1 //===- FileCheck.cpp - Check that File's Contents match what is expected --===//
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 // FileCheck does a line-by line check of a file that validates whether it
10 // contains the expected content. This is useful for regression tests etc.
11 //
12 // This file implements most of the API that will be used by the FileCheck utility
13 // as well as various unittests.
14 //===----------------------------------------------------------------------===//
15
16 #include "llvm/FileCheck/FileCheck.h"
17 #include "FileCheckImpl.h"
18 #include "llvm/ADT/STLExtras.h"
19 #include "llvm/ADT/StringSet.h"
20 #include "llvm/ADT/Twine.h"
21 #include "llvm/Support/CheckedArithmetic.h"
22 #include "llvm/Support/FormatVariadic.h"
23 #include <cstdint>
24 #include <list>
25 #include <set>
26 #include <tuple>
27 #include <utility>
28
29 using namespace llvm;
30
toString() const31 StringRef ExpressionFormat::toString() const {
32 switch (Value) {
33 case Kind::NoFormat:
34 return StringRef("<none>");
35 case Kind::Unsigned:
36 return StringRef("%u");
37 case Kind::Signed:
38 return StringRef("%d");
39 case Kind::HexUpper:
40 return StringRef("%X");
41 case Kind::HexLower:
42 return StringRef("%x");
43 }
44 llvm_unreachable("unknown expression format");
45 }
46
getWildcardRegex() const47 Expected<std::string> ExpressionFormat::getWildcardRegex() const {
48 StringRef AlternateFormPrefix = AlternateForm ? StringRef("0x") : StringRef();
49
50 auto CreatePrecisionRegex = [&](StringRef S) {
51 return (Twine(AlternateFormPrefix) + S + Twine('{') + Twine(Precision) +
52 "}")
53 .str();
54 };
55
56 switch (Value) {
57 case Kind::Unsigned:
58 if (Precision)
59 return CreatePrecisionRegex("([1-9][0-9]*)?[0-9]");
60 return std::string("[0-9]+");
61 case Kind::Signed:
62 if (Precision)
63 return CreatePrecisionRegex("-?([1-9][0-9]*)?[0-9]");
64 return std::string("-?[0-9]+");
65 case Kind::HexUpper:
66 if (Precision)
67 return CreatePrecisionRegex("([1-9A-F][0-9A-F]*)?[0-9A-F]");
68 return (Twine(AlternateFormPrefix) + Twine("[0-9A-F]+")).str();
69 case Kind::HexLower:
70 if (Precision)
71 return CreatePrecisionRegex("([1-9a-f][0-9a-f]*)?[0-9a-f]");
72 return (Twine(AlternateFormPrefix) + Twine("[0-9a-f]+")).str();
73 default:
74 return createStringError(std::errc::invalid_argument,
75 "trying to match value with invalid format");
76 }
77 }
78
79 Expected<std::string>
getMatchingString(ExpressionValue IntegerValue) const80 ExpressionFormat::getMatchingString(ExpressionValue IntegerValue) const {
81 uint64_t AbsoluteValue;
82 StringRef SignPrefix = IntegerValue.isNegative() ? "-" : "";
83
84 if (Value == Kind::Signed) {
85 Expected<int64_t> SignedValue = IntegerValue.getSignedValue();
86 if (!SignedValue)
87 return SignedValue.takeError();
88 if (*SignedValue < 0)
89 AbsoluteValue = cantFail(IntegerValue.getAbsolute().getUnsignedValue());
90 else
91 AbsoluteValue = *SignedValue;
92 } else {
93 Expected<uint64_t> UnsignedValue = IntegerValue.getUnsignedValue();
94 if (!UnsignedValue)
95 return UnsignedValue.takeError();
96 AbsoluteValue = *UnsignedValue;
97 }
98
99 std::string AbsoluteValueStr;
100 switch (Value) {
101 case Kind::Unsigned:
102 case Kind::Signed:
103 AbsoluteValueStr = utostr(AbsoluteValue);
104 break;
105 case Kind::HexUpper:
106 case Kind::HexLower:
107 AbsoluteValueStr = utohexstr(AbsoluteValue, Value == Kind::HexLower);
108 break;
109 default:
110 return createStringError(std::errc::invalid_argument,
111 "trying to match value with invalid format");
112 }
113
114 StringRef AlternateFormPrefix = AlternateForm ? StringRef("0x") : StringRef();
115
116 if (Precision > AbsoluteValueStr.size()) {
117 unsigned LeadingZeros = Precision - AbsoluteValueStr.size();
118 return (Twine(SignPrefix) + Twine(AlternateFormPrefix) +
119 std::string(LeadingZeros, '0') + AbsoluteValueStr)
120 .str();
121 }
122
123 return (Twine(SignPrefix) + Twine(AlternateFormPrefix) + AbsoluteValueStr)
124 .str();
125 }
126
127 Expected<ExpressionValue>
valueFromStringRepr(StringRef StrVal,const SourceMgr & SM) const128 ExpressionFormat::valueFromStringRepr(StringRef StrVal,
129 const SourceMgr &SM) const {
130 bool ValueIsSigned = Value == Kind::Signed;
131 // Both the FileCheck utility and library only call this method with a valid
132 // value in StrVal. This is guaranteed by the regex returned by
133 // getWildcardRegex() above. Only underflow and overflow errors can thus
134 // occur. However new uses of this method could be added in the future so
135 // the error message does not make assumptions about StrVal.
136 StringRef IntegerParseErrorStr = "unable to represent numeric value";
137 if (ValueIsSigned) {
138 int64_t SignedValue;
139
140 if (StrVal.getAsInteger(10, SignedValue))
141 return ErrorDiagnostic::get(SM, StrVal, IntegerParseErrorStr);
142
143 return ExpressionValue(SignedValue);
144 }
145
146 bool Hex = Value == Kind::HexUpper || Value == Kind::HexLower;
147 uint64_t UnsignedValue;
148 bool MissingFormPrefix = AlternateForm && !StrVal.consume_front("0x");
149 if (StrVal.getAsInteger(Hex ? 16 : 10, UnsignedValue))
150 return ErrorDiagnostic::get(SM, StrVal, IntegerParseErrorStr);
151
152 // Error out for a missing prefix only now that we know we have an otherwise
153 // valid integer. For example, "-0x18" is reported above instead.
154 if (MissingFormPrefix)
155 return ErrorDiagnostic::get(SM, StrVal, "missing alternate form prefix");
156
157 return ExpressionValue(UnsignedValue);
158 }
159
getAsSigned(uint64_t UnsignedValue)160 static int64_t getAsSigned(uint64_t UnsignedValue) {
161 // Use memcpy to reinterpret the bitpattern in Value since casting to
162 // signed is implementation-defined if the unsigned value is too big to be
163 // represented in the signed type and using an union violates type aliasing
164 // rules.
165 int64_t SignedValue;
166 memcpy(&SignedValue, &UnsignedValue, sizeof(SignedValue));
167 return SignedValue;
168 }
169
getSignedValue() const170 Expected<int64_t> ExpressionValue::getSignedValue() const {
171 if (Negative)
172 return getAsSigned(Value);
173
174 if (Value > (uint64_t)std::numeric_limits<int64_t>::max())
175 return make_error<OverflowError>();
176
177 // Value is in the representable range of int64_t so we can use cast.
178 return static_cast<int64_t>(Value);
179 }
180
getUnsignedValue() const181 Expected<uint64_t> ExpressionValue::getUnsignedValue() const {
182 if (Negative)
183 return make_error<OverflowError>();
184
185 return Value;
186 }
187
getAbsolute() const188 ExpressionValue ExpressionValue::getAbsolute() const {
189 if (!Negative)
190 return *this;
191
192 int64_t SignedValue = getAsSigned(Value);
193 int64_t MaxInt64 = std::numeric_limits<int64_t>::max();
194 // Absolute value can be represented as int64_t.
195 if (SignedValue >= -MaxInt64)
196 return ExpressionValue(-getAsSigned(Value));
197
198 // -X == -(max int64_t + Rem), negate each component independently.
199 SignedValue += MaxInt64;
200 uint64_t RemainingValueAbsolute = -SignedValue;
201 return ExpressionValue(MaxInt64 + RemainingValueAbsolute);
202 }
203
operator +(const ExpressionValue & LeftOperand,const ExpressionValue & RightOperand)204 Expected<ExpressionValue> llvm::operator+(const ExpressionValue &LeftOperand,
205 const ExpressionValue &RightOperand) {
206 if (LeftOperand.isNegative() && RightOperand.isNegative()) {
207 int64_t LeftValue = cantFail(LeftOperand.getSignedValue());
208 int64_t RightValue = cantFail(RightOperand.getSignedValue());
209 std::optional<int64_t> Result = checkedAdd<int64_t>(LeftValue, RightValue);
210 if (!Result)
211 return make_error<OverflowError>();
212
213 return ExpressionValue(*Result);
214 }
215
216 // (-A) + B == B - A.
217 if (LeftOperand.isNegative())
218 return RightOperand - LeftOperand.getAbsolute();
219
220 // A + (-B) == A - B.
221 if (RightOperand.isNegative())
222 return LeftOperand - RightOperand.getAbsolute();
223
224 // Both values are positive at this point.
225 uint64_t LeftValue = cantFail(LeftOperand.getUnsignedValue());
226 uint64_t RightValue = cantFail(RightOperand.getUnsignedValue());
227 std::optional<uint64_t> Result =
228 checkedAddUnsigned<uint64_t>(LeftValue, RightValue);
229 if (!Result)
230 return make_error<OverflowError>();
231
232 return ExpressionValue(*Result);
233 }
234
operator -(const ExpressionValue & LeftOperand,const ExpressionValue & RightOperand)235 Expected<ExpressionValue> llvm::operator-(const ExpressionValue &LeftOperand,
236 const ExpressionValue &RightOperand) {
237 // Result will be negative and thus might underflow.
238 if (LeftOperand.isNegative() && !RightOperand.isNegative()) {
239 int64_t LeftValue = cantFail(LeftOperand.getSignedValue());
240 uint64_t RightValue = cantFail(RightOperand.getUnsignedValue());
241 // Result <= -1 - (max int64_t) which overflows on 1- and 2-complement.
242 if (RightValue > (uint64_t)std::numeric_limits<int64_t>::max())
243 return make_error<OverflowError>();
244 std::optional<int64_t> Result =
245 checkedSub(LeftValue, static_cast<int64_t>(RightValue));
246 if (!Result)
247 return make_error<OverflowError>();
248
249 return ExpressionValue(*Result);
250 }
251
252 // (-A) - (-B) == B - A.
253 if (LeftOperand.isNegative())
254 return RightOperand.getAbsolute() - LeftOperand.getAbsolute();
255
256 // A - (-B) == A + B.
257 if (RightOperand.isNegative())
258 return LeftOperand + RightOperand.getAbsolute();
259
260 // Both values are positive at this point.
261 uint64_t LeftValue = cantFail(LeftOperand.getUnsignedValue());
262 uint64_t RightValue = cantFail(RightOperand.getUnsignedValue());
263 if (LeftValue >= RightValue)
264 return ExpressionValue(LeftValue - RightValue);
265 else {
266 uint64_t AbsoluteDifference = RightValue - LeftValue;
267 uint64_t MaxInt64 = std::numeric_limits<int64_t>::max();
268 // Value might underflow.
269 if (AbsoluteDifference > MaxInt64) {
270 AbsoluteDifference -= MaxInt64;
271 int64_t Result = -MaxInt64;
272 int64_t MinInt64 = std::numeric_limits<int64_t>::min();
273 // Underflow, tested by:
274 // abs(Result + (max int64_t)) > abs((min int64_t) + (max int64_t))
275 if (AbsoluteDifference > static_cast<uint64_t>(-(MinInt64 - Result)))
276 return make_error<OverflowError>();
277 Result -= static_cast<int64_t>(AbsoluteDifference);
278 return ExpressionValue(Result);
279 }
280
281 return ExpressionValue(-static_cast<int64_t>(AbsoluteDifference));
282 }
283 }
284
operator *(const ExpressionValue & LeftOperand,const ExpressionValue & RightOperand)285 Expected<ExpressionValue> llvm::operator*(const ExpressionValue &LeftOperand,
286 const ExpressionValue &RightOperand) {
287 // -A * -B == A * B
288 if (LeftOperand.isNegative() && RightOperand.isNegative())
289 return LeftOperand.getAbsolute() * RightOperand.getAbsolute();
290
291 // A * -B == -B * A
292 if (RightOperand.isNegative())
293 return RightOperand * LeftOperand;
294
295 assert(!RightOperand.isNegative() && "Unexpected negative operand!");
296
297 // Result will be negative and can underflow.
298 if (LeftOperand.isNegative()) {
299 auto Result = LeftOperand.getAbsolute() * RightOperand.getAbsolute();
300 if (!Result)
301 return Result;
302
303 return ExpressionValue(0) - *Result;
304 }
305
306 // Result will be positive and can overflow.
307 uint64_t LeftValue = cantFail(LeftOperand.getUnsignedValue());
308 uint64_t RightValue = cantFail(RightOperand.getUnsignedValue());
309 std::optional<uint64_t> Result =
310 checkedMulUnsigned<uint64_t>(LeftValue, RightValue);
311 if (!Result)
312 return make_error<OverflowError>();
313
314 return ExpressionValue(*Result);
315 }
316
operator /(const ExpressionValue & LeftOperand,const ExpressionValue & RightOperand)317 Expected<ExpressionValue> llvm::operator/(const ExpressionValue &LeftOperand,
318 const ExpressionValue &RightOperand) {
319 // -A / -B == A / B
320 if (LeftOperand.isNegative() && RightOperand.isNegative())
321 return LeftOperand.getAbsolute() / RightOperand.getAbsolute();
322
323 // Check for divide by zero.
324 if (RightOperand == ExpressionValue(0))
325 return make_error<OverflowError>();
326
327 // Result will be negative and can underflow.
328 if (LeftOperand.isNegative() || RightOperand.isNegative())
329 return ExpressionValue(0) -
330 cantFail(LeftOperand.getAbsolute() / RightOperand.getAbsolute());
331
332 uint64_t LeftValue = cantFail(LeftOperand.getUnsignedValue());
333 uint64_t RightValue = cantFail(RightOperand.getUnsignedValue());
334 return ExpressionValue(LeftValue / RightValue);
335 }
336
max(const ExpressionValue & LeftOperand,const ExpressionValue & RightOperand)337 Expected<ExpressionValue> llvm::max(const ExpressionValue &LeftOperand,
338 const ExpressionValue &RightOperand) {
339 if (LeftOperand.isNegative() && RightOperand.isNegative()) {
340 int64_t LeftValue = cantFail(LeftOperand.getSignedValue());
341 int64_t RightValue = cantFail(RightOperand.getSignedValue());
342 return ExpressionValue(std::max(LeftValue, RightValue));
343 }
344
345 if (!LeftOperand.isNegative() && !RightOperand.isNegative()) {
346 uint64_t LeftValue = cantFail(LeftOperand.getUnsignedValue());
347 uint64_t RightValue = cantFail(RightOperand.getUnsignedValue());
348 return ExpressionValue(std::max(LeftValue, RightValue));
349 }
350
351 if (LeftOperand.isNegative())
352 return RightOperand;
353
354 return LeftOperand;
355 }
356
min(const ExpressionValue & LeftOperand,const ExpressionValue & RightOperand)357 Expected<ExpressionValue> llvm::min(const ExpressionValue &LeftOperand,
358 const ExpressionValue &RightOperand) {
359 if (cantFail(max(LeftOperand, RightOperand)) == LeftOperand)
360 return RightOperand;
361
362 return LeftOperand;
363 }
364
eval() const365 Expected<ExpressionValue> NumericVariableUse::eval() const {
366 std::optional<ExpressionValue> Value = Variable->getValue();
367 if (Value)
368 return *Value;
369
370 return make_error<UndefVarError>(getExpressionStr());
371 }
372
eval() const373 Expected<ExpressionValue> BinaryOperation::eval() const {
374 Expected<ExpressionValue> LeftOp = LeftOperand->eval();
375 Expected<ExpressionValue> RightOp = RightOperand->eval();
376
377 // Bubble up any error (e.g. undefined variables) in the recursive
378 // evaluation.
379 if (!LeftOp || !RightOp) {
380 Error Err = Error::success();
381 if (!LeftOp)
382 Err = joinErrors(std::move(Err), LeftOp.takeError());
383 if (!RightOp)
384 Err = joinErrors(std::move(Err), RightOp.takeError());
385 return std::move(Err);
386 }
387
388 return EvalBinop(*LeftOp, *RightOp);
389 }
390
391 Expected<ExpressionFormat>
getImplicitFormat(const SourceMgr & SM) const392 BinaryOperation::getImplicitFormat(const SourceMgr &SM) const {
393 Expected<ExpressionFormat> LeftFormat = LeftOperand->getImplicitFormat(SM);
394 Expected<ExpressionFormat> RightFormat = RightOperand->getImplicitFormat(SM);
395 if (!LeftFormat || !RightFormat) {
396 Error Err = Error::success();
397 if (!LeftFormat)
398 Err = joinErrors(std::move(Err), LeftFormat.takeError());
399 if (!RightFormat)
400 Err = joinErrors(std::move(Err), RightFormat.takeError());
401 return std::move(Err);
402 }
403
404 if (*LeftFormat != ExpressionFormat::Kind::NoFormat &&
405 *RightFormat != ExpressionFormat::Kind::NoFormat &&
406 *LeftFormat != *RightFormat)
407 return ErrorDiagnostic::get(
408 SM, getExpressionStr(),
409 "implicit format conflict between '" + LeftOperand->getExpressionStr() +
410 "' (" + LeftFormat->toString() + ") and '" +
411 RightOperand->getExpressionStr() + "' (" + RightFormat->toString() +
412 "), need an explicit format specifier");
413
414 return *LeftFormat != ExpressionFormat::Kind::NoFormat ? *LeftFormat
415 : *RightFormat;
416 }
417
getResult() const418 Expected<std::string> NumericSubstitution::getResult() const {
419 assert(ExpressionPointer->getAST() != nullptr &&
420 "Substituting empty expression");
421 Expected<ExpressionValue> EvaluatedValue =
422 ExpressionPointer->getAST()->eval();
423 if (!EvaluatedValue)
424 return EvaluatedValue.takeError();
425 ExpressionFormat Format = ExpressionPointer->getFormat();
426 return Format.getMatchingString(*EvaluatedValue);
427 }
428
getResult() const429 Expected<std::string> StringSubstitution::getResult() const {
430 // Look up the value and escape it so that we can put it into the regex.
431 Expected<StringRef> VarVal = Context->getPatternVarValue(FromStr);
432 if (!VarVal)
433 return VarVal.takeError();
434 return Regex::escape(*VarVal);
435 }
436
isValidVarNameStart(char C)437 bool Pattern::isValidVarNameStart(char C) { return C == '_' || isAlpha(C); }
438
439 Expected<Pattern::VariableProperties>
parseVariable(StringRef & Str,const SourceMgr & SM)440 Pattern::parseVariable(StringRef &Str, const SourceMgr &SM) {
441 if (Str.empty())
442 return ErrorDiagnostic::get(SM, Str, "empty variable name");
443
444 size_t I = 0;
445 bool IsPseudo = Str[0] == '@';
446
447 // Global vars start with '$'.
448 if (Str[0] == '$' || IsPseudo)
449 ++I;
450
451 if (!isValidVarNameStart(Str[I++]))
452 return ErrorDiagnostic::get(SM, Str, "invalid variable name");
453
454 for (size_t E = Str.size(); I != E; ++I)
455 // Variable names are composed of alphanumeric characters and underscores.
456 if (Str[I] != '_' && !isAlnum(Str[I]))
457 break;
458
459 StringRef Name = Str.take_front(I);
460 Str = Str.substr(I);
461 return VariableProperties {Name, IsPseudo};
462 }
463
464 // StringRef holding all characters considered as horizontal whitespaces by
465 // FileCheck input canonicalization.
466 constexpr StringLiteral SpaceChars = " \t";
467
468 // Parsing helper function that strips the first character in S and returns it.
popFront(StringRef & S)469 static char popFront(StringRef &S) {
470 char C = S.front();
471 S = S.drop_front();
472 return C;
473 }
474
475 char OverflowError::ID = 0;
476 char UndefVarError::ID = 0;
477 char ErrorDiagnostic::ID = 0;
478 char NotFoundError::ID = 0;
479 char ErrorReported::ID = 0;
480
parseNumericVariableDefinition(StringRef & Expr,FileCheckPatternContext * Context,std::optional<size_t> LineNumber,ExpressionFormat ImplicitFormat,const SourceMgr & SM)481 Expected<NumericVariable *> Pattern::parseNumericVariableDefinition(
482 StringRef &Expr, FileCheckPatternContext *Context,
483 std::optional<size_t> LineNumber, ExpressionFormat ImplicitFormat,
484 const SourceMgr &SM) {
485 Expected<VariableProperties> ParseVarResult = parseVariable(Expr, SM);
486 if (!ParseVarResult)
487 return ParseVarResult.takeError();
488 StringRef Name = ParseVarResult->Name;
489
490 if (ParseVarResult->IsPseudo)
491 return ErrorDiagnostic::get(
492 SM, Name, "definition of pseudo numeric variable unsupported");
493
494 // Detect collisions between string and numeric variables when the latter
495 // is created later than the former.
496 if (Context->DefinedVariableTable.find(Name) !=
497 Context->DefinedVariableTable.end())
498 return ErrorDiagnostic::get(
499 SM, Name, "string variable with name '" + Name + "' already exists");
500
501 Expr = Expr.ltrim(SpaceChars);
502 if (!Expr.empty())
503 return ErrorDiagnostic::get(
504 SM, Expr, "unexpected characters after numeric variable name");
505
506 NumericVariable *DefinedNumericVariable;
507 auto VarTableIter = Context->GlobalNumericVariableTable.find(Name);
508 if (VarTableIter != Context->GlobalNumericVariableTable.end()) {
509 DefinedNumericVariable = VarTableIter->second;
510 if (DefinedNumericVariable->getImplicitFormat() != ImplicitFormat)
511 return ErrorDiagnostic::get(
512 SM, Expr, "format different from previous variable definition");
513 } else
514 DefinedNumericVariable =
515 Context->makeNumericVariable(Name, ImplicitFormat, LineNumber);
516
517 return DefinedNumericVariable;
518 }
519
parseNumericVariableUse(StringRef Name,bool IsPseudo,std::optional<size_t> LineNumber,FileCheckPatternContext * Context,const SourceMgr & SM)520 Expected<std::unique_ptr<NumericVariableUse>> Pattern::parseNumericVariableUse(
521 StringRef Name, bool IsPseudo, std::optional<size_t> LineNumber,
522 FileCheckPatternContext *Context, const SourceMgr &SM) {
523 if (IsPseudo && !Name.equals("@LINE"))
524 return ErrorDiagnostic::get(
525 SM, Name, "invalid pseudo numeric variable '" + Name + "'");
526
527 // Numeric variable definitions and uses are parsed in the order in which
528 // they appear in the CHECK patterns. For each definition, the pointer to the
529 // class instance of the corresponding numeric variable definition is stored
530 // in GlobalNumericVariableTable in parsePattern. Therefore, if the pointer
531 // we get below is null, it means no such variable was defined before. When
532 // that happens, we create a dummy variable so that parsing can continue. All
533 // uses of undefined variables, whether string or numeric, are then diagnosed
534 // in printNoMatch() after failing to match.
535 auto VarTableIter = Context->GlobalNumericVariableTable.find(Name);
536 NumericVariable *NumericVariable;
537 if (VarTableIter != Context->GlobalNumericVariableTable.end())
538 NumericVariable = VarTableIter->second;
539 else {
540 NumericVariable = Context->makeNumericVariable(
541 Name, ExpressionFormat(ExpressionFormat::Kind::Unsigned));
542 Context->GlobalNumericVariableTable[Name] = NumericVariable;
543 }
544
545 std::optional<size_t> DefLineNumber = NumericVariable->getDefLineNumber();
546 if (DefLineNumber && LineNumber && *DefLineNumber == *LineNumber)
547 return ErrorDiagnostic::get(
548 SM, Name,
549 "numeric variable '" + Name +
550 "' defined earlier in the same CHECK directive");
551
552 return std::make_unique<NumericVariableUse>(Name, NumericVariable);
553 }
554
parseNumericOperand(StringRef & Expr,AllowedOperand AO,bool MaybeInvalidConstraint,std::optional<size_t> LineNumber,FileCheckPatternContext * Context,const SourceMgr & SM)555 Expected<std::unique_ptr<ExpressionAST>> Pattern::parseNumericOperand(
556 StringRef &Expr, AllowedOperand AO, bool MaybeInvalidConstraint,
557 std::optional<size_t> LineNumber, FileCheckPatternContext *Context,
558 const SourceMgr &SM) {
559 if (Expr.startswith("(")) {
560 if (AO != AllowedOperand::Any)
561 return ErrorDiagnostic::get(
562 SM, Expr, "parenthesized expression not permitted here");
563 return parseParenExpr(Expr, LineNumber, Context, SM);
564 }
565
566 if (AO == AllowedOperand::LineVar || AO == AllowedOperand::Any) {
567 // Try to parse as a numeric variable use.
568 Expected<Pattern::VariableProperties> ParseVarResult =
569 parseVariable(Expr, SM);
570 if (ParseVarResult) {
571 // Try to parse a function call.
572 if (Expr.ltrim(SpaceChars).startswith("(")) {
573 if (AO != AllowedOperand::Any)
574 return ErrorDiagnostic::get(SM, ParseVarResult->Name,
575 "unexpected function call");
576
577 return parseCallExpr(Expr, ParseVarResult->Name, LineNumber, Context,
578 SM);
579 }
580
581 return parseNumericVariableUse(ParseVarResult->Name,
582 ParseVarResult->IsPseudo, LineNumber,
583 Context, SM);
584 }
585
586 if (AO == AllowedOperand::LineVar)
587 return ParseVarResult.takeError();
588 // Ignore the error and retry parsing as a literal.
589 consumeError(ParseVarResult.takeError());
590 }
591
592 // Otherwise, parse it as a literal.
593 int64_t SignedLiteralValue;
594 uint64_t UnsignedLiteralValue;
595 StringRef SaveExpr = Expr;
596 // Accept both signed and unsigned literal, default to signed literal.
597 if (!Expr.consumeInteger((AO == AllowedOperand::LegacyLiteral) ? 10 : 0,
598 UnsignedLiteralValue))
599 return std::make_unique<ExpressionLiteral>(SaveExpr.drop_back(Expr.size()),
600 UnsignedLiteralValue);
601 Expr = SaveExpr;
602 if (AO == AllowedOperand::Any && !Expr.consumeInteger(0, SignedLiteralValue))
603 return std::make_unique<ExpressionLiteral>(SaveExpr.drop_back(Expr.size()),
604 SignedLiteralValue);
605
606 return ErrorDiagnostic::get(
607 SM, Expr,
608 Twine("invalid ") +
609 (MaybeInvalidConstraint ? "matching constraint or " : "") +
610 "operand format");
611 }
612
613 Expected<std::unique_ptr<ExpressionAST>>
parseParenExpr(StringRef & Expr,std::optional<size_t> LineNumber,FileCheckPatternContext * Context,const SourceMgr & SM)614 Pattern::parseParenExpr(StringRef &Expr, std::optional<size_t> LineNumber,
615 FileCheckPatternContext *Context, const SourceMgr &SM) {
616 Expr = Expr.ltrim(SpaceChars);
617 assert(Expr.startswith("("));
618
619 // Parse right operand.
620 Expr.consume_front("(");
621 Expr = Expr.ltrim(SpaceChars);
622 if (Expr.empty())
623 return ErrorDiagnostic::get(SM, Expr, "missing operand in expression");
624
625 // Note: parseNumericOperand handles nested opening parentheses.
626 Expected<std::unique_ptr<ExpressionAST>> SubExprResult = parseNumericOperand(
627 Expr, AllowedOperand::Any, /*MaybeInvalidConstraint=*/false, LineNumber,
628 Context, SM);
629 Expr = Expr.ltrim(SpaceChars);
630 while (SubExprResult && !Expr.empty() && !Expr.startswith(")")) {
631 StringRef OrigExpr = Expr;
632 SubExprResult = parseBinop(OrigExpr, Expr, std::move(*SubExprResult), false,
633 LineNumber, Context, SM);
634 Expr = Expr.ltrim(SpaceChars);
635 }
636 if (!SubExprResult)
637 return SubExprResult;
638
639 if (!Expr.consume_front(")")) {
640 return ErrorDiagnostic::get(SM, Expr,
641 "missing ')' at end of nested expression");
642 }
643 return SubExprResult;
644 }
645
646 Expected<std::unique_ptr<ExpressionAST>>
parseBinop(StringRef Expr,StringRef & RemainingExpr,std::unique_ptr<ExpressionAST> LeftOp,bool IsLegacyLineExpr,std::optional<size_t> LineNumber,FileCheckPatternContext * Context,const SourceMgr & SM)647 Pattern::parseBinop(StringRef Expr, StringRef &RemainingExpr,
648 std::unique_ptr<ExpressionAST> LeftOp,
649 bool IsLegacyLineExpr, std::optional<size_t> LineNumber,
650 FileCheckPatternContext *Context, const SourceMgr &SM) {
651 RemainingExpr = RemainingExpr.ltrim(SpaceChars);
652 if (RemainingExpr.empty())
653 return std::move(LeftOp);
654
655 // Check if this is a supported operation and select a function to perform
656 // it.
657 SMLoc OpLoc = SMLoc::getFromPointer(RemainingExpr.data());
658 char Operator = popFront(RemainingExpr);
659 binop_eval_t EvalBinop;
660 switch (Operator) {
661 case '+':
662 EvalBinop = operator+;
663 break;
664 case '-':
665 EvalBinop = operator-;
666 break;
667 default:
668 return ErrorDiagnostic::get(
669 SM, OpLoc, Twine("unsupported operation '") + Twine(Operator) + "'");
670 }
671
672 // Parse right operand.
673 RemainingExpr = RemainingExpr.ltrim(SpaceChars);
674 if (RemainingExpr.empty())
675 return ErrorDiagnostic::get(SM, RemainingExpr,
676 "missing operand in expression");
677 // The second operand in a legacy @LINE expression is always a literal.
678 AllowedOperand AO =
679 IsLegacyLineExpr ? AllowedOperand::LegacyLiteral : AllowedOperand::Any;
680 Expected<std::unique_ptr<ExpressionAST>> RightOpResult =
681 parseNumericOperand(RemainingExpr, AO, /*MaybeInvalidConstraint=*/false,
682 LineNumber, Context, SM);
683 if (!RightOpResult)
684 return RightOpResult;
685
686 Expr = Expr.drop_back(RemainingExpr.size());
687 return std::make_unique<BinaryOperation>(Expr, EvalBinop, std::move(LeftOp),
688 std::move(*RightOpResult));
689 }
690
691 Expected<std::unique_ptr<ExpressionAST>>
parseCallExpr(StringRef & Expr,StringRef FuncName,std::optional<size_t> LineNumber,FileCheckPatternContext * Context,const SourceMgr & SM)692 Pattern::parseCallExpr(StringRef &Expr, StringRef FuncName,
693 std::optional<size_t> LineNumber,
694 FileCheckPatternContext *Context, const SourceMgr &SM) {
695 Expr = Expr.ltrim(SpaceChars);
696 assert(Expr.startswith("("));
697
698 auto OptFunc = StringSwitch<binop_eval_t>(FuncName)
699 .Case("add", operator+)
700 .Case("div", operator/)
701 .Case("max", max)
702 .Case("min", min)
703 .Case("mul", operator*)
704 .Case("sub", operator-)
705 .Default(nullptr);
706
707 if (!OptFunc)
708 return ErrorDiagnostic::get(
709 SM, FuncName, Twine("call to undefined function '") + FuncName + "'");
710
711 Expr.consume_front("(");
712 Expr = Expr.ltrim(SpaceChars);
713
714 // Parse call arguments, which are comma separated.
715 SmallVector<std::unique_ptr<ExpressionAST>, 4> Args;
716 while (!Expr.empty() && !Expr.startswith(")")) {
717 if (Expr.startswith(","))
718 return ErrorDiagnostic::get(SM, Expr, "missing argument");
719
720 // Parse the argument, which is an arbitary expression.
721 StringRef OuterBinOpExpr = Expr;
722 Expected<std::unique_ptr<ExpressionAST>> Arg = parseNumericOperand(
723 Expr, AllowedOperand::Any, /*MaybeInvalidConstraint=*/false, LineNumber,
724 Context, SM);
725 while (Arg && !Expr.empty()) {
726 Expr = Expr.ltrim(SpaceChars);
727 // Have we reached an argument terminator?
728 if (Expr.startswith(",") || Expr.startswith(")"))
729 break;
730
731 // Arg = Arg <op> <expr>
732 Arg = parseBinop(OuterBinOpExpr, Expr, std::move(*Arg), false, LineNumber,
733 Context, SM);
734 }
735
736 // Prefer an expression error over a generic invalid argument message.
737 if (!Arg)
738 return Arg.takeError();
739 Args.push_back(std::move(*Arg));
740
741 // Have we parsed all available arguments?
742 Expr = Expr.ltrim(SpaceChars);
743 if (!Expr.consume_front(","))
744 break;
745
746 Expr = Expr.ltrim(SpaceChars);
747 if (Expr.startswith(")"))
748 return ErrorDiagnostic::get(SM, Expr, "missing argument");
749 }
750
751 if (!Expr.consume_front(")"))
752 return ErrorDiagnostic::get(SM, Expr,
753 "missing ')' at end of call expression");
754
755 const unsigned NumArgs = Args.size();
756 if (NumArgs == 2)
757 return std::make_unique<BinaryOperation>(Expr, *OptFunc, std::move(Args[0]),
758 std::move(Args[1]));
759
760 // TODO: Support more than binop_eval_t.
761 return ErrorDiagnostic::get(SM, FuncName,
762 Twine("function '") + FuncName +
763 Twine("' takes 2 arguments but ") +
764 Twine(NumArgs) + " given");
765 }
766
parseNumericSubstitutionBlock(StringRef Expr,std::optional<NumericVariable * > & DefinedNumericVariable,bool IsLegacyLineExpr,std::optional<size_t> LineNumber,FileCheckPatternContext * Context,const SourceMgr & SM)767 Expected<std::unique_ptr<Expression>> Pattern::parseNumericSubstitutionBlock(
768 StringRef Expr, std::optional<NumericVariable *> &DefinedNumericVariable,
769 bool IsLegacyLineExpr, std::optional<size_t> LineNumber,
770 FileCheckPatternContext *Context, const SourceMgr &SM) {
771 std::unique_ptr<ExpressionAST> ExpressionASTPointer = nullptr;
772 StringRef DefExpr = StringRef();
773 DefinedNumericVariable = std::nullopt;
774 ExpressionFormat ExplicitFormat = ExpressionFormat();
775 unsigned Precision = 0;
776
777 // Parse format specifier (NOTE: ',' is also an argument seperator).
778 size_t FormatSpecEnd = Expr.find(',');
779 size_t FunctionStart = Expr.find('(');
780 if (FormatSpecEnd != StringRef::npos && FormatSpecEnd < FunctionStart) {
781 StringRef FormatExpr = Expr.take_front(FormatSpecEnd);
782 Expr = Expr.drop_front(FormatSpecEnd + 1);
783 FormatExpr = FormatExpr.trim(SpaceChars);
784 if (!FormatExpr.consume_front("%"))
785 return ErrorDiagnostic::get(
786 SM, FormatExpr,
787 "invalid matching format specification in expression");
788
789 // Parse alternate form flag.
790 SMLoc AlternateFormFlagLoc = SMLoc::getFromPointer(FormatExpr.data());
791 bool AlternateForm = FormatExpr.consume_front("#");
792
793 // Parse precision.
794 if (FormatExpr.consume_front(".")) {
795 if (FormatExpr.consumeInteger(10, Precision))
796 return ErrorDiagnostic::get(SM, FormatExpr,
797 "invalid precision in format specifier");
798 }
799
800 if (!FormatExpr.empty()) {
801 // Check for unknown matching format specifier and set matching format in
802 // class instance representing this expression.
803 SMLoc FmtLoc = SMLoc::getFromPointer(FormatExpr.data());
804 switch (popFront(FormatExpr)) {
805 case 'u':
806 ExplicitFormat =
807 ExpressionFormat(ExpressionFormat::Kind::Unsigned, Precision);
808 break;
809 case 'd':
810 ExplicitFormat =
811 ExpressionFormat(ExpressionFormat::Kind::Signed, Precision);
812 break;
813 case 'x':
814 ExplicitFormat = ExpressionFormat(ExpressionFormat::Kind::HexLower,
815 Precision, AlternateForm);
816 break;
817 case 'X':
818 ExplicitFormat = ExpressionFormat(ExpressionFormat::Kind::HexUpper,
819 Precision, AlternateForm);
820 break;
821 default:
822 return ErrorDiagnostic::get(SM, FmtLoc,
823 "invalid format specifier in expression");
824 }
825 }
826
827 if (AlternateForm && ExplicitFormat != ExpressionFormat::Kind::HexLower &&
828 ExplicitFormat != ExpressionFormat::Kind::HexUpper)
829 return ErrorDiagnostic::get(
830 SM, AlternateFormFlagLoc,
831 "alternate form only supported for hex values");
832
833 FormatExpr = FormatExpr.ltrim(SpaceChars);
834 if (!FormatExpr.empty())
835 return ErrorDiagnostic::get(
836 SM, FormatExpr,
837 "invalid matching format specification in expression");
838 }
839
840 // Save variable definition expression if any.
841 size_t DefEnd = Expr.find(':');
842 if (DefEnd != StringRef::npos) {
843 DefExpr = Expr.substr(0, DefEnd);
844 Expr = Expr.substr(DefEnd + 1);
845 }
846
847 // Parse matching constraint.
848 Expr = Expr.ltrim(SpaceChars);
849 bool HasParsedValidConstraint = false;
850 if (Expr.consume_front("=="))
851 HasParsedValidConstraint = true;
852
853 // Parse the expression itself.
854 Expr = Expr.ltrim(SpaceChars);
855 if (Expr.empty()) {
856 if (HasParsedValidConstraint)
857 return ErrorDiagnostic::get(
858 SM, Expr, "empty numeric expression should not have a constraint");
859 } else {
860 Expr = Expr.rtrim(SpaceChars);
861 StringRef OuterBinOpExpr = Expr;
862 // The first operand in a legacy @LINE expression is always the @LINE
863 // pseudo variable.
864 AllowedOperand AO =
865 IsLegacyLineExpr ? AllowedOperand::LineVar : AllowedOperand::Any;
866 Expected<std::unique_ptr<ExpressionAST>> ParseResult = parseNumericOperand(
867 Expr, AO, !HasParsedValidConstraint, LineNumber, Context, SM);
868 while (ParseResult && !Expr.empty()) {
869 ParseResult = parseBinop(OuterBinOpExpr, Expr, std::move(*ParseResult),
870 IsLegacyLineExpr, LineNumber, Context, SM);
871 // Legacy @LINE expressions only allow 2 operands.
872 if (ParseResult && IsLegacyLineExpr && !Expr.empty())
873 return ErrorDiagnostic::get(
874 SM, Expr,
875 "unexpected characters at end of expression '" + Expr + "'");
876 }
877 if (!ParseResult)
878 return ParseResult.takeError();
879 ExpressionASTPointer = std::move(*ParseResult);
880 }
881
882 // Select format of the expression, i.e. (i) its explicit format, if any,
883 // otherwise (ii) its implicit format, if any, otherwise (iii) the default
884 // format (unsigned). Error out in case of conflicting implicit format
885 // without explicit format.
886 ExpressionFormat Format;
887 if (ExplicitFormat)
888 Format = ExplicitFormat;
889 else if (ExpressionASTPointer) {
890 Expected<ExpressionFormat> ImplicitFormat =
891 ExpressionASTPointer->getImplicitFormat(SM);
892 if (!ImplicitFormat)
893 return ImplicitFormat.takeError();
894 Format = *ImplicitFormat;
895 }
896 if (!Format)
897 Format = ExpressionFormat(ExpressionFormat::Kind::Unsigned, Precision);
898
899 std::unique_ptr<Expression> ExpressionPointer =
900 std::make_unique<Expression>(std::move(ExpressionASTPointer), Format);
901
902 // Parse the numeric variable definition.
903 if (DefEnd != StringRef::npos) {
904 DefExpr = DefExpr.ltrim(SpaceChars);
905 Expected<NumericVariable *> ParseResult = parseNumericVariableDefinition(
906 DefExpr, Context, LineNumber, ExpressionPointer->getFormat(), SM);
907
908 if (!ParseResult)
909 return ParseResult.takeError();
910 DefinedNumericVariable = *ParseResult;
911 }
912
913 return std::move(ExpressionPointer);
914 }
915
parsePattern(StringRef PatternStr,StringRef Prefix,SourceMgr & SM,const FileCheckRequest & Req)916 bool Pattern::parsePattern(StringRef PatternStr, StringRef Prefix,
917 SourceMgr &SM, const FileCheckRequest &Req) {
918 bool MatchFullLinesHere = Req.MatchFullLines && CheckTy != Check::CheckNot;
919 IgnoreCase = Req.IgnoreCase;
920
921 PatternLoc = SMLoc::getFromPointer(PatternStr.data());
922
923 if (!(Req.NoCanonicalizeWhiteSpace && Req.MatchFullLines))
924 // Ignore trailing whitespace.
925 while (!PatternStr.empty() &&
926 (PatternStr.back() == ' ' || PatternStr.back() == '\t'))
927 PatternStr = PatternStr.substr(0, PatternStr.size() - 1);
928
929 // Check that there is something on the line.
930 if (PatternStr.empty() && CheckTy != Check::CheckEmpty) {
931 SM.PrintMessage(PatternLoc, SourceMgr::DK_Error,
932 "found empty check string with prefix '" + Prefix + ":'");
933 return true;
934 }
935
936 if (!PatternStr.empty() && CheckTy == Check::CheckEmpty) {
937 SM.PrintMessage(
938 PatternLoc, SourceMgr::DK_Error,
939 "found non-empty check string for empty check with prefix '" + Prefix +
940 ":'");
941 return true;
942 }
943
944 if (CheckTy == Check::CheckEmpty) {
945 RegExStr = "(\n$)";
946 return false;
947 }
948
949 // If literal check, set fixed string.
950 if (CheckTy.isLiteralMatch()) {
951 FixedStr = PatternStr;
952 return false;
953 }
954
955 // Check to see if this is a fixed string, or if it has regex pieces.
956 if (!MatchFullLinesHere &&
957 (PatternStr.size() < 2 ||
958 (!PatternStr.contains("{{") && !PatternStr.contains("[[")))) {
959 FixedStr = PatternStr;
960 return false;
961 }
962
963 if (MatchFullLinesHere) {
964 RegExStr += '^';
965 if (!Req.NoCanonicalizeWhiteSpace)
966 RegExStr += " *";
967 }
968
969 // Paren value #0 is for the fully matched string. Any new parenthesized
970 // values add from there.
971 unsigned CurParen = 1;
972
973 // Otherwise, there is at least one regex piece. Build up the regex pattern
974 // by escaping scary characters in fixed strings, building up one big regex.
975 while (!PatternStr.empty()) {
976 // RegEx matches.
977 if (PatternStr.startswith("{{")) {
978 // This is the start of a regex match. Scan for the }}.
979 size_t End = PatternStr.find("}}");
980 if (End == StringRef::npos) {
981 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
982 SourceMgr::DK_Error,
983 "found start of regex string with no end '}}'");
984 return true;
985 }
986
987 // Enclose {{}} patterns in parens just like [[]] even though we're not
988 // capturing the result for any purpose. This is required in case the
989 // expression contains an alternation like: CHECK: abc{{x|z}}def. We
990 // want this to turn into: "abc(x|z)def" not "abcx|zdef".
991 RegExStr += '(';
992 ++CurParen;
993
994 if (AddRegExToRegEx(PatternStr.substr(2, End - 2), CurParen, SM))
995 return true;
996 RegExStr += ')';
997
998 PatternStr = PatternStr.substr(End + 2);
999 continue;
1000 }
1001
1002 // String and numeric substitution blocks. Pattern substitution blocks come
1003 // in two forms: [[foo:.*]] and [[foo]]. The former matches .* (or some
1004 // other regex) and assigns it to the string variable 'foo'. The latter
1005 // substitutes foo's value. Numeric substitution blocks recognize the same
1006 // form as string ones, but start with a '#' sign after the double
1007 // brackets. They also accept a combined form which sets a numeric variable
1008 // to the evaluation of an expression. Both string and numeric variable
1009 // names must satisfy the regular expression "[a-zA-Z_][0-9a-zA-Z_]*" to be
1010 // valid, as this helps catch some common errors. If there are extra '['s
1011 // before the "[[", treat them literally.
1012 if (PatternStr.startswith("[[") && !PatternStr.startswith("[[[")) {
1013 StringRef UnparsedPatternStr = PatternStr.substr(2);
1014 // Find the closing bracket pair ending the match. End is going to be an
1015 // offset relative to the beginning of the match string.
1016 size_t End = FindRegexVarEnd(UnparsedPatternStr, SM);
1017 StringRef MatchStr = UnparsedPatternStr.substr(0, End);
1018 bool IsNumBlock = MatchStr.consume_front("#");
1019
1020 if (End == StringRef::npos) {
1021 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
1022 SourceMgr::DK_Error,
1023 "Invalid substitution block, no ]] found");
1024 return true;
1025 }
1026 // Strip the substitution block we are parsing. End points to the start
1027 // of the "]]" closing the expression so account for it in computing the
1028 // index of the first unparsed character.
1029 PatternStr = UnparsedPatternStr.substr(End + 2);
1030
1031 bool IsDefinition = false;
1032 bool SubstNeeded = false;
1033 // Whether the substitution block is a legacy use of @LINE with string
1034 // substitution block syntax.
1035 bool IsLegacyLineExpr = false;
1036 StringRef DefName;
1037 StringRef SubstStr;
1038 StringRef MatchRegexp;
1039 std::string WildcardRegexp;
1040 size_t SubstInsertIdx = RegExStr.size();
1041
1042 // Parse string variable or legacy @LINE expression.
1043 if (!IsNumBlock) {
1044 size_t VarEndIdx = MatchStr.find(':');
1045 size_t SpacePos = MatchStr.substr(0, VarEndIdx).find_first_of(" \t");
1046 if (SpacePos != StringRef::npos) {
1047 SM.PrintMessage(SMLoc::getFromPointer(MatchStr.data() + SpacePos),
1048 SourceMgr::DK_Error, "unexpected whitespace");
1049 return true;
1050 }
1051
1052 // Get the name (e.g. "foo") and verify it is well formed.
1053 StringRef OrigMatchStr = MatchStr;
1054 Expected<Pattern::VariableProperties> ParseVarResult =
1055 parseVariable(MatchStr, SM);
1056 if (!ParseVarResult) {
1057 logAllUnhandledErrors(ParseVarResult.takeError(), errs());
1058 return true;
1059 }
1060 StringRef Name = ParseVarResult->Name;
1061 bool IsPseudo = ParseVarResult->IsPseudo;
1062
1063 IsDefinition = (VarEndIdx != StringRef::npos);
1064 SubstNeeded = !IsDefinition;
1065 if (IsDefinition) {
1066 if ((IsPseudo || !MatchStr.consume_front(":"))) {
1067 SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
1068 SourceMgr::DK_Error,
1069 "invalid name in string variable definition");
1070 return true;
1071 }
1072
1073 // Detect collisions between string and numeric variables when the
1074 // former is created later than the latter.
1075 if (Context->GlobalNumericVariableTable.find(Name) !=
1076 Context->GlobalNumericVariableTable.end()) {
1077 SM.PrintMessage(
1078 SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
1079 "numeric variable with name '" + Name + "' already exists");
1080 return true;
1081 }
1082 DefName = Name;
1083 MatchRegexp = MatchStr;
1084 } else {
1085 if (IsPseudo) {
1086 MatchStr = OrigMatchStr;
1087 IsLegacyLineExpr = IsNumBlock = true;
1088 } else {
1089 if (!MatchStr.empty()) {
1090 SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
1091 SourceMgr::DK_Error,
1092 "invalid name in string variable use");
1093 return true;
1094 }
1095 SubstStr = Name;
1096 }
1097 }
1098 }
1099
1100 // Parse numeric substitution block.
1101 std::unique_ptr<Expression> ExpressionPointer;
1102 std::optional<NumericVariable *> DefinedNumericVariable;
1103 if (IsNumBlock) {
1104 Expected<std::unique_ptr<Expression>> ParseResult =
1105 parseNumericSubstitutionBlock(MatchStr, DefinedNumericVariable,
1106 IsLegacyLineExpr, LineNumber, Context,
1107 SM);
1108 if (!ParseResult) {
1109 logAllUnhandledErrors(ParseResult.takeError(), errs());
1110 return true;
1111 }
1112 ExpressionPointer = std::move(*ParseResult);
1113 SubstNeeded = ExpressionPointer->getAST() != nullptr;
1114 if (DefinedNumericVariable) {
1115 IsDefinition = true;
1116 DefName = (*DefinedNumericVariable)->getName();
1117 }
1118 if (SubstNeeded)
1119 SubstStr = MatchStr;
1120 else {
1121 ExpressionFormat Format = ExpressionPointer->getFormat();
1122 WildcardRegexp = cantFail(Format.getWildcardRegex());
1123 MatchRegexp = WildcardRegexp;
1124 }
1125 }
1126
1127 // Handle variable definition: [[<def>:(...)]] and [[#(...)<def>:(...)]].
1128 if (IsDefinition) {
1129 RegExStr += '(';
1130 ++SubstInsertIdx;
1131
1132 if (IsNumBlock) {
1133 NumericVariableMatch NumericVariableDefinition = {
1134 *DefinedNumericVariable, CurParen};
1135 NumericVariableDefs[DefName] = NumericVariableDefinition;
1136 // This store is done here rather than in match() to allow
1137 // parseNumericVariableUse() to get the pointer to the class instance
1138 // of the right variable definition corresponding to a given numeric
1139 // variable use.
1140 Context->GlobalNumericVariableTable[DefName] =
1141 *DefinedNumericVariable;
1142 } else {
1143 VariableDefs[DefName] = CurParen;
1144 // Mark string variable as defined to detect collisions between
1145 // string and numeric variables in parseNumericVariableUse() and
1146 // defineCmdlineVariables() when the latter is created later than the
1147 // former. We cannot reuse GlobalVariableTable for this by populating
1148 // it with an empty string since we would then lose the ability to
1149 // detect the use of an undefined variable in match().
1150 Context->DefinedVariableTable[DefName] = true;
1151 }
1152
1153 ++CurParen;
1154 }
1155
1156 if (!MatchRegexp.empty() && AddRegExToRegEx(MatchRegexp, CurParen, SM))
1157 return true;
1158
1159 if (IsDefinition)
1160 RegExStr += ')';
1161
1162 // Handle substitutions: [[foo]] and [[#<foo expr>]].
1163 if (SubstNeeded) {
1164 // Handle substitution of string variables that were defined earlier on
1165 // the same line by emitting a backreference. Expressions do not
1166 // support substituting a numeric variable defined on the same line.
1167 if (!IsNumBlock && VariableDefs.find(SubstStr) != VariableDefs.end()) {
1168 unsigned CaptureParenGroup = VariableDefs[SubstStr];
1169 if (CaptureParenGroup < 1 || CaptureParenGroup > 9) {
1170 SM.PrintMessage(SMLoc::getFromPointer(SubstStr.data()),
1171 SourceMgr::DK_Error,
1172 "Can't back-reference more than 9 variables");
1173 return true;
1174 }
1175 AddBackrefToRegEx(CaptureParenGroup);
1176 } else {
1177 // Handle substitution of string variables ([[<var>]]) defined in
1178 // previous CHECK patterns, and substitution of expressions.
1179 Substitution *Substitution =
1180 IsNumBlock
1181 ? Context->makeNumericSubstitution(
1182 SubstStr, std::move(ExpressionPointer), SubstInsertIdx)
1183 : Context->makeStringSubstitution(SubstStr, SubstInsertIdx);
1184 Substitutions.push_back(Substitution);
1185 }
1186 }
1187
1188 continue;
1189 }
1190
1191 // Handle fixed string matches.
1192 // Find the end, which is the start of the next regex.
1193 size_t FixedMatchEnd =
1194 std::min(PatternStr.find("{{", 1), PatternStr.find("[[", 1));
1195 RegExStr += Regex::escape(PatternStr.substr(0, FixedMatchEnd));
1196 PatternStr = PatternStr.substr(FixedMatchEnd);
1197 }
1198
1199 if (MatchFullLinesHere) {
1200 if (!Req.NoCanonicalizeWhiteSpace)
1201 RegExStr += " *";
1202 RegExStr += '$';
1203 }
1204
1205 return false;
1206 }
1207
AddRegExToRegEx(StringRef RS,unsigned & CurParen,SourceMgr & SM)1208 bool Pattern::AddRegExToRegEx(StringRef RS, unsigned &CurParen, SourceMgr &SM) {
1209 Regex R(RS);
1210 std::string Error;
1211 if (!R.isValid(Error)) {
1212 SM.PrintMessage(SMLoc::getFromPointer(RS.data()), SourceMgr::DK_Error,
1213 "invalid regex: " + Error);
1214 return true;
1215 }
1216
1217 RegExStr += RS.str();
1218 CurParen += R.getNumMatches();
1219 return false;
1220 }
1221
AddBackrefToRegEx(unsigned BackrefNum)1222 void Pattern::AddBackrefToRegEx(unsigned BackrefNum) {
1223 assert(BackrefNum >= 1 && BackrefNum <= 9 && "Invalid backref number");
1224 std::string Backref = std::string("\\") + std::string(1, '0' + BackrefNum);
1225 RegExStr += Backref;
1226 }
1227
match(StringRef Buffer,const SourceMgr & SM) const1228 Pattern::MatchResult Pattern::match(StringRef Buffer,
1229 const SourceMgr &SM) const {
1230 // If this is the EOF pattern, match it immediately.
1231 if (CheckTy == Check::CheckEOF)
1232 return MatchResult(Buffer.size(), 0, Error::success());
1233
1234 // If this is a fixed string pattern, just match it now.
1235 if (!FixedStr.empty()) {
1236 size_t Pos =
1237 IgnoreCase ? Buffer.find_insensitive(FixedStr) : Buffer.find(FixedStr);
1238 if (Pos == StringRef::npos)
1239 return make_error<NotFoundError>();
1240 return MatchResult(Pos, /*MatchLen=*/FixedStr.size(), Error::success());
1241 }
1242
1243 // Regex match.
1244
1245 // If there are substitutions, we need to create a temporary string with the
1246 // actual value.
1247 StringRef RegExToMatch = RegExStr;
1248 std::string TmpStr;
1249 if (!Substitutions.empty()) {
1250 TmpStr = RegExStr;
1251 if (LineNumber)
1252 Context->LineVariable->setValue(ExpressionValue(*LineNumber));
1253
1254 size_t InsertOffset = 0;
1255 // Substitute all string variables and expressions whose values are only
1256 // now known. Use of string variables defined on the same line are handled
1257 // by back-references.
1258 Error Errs = Error::success();
1259 for (const auto &Substitution : Substitutions) {
1260 // Substitute and check for failure (e.g. use of undefined variable).
1261 Expected<std::string> Value = Substitution->getResult();
1262 if (!Value) {
1263 // Convert to an ErrorDiagnostic to get location information. This is
1264 // done here rather than printMatch/printNoMatch since now we know which
1265 // substitution block caused the overflow.
1266 Errs = joinErrors(std::move(Errs),
1267 handleErrors(
1268 Value.takeError(),
1269 [&](const OverflowError &E) {
1270 return ErrorDiagnostic::get(
1271 SM, Substitution->getFromString(),
1272 "unable to substitute variable or "
1273 "numeric expression: overflow error");
1274 },
1275 [&SM](const UndefVarError &E) {
1276 return ErrorDiagnostic::get(SM, E.getVarName(),
1277 E.message());
1278 }));
1279 continue;
1280 }
1281
1282 // Plop it into the regex at the adjusted offset.
1283 TmpStr.insert(TmpStr.begin() + Substitution->getIndex() + InsertOffset,
1284 Value->begin(), Value->end());
1285 InsertOffset += Value->size();
1286 }
1287 if (Errs)
1288 return std::move(Errs);
1289
1290 // Match the newly constructed regex.
1291 RegExToMatch = TmpStr;
1292 }
1293
1294 SmallVector<StringRef, 4> MatchInfo;
1295 unsigned int Flags = Regex::Newline;
1296 if (IgnoreCase)
1297 Flags |= Regex::IgnoreCase;
1298 if (!Regex(RegExToMatch, Flags).match(Buffer, &MatchInfo))
1299 return make_error<NotFoundError>();
1300
1301 // Successful regex match.
1302 assert(!MatchInfo.empty() && "Didn't get any match");
1303 StringRef FullMatch = MatchInfo[0];
1304
1305 // If this defines any string variables, remember their values.
1306 for (const auto &VariableDef : VariableDefs) {
1307 assert(VariableDef.second < MatchInfo.size() && "Internal paren error");
1308 Context->GlobalVariableTable[VariableDef.first] =
1309 MatchInfo[VariableDef.second];
1310 }
1311
1312 // Like CHECK-NEXT, CHECK-EMPTY's match range is considered to start after
1313 // the required preceding newline, which is consumed by the pattern in the
1314 // case of CHECK-EMPTY but not CHECK-NEXT.
1315 size_t MatchStartSkip = CheckTy == Check::CheckEmpty;
1316 Match TheMatch;
1317 TheMatch.Pos = FullMatch.data() - Buffer.data() + MatchStartSkip;
1318 TheMatch.Len = FullMatch.size() - MatchStartSkip;
1319
1320 // If this defines any numeric variables, remember their values.
1321 for (const auto &NumericVariableDef : NumericVariableDefs) {
1322 const NumericVariableMatch &NumericVariableMatch =
1323 NumericVariableDef.getValue();
1324 unsigned CaptureParenGroup = NumericVariableMatch.CaptureParenGroup;
1325 assert(CaptureParenGroup < MatchInfo.size() && "Internal paren error");
1326 NumericVariable *DefinedNumericVariable =
1327 NumericVariableMatch.DefinedNumericVariable;
1328
1329 StringRef MatchedValue = MatchInfo[CaptureParenGroup];
1330 ExpressionFormat Format = DefinedNumericVariable->getImplicitFormat();
1331 Expected<ExpressionValue> Value =
1332 Format.valueFromStringRepr(MatchedValue, SM);
1333 if (!Value)
1334 return MatchResult(TheMatch, Value.takeError());
1335 DefinedNumericVariable->setValue(*Value, MatchedValue);
1336 }
1337
1338 return MatchResult(TheMatch, Error::success());
1339 }
1340
computeMatchDistance(StringRef Buffer) const1341 unsigned Pattern::computeMatchDistance(StringRef Buffer) const {
1342 // Just compute the number of matching characters. For regular expressions, we
1343 // just compare against the regex itself and hope for the best.
1344 //
1345 // FIXME: One easy improvement here is have the regex lib generate a single
1346 // example regular expression which matches, and use that as the example
1347 // string.
1348 StringRef ExampleString(FixedStr);
1349 if (ExampleString.empty())
1350 ExampleString = RegExStr;
1351
1352 // Only compare up to the first line in the buffer, or the string size.
1353 StringRef BufferPrefix = Buffer.substr(0, ExampleString.size());
1354 BufferPrefix = BufferPrefix.split('\n').first;
1355 return BufferPrefix.edit_distance(ExampleString);
1356 }
1357
printSubstitutions(const SourceMgr & SM,StringRef Buffer,SMRange Range,FileCheckDiag::MatchType MatchTy,std::vector<FileCheckDiag> * Diags) const1358 void Pattern::printSubstitutions(const SourceMgr &SM, StringRef Buffer,
1359 SMRange Range,
1360 FileCheckDiag::MatchType MatchTy,
1361 std::vector<FileCheckDiag> *Diags) const {
1362 // Print what we know about substitutions.
1363 if (!Substitutions.empty()) {
1364 for (const auto &Substitution : Substitutions) {
1365 SmallString<256> Msg;
1366 raw_svector_ostream OS(Msg);
1367
1368 Expected<std::string> MatchedValue = Substitution->getResult();
1369 // Substitution failures are handled in printNoMatch().
1370 if (!MatchedValue) {
1371 consumeError(MatchedValue.takeError());
1372 continue;
1373 }
1374
1375 OS << "with \"";
1376 OS.write_escaped(Substitution->getFromString()) << "\" equal to \"";
1377 OS.write_escaped(*MatchedValue) << "\"";
1378
1379 // We report only the start of the match/search range to suggest we are
1380 // reporting the substitutions as set at the start of the match/search.
1381 // Indicating a non-zero-length range might instead seem to imply that the
1382 // substitution matches or was captured from exactly that range.
1383 if (Diags)
1384 Diags->emplace_back(SM, CheckTy, getLoc(), MatchTy,
1385 SMRange(Range.Start, Range.Start), OS.str());
1386 else
1387 SM.PrintMessage(Range.Start, SourceMgr::DK_Note, OS.str());
1388 }
1389 }
1390 }
1391
printVariableDefs(const SourceMgr & SM,FileCheckDiag::MatchType MatchTy,std::vector<FileCheckDiag> * Diags) const1392 void Pattern::printVariableDefs(const SourceMgr &SM,
1393 FileCheckDiag::MatchType MatchTy,
1394 std::vector<FileCheckDiag> *Diags) const {
1395 if (VariableDefs.empty() && NumericVariableDefs.empty())
1396 return;
1397 // Build list of variable captures.
1398 struct VarCapture {
1399 StringRef Name;
1400 SMRange Range;
1401 };
1402 SmallVector<VarCapture, 2> VarCaptures;
1403 for (const auto &VariableDef : VariableDefs) {
1404 VarCapture VC;
1405 VC.Name = VariableDef.first;
1406 StringRef Value = Context->GlobalVariableTable[VC.Name];
1407 SMLoc Start = SMLoc::getFromPointer(Value.data());
1408 SMLoc End = SMLoc::getFromPointer(Value.data() + Value.size());
1409 VC.Range = SMRange(Start, End);
1410 VarCaptures.push_back(VC);
1411 }
1412 for (const auto &VariableDef : NumericVariableDefs) {
1413 VarCapture VC;
1414 VC.Name = VariableDef.getKey();
1415 std::optional<StringRef> StrValue =
1416 VariableDef.getValue().DefinedNumericVariable->getStringValue();
1417 if (!StrValue)
1418 continue;
1419 SMLoc Start = SMLoc::getFromPointer(StrValue->data());
1420 SMLoc End = SMLoc::getFromPointer(StrValue->data() + StrValue->size());
1421 VC.Range = SMRange(Start, End);
1422 VarCaptures.push_back(VC);
1423 }
1424 // Sort variable captures by the order in which they matched the input.
1425 // Ranges shouldn't be overlapping, so we can just compare the start.
1426 llvm::sort(VarCaptures, [](const VarCapture &A, const VarCapture &B) {
1427 if (&A == &B)
1428 return false;
1429 assert(A.Range.Start != B.Range.Start &&
1430 "unexpected overlapping variable captures");
1431 return A.Range.Start.getPointer() < B.Range.Start.getPointer();
1432 });
1433 // Create notes for the sorted captures.
1434 for (const VarCapture &VC : VarCaptures) {
1435 SmallString<256> Msg;
1436 raw_svector_ostream OS(Msg);
1437 OS << "captured var \"" << VC.Name << "\"";
1438 if (Diags)
1439 Diags->emplace_back(SM, CheckTy, getLoc(), MatchTy, VC.Range, OS.str());
1440 else
1441 SM.PrintMessage(VC.Range.Start, SourceMgr::DK_Note, OS.str(), VC.Range);
1442 }
1443 }
1444
ProcessMatchResult(FileCheckDiag::MatchType MatchTy,const SourceMgr & SM,SMLoc Loc,Check::FileCheckType CheckTy,StringRef Buffer,size_t Pos,size_t Len,std::vector<FileCheckDiag> * Diags,bool AdjustPrevDiags=false)1445 static SMRange ProcessMatchResult(FileCheckDiag::MatchType MatchTy,
1446 const SourceMgr &SM, SMLoc Loc,
1447 Check::FileCheckType CheckTy,
1448 StringRef Buffer, size_t Pos, size_t Len,
1449 std::vector<FileCheckDiag> *Diags,
1450 bool AdjustPrevDiags = false) {
1451 SMLoc Start = SMLoc::getFromPointer(Buffer.data() + Pos);
1452 SMLoc End = SMLoc::getFromPointer(Buffer.data() + Pos + Len);
1453 SMRange Range(Start, End);
1454 if (Diags) {
1455 if (AdjustPrevDiags) {
1456 SMLoc CheckLoc = Diags->rbegin()->CheckLoc;
1457 for (auto I = Diags->rbegin(), E = Diags->rend();
1458 I != E && I->CheckLoc == CheckLoc; ++I)
1459 I->MatchTy = MatchTy;
1460 } else
1461 Diags->emplace_back(SM, CheckTy, Loc, MatchTy, Range);
1462 }
1463 return Range;
1464 }
1465
printFuzzyMatch(const SourceMgr & SM,StringRef Buffer,std::vector<FileCheckDiag> * Diags) const1466 void Pattern::printFuzzyMatch(const SourceMgr &SM, StringRef Buffer,
1467 std::vector<FileCheckDiag> *Diags) const {
1468 // Attempt to find the closest/best fuzzy match. Usually an error happens
1469 // because some string in the output didn't exactly match. In these cases, we
1470 // would like to show the user a best guess at what "should have" matched, to
1471 // save them having to actually check the input manually.
1472 size_t NumLinesForward = 0;
1473 size_t Best = StringRef::npos;
1474 double BestQuality = 0;
1475
1476 // Use an arbitrary 4k limit on how far we will search.
1477 for (size_t i = 0, e = std::min(size_t(4096), Buffer.size()); i != e; ++i) {
1478 if (Buffer[i] == '\n')
1479 ++NumLinesForward;
1480
1481 // Patterns have leading whitespace stripped, so skip whitespace when
1482 // looking for something which looks like a pattern.
1483 if (Buffer[i] == ' ' || Buffer[i] == '\t')
1484 continue;
1485
1486 // Compute the "quality" of this match as an arbitrary combination of the
1487 // match distance and the number of lines skipped to get to this match.
1488 unsigned Distance = computeMatchDistance(Buffer.substr(i));
1489 double Quality = Distance + (NumLinesForward / 100.);
1490
1491 if (Quality < BestQuality || Best == StringRef::npos) {
1492 Best = i;
1493 BestQuality = Quality;
1494 }
1495 }
1496
1497 // Print the "possible intended match here" line if we found something
1498 // reasonable and not equal to what we showed in the "scanning from here"
1499 // line.
1500 if (Best && Best != StringRef::npos && BestQuality < 50) {
1501 SMRange MatchRange =
1502 ProcessMatchResult(FileCheckDiag::MatchFuzzy, SM, getLoc(),
1503 getCheckTy(), Buffer, Best, 0, Diags);
1504 SM.PrintMessage(MatchRange.Start, SourceMgr::DK_Note,
1505 "possible intended match here");
1506
1507 // FIXME: If we wanted to be really friendly we would show why the match
1508 // failed, as it can be hard to spot simple one character differences.
1509 }
1510 }
1511
1512 Expected<StringRef>
getPatternVarValue(StringRef VarName)1513 FileCheckPatternContext::getPatternVarValue(StringRef VarName) {
1514 auto VarIter = GlobalVariableTable.find(VarName);
1515 if (VarIter == GlobalVariableTable.end())
1516 return make_error<UndefVarError>(VarName);
1517
1518 return VarIter->second;
1519 }
1520
1521 template <class... Types>
makeNumericVariable(Types...args)1522 NumericVariable *FileCheckPatternContext::makeNumericVariable(Types... args) {
1523 NumericVariables.push_back(std::make_unique<NumericVariable>(args...));
1524 return NumericVariables.back().get();
1525 }
1526
1527 Substitution *
makeStringSubstitution(StringRef VarName,size_t InsertIdx)1528 FileCheckPatternContext::makeStringSubstitution(StringRef VarName,
1529 size_t InsertIdx) {
1530 Substitutions.push_back(
1531 std::make_unique<StringSubstitution>(this, VarName, InsertIdx));
1532 return Substitutions.back().get();
1533 }
1534
makeNumericSubstitution(StringRef ExpressionStr,std::unique_ptr<Expression> Expression,size_t InsertIdx)1535 Substitution *FileCheckPatternContext::makeNumericSubstitution(
1536 StringRef ExpressionStr, std::unique_ptr<Expression> Expression,
1537 size_t InsertIdx) {
1538 Substitutions.push_back(std::make_unique<NumericSubstitution>(
1539 this, ExpressionStr, std::move(Expression), InsertIdx));
1540 return Substitutions.back().get();
1541 }
1542
FindRegexVarEnd(StringRef Str,SourceMgr & SM)1543 size_t Pattern::FindRegexVarEnd(StringRef Str, SourceMgr &SM) {
1544 // Offset keeps track of the current offset within the input Str
1545 size_t Offset = 0;
1546 // [...] Nesting depth
1547 size_t BracketDepth = 0;
1548
1549 while (!Str.empty()) {
1550 if (Str.startswith("]]") && BracketDepth == 0)
1551 return Offset;
1552 if (Str[0] == '\\') {
1553 // Backslash escapes the next char within regexes, so skip them both.
1554 Str = Str.substr(2);
1555 Offset += 2;
1556 } else {
1557 switch (Str[0]) {
1558 default:
1559 break;
1560 case '[':
1561 BracketDepth++;
1562 break;
1563 case ']':
1564 if (BracketDepth == 0) {
1565 SM.PrintMessage(SMLoc::getFromPointer(Str.data()),
1566 SourceMgr::DK_Error,
1567 "missing closing \"]\" for regex variable");
1568 exit(1);
1569 }
1570 BracketDepth--;
1571 break;
1572 }
1573 Str = Str.substr(1);
1574 Offset++;
1575 }
1576 }
1577
1578 return StringRef::npos;
1579 }
1580
CanonicalizeFile(MemoryBuffer & MB,SmallVectorImpl<char> & OutputBuffer)1581 StringRef FileCheck::CanonicalizeFile(MemoryBuffer &MB,
1582 SmallVectorImpl<char> &OutputBuffer) {
1583 OutputBuffer.reserve(MB.getBufferSize());
1584
1585 for (const char *Ptr = MB.getBufferStart(), *End = MB.getBufferEnd();
1586 Ptr != End; ++Ptr) {
1587 // Eliminate trailing dosish \r.
1588 if (Ptr <= End - 2 && Ptr[0] == '\r' && Ptr[1] == '\n') {
1589 continue;
1590 }
1591
1592 // If current char is not a horizontal whitespace or if horizontal
1593 // whitespace canonicalization is disabled, dump it to output as is.
1594 if (Req.NoCanonicalizeWhiteSpace || (*Ptr != ' ' && *Ptr != '\t')) {
1595 OutputBuffer.push_back(*Ptr);
1596 continue;
1597 }
1598
1599 // Otherwise, add one space and advance over neighboring space.
1600 OutputBuffer.push_back(' ');
1601 while (Ptr + 1 != End && (Ptr[1] == ' ' || Ptr[1] == '\t'))
1602 ++Ptr;
1603 }
1604
1605 // Add a null byte and then return all but that byte.
1606 OutputBuffer.push_back('\0');
1607 return StringRef(OutputBuffer.data(), OutputBuffer.size() - 1);
1608 }
1609
FileCheckDiag(const SourceMgr & SM,const Check::FileCheckType & CheckTy,SMLoc CheckLoc,MatchType MatchTy,SMRange InputRange,StringRef Note)1610 FileCheckDiag::FileCheckDiag(const SourceMgr &SM,
1611 const Check::FileCheckType &CheckTy,
1612 SMLoc CheckLoc, MatchType MatchTy,
1613 SMRange InputRange, StringRef Note)
1614 : CheckTy(CheckTy), CheckLoc(CheckLoc), MatchTy(MatchTy), Note(Note) {
1615 auto Start = SM.getLineAndColumn(InputRange.Start);
1616 auto End = SM.getLineAndColumn(InputRange.End);
1617 InputStartLine = Start.first;
1618 InputStartCol = Start.second;
1619 InputEndLine = End.first;
1620 InputEndCol = End.second;
1621 }
1622
IsPartOfWord(char c)1623 static bool IsPartOfWord(char c) {
1624 return (isAlnum(c) || c == '-' || c == '_');
1625 }
1626
setCount(int C)1627 Check::FileCheckType &Check::FileCheckType::setCount(int C) {
1628 assert(Count > 0 && "zero and negative counts are not supported");
1629 assert((C == 1 || Kind == CheckPlain) &&
1630 "count supported only for plain CHECK directives");
1631 Count = C;
1632 return *this;
1633 }
1634
getModifiersDescription() const1635 std::string Check::FileCheckType::getModifiersDescription() const {
1636 if (Modifiers.none())
1637 return "";
1638 std::string Ret;
1639 raw_string_ostream OS(Ret);
1640 OS << '{';
1641 if (isLiteralMatch())
1642 OS << "LITERAL";
1643 OS << '}';
1644 return OS.str();
1645 }
1646
getDescription(StringRef Prefix) const1647 std::string Check::FileCheckType::getDescription(StringRef Prefix) const {
1648 // Append directive modifiers.
1649 auto WithModifiers = [this, Prefix](StringRef Str) -> std::string {
1650 return (Prefix + Str + getModifiersDescription()).str();
1651 };
1652
1653 switch (Kind) {
1654 case Check::CheckNone:
1655 return "invalid";
1656 case Check::CheckMisspelled:
1657 return "misspelled";
1658 case Check::CheckPlain:
1659 if (Count > 1)
1660 return WithModifiers("-COUNT");
1661 return WithModifiers("");
1662 case Check::CheckNext:
1663 return WithModifiers("-NEXT");
1664 case Check::CheckSame:
1665 return WithModifiers("-SAME");
1666 case Check::CheckNot:
1667 return WithModifiers("-NOT");
1668 case Check::CheckDAG:
1669 return WithModifiers("-DAG");
1670 case Check::CheckLabel:
1671 return WithModifiers("-LABEL");
1672 case Check::CheckEmpty:
1673 return WithModifiers("-EMPTY");
1674 case Check::CheckComment:
1675 return std::string(Prefix);
1676 case Check::CheckEOF:
1677 return "implicit EOF";
1678 case Check::CheckBadNot:
1679 return "bad NOT";
1680 case Check::CheckBadCount:
1681 return "bad COUNT";
1682 }
1683 llvm_unreachable("unknown FileCheckType");
1684 }
1685
1686 static std::pair<Check::FileCheckType, StringRef>
FindCheckType(const FileCheckRequest & Req,StringRef Buffer,StringRef Prefix,bool & Misspelled)1687 FindCheckType(const FileCheckRequest &Req, StringRef Buffer, StringRef Prefix,
1688 bool &Misspelled) {
1689 if (Buffer.size() <= Prefix.size())
1690 return {Check::CheckNone, StringRef()};
1691
1692 StringRef Rest = Buffer.drop_front(Prefix.size());
1693 // Check for comment.
1694 if (llvm::is_contained(Req.CommentPrefixes, Prefix)) {
1695 if (Rest.consume_front(":"))
1696 return {Check::CheckComment, Rest};
1697 // Ignore a comment prefix if it has a suffix like "-NOT".
1698 return {Check::CheckNone, StringRef()};
1699 }
1700
1701 auto ConsumeModifiers = [&](Check::FileCheckType Ret)
1702 -> std::pair<Check::FileCheckType, StringRef> {
1703 if (Rest.consume_front(":"))
1704 return {Ret, Rest};
1705 if (!Rest.consume_front("{"))
1706 return {Check::CheckNone, StringRef()};
1707
1708 // Parse the modifiers, speparated by commas.
1709 do {
1710 // Allow whitespace in modifiers list.
1711 Rest = Rest.ltrim();
1712 if (Rest.consume_front("LITERAL"))
1713 Ret.setLiteralMatch();
1714 else
1715 return {Check::CheckNone, Rest};
1716 // Allow whitespace in modifiers list.
1717 Rest = Rest.ltrim();
1718 } while (Rest.consume_front(","));
1719 if (!Rest.consume_front("}:"))
1720 return {Check::CheckNone, Rest};
1721 return {Ret, Rest};
1722 };
1723
1724 // Verify that the prefix is followed by directive modifiers or a colon.
1725 if (Rest.consume_front(":"))
1726 return {Check::CheckPlain, Rest};
1727 if (Rest.front() == '{')
1728 return ConsumeModifiers(Check::CheckPlain);
1729
1730 if (Rest.consume_front("_"))
1731 Misspelled = true;
1732 else if (!Rest.consume_front("-"))
1733 return {Check::CheckNone, StringRef()};
1734
1735 if (Rest.consume_front("COUNT-")) {
1736 int64_t Count;
1737 if (Rest.consumeInteger(10, Count))
1738 // Error happened in parsing integer.
1739 return {Check::CheckBadCount, Rest};
1740 if (Count <= 0 || Count > INT32_MAX)
1741 return {Check::CheckBadCount, Rest};
1742 if (Rest.front() != ':' && Rest.front() != '{')
1743 return {Check::CheckBadCount, Rest};
1744 return ConsumeModifiers(
1745 Check::FileCheckType(Check::CheckPlain).setCount(Count));
1746 }
1747
1748 // You can't combine -NOT with another suffix.
1749 if (Rest.startswith("DAG-NOT:") || Rest.startswith("NOT-DAG:") ||
1750 Rest.startswith("NEXT-NOT:") || Rest.startswith("NOT-NEXT:") ||
1751 Rest.startswith("SAME-NOT:") || Rest.startswith("NOT-SAME:") ||
1752 Rest.startswith("EMPTY-NOT:") || Rest.startswith("NOT-EMPTY:"))
1753 return {Check::CheckBadNot, Rest};
1754
1755 if (Rest.consume_front("NEXT"))
1756 return ConsumeModifiers(Check::CheckNext);
1757
1758 if (Rest.consume_front("SAME"))
1759 return ConsumeModifiers(Check::CheckSame);
1760
1761 if (Rest.consume_front("NOT"))
1762 return ConsumeModifiers(Check::CheckNot);
1763
1764 if (Rest.consume_front("DAG"))
1765 return ConsumeModifiers(Check::CheckDAG);
1766
1767 if (Rest.consume_front("LABEL"))
1768 return ConsumeModifiers(Check::CheckLabel);
1769
1770 if (Rest.consume_front("EMPTY"))
1771 return ConsumeModifiers(Check::CheckEmpty);
1772
1773 return {Check::CheckNone, Rest};
1774 }
1775
1776 static std::pair<Check::FileCheckType, StringRef>
FindCheckType(const FileCheckRequest & Req,StringRef Buffer,StringRef Prefix)1777 FindCheckType(const FileCheckRequest &Req, StringRef Buffer, StringRef Prefix) {
1778 bool Misspelled = false;
1779 auto Res = FindCheckType(Req, Buffer, Prefix, Misspelled);
1780 if (Res.first != Check::CheckNone && Misspelled)
1781 return {Check::CheckMisspelled, Res.second};
1782 return Res;
1783 }
1784
1785 // From the given position, find the next character after the word.
SkipWord(StringRef Str,size_t Loc)1786 static size_t SkipWord(StringRef Str, size_t Loc) {
1787 while (Loc < Str.size() && IsPartOfWord(Str[Loc]))
1788 ++Loc;
1789 return Loc;
1790 }
1791
1792 /// Searches the buffer for the first prefix in the prefix regular expression.
1793 ///
1794 /// This searches the buffer using the provided regular expression, however it
1795 /// enforces constraints beyond that:
1796 /// 1) The found prefix must not be a suffix of something that looks like
1797 /// a valid prefix.
1798 /// 2) The found prefix must be followed by a valid check type suffix using \c
1799 /// FindCheckType above.
1800 ///
1801 /// \returns a pair of StringRefs into the Buffer, which combines:
1802 /// - the first match of the regular expression to satisfy these two is
1803 /// returned,
1804 /// otherwise an empty StringRef is returned to indicate failure.
1805 /// - buffer rewound to the location right after parsed suffix, for parsing
1806 /// to continue from
1807 ///
1808 /// If this routine returns a valid prefix, it will also shrink \p Buffer to
1809 /// start at the beginning of the returned prefix, increment \p LineNumber for
1810 /// each new line consumed from \p Buffer, and set \p CheckTy to the type of
1811 /// check found by examining the suffix.
1812 ///
1813 /// If no valid prefix is found, the state of Buffer, LineNumber, and CheckTy
1814 /// is unspecified.
1815 static std::pair<StringRef, StringRef>
FindFirstMatchingPrefix(const FileCheckRequest & Req,Regex & PrefixRE,StringRef & Buffer,unsigned & LineNumber,Check::FileCheckType & CheckTy)1816 FindFirstMatchingPrefix(const FileCheckRequest &Req, Regex &PrefixRE,
1817 StringRef &Buffer, unsigned &LineNumber,
1818 Check::FileCheckType &CheckTy) {
1819 SmallVector<StringRef, 2> Matches;
1820
1821 while (!Buffer.empty()) {
1822 // Find the first (longest) match using the RE.
1823 if (!PrefixRE.match(Buffer, &Matches))
1824 // No match at all, bail.
1825 return {StringRef(), StringRef()};
1826
1827 StringRef Prefix = Matches[0];
1828 Matches.clear();
1829
1830 assert(Prefix.data() >= Buffer.data() &&
1831 Prefix.data() < Buffer.data() + Buffer.size() &&
1832 "Prefix doesn't start inside of buffer!");
1833 size_t Loc = Prefix.data() - Buffer.data();
1834 StringRef Skipped = Buffer.substr(0, Loc);
1835 Buffer = Buffer.drop_front(Loc);
1836 LineNumber += Skipped.count('\n');
1837
1838 // Check that the matched prefix isn't a suffix of some other check-like
1839 // word.
1840 // FIXME: This is a very ad-hoc check. it would be better handled in some
1841 // other way. Among other things it seems hard to distinguish between
1842 // intentional and unintentional uses of this feature.
1843 if (Skipped.empty() || !IsPartOfWord(Skipped.back())) {
1844 // Now extract the type.
1845 StringRef AfterSuffix;
1846 std::tie(CheckTy, AfterSuffix) = FindCheckType(Req, Buffer, Prefix);
1847
1848 // If we've found a valid check type for this prefix, we're done.
1849 if (CheckTy != Check::CheckNone)
1850 return {Prefix, AfterSuffix};
1851 }
1852
1853 // If we didn't successfully find a prefix, we need to skip this invalid
1854 // prefix and continue scanning. We directly skip the prefix that was
1855 // matched and any additional parts of that check-like word.
1856 Buffer = Buffer.drop_front(SkipWord(Buffer, Prefix.size()));
1857 }
1858
1859 // We ran out of buffer while skipping partial matches so give up.
1860 return {StringRef(), StringRef()};
1861 }
1862
createLineVariable()1863 void FileCheckPatternContext::createLineVariable() {
1864 assert(!LineVariable && "@LINE pseudo numeric variable already created");
1865 StringRef LineName = "@LINE";
1866 LineVariable = makeNumericVariable(
1867 LineName, ExpressionFormat(ExpressionFormat::Kind::Unsigned));
1868 GlobalNumericVariableTable[LineName] = LineVariable;
1869 }
1870
FileCheck(FileCheckRequest Req)1871 FileCheck::FileCheck(FileCheckRequest Req)
1872 : Req(Req), PatternContext(std::make_unique<FileCheckPatternContext>()),
1873 CheckStrings(std::make_unique<std::vector<FileCheckString>>()) {}
1874
1875 FileCheck::~FileCheck() = default;
1876
readCheckFile(SourceMgr & SM,StringRef Buffer,Regex & PrefixRE,std::pair<unsigned,unsigned> * ImpPatBufferIDRange)1877 bool FileCheck::readCheckFile(
1878 SourceMgr &SM, StringRef Buffer, Regex &PrefixRE,
1879 std::pair<unsigned, unsigned> *ImpPatBufferIDRange) {
1880 if (ImpPatBufferIDRange)
1881 ImpPatBufferIDRange->first = ImpPatBufferIDRange->second = 0;
1882
1883 Error DefineError =
1884 PatternContext->defineCmdlineVariables(Req.GlobalDefines, SM);
1885 if (DefineError) {
1886 logAllUnhandledErrors(std::move(DefineError), errs());
1887 return true;
1888 }
1889
1890 PatternContext->createLineVariable();
1891
1892 std::vector<Pattern> ImplicitNegativeChecks;
1893 for (StringRef PatternString : Req.ImplicitCheckNot) {
1894 // Create a buffer with fake command line content in order to display the
1895 // command line option responsible for the specific implicit CHECK-NOT.
1896 std::string Prefix = "-implicit-check-not='";
1897 std::string Suffix = "'";
1898 std::unique_ptr<MemoryBuffer> CmdLine = MemoryBuffer::getMemBufferCopy(
1899 (Prefix + PatternString + Suffix).str(), "command line");
1900
1901 StringRef PatternInBuffer =
1902 CmdLine->getBuffer().substr(Prefix.size(), PatternString.size());
1903 unsigned BufferID = SM.AddNewSourceBuffer(std::move(CmdLine), SMLoc());
1904 if (ImpPatBufferIDRange) {
1905 if (ImpPatBufferIDRange->first == ImpPatBufferIDRange->second) {
1906 ImpPatBufferIDRange->first = BufferID;
1907 ImpPatBufferIDRange->second = BufferID + 1;
1908 } else {
1909 assert(BufferID == ImpPatBufferIDRange->second &&
1910 "expected consecutive source buffer IDs");
1911 ++ImpPatBufferIDRange->second;
1912 }
1913 }
1914
1915 ImplicitNegativeChecks.push_back(
1916 Pattern(Check::CheckNot, PatternContext.get()));
1917 ImplicitNegativeChecks.back().parsePattern(PatternInBuffer,
1918 "IMPLICIT-CHECK", SM, Req);
1919 }
1920
1921 std::vector<Pattern> DagNotMatches = ImplicitNegativeChecks;
1922
1923 // LineNumber keeps track of the line on which CheckPrefix instances are
1924 // found.
1925 unsigned LineNumber = 1;
1926
1927 std::set<StringRef> PrefixesNotFound(Req.CheckPrefixes.begin(),
1928 Req.CheckPrefixes.end());
1929 const size_t DistinctPrefixes = PrefixesNotFound.size();
1930 while (true) {
1931 Check::FileCheckType CheckTy;
1932
1933 // See if a prefix occurs in the memory buffer.
1934 StringRef UsedPrefix;
1935 StringRef AfterSuffix;
1936 std::tie(UsedPrefix, AfterSuffix) =
1937 FindFirstMatchingPrefix(Req, PrefixRE, Buffer, LineNumber, CheckTy);
1938 if (UsedPrefix.empty())
1939 break;
1940 if (CheckTy != Check::CheckComment)
1941 PrefixesNotFound.erase(UsedPrefix);
1942
1943 assert(UsedPrefix.data() == Buffer.data() &&
1944 "Failed to move Buffer's start forward, or pointed prefix outside "
1945 "of the buffer!");
1946 assert(AfterSuffix.data() >= Buffer.data() &&
1947 AfterSuffix.data() < Buffer.data() + Buffer.size() &&
1948 "Parsing after suffix doesn't start inside of buffer!");
1949
1950 // Location to use for error messages.
1951 const char *UsedPrefixStart = UsedPrefix.data();
1952
1953 // Skip the buffer to the end of parsed suffix (or just prefix, if no good
1954 // suffix was processed).
1955 Buffer = AfterSuffix.empty() ? Buffer.drop_front(UsedPrefix.size())
1956 : AfterSuffix;
1957
1958 // Complain about misspelled directives.
1959 if (CheckTy == Check::CheckMisspelled) {
1960 StringRef UsedDirective(UsedPrefix.data(),
1961 AfterSuffix.data() - UsedPrefix.data());
1962 SM.PrintMessage(SMLoc::getFromPointer(UsedDirective.data()),
1963 SourceMgr::DK_Error,
1964 "misspelled directive '" + UsedDirective + "'");
1965 return true;
1966 }
1967
1968 // Complain about useful-looking but unsupported suffixes.
1969 if (CheckTy == Check::CheckBadNot) {
1970 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Error,
1971 "unsupported -NOT combo on prefix '" + UsedPrefix + "'");
1972 return true;
1973 }
1974
1975 // Complain about invalid count specification.
1976 if (CheckTy == Check::CheckBadCount) {
1977 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Error,
1978 "invalid count in -COUNT specification on prefix '" +
1979 UsedPrefix + "'");
1980 return true;
1981 }
1982
1983 // Okay, we found the prefix, yay. Remember the rest of the line, but ignore
1984 // leading whitespace.
1985 if (!(Req.NoCanonicalizeWhiteSpace && Req.MatchFullLines))
1986 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t"));
1987
1988 // Scan ahead to the end of line.
1989 size_t EOL = Buffer.find_first_of("\n\r");
1990
1991 // Remember the location of the start of the pattern, for diagnostics.
1992 SMLoc PatternLoc = SMLoc::getFromPointer(Buffer.data());
1993
1994 // Extract the pattern from the buffer.
1995 StringRef PatternBuffer = Buffer.substr(0, EOL);
1996 Buffer = Buffer.substr(EOL);
1997
1998 // If this is a comment, we're done.
1999 if (CheckTy == Check::CheckComment)
2000 continue;
2001
2002 // Parse the pattern.
2003 Pattern P(CheckTy, PatternContext.get(), LineNumber);
2004 if (P.parsePattern(PatternBuffer, UsedPrefix, SM, Req))
2005 return true;
2006
2007 // Verify that CHECK-LABEL lines do not define or use variables
2008 if ((CheckTy == Check::CheckLabel) && P.hasVariable()) {
2009 SM.PrintMessage(
2010 SMLoc::getFromPointer(UsedPrefixStart), SourceMgr::DK_Error,
2011 "found '" + UsedPrefix + "-LABEL:'"
2012 " with variable definition or use");
2013 return true;
2014 }
2015
2016 // Verify that CHECK-NEXT/SAME/EMPTY lines have at least one CHECK line before them.
2017 if ((CheckTy == Check::CheckNext || CheckTy == Check::CheckSame ||
2018 CheckTy == Check::CheckEmpty) &&
2019 CheckStrings->empty()) {
2020 StringRef Type = CheckTy == Check::CheckNext
2021 ? "NEXT"
2022 : CheckTy == Check::CheckEmpty ? "EMPTY" : "SAME";
2023 SM.PrintMessage(SMLoc::getFromPointer(UsedPrefixStart),
2024 SourceMgr::DK_Error,
2025 "found '" + UsedPrefix + "-" + Type +
2026 "' without previous '" + UsedPrefix + ": line");
2027 return true;
2028 }
2029
2030 // Handle CHECK-DAG/-NOT.
2031 if (CheckTy == Check::CheckDAG || CheckTy == Check::CheckNot) {
2032 DagNotMatches.push_back(P);
2033 continue;
2034 }
2035
2036 // Okay, add the string we captured to the output vector and move on.
2037 CheckStrings->emplace_back(P, UsedPrefix, PatternLoc);
2038 std::swap(DagNotMatches, CheckStrings->back().DagNotStrings);
2039 DagNotMatches = ImplicitNegativeChecks;
2040 }
2041
2042 // When there are no used prefixes we report an error except in the case that
2043 // no prefix is specified explicitly but -implicit-check-not is specified.
2044 const bool NoPrefixesFound = PrefixesNotFound.size() == DistinctPrefixes;
2045 const bool SomePrefixesUnexpectedlyNotUsed =
2046 !Req.AllowUnusedPrefixes && !PrefixesNotFound.empty();
2047 if ((NoPrefixesFound || SomePrefixesUnexpectedlyNotUsed) &&
2048 (ImplicitNegativeChecks.empty() || !Req.IsDefaultCheckPrefix)) {
2049 errs() << "error: no check strings found with prefix"
2050 << (PrefixesNotFound.size() > 1 ? "es " : " ");
2051 bool First = true;
2052 for (StringRef MissingPrefix : PrefixesNotFound) {
2053 if (!First)
2054 errs() << ", ";
2055 errs() << "\'" << MissingPrefix << ":'";
2056 First = false;
2057 }
2058 errs() << '\n';
2059 return true;
2060 }
2061
2062 // Add an EOF pattern for any trailing --implicit-check-not/CHECK-DAG/-NOTs,
2063 // and use the first prefix as a filler for the error message.
2064 if (!DagNotMatches.empty()) {
2065 CheckStrings->emplace_back(
2066 Pattern(Check::CheckEOF, PatternContext.get(), LineNumber + 1),
2067 *Req.CheckPrefixes.begin(), SMLoc::getFromPointer(Buffer.data()));
2068 std::swap(DagNotMatches, CheckStrings->back().DagNotStrings);
2069 }
2070
2071 return false;
2072 }
2073
2074 /// Returns either (1) \c ErrorSuccess if there was no error or (2)
2075 /// \c ErrorReported if an error was reported, such as an unexpected match.
printMatch(bool ExpectedMatch,const SourceMgr & SM,StringRef Prefix,SMLoc Loc,const Pattern & Pat,int MatchedCount,StringRef Buffer,Pattern::MatchResult MatchResult,const FileCheckRequest & Req,std::vector<FileCheckDiag> * Diags)2076 static Error printMatch(bool ExpectedMatch, const SourceMgr &SM,
2077 StringRef Prefix, SMLoc Loc, const Pattern &Pat,
2078 int MatchedCount, StringRef Buffer,
2079 Pattern::MatchResult MatchResult,
2080 const FileCheckRequest &Req,
2081 std::vector<FileCheckDiag> *Diags) {
2082 // Suppress some verbosity if there's no error.
2083 bool HasError = !ExpectedMatch || MatchResult.TheError;
2084 bool PrintDiag = true;
2085 if (!HasError) {
2086 if (!Req.Verbose)
2087 return ErrorReported::reportedOrSuccess(HasError);
2088 if (!Req.VerboseVerbose && Pat.getCheckTy() == Check::CheckEOF)
2089 return ErrorReported::reportedOrSuccess(HasError);
2090 // Due to their verbosity, we don't print verbose diagnostics here if we're
2091 // gathering them for Diags to be rendered elsewhere, but we always print
2092 // other diagnostics.
2093 PrintDiag = !Diags;
2094 }
2095
2096 // Add "found" diagnostic, substitutions, and variable definitions to Diags.
2097 FileCheckDiag::MatchType MatchTy = ExpectedMatch
2098 ? FileCheckDiag::MatchFoundAndExpected
2099 : FileCheckDiag::MatchFoundButExcluded;
2100 SMRange MatchRange = ProcessMatchResult(MatchTy, SM, Loc, Pat.getCheckTy(),
2101 Buffer, MatchResult.TheMatch->Pos,
2102 MatchResult.TheMatch->Len, Diags);
2103 if (Diags) {
2104 Pat.printSubstitutions(SM, Buffer, MatchRange, MatchTy, Diags);
2105 Pat.printVariableDefs(SM, MatchTy, Diags);
2106 }
2107 if (!PrintDiag) {
2108 assert(!HasError && "expected to report more diagnostics for error");
2109 return ErrorReported::reportedOrSuccess(HasError);
2110 }
2111
2112 // Print the match.
2113 std::string Message = formatv("{0}: {1} string found in input",
2114 Pat.getCheckTy().getDescription(Prefix),
2115 (ExpectedMatch ? "expected" : "excluded"))
2116 .str();
2117 if (Pat.getCount() > 1)
2118 Message += formatv(" ({0} out of {1})", MatchedCount, Pat.getCount()).str();
2119 SM.PrintMessage(
2120 Loc, ExpectedMatch ? SourceMgr::DK_Remark : SourceMgr::DK_Error, Message);
2121 SM.PrintMessage(MatchRange.Start, SourceMgr::DK_Note, "found here",
2122 {MatchRange});
2123
2124 // Print additional information, which can be useful even if there are errors.
2125 Pat.printSubstitutions(SM, Buffer, MatchRange, MatchTy, nullptr);
2126 Pat.printVariableDefs(SM, MatchTy, nullptr);
2127
2128 // Print errors and add them to Diags. We report these errors after the match
2129 // itself because we found them after the match. If we had found them before
2130 // the match, we'd be in printNoMatch.
2131 handleAllErrors(std::move(MatchResult.TheError),
2132 [&](const ErrorDiagnostic &E) {
2133 E.log(errs());
2134 if (Diags) {
2135 Diags->emplace_back(SM, Pat.getCheckTy(), Loc,
2136 FileCheckDiag::MatchFoundErrorNote,
2137 E.getRange(), E.getMessage().str());
2138 }
2139 });
2140 return ErrorReported::reportedOrSuccess(HasError);
2141 }
2142
2143 /// Returns either (1) \c ErrorSuccess if there was no error, or (2)
2144 /// \c ErrorReported if an error was reported, such as an expected match not
2145 /// found.
printNoMatch(bool ExpectedMatch,const SourceMgr & SM,StringRef Prefix,SMLoc Loc,const Pattern & Pat,int MatchedCount,StringRef Buffer,Error MatchError,bool VerboseVerbose,std::vector<FileCheckDiag> * Diags)2146 static Error printNoMatch(bool ExpectedMatch, const SourceMgr &SM,
2147 StringRef Prefix, SMLoc Loc, const Pattern &Pat,
2148 int MatchedCount, StringRef Buffer, Error MatchError,
2149 bool VerboseVerbose,
2150 std::vector<FileCheckDiag> *Diags) {
2151 // Print any pattern errors, and record them to be added to Diags later.
2152 bool HasError = ExpectedMatch;
2153 bool HasPatternError = false;
2154 FileCheckDiag::MatchType MatchTy = ExpectedMatch
2155 ? FileCheckDiag::MatchNoneButExpected
2156 : FileCheckDiag::MatchNoneAndExcluded;
2157 SmallVector<std::string, 4> ErrorMsgs;
2158 handleAllErrors(
2159 std::move(MatchError),
2160 [&](const ErrorDiagnostic &E) {
2161 HasError = HasPatternError = true;
2162 MatchTy = FileCheckDiag::MatchNoneForInvalidPattern;
2163 E.log(errs());
2164 if (Diags)
2165 ErrorMsgs.push_back(E.getMessage().str());
2166 },
2167 // NotFoundError is why printNoMatch was invoked.
2168 [](const NotFoundError &E) {});
2169
2170 // Suppress some verbosity if there's no error.
2171 bool PrintDiag = true;
2172 if (!HasError) {
2173 if (!VerboseVerbose)
2174 return ErrorReported::reportedOrSuccess(HasError);
2175 // Due to their verbosity, we don't print verbose diagnostics here if we're
2176 // gathering them for Diags to be rendered elsewhere, but we always print
2177 // other diagnostics.
2178 PrintDiag = !Diags;
2179 }
2180
2181 // Add "not found" diagnostic, substitutions, and pattern errors to Diags.
2182 //
2183 // We handle Diags a little differently than the errors we print directly:
2184 // we add the "not found" diagnostic to Diags even if there are pattern
2185 // errors. The reason is that we need to attach pattern errors as notes
2186 // somewhere in the input, and the input search range from the "not found"
2187 // diagnostic is all we have to anchor them.
2188 SMRange SearchRange = ProcessMatchResult(MatchTy, SM, Loc, Pat.getCheckTy(),
2189 Buffer, 0, Buffer.size(), Diags);
2190 if (Diags) {
2191 SMRange NoteRange = SMRange(SearchRange.Start, SearchRange.Start);
2192 for (StringRef ErrorMsg : ErrorMsgs)
2193 Diags->emplace_back(SM, Pat.getCheckTy(), Loc, MatchTy, NoteRange,
2194 ErrorMsg);
2195 Pat.printSubstitutions(SM, Buffer, SearchRange, MatchTy, Diags);
2196 }
2197 if (!PrintDiag) {
2198 assert(!HasError && "expected to report more diagnostics for error");
2199 return ErrorReported::reportedOrSuccess(HasError);
2200 }
2201
2202 // Print "not found" diagnostic, except that's implied if we already printed a
2203 // pattern error.
2204 if (!HasPatternError) {
2205 std::string Message = formatv("{0}: {1} string not found in input",
2206 Pat.getCheckTy().getDescription(Prefix),
2207 (ExpectedMatch ? "expected" : "excluded"))
2208 .str();
2209 if (Pat.getCount() > 1)
2210 Message +=
2211 formatv(" ({0} out of {1})", MatchedCount, Pat.getCount()).str();
2212 SM.PrintMessage(Loc,
2213 ExpectedMatch ? SourceMgr::DK_Error : SourceMgr::DK_Remark,
2214 Message);
2215 SM.PrintMessage(SearchRange.Start, SourceMgr::DK_Note,
2216 "scanning from here");
2217 }
2218
2219 // Print additional information, which can be useful even after a pattern
2220 // error.
2221 Pat.printSubstitutions(SM, Buffer, SearchRange, MatchTy, nullptr);
2222 if (ExpectedMatch)
2223 Pat.printFuzzyMatch(SM, Buffer, Diags);
2224 return ErrorReported::reportedOrSuccess(HasError);
2225 }
2226
2227 /// Returns either (1) \c ErrorSuccess if there was no error, or (2)
2228 /// \c ErrorReported if an error was reported.
reportMatchResult(bool ExpectedMatch,const SourceMgr & SM,StringRef Prefix,SMLoc Loc,const Pattern & Pat,int MatchedCount,StringRef Buffer,Pattern::MatchResult MatchResult,const FileCheckRequest & Req,std::vector<FileCheckDiag> * Diags)2229 static Error reportMatchResult(bool ExpectedMatch, const SourceMgr &SM,
2230 StringRef Prefix, SMLoc Loc, const Pattern &Pat,
2231 int MatchedCount, StringRef Buffer,
2232 Pattern::MatchResult MatchResult,
2233 const FileCheckRequest &Req,
2234 std::vector<FileCheckDiag> *Diags) {
2235 if (MatchResult.TheMatch)
2236 return printMatch(ExpectedMatch, SM, Prefix, Loc, Pat, MatchedCount, Buffer,
2237 std::move(MatchResult), Req, Diags);
2238 return printNoMatch(ExpectedMatch, SM, Prefix, Loc, Pat, MatchedCount, Buffer,
2239 std::move(MatchResult.TheError), Req.VerboseVerbose,
2240 Diags);
2241 }
2242
2243 /// Counts the number of newlines in the specified range.
CountNumNewlinesBetween(StringRef Range,const char * & FirstNewLine)2244 static unsigned CountNumNewlinesBetween(StringRef Range,
2245 const char *&FirstNewLine) {
2246 unsigned NumNewLines = 0;
2247 while (true) {
2248 // Scan for newline.
2249 Range = Range.substr(Range.find_first_of("\n\r"));
2250 if (Range.empty())
2251 return NumNewLines;
2252
2253 ++NumNewLines;
2254
2255 // Handle \n\r and \r\n as a single newline.
2256 if (Range.size() > 1 && (Range[1] == '\n' || Range[1] == '\r') &&
2257 (Range[0] != Range[1]))
2258 Range = Range.substr(1);
2259 Range = Range.substr(1);
2260
2261 if (NumNewLines == 1)
2262 FirstNewLine = Range.begin();
2263 }
2264 }
2265
Check(const SourceMgr & SM,StringRef Buffer,bool IsLabelScanMode,size_t & MatchLen,FileCheckRequest & Req,std::vector<FileCheckDiag> * Diags) const2266 size_t FileCheckString::Check(const SourceMgr &SM, StringRef Buffer,
2267 bool IsLabelScanMode, size_t &MatchLen,
2268 FileCheckRequest &Req,
2269 std::vector<FileCheckDiag> *Diags) const {
2270 size_t LastPos = 0;
2271 std::vector<const Pattern *> NotStrings;
2272
2273 // IsLabelScanMode is true when we are scanning forward to find CHECK-LABEL
2274 // bounds; we have not processed variable definitions within the bounded block
2275 // yet so cannot handle any final CHECK-DAG yet; this is handled when going
2276 // over the block again (including the last CHECK-LABEL) in normal mode.
2277 if (!IsLabelScanMode) {
2278 // Match "dag strings" (with mixed "not strings" if any).
2279 LastPos = CheckDag(SM, Buffer, NotStrings, Req, Diags);
2280 if (LastPos == StringRef::npos)
2281 return StringRef::npos;
2282 }
2283
2284 // Match itself from the last position after matching CHECK-DAG.
2285 size_t LastMatchEnd = LastPos;
2286 size_t FirstMatchPos = 0;
2287 // Go match the pattern Count times. Majority of patterns only match with
2288 // count 1 though.
2289 assert(Pat.getCount() != 0 && "pattern count can not be zero");
2290 for (int i = 1; i <= Pat.getCount(); i++) {
2291 StringRef MatchBuffer = Buffer.substr(LastMatchEnd);
2292 // get a match at current start point
2293 Pattern::MatchResult MatchResult = Pat.match(MatchBuffer, SM);
2294
2295 // report
2296 if (Error Err = reportMatchResult(/*ExpectedMatch=*/true, SM, Prefix, Loc,
2297 Pat, i, MatchBuffer,
2298 std::move(MatchResult), Req, Diags)) {
2299 cantFail(handleErrors(std::move(Err), [&](const ErrorReported &E) {}));
2300 return StringRef::npos;
2301 }
2302
2303 size_t MatchPos = MatchResult.TheMatch->Pos;
2304 if (i == 1)
2305 FirstMatchPos = LastPos + MatchPos;
2306
2307 // move start point after the match
2308 LastMatchEnd += MatchPos + MatchResult.TheMatch->Len;
2309 }
2310 // Full match len counts from first match pos.
2311 MatchLen = LastMatchEnd - FirstMatchPos;
2312
2313 // Similar to the above, in "label-scan mode" we can't yet handle CHECK-NEXT
2314 // or CHECK-NOT
2315 if (!IsLabelScanMode) {
2316 size_t MatchPos = FirstMatchPos - LastPos;
2317 StringRef MatchBuffer = Buffer.substr(LastPos);
2318 StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
2319
2320 // If this check is a "CHECK-NEXT", verify that the previous match was on
2321 // the previous line (i.e. that there is one newline between them).
2322 if (CheckNext(SM, SkippedRegion)) {
2323 ProcessMatchResult(FileCheckDiag::MatchFoundButWrongLine, SM, Loc,
2324 Pat.getCheckTy(), MatchBuffer, MatchPos, MatchLen,
2325 Diags, Req.Verbose);
2326 return StringRef::npos;
2327 }
2328
2329 // If this check is a "CHECK-SAME", verify that the previous match was on
2330 // the same line (i.e. that there is no newline between them).
2331 if (CheckSame(SM, SkippedRegion)) {
2332 ProcessMatchResult(FileCheckDiag::MatchFoundButWrongLine, SM, Loc,
2333 Pat.getCheckTy(), MatchBuffer, MatchPos, MatchLen,
2334 Diags, Req.Verbose);
2335 return StringRef::npos;
2336 }
2337
2338 // If this match had "not strings", verify that they don't exist in the
2339 // skipped region.
2340 if (CheckNot(SM, SkippedRegion, NotStrings, Req, Diags))
2341 return StringRef::npos;
2342 }
2343
2344 return FirstMatchPos;
2345 }
2346
CheckNext(const SourceMgr & SM,StringRef Buffer) const2347 bool FileCheckString::CheckNext(const SourceMgr &SM, StringRef Buffer) const {
2348 if (Pat.getCheckTy() != Check::CheckNext &&
2349 Pat.getCheckTy() != Check::CheckEmpty)
2350 return false;
2351
2352 Twine CheckName =
2353 Prefix +
2354 Twine(Pat.getCheckTy() == Check::CheckEmpty ? "-EMPTY" : "-NEXT");
2355
2356 // Count the number of newlines between the previous match and this one.
2357 const char *FirstNewLine = nullptr;
2358 unsigned NumNewLines = CountNumNewlinesBetween(Buffer, FirstNewLine);
2359
2360 if (NumNewLines == 0) {
2361 SM.PrintMessage(Loc, SourceMgr::DK_Error,
2362 CheckName + ": is on the same line as previous match");
2363 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()), SourceMgr::DK_Note,
2364 "'next' match was here");
2365 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
2366 "previous match ended here");
2367 return true;
2368 }
2369
2370 if (NumNewLines != 1) {
2371 SM.PrintMessage(Loc, SourceMgr::DK_Error,
2372 CheckName +
2373 ": is not on the line after the previous match");
2374 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()), SourceMgr::DK_Note,
2375 "'next' match was here");
2376 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
2377 "previous match ended here");
2378 SM.PrintMessage(SMLoc::getFromPointer(FirstNewLine), SourceMgr::DK_Note,
2379 "non-matching line after previous match is here");
2380 return true;
2381 }
2382
2383 return false;
2384 }
2385
CheckSame(const SourceMgr & SM,StringRef Buffer) const2386 bool FileCheckString::CheckSame(const SourceMgr &SM, StringRef Buffer) const {
2387 if (Pat.getCheckTy() != Check::CheckSame)
2388 return false;
2389
2390 // Count the number of newlines between the previous match and this one.
2391 const char *FirstNewLine = nullptr;
2392 unsigned NumNewLines = CountNumNewlinesBetween(Buffer, FirstNewLine);
2393
2394 if (NumNewLines != 0) {
2395 SM.PrintMessage(Loc, SourceMgr::DK_Error,
2396 Prefix +
2397 "-SAME: is not on the same line as the previous match");
2398 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()), SourceMgr::DK_Note,
2399 "'next' match was here");
2400 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
2401 "previous match ended here");
2402 return true;
2403 }
2404
2405 return false;
2406 }
2407
CheckNot(const SourceMgr & SM,StringRef Buffer,const std::vector<const Pattern * > & NotStrings,const FileCheckRequest & Req,std::vector<FileCheckDiag> * Diags) const2408 bool FileCheckString::CheckNot(const SourceMgr &SM, StringRef Buffer,
2409 const std::vector<const Pattern *> &NotStrings,
2410 const FileCheckRequest &Req,
2411 std::vector<FileCheckDiag> *Diags) const {
2412 bool DirectiveFail = false;
2413 for (const Pattern *Pat : NotStrings) {
2414 assert((Pat->getCheckTy() == Check::CheckNot) && "Expect CHECK-NOT!");
2415 Pattern::MatchResult MatchResult = Pat->match(Buffer, SM);
2416 if (Error Err = reportMatchResult(/*ExpectedMatch=*/false, SM, Prefix,
2417 Pat->getLoc(), *Pat, 1, Buffer,
2418 std::move(MatchResult), Req, Diags)) {
2419 cantFail(handleErrors(std::move(Err), [&](const ErrorReported &E) {}));
2420 DirectiveFail = true;
2421 continue;
2422 }
2423 }
2424 return DirectiveFail;
2425 }
2426
CheckDag(const SourceMgr & SM,StringRef Buffer,std::vector<const Pattern * > & NotStrings,const FileCheckRequest & Req,std::vector<FileCheckDiag> * Diags) const2427 size_t FileCheckString::CheckDag(const SourceMgr &SM, StringRef Buffer,
2428 std::vector<const Pattern *> &NotStrings,
2429 const FileCheckRequest &Req,
2430 std::vector<FileCheckDiag> *Diags) const {
2431 if (DagNotStrings.empty())
2432 return 0;
2433
2434 // The start of the search range.
2435 size_t StartPos = 0;
2436
2437 struct MatchRange {
2438 size_t Pos;
2439 size_t End;
2440 };
2441 // A sorted list of ranges for non-overlapping CHECK-DAG matches. Match
2442 // ranges are erased from this list once they are no longer in the search
2443 // range.
2444 std::list<MatchRange> MatchRanges;
2445
2446 // We need PatItr and PatEnd later for detecting the end of a CHECK-DAG
2447 // group, so we don't use a range-based for loop here.
2448 for (auto PatItr = DagNotStrings.begin(), PatEnd = DagNotStrings.end();
2449 PatItr != PatEnd; ++PatItr) {
2450 const Pattern &Pat = *PatItr;
2451 assert((Pat.getCheckTy() == Check::CheckDAG ||
2452 Pat.getCheckTy() == Check::CheckNot) &&
2453 "Invalid CHECK-DAG or CHECK-NOT!");
2454
2455 if (Pat.getCheckTy() == Check::CheckNot) {
2456 NotStrings.push_back(&Pat);
2457 continue;
2458 }
2459
2460 assert((Pat.getCheckTy() == Check::CheckDAG) && "Expect CHECK-DAG!");
2461
2462 // CHECK-DAG always matches from the start.
2463 size_t MatchLen = 0, MatchPos = StartPos;
2464
2465 // Search for a match that doesn't overlap a previous match in this
2466 // CHECK-DAG group.
2467 for (auto MI = MatchRanges.begin(), ME = MatchRanges.end(); true; ++MI) {
2468 StringRef MatchBuffer = Buffer.substr(MatchPos);
2469 Pattern::MatchResult MatchResult = Pat.match(MatchBuffer, SM);
2470 // With a group of CHECK-DAGs, a single mismatching means the match on
2471 // that group of CHECK-DAGs fails immediately.
2472 if (MatchResult.TheError || Req.VerboseVerbose) {
2473 if (Error Err = reportMatchResult(/*ExpectedMatch=*/true, SM, Prefix,
2474 Pat.getLoc(), Pat, 1, MatchBuffer,
2475 std::move(MatchResult), Req, Diags)) {
2476 cantFail(
2477 handleErrors(std::move(Err), [&](const ErrorReported &E) {}));
2478 return StringRef::npos;
2479 }
2480 }
2481 MatchLen = MatchResult.TheMatch->Len;
2482 // Re-calc it as the offset relative to the start of the original
2483 // string.
2484 MatchPos += MatchResult.TheMatch->Pos;
2485 MatchRange M{MatchPos, MatchPos + MatchLen};
2486 if (Req.AllowDeprecatedDagOverlap) {
2487 // We don't need to track all matches in this mode, so we just maintain
2488 // one match range that encompasses the current CHECK-DAG group's
2489 // matches.
2490 if (MatchRanges.empty())
2491 MatchRanges.insert(MatchRanges.end(), M);
2492 else {
2493 auto Block = MatchRanges.begin();
2494 Block->Pos = std::min(Block->Pos, M.Pos);
2495 Block->End = std::max(Block->End, M.End);
2496 }
2497 break;
2498 }
2499 // Iterate previous matches until overlapping match or insertion point.
2500 bool Overlap = false;
2501 for (; MI != ME; ++MI) {
2502 if (M.Pos < MI->End) {
2503 // !Overlap => New match has no overlap and is before this old match.
2504 // Overlap => New match overlaps this old match.
2505 Overlap = MI->Pos < M.End;
2506 break;
2507 }
2508 }
2509 if (!Overlap) {
2510 // Insert non-overlapping match into list.
2511 MatchRanges.insert(MI, M);
2512 break;
2513 }
2514 if (Req.VerboseVerbose) {
2515 // Due to their verbosity, we don't print verbose diagnostics here if
2516 // we're gathering them for a different rendering, but we always print
2517 // other diagnostics.
2518 if (!Diags) {
2519 SMLoc OldStart = SMLoc::getFromPointer(Buffer.data() + MI->Pos);
2520 SMLoc OldEnd = SMLoc::getFromPointer(Buffer.data() + MI->End);
2521 SMRange OldRange(OldStart, OldEnd);
2522 SM.PrintMessage(OldStart, SourceMgr::DK_Note,
2523 "match discarded, overlaps earlier DAG match here",
2524 {OldRange});
2525 } else {
2526 SMLoc CheckLoc = Diags->rbegin()->CheckLoc;
2527 for (auto I = Diags->rbegin(), E = Diags->rend();
2528 I != E && I->CheckLoc == CheckLoc; ++I)
2529 I->MatchTy = FileCheckDiag::MatchFoundButDiscarded;
2530 }
2531 }
2532 MatchPos = MI->End;
2533 }
2534 if (!Req.VerboseVerbose)
2535 cantFail(printMatch(
2536 /*ExpectedMatch=*/true, SM, Prefix, Pat.getLoc(), Pat, 1, Buffer,
2537 Pattern::MatchResult(MatchPos, MatchLen, Error::success()), Req,
2538 Diags));
2539
2540 // Handle the end of a CHECK-DAG group.
2541 if (std::next(PatItr) == PatEnd ||
2542 std::next(PatItr)->getCheckTy() == Check::CheckNot) {
2543 if (!NotStrings.empty()) {
2544 // If there are CHECK-NOTs between two CHECK-DAGs or from CHECK to
2545 // CHECK-DAG, verify that there are no 'not' strings occurred in that
2546 // region.
2547 StringRef SkippedRegion =
2548 Buffer.slice(StartPos, MatchRanges.begin()->Pos);
2549 if (CheckNot(SM, SkippedRegion, NotStrings, Req, Diags))
2550 return StringRef::npos;
2551 // Clear "not strings".
2552 NotStrings.clear();
2553 }
2554 // All subsequent CHECK-DAGs and CHECK-NOTs should be matched from the
2555 // end of this CHECK-DAG group's match range.
2556 StartPos = MatchRanges.rbegin()->End;
2557 // Don't waste time checking for (impossible) overlaps before that.
2558 MatchRanges.clear();
2559 }
2560 }
2561
2562 return StartPos;
2563 }
2564
ValidatePrefixes(StringRef Kind,StringSet<> & UniquePrefixes,ArrayRef<StringRef> SuppliedPrefixes)2565 static bool ValidatePrefixes(StringRef Kind, StringSet<> &UniquePrefixes,
2566 ArrayRef<StringRef> SuppliedPrefixes) {
2567 for (StringRef Prefix : SuppliedPrefixes) {
2568 if (Prefix.empty()) {
2569 errs() << "error: supplied " << Kind << " prefix must not be the empty "
2570 << "string\n";
2571 return false;
2572 }
2573 static const Regex Validator("^[a-zA-Z0-9_-]*$");
2574 if (!Validator.match(Prefix)) {
2575 errs() << "error: supplied " << Kind << " prefix must start with a "
2576 << "letter and contain only alphanumeric characters, hyphens, and "
2577 << "underscores: '" << Prefix << "'\n";
2578 return false;
2579 }
2580 if (!UniquePrefixes.insert(Prefix).second) {
2581 errs() << "error: supplied " << Kind << " prefix must be unique among "
2582 << "check and comment prefixes: '" << Prefix << "'\n";
2583 return false;
2584 }
2585 }
2586 return true;
2587 }
2588
2589 static const char *DefaultCheckPrefixes[] = {"CHECK"};
2590 static const char *DefaultCommentPrefixes[] = {"COM", "RUN"};
2591
ValidateCheckPrefixes()2592 bool FileCheck::ValidateCheckPrefixes() {
2593 StringSet<> UniquePrefixes;
2594 // Add default prefixes to catch user-supplied duplicates of them below.
2595 if (Req.CheckPrefixes.empty()) {
2596 for (const char *Prefix : DefaultCheckPrefixes)
2597 UniquePrefixes.insert(Prefix);
2598 }
2599 if (Req.CommentPrefixes.empty()) {
2600 for (const char *Prefix : DefaultCommentPrefixes)
2601 UniquePrefixes.insert(Prefix);
2602 }
2603 // Do not validate the default prefixes, or diagnostics about duplicates might
2604 // incorrectly indicate that they were supplied by the user.
2605 if (!ValidatePrefixes("check", UniquePrefixes, Req.CheckPrefixes))
2606 return false;
2607 if (!ValidatePrefixes("comment", UniquePrefixes, Req.CommentPrefixes))
2608 return false;
2609 return true;
2610 }
2611
buildCheckPrefixRegex()2612 Regex FileCheck::buildCheckPrefixRegex() {
2613 if (Req.CheckPrefixes.empty()) {
2614 for (const char *Prefix : DefaultCheckPrefixes)
2615 Req.CheckPrefixes.push_back(Prefix);
2616 Req.IsDefaultCheckPrefix = true;
2617 }
2618 if (Req.CommentPrefixes.empty()) {
2619 for (const char *Prefix : DefaultCommentPrefixes)
2620 Req.CommentPrefixes.push_back(Prefix);
2621 }
2622
2623 // We already validated the contents of CheckPrefixes and CommentPrefixes so
2624 // just concatenate them as alternatives.
2625 SmallString<32> PrefixRegexStr;
2626 for (size_t I = 0, E = Req.CheckPrefixes.size(); I != E; ++I) {
2627 if (I != 0)
2628 PrefixRegexStr.push_back('|');
2629 PrefixRegexStr.append(Req.CheckPrefixes[I]);
2630 }
2631 for (StringRef Prefix : Req.CommentPrefixes) {
2632 PrefixRegexStr.push_back('|');
2633 PrefixRegexStr.append(Prefix);
2634 }
2635
2636 return Regex(PrefixRegexStr);
2637 }
2638
defineCmdlineVariables(ArrayRef<StringRef> CmdlineDefines,SourceMgr & SM)2639 Error FileCheckPatternContext::defineCmdlineVariables(
2640 ArrayRef<StringRef> CmdlineDefines, SourceMgr &SM) {
2641 assert(GlobalVariableTable.empty() && GlobalNumericVariableTable.empty() &&
2642 "Overriding defined variable with command-line variable definitions");
2643
2644 if (CmdlineDefines.empty())
2645 return Error::success();
2646
2647 // Create a string representing the vector of command-line definitions. Each
2648 // definition is on its own line and prefixed with a definition number to
2649 // clarify which definition a given diagnostic corresponds to.
2650 unsigned I = 0;
2651 Error Errs = Error::success();
2652 std::string CmdlineDefsDiag;
2653 SmallVector<std::pair<size_t, size_t>, 4> CmdlineDefsIndices;
2654 for (StringRef CmdlineDef : CmdlineDefines) {
2655 std::string DefPrefix = ("Global define #" + Twine(++I) + ": ").str();
2656 size_t EqIdx = CmdlineDef.find('=');
2657 if (EqIdx == StringRef::npos) {
2658 CmdlineDefsIndices.push_back(std::make_pair(CmdlineDefsDiag.size(), 0));
2659 continue;
2660 }
2661 // Numeric variable definition.
2662 if (CmdlineDef[0] == '#') {
2663 // Append a copy of the command-line definition adapted to use the same
2664 // format as in the input file to be able to reuse
2665 // parseNumericSubstitutionBlock.
2666 CmdlineDefsDiag += (DefPrefix + CmdlineDef + " (parsed as: [[").str();
2667 std::string SubstitutionStr = std::string(CmdlineDef);
2668 SubstitutionStr[EqIdx] = ':';
2669 CmdlineDefsIndices.push_back(
2670 std::make_pair(CmdlineDefsDiag.size(), SubstitutionStr.size()));
2671 CmdlineDefsDiag += (SubstitutionStr + Twine("]])\n")).str();
2672 } else {
2673 CmdlineDefsDiag += DefPrefix;
2674 CmdlineDefsIndices.push_back(
2675 std::make_pair(CmdlineDefsDiag.size(), CmdlineDef.size()));
2676 CmdlineDefsDiag += (CmdlineDef + "\n").str();
2677 }
2678 }
2679
2680 // Create a buffer with fake command line content in order to display
2681 // parsing diagnostic with location information and point to the
2682 // global definition with invalid syntax.
2683 std::unique_ptr<MemoryBuffer> CmdLineDefsDiagBuffer =
2684 MemoryBuffer::getMemBufferCopy(CmdlineDefsDiag, "Global defines");
2685 StringRef CmdlineDefsDiagRef = CmdLineDefsDiagBuffer->getBuffer();
2686 SM.AddNewSourceBuffer(std::move(CmdLineDefsDiagBuffer), SMLoc());
2687
2688 for (std::pair<size_t, size_t> CmdlineDefIndices : CmdlineDefsIndices) {
2689 StringRef CmdlineDef = CmdlineDefsDiagRef.substr(CmdlineDefIndices.first,
2690 CmdlineDefIndices.second);
2691 if (CmdlineDef.empty()) {
2692 Errs = joinErrors(
2693 std::move(Errs),
2694 ErrorDiagnostic::get(SM, CmdlineDef,
2695 "missing equal sign in global definition"));
2696 continue;
2697 }
2698
2699 // Numeric variable definition.
2700 if (CmdlineDef[0] == '#') {
2701 // Now parse the definition both to check that the syntax is correct and
2702 // to create the necessary class instance.
2703 StringRef CmdlineDefExpr = CmdlineDef.substr(1);
2704 std::optional<NumericVariable *> DefinedNumericVariable;
2705 Expected<std::unique_ptr<Expression>> ExpressionResult =
2706 Pattern::parseNumericSubstitutionBlock(CmdlineDefExpr,
2707 DefinedNumericVariable, false,
2708 std::nullopt, this, SM);
2709 if (!ExpressionResult) {
2710 Errs = joinErrors(std::move(Errs), ExpressionResult.takeError());
2711 continue;
2712 }
2713 std::unique_ptr<Expression> Expression = std::move(*ExpressionResult);
2714 // Now evaluate the expression whose value this variable should be set
2715 // to, since the expression of a command-line variable definition should
2716 // only use variables defined earlier on the command-line. If not, this
2717 // is an error and we report it.
2718 Expected<ExpressionValue> Value = Expression->getAST()->eval();
2719 if (!Value) {
2720 Errs = joinErrors(std::move(Errs), Value.takeError());
2721 continue;
2722 }
2723
2724 assert(DefinedNumericVariable && "No variable defined");
2725 (*DefinedNumericVariable)->setValue(*Value);
2726
2727 // Record this variable definition.
2728 GlobalNumericVariableTable[(*DefinedNumericVariable)->getName()] =
2729 *DefinedNumericVariable;
2730 } else {
2731 // String variable definition.
2732 std::pair<StringRef, StringRef> CmdlineNameVal = CmdlineDef.split('=');
2733 StringRef CmdlineName = CmdlineNameVal.first;
2734 StringRef OrigCmdlineName = CmdlineName;
2735 Expected<Pattern::VariableProperties> ParseVarResult =
2736 Pattern::parseVariable(CmdlineName, SM);
2737 if (!ParseVarResult) {
2738 Errs = joinErrors(std::move(Errs), ParseVarResult.takeError());
2739 continue;
2740 }
2741 // Check that CmdlineName does not denote a pseudo variable is only
2742 // composed of the parsed numeric variable. This catches cases like
2743 // "FOO+2" in a "FOO+2=10" definition.
2744 if (ParseVarResult->IsPseudo || !CmdlineName.empty()) {
2745 Errs = joinErrors(std::move(Errs),
2746 ErrorDiagnostic::get(
2747 SM, OrigCmdlineName,
2748 "invalid name in string variable definition '" +
2749 OrigCmdlineName + "'"));
2750 continue;
2751 }
2752 StringRef Name = ParseVarResult->Name;
2753
2754 // Detect collisions between string and numeric variables when the former
2755 // is created later than the latter.
2756 if (GlobalNumericVariableTable.find(Name) !=
2757 GlobalNumericVariableTable.end()) {
2758 Errs = joinErrors(std::move(Errs),
2759 ErrorDiagnostic::get(SM, Name,
2760 "numeric variable with name '" +
2761 Name + "' already exists"));
2762 continue;
2763 }
2764 GlobalVariableTable.insert(CmdlineNameVal);
2765 // Mark the string variable as defined to detect collisions between
2766 // string and numeric variables in defineCmdlineVariables when the latter
2767 // is created later than the former. We cannot reuse GlobalVariableTable
2768 // for this by populating it with an empty string since we would then
2769 // lose the ability to detect the use of an undefined variable in
2770 // match().
2771 DefinedVariableTable[Name] = true;
2772 }
2773 }
2774
2775 return Errs;
2776 }
2777
clearLocalVars()2778 void FileCheckPatternContext::clearLocalVars() {
2779 SmallVector<StringRef, 16> LocalPatternVars, LocalNumericVars;
2780 for (const StringMapEntry<StringRef> &Var : GlobalVariableTable)
2781 if (Var.first()[0] != '$')
2782 LocalPatternVars.push_back(Var.first());
2783
2784 // Numeric substitution reads the value of a variable directly, not via
2785 // GlobalNumericVariableTable. Therefore, we clear local variables by
2786 // clearing their value which will lead to a numeric substitution failure. We
2787 // also mark the variable for removal from GlobalNumericVariableTable since
2788 // this is what defineCmdlineVariables checks to decide that no global
2789 // variable has been defined.
2790 for (const auto &Var : GlobalNumericVariableTable)
2791 if (Var.first()[0] != '$') {
2792 Var.getValue()->clearValue();
2793 LocalNumericVars.push_back(Var.first());
2794 }
2795
2796 for (const auto &Var : LocalPatternVars)
2797 GlobalVariableTable.erase(Var);
2798 for (const auto &Var : LocalNumericVars)
2799 GlobalNumericVariableTable.erase(Var);
2800 }
2801
checkInput(SourceMgr & SM,StringRef Buffer,std::vector<FileCheckDiag> * Diags)2802 bool FileCheck::checkInput(SourceMgr &SM, StringRef Buffer,
2803 std::vector<FileCheckDiag> *Diags) {
2804 bool ChecksFailed = false;
2805
2806 unsigned i = 0, j = 0, e = CheckStrings->size();
2807 while (true) {
2808 StringRef CheckRegion;
2809 if (j == e) {
2810 CheckRegion = Buffer;
2811 } else {
2812 const FileCheckString &CheckLabelStr = (*CheckStrings)[j];
2813 if (CheckLabelStr.Pat.getCheckTy() != Check::CheckLabel) {
2814 ++j;
2815 continue;
2816 }
2817
2818 // Scan to next CHECK-LABEL match, ignoring CHECK-NOT and CHECK-DAG
2819 size_t MatchLabelLen = 0;
2820 size_t MatchLabelPos =
2821 CheckLabelStr.Check(SM, Buffer, true, MatchLabelLen, Req, Diags);
2822 if (MatchLabelPos == StringRef::npos)
2823 // Immediately bail if CHECK-LABEL fails, nothing else we can do.
2824 return false;
2825
2826 CheckRegion = Buffer.substr(0, MatchLabelPos + MatchLabelLen);
2827 Buffer = Buffer.substr(MatchLabelPos + MatchLabelLen);
2828 ++j;
2829 }
2830
2831 // Do not clear the first region as it's the one before the first
2832 // CHECK-LABEL and it would clear variables defined on the command-line
2833 // before they get used.
2834 if (i != 0 && Req.EnableVarScope)
2835 PatternContext->clearLocalVars();
2836
2837 for (; i != j; ++i) {
2838 const FileCheckString &CheckStr = (*CheckStrings)[i];
2839
2840 // Check each string within the scanned region, including a second check
2841 // of any final CHECK-LABEL (to verify CHECK-NOT and CHECK-DAG)
2842 size_t MatchLen = 0;
2843 size_t MatchPos =
2844 CheckStr.Check(SM, CheckRegion, false, MatchLen, Req, Diags);
2845
2846 if (MatchPos == StringRef::npos) {
2847 ChecksFailed = true;
2848 i = j;
2849 break;
2850 }
2851
2852 CheckRegion = CheckRegion.substr(MatchPos + MatchLen);
2853 }
2854
2855 if (j == e)
2856 break;
2857 }
2858
2859 // Success if no checks failed.
2860 return !ChecksFailed;
2861 }
2862