1 | //===------------------------- ItaniumDemangle.cpp ------------------------===// |
2 | // |
3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
4 | // See https://llvm.org/LICENSE.txt for license information. |
5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
6 | // |
7 | //===----------------------------------------------------------------------===// |
8 | |
9 | // FIXME: (possibly) incomplete list of features that clang mangles that this |
10 | // file does not yet support: |
11 | // - C++ modules TS |
12 | |
13 | #include "llvm/Demangle/Demangle.h" |
14 | #include "llvm/Demangle/ItaniumDemangle.h" |
15 | |
16 | #include <cassert> |
17 | #include <cctype> |
18 | #include <cstdio> |
19 | #include <cstdlib> |
20 | #include <cstring> |
21 | #include <exception> |
22 | #include <functional> |
23 | #include <utility> |
24 | |
25 | using namespace llvm; |
26 | using namespace llvm::itanium_demangle; |
27 | |
28 | constexpr const char *itanium_demangle::FloatData<float>::spec; |
29 | constexpr const char *itanium_demangle::FloatData<double>::spec; |
30 | constexpr const char *itanium_demangle::FloatData<long double>::spec; |
31 | |
32 | // <discriminator> := _ <non-negative number> # when number < 10 |
33 | // := __ <non-negative number> _ # when number >= 10 |
34 | // extension := decimal-digit+ # at the end of string |
35 | const char *itanium_demangle::parse_discriminator(const char *first, |
36 | const char *last) { |
37 | // parse but ignore discriminator |
38 | if (first != last) { |
39 | if (*first == '_') { |
40 | const char *t1 = first + 1; |
41 | if (t1 != last) { |
42 | if (std::isdigit(*t1)) |
43 | first = t1 + 1; |
44 | else if (*t1 == '_') { |
45 | for (++t1; t1 != last && std::isdigit(*t1); ++t1) |
46 | ; |
47 | if (t1 != last && *t1 == '_') |
48 | first = t1 + 1; |
49 | } |
50 | } |
51 | } else if (std::isdigit(*first)) { |
52 | const char *t1 = first + 1; |
53 | for (; t1 != last && std::isdigit(*t1); ++t1) |
54 | ; |
55 | if (t1 == last) |
56 | first = last; |
57 | } |
58 | } |
59 | return first; |
60 | } |
61 | |
62 | #ifndef NDEBUG |
63 | namespace { |
64 | struct DumpVisitor { |
65 | unsigned Depth = 0; |
66 | bool PendingNewline = false; |
67 | |
68 | template<typename NodeT> static constexpr bool wantsNewline(const NodeT *) { |
69 | return true; |
70 | } |
71 | static bool wantsNewline(NodeArray A) { return !A.empty(); } |
72 | static constexpr bool wantsNewline(...) { return false; } |
73 | |
74 | template<typename ...Ts> static bool anyWantNewline(Ts ...Vs) { |
75 | for (bool B : {wantsNewline(Vs)...}) |
76 | if (B) |
77 | return true; |
78 | return false; |
79 | } |
80 | |
81 | void printStr(const char *S) { fprintf(stderr, "%s" , S); } |
82 | void print(std::string_view SV) { |
83 | fprintf(stderr, "\"%.*s\"" , (int)SV.size(), SV.data()); |
84 | } |
85 | void print(const Node *N) { |
86 | if (N) |
87 | N->visit(std::ref(*this)); |
88 | else |
89 | printStr("<null>" ); |
90 | } |
91 | void print(NodeArray A) { |
92 | ++Depth; |
93 | printStr("{" ); |
94 | bool First = true; |
95 | for (const Node *N : A) { |
96 | if (First) |
97 | print(N); |
98 | else |
99 | printWithComma(N); |
100 | First = false; |
101 | } |
102 | printStr("}" ); |
103 | --Depth; |
104 | } |
105 | |
106 | // Overload used when T is exactly 'bool', not merely convertible to 'bool'. |
107 | void print(bool B) { printStr(B ? "true" : "false" ); } |
108 | |
109 | template <class T> std::enable_if_t<std::is_unsigned<T>::value> print(T N) { |
110 | fprintf(stderr, "%llu" , (unsigned long long)N); |
111 | } |
112 | |
113 | template <class T> std::enable_if_t<std::is_signed<T>::value> print(T N) { |
114 | fprintf(stderr, "%lld" , (long long)N); |
115 | } |
116 | |
117 | void print(ReferenceKind RK) { |
118 | switch (RK) { |
119 | case ReferenceKind::LValue: |
120 | return printStr("ReferenceKind::LValue" ); |
121 | case ReferenceKind::RValue: |
122 | return printStr("ReferenceKind::RValue" ); |
123 | } |
124 | } |
125 | void print(FunctionRefQual RQ) { |
126 | switch (RQ) { |
127 | case FunctionRefQual::FrefQualNone: |
128 | return printStr("FunctionRefQual::FrefQualNone" ); |
129 | case FunctionRefQual::FrefQualLValue: |
130 | return printStr("FunctionRefQual::FrefQualLValue" ); |
131 | case FunctionRefQual::FrefQualRValue: |
132 | return printStr("FunctionRefQual::FrefQualRValue" ); |
133 | } |
134 | } |
135 | void print(Qualifiers Qs) { |
136 | if (!Qs) return printStr("QualNone" ); |
137 | struct QualName { Qualifiers Q; const char *Name; } Names[] = { |
138 | {QualConst, "QualConst" }, |
139 | {QualVolatile, "QualVolatile" }, |
140 | {QualRestrict, "QualRestrict" }, |
141 | }; |
142 | for (QualName Name : Names) { |
143 | if (Qs & Name.Q) { |
144 | printStr(Name.Name); |
145 | Qs = Qualifiers(Qs & ~Name.Q); |
146 | if (Qs) printStr(" | " ); |
147 | } |
148 | } |
149 | } |
150 | void print(SpecialSubKind SSK) { |
151 | switch (SSK) { |
152 | case SpecialSubKind::allocator: |
153 | return printStr("SpecialSubKind::allocator" ); |
154 | case SpecialSubKind::basic_string: |
155 | return printStr("SpecialSubKind::basic_string" ); |
156 | case SpecialSubKind::string: |
157 | return printStr("SpecialSubKind::string" ); |
158 | case SpecialSubKind::istream: |
159 | return printStr("SpecialSubKind::istream" ); |
160 | case SpecialSubKind::ostream: |
161 | return printStr("SpecialSubKind::ostream" ); |
162 | case SpecialSubKind::iostream: |
163 | return printStr("SpecialSubKind::iostream" ); |
164 | } |
165 | } |
166 | void print(TemplateParamKind TPK) { |
167 | switch (TPK) { |
168 | case TemplateParamKind::Type: |
169 | return printStr("TemplateParamKind::Type" ); |
170 | case TemplateParamKind::NonType: |
171 | return printStr("TemplateParamKind::NonType" ); |
172 | case TemplateParamKind::Template: |
173 | return printStr("TemplateParamKind::Template" ); |
174 | } |
175 | } |
176 | void print(Node::Prec P) { |
177 | switch (P) { |
178 | case Node::Prec::Primary: |
179 | return printStr("Node::Prec::Primary" ); |
180 | case Node::Prec::Postfix: |
181 | return printStr("Node::Prec::Postfix" ); |
182 | case Node::Prec::Unary: |
183 | return printStr("Node::Prec::Unary" ); |
184 | case Node::Prec::Cast: |
185 | return printStr("Node::Prec::Cast" ); |
186 | case Node::Prec::PtrMem: |
187 | return printStr("Node::Prec::PtrMem" ); |
188 | case Node::Prec::Multiplicative: |
189 | return printStr("Node::Prec::Multiplicative" ); |
190 | case Node::Prec::Additive: |
191 | return printStr("Node::Prec::Additive" ); |
192 | case Node::Prec::Shift: |
193 | return printStr("Node::Prec::Shift" ); |
194 | case Node::Prec::Spaceship: |
195 | return printStr("Node::Prec::Spaceship" ); |
196 | case Node::Prec::Relational: |
197 | return printStr("Node::Prec::Relational" ); |
198 | case Node::Prec::Equality: |
199 | return printStr("Node::Prec::Equality" ); |
200 | case Node::Prec::And: |
201 | return printStr("Node::Prec::And" ); |
202 | case Node::Prec::Xor: |
203 | return printStr("Node::Prec::Xor" ); |
204 | case Node::Prec::Ior: |
205 | return printStr("Node::Prec::Ior" ); |
206 | case Node::Prec::AndIf: |
207 | return printStr("Node::Prec::AndIf" ); |
208 | case Node::Prec::OrIf: |
209 | return printStr("Node::Prec::OrIf" ); |
210 | case Node::Prec::Conditional: |
211 | return printStr("Node::Prec::Conditional" ); |
212 | case Node::Prec::Assign: |
213 | return printStr("Node::Prec::Assign" ); |
214 | case Node::Prec::Comma: |
215 | return printStr("Node::Prec::Comma" ); |
216 | case Node::Prec::Default: |
217 | return printStr("Node::Prec::Default" ); |
218 | } |
219 | } |
220 | |
221 | void newLine() { |
222 | printStr("\n" ); |
223 | for (unsigned I = 0; I != Depth; ++I) |
224 | printStr(" " ); |
225 | PendingNewline = false; |
226 | } |
227 | |
228 | template<typename T> void printWithPendingNewline(T V) { |
229 | print(V); |
230 | if (wantsNewline(V)) |
231 | PendingNewline = true; |
232 | } |
233 | |
234 | template<typename T> void printWithComma(T V) { |
235 | if (PendingNewline || wantsNewline(V)) { |
236 | printStr("," ); |
237 | newLine(); |
238 | } else { |
239 | printStr(", " ); |
240 | } |
241 | |
242 | printWithPendingNewline(V); |
243 | } |
244 | |
245 | struct CtorArgPrinter { |
246 | DumpVisitor &Visitor; |
247 | |
248 | template<typename T, typename ...Rest> void operator()(T V, Rest ...Vs) { |
249 | if (Visitor.anyWantNewline(V, Vs...)) |
250 | Visitor.newLine(); |
251 | Visitor.printWithPendingNewline(V); |
252 | int PrintInOrder[] = { (Visitor.printWithComma(Vs), 0)..., 0 }; |
253 | (void)PrintInOrder; |
254 | } |
255 | }; |
256 | |
257 | template<typename NodeT> void operator()(const NodeT *Node) { |
258 | Depth += 2; |
259 | fprintf(stderr, "%s(" , itanium_demangle::NodeKind<NodeT>::name()); |
260 | Node->match(CtorArgPrinter{*this}); |
261 | fprintf(stderr, ")" ); |
262 | Depth -= 2; |
263 | } |
264 | |
265 | void operator()(const ForwardTemplateReference *Node) { |
266 | Depth += 2; |
267 | fprintf(stderr, "ForwardTemplateReference(" ); |
268 | if (Node->Ref && !Node->Printing) { |
269 | Node->Printing = true; |
270 | CtorArgPrinter{*this}(Node->Ref); |
271 | Node->Printing = false; |
272 | } else { |
273 | CtorArgPrinter{*this}(Node->Index); |
274 | } |
275 | fprintf(stderr, ")" ); |
276 | Depth -= 2; |
277 | } |
278 | }; |
279 | } |
280 | |
281 | void itanium_demangle::Node::dump() const { |
282 | DumpVisitor V; |
283 | visit(std::ref(V)); |
284 | V.newLine(); |
285 | } |
286 | #endif |
287 | |
288 | namespace { |
289 | class BumpPointerAllocator { |
290 | struct BlockMeta { |
291 | BlockMeta* Next; |
292 | size_t Current; |
293 | }; |
294 | |
295 | static constexpr size_t AllocSize = 4096; |
296 | static constexpr size_t UsableAllocSize = AllocSize - sizeof(BlockMeta); |
297 | |
298 | alignas(long double) char InitialBuffer[AllocSize]; |
299 | BlockMeta* BlockList = nullptr; |
300 | |
301 | void grow() { |
302 | char* NewMeta = static_cast<char *>(std::malloc(size: AllocSize)); |
303 | if (NewMeta == nullptr) |
304 | std::terminate(); |
305 | BlockList = new (NewMeta) BlockMeta{.Next: BlockList, .Current: 0}; |
306 | } |
307 | |
308 | void* allocateMassive(size_t NBytes) { |
309 | NBytes += sizeof(BlockMeta); |
310 | BlockMeta* NewMeta = reinterpret_cast<BlockMeta*>(std::malloc(size: NBytes)); |
311 | if (NewMeta == nullptr) |
312 | std::terminate(); |
313 | BlockList->Next = new (NewMeta) BlockMeta{.Next: BlockList->Next, .Current: 0}; |
314 | return static_cast<void*>(NewMeta + 1); |
315 | } |
316 | |
317 | public: |
318 | BumpPointerAllocator() |
319 | : BlockList(new (InitialBuffer) BlockMeta{.Next: nullptr, .Current: 0}) {} |
320 | |
321 | void* allocate(size_t N) { |
322 | N = (N + 15u) & ~15u; |
323 | if (N + BlockList->Current >= UsableAllocSize) { |
324 | if (N > UsableAllocSize) |
325 | return allocateMassive(NBytes: N); |
326 | grow(); |
327 | } |
328 | BlockList->Current += N; |
329 | return static_cast<void*>(reinterpret_cast<char*>(BlockList + 1) + |
330 | BlockList->Current - N); |
331 | } |
332 | |
333 | void reset() { |
334 | while (BlockList) { |
335 | BlockMeta* Tmp = BlockList; |
336 | BlockList = BlockList->Next; |
337 | if (reinterpret_cast<char*>(Tmp) != InitialBuffer) |
338 | std::free(ptr: Tmp); |
339 | } |
340 | BlockList = new (InitialBuffer) BlockMeta{.Next: nullptr, .Current: 0}; |
341 | } |
342 | |
343 | ~BumpPointerAllocator() { reset(); } |
344 | }; |
345 | |
346 | class DefaultAllocator { |
347 | BumpPointerAllocator Alloc; |
348 | |
349 | public: |
350 | void reset() { Alloc.reset(); } |
351 | |
352 | template<typename T, typename ...Args> T *makeNode(Args &&...args) { |
353 | return new (Alloc.allocate(N: sizeof(T))) |
354 | T(std::forward<Args>(args)...); |
355 | } |
356 | |
357 | void *allocateNodeArray(size_t sz) { |
358 | return Alloc.allocate(N: sizeof(Node *) * sz); |
359 | } |
360 | }; |
361 | } // unnamed namespace |
362 | |
363 | //===----------------------------------------------------------------------===// |
364 | // Code beyond this point should not be synchronized with libc++abi. |
365 | //===----------------------------------------------------------------------===// |
366 | |
367 | using Demangler = itanium_demangle::ManglingParser<DefaultAllocator>; |
368 | |
369 | char *llvm::itaniumDemangle(std::string_view MangledName, bool ParseParams) { |
370 | if (MangledName.empty()) |
371 | return nullptr; |
372 | |
373 | Demangler Parser(MangledName.data(), |
374 | MangledName.data() + MangledName.length()); |
375 | Node *AST = Parser.parse(ParseParams); |
376 | if (!AST) |
377 | return nullptr; |
378 | |
379 | OutputBuffer OB; |
380 | assert(Parser.ForwardTemplateRefs.empty()); |
381 | AST->print(OB); |
382 | OB += '\0'; |
383 | return OB.getBuffer(); |
384 | } |
385 | |
386 | ItaniumPartialDemangler::ItaniumPartialDemangler() |
387 | : RootNode(nullptr), Context(new Demangler{nullptr, nullptr}) {} |
388 | |
389 | ItaniumPartialDemangler::~ItaniumPartialDemangler() { |
390 | delete static_cast<Demangler *>(Context); |
391 | } |
392 | |
393 | ItaniumPartialDemangler::ItaniumPartialDemangler( |
394 | ItaniumPartialDemangler &&Other) |
395 | : RootNode(Other.RootNode), Context(Other.Context) { |
396 | Other.Context = Other.RootNode = nullptr; |
397 | } |
398 | |
399 | ItaniumPartialDemangler &ItaniumPartialDemangler:: |
400 | operator=(ItaniumPartialDemangler &&Other) { |
401 | std::swap(a&: RootNode, b&: Other.RootNode); |
402 | std::swap(a&: Context, b&: Other.Context); |
403 | return *this; |
404 | } |
405 | |
406 | // Demangle MangledName into an AST, storing it into this->RootNode. |
407 | bool ItaniumPartialDemangler::partialDemangle(const char *MangledName) { |
408 | Demangler *Parser = static_cast<Demangler *>(Context); |
409 | size_t Len = std::strlen(s: MangledName); |
410 | Parser->reset(First_: MangledName, Last_: MangledName + Len); |
411 | RootNode = Parser->parse(); |
412 | return RootNode == nullptr; |
413 | } |
414 | |
415 | static char *printNode(const Node *RootNode, char *Buf, size_t *N) { |
416 | OutputBuffer OB(Buf, N); |
417 | RootNode->print(OB); |
418 | OB += '\0'; |
419 | if (N != nullptr) |
420 | *N = OB.getCurrentPosition(); |
421 | return OB.getBuffer(); |
422 | } |
423 | |
424 | char *ItaniumPartialDemangler::getFunctionBaseName(char *Buf, size_t *N) const { |
425 | if (!isFunction()) |
426 | return nullptr; |
427 | |
428 | const Node *Name = static_cast<const FunctionEncoding *>(RootNode)->getName(); |
429 | |
430 | while (true) { |
431 | switch (Name->getKind()) { |
432 | case Node::KAbiTagAttr: |
433 | Name = static_cast<const AbiTagAttr *>(Name)->Base; |
434 | continue; |
435 | case Node::KModuleEntity: |
436 | Name = static_cast<const ModuleEntity *>(Name)->Name; |
437 | continue; |
438 | case Node::KNestedName: |
439 | Name = static_cast<const NestedName *>(Name)->Name; |
440 | continue; |
441 | case Node::KLocalName: |
442 | Name = static_cast<const LocalName *>(Name)->Entity; |
443 | continue; |
444 | case Node::KNameWithTemplateArgs: |
445 | Name = static_cast<const NameWithTemplateArgs *>(Name)->Name; |
446 | continue; |
447 | default: |
448 | return printNode(RootNode: Name, Buf, N); |
449 | } |
450 | } |
451 | } |
452 | |
453 | char *ItaniumPartialDemangler::getFunctionDeclContextName(char *Buf, |
454 | size_t *N) const { |
455 | if (!isFunction()) |
456 | return nullptr; |
457 | const Node *Name = static_cast<const FunctionEncoding *>(RootNode)->getName(); |
458 | |
459 | OutputBuffer OB(Buf, N); |
460 | |
461 | KeepGoingLocalFunction: |
462 | while (true) { |
463 | if (Name->getKind() == Node::KAbiTagAttr) { |
464 | Name = static_cast<const AbiTagAttr *>(Name)->Base; |
465 | continue; |
466 | } |
467 | if (Name->getKind() == Node::KNameWithTemplateArgs) { |
468 | Name = static_cast<const NameWithTemplateArgs *>(Name)->Name; |
469 | continue; |
470 | } |
471 | break; |
472 | } |
473 | |
474 | if (Name->getKind() == Node::KModuleEntity) |
475 | Name = static_cast<const ModuleEntity *>(Name)->Name; |
476 | |
477 | switch (Name->getKind()) { |
478 | case Node::KNestedName: |
479 | static_cast<const NestedName *>(Name)->Qual->print(OB); |
480 | break; |
481 | case Node::KLocalName: { |
482 | auto *LN = static_cast<const LocalName *>(Name); |
483 | LN->Encoding->print(OB); |
484 | OB += "::" ; |
485 | Name = LN->Entity; |
486 | goto KeepGoingLocalFunction; |
487 | } |
488 | default: |
489 | break; |
490 | } |
491 | OB += '\0'; |
492 | if (N != nullptr) |
493 | *N = OB.getCurrentPosition(); |
494 | return OB.getBuffer(); |
495 | } |
496 | |
497 | char *ItaniumPartialDemangler::getFunctionName(char *Buf, size_t *N) const { |
498 | if (!isFunction()) |
499 | return nullptr; |
500 | auto *Name = static_cast<FunctionEncoding *>(RootNode)->getName(); |
501 | return printNode(RootNode: Name, Buf, N); |
502 | } |
503 | |
504 | char *ItaniumPartialDemangler::getFunctionParameters(char *Buf, |
505 | size_t *N) const { |
506 | if (!isFunction()) |
507 | return nullptr; |
508 | NodeArray Params = static_cast<FunctionEncoding *>(RootNode)->getParams(); |
509 | |
510 | OutputBuffer OB(Buf, N); |
511 | |
512 | OB += '('; |
513 | Params.printWithComma(OB); |
514 | OB += ')'; |
515 | OB += '\0'; |
516 | if (N != nullptr) |
517 | *N = OB.getCurrentPosition(); |
518 | return OB.getBuffer(); |
519 | } |
520 | |
521 | char *ItaniumPartialDemangler::getFunctionReturnType( |
522 | char *Buf, size_t *N) const { |
523 | if (!isFunction()) |
524 | return nullptr; |
525 | |
526 | OutputBuffer OB(Buf, N); |
527 | |
528 | if (const Node *Ret = |
529 | static_cast<const FunctionEncoding *>(RootNode)->getReturnType()) |
530 | Ret->print(OB); |
531 | |
532 | OB += '\0'; |
533 | if (N != nullptr) |
534 | *N = OB.getCurrentPosition(); |
535 | return OB.getBuffer(); |
536 | } |
537 | |
538 | char *ItaniumPartialDemangler::finishDemangle(char *Buf, size_t *N) const { |
539 | assert(RootNode != nullptr && "must call partialDemangle()" ); |
540 | return printNode(RootNode: static_cast<Node *>(RootNode), Buf, N); |
541 | } |
542 | |
543 | bool ItaniumPartialDemangler::hasFunctionQualifiers() const { |
544 | assert(RootNode != nullptr && "must call partialDemangle()" ); |
545 | if (!isFunction()) |
546 | return false; |
547 | auto *E = static_cast<const FunctionEncoding *>(RootNode); |
548 | return E->getCVQuals() != QualNone || E->getRefQual() != FrefQualNone; |
549 | } |
550 | |
551 | bool ItaniumPartialDemangler::isCtorOrDtor() const { |
552 | const Node *N = static_cast<const Node *>(RootNode); |
553 | while (N) { |
554 | switch (N->getKind()) { |
555 | default: |
556 | return false; |
557 | case Node::KCtorDtorName: |
558 | return true; |
559 | |
560 | case Node::KAbiTagAttr: |
561 | N = static_cast<const AbiTagAttr *>(N)->Base; |
562 | break; |
563 | case Node::KFunctionEncoding: |
564 | N = static_cast<const FunctionEncoding *>(N)->getName(); |
565 | break; |
566 | case Node::KLocalName: |
567 | N = static_cast<const LocalName *>(N)->Entity; |
568 | break; |
569 | case Node::KNameWithTemplateArgs: |
570 | N = static_cast<const NameWithTemplateArgs *>(N)->Name; |
571 | break; |
572 | case Node::KNestedName: |
573 | N = static_cast<const NestedName *>(N)->Name; |
574 | break; |
575 | case Node::KModuleEntity: |
576 | N = static_cast<const ModuleEntity *>(N)->Name; |
577 | break; |
578 | } |
579 | } |
580 | return false; |
581 | } |
582 | |
583 | bool ItaniumPartialDemangler::isFunction() const { |
584 | assert(RootNode != nullptr && "must call partialDemangle()" ); |
585 | return static_cast<const Node *>(RootNode)->getKind() == |
586 | Node::KFunctionEncoding; |
587 | } |
588 | |
589 | bool ItaniumPartialDemangler::isSpecialName() const { |
590 | assert(RootNode != nullptr && "must call partialDemangle()" ); |
591 | auto K = static_cast<const Node *>(RootNode)->getKind(); |
592 | return K == Node::KSpecialName || K == Node::KCtorVtableSpecialName; |
593 | } |
594 | |
595 | bool ItaniumPartialDemangler::isData() const { |
596 | return !isFunction() && !isSpecialName(); |
597 | } |
598 | |