197 | PreloadState* preload, | |||
| 3198 | GreedyLoopState* greedy_loop_state, | |||
| 3199 | int text_length) { | |||
| 3200 | RegExpMacroAssembler* macro_assembler = compiler->macro_assembler(); | |||
| 3201 | // Here we have special handling for greedy loops containing only text nodes | |||
| 3202 | // and other simple nodes. These are handled by pushing the current | |||
| 3203 | // position on the stack and then incrementing the current position each | |||
| 3204 | // time around the switch. On backtrack we decrement the current position | |||
| 3205 | // and check it against the pushed value. This avoids pushing backtrack | |||
| 3206 | // information for each iteration of the loop, which could take up a lot of | |||
| 3207 | // space. | |||
| 3208 | DCHECK(trace->stop_node() == nullptr)do { static_assert( mozilla::detail::AssertionConditionType< decltype(trace->stop_node() == nullptr)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(trace->stop_node() == nullptr ))), 0))) { do { } while (false); MOZ_ReportAssertionFailure( "trace->stop_node() == nullptr", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3208); AnnotateMozCrashReason("MOZ_ASSERT" "(" "trace->stop_node() == nullptr" ")"); do { *((volatile int*)__null) = 3208; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3209 | macro_assembler->PushCurrentPosition(); | |||
| 3210 | Label greedy_match_failed; | |||
| 3211 | Trace greedy_match_trace; | |||
| 3212 | if (not_at_start()) greedy_match_trace.set_at_start(Trace::FALSE_VALUE); | |||
| 3213 | greedy_match_trace.set_backtrack(&greedy_match_failed); | |||
| 3214 | Label loop_label; | |||
| 3215 | macro_assembler->Bind(&loop_label); | |||
| 3216 | greedy_match_trace.set_stop_node(this); | |||
| 3217 | greedy_match_trace.set_loop_label(&loop_label); | |||
| 3218 | alternatives_->at(0).node()->Emit(compiler, &greedy_match_trace); | |||
| 3219 | macro_assembler->Bind(&greedy_match_failed); | |||
| 3220 | ||||
| 3221 | Label second_choice; // For use in greedy matches. | |||
| 3222 | macro_assembler->Bind(&second_choice); | |||
| 3223 | ||||
| 3224 | Trace* new_trace = greedy_loop_state->counter_backtrack_trace(); | |||
| 3225 | ||||
| 3226 | EmitChoices(compiler, alt_gens, 1, new_trace, preload); | |||
| 3227 | ||||
| 3228 | macro_assembler->Bind(greedy_loop_state->label()); | |||
| 3229 | // If we have unwound to the bottom then backtrack. | |||
| 3230 | macro_assembler->CheckGreedyLoop(trace->backtrack()); | |||
| 3231 | // Otherwise try the second priority at an earlier position. | |||
| 3232 | macro_assembler->AdvanceCurrentPosition(-text_length); | |||
| 3233 | macro_assembler->GoTo(&second_choice); | |||
| 3234 | return new_trace; | |||
| 3235 | } | |||
| 3236 | ||||
| 3237 | int ChoiceNode::EmitOptimizedUnanchoredSearch(RegExpCompiler* compiler, | |||
| 3238 | Trace* trace) { | |||
| 3239 | int eats_at_least = PreloadState::kEatsAtLeastNotYetInitialized; | |||
| 3240 | if (alternatives_->length() != 2) return eats_at_least; | |||
| 3241 | ||||
| 3242 | GuardedAlternative alt1 = alternatives_->at(1); | |||
| 3243 | if (alt1.guards() != nullptr && alt1.guards()->length() != 0) { | |||
| 3244 | return eats_at_least; | |||
| 3245 | } | |||
| 3246 | RegExpNode* eats_anything_node = alt1.node(); | |||
| 3247 | if (eats_anything_node->GetSuccessorOfOmnivorousTextNode(compiler) != this) { | |||
| 3248 | return eats_at_least; | |||
| 3249 | } | |||
| 3250 | ||||
| 3251 | // Really we should be creating a new trace when we execute this function, | |||
| 3252 | // but there is no need, because the code it generates cannot backtrack, and | |||
| 3253 | // we always arrive here with a trivial trace (since it's the entry to a | |||
| 3254 | // loop. That also implies that there are no preloaded characters, which is | |||
| 3255 | // good, because it means we won't be violating any assumptions by | |||
| 3256 | // overwriting those characters with new load instructions. | |||
| 3257 | DCHECK(trace->is_trivial())do { static_assert( mozilla::detail::AssertionConditionType< decltype(trace->is_trivial())>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(trace->is_trivial()))), 0 ))) { do { } while (false); MOZ_ReportAssertionFailure("trace->is_trivial()" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3257); AnnotateMozCrashReason("MOZ_ASSERT" "(" "trace->is_trivial()" ")"); do { *((volatile int*)__null) = 3257; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3258 | ||||
| 3259 | RegExpMacroAssembler* macro_assembler = compiler->macro_assembler(); | |||
| 3260 | Isolate* isolate = macro_assembler->isolate(); | |||
| 3261 | // At this point we know that we are at a non-greedy loop that will eat | |||
| 3262 | // any character one at a time. Any non-anchored regexp has such a | |||
| 3263 | // loop prepended to it in order to find where it starts. We look for | |||
| 3264 | // a pattern of the form ...abc... where we can look 6 characters ahead | |||
| 3265 | // and step forwards 3 if the character is not one of abc. Abc need | |||
| 3266 | // not be atoms, they can be any reasonably limited character class or | |||
| 3267 | // small alternation. | |||
| 3268 | BoyerMooreLookahead* bm = bm_info(false); | |||
| 3269 | if (bm == nullptr) { | |||
| 3270 | eats_at_least = std::min(kMaxLookaheadForBoyerMoore, EatsAtLeast(false)); | |||
| 3271 | if (eats_at_least >= 1) { | |||
| 3272 | bm = zone()->New<BoyerMooreLookahead>(eats_at_least, compiler, zone()); | |||
| 3273 | GuardedAlternative alt0 = alternatives_->at(0); | |||
| 3274 | alt0.node()->FillInBMInfo(isolate, 0, kRecursionBudget, bm, false); | |||
| 3275 | } | |||
| 3276 | } | |||
| 3277 | if (bm != nullptr) { | |||
| 3278 | bm->EmitSkipInstructions(macro_assembler); | |||
| 3279 | } | |||
| 3280 | return eats_at_least; | |||
| 3281 | } | |||
| 3282 | ||||
| 3283 | void ChoiceNode::EmitChoices(RegExpCompiler* compiler, | |||
| 3284 | AlternativeGenerationList* alt_gens, | |||
| 3285 | int first_choice, Trace* trace, | |||
| 3286 | PreloadState* preload) { | |||
| 3287 | RegExpMacroAssembler* macro_assembler = compiler->macro_assembler(); | |||
| 3288 | SetUpPreLoad(compiler, trace, preload); | |||
| 3289 | ||||
| 3290 | // For now we just call all choices one after the other. The idea ultimately | |||
| 3291 | // is to use the Dispatch table to try only the relevant ones. | |||
| 3292 | int choice_count = alternatives_->length(); | |||
| 3293 | ||||
| 3294 | int new_flush_budget = trace->flush_budget() / choice_count; | |||
| 3295 | ||||
| 3296 | for (int i = first_choice; i < choice_count; i++) { | |||
| 3297 | bool is_last = i == choice_count - 1; | |||
| 3298 | bool fall_through_on_failure = !is_last; | |||
| 3299 | GuardedAlternative alternative = alternatives_->at(i); | |||
| 3300 | AlternativeGeneration* alt_gen = alt_gens->at(i); | |||
| 3301 | alt_gen->quick_check_details.set_characters(preload->preload_characters_); | |||
| 3302 | ZoneList<Guard*>* guards = alternative.guards(); | |||
| 3303 | int guard_count = (guards == nullptr) ? 0 : guards->length(); | |||
| 3304 | Trace new_trace(*trace); | |||
| 3305 | new_trace.set_characters_preloaded( | |||
| 3306 | preload->preload_is_current_
| |||
| 3307 | if (preload->preload_has_checked_bounds_
| |||
| 3308 | new_trace.set_bound_checked_up_to(preload->preload_characters_); | |||
| 3309 | } | |||
| 3310 | new_trace.quick_check_performed()->Clear(); | |||
| 3311 | if (not_at_start_) new_trace.set_at_start(Trace::FALSE_VALUE); | |||
| 3312 | if (!is_last
| |||
| 3313 | new_trace.set_backtrack(&alt_gen->after); | |||
| 3314 | } | |||
| 3315 | alt_gen->expects_preload = preload->preload_is_current_; | |||
| 3316 | bool generate_full_check_inline = false; | |||
| 3317 | if (compiler->optimize() && | |||
| 3318 | try_to_emit_quick_check_for_alternative(i == 0) && | |||
| 3319 | alternative.node()->EmitQuickCheck( | |||
| 3320 | compiler, trace, &new_trace, preload->preload_has_checked_bounds_, | |||
| 3321 | &alt_gen->possible_success, &alt_gen->quick_check_details, | |||
| 3322 | fall_through_on_failure, this)) { | |||
| 3323 | // Quick check was generated for this choice. | |||
| 3324 | preload->preload_is_current_ = true; | |||
| 3325 | preload->preload_has_checked_bounds_ = true; | |||
| 3326 | // If we generated the quick check to fall through on possible success, | |||
| 3327 | // we now need to generate the full check inline. | |||
| 3328 | if (!fall_through_on_failure) { | |||
| 3329 | macro_assembler->Bind(&alt_gen->possible_success); | |||
| 3330 | new_trace.set_quick_check_performed(&alt_gen->quick_check_details); | |||
| 3331 | new_trace.set_characters_preloaded(preload->preload_characters_); | |||
| 3332 | new_trace.set_bound_checked_up_to(preload->preload_characters_); | |||
| 3333 | generate_full_check_inline = true; | |||
| 3334 | } | |||
| 3335 | } else if (alt_gen->quick_check_details.cannot_match()) { | |||
| 3336 | if (!fall_through_on_failure) { | |||
| 3337 | macro_assembler->GoTo(trace->backtrack()); | |||
| 3338 | } | |||
| 3339 | continue; | |||
| 3340 | } else { | |||
| 3341 | // No quick check was generated. Put the full code here. | |||
| 3342 | // If this is not the first choice then there could be slow checks from | |||
| 3343 | // previous cases that go here when they fail. There's no reason to | |||
| 3344 | // insist that they preload characters since the slow check we are about | |||
| 3345 | // to generate probably can't use it. | |||
| 3346 | if (i != first_choice) { | |||
| 3347 | alt_gen->expects_preload = false; | |||
| 3348 | new_trace.InvalidateCurrentCharacter(); | |||
| 3349 | } | |||
| 3350 | generate_full_check_inline = true; | |||
| 3351 | } | |||
| 3352 | if (generate_full_check_inline) { | |||
| 3353 | if (new_trace.actions() != nullptr) { | |||
| 3354 | new_trace.set_flush_budget(new_flush_budget); | |||
| 3355 | } | |||
| 3356 | for (int j = 0; j < guard_count; j++) { | |||
| 3357 | GenerateGuard(macro_assembler, guards->at(j), &new_trace); | |||
| 3358 | } | |||
| 3359 | alternative.node()->Emit(compiler, &new_trace); | |||
| 3360 | preload->preload_is_current_ = false; | |||
| 3361 | } | |||
| 3362 | macro_assembler->Bind(&alt_gen->after); | |||
| 3363 | } | |||
| 3364 | } | |||
| 3365 | ||||
| 3366 | void ChoiceNode::EmitOutOfLineContinuation(RegExpCompiler* compiler, | |||
| 3367 | Trace* trace, | |||
| 3368 | GuardedAlternative alternative, | |||
| 3369 | AlternativeGeneration* alt_gen, | |||
| 3370 | int preload_characters, | |||
| 3371 | bool next_expects_preload) { | |||
| 3372 | if (!alt_gen->possible_success.is_linked()) return; | |||
| 3373 | ||||
| 3374 | RegExpMacroAssembler* macro_assembler = compiler->macro_assembler(); | |||
| 3375 | macro_assembler->Bind(&alt_gen->possible_success); | |||
| 3376 | Trace out_of_line_trace(*trace); | |||
| 3377 | out_of_line_trace.set_characters_preloaded(preload_characters); | |||
| 3378 | out_of_line_trace.set_quick_check_performed(&alt_gen->quick_check_details); | |||
| 3379 | if (not_at_start_) out_of_line_trace.set_at_start(Trace::FALSE_VALUE); | |||
| 3380 | ZoneList<Guard*>* guards = alternative.guards(); | |||
| 3381 | int guard_count = (guards == nullptr) ? 0 : guards->length(); | |||
| 3382 | if (next_expects_preload) { | |||
| 3383 | Label reload_current_char; | |||
| 3384 | out_of_line_trace.set_backtrack(&reload_current_char); | |||
| 3385 | for (int j = 0; j < guard_count; j++) { | |||
| 3386 | GenerateGuard(macro_assembler, guards->at(j), &out_of_line_trace); | |||
| 3387 | } | |||
| 3388 | alternative.node()->Emit(compiler, &out_of_line_trace); | |||
| 3389 | macro_assembler->Bind(&reload_current_char); | |||
| 3390 | // Reload the current character, since the next quick check expects that. | |||
| 3391 | // We don't need to check bounds here because we only get into this | |||
| 3392 | // code through a quick check which already did the checked load. | |||
| 3393 | macro_assembler->LoadCurrentCharacter(trace->cp_offset(), nullptr, false, | |||
| 3394 | preload_characters); | |||
| 3395 | macro_assembler->GoTo(&(alt_gen->after)); | |||
| 3396 | } else { | |||
| 3397 | out_of_line_trace.set_backtrack(&(alt_gen->after)); | |||
| 3398 | for (int j = 0; j < guard_count; j++) { | |||
| 3399 | GenerateGuard(macro_assembler, guards->at(j), &out_of_line_trace); | |||
| 3400 | } | |||
| 3401 | alternative.node()->Emit(compiler, &out_of_line_trace); | |||
| 3402 | } | |||
| 3403 | } | |||
| 3404 | ||||
| 3405 | void ActionNode::Emit(RegExpCompiler* compiler, Trace* trace) { | |||
| 3406 | RegExpMacroAssembler* assembler = compiler->macro_assembler(); | |||
| 3407 | LimitResult limit_result = LimitVersions(compiler, trace); | |||
| 3408 | if (limit_result == DONE) return; | |||
| 3409 | DCHECK(limit_result == CONTINUE)do { static_assert( mozilla::detail::AssertionConditionType< decltype(limit_result == CONTINUE)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(limit_result == CONTINUE))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("limit_result == CONTINUE" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3409); AnnotateMozCrashReason("MOZ_ASSERT" "(" "limit_result == CONTINUE" ")"); do { *((volatile int*)__null) = 3409; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3410 | ||||
| 3411 | RecursionCheck rc(compiler); | |||
| 3412 | ||||
| 3413 | switch (action_type_) { | |||
| 3414 | case STORE_POSITION: { | |||
| 3415 | Trace::DeferredCapture new_capture(data_.u_position_register.reg, | |||
| 3416 | data_.u_position_register.is_capture, | |||
| 3417 | trace); | |||
| 3418 | Trace new_trace = *trace; | |||
| 3419 | new_trace.add_action(&new_capture); | |||
| 3420 | on_success()->Emit(compiler, &new_trace); | |||
| 3421 | break; | |||
| 3422 | } | |||
| 3423 | case INCREMENT_REGISTER: { | |||
| 3424 | Trace::DeferredIncrementRegister new_increment( | |||
| 3425 | data_.u_increment_register.reg); | |||
| 3426 | Trace new_trace = *trace; | |||
| 3427 | new_trace.add_action(&new_increment); | |||
| 3428 | on_success()->Emit(compiler, &new_trace); | |||
| 3429 | break; | |||
| 3430 | } | |||
| 3431 | case SET_REGISTER_FOR_LOOP: { | |||
| 3432 | Trace::DeferredSetRegisterForLoop new_set(data_.u_store_register.reg, | |||
| 3433 | data_.u_store_register.value); | |||
| 3434 | Trace new_trace = *trace; | |||
| 3435 | new_trace.add_action(&new_set); | |||
| 3436 | on_success()->Emit(compiler, &new_trace); | |||
| 3437 | break; | |||
| 3438 | } | |||
| 3439 | case CLEAR_CAPTURES: { | |||
| 3440 | Trace::DeferredClearCaptures new_capture(Interval( | |||
| 3441 | data_.u_clear_captures.range_from, data_.u_clear_captures.range_to)); | |||
| 3442 | Trace new_trace = *trace; | |||
| 3443 | new_trace.add_action(&new_capture); | |||
| 3444 | on_success()->Emit(compiler, &new_trace); | |||
| 3445 | break; | |||
| 3446 | } | |||
| 3447 | case BEGIN_POSITIVE_SUBMATCH: | |||
| 3448 | case BEGIN_NEGATIVE_SUBMATCH: | |||
| 3449 | if (!trace->is_trivial()) { | |||
| 3450 | trace->Flush(compiler, this); | |||
| 3451 | } else { | |||
| 3452 | assembler->WriteCurrentPositionToRegister( | |||
| 3453 | data_.u_submatch.current_position_register, 0); | |||
| 3454 | assembler->WriteStackPointerToRegister( | |||
| 3455 | data_.u_submatch.stack_pointer_register); | |||
| 3456 | on_success()->Emit(compiler, trace); | |||
| 3457 | } | |||
| 3458 | break; | |||
| 3459 | case EMPTY_MATCH_CHECK: { | |||
| 3460 | int start_pos_reg = data_.u_empty_match_check.start_register; | |||
| 3461 | int stored_pos = 0; | |||
| 3462 | int rep_reg = data_.u_empty_match_check.repetition_register; | |||
| 3463 | bool has_minimum = (rep_reg != RegExpCompiler::kNoRegister); | |||
| 3464 | bool know_dist = trace->GetStoredPosition(start_pos_reg, &stored_pos); | |||
| 3465 | if (know_dist && !has_minimum && stored_pos == trace->cp_offset()) { | |||
| 3466 | // If we know we haven't advanced and there is no minimum we | |||
| 3467 | // can just backtrack immediately. | |||
| 3468 | assembler->GoTo(trace->backtrack()); | |||
| 3469 | } else if (know_dist && stored_pos < trace->cp_offset()) { | |||
| 3470 | // If we know we've advanced we can generate the continuation | |||
| 3471 | // immediately. | |||
| 3472 | on_success()->Emit(compiler, trace); | |||
| 3473 | } else if (!trace->is_trivial()) { | |||
| 3474 | trace->Flush(compiler, this); | |||
| 3475 | } else { | |||
| 3476 | Label skip_empty_check; | |||
| 3477 | // If we have a minimum number of repetitions we check the current | |||
| 3478 | // number first and skip the empty check if it's not enough. | |||
| 3479 | if (has_minimum) { | |||
| 3480 | int limit = data_.u_empty_match_check.repetition_limit; | |||
| 3481 | assembler->IfRegisterLT(rep_reg, limit, &skip_empty_check); | |||
| 3482 | } | |||
| 3483 | // If the match is empty we bail out, otherwise we fall through | |||
| 3484 | // to the on-success continuation. | |||
| 3485 | assembler->IfRegisterEqPos(data_.u_empty_match_check.start_register, | |||
| 3486 | trace->backtrack()); | |||
| 3487 | assembler->Bind(&skip_empty_check); | |||
| 3488 | on_success()->Emit(compiler, trace); | |||
| 3489 | } | |||
| 3490 | break; | |||
| 3491 | } | |||
| 3492 | case POSITIVE_SUBMATCH_SUCCESS: { | |||
| 3493 | if (!trace->is_trivial()) { | |||
| 3494 | trace->Flush(compiler, this); | |||
| 3495 | return; | |||
| 3496 | } | |||
| 3497 | assembler->ReadCurrentPositionFromRegister( | |||
| 3498 | data_.u_submatch.current_position_register); | |||
| 3499 | assembler->ReadStackPointerFromRegister( | |||
| 3500 | data_.u_submatch.stack_pointer_register); | |||
| 3501 | int clear_register_count = data_.u_submatch.clear_register_count; | |||
| 3502 | if (clear_register_count == 0) { | |||
| 3503 | on_success()->Emit(compiler, trace); | |||
| 3504 | return; | |||
| 3505 | } | |||
| 3506 | int clear_registers_from = data_.u_submatch.clear_register_from; | |||
| 3507 | Label clear_registers_backtrack; | |||
| 3508 | Trace new_trace = *trace; | |||
| 3509 | new_trace.set_backtrack(&clear_registers_backtrack); | |||
| 3510 | on_success()->Emit(compiler, &new_trace); | |||
| 3511 | ||||
| 3512 | assembler->Bind(&clear_registers_backtrack); | |||
| 3513 | int clear_registers_to = clear_registers_from + clear_register_count - 1; | |||
| 3514 | assembler->ClearRegisters(clear_registers_from, clear_registers_to); | |||
| 3515 | ||||
| 3516 | DCHECK(trace->backtrack() == nullptr)do { static_assert( mozilla::detail::AssertionConditionType< decltype(trace->backtrack() == nullptr)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(trace->backtrack() == nullptr ))), 0))) { do { } while (false); MOZ_ReportAssertionFailure( "trace->backtrack() == nullptr", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3516); AnnotateMozCrashReason("MOZ_ASSERT" "(" "trace->backtrack() == nullptr" ")"); do { *((volatile int*)__null) = 3516; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3517 | assembler->Backtrack(); | |||
| 3518 | return; | |||
| 3519 | } | |||
| 3520 | case MODIFY_FLAGS: { | |||
| 3521 | compiler->set_flags(flags()); | |||
| 3522 | on_success()->Emit(compiler, trace); | |||
| 3523 | break; | |||
| 3524 | } | |||
| 3525 | default: | |||
| 3526 | UNREACHABLE()do { do { } while (false); MOZ_ReportCrash("" "unreachable code" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3526); AnnotateMozCrashReason("MOZ_CRASH(" "unreachable code" ")"); do { *((volatile int*)__null) = 3526; __attribute__((nomerge )) ::abort(); } while (false); } while (false); | |||
| 3527 | } | |||
| 3528 | } | |||
| 3529 | ||||
| 3530 | void BackReferenceNode::Emit(RegExpCompiler* compiler, Trace* trace) { | |||
| 3531 | RegExpMacroAssembler* assembler = compiler->macro_assembler(); | |||
| 3532 | if (!trace->is_trivial()) { | |||
| 3533 | trace->Flush(compiler, this); | |||
| 3534 | return; | |||
| 3535 | } | |||
| 3536 | ||||
| 3537 | LimitResult limit_result = LimitVersions(compiler, trace); | |||
| 3538 | if (limit_result == DONE) return; | |||
| 3539 | DCHECK(limit_result == CONTINUE)do { static_assert( mozilla::detail::AssertionConditionType< decltype(limit_result == CONTINUE)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(limit_result == CONTINUE))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("limit_result == CONTINUE" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3539); AnnotateMozCrashReason("MOZ_ASSERT" "(" "limit_result == CONTINUE" ")"); do { *((volatile int*)__null) = 3539; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3540 | ||||
| 3541 | RecursionCheck rc(compiler); | |||
| 3542 | ||||
| 3543 | DCHECK_EQ(start_reg_ + 1, end_reg_)do { static_assert( mozilla::detail::AssertionConditionType< decltype((start_reg_ + 1) == (end_reg_))>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!((start_reg_ + 1) == (end_reg_ )))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("(start_reg_ + 1) == (end_reg_)", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3543); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(start_reg_ + 1) == (end_reg_)" ")"); do { *((volatile int*)__null) = 3543; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3544 | if (IsIgnoreCase(compiler->flags())) { | |||
| 3545 | bool unicode = IsEitherUnicode(compiler->flags()); | |||
| 3546 | assembler->CheckNotBackReferenceIgnoreCase(start_reg_, read_backward(), | |||
| 3547 | unicode, trace->backtrack()); | |||
| 3548 | } else { | |||
| 3549 | assembler->CheckNotBackReference(start_reg_, read_backward(), | |||
| 3550 | trace->backtrack()); | |||
| 3551 | } | |||
| 3552 | // We are going to advance backward, so we may end up at the start. | |||
| 3553 | if (read_backward()) trace->set_at_start(Trace::UNKNOWN); | |||
| 3554 | ||||
| 3555 | // Check that the back reference does not end inside a surrogate pair. | |||
| 3556 | if (IsEitherUnicode(compiler->flags()) && !compiler->one_byte()) { | |||
| 3557 | assembler->CheckNotInSurrogatePair(trace->cp_offset(), trace->backtrack()); | |||
| 3558 | } | |||
| 3559 | on_success()->Emit(compiler, trace); | |||
| 3560 | } | |||
| 3561 | ||||
| 3562 | void TextNode::CalculateOffsets() { | |||
| 3563 | int element_count = elements()->length(); | |||
| 3564 | // Set up the offsets of the elements relative to the start. This is a fixed | |||
| 3565 | // quantity since a TextNode can only contain fixed-width things. | |||
| 3566 | int cp_offset = 0; | |||
| 3567 | for (int i = 0; i < element_count; i++) { | |||
| 3568 | TextElement& elm = elements()->at(i); | |||
| 3569 | elm.set_cp_offset(cp_offset); | |||
| 3570 | cp_offset += elm.length(); | |||
| 3571 | } | |||
| 3572 | } | |||
| 3573 | ||||
| 3574 | namespace { | |||
| 3575 | ||||
| 3576 | // Assertion propagation moves information about assertions such as | |||
| 3577 | // \b to the affected nodes. For instance, in /.\b./ information must | |||
| 3578 | // be propagated to the first '.' that whatever follows needs to know | |||
| 3579 | // if it matched a word or a non-word, and to the second '.' that it | |||
| 3580 | // has to check if it succeeds a word or non-word. In this case the | |||
| 3581 | // result will be something like: | |||
| 3582 | // | |||
| 3583 | // +-------+ +------------+ | |||
| 3584 | // | . | | . | | |||
| 3585 | // +-------+ ---> +------------+ | |||
| 3586 | // | word? | | check word | | |||
| 3587 | // +-------+ +------------+ | |||
| 3588 | class AssertionPropagator : public AllStatic { | |||
| 3589 | public: | |||
| 3590 | static void VisitText(TextNode* that) {} | |||
| 3591 | ||||
| 3592 | static void VisitAction(ActionNode* that) { | |||
| 3593 | // If the next node is interested in what it follows then this node | |||
| 3594 | // has to be interested too so it can pass the information on. | |||
| 3595 | that->info()->AddFromFollowing(that->on_success()->info()); | |||
| 3596 | } | |||
| 3597 | ||||
| 3598 | static void VisitChoice(ChoiceNode* that, int i) { | |||
| 3599 | // Anything the following nodes need to know has to be known by | |||
| 3600 | // this node also, so it can pass it on. | |||
| 3601 | that->info()->AddFromFollowing(that->alternatives()->at(i).node()->info()); | |||
| 3602 | } | |||
| 3603 | ||||
| 3604 | static void VisitLoopChoiceContinueNode(LoopChoiceNode* that) { | |||
| 3605 | that->info()->AddFromFollowing(that->continue_node()->info()); | |||
| 3606 | } | |||
| 3607 | ||||
| 3608 | static void VisitLoopChoiceLoopNode(LoopChoiceNode* that) { | |||
| 3609 | that->info()->AddFromFollowing(that->loop_node()->info()); | |||
| 3610 | } | |||
| 3611 | ||||
| 3612 | static void VisitNegativeLookaroundChoiceLookaroundNode( | |||
| 3613 | NegativeLookaroundChoiceNode* that) { | |||
| 3614 | VisitChoice(that, NegativeLookaroundChoiceNode::kLookaroundIndex); | |||
| 3615 | } | |||
| 3616 | ||||
| 3617 | static void VisitNegativeLookaroundChoiceContinueNode( | |||
| 3618 | NegativeLookaroundChoiceNode* that) { | |||
| 3619 | VisitChoice(that, NegativeLookaroundChoiceNode::kContinueIndex); | |||
| 3620 | } | |||
| 3621 | ||||
| 3622 | static void VisitBackReference(BackReferenceNode* that) {} | |||
| 3623 | ||||
| 3624 | static void VisitAssertion(AssertionNode* that) {} | |||
| 3625 | }; | |||
| 3626 | ||||
| 3627 | // Propagates information about the minimum size of successful matches from | |||
| 3628 | // successor nodes to their predecessors. Note that all eats_at_least values | |||
| 3629 | // are initialized to zero before analysis. | |||
| 3630 | class EatsAtLeastPropagator : public AllStatic { | |||
| 3631 | public: | |||
| 3632 | static void VisitText(TextNode* that) { | |||
| 3633 | // The eats_at_least value is not used if reading backward. | |||
| 3634 | if (!that->read_backward()) { | |||
| 3635 | // We are not at the start after this node, and thus we can use the | |||
| 3636 | // successor's eats_at_least_from_not_start value. | |||
| 3637 | uint8_t eats_at_least = base::saturated_cast<uint8_t>( | |||
| 3638 | that->Length() + that->on_success() | |||
| 3639 | ->eats_at_least_info() | |||
| 3640 | ->eats_at_least_from_not_start); | |||
| 3641 | that->set_eats_at_least_info(EatsAtLeastInfo(eats_at_least)); | |||
| 3642 | } | |||
| 3643 | } | |||
| 3644 | ||||
| 3645 | static void VisitAction(ActionNode* that) { | |||
| 3646 | switch (that->action_type()) { | |||
| 3647 | case ActionNode::BEGIN_POSITIVE_SUBMATCH: | |||
| 3648 | case ActionNode::POSITIVE_SUBMATCH_SUCCESS: | |||
| 3649 | // We do not propagate eats_at_least data through positive lookarounds, | |||
| 3650 | // because they rewind input. | |||
| 3651 | // TODO(v8:11859) Potential approaches for fixing this include: | |||
| 3652 | // 1. Add a dedicated choice node for positive lookaround, similar to | |||
| 3653 | // NegativeLookaroundChoiceNode. | |||
| 3654 | // 2. Add an eats_at_least_inside_loop field to EatsAtLeastInfo, which | |||
| 3655 | // is <= eats_at_least_from_possibly_start, and use that value in | |||
| 3656 | // EatsAtLeastFromLoopEntry. | |||
| 3657 | DCHECK(that->eats_at_least_info()->IsZero())do { static_assert( mozilla::detail::AssertionConditionType< decltype(that->eats_at_least_info()->IsZero())>::isValid , "invalid assertion condition"); if ((__builtin_expect(!!(!( !!(that->eats_at_least_info()->IsZero()))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("that->eats_at_least_info()->IsZero()" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3657); AnnotateMozCrashReason("MOZ_ASSERT" "(" "that->eats_at_least_info()->IsZero()" ")"); do { *((volatile int*)__null) = 3657; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3658 | break; | |||
| 3659 | case ActionNode::SET_REGISTER_FOR_LOOP: | |||
| 3660 | // SET_REGISTER_FOR_LOOP indicates a loop entry point, which means the | |||
| 3661 | // loop body will run at least the minimum number of times before the | |||
| 3662 | // continuation case can run. | |||
| 3663 | that->set_eats_at_least_info( | |||
| 3664 | that->on_success()->EatsAtLeastFromLoopEntry()); | |||
| 3665 | break; | |||
| 3666 | case ActionNode::BEGIN_NEGATIVE_SUBMATCH: | |||
| 3667 | default: | |||
| 3668 | // Otherwise, the current node eats at least as much as its successor. | |||
| 3669 | // Note: we can propagate eats_at_least data for BEGIN_NEGATIVE_SUBMATCH | |||
| 3670 | // because NegativeLookaroundChoiceNode ignores its lookaround successor | |||
| 3671 | // when computing eats-at-least and quick check information. | |||
| 3672 | that->set_eats_at_least_info(*that->on_success()->eats_at_least_info()); | |||
| 3673 | break; | |||
| 3674 | } | |||
| 3675 | } | |||
| 3676 | ||||
| 3677 | static void VisitChoice(ChoiceNode* that, int i) { | |||
| 3678 | // The minimum possible match from a choice node is the minimum of its | |||
| 3679 | // successors. | |||
| 3680 | EatsAtLeastInfo eats_at_least = | |||
| 3681 | i == 0 ? EatsAtLeastInfo(UINT8_MAX(255)) : *that->eats_at_least_info(); | |||
| 3682 | eats_at_least.SetMin( | |||
| 3683 | *that->alternatives()->at(i).node()->eats_at_least_info()); | |||
| 3684 | that->set_eats_at_least_info(eats_at_least); | |||
| 3685 | } | |||
| 3686 | ||||
| 3687 | static void VisitLoopChoiceContinueNode(LoopChoiceNode* that) { | |||
| 3688 | if (!that->read_backward()) { | |||
| 3689 | that->set_eats_at_least_info( | |||
| 3690 | *that->continue_node()->eats_at_least_info()); | |||
| 3691 | } | |||
| 3692 | } | |||
| 3693 | ||||
| 3694 | static void VisitLoopChoiceLoopNode(LoopChoiceNode* that) {} | |||
| 3695 | ||||
| 3696 | static void VisitNegativeLookaroundChoiceLookaroundNode( | |||
| 3697 | NegativeLookaroundChoiceNode* that) {} | |||
| 3698 | ||||
| 3699 | static void VisitNegativeLookaroundChoiceContinueNode( | |||
| 3700 | NegativeLookaroundChoiceNode* that) { | |||
| 3701 | that->set_eats_at_least_info(*that->continue_node()->eats_at_least_info()); | |||
| 3702 | } | |||
| 3703 | ||||
| 3704 | static void VisitBackReference(BackReferenceNode* that) { | |||
| 3705 | if (!that->read_backward()) { | |||
| 3706 | that->set_eats_at_least_info(*that->on_success()->eats_at_least_info()); | |||
| 3707 | } | |||
| 3708 | } | |||
| 3709 | ||||
| 3710 | static void VisitAssertion(AssertionNode* that) { | |||
| 3711 | EatsAtLeastInfo eats_at_least = *that->on_success()->eats_at_least_info(); | |||
| 3712 | if (that->assertion_type() == AssertionNode::AT_START) { | |||
| 3713 | // If we know we are not at the start and we are asked "how many | |||
| 3714 | // characters will you match if you succeed?" then we can answer anything | |||
| 3715 | // since false implies false. So let's just set the max answer | |||
| 3716 | // (UINT8_MAX) since that won't prevent us from preloading a lot of | |||
| 3717 | // characters for the other branches in the node graph. | |||
| 3718 | eats_at_least.eats_at_least_from_not_start = UINT8_MAX(255); | |||
| 3719 | } | |||
| 3720 | that->set_eats_at_least_info(eats_at_least); | |||
| 3721 | } | |||
| 3722 | }; | |||
| 3723 | ||||
| 3724 | } // namespace | |||
| 3725 | ||||
| 3726 | // ------------------------------------------------------------------- | |||
| 3727 | // Analysis | |||
| 3728 | ||||
| 3729 | // Iterates the node graph and provides the opportunity for propagators to set | |||
| 3730 | // values that depend on successor nodes. | |||
| 3731 | template <typename... Propagators> | |||
| 3732 | class Analysis : public NodeVisitor { | |||
| 3733 | public: | |||
| 3734 | Analysis(Isolate* isolate, bool is_one_byte, RegExpFlags flags) | |||
| 3735 | : isolate_(isolate), | |||
| 3736 | is_one_byte_(is_one_byte), | |||
| 3737 | flags_(flags), | |||
| 3738 | error_(RegExpError::kNone) {} | |||
| 3739 | ||||
| 3740 | void EnsureAnalyzed(RegExpNode* that) { | |||
| 3741 | StackLimitCheck check(isolate()); | |||
| 3742 | if (check.HasOverflowed()) { | |||
| 3743 | if (v8_flagsjs::jit::JitOptions.correctness_fuzzer_suppressions) { | |||
| 3744 | FATAL("Analysis: Aborting on stack overflow")do { do { } while (false); MOZ_ReportCrash("" "Analysis: Aborting on stack overflow" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3744); AnnotateMozCrashReason("MOZ_CRASH(" "Analysis: Aborting on stack overflow" ")"); do { *((volatile int*)__null) = 3744; __attribute__((nomerge )) ::abort(); } while (false); } while (false); | |||
| 3745 | } | |||
| 3746 | fail(RegExpError::kAnalysisStackOverflow); | |||
| 3747 | return; | |||
| 3748 | } | |||
| 3749 | if (that->info()->been_analyzed || that->info()->being_analyzed) return; | |||
| 3750 | that->info()->being_analyzed = true; | |||
| 3751 | that->Accept(this); | |||
| 3752 | that->info()->being_analyzed = false; | |||
| 3753 | that->info()->been_analyzed = true; | |||
| 3754 | } | |||
| 3755 | ||||
| 3756 | bool has_failed() { return error_ != RegExpError::kNone; } | |||
| 3757 | RegExpError error() { | |||
| 3758 | DCHECK(error_ != RegExpError::kNone)do { static_assert( mozilla::detail::AssertionConditionType< decltype(error_ != RegExpError::kNone)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(error_ != RegExpError::kNone ))), 0))) { do { } while (false); MOZ_ReportAssertionFailure( "error_ != RegExpError::kNone", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3758); AnnotateMozCrashReason("MOZ_ASSERT" "(" "error_ != RegExpError::kNone" ")"); do { *((volatile int*)__null) = 3758; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3759 | return error_; | |||
| 3760 | } | |||
| 3761 | void fail(RegExpError error) { error_ = error; } | |||
| 3762 | ||||
| 3763 | Isolate* isolate() const { return isolate_; } | |||
| 3764 | ||||
| 3765 | void VisitEnd(EndNode* that) override { | |||
| 3766 | // nothing to do | |||
| 3767 | } | |||
| 3768 | ||||
| 3769 | // Used to call the given static function on each propagator / variadic template | |||
| 3770 | // argument. | |||
| 3771 | #define STATIC_FOR_EACH(expr) \ | |||
| 3772 | do { \ | |||
| 3773 | int dummy[] = {((expr), 0)...}; \ | |||
| 3774 | USE(dummy)do { ::v8::base::Use unused_tmp_array_for_use_macro[]{dummy}; (void)unused_tmp_array_for_use_macro; } while (false); \ | |||
| 3775 | } while (false) | |||
| 3776 | ||||
| 3777 | void VisitText(TextNode* that) override { | |||
| 3778 | that->MakeCaseIndependent(isolate(), is_one_byte_, flags()); | |||
| 3779 | EnsureAnalyzed(that->on_success()); | |||
| 3780 | if (has_failed()) return; | |||
| 3781 | that->CalculateOffsets(); | |||
| 3782 | STATIC_FOR_EACH(Propagators::VisitText(that)); | |||
| 3783 | } | |||
| 3784 | ||||
| 3785 | void VisitAction(ActionNode* that) override { | |||
| 3786 | if (that->action_type() == ActionNode::MODIFY_FLAGS) { | |||
| 3787 | set_flags(that->flags()); | |||
| 3788 | } | |||
| 3789 | EnsureAnalyzed(that->on_success()); | |||
| 3790 | if (has_failed()) return; | |||
| 3791 | STATIC_FOR_EACH(Propagators::VisitAction(that)); | |||
| 3792 | } | |||
| 3793 | ||||
| 3794 | void VisitChoice(ChoiceNode* that) override { | |||
| 3795 | for (int i = 0; i < that->alternatives()->length(); i++) { | |||
| 3796 | EnsureAnalyzed(that->alternatives()->at(i).node()); | |||
| 3797 | if (has_failed()) return; | |||
| 3798 | STATIC_FOR_EACH(Propagators::VisitChoice(that, i)); | |||
| 3799 | } | |||
| 3800 | } | |||
| 3801 | ||||
| 3802 | void VisitLoopChoice(LoopChoiceNode* that) override { | |||
| 3803 | DCHECK_EQ(that->alternatives()->length(), 2)do { static_assert( mozilla::detail::AssertionConditionType< decltype((that->alternatives()->length()) == (2))>:: isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!((that->alternatives()->length()) == (2)))), 0) )) { do { } while (false); MOZ_ReportAssertionFailure("(that->alternatives()->length()) == (2)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3803); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(that->alternatives()->length()) == (2)" ")"); do { *((volatile int*)__null) = 3803; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); // Just loop and continue. | |||
| 3804 | ||||
| 3805 | // First propagate all information from the continuation node. | |||
| 3806 | EnsureAnalyzed(that->continue_node()); | |||
| 3807 | if (has_failed()) return; | |||
| 3808 | STATIC_FOR_EACH(Propagators::VisitLoopChoiceContinueNode(that)); | |||
| 3809 | ||||
| 3810 | // Check the loop last since it may need the value of this node | |||
| 3811 | // to get a correct result. | |||
| 3812 | EnsureAnalyzed(that->loop_node()); | |||
| 3813 | if (has_failed()) return; | |||
| 3814 | STATIC_FOR_EACH(Propagators::VisitLoopChoiceLoopNode(that)); | |||
| 3815 | } | |||
| 3816 | ||||
| 3817 | void VisitNegativeLookaroundChoice( | |||
| 3818 | NegativeLookaroundChoiceNode* that) override { | |||
| 3819 | DCHECK_EQ(that->alternatives()->length(), 2)do { static_assert( mozilla::detail::AssertionConditionType< decltype((that->alternatives()->length()) == (2))>:: isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!((that->alternatives()->length()) == (2)))), 0) )) { do { } while (false); MOZ_ReportAssertionFailure("(that->alternatives()->length()) == (2)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3819); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(that->alternatives()->length()) == (2)" ")"); do { *((volatile int*)__null) = 3819; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); // Lookaround and continue. | |||
| 3820 | ||||
| 3821 | EnsureAnalyzed(that->lookaround_node()); | |||
| 3822 | if (has_failed()) return; | |||
| 3823 | STATIC_FOR_EACH( | |||
| 3824 | Propagators::VisitNegativeLookaroundChoiceLookaroundNode(that)); | |||
| 3825 | ||||
| 3826 | EnsureAnalyzed(that->continue_node()); | |||
| 3827 | if (has_failed()) return; | |||
| 3828 | STATIC_FOR_EACH( | |||
| 3829 | Propagators::VisitNegativeLookaroundChoiceContinueNode(that)); | |||
| 3830 | } | |||
| 3831 | ||||
| 3832 | void VisitBackReference(BackReferenceNode* that) override { | |||
| 3833 | EnsureAnalyzed(that->on_success()); | |||
| 3834 | if (has_failed()) return; | |||
| 3835 | STATIC_FOR_EACH(Propagators::VisitBackReference(that)); | |||
| 3836 | } | |||
| 3837 | ||||
| 3838 | void VisitAssertion(AssertionNode* that) override { | |||
| 3839 | EnsureAnalyzed(that->on_success()); | |||
| 3840 | if (has_failed()) return; | |||
| 3841 | STATIC_FOR_EACH(Propagators::VisitAssertion(that)); | |||
| 3842 | } | |||
| 3843 | ||||
| 3844 | #undef STATIC_FOR_EACH | |||
| 3845 | ||||
| 3846 | private: | |||
| 3847 | RegExpFlags flags() const { return flags_; } | |||
| 3848 | void set_flags(RegExpFlags flags) { flags_ = flags; } | |||
| 3849 | ||||
| 3850 | Isolate* isolate_; | |||
| 3851 | const bool is_one_byte_; | |||
| 3852 | RegExpFlags flags_; | |||
| 3853 | RegExpError error_; | |||
| 3854 | ||||
| 3855 | DISALLOW_IMPLICIT_CONSTRUCTORS(Analysis)Analysis() = delete; Analysis(const Analysis&) = delete; Analysis & operator=(const Analysis&) = delete; | |||
| 3856 | }; | |||
| 3857 | ||||
| 3858 | RegExpError AnalyzeRegExp(Isolate* isolate, bool is_one_byte, RegExpFlags flags, | |||
| 3859 | RegExpNode* node) { | |||
| 3860 | Analysis<AssertionPropagator, EatsAtLeastPropagator> analysis( | |||
| 3861 | isolate, is_one_byte, flags); | |||
| 3862 | DCHECK_EQ(node->info()->been_analyzed, false)do { static_assert( mozilla::detail::AssertionConditionType< decltype((node->info()->been_analyzed) == (false))>:: isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!((node->info()->been_analyzed) == (false)))), 0 ))) { do { } while (false); MOZ_ReportAssertionFailure("(node->info()->been_analyzed) == (false)" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3862); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(node->info()->been_analyzed) == (false)" ")"); do { *((volatile int*)__null) = 3862; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3863 | analysis.EnsureAnalyzed(node); | |||
| 3864 | DCHECK_IMPLIES(analysis.has_failed(), analysis.error() != RegExpError::kNone)do { if (analysis.has_failed()) { do { static_assert( mozilla ::detail::AssertionConditionType<decltype(analysis.error() != RegExpError::kNone)>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(analysis.error() != RegExpError ::kNone))), 0))) { do { } while (false); MOZ_ReportAssertionFailure ("analysis.error() != RegExpError::kNone", "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3864); AnnotateMozCrashReason("MOZ_ASSERT" "(" "analysis.error() != RegExpError::kNone" ")"); do { *((volatile int*)__null) = 3864; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); } } while ( false); | |||
| 3865 | return analysis.has_failed() ? analysis.error() : RegExpError::kNone; | |||
| 3866 | } | |||
| 3867 | ||||
| 3868 | void BackReferenceNode::FillInBMInfo(Isolate* isolate, int offset, int budget, | |||
| 3869 | BoyerMooreLookahead* bm, | |||
| 3870 | bool not_at_start) { | |||
| 3871 | // Working out the set of characters that a backreference can match is too | |||
| 3872 | // hard, so we just say that any character can match. | |||
| 3873 | bm->SetRest(offset); | |||
| 3874 | SaveBMInfo(bm, not_at_start, offset); | |||
| 3875 | } | |||
| 3876 | ||||
| 3877 | static_assert(BoyerMoorePositionInfo::kMapSize == | |||
| 3878 | RegExpMacroAssembler::kTableSize); | |||
| 3879 | ||||
| 3880 | void ChoiceNode::FillInBMInfo(Isolate* isolate, int offset, int budget, | |||
| 3881 | BoyerMooreLookahead* bm, bool not_at_start) { | |||
| 3882 | ZoneList<GuardedAlternative>* alts = alternatives(); | |||
| 3883 | budget = (budget - 1) / alts->length(); | |||
| 3884 | for (int i = 0; i < alts->length(); i++) { | |||
| 3885 | GuardedAlternative& alt = alts->at(i); | |||
| 3886 | if (alt.guards() != nullptr && alt.guards()->length() != 0) { | |||
| 3887 | bm->SetRest(offset); // Give up trying to fill in info. | |||
| 3888 | SaveBMInfo(bm, not_at_start, offset); | |||
| 3889 | return; | |||
| 3890 | } | |||
| 3891 | alt.node()->FillInBMInfo(isolate, offset, budget, bm, not_at_start); | |||
| 3892 | } | |||
| 3893 | SaveBMInfo(bm, not_at_start, offset); | |||
| 3894 | } | |||
| 3895 | ||||
| 3896 | void TextNode::FillInBMInfo(Isolate* isolate, int initial_offset, int budget, | |||
| 3897 | BoyerMooreLookahead* bm, bool not_at_start) { | |||
| 3898 | if (initial_offset >= bm->length()) return; | |||
| 3899 | int offset = initial_offset; | |||
| 3900 | int max_char = bm->max_char(); | |||
| 3901 | for (int i = 0; i < elements()->length(); i++) { | |||
| 3902 | if (offset >= bm->length()) { | |||
| 3903 | if (initial_offset == 0) set_bm_info(not_at_start, bm); | |||
| 3904 | return; | |||
| 3905 | } | |||
| 3906 | TextElement text = elements()->at(i); | |||
| 3907 | if (text.text_type() == TextElement::ATOM) { | |||
| 3908 | RegExpAtom* atom = text.atom(); | |||
| 3909 | for (int j = 0; j < atom->length(); j++, offset++) { | |||
| 3910 | if (offset >= bm->length()) { | |||
| 3911 | if (initial_offset == 0) set_bm_info(not_at_start, bm); | |||
| 3912 | return; | |||
| 3913 | } | |||
| 3914 | base::uc16 character = atom->data()[j]; | |||
| 3915 | if (IsIgnoreCase(bm->compiler()->flags())) { | |||
| 3916 | unibrow::uchar chars[4]; | |||
| 3917 | int length = GetCaseIndependentLetters(isolate, character, | |||
| 3918 | bm->compiler(), chars, 4); | |||
| 3919 | for (int k = 0; k < length; k++) { | |||
| 3920 | bm->Set(offset, chars[k]); | |||
| 3921 | } | |||
| 3922 | } else { | |||
| 3923 | if (character <= max_char) bm->Set(offset, character); | |||
| 3924 | } | |||
| 3925 | } | |||
| 3926 | } else { | |||
| 3927 | DCHECK_EQ(TextElement::CLASS_RANGES, text.text_type())do { static_assert( mozilla::detail::AssertionConditionType< decltype((TextElement::CLASS_RANGES) == (text.text_type()))> ::isValid, "invalid assertion condition"); if ((__builtin_expect (!!(!(!!((TextElement::CLASS_RANGES) == (text.text_type())))) , 0))) { do { } while (false); MOZ_ReportAssertionFailure("(TextElement::CLASS_RANGES) == (text.text_type())" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3927); AnnotateMozCrashReason("MOZ_ASSERT" "(" "(TextElement::CLASS_RANGES) == (text.text_type())" ")"); do { *((volatile int*)__null) = 3927; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3928 | RegExpClassRanges* class_ranges = text.class_ranges(); | |||
| 3929 | ZoneList<CharacterRange>* ranges = class_ranges->ranges(zone()); | |||
| 3930 | if (class_ranges->is_negated()) { | |||
| 3931 | bm->SetAll(offset); | |||
| 3932 | } else { | |||
| 3933 | for (int k = 0; k < ranges->length(); k++) { | |||
| 3934 | CharacterRange& range = ranges->at(k); | |||
| 3935 | if (static_cast<int>(range.from()) > max_char) continue; | |||
| 3936 | int to = std::min(max_char, static_cast<int>(range.to())); | |||
| 3937 | bm->SetInterval(offset, Interval(range.from(), to)); | |||
| 3938 | } | |||
| 3939 | } | |||
| 3940 | offset++; | |||
| 3941 | } | |||
| 3942 | } | |||
| 3943 | if (offset >= bm->length()) { | |||
| 3944 | if (initial_offset == 0) set_bm_info(not_at_start, bm); | |||
| 3945 | return; | |||
| 3946 | } | |||
| 3947 | on_success()->FillInBMInfo(isolate, offset, budget - 1, bm, | |||
| 3948 | true); // Not at start after a text node. | |||
| 3949 | if (initial_offset == 0) set_bm_info(not_at_start, bm); | |||
| 3950 | } | |||
| 3951 | ||||
| 3952 | RegExpNode* RegExpCompiler::OptionallyStepBackToLeadSurrogate( | |||
| 3953 | RegExpNode* on_success) { | |||
| 3954 | DCHECK(!read_backward())do { static_assert( mozilla::detail::AssertionConditionType< decltype(!read_backward())>::isValid, "invalid assertion condition" ); if ((__builtin_expect(!!(!(!!(!read_backward()))), 0))) { do { } while (false); MOZ_ReportAssertionFailure("!read_backward()" , "/var/lib/jenkins/workspace/firefox-scan-build/js/src/irregexp/imported/regexp-compiler.cc" , 3954); AnnotateMozCrashReason("MOZ_ASSERT" "(" "!read_backward()" ")"); do { *((volatile int*)__null) = 3954; __attribute__((nomerge )) ::abort(); } while (false); } } while (false); | |||
| 3955 | ZoneList<CharacterRange>* lead_surrogates = CharacterRange::List( | |||
| 3956 | zone(), CharacterRange::Range(kLeadSurrogateStart, kLeadSurrogateEnd)); | |||
| 3957 | ZoneList<CharacterRange>* trail_surrogates = CharacterRange::List( | |||
| 3958 | zone(), CharacterRange::Range(kTrailSurrogateStart, kTrailSurrogateEnd)); | |||
| 3959 | ||||
| 3960 | ChoiceNode* optional_step_back = zone()->New<ChoiceNode>(2, zone()); | |||
| 3961 | ||||
| 3962 | int stack_register = UnicodeLookaroundStackRegister(); | |||
| 3963 | int position_register = UnicodeLookaroundPositionRegister(); | |||
| 3964 | RegExpNode* step_back = TextNode::CreateForCharacterRanges( | |||
| 3965 | zone(), lead_surrogates, true, on_success); | |||
| 3966 | RegExpLookaround::Builder builder(true, step_back, stack_register, | |||
| 3967 | position_register); | |||
| 3968 | RegExpNode* match_trail = TextNode::CreateForCharacterRanges( | |||
| 3969 | zone(), trail_surrogates, false, builder.on_match_success()); | |||
| 3970 | ||||
| 3971 | optional_step_back->AddAlternative( | |||
| 3972 | GuardedAlternative(builder.ForMatch(match_trail))); | |||
| 3973 | optional_step_back->AddAlternative(GuardedAlternative(on_success)); | |||
| 3974 | ||||
| 3975 | return optional_step_back; | |||
| 3976 | } | |||
| 3977 | ||||
| 3978 | RegExpNode* RegExpCompiler::PreprocessRegExp(RegExpCompileData* data, | |||
| 3979 | bool is_one_byte) { | |||
| 3980 | // Wrap the body of the regexp in capture #0. | |||
| 3981 | RegExpNode* captured_body = | |||
| 3982 | RegExpCapture::ToNode(data->tree, 0, this, accept()); | |||
| 3983 | RegExpNode* node = captured_body; | |||
| 3984 | if (!data->tree->IsAnchoredAtStart() && !IsSticky(flags())) { | |||
| 3985 | // Add a .*? at the beginning, outside the body capture, unless | |||
| 3986 | // this expression is anchored at the beginning or sticky. | |||
| 3987 | RegExpNode* loop_node = RegExpQuantifier::ToNode( | |||
| 3988 | 0, RegExpTree::kInfinity, false, | |||
| 3989 | zone()->New<RegExpClassRanges>(StandardCharacterSet::kEverything), this, | |||
| 3990 | captured_body, data->contains_anchor); | |||
| 3991 | ||||
| 3992 | if (data->contains_anchor) { | |||
| 3993 | // Unroll loop once, to take care of the case that might start | |||
| 3994 | // at the start of input. | |||
| 3995 | ChoiceNode* first_step_node = zone()->New<ChoiceNode>(2, zone()); | |||
| 3996 | first_step_node->AddAlternative(GuardedAlternative(captured_body)); | |||
| 3997 | first_step_node->AddAlternative(GuardedAlternative(zone()->New<TextNode>( | |||
| 3998 | zone()->New<RegExpClassRanges>(StandardCharacterSet::kEverything), | |||
| 3999 | false, loop_node))); | |||
| 4000 | node = first_step_node; | |||
| 4001 | } else { | |||
| 4002 | node = loop_node; | |||
| 4003 | } | |||
| 4004 | } | |||
| 4005 | if (is_one_byte) { | |||
| 4006 | node = node->FilterOneByte(RegExpCompiler::kMaxRecursion, this); | |||
| 4007 | // Do it again to propagate the new nodes to places where they were not | |||
| 4008 | // put because they had not been calculated yet. | |||
| 4009 | if (node != nullptr) { | |||
| 4010 | node = node->FilterOneByte(RegExpCompiler::kMaxRecursion, this); | |||
| 4011 | } | |||
| 4012 | } else if (IsEitherUnicode(flags()) && | |||
| 4013 | (IsGlobal(flags()) || IsSticky(flags()))) { | |||
| 4014 | node = OptionallyStepBackToLeadSurrogate(node); | |||
| 4015 | } | |||
| 4016 | ||||
| 4017 | if (node == nullptr) node = zone()->New<EndNode>(EndNode::BACKTRACK, zone()); | |||
| 4018 | return node; | |||
| 4019 | } | |||
| 4020 | ||||
| 4021 | void RegExpCompiler::ToNodeCheckForStackOverflow() { | |||
| 4022 | if (StackLimitCheck{isolate()}.HasOverflowed()) { | |||
| 4023 | V8::FatalProcessOutOfMemory(isolate(), "RegExpCompiler"); | |||
| 4024 | } | |||
| 4025 | } | |||
| 4026 | ||||
| 4027 | } // namespace internal | |||
| 4028 | } // namespace v8 |