[ARM] Rejig MVE load store tests. NFC
[llvm-core.git] / lib / Support / FileCheck.cpp
blobdb0231523025382b9739f3cdc99da7683fe9112b
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.
12 // This file implements most of the API that will be used by the FileCheck utility
13 // as well as various unittests.
14 //===----------------------------------------------------------------------===//
16 #include "llvm/Support/FileCheck.h"
17 #include "llvm/ADT/StringSet.h"
18 #include "llvm/Support/FormatVariadic.h"
19 #include <cstdint>
20 #include <list>
21 #include <map>
22 #include <tuple>
23 #include <utility>
25 using namespace llvm;
27 bool FileCheckNumericVariable::isValueKnownAtMatchTime() const {
28 if (Value)
29 return true;
31 return ExpressionAST != nullptr;
34 void FileCheckNumericVariable::setValue(uint64_t NewValue) {
35 if (ExpressionAST != nullptr) {
36 // Caller is expected to call setValue only if substitution was successful.
37 assert(NewValue == cantFail(ExpressionAST->eval(),
38 "Failed to evaluate associated expression when "
39 "sanity checking value") &&
40 "Value being set to different from variable evaluation");
42 Value = NewValue;
43 // Clear pointer to AST to ensure it is not used after the numeric
44 // substitution defining this variable is processed since it's the
45 // substitution that owns the pointer.
46 ExpressionAST = nullptr;
49 Expected<uint64_t> FileCheckNumericVariableUse::eval() const {
50 Optional<uint64_t> Value = NumericVariable->getValue();
51 if (Value)
52 return *Value;
54 FileCheckExpressionAST *ExpressionAST = NumericVariable->getExpressionAST();
55 if (!ExpressionAST)
56 return make_error<FileCheckUndefVarError>(Name);
58 return ExpressionAST->eval();
61 Expected<uint64_t> FileCheckASTBinop::eval() const {
62 Expected<uint64_t> LeftOp = LeftOperand->eval();
63 Expected<uint64_t> RightOp = RightOperand->eval();
65 // Bubble up any error (e.g. undefined variables) in the recursive
66 // evaluation.
67 if (!LeftOp || !RightOp) {
68 Error Err = Error::success();
69 if (!LeftOp)
70 Err = joinErrors(std::move(Err), LeftOp.takeError());
71 if (!RightOp)
72 Err = joinErrors(std::move(Err), RightOp.takeError());
73 return std::move(Err);
76 return EvalBinop(*LeftOp, *RightOp);
79 Expected<std::string> FileCheckNumericSubstitution::getResult() const {
80 Expected<uint64_t> EvaluatedValue = ExpressionAST->eval();
81 if (!EvaluatedValue)
82 return EvaluatedValue.takeError();
83 return utostr(*EvaluatedValue);
86 Expected<std::string> FileCheckStringSubstitution::getResult() const {
87 // Look up the value and escape it so that we can put it into the regex.
88 Expected<StringRef> VarVal = Context->getPatternVarValue(FromStr);
89 if (!VarVal)
90 return VarVal.takeError();
91 return Regex::escape(*VarVal);
94 bool FileCheckPattern::isValidVarNameStart(char C) {
95 return C == '_' || isalpha(C);
98 Expected<FileCheckPattern::VariableProperties>
99 FileCheckPattern::parseVariable(StringRef &Str, const SourceMgr &SM) {
100 if (Str.empty())
101 return FileCheckErrorDiagnostic::get(SM, Str, "empty variable name");
103 bool ParsedOneChar = false;
104 unsigned I = 0;
105 bool IsPseudo = Str[0] == '@';
107 // Global vars start with '$'.
108 if (Str[0] == '$' || IsPseudo)
109 ++I;
111 for (unsigned E = Str.size(); I != E; ++I) {
112 if (!ParsedOneChar && !isValidVarNameStart(Str[I]))
113 return FileCheckErrorDiagnostic::get(SM, Str, "invalid variable name");
115 // Variable names are composed of alphanumeric characters and underscores.
116 if (Str[I] != '_' && !isalnum(Str[I]))
117 break;
118 ParsedOneChar = true;
121 StringRef Name = Str.take_front(I);
122 Str = Str.substr(I);
123 return VariableProperties {Name, IsPseudo};
126 // StringRef holding all characters considered as horizontal whitespaces by
127 // FileCheck input canonicalization.
128 StringRef SpaceChars = " \t";
130 // Parsing helper function that strips the first character in S and returns it.
131 static char popFront(StringRef &S) {
132 char C = S.front();
133 S = S.drop_front();
134 return C;
137 char FileCheckUndefVarError::ID = 0;
138 char FileCheckErrorDiagnostic::ID = 0;
139 char FileCheckNotFoundError::ID = 0;
141 Expected<FileCheckNumericVariable *>
142 FileCheckPattern::parseNumericVariableDefinition(
143 StringRef &Expr, FileCheckPatternContext *Context,
144 Optional<size_t> LineNumber, FileCheckExpressionAST *ExpressionAST,
145 const SourceMgr &SM) {
146 Expected<VariableProperties> ParseVarResult = parseVariable(Expr, SM);
147 if (!ParseVarResult)
148 return ParseVarResult.takeError();
149 StringRef Name = ParseVarResult->Name;
151 if (ParseVarResult->IsPseudo)
152 return FileCheckErrorDiagnostic::get(
153 SM, Name, "definition of pseudo numeric variable unsupported");
155 // Detect collisions between string and numeric variables when the latter
156 // is created later than the former.
157 if (Context->DefinedVariableTable.find(Name) !=
158 Context->DefinedVariableTable.end())
159 return FileCheckErrorDiagnostic::get(
160 SM, Name, "string variable with name '" + Name + "' already exists");
162 Expr = Expr.ltrim(SpaceChars);
163 if (!Expr.empty())
164 return FileCheckErrorDiagnostic::get(
165 SM, Expr, "unexpected characters after numeric variable name");
167 FileCheckNumericVariable *DefinedNumericVariable;
168 auto VarTableIter = Context->GlobalNumericVariableTable.find(Name);
169 if (VarTableIter != Context->GlobalNumericVariableTable.end())
170 DefinedNumericVariable = VarTableIter->second;
171 else
172 DefinedNumericVariable =
173 Context->makeNumericVariable(Name, LineNumber, ExpressionAST);
175 return DefinedNumericVariable;
178 Expected<std::unique_ptr<FileCheckNumericVariableUse>>
179 FileCheckPattern::parseNumericVariableUse(StringRef Name, bool IsPseudo,
180 Optional<size_t> LineNumber,
181 FileCheckPatternContext *Context,
182 const SourceMgr &SM) {
183 if (IsPseudo && !Name.equals("@LINE"))
184 return FileCheckErrorDiagnostic::get(
185 SM, Name, "invalid pseudo numeric variable '" + Name + "'");
187 // Numeric variable definitions and uses are parsed in the order in which
188 // they appear in the CHECK patterns. For each definition, the pointer to the
189 // class instance of the corresponding numeric variable definition is stored
190 // in GlobalNumericVariableTable in parsePattern. Therefore, if the pointer
191 // we get below is null, it means no such variable was defined before. When
192 // that happens, we create a dummy variable so that parsing can continue. All
193 // uses of undefined variables, whether string or numeric, are then diagnosed
194 // in printSubstitutions() after failing to match.
195 auto VarTableIter = Context->GlobalNumericVariableTable.find(Name);
196 FileCheckNumericVariable *NumericVariable;
197 if (VarTableIter != Context->GlobalNumericVariableTable.end())
198 NumericVariable = VarTableIter->second;
199 else {
200 NumericVariable = Context->makeNumericVariable(Name);
201 Context->GlobalNumericVariableTable[Name] = NumericVariable;
204 Optional<size_t> DefLineNumber = NumericVariable->getDefLineNumber();
205 if (DefLineNumber && LineNumber && *DefLineNumber == *LineNumber &&
206 !NumericVariable->isValueKnownAtMatchTime())
207 return FileCheckErrorDiagnostic::get(
208 SM, Name,
209 "numeric variable '" + Name +
210 "' defined from input on the same line as used");
212 return llvm::make_unique<FileCheckNumericVariableUse>(Name, NumericVariable);
215 Expected<std::unique_ptr<FileCheckExpressionAST>>
216 FileCheckPattern::parseNumericOperand(StringRef &Expr, AllowedOperand AO,
217 Optional<size_t> LineNumber,
218 FileCheckPatternContext *Context,
219 const SourceMgr &SM) {
220 if (AO == AllowedOperand::LineVar || AO == AllowedOperand::Any) {
221 // Try to parse as a numeric variable use.
222 Expected<FileCheckPattern::VariableProperties> ParseVarResult =
223 parseVariable(Expr, SM);
224 if (ParseVarResult)
225 return parseNumericVariableUse(ParseVarResult->Name,
226 ParseVarResult->IsPseudo, LineNumber,
227 Context, SM);
228 if (AO == AllowedOperand::LineVar)
229 return ParseVarResult.takeError();
230 // Ignore the error and retry parsing as a literal.
231 consumeError(ParseVarResult.takeError());
234 // Otherwise, parse it as a literal.
235 uint64_t LiteralValue;
236 if (!Expr.consumeInteger(/*Radix=*/10, LiteralValue))
237 return llvm::make_unique<FileCheckExpressionLiteral>(LiteralValue);
239 return FileCheckErrorDiagnostic::get(SM, Expr,
240 "invalid operand format '" + Expr + "'");
243 static uint64_t add(uint64_t LeftOp, uint64_t RightOp) {
244 return LeftOp + RightOp;
247 static uint64_t sub(uint64_t LeftOp, uint64_t RightOp) {
248 return LeftOp - RightOp;
251 Expected<std::unique_ptr<FileCheckExpressionAST>> FileCheckPattern::parseBinop(
252 StringRef &Expr, std::unique_ptr<FileCheckExpressionAST> LeftOp,
253 bool IsLegacyLineExpr, Optional<size_t> LineNumber,
254 FileCheckPatternContext *Context, const SourceMgr &SM) {
255 Expr = Expr.ltrim(SpaceChars);
256 if (Expr.empty())
257 return std::move(LeftOp);
259 // Check if this is a supported operation and select a function to perform
260 // it.
261 SMLoc OpLoc = SMLoc::getFromPointer(Expr.data());
262 char Operator = popFront(Expr);
263 binop_eval_t EvalBinop;
264 switch (Operator) {
265 case '+':
266 EvalBinop = add;
267 break;
268 case '-':
269 EvalBinop = sub;
270 break;
271 default:
272 return FileCheckErrorDiagnostic::get(
273 SM, OpLoc, Twine("unsupported operation '") + Twine(Operator) + "'");
276 // Parse right operand.
277 Expr = Expr.ltrim(SpaceChars);
278 if (Expr.empty())
279 return FileCheckErrorDiagnostic::get(SM, Expr,
280 "missing operand in expression");
281 // The second operand in a legacy @LINE expression is always a literal.
282 AllowedOperand AO =
283 IsLegacyLineExpr ? AllowedOperand::Literal : AllowedOperand::Any;
284 Expected<std::unique_ptr<FileCheckExpressionAST>> RightOpResult =
285 parseNumericOperand(Expr, AO, LineNumber, Context, SM);
286 if (!RightOpResult)
287 return RightOpResult;
289 Expr = Expr.ltrim(SpaceChars);
290 return llvm::make_unique<FileCheckASTBinop>(EvalBinop, std::move(LeftOp),
291 std::move(*RightOpResult));
294 Expected<std::unique_ptr<FileCheckExpressionAST>>
295 FileCheckPattern::parseNumericSubstitutionBlock(
296 StringRef Expr,
297 Optional<FileCheckNumericVariable *> &DefinedNumericVariable,
298 bool IsLegacyLineExpr, Optional<size_t> LineNumber,
299 FileCheckPatternContext *Context, const SourceMgr &SM) {
300 std::unique_ptr<FileCheckExpressionAST> ExpressionAST = nullptr;
301 StringRef DefExpr = StringRef();
302 DefinedNumericVariable = None;
303 // Save variable definition expression if any.
304 size_t DefEnd = Expr.find(':');
305 if (DefEnd != StringRef::npos) {
306 DefExpr = Expr.substr(0, DefEnd);
307 Expr = Expr.substr(DefEnd + 1);
310 // Parse the expression itself.
311 Expr = Expr.ltrim(SpaceChars);
312 if (!Expr.empty()) {
313 // The first operand in a legacy @LINE expression is always the @LINE
314 // pseudo variable.
315 AllowedOperand AO =
316 IsLegacyLineExpr ? AllowedOperand::LineVar : AllowedOperand::Any;
317 Expected<std::unique_ptr<FileCheckExpressionAST>> ParseResult =
318 parseNumericOperand(Expr, AO, LineNumber, Context, SM);
319 while (ParseResult && !Expr.empty()) {
320 ParseResult = parseBinop(Expr, std::move(*ParseResult), IsLegacyLineExpr,
321 LineNumber, Context, SM);
322 // Legacy @LINE expressions only allow 2 operands.
323 if (ParseResult && IsLegacyLineExpr && !Expr.empty())
324 return FileCheckErrorDiagnostic::get(
325 SM, Expr,
326 "unexpected characters at end of expression '" + Expr + "'");
328 if (!ParseResult)
329 return ParseResult;
330 ExpressionAST = std::move(*ParseResult);
333 // Parse the numeric variable definition.
334 if (DefEnd != StringRef::npos) {
335 DefExpr = DefExpr.ltrim(SpaceChars);
336 Expected<FileCheckNumericVariable *> ParseResult =
337 parseNumericVariableDefinition(DefExpr, Context, LineNumber,
338 ExpressionAST.get(), SM);
340 if (!ParseResult)
341 return ParseResult.takeError();
342 DefinedNumericVariable = *ParseResult;
345 return std::move(ExpressionAST);
348 bool FileCheckPattern::parsePattern(StringRef PatternStr, StringRef Prefix,
349 SourceMgr &SM,
350 const FileCheckRequest &Req) {
351 bool MatchFullLinesHere = Req.MatchFullLines && CheckTy != Check::CheckNot;
353 PatternLoc = SMLoc::getFromPointer(PatternStr.data());
355 if (!(Req.NoCanonicalizeWhiteSpace && Req.MatchFullLines))
356 // Ignore trailing whitespace.
357 while (!PatternStr.empty() &&
358 (PatternStr.back() == ' ' || PatternStr.back() == '\t'))
359 PatternStr = PatternStr.substr(0, PatternStr.size() - 1);
361 // Check that there is something on the line.
362 if (PatternStr.empty() && CheckTy != Check::CheckEmpty) {
363 SM.PrintMessage(PatternLoc, SourceMgr::DK_Error,
364 "found empty check string with prefix '" + Prefix + ":'");
365 return true;
368 if (!PatternStr.empty() && CheckTy == Check::CheckEmpty) {
369 SM.PrintMessage(
370 PatternLoc, SourceMgr::DK_Error,
371 "found non-empty check string for empty check with prefix '" + Prefix +
372 ":'");
373 return true;
376 if (CheckTy == Check::CheckEmpty) {
377 RegExStr = "(\n$)";
378 return false;
381 // Check to see if this is a fixed string, or if it has regex pieces.
382 if (!MatchFullLinesHere &&
383 (PatternStr.size() < 2 || (PatternStr.find("{{") == StringRef::npos &&
384 PatternStr.find("[[") == StringRef::npos))) {
385 FixedStr = PatternStr;
386 return false;
389 if (MatchFullLinesHere) {
390 RegExStr += '^';
391 if (!Req.NoCanonicalizeWhiteSpace)
392 RegExStr += " *";
395 // Paren value #0 is for the fully matched string. Any new parenthesized
396 // values add from there.
397 unsigned CurParen = 1;
399 // Otherwise, there is at least one regex piece. Build up the regex pattern
400 // by escaping scary characters in fixed strings, building up one big regex.
401 while (!PatternStr.empty()) {
402 // RegEx matches.
403 if (PatternStr.startswith("{{")) {
404 // This is the start of a regex match. Scan for the }}.
405 size_t End = PatternStr.find("}}");
406 if (End == StringRef::npos) {
407 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
408 SourceMgr::DK_Error,
409 "found start of regex string with no end '}}'");
410 return true;
413 // Enclose {{}} patterns in parens just like [[]] even though we're not
414 // capturing the result for any purpose. This is required in case the
415 // expression contains an alternation like: CHECK: abc{{x|z}}def. We
416 // want this to turn into: "abc(x|z)def" not "abcx|zdef".
417 RegExStr += '(';
418 ++CurParen;
420 if (AddRegExToRegEx(PatternStr.substr(2, End - 2), CurParen, SM))
421 return true;
422 RegExStr += ')';
424 PatternStr = PatternStr.substr(End + 2);
425 continue;
428 // String and numeric substitution blocks. Pattern substitution blocks come
429 // in two forms: [[foo:.*]] and [[foo]]. The former matches .* (or some
430 // other regex) and assigns it to the string variable 'foo'. The latter
431 // substitutes foo's value. Numeric substitution blocks recognize the same
432 // form as string ones, but start with a '#' sign after the double
433 // brackets. They also accept a combined form which sets a numeric variable
434 // to the evaluation of an expression. Both string and numeric variable
435 // names must satisfy the regular expression "[a-zA-Z_][0-9a-zA-Z_]*" to be
436 // valid, as this helps catch some common errors.
437 if (PatternStr.startswith("[[")) {
438 StringRef UnparsedPatternStr = PatternStr.substr(2);
439 // Find the closing bracket pair ending the match. End is going to be an
440 // offset relative to the beginning of the match string.
441 size_t End = FindRegexVarEnd(UnparsedPatternStr, SM);
442 StringRef MatchStr = UnparsedPatternStr.substr(0, End);
443 bool IsNumBlock = MatchStr.consume_front("#");
445 if (End == StringRef::npos) {
446 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
447 SourceMgr::DK_Error,
448 "Invalid substitution block, no ]] found");
449 return true;
451 // Strip the substitution block we are parsing. End points to the start
452 // of the "]]" closing the expression so account for it in computing the
453 // index of the first unparsed character.
454 PatternStr = UnparsedPatternStr.substr(End + 2);
456 bool IsDefinition = false;
457 bool SubstNeeded = false;
458 // Whether the substitution block is a legacy use of @LINE with string
459 // substitution block syntax.
460 bool IsLegacyLineExpr = false;
461 StringRef DefName;
462 StringRef SubstStr;
463 StringRef MatchRegexp;
464 size_t SubstInsertIdx = RegExStr.size();
466 // Parse string variable or legacy @LINE expression.
467 if (!IsNumBlock) {
468 size_t VarEndIdx = MatchStr.find(":");
469 size_t SpacePos = MatchStr.substr(0, VarEndIdx).find_first_of(" \t");
470 if (SpacePos != StringRef::npos) {
471 SM.PrintMessage(SMLoc::getFromPointer(MatchStr.data() + SpacePos),
472 SourceMgr::DK_Error, "unexpected whitespace");
473 return true;
476 // Get the name (e.g. "foo") and verify it is well formed.
477 StringRef OrigMatchStr = MatchStr;
478 Expected<FileCheckPattern::VariableProperties> ParseVarResult =
479 parseVariable(MatchStr, SM);
480 if (!ParseVarResult) {
481 logAllUnhandledErrors(ParseVarResult.takeError(), errs());
482 return true;
484 StringRef Name = ParseVarResult->Name;
485 bool IsPseudo = ParseVarResult->IsPseudo;
487 IsDefinition = (VarEndIdx != StringRef::npos);
488 SubstNeeded = !IsDefinition;
489 if (IsDefinition) {
490 if ((IsPseudo || !MatchStr.consume_front(":"))) {
491 SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
492 SourceMgr::DK_Error,
493 "invalid name in string variable definition");
494 return true;
497 // Detect collisions between string and numeric variables when the
498 // former is created later than the latter.
499 if (Context->GlobalNumericVariableTable.find(Name) !=
500 Context->GlobalNumericVariableTable.end()) {
501 SM.PrintMessage(
502 SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
503 "numeric variable with name '" + Name + "' already exists");
504 return true;
506 DefName = Name;
507 MatchRegexp = MatchStr;
508 } else {
509 if (IsPseudo) {
510 MatchStr = OrigMatchStr;
511 IsLegacyLineExpr = IsNumBlock = true;
512 } else
513 SubstStr = Name;
517 // Parse numeric substitution block.
518 std::unique_ptr<FileCheckExpressionAST> ExpressionAST;
519 Optional<FileCheckNumericVariable *> DefinedNumericVariable;
520 if (IsNumBlock) {
521 Expected<std::unique_ptr<FileCheckExpressionAST>> ParseResult =
522 parseNumericSubstitutionBlock(MatchStr, DefinedNumericVariable,
523 IsLegacyLineExpr, LineNumber, Context,
524 SM);
525 if (!ParseResult) {
526 logAllUnhandledErrors(ParseResult.takeError(), errs());
527 return true;
529 ExpressionAST = std::move(*ParseResult);
530 SubstNeeded = ExpressionAST != nullptr;
531 if (DefinedNumericVariable) {
532 IsDefinition = true;
533 DefName = (*DefinedNumericVariable)->getName();
535 if (SubstNeeded)
536 SubstStr = MatchStr;
537 else
538 MatchRegexp = "[0-9]+";
541 // Handle variable definition: [[<def>:(...)]] and [[#(...)<def>:(...)]].
542 if (IsDefinition) {
543 RegExStr += '(';
544 ++SubstInsertIdx;
546 if (IsNumBlock) {
547 FileCheckNumericVariableMatch NumericVariableDefinition = {
548 *DefinedNumericVariable, CurParen};
549 NumericVariableDefs[DefName] = NumericVariableDefinition;
550 // This store is done here rather than in match() to allow
551 // parseNumericVariableUse() to get the pointer to the class instance
552 // of the right variable definition corresponding to a given numeric
553 // variable use.
554 Context->GlobalNumericVariableTable[DefName] =
555 *DefinedNumericVariable;
556 } else {
557 VariableDefs[DefName] = CurParen;
558 // Mark string variable as defined to detect collisions between
559 // string and numeric variables in parseNumericVariableUse() and
560 // defineCmdlineVariables() when the latter is created later than the
561 // former. We cannot reuse GlobalVariableTable for this by populating
562 // it with an empty string since we would then lose the ability to
563 // detect the use of an undefined variable in match().
564 Context->DefinedVariableTable[DefName] = true;
567 ++CurParen;
570 if (!MatchRegexp.empty() && AddRegExToRegEx(MatchRegexp, CurParen, SM))
571 return true;
573 if (IsDefinition)
574 RegExStr += ')';
576 // Handle substitutions: [[foo]] and [[#<foo expr>]].
577 if (SubstNeeded) {
578 // Handle substitution of string variables that were defined earlier on
579 // the same line by emitting a backreference. Expressions do not
580 // support substituting a numeric variable defined on the same line.
581 if (!IsNumBlock && VariableDefs.find(SubstStr) != VariableDefs.end()) {
582 unsigned CaptureParenGroup = VariableDefs[SubstStr];
583 if (CaptureParenGroup < 1 || CaptureParenGroup > 9) {
584 SM.PrintMessage(SMLoc::getFromPointer(SubstStr.data()),
585 SourceMgr::DK_Error,
586 "Can't back-reference more than 9 variables");
587 return true;
589 AddBackrefToRegEx(CaptureParenGroup);
590 } else {
591 // Handle substitution of string variables ([[<var>]]) defined in
592 // previous CHECK patterns, and substitution of expressions.
593 FileCheckSubstitution *Substitution =
594 IsNumBlock
595 ? Context->makeNumericSubstitution(
596 SubstStr, std::move(ExpressionAST), SubstInsertIdx)
597 : Context->makeStringSubstitution(SubstStr, SubstInsertIdx);
598 Substitutions.push_back(Substitution);
603 // Handle fixed string matches.
604 // Find the end, which is the start of the next regex.
605 size_t FixedMatchEnd = PatternStr.find("{{");
606 FixedMatchEnd = std::min(FixedMatchEnd, PatternStr.find("[["));
607 RegExStr += Regex::escape(PatternStr.substr(0, FixedMatchEnd));
608 PatternStr = PatternStr.substr(FixedMatchEnd);
611 if (MatchFullLinesHere) {
612 if (!Req.NoCanonicalizeWhiteSpace)
613 RegExStr += " *";
614 RegExStr += '$';
617 return false;
620 bool FileCheckPattern::AddRegExToRegEx(StringRef RS, unsigned &CurParen, SourceMgr &SM) {
621 Regex R(RS);
622 std::string Error;
623 if (!R.isValid(Error)) {
624 SM.PrintMessage(SMLoc::getFromPointer(RS.data()), SourceMgr::DK_Error,
625 "invalid regex: " + Error);
626 return true;
629 RegExStr += RS.str();
630 CurParen += R.getNumMatches();
631 return false;
634 void FileCheckPattern::AddBackrefToRegEx(unsigned BackrefNum) {
635 assert(BackrefNum >= 1 && BackrefNum <= 9 && "Invalid backref number");
636 std::string Backref = std::string("\\") + std::string(1, '0' + BackrefNum);
637 RegExStr += Backref;
640 Expected<size_t> FileCheckPattern::match(StringRef Buffer, size_t &MatchLen,
641 const SourceMgr &SM) const {
642 // If this is the EOF pattern, match it immediately.
643 if (CheckTy == Check::CheckEOF) {
644 MatchLen = 0;
645 return Buffer.size();
648 // If this is a fixed string pattern, just match it now.
649 if (!FixedStr.empty()) {
650 MatchLen = FixedStr.size();
651 size_t Pos = Buffer.find(FixedStr);
652 if (Pos == StringRef::npos)
653 return make_error<FileCheckNotFoundError>();
654 return Pos;
657 // Regex match.
659 // If there are substitutions, we need to create a temporary string with the
660 // actual value.
661 StringRef RegExToMatch = RegExStr;
662 std::string TmpStr;
663 if (!Substitutions.empty()) {
664 TmpStr = RegExStr;
665 if (LineNumber)
666 Context->LineVariable->setValue(*LineNumber);
668 size_t InsertOffset = 0;
669 // Substitute all string variables and expressions whose values are only
670 // now known. Use of string variables defined on the same line are handled
671 // by back-references.
672 for (const auto &Substitution : Substitutions) {
673 // Substitute and check for failure (e.g. use of undefined variable).
674 Expected<std::string> Value = Substitution->getResult();
675 if (!Value)
676 return Value.takeError();
678 // Plop it into the regex at the adjusted offset.
679 TmpStr.insert(TmpStr.begin() + Substitution->getIndex() + InsertOffset,
680 Value->begin(), Value->end());
681 InsertOffset += Value->size();
684 // Match the newly constructed regex.
685 RegExToMatch = TmpStr;
688 SmallVector<StringRef, 4> MatchInfo;
689 if (!Regex(RegExToMatch, Regex::Newline).match(Buffer, &MatchInfo))
690 return make_error<FileCheckNotFoundError>();
692 // Successful regex match.
693 assert(!MatchInfo.empty() && "Didn't get any match");
694 StringRef FullMatch = MatchInfo[0];
696 // If this defines any string variables, remember their values.
697 for (const auto &VariableDef : VariableDefs) {
698 assert(VariableDef.second < MatchInfo.size() && "Internal paren error");
699 Context->GlobalVariableTable[VariableDef.first] =
700 MatchInfo[VariableDef.second];
703 // If this defines any numeric variables, remember their values.
704 for (const auto &NumericVariableDef : NumericVariableDefs) {
705 const FileCheckNumericVariableMatch &NumericVariableMatch =
706 NumericVariableDef.getValue();
707 unsigned CaptureParenGroup = NumericVariableMatch.CaptureParenGroup;
708 assert(CaptureParenGroup < MatchInfo.size() && "Internal paren error");
709 FileCheckNumericVariable *DefinedNumericVariable =
710 NumericVariableMatch.DefinedNumericVariable;
712 StringRef MatchedValue = MatchInfo[CaptureParenGroup];
713 uint64_t Val;
714 if (MatchedValue.getAsInteger(10, Val))
715 return FileCheckErrorDiagnostic::get(SM, MatchedValue,
716 "Unable to represent numeric value");
717 DefinedNumericVariable->setValue(Val);
720 // Like CHECK-NEXT, CHECK-EMPTY's match range is considered to start after
721 // the required preceding newline, which is consumed by the pattern in the
722 // case of CHECK-EMPTY but not CHECK-NEXT.
723 size_t MatchStartSkip = CheckTy == Check::CheckEmpty;
724 MatchLen = FullMatch.size() - MatchStartSkip;
725 return FullMatch.data() - Buffer.data() + MatchStartSkip;
728 unsigned FileCheckPattern::computeMatchDistance(StringRef Buffer) const {
729 // Just compute the number of matching characters. For regular expressions, we
730 // just compare against the regex itself and hope for the best.
732 // FIXME: One easy improvement here is have the regex lib generate a single
733 // example regular expression which matches, and use that as the example
734 // string.
735 StringRef ExampleString(FixedStr);
736 if (ExampleString.empty())
737 ExampleString = RegExStr;
739 // Only compare up to the first line in the buffer, or the string size.
740 StringRef BufferPrefix = Buffer.substr(0, ExampleString.size());
741 BufferPrefix = BufferPrefix.split('\n').first;
742 return BufferPrefix.edit_distance(ExampleString);
745 void FileCheckPattern::printSubstitutions(const SourceMgr &SM, StringRef Buffer,
746 SMRange MatchRange) const {
747 // Print what we know about substitutions.
748 if (!Substitutions.empty()) {
749 for (const auto &Substitution : Substitutions) {
750 SmallString<256> Msg;
751 raw_svector_ostream OS(Msg);
752 Expected<std::string> MatchedValue = Substitution->getResult();
754 // Substitution failed or is not known at match time, print the undefined
755 // variables it uses.
756 if (!MatchedValue) {
757 bool UndefSeen = false;
758 handleAllErrors(MatchedValue.takeError(),
759 [](const FileCheckNotFoundError &E) {},
760 // Handled in PrintNoMatch().
761 [](const FileCheckErrorDiagnostic &E) {},
762 [&](const FileCheckUndefVarError &E) {
763 if (!UndefSeen) {
764 OS << "uses undefined variable(s):";
765 UndefSeen = true;
767 OS << " ";
768 E.log(OS);
770 } else {
771 // Substitution succeeded. Print substituted value.
772 OS << "with \"";
773 OS.write_escaped(Substitution->getFromString()) << "\" equal to \"";
774 OS.write_escaped(*MatchedValue) << "\"";
777 if (MatchRange.isValid())
778 SM.PrintMessage(MatchRange.Start, SourceMgr::DK_Note, OS.str(),
779 {MatchRange});
780 else
781 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()),
782 SourceMgr::DK_Note, OS.str());
787 static SMRange ProcessMatchResult(FileCheckDiag::MatchType MatchTy,
788 const SourceMgr &SM, SMLoc Loc,
789 Check::FileCheckType CheckTy,
790 StringRef Buffer, size_t Pos, size_t Len,
791 std::vector<FileCheckDiag> *Diags,
792 bool AdjustPrevDiag = false) {
793 SMLoc Start = SMLoc::getFromPointer(Buffer.data() + Pos);
794 SMLoc End = SMLoc::getFromPointer(Buffer.data() + Pos + Len);
795 SMRange Range(Start, End);
796 if (Diags) {
797 if (AdjustPrevDiag)
798 Diags->rbegin()->MatchTy = MatchTy;
799 else
800 Diags->emplace_back(SM, CheckTy, Loc, MatchTy, Range);
802 return Range;
805 void FileCheckPattern::printFuzzyMatch(
806 const SourceMgr &SM, StringRef Buffer,
807 std::vector<FileCheckDiag> *Diags) const {
808 // Attempt to find the closest/best fuzzy match. Usually an error happens
809 // because some string in the output didn't exactly match. In these cases, we
810 // would like to show the user a best guess at what "should have" matched, to
811 // save them having to actually check the input manually.
812 size_t NumLinesForward = 0;
813 size_t Best = StringRef::npos;
814 double BestQuality = 0;
816 // Use an arbitrary 4k limit on how far we will search.
817 for (size_t i = 0, e = std::min(size_t(4096), Buffer.size()); i != e; ++i) {
818 if (Buffer[i] == '\n')
819 ++NumLinesForward;
821 // Patterns have leading whitespace stripped, so skip whitespace when
822 // looking for something which looks like a pattern.
823 if (Buffer[i] == ' ' || Buffer[i] == '\t')
824 continue;
826 // Compute the "quality" of this match as an arbitrary combination of the
827 // match distance and the number of lines skipped to get to this match.
828 unsigned Distance = computeMatchDistance(Buffer.substr(i));
829 double Quality = Distance + (NumLinesForward / 100.);
831 if (Quality < BestQuality || Best == StringRef::npos) {
832 Best = i;
833 BestQuality = Quality;
837 // Print the "possible intended match here" line if we found something
838 // reasonable and not equal to what we showed in the "scanning from here"
839 // line.
840 if (Best && Best != StringRef::npos && BestQuality < 50) {
841 SMRange MatchRange =
842 ProcessMatchResult(FileCheckDiag::MatchFuzzy, SM, getLoc(),
843 getCheckTy(), Buffer, Best, 0, Diags);
844 SM.PrintMessage(MatchRange.Start, SourceMgr::DK_Note,
845 "possible intended match here");
847 // FIXME: If we wanted to be really friendly we would show why the match
848 // failed, as it can be hard to spot simple one character differences.
852 Expected<StringRef>
853 FileCheckPatternContext::getPatternVarValue(StringRef VarName) {
854 auto VarIter = GlobalVariableTable.find(VarName);
855 if (VarIter == GlobalVariableTable.end())
856 return make_error<FileCheckUndefVarError>(VarName);
858 return VarIter->second;
861 template <class... Types>
862 FileCheckNumericVariable *
863 FileCheckPatternContext::makeNumericVariable(Types... args) {
864 NumericVariables.push_back(
865 llvm::make_unique<FileCheckNumericVariable>(args...));
866 return NumericVariables.back().get();
869 FileCheckSubstitution *
870 FileCheckPatternContext::makeStringSubstitution(StringRef VarName,
871 size_t InsertIdx) {
872 Substitutions.push_back(
873 llvm::make_unique<FileCheckStringSubstitution>(this, VarName, InsertIdx));
874 return Substitutions.back().get();
877 FileCheckSubstitution *FileCheckPatternContext::makeNumericSubstitution(
878 StringRef ExpressionStr,
879 std::unique_ptr<FileCheckExpressionAST> ExpressionAST, size_t InsertIdx) {
880 Substitutions.push_back(llvm::make_unique<FileCheckNumericSubstitution>(
881 this, ExpressionStr, std::move(ExpressionAST), InsertIdx));
882 return Substitutions.back().get();
885 size_t FileCheckPattern::FindRegexVarEnd(StringRef Str, SourceMgr &SM) {
886 // Offset keeps track of the current offset within the input Str
887 size_t Offset = 0;
888 // [...] Nesting depth
889 size_t BracketDepth = 0;
891 while (!Str.empty()) {
892 if (Str.startswith("]]") && BracketDepth == 0)
893 return Offset;
894 if (Str[0] == '\\') {
895 // Backslash escapes the next char within regexes, so skip them both.
896 Str = Str.substr(2);
897 Offset += 2;
898 } else {
899 switch (Str[0]) {
900 default:
901 break;
902 case '[':
903 BracketDepth++;
904 break;
905 case ']':
906 if (BracketDepth == 0) {
907 SM.PrintMessage(SMLoc::getFromPointer(Str.data()),
908 SourceMgr::DK_Error,
909 "missing closing \"]\" for regex variable");
910 exit(1);
912 BracketDepth--;
913 break;
915 Str = Str.substr(1);
916 Offset++;
920 return StringRef::npos;
923 StringRef FileCheck::CanonicalizeFile(MemoryBuffer &MB,
924 SmallVectorImpl<char> &OutputBuffer) {
925 OutputBuffer.reserve(MB.getBufferSize());
927 for (const char *Ptr = MB.getBufferStart(), *End = MB.getBufferEnd();
928 Ptr != End; ++Ptr) {
929 // Eliminate trailing dosish \r.
930 if (Ptr <= End - 2 && Ptr[0] == '\r' && Ptr[1] == '\n') {
931 continue;
934 // If current char is not a horizontal whitespace or if horizontal
935 // whitespace canonicalization is disabled, dump it to output as is.
936 if (Req.NoCanonicalizeWhiteSpace || (*Ptr != ' ' && *Ptr != '\t')) {
937 OutputBuffer.push_back(*Ptr);
938 continue;
941 // Otherwise, add one space and advance over neighboring space.
942 OutputBuffer.push_back(' ');
943 while (Ptr + 1 != End && (Ptr[1] == ' ' || Ptr[1] == '\t'))
944 ++Ptr;
947 // Add a null byte and then return all but that byte.
948 OutputBuffer.push_back('\0');
949 return StringRef(OutputBuffer.data(), OutputBuffer.size() - 1);
952 FileCheckDiag::FileCheckDiag(const SourceMgr &SM,
953 const Check::FileCheckType &CheckTy,
954 SMLoc CheckLoc, MatchType MatchTy,
955 SMRange InputRange)
956 : CheckTy(CheckTy), MatchTy(MatchTy) {
957 auto Start = SM.getLineAndColumn(InputRange.Start);
958 auto End = SM.getLineAndColumn(InputRange.End);
959 InputStartLine = Start.first;
960 InputStartCol = Start.second;
961 InputEndLine = End.first;
962 InputEndCol = End.second;
963 Start = SM.getLineAndColumn(CheckLoc);
964 CheckLine = Start.first;
965 CheckCol = Start.second;
968 static bool IsPartOfWord(char c) {
969 return (isalnum(c) || c == '-' || c == '_');
972 Check::FileCheckType &Check::FileCheckType::setCount(int C) {
973 assert(Count > 0 && "zero and negative counts are not supported");
974 assert((C == 1 || Kind == CheckPlain) &&
975 "count supported only for plain CHECK directives");
976 Count = C;
977 return *this;
980 std::string Check::FileCheckType::getDescription(StringRef Prefix) const {
981 switch (Kind) {
982 case Check::CheckNone:
983 return "invalid";
984 case Check::CheckPlain:
985 if (Count > 1)
986 return Prefix.str() + "-COUNT";
987 return Prefix;
988 case Check::CheckNext:
989 return Prefix.str() + "-NEXT";
990 case Check::CheckSame:
991 return Prefix.str() + "-SAME";
992 case Check::CheckNot:
993 return Prefix.str() + "-NOT";
994 case Check::CheckDAG:
995 return Prefix.str() + "-DAG";
996 case Check::CheckLabel:
997 return Prefix.str() + "-LABEL";
998 case Check::CheckEmpty:
999 return Prefix.str() + "-EMPTY";
1000 case Check::CheckEOF:
1001 return "implicit EOF";
1002 case Check::CheckBadNot:
1003 return "bad NOT";
1004 case Check::CheckBadCount:
1005 return "bad COUNT";
1007 llvm_unreachable("unknown FileCheckType");
1010 static std::pair<Check::FileCheckType, StringRef>
1011 FindCheckType(StringRef Buffer, StringRef Prefix) {
1012 if (Buffer.size() <= Prefix.size())
1013 return {Check::CheckNone, StringRef()};
1015 char NextChar = Buffer[Prefix.size()];
1017 StringRef Rest = Buffer.drop_front(Prefix.size() + 1);
1018 // Verify that the : is present after the prefix.
1019 if (NextChar == ':')
1020 return {Check::CheckPlain, Rest};
1022 if (NextChar != '-')
1023 return {Check::CheckNone, StringRef()};
1025 if (Rest.consume_front("COUNT-")) {
1026 int64_t Count;
1027 if (Rest.consumeInteger(10, Count))
1028 // Error happened in parsing integer.
1029 return {Check::CheckBadCount, Rest};
1030 if (Count <= 0 || Count > INT32_MAX)
1031 return {Check::CheckBadCount, Rest};
1032 if (!Rest.consume_front(":"))
1033 return {Check::CheckBadCount, Rest};
1034 return {Check::FileCheckType(Check::CheckPlain).setCount(Count), Rest};
1037 if (Rest.consume_front("NEXT:"))
1038 return {Check::CheckNext, Rest};
1040 if (Rest.consume_front("SAME:"))
1041 return {Check::CheckSame, Rest};
1043 if (Rest.consume_front("NOT:"))
1044 return {Check::CheckNot, Rest};
1046 if (Rest.consume_front("DAG:"))
1047 return {Check::CheckDAG, Rest};
1049 if (Rest.consume_front("LABEL:"))
1050 return {Check::CheckLabel, Rest};
1052 if (Rest.consume_front("EMPTY:"))
1053 return {Check::CheckEmpty, Rest};
1055 // You can't combine -NOT with another suffix.
1056 if (Rest.startswith("DAG-NOT:") || Rest.startswith("NOT-DAG:") ||
1057 Rest.startswith("NEXT-NOT:") || Rest.startswith("NOT-NEXT:") ||
1058 Rest.startswith("SAME-NOT:") || Rest.startswith("NOT-SAME:") ||
1059 Rest.startswith("EMPTY-NOT:") || Rest.startswith("NOT-EMPTY:"))
1060 return {Check::CheckBadNot, Rest};
1062 return {Check::CheckNone, Rest};
1065 // From the given position, find the next character after the word.
1066 static size_t SkipWord(StringRef Str, size_t Loc) {
1067 while (Loc < Str.size() && IsPartOfWord(Str[Loc]))
1068 ++Loc;
1069 return Loc;
1072 /// Searches the buffer for the first prefix in the prefix regular expression.
1074 /// This searches the buffer using the provided regular expression, however it
1075 /// enforces constraints beyond that:
1076 /// 1) The found prefix must not be a suffix of something that looks like
1077 /// a valid prefix.
1078 /// 2) The found prefix must be followed by a valid check type suffix using \c
1079 /// FindCheckType above.
1081 /// \returns a pair of StringRefs into the Buffer, which combines:
1082 /// - the first match of the regular expression to satisfy these two is
1083 /// returned,
1084 /// otherwise an empty StringRef is returned to indicate failure.
1085 /// - buffer rewound to the location right after parsed suffix, for parsing
1086 /// to continue from
1088 /// If this routine returns a valid prefix, it will also shrink \p Buffer to
1089 /// start at the beginning of the returned prefix, increment \p LineNumber for
1090 /// each new line consumed from \p Buffer, and set \p CheckTy to the type of
1091 /// check found by examining the suffix.
1093 /// If no valid prefix is found, the state of Buffer, LineNumber, and CheckTy
1094 /// is unspecified.
1095 static std::pair<StringRef, StringRef>
1096 FindFirstMatchingPrefix(Regex &PrefixRE, StringRef &Buffer,
1097 unsigned &LineNumber, Check::FileCheckType &CheckTy) {
1098 SmallVector<StringRef, 2> Matches;
1100 while (!Buffer.empty()) {
1101 // Find the first (longest) match using the RE.
1102 if (!PrefixRE.match(Buffer, &Matches))
1103 // No match at all, bail.
1104 return {StringRef(), StringRef()};
1106 StringRef Prefix = Matches[0];
1107 Matches.clear();
1109 assert(Prefix.data() >= Buffer.data() &&
1110 Prefix.data() < Buffer.data() + Buffer.size() &&
1111 "Prefix doesn't start inside of buffer!");
1112 size_t Loc = Prefix.data() - Buffer.data();
1113 StringRef Skipped = Buffer.substr(0, Loc);
1114 Buffer = Buffer.drop_front(Loc);
1115 LineNumber += Skipped.count('\n');
1117 // Check that the matched prefix isn't a suffix of some other check-like
1118 // word.
1119 // FIXME: This is a very ad-hoc check. it would be better handled in some
1120 // other way. Among other things it seems hard to distinguish between
1121 // intentional and unintentional uses of this feature.
1122 if (Skipped.empty() || !IsPartOfWord(Skipped.back())) {
1123 // Now extract the type.
1124 StringRef AfterSuffix;
1125 std::tie(CheckTy, AfterSuffix) = FindCheckType(Buffer, Prefix);
1127 // If we've found a valid check type for this prefix, we're done.
1128 if (CheckTy != Check::CheckNone)
1129 return {Prefix, AfterSuffix};
1132 // If we didn't successfully find a prefix, we need to skip this invalid
1133 // prefix and continue scanning. We directly skip the prefix that was
1134 // matched and any additional parts of that check-like word.
1135 Buffer = Buffer.drop_front(SkipWord(Buffer, Prefix.size()));
1138 // We ran out of buffer while skipping partial matches so give up.
1139 return {StringRef(), StringRef()};
1142 void FileCheckPatternContext::createLineVariable() {
1143 assert(!LineVariable && "@LINE pseudo numeric variable already created");
1144 StringRef LineName = "@LINE";
1145 LineVariable = makeNumericVariable(LineName);
1146 GlobalNumericVariableTable[LineName] = LineVariable;
1149 bool FileCheck::ReadCheckFile(SourceMgr &SM, StringRef Buffer, Regex &PrefixRE,
1150 std::vector<FileCheckString> &CheckStrings) {
1151 Error DefineError =
1152 PatternContext.defineCmdlineVariables(Req.GlobalDefines, SM);
1153 if (DefineError) {
1154 logAllUnhandledErrors(std::move(DefineError), errs());
1155 return true;
1158 PatternContext.createLineVariable();
1160 std::vector<FileCheckPattern> ImplicitNegativeChecks;
1161 for (const auto &PatternString : Req.ImplicitCheckNot) {
1162 // Create a buffer with fake command line content in order to display the
1163 // command line option responsible for the specific implicit CHECK-NOT.
1164 std::string Prefix = "-implicit-check-not='";
1165 std::string Suffix = "'";
1166 std::unique_ptr<MemoryBuffer> CmdLine = MemoryBuffer::getMemBufferCopy(
1167 Prefix + PatternString + Suffix, "command line");
1169 StringRef PatternInBuffer =
1170 CmdLine->getBuffer().substr(Prefix.size(), PatternString.size());
1171 SM.AddNewSourceBuffer(std::move(CmdLine), SMLoc());
1173 ImplicitNegativeChecks.push_back(
1174 FileCheckPattern(Check::CheckNot, &PatternContext));
1175 ImplicitNegativeChecks.back().parsePattern(PatternInBuffer,
1176 "IMPLICIT-CHECK", SM, Req);
1179 std::vector<FileCheckPattern> DagNotMatches = ImplicitNegativeChecks;
1181 // LineNumber keeps track of the line on which CheckPrefix instances are
1182 // found.
1183 unsigned LineNumber = 1;
1185 while (1) {
1186 Check::FileCheckType CheckTy;
1188 // See if a prefix occurs in the memory buffer.
1189 StringRef UsedPrefix;
1190 StringRef AfterSuffix;
1191 std::tie(UsedPrefix, AfterSuffix) =
1192 FindFirstMatchingPrefix(PrefixRE, Buffer, LineNumber, CheckTy);
1193 if (UsedPrefix.empty())
1194 break;
1195 assert(UsedPrefix.data() == Buffer.data() &&
1196 "Failed to move Buffer's start forward, or pointed prefix outside "
1197 "of the buffer!");
1198 assert(AfterSuffix.data() >= Buffer.data() &&
1199 AfterSuffix.data() < Buffer.data() + Buffer.size() &&
1200 "Parsing after suffix doesn't start inside of buffer!");
1202 // Location to use for error messages.
1203 const char *UsedPrefixStart = UsedPrefix.data();
1205 // Skip the buffer to the end of parsed suffix (or just prefix, if no good
1206 // suffix was processed).
1207 Buffer = AfterSuffix.empty() ? Buffer.drop_front(UsedPrefix.size())
1208 : AfterSuffix;
1210 // Complain about useful-looking but unsupported suffixes.
1211 if (CheckTy == Check::CheckBadNot) {
1212 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Error,
1213 "unsupported -NOT combo on prefix '" + UsedPrefix + "'");
1214 return true;
1217 // Complain about invalid count specification.
1218 if (CheckTy == Check::CheckBadCount) {
1219 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Error,
1220 "invalid count in -COUNT specification on prefix '" +
1221 UsedPrefix + "'");
1222 return true;
1225 // Okay, we found the prefix, yay. Remember the rest of the line, but ignore
1226 // leading whitespace.
1227 if (!(Req.NoCanonicalizeWhiteSpace && Req.MatchFullLines))
1228 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t"));
1230 // Scan ahead to the end of line.
1231 size_t EOL = Buffer.find_first_of("\n\r");
1233 // Remember the location of the start of the pattern, for diagnostics.
1234 SMLoc PatternLoc = SMLoc::getFromPointer(Buffer.data());
1236 // Parse the pattern.
1237 FileCheckPattern P(CheckTy, &PatternContext, LineNumber);
1238 if (P.parsePattern(Buffer.substr(0, EOL), UsedPrefix, SM, Req))
1239 return true;
1241 // Verify that CHECK-LABEL lines do not define or use variables
1242 if ((CheckTy == Check::CheckLabel) && P.hasVariable()) {
1243 SM.PrintMessage(
1244 SMLoc::getFromPointer(UsedPrefixStart), SourceMgr::DK_Error,
1245 "found '" + UsedPrefix + "-LABEL:'"
1246 " with variable definition or use");
1247 return true;
1250 Buffer = Buffer.substr(EOL);
1252 // Verify that CHECK-NEXT/SAME/EMPTY lines have at least one CHECK line before them.
1253 if ((CheckTy == Check::CheckNext || CheckTy == Check::CheckSame ||
1254 CheckTy == Check::CheckEmpty) &&
1255 CheckStrings.empty()) {
1256 StringRef Type = CheckTy == Check::CheckNext
1257 ? "NEXT"
1258 : CheckTy == Check::CheckEmpty ? "EMPTY" : "SAME";
1259 SM.PrintMessage(SMLoc::getFromPointer(UsedPrefixStart),
1260 SourceMgr::DK_Error,
1261 "found '" + UsedPrefix + "-" + Type +
1262 "' without previous '" + UsedPrefix + ": line");
1263 return true;
1266 // Handle CHECK-DAG/-NOT.
1267 if (CheckTy == Check::CheckDAG || CheckTy == Check::CheckNot) {
1268 DagNotMatches.push_back(P);
1269 continue;
1272 // Okay, add the string we captured to the output vector and move on.
1273 CheckStrings.emplace_back(P, UsedPrefix, PatternLoc);
1274 std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
1275 DagNotMatches = ImplicitNegativeChecks;
1278 // Add an EOF pattern for any trailing CHECK-DAG/-NOTs, and use the first
1279 // prefix as a filler for the error message.
1280 if (!DagNotMatches.empty()) {
1281 CheckStrings.emplace_back(
1282 FileCheckPattern(Check::CheckEOF, &PatternContext, LineNumber + 1),
1283 *Req.CheckPrefixes.begin(), SMLoc::getFromPointer(Buffer.data()));
1284 std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
1287 if (CheckStrings.empty()) {
1288 errs() << "error: no check strings found with prefix"
1289 << (Req.CheckPrefixes.size() > 1 ? "es " : " ");
1290 auto I = Req.CheckPrefixes.begin();
1291 auto E = Req.CheckPrefixes.end();
1292 if (I != E) {
1293 errs() << "\'" << *I << ":'";
1294 ++I;
1296 for (; I != E; ++I)
1297 errs() << ", \'" << *I << ":'";
1299 errs() << '\n';
1300 return true;
1303 return false;
1306 static void PrintMatch(bool ExpectedMatch, const SourceMgr &SM,
1307 StringRef Prefix, SMLoc Loc, const FileCheckPattern &Pat,
1308 int MatchedCount, StringRef Buffer, size_t MatchPos,
1309 size_t MatchLen, const FileCheckRequest &Req,
1310 std::vector<FileCheckDiag> *Diags) {
1311 bool PrintDiag = true;
1312 if (ExpectedMatch) {
1313 if (!Req.Verbose)
1314 return;
1315 if (!Req.VerboseVerbose && Pat.getCheckTy() == Check::CheckEOF)
1316 return;
1317 // Due to their verbosity, we don't print verbose diagnostics here if we're
1318 // gathering them for a different rendering, but we always print other
1319 // diagnostics.
1320 PrintDiag = !Diags;
1322 SMRange MatchRange = ProcessMatchResult(
1323 ExpectedMatch ? FileCheckDiag::MatchFoundAndExpected
1324 : FileCheckDiag::MatchFoundButExcluded,
1325 SM, Loc, Pat.getCheckTy(), Buffer, MatchPos, MatchLen, Diags);
1326 if (!PrintDiag)
1327 return;
1329 std::string Message = formatv("{0}: {1} string found in input",
1330 Pat.getCheckTy().getDescription(Prefix),
1331 (ExpectedMatch ? "expected" : "excluded"))
1332 .str();
1333 if (Pat.getCount() > 1)
1334 Message += formatv(" ({0} out of {1})", MatchedCount, Pat.getCount()).str();
1336 SM.PrintMessage(
1337 Loc, ExpectedMatch ? SourceMgr::DK_Remark : SourceMgr::DK_Error, Message);
1338 SM.PrintMessage(MatchRange.Start, SourceMgr::DK_Note, "found here",
1339 {MatchRange});
1340 Pat.printSubstitutions(SM, Buffer, MatchRange);
1343 static void PrintMatch(bool ExpectedMatch, const SourceMgr &SM,
1344 const FileCheckString &CheckStr, int MatchedCount,
1345 StringRef Buffer, size_t MatchPos, size_t MatchLen,
1346 FileCheckRequest &Req,
1347 std::vector<FileCheckDiag> *Diags) {
1348 PrintMatch(ExpectedMatch, SM, CheckStr.Prefix, CheckStr.Loc, CheckStr.Pat,
1349 MatchedCount, Buffer, MatchPos, MatchLen, Req, Diags);
1352 static void PrintNoMatch(bool ExpectedMatch, const SourceMgr &SM,
1353 StringRef Prefix, SMLoc Loc,
1354 const FileCheckPattern &Pat, int MatchedCount,
1355 StringRef Buffer, bool VerboseVerbose,
1356 std::vector<FileCheckDiag> *Diags, Error MatchErrors) {
1357 assert(MatchErrors && "Called on successful match");
1358 bool PrintDiag = true;
1359 if (!ExpectedMatch) {
1360 if (!VerboseVerbose) {
1361 consumeError(std::move(MatchErrors));
1362 return;
1364 // Due to their verbosity, we don't print verbose diagnostics here if we're
1365 // gathering them for a different rendering, but we always print other
1366 // diagnostics.
1367 PrintDiag = !Diags;
1370 // If the current position is at the end of a line, advance to the start of
1371 // the next line.
1372 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t\n\r"));
1373 SMRange SearchRange = ProcessMatchResult(
1374 ExpectedMatch ? FileCheckDiag::MatchNoneButExpected
1375 : FileCheckDiag::MatchNoneAndExcluded,
1376 SM, Loc, Pat.getCheckTy(), Buffer, 0, Buffer.size(), Diags);
1377 if (!PrintDiag) {
1378 consumeError(std::move(MatchErrors));
1379 return;
1382 MatchErrors =
1383 handleErrors(std::move(MatchErrors),
1384 [](const FileCheckErrorDiagnostic &E) { E.log(errs()); });
1386 // No problem matching the string per se.
1387 if (!MatchErrors)
1388 return;
1389 consumeError(std::move(MatchErrors));
1391 // Print "not found" diagnostic.
1392 std::string Message = formatv("{0}: {1} string not found in input",
1393 Pat.getCheckTy().getDescription(Prefix),
1394 (ExpectedMatch ? "expected" : "excluded"))
1395 .str();
1396 if (Pat.getCount() > 1)
1397 Message += formatv(" ({0} out of {1})", MatchedCount, Pat.getCount()).str();
1398 SM.PrintMessage(
1399 Loc, ExpectedMatch ? SourceMgr::DK_Error : SourceMgr::DK_Remark, Message);
1401 // Print the "scanning from here" line.
1402 SM.PrintMessage(SearchRange.Start, SourceMgr::DK_Note, "scanning from here");
1404 // Allow the pattern to print additional information if desired.
1405 Pat.printSubstitutions(SM, Buffer);
1407 if (ExpectedMatch)
1408 Pat.printFuzzyMatch(SM, Buffer, Diags);
1411 static void PrintNoMatch(bool ExpectedMatch, const SourceMgr &SM,
1412 const FileCheckString &CheckStr, int MatchedCount,
1413 StringRef Buffer, bool VerboseVerbose,
1414 std::vector<FileCheckDiag> *Diags, Error MatchErrors) {
1415 PrintNoMatch(ExpectedMatch, SM, CheckStr.Prefix, CheckStr.Loc, CheckStr.Pat,
1416 MatchedCount, Buffer, VerboseVerbose, Diags,
1417 std::move(MatchErrors));
1420 /// Counts the number of newlines in the specified range.
1421 static unsigned CountNumNewlinesBetween(StringRef Range,
1422 const char *&FirstNewLine) {
1423 unsigned NumNewLines = 0;
1424 while (1) {
1425 // Scan for newline.
1426 Range = Range.substr(Range.find_first_of("\n\r"));
1427 if (Range.empty())
1428 return NumNewLines;
1430 ++NumNewLines;
1432 // Handle \n\r and \r\n as a single newline.
1433 if (Range.size() > 1 && (Range[1] == '\n' || Range[1] == '\r') &&
1434 (Range[0] != Range[1]))
1435 Range = Range.substr(1);
1436 Range = Range.substr(1);
1438 if (NumNewLines == 1)
1439 FirstNewLine = Range.begin();
1443 size_t FileCheckString::Check(const SourceMgr &SM, StringRef Buffer,
1444 bool IsLabelScanMode, size_t &MatchLen,
1445 FileCheckRequest &Req,
1446 std::vector<FileCheckDiag> *Diags) const {
1447 size_t LastPos = 0;
1448 std::vector<const FileCheckPattern *> NotStrings;
1450 // IsLabelScanMode is true when we are scanning forward to find CHECK-LABEL
1451 // bounds; we have not processed variable definitions within the bounded block
1452 // yet so cannot handle any final CHECK-DAG yet; this is handled when going
1453 // over the block again (including the last CHECK-LABEL) in normal mode.
1454 if (!IsLabelScanMode) {
1455 // Match "dag strings" (with mixed "not strings" if any).
1456 LastPos = CheckDag(SM, Buffer, NotStrings, Req, Diags);
1457 if (LastPos == StringRef::npos)
1458 return StringRef::npos;
1461 // Match itself from the last position after matching CHECK-DAG.
1462 size_t LastMatchEnd = LastPos;
1463 size_t FirstMatchPos = 0;
1464 // Go match the pattern Count times. Majority of patterns only match with
1465 // count 1 though.
1466 assert(Pat.getCount() != 0 && "pattern count can not be zero");
1467 for (int i = 1; i <= Pat.getCount(); i++) {
1468 StringRef MatchBuffer = Buffer.substr(LastMatchEnd);
1469 size_t CurrentMatchLen;
1470 // get a match at current start point
1471 Expected<size_t> MatchResult = Pat.match(MatchBuffer, CurrentMatchLen, SM);
1473 // report
1474 if (!MatchResult) {
1475 PrintNoMatch(true, SM, *this, i, MatchBuffer, Req.VerboseVerbose, Diags,
1476 MatchResult.takeError());
1477 return StringRef::npos;
1479 size_t MatchPos = *MatchResult;
1480 PrintMatch(true, SM, *this, i, MatchBuffer, MatchPos, CurrentMatchLen, Req,
1481 Diags);
1482 if (i == 1)
1483 FirstMatchPos = LastPos + MatchPos;
1485 // move start point after the match
1486 LastMatchEnd += MatchPos + CurrentMatchLen;
1488 // Full match len counts from first match pos.
1489 MatchLen = LastMatchEnd - FirstMatchPos;
1491 // Similar to the above, in "label-scan mode" we can't yet handle CHECK-NEXT
1492 // or CHECK-NOT
1493 if (!IsLabelScanMode) {
1494 size_t MatchPos = FirstMatchPos - LastPos;
1495 StringRef MatchBuffer = Buffer.substr(LastPos);
1496 StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
1498 // If this check is a "CHECK-NEXT", verify that the previous match was on
1499 // the previous line (i.e. that there is one newline between them).
1500 if (CheckNext(SM, SkippedRegion)) {
1501 ProcessMatchResult(FileCheckDiag::MatchFoundButWrongLine, SM, Loc,
1502 Pat.getCheckTy(), MatchBuffer, MatchPos, MatchLen,
1503 Diags, Req.Verbose);
1504 return StringRef::npos;
1507 // If this check is a "CHECK-SAME", verify that the previous match was on
1508 // the same line (i.e. that there is no newline between them).
1509 if (CheckSame(SM, SkippedRegion)) {
1510 ProcessMatchResult(FileCheckDiag::MatchFoundButWrongLine, SM, Loc,
1511 Pat.getCheckTy(), MatchBuffer, MatchPos, MatchLen,
1512 Diags, Req.Verbose);
1513 return StringRef::npos;
1516 // If this match had "not strings", verify that they don't exist in the
1517 // skipped region.
1518 if (CheckNot(SM, SkippedRegion, NotStrings, Req, Diags))
1519 return StringRef::npos;
1522 return FirstMatchPos;
1525 bool FileCheckString::CheckNext(const SourceMgr &SM, StringRef Buffer) const {
1526 if (Pat.getCheckTy() != Check::CheckNext &&
1527 Pat.getCheckTy() != Check::CheckEmpty)
1528 return false;
1530 Twine CheckName =
1531 Prefix +
1532 Twine(Pat.getCheckTy() == Check::CheckEmpty ? "-EMPTY" : "-NEXT");
1534 // Count the number of newlines between the previous match and this one.
1535 const char *FirstNewLine = nullptr;
1536 unsigned NumNewLines = CountNumNewlinesBetween(Buffer, FirstNewLine);
1538 if (NumNewLines == 0) {
1539 SM.PrintMessage(Loc, SourceMgr::DK_Error,
1540 CheckName + ": is on the same line as previous match");
1541 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()), SourceMgr::DK_Note,
1542 "'next' match was here");
1543 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
1544 "previous match ended here");
1545 return true;
1548 if (NumNewLines != 1) {
1549 SM.PrintMessage(Loc, SourceMgr::DK_Error,
1550 CheckName +
1551 ": is not on the line after the previous match");
1552 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()), SourceMgr::DK_Note,
1553 "'next' match was here");
1554 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
1555 "previous match ended here");
1556 SM.PrintMessage(SMLoc::getFromPointer(FirstNewLine), SourceMgr::DK_Note,
1557 "non-matching line after previous match is here");
1558 return true;
1561 return false;
1564 bool FileCheckString::CheckSame(const SourceMgr &SM, StringRef Buffer) const {
1565 if (Pat.getCheckTy() != Check::CheckSame)
1566 return false;
1568 // Count the number of newlines between the previous match and this one.
1569 const char *FirstNewLine = nullptr;
1570 unsigned NumNewLines = CountNumNewlinesBetween(Buffer, FirstNewLine);
1572 if (NumNewLines != 0) {
1573 SM.PrintMessage(Loc, SourceMgr::DK_Error,
1574 Prefix +
1575 "-SAME: is not on the same line as the previous match");
1576 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()), SourceMgr::DK_Note,
1577 "'next' match was here");
1578 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
1579 "previous match ended here");
1580 return true;
1583 return false;
1586 bool FileCheckString::CheckNot(
1587 const SourceMgr &SM, StringRef Buffer,
1588 const std::vector<const FileCheckPattern *> &NotStrings,
1589 const FileCheckRequest &Req, std::vector<FileCheckDiag> *Diags) const {
1590 for (const FileCheckPattern *Pat : NotStrings) {
1591 assert((Pat->getCheckTy() == Check::CheckNot) && "Expect CHECK-NOT!");
1593 size_t MatchLen = 0;
1594 Expected<size_t> MatchResult = Pat->match(Buffer, MatchLen, SM);
1596 if (!MatchResult) {
1597 PrintNoMatch(false, SM, Prefix, Pat->getLoc(), *Pat, 1, Buffer,
1598 Req.VerboseVerbose, Diags, MatchResult.takeError());
1599 continue;
1601 size_t Pos = *MatchResult;
1603 PrintMatch(false, SM, Prefix, Pat->getLoc(), *Pat, 1, Buffer, Pos, MatchLen,
1604 Req, Diags);
1606 return true;
1609 return false;
1612 size_t
1613 FileCheckString::CheckDag(const SourceMgr &SM, StringRef Buffer,
1614 std::vector<const FileCheckPattern *> &NotStrings,
1615 const FileCheckRequest &Req,
1616 std::vector<FileCheckDiag> *Diags) const {
1617 if (DagNotStrings.empty())
1618 return 0;
1620 // The start of the search range.
1621 size_t StartPos = 0;
1623 struct MatchRange {
1624 size_t Pos;
1625 size_t End;
1627 // A sorted list of ranges for non-overlapping CHECK-DAG matches. Match
1628 // ranges are erased from this list once they are no longer in the search
1629 // range.
1630 std::list<MatchRange> MatchRanges;
1632 // We need PatItr and PatEnd later for detecting the end of a CHECK-DAG
1633 // group, so we don't use a range-based for loop here.
1634 for (auto PatItr = DagNotStrings.begin(), PatEnd = DagNotStrings.end();
1635 PatItr != PatEnd; ++PatItr) {
1636 const FileCheckPattern &Pat = *PatItr;
1637 assert((Pat.getCheckTy() == Check::CheckDAG ||
1638 Pat.getCheckTy() == Check::CheckNot) &&
1639 "Invalid CHECK-DAG or CHECK-NOT!");
1641 if (Pat.getCheckTy() == Check::CheckNot) {
1642 NotStrings.push_back(&Pat);
1643 continue;
1646 assert((Pat.getCheckTy() == Check::CheckDAG) && "Expect CHECK-DAG!");
1648 // CHECK-DAG always matches from the start.
1649 size_t MatchLen = 0, MatchPos = StartPos;
1651 // Search for a match that doesn't overlap a previous match in this
1652 // CHECK-DAG group.
1653 for (auto MI = MatchRanges.begin(), ME = MatchRanges.end(); true; ++MI) {
1654 StringRef MatchBuffer = Buffer.substr(MatchPos);
1655 Expected<size_t> MatchResult = Pat.match(MatchBuffer, MatchLen, SM);
1656 // With a group of CHECK-DAGs, a single mismatching means the match on
1657 // that group of CHECK-DAGs fails immediately.
1658 if (!MatchResult) {
1659 PrintNoMatch(true, SM, Prefix, Pat.getLoc(), Pat, 1, MatchBuffer,
1660 Req.VerboseVerbose, Diags, MatchResult.takeError());
1661 return StringRef::npos;
1663 size_t MatchPosBuf = *MatchResult;
1664 // Re-calc it as the offset relative to the start of the original string.
1665 MatchPos += MatchPosBuf;
1666 if (Req.VerboseVerbose)
1667 PrintMatch(true, SM, Prefix, Pat.getLoc(), Pat, 1, Buffer, MatchPos,
1668 MatchLen, Req, Diags);
1669 MatchRange M{MatchPos, MatchPos + MatchLen};
1670 if (Req.AllowDeprecatedDagOverlap) {
1671 // We don't need to track all matches in this mode, so we just maintain
1672 // one match range that encompasses the current CHECK-DAG group's
1673 // matches.
1674 if (MatchRanges.empty())
1675 MatchRanges.insert(MatchRanges.end(), M);
1676 else {
1677 auto Block = MatchRanges.begin();
1678 Block->Pos = std::min(Block->Pos, M.Pos);
1679 Block->End = std::max(Block->End, M.End);
1681 break;
1683 // Iterate previous matches until overlapping match or insertion point.
1684 bool Overlap = false;
1685 for (; MI != ME; ++MI) {
1686 if (M.Pos < MI->End) {
1687 // !Overlap => New match has no overlap and is before this old match.
1688 // Overlap => New match overlaps this old match.
1689 Overlap = MI->Pos < M.End;
1690 break;
1693 if (!Overlap) {
1694 // Insert non-overlapping match into list.
1695 MatchRanges.insert(MI, M);
1696 break;
1698 if (Req.VerboseVerbose) {
1699 // Due to their verbosity, we don't print verbose diagnostics here if
1700 // we're gathering them for a different rendering, but we always print
1701 // other diagnostics.
1702 if (!Diags) {
1703 SMLoc OldStart = SMLoc::getFromPointer(Buffer.data() + MI->Pos);
1704 SMLoc OldEnd = SMLoc::getFromPointer(Buffer.data() + MI->End);
1705 SMRange OldRange(OldStart, OldEnd);
1706 SM.PrintMessage(OldStart, SourceMgr::DK_Note,
1707 "match discarded, overlaps earlier DAG match here",
1708 {OldRange});
1709 } else
1710 Diags->rbegin()->MatchTy = FileCheckDiag::MatchFoundButDiscarded;
1712 MatchPos = MI->End;
1714 if (!Req.VerboseVerbose)
1715 PrintMatch(true, SM, Prefix, Pat.getLoc(), Pat, 1, Buffer, MatchPos,
1716 MatchLen, Req, Diags);
1718 // Handle the end of a CHECK-DAG group.
1719 if (std::next(PatItr) == PatEnd ||
1720 std::next(PatItr)->getCheckTy() == Check::CheckNot) {
1721 if (!NotStrings.empty()) {
1722 // If there are CHECK-NOTs between two CHECK-DAGs or from CHECK to
1723 // CHECK-DAG, verify that there are no 'not' strings occurred in that
1724 // region.
1725 StringRef SkippedRegion =
1726 Buffer.slice(StartPos, MatchRanges.begin()->Pos);
1727 if (CheckNot(SM, SkippedRegion, NotStrings, Req, Diags))
1728 return StringRef::npos;
1729 // Clear "not strings".
1730 NotStrings.clear();
1732 // All subsequent CHECK-DAGs and CHECK-NOTs should be matched from the
1733 // end of this CHECK-DAG group's match range.
1734 StartPos = MatchRanges.rbegin()->End;
1735 // Don't waste time checking for (impossible) overlaps before that.
1736 MatchRanges.clear();
1740 return StartPos;
1743 // A check prefix must contain only alphanumeric, hyphens and underscores.
1744 static bool ValidateCheckPrefix(StringRef CheckPrefix) {
1745 Regex Validator("^[a-zA-Z0-9_-]*$");
1746 return Validator.match(CheckPrefix);
1749 bool FileCheck::ValidateCheckPrefixes() {
1750 StringSet<> PrefixSet;
1752 for (StringRef Prefix : Req.CheckPrefixes) {
1753 // Reject empty prefixes.
1754 if (Prefix == "")
1755 return false;
1757 if (!PrefixSet.insert(Prefix).second)
1758 return false;
1760 if (!ValidateCheckPrefix(Prefix))
1761 return false;
1764 return true;
1767 Regex FileCheck::buildCheckPrefixRegex() {
1768 // I don't think there's a way to specify an initial value for cl::list,
1769 // so if nothing was specified, add the default
1770 if (Req.CheckPrefixes.empty())
1771 Req.CheckPrefixes.push_back("CHECK");
1773 // We already validated the contents of CheckPrefixes so just concatenate
1774 // them as alternatives.
1775 SmallString<32> PrefixRegexStr;
1776 for (StringRef Prefix : Req.CheckPrefixes) {
1777 if (Prefix != Req.CheckPrefixes.front())
1778 PrefixRegexStr.push_back('|');
1780 PrefixRegexStr.append(Prefix);
1783 return Regex(PrefixRegexStr);
1786 Error FileCheckPatternContext::defineCmdlineVariables(
1787 std::vector<std::string> &CmdlineDefines, SourceMgr &SM) {
1788 assert(GlobalVariableTable.empty() && GlobalNumericVariableTable.empty() &&
1789 "Overriding defined variable with command-line variable definitions");
1791 if (CmdlineDefines.empty())
1792 return Error::success();
1794 // Create a string representing the vector of command-line definitions. Each
1795 // definition is on its own line and prefixed with a definition number to
1796 // clarify which definition a given diagnostic corresponds to.
1797 unsigned I = 0;
1798 Error Errs = Error::success();
1799 std::string CmdlineDefsDiag;
1800 SmallVector<std::pair<size_t, size_t>, 4> CmdlineDefsIndices;
1801 for (StringRef CmdlineDef : CmdlineDefines) {
1802 std::string DefPrefix = ("Global define #" + Twine(++I) + ": ").str();
1803 size_t EqIdx = CmdlineDef.find('=');
1804 if (EqIdx == StringRef::npos) {
1805 CmdlineDefsIndices.push_back(std::make_pair(CmdlineDefsDiag.size(), 0));
1806 continue;
1808 // Numeric variable definition.
1809 if (CmdlineDef[0] == '#') {
1810 // Append a copy of the command-line definition adapted to use the same
1811 // format as in the input file to be able to reuse
1812 // parseNumericSubstitutionBlock.
1813 CmdlineDefsDiag += (DefPrefix + CmdlineDef + " (parsed as: [[").str();
1814 std::string SubstitutionStr = CmdlineDef;
1815 SubstitutionStr[EqIdx] = ':';
1816 CmdlineDefsIndices.push_back(
1817 std::make_pair(CmdlineDefsDiag.size(), SubstitutionStr.size()));
1818 CmdlineDefsDiag += (SubstitutionStr + Twine("]])\n")).str();
1819 } else {
1820 CmdlineDefsDiag += DefPrefix;
1821 CmdlineDefsIndices.push_back(
1822 std::make_pair(CmdlineDefsDiag.size(), CmdlineDef.size()));
1823 CmdlineDefsDiag += (CmdlineDef + "\n").str();
1827 // Create a buffer with fake command line content in order to display
1828 // parsing diagnostic with location information and point to the
1829 // global definition with invalid syntax.
1830 std::unique_ptr<MemoryBuffer> CmdLineDefsDiagBuffer =
1831 MemoryBuffer::getMemBufferCopy(CmdlineDefsDiag, "Global defines");
1832 StringRef CmdlineDefsDiagRef = CmdLineDefsDiagBuffer->getBuffer();
1833 SM.AddNewSourceBuffer(std::move(CmdLineDefsDiagBuffer), SMLoc());
1835 for (std::pair<size_t, size_t> CmdlineDefIndices : CmdlineDefsIndices) {
1836 StringRef CmdlineDef = CmdlineDefsDiagRef.substr(CmdlineDefIndices.first,
1837 CmdlineDefIndices.second);
1838 if (CmdlineDef.empty()) {
1839 Errs = joinErrors(
1840 std::move(Errs),
1841 FileCheckErrorDiagnostic::get(
1842 SM, CmdlineDef, "missing equal sign in global definition"));
1843 continue;
1846 // Numeric variable definition.
1847 if (CmdlineDef[0] == '#') {
1848 // Now parse the definition both to check that the syntax is correct and
1849 // to create the necessary class instance.
1850 StringRef CmdlineDefExpr = CmdlineDef.substr(1);
1851 Optional<FileCheckNumericVariable *> DefinedNumericVariable;
1852 Expected<std::unique_ptr<FileCheckExpressionAST>> ExpressionASTResult =
1853 FileCheckPattern::parseNumericSubstitutionBlock(
1854 CmdlineDefExpr, DefinedNumericVariable, false, None, this, SM);
1855 if (!ExpressionASTResult) {
1856 Errs = joinErrors(std::move(Errs), ExpressionASTResult.takeError());
1857 continue;
1859 std::unique_ptr<FileCheckExpressionAST> ExpressionAST =
1860 std::move(*ExpressionASTResult);
1861 // Now evaluate the expression whose value this variable should be set
1862 // to, since the expression of a command-line variable definition should
1863 // only use variables defined earlier on the command-line. If not, this
1864 // is an error and we report it.
1865 Expected<uint64_t> Value = ExpressionAST->eval();
1866 if (!Value) {
1867 Errs = joinErrors(std::move(Errs), Value.takeError());
1868 continue;
1871 assert(DefinedNumericVariable && "No variable defined");
1872 (*DefinedNumericVariable)->setValue(*Value);
1874 // Record this variable definition.
1875 GlobalNumericVariableTable[(*DefinedNumericVariable)->getName()] =
1876 *DefinedNumericVariable;
1877 } else {
1878 // String variable definition.
1879 std::pair<StringRef, StringRef> CmdlineNameVal = CmdlineDef.split('=');
1880 StringRef CmdlineName = CmdlineNameVal.first;
1881 StringRef OrigCmdlineName = CmdlineName;
1882 Expected<FileCheckPattern::VariableProperties> ParseVarResult =
1883 FileCheckPattern::parseVariable(CmdlineName, SM);
1884 if (!ParseVarResult) {
1885 Errs = joinErrors(std::move(Errs), ParseVarResult.takeError());
1886 continue;
1888 // Check that CmdlineName does not denote a pseudo variable is only
1889 // composed of the parsed numeric variable. This catches cases like
1890 // "FOO+2" in a "FOO+2=10" definition.
1891 if (ParseVarResult->IsPseudo || !CmdlineName.empty()) {
1892 Errs = joinErrors(std::move(Errs),
1893 FileCheckErrorDiagnostic::get(
1894 SM, OrigCmdlineName,
1895 "invalid name in string variable definition '" +
1896 OrigCmdlineName + "'"));
1897 continue;
1899 StringRef Name = ParseVarResult->Name;
1901 // Detect collisions between string and numeric variables when the former
1902 // is created later than the latter.
1903 if (GlobalNumericVariableTable.find(Name) !=
1904 GlobalNumericVariableTable.end()) {
1905 Errs = joinErrors(std::move(Errs), FileCheckErrorDiagnostic::get(
1906 SM, Name,
1907 "numeric variable with name '" +
1908 Name + "' already exists"));
1909 continue;
1911 GlobalVariableTable.insert(CmdlineNameVal);
1912 // Mark the string variable as defined to detect collisions between
1913 // string and numeric variables in defineCmdlineVariables when the latter
1914 // is created later than the former. We cannot reuse GlobalVariableTable
1915 // for this by populating it with an empty string since we would then
1916 // lose the ability to detect the use of an undefined variable in
1917 // match().
1918 DefinedVariableTable[Name] = true;
1922 return Errs;
1925 void FileCheckPatternContext::clearLocalVars() {
1926 SmallVector<StringRef, 16> LocalPatternVars, LocalNumericVars;
1927 for (const StringMapEntry<StringRef> &Var : GlobalVariableTable)
1928 if (Var.first()[0] != '$')
1929 LocalPatternVars.push_back(Var.first());
1931 // Numeric substitution reads the value of a variable directly, not via
1932 // GlobalNumericVariableTable. Therefore, we clear local variables by
1933 // clearing their value which will lead to a numeric substitution failure. We
1934 // also mark the variable for removal from GlobalNumericVariableTable since
1935 // this is what defineCmdlineVariables checks to decide that no global
1936 // variable has been defined.
1937 for (const auto &Var : GlobalNumericVariableTable)
1938 if (Var.first()[0] != '$') {
1939 Var.getValue()->clearValue();
1940 LocalNumericVars.push_back(Var.first());
1943 for (const auto &Var : LocalPatternVars)
1944 GlobalVariableTable.erase(Var);
1945 for (const auto &Var : LocalNumericVars)
1946 GlobalNumericVariableTable.erase(Var);
1949 bool FileCheck::CheckInput(SourceMgr &SM, StringRef Buffer,
1950 ArrayRef<FileCheckString> CheckStrings,
1951 std::vector<FileCheckDiag> *Diags) {
1952 bool ChecksFailed = false;
1954 unsigned i = 0, j = 0, e = CheckStrings.size();
1955 while (true) {
1956 StringRef CheckRegion;
1957 if (j == e) {
1958 CheckRegion = Buffer;
1959 } else {
1960 const FileCheckString &CheckLabelStr = CheckStrings[j];
1961 if (CheckLabelStr.Pat.getCheckTy() != Check::CheckLabel) {
1962 ++j;
1963 continue;
1966 // Scan to next CHECK-LABEL match, ignoring CHECK-NOT and CHECK-DAG
1967 size_t MatchLabelLen = 0;
1968 size_t MatchLabelPos =
1969 CheckLabelStr.Check(SM, Buffer, true, MatchLabelLen, Req, Diags);
1970 if (MatchLabelPos == StringRef::npos)
1971 // Immediately bail if CHECK-LABEL fails, nothing else we can do.
1972 return false;
1974 CheckRegion = Buffer.substr(0, MatchLabelPos + MatchLabelLen);
1975 Buffer = Buffer.substr(MatchLabelPos + MatchLabelLen);
1976 ++j;
1979 // Do not clear the first region as it's the one before the first
1980 // CHECK-LABEL and it would clear variables defined on the command-line
1981 // before they get used.
1982 if (i != 0 && Req.EnableVarScope)
1983 PatternContext.clearLocalVars();
1985 for (; i != j; ++i) {
1986 const FileCheckString &CheckStr = CheckStrings[i];
1988 // Check each string within the scanned region, including a second check
1989 // of any final CHECK-LABEL (to verify CHECK-NOT and CHECK-DAG)
1990 size_t MatchLen = 0;
1991 size_t MatchPos =
1992 CheckStr.Check(SM, CheckRegion, false, MatchLen, Req, Diags);
1994 if (MatchPos == StringRef::npos) {
1995 ChecksFailed = true;
1996 i = j;
1997 break;
2000 CheckRegion = CheckRegion.substr(MatchPos + MatchLen);
2003 if (j == e)
2004 break;
2007 // Success if no checks failed.
2008 return !ChecksFailed;