LLVM  9.0.0svn
FileCheck.cpp
Go to the documentation of this file.
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/Support/FileCheck.h"
17 #include "llvm/ADT/StringSet.h"
19 #include <cstdint>
20 #include <list>
21 #include <map>
22 #include <tuple>
23 #include <utility>
24 
25 using namespace llvm;
26 
27 void FileCheckNumericVariable::setValue(uint64_t NewValue) {
28  assert(!Value && "Overwriting numeric variable's value is not allowed");
29  Value = NewValue;
30 }
31 
33  if (!Value)
34  return;
35  Value = None;
36 }
37 
39  Optional<uint64_t> Value = NumericVariable->getValue();
40  if (Value)
41  return *Value;
42  return make_error<FileCheckUndefVarError>(Name);
43 }
44 
46  Expected<uint64_t> LeftOp = LeftOperand->eval();
47  Expected<uint64_t> RightOp = RightOperand->eval();
48 
49  // Bubble up any error (e.g. undefined variables) in the recursive
50  // evaluation.
51  if (!LeftOp || !RightOp) {
52  Error Err = Error::success();
53  if (!LeftOp)
54  Err = joinErrors(std::move(Err), LeftOp.takeError());
55  if (!RightOp)
56  Err = joinErrors(std::move(Err), RightOp.takeError());
57  return std::move(Err);
58  }
59 
60  return EvalBinop(*LeftOp, *RightOp);
61 }
62 
64  Expected<uint64_t> EvaluatedValue = ExpressionAST->eval();
65  if (!EvaluatedValue)
66  return EvaluatedValue.takeError();
67  return utostr(*EvaluatedValue);
68 }
69 
71  // Look up the value and escape it so that we can put it into the regex.
72  Expected<StringRef> VarVal = Context->getPatternVarValue(FromStr);
73  if (!VarVal)
74  return VarVal.takeError();
75  return Regex::escape(*VarVal);
76 }
77 
79  return C == '_' || isalpha(C);
80 }
81 
84  if (Str.empty())
85  return FileCheckErrorDiagnostic::get(SM, Str, "empty variable name");
86 
87  bool ParsedOneChar = false;
88  unsigned I = 0;
89  bool IsPseudo = Str[0] == '@';
90 
91  // Global vars start with '$'.
92  if (Str[0] == '$' || IsPseudo)
93  ++I;
94 
95  for (unsigned E = Str.size(); I != E; ++I) {
96  if (!ParsedOneChar && !isValidVarNameStart(Str[I]))
97  return FileCheckErrorDiagnostic::get(SM, Str, "invalid variable name");
98 
99  // Variable names are composed of alphanumeric characters and underscores.
100  if (Str[I] != '_' && !isalnum(Str[I]))
101  break;
102  ParsedOneChar = true;
103  }
104 
105  StringRef Name = Str.take_front(I);
106  Str = Str.substr(I);
107  return VariableProperties {Name, IsPseudo};
108 }
109 
110 // StringRef holding all characters considered as horizontal whitespaces by
111 // FileCheck input canonicalization.
113 
114 // Parsing helper function that strips the first character in S and returns it.
115 static char popFront(StringRef &S) {
116  char C = S.front();
117  S = S.drop_front();
118  return C;
119 }
120 
124 
128  Optional<size_t> LineNumber, const SourceMgr &SM) {
129  Expected<VariableProperties> ParseVarResult = parseVariable(Expr, SM);
130  if (!ParseVarResult)
131  return ParseVarResult.takeError();
132  StringRef Name = ParseVarResult->Name;
133 
134  if (ParseVarResult->IsPseudo)
136  SM, Name, "definition of pseudo numeric variable unsupported");
137 
138  // Detect collisions between string and numeric variables when the latter
139  // is created later than the former.
140  if (Context->DefinedVariableTable.find(Name) !=
141  Context->DefinedVariableTable.end())
143  SM, Name, "string variable with name '" + Name + "' already exists");
144 
145  Expr = Expr.ltrim(SpaceChars);
146  if (!Expr.empty())
148  SM, Expr, "unexpected characters after numeric variable name");
149 
150  FileCheckNumericVariable *DefinedNumericVariable;
151  auto VarTableIter = Context->GlobalNumericVariableTable.find(Name);
152  if (VarTableIter != Context->GlobalNumericVariableTable.end())
153  DefinedNumericVariable = VarTableIter->second;
154  else
155  DefinedNumericVariable = Context->makeNumericVariable(Name, LineNumber);
156 
157  return DefinedNumericVariable;
158 }
159 
161 FileCheckPattern::parseNumericVariableUse(StringRef Name, bool IsPseudo,
162  const SourceMgr &SM) const {
163  if (IsPseudo && !Name.equals("@LINE"))
165  SM, Name, "invalid pseudo numeric variable '" + Name + "'");
166 
167  // Numeric variable definitions and uses are parsed in the order in which
168  // they appear in the CHECK patterns. For each definition, the pointer to the
169  // class instance of the corresponding numeric variable definition is stored
170  // in GlobalNumericVariableTable in parsePattern. Therefore, if the pointer
171  // we get below is null, it means no such variable was defined before. When
172  // that happens, we create a dummy variable so that parsing can continue. All
173  // uses of undefined variables, whether string or numeric, are then diagnosed
174  // in printSubstitutions() after failing to match.
175  auto VarTableIter = Context->GlobalNumericVariableTable.find(Name);
176  FileCheckNumericVariable *NumericVariable;
177  if (VarTableIter != Context->GlobalNumericVariableTable.end())
178  NumericVariable = VarTableIter->second;
179  else {
180  NumericVariable = Context->makeNumericVariable(Name);
181  Context->GlobalNumericVariableTable[Name] = NumericVariable;
182  }
183 
184  Optional<size_t> DefLineNumber = NumericVariable->getDefLineNumber();
185  if (DefLineNumber && LineNumber && *DefLineNumber == *LineNumber)
187  SM, Name,
188  "numeric variable '" + Name + "' defined on the same line as used");
189 
190  return llvm::make_unique<FileCheckNumericVariableUse>(Name, NumericVariable);
191 }
192 
194 FileCheckPattern::parseNumericOperand(StringRef &Expr, AllowedOperand AO,
195  const SourceMgr &SM) const {
196  if (AO == AllowedOperand::LineVar || AO == AllowedOperand::Any) {
197  // Try to parse as a numeric variable use.
199  parseVariable(Expr, SM);
200  if (ParseVarResult)
201  return parseNumericVariableUse(ParseVarResult->Name,
202  ParseVarResult->IsPseudo, SM);
203  if (AO == AllowedOperand::LineVar)
204  return ParseVarResult.takeError();
205  // Ignore the error and retry parsing as a literal.
206  consumeError(ParseVarResult.takeError());
207  }
208 
209  // Otherwise, parse it as a literal.
210  uint64_t LiteralValue;
211  if (!Expr.consumeInteger(/*Radix=*/10, LiteralValue))
212  return llvm::make_unique<FileCheckExpressionLiteral>(LiteralValue);
213 
214  return FileCheckErrorDiagnostic::get(SM, Expr,
215  "invalid operand format '" + Expr + "'");
216 }
217 
218 static uint64_t add(uint64_t LeftOp, uint64_t RightOp) {
219  return LeftOp + RightOp;
220 }
221 
222 static uint64_t sub(uint64_t LeftOp, uint64_t RightOp) {
223  return LeftOp - RightOp;
224 }
225 
227 FileCheckPattern::parseBinop(StringRef &Expr,
228  std::unique_ptr<FileCheckExpressionAST> LeftOp,
229  bool IsLegacyLineExpr, const SourceMgr &SM) const {
230  Expr = Expr.ltrim(SpaceChars);
231  if (Expr.empty())
232  return std::move(LeftOp);
233 
234  // Check if this is a supported operation and select a function to perform
235  // it.
236  SMLoc OpLoc = SMLoc::getFromPointer(Expr.data());
237  char Operator = popFront(Expr);
238  binop_eval_t EvalBinop;
239  switch (Operator) {
240  case '+':
241  EvalBinop = add;
242  break;
243  case '-':
244  EvalBinop = sub;
245  break;
246  default:
248  SM, OpLoc, Twine("unsupported operation '") + Twine(Operator) + "'");
249  }
250 
251  // Parse right operand.
252  Expr = Expr.ltrim(SpaceChars);
253  if (Expr.empty())
254  return FileCheckErrorDiagnostic::get(SM, Expr,
255  "missing operand in expression");
256  // The second operand in a legacy @LINE expression is always a literal.
257  AllowedOperand AO =
258  IsLegacyLineExpr ? AllowedOperand::Literal : AllowedOperand::Any;
260  parseNumericOperand(Expr, AO, SM);
261  if (!RightOpResult)
262  return RightOpResult;
263 
264  Expr = Expr.ltrim(SpaceChars);
265  return llvm::make_unique<FileCheckASTBinop>(EvalBinop, std::move(LeftOp),
266  std::move(*RightOpResult));
267 }
268 
271  StringRef Expr,
272  Optional<FileCheckNumericVariable *> &DefinedNumericVariable,
273  bool IsLegacyLineExpr, const SourceMgr &SM) const {
274  // Parse the numeric variable definition.
275  DefinedNumericVariable = None;
276  size_t DefEnd = Expr.find(':');
277  if (DefEnd != StringRef::npos) {
278  StringRef DefExpr = Expr.substr(0, DefEnd);
279  StringRef UseExpr = Expr.substr(DefEnd + 1);
280 
281  UseExpr = UseExpr.ltrim(SpaceChars);
282  if (!UseExpr.empty())
284  SM, UseExpr,
285  "unexpected string after variable definition: '" + UseExpr + "'");
286 
287  DefExpr = DefExpr.ltrim(SpaceChars);
289  parseNumericVariableDefinition(DefExpr, Context, LineNumber, SM);
290  if (!ParseResult)
291  return ParseResult.takeError();
292  DefinedNumericVariable = *ParseResult;
293 
294  return nullptr;
295  }
296 
297  // Parse the expression itself.
298  Expr = Expr.ltrim(SpaceChars);
299  // The first operand in a legacy @LINE expression is always the @LINE pseudo
300  // variable.
301  AllowedOperand AO =
302  IsLegacyLineExpr ? AllowedOperand::LineVar : AllowedOperand::Any;
304  parseNumericOperand(Expr, AO, SM);
305  while (ParseResult && !Expr.empty()) {
306  ParseResult =
307  parseBinop(Expr, std::move(*ParseResult), IsLegacyLineExpr, SM);
308  // Legacy @LINE expressions only allow 2 operands.
309  if (ParseResult && IsLegacyLineExpr && !Expr.empty())
311  SM, Expr,
312  "unexpected characters at end of expression '" + Expr + "'");
313  }
314  if (!ParseResult)
315  return ParseResult;
316  return std::move(*ParseResult);
317 }
318 
320  SourceMgr &SM,
321  const FileCheckRequest &Req) {
322  bool MatchFullLinesHere = Req.MatchFullLines && CheckTy != Check::CheckNot;
323 
324  PatternLoc = SMLoc::getFromPointer(PatternStr.data());
325 
326  if (!(Req.NoCanonicalizeWhiteSpace && Req.MatchFullLines))
327  // Ignore trailing whitespace.
328  while (!PatternStr.empty() &&
329  (PatternStr.back() == ' ' || PatternStr.back() == '\t'))
330  PatternStr = PatternStr.substr(0, PatternStr.size() - 1);
331 
332  // Check that there is something on the line.
333  if (PatternStr.empty() && CheckTy != Check::CheckEmpty) {
334  SM.PrintMessage(PatternLoc, SourceMgr::DK_Error,
335  "found empty check string with prefix '" + Prefix + ":'");
336  return true;
337  }
338 
339  if (!PatternStr.empty() && CheckTy == Check::CheckEmpty) {
340  SM.PrintMessage(
341  PatternLoc, SourceMgr::DK_Error,
342  "found non-empty check string for empty check with prefix '" + Prefix +
343  ":'");
344  return true;
345  }
346 
347  if (CheckTy == Check::CheckEmpty) {
348  RegExStr = "(\n$)";
349  return false;
350  }
351 
352  // Check to see if this is a fixed string, or if it has regex pieces.
353  if (!MatchFullLinesHere &&
354  (PatternStr.size() < 2 || (PatternStr.find("{{") == StringRef::npos &&
355  PatternStr.find("[[") == StringRef::npos))) {
356  FixedStr = PatternStr;
357  return false;
358  }
359 
360  if (MatchFullLinesHere) {
361  RegExStr += '^';
362  if (!Req.NoCanonicalizeWhiteSpace)
363  RegExStr += " *";
364  }
365 
366  // Paren value #0 is for the fully matched string. Any new parenthesized
367  // values add from there.
368  unsigned CurParen = 1;
369 
370  // Otherwise, there is at least one regex piece. Build up the regex pattern
371  // by escaping scary characters in fixed strings, building up one big regex.
372  while (!PatternStr.empty()) {
373  // RegEx matches.
374  if (PatternStr.startswith("{{")) {
375  // This is the start of a regex match. Scan for the }}.
376  size_t End = PatternStr.find("}}");
377  if (End == StringRef::npos) {
378  SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
380  "found start of regex string with no end '}}'");
381  return true;
382  }
383 
384  // Enclose {{}} patterns in parens just like [[]] even though we're not
385  // capturing the result for any purpose. This is required in case the
386  // expression contains an alternation like: CHECK: abc{{x|z}}def. We
387  // want this to turn into: "abc(x|z)def" not "abcx|zdef".
388  RegExStr += '(';
389  ++CurParen;
390 
391  if (AddRegExToRegEx(PatternStr.substr(2, End - 2), CurParen, SM))
392  return true;
393  RegExStr += ')';
394 
395  PatternStr = PatternStr.substr(End + 2);
396  continue;
397  }
398 
399  // String and numeric substitution blocks. String substitution blocks come
400  // in two forms: [[foo:.*]] and [[foo]]. The former matches .* (or some
401  // other regex) and assigns it to the string variable 'foo'. The latter
402  // substitutes foo's value. Numeric substitution blocks work the same way
403  // as string ones, but start with a '#' sign after the double brackets.
404  // Both string and numeric variable names must satisfy the regular
405  // expression "[a-zA-Z_][0-9a-zA-Z_]*" to be valid, as this helps catch
406  // some common errors.
407  if (PatternStr.startswith("[[")) {
408  StringRef UnparsedPatternStr = PatternStr.substr(2);
409  // Find the closing bracket pair ending the match. End is going to be an
410  // offset relative to the beginning of the match string.
411  size_t End = FindRegexVarEnd(UnparsedPatternStr, SM);
412  StringRef MatchStr = UnparsedPatternStr.substr(0, End);
413  bool IsNumBlock = MatchStr.consume_front("#");
414 
415  if (End == StringRef::npos) {
416  SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
418  "Invalid substitution block, no ]] found");
419  return true;
420  }
421  // Strip the substitution block we are parsing. End points to the start
422  // of the "]]" closing the expression so account for it in computing the
423  // index of the first unparsed character.
424  PatternStr = UnparsedPatternStr.substr(End + 2);
425 
426  bool IsDefinition = false;
427  // Whether the substitution block is a legacy use of @LINE with string
428  // substitution block syntax.
429  bool IsLegacyLineExpr = false;
430  StringRef DefName;
431  StringRef SubstStr;
432  StringRef MatchRegexp;
433  size_t SubstInsertIdx = RegExStr.size();
434 
435  // Parse string variable or legacy @LINE expression.
436  if (!IsNumBlock) {
437  size_t VarEndIdx = MatchStr.find(":");
438  size_t SpacePos = MatchStr.substr(0, VarEndIdx).find_first_of(" \t");
439  if (SpacePos != StringRef::npos) {
440  SM.PrintMessage(SMLoc::getFromPointer(MatchStr.data() + SpacePos),
441  SourceMgr::DK_Error, "unexpected whitespace");
442  return true;
443  }
444 
445  // Get the name (e.g. "foo") and verify it is well formed.
446  StringRef OrigMatchStr = MatchStr;
448  parseVariable(MatchStr, SM);
449  if (!ParseVarResult) {
450  logAllUnhandledErrors(ParseVarResult.takeError(), errs());
451  return true;
452  }
453  StringRef Name = ParseVarResult->Name;
454  bool IsPseudo = ParseVarResult->IsPseudo;
455 
456  IsDefinition = (VarEndIdx != StringRef::npos);
457  if (IsDefinition) {
458  if ((IsPseudo || !MatchStr.consume_front(":"))) {
461  "invalid name in string variable definition");
462  return true;
463  }
464 
465  // Detect collisions between string and numeric variables when the
466  // former is created later than the latter.
467  if (Context->GlobalNumericVariableTable.find(Name) !=
468  Context->GlobalNumericVariableTable.end()) {
469  SM.PrintMessage(
471  "numeric variable with name '" + Name + "' already exists");
472  return true;
473  }
474  DefName = Name;
475  MatchRegexp = MatchStr;
476  } else {
477  if (IsPseudo) {
478  MatchStr = OrigMatchStr;
479  IsLegacyLineExpr = IsNumBlock = true;
480  } else
481  SubstStr = Name;
482  }
483  }
484 
485  // Parse numeric substitution block.
486  std::unique_ptr<FileCheckExpressionAST> ExpressionAST;
487  Optional<FileCheckNumericVariable *> DefinedNumericVariable;
488  if (IsNumBlock) {
490  parseNumericSubstitutionBlock(MatchStr, DefinedNumericVariable,
491  IsLegacyLineExpr, SM);
492  if (!ParseResult) {
493  logAllUnhandledErrors(ParseResult.takeError(), errs());
494  return true;
495  }
496  ExpressionAST = std::move(*ParseResult);
497  if (DefinedNumericVariable) {
498  IsDefinition = true;
499  DefName = (*DefinedNumericVariable)->getName();
500  MatchRegexp = StringRef("[0-9]+");
501  } else
502  SubstStr = MatchStr;
503  }
504 
505  // Handle substitutions: [[foo]] and [[#<foo expr>]].
506  if (!IsDefinition) {
507  // Handle substitution of string variables that were defined earlier on
508  // the same line by emitting a backreference. Expressions do not
509  // support substituting a numeric variable defined on the same line.
510  if (!IsNumBlock && VariableDefs.find(SubstStr) != VariableDefs.end()) {
511  unsigned CaptureParenGroup = VariableDefs[SubstStr];
512  if (CaptureParenGroup < 1 || CaptureParenGroup > 9) {
513  SM.PrintMessage(SMLoc::getFromPointer(SubstStr.data()),
515  "Can't back-reference more than 9 variables");
516  return true;
517  }
518  AddBackrefToRegEx(CaptureParenGroup);
519  } else {
520  // Handle substitution of string variables ([[<var>]]) defined in
521  // previous CHECK patterns, and substitution of expressions.
522  FileCheckSubstitution *Substitution =
523  IsNumBlock
524  ? Context->makeNumericSubstitution(
525  SubstStr, std::move(ExpressionAST), SubstInsertIdx)
526  : Context->makeStringSubstitution(SubstStr, SubstInsertIdx);
527  Substitutions.push_back(Substitution);
528  }
529  continue;
530  }
531 
532  // Handle variable definitions: [[<def>:(...)]] and
533  // [[#(...)<def>:(...)]].
534  if (IsNumBlock) {
535  FileCheckNumericVariableMatch NumericVariableDefinition = {
536  *DefinedNumericVariable, CurParen};
537  NumericVariableDefs[DefName] = NumericVariableDefinition;
538  // This store is done here rather than in match() to allow
539  // parseNumericVariableUse() to get the pointer to the class instance
540  // of the right variable definition corresponding to a given numeric
541  // variable use.
542  Context->GlobalNumericVariableTable[DefName] = *DefinedNumericVariable;
543  } else {
544  VariableDefs[DefName] = CurParen;
545  // Mark the string variable as defined to detect collisions between
546  // string and numeric variables in parseNumericVariableUse() and
547  // DefineCmdlineVariables() when the latter is created later than the
548  // former. We cannot reuse GlobalVariableTable for this by populating
549  // it with an empty string since we would then lose the ability to
550  // detect the use of an undefined variable in match().
551  Context->DefinedVariableTable[DefName] = true;
552  }
553  RegExStr += '(';
554  ++CurParen;
555 
556  if (AddRegExToRegEx(MatchRegexp, CurParen, SM))
557  return true;
558 
559  RegExStr += ')';
560  }
561 
562  // Handle fixed string matches.
563  // Find the end, which is the start of the next regex.
564  size_t FixedMatchEnd = PatternStr.find("{{");
565  FixedMatchEnd = std::min(FixedMatchEnd, PatternStr.find("[["));
566  RegExStr += Regex::escape(PatternStr.substr(0, FixedMatchEnd));
567  PatternStr = PatternStr.substr(FixedMatchEnd);
568  }
569 
570  if (MatchFullLinesHere) {
571  if (!Req.NoCanonicalizeWhiteSpace)
572  RegExStr += " *";
573  RegExStr += '$';
574  }
575 
576  return false;
577 }
578 
579 bool FileCheckPattern::AddRegExToRegEx(StringRef RS, unsigned &CurParen, SourceMgr &SM) {
580  Regex R(RS);
581  std::string Error;
582  if (!R.isValid(Error)) {
584  "invalid regex: " + Error);
585  return true;
586  }
587 
588  RegExStr += RS.str();
589  CurParen += R.getNumMatches();
590  return false;
591 }
592 
593 void FileCheckPattern::AddBackrefToRegEx(unsigned BackrefNum) {
594  assert(BackrefNum >= 1 && BackrefNum <= 9 && "Invalid backref number");
595  std::string Backref = std::string("\\") + std::string(1, '0' + BackrefNum);
596  RegExStr += Backref;
597 }
598 
600  const SourceMgr &SM) const {
601  // If this is the EOF pattern, match it immediately.
602  if (CheckTy == Check::CheckEOF) {
603  MatchLen = 0;
604  return Buffer.size();
605  }
606 
607  // If this is a fixed string pattern, just match it now.
608  if (!FixedStr.empty()) {
609  MatchLen = FixedStr.size();
610  size_t Pos = Buffer.find(FixedStr);
611  if (Pos == StringRef::npos)
612  return make_error<FileCheckNotFoundError>();
613  return Pos;
614  }
615 
616  // Regex match.
617 
618  // If there are substitutions, we need to create a temporary string with the
619  // actual value.
620  StringRef RegExToMatch = RegExStr;
621  std::string TmpStr;
622  if (!Substitutions.empty()) {
623  TmpStr = RegExStr;
624  if (LineNumber)
625  Context->LineVariable->setValue(*LineNumber);
626 
627  size_t InsertOffset = 0;
628  // Substitute all string variables and expressions whose values are only
629  // now known. Use of string variables defined on the same line are handled
630  // by back-references.
631  for (const auto &Substitution : Substitutions) {
632  // Substitute and check for failure (e.g. use of undefined variable).
633  Expected<std::string> Value = Substitution->getResult();
634  if (!Value) {
635  Context->LineVariable->clearValue();
636  return Value.takeError();
637  }
638 
639  // Plop it into the regex at the adjusted offset.
640  TmpStr.insert(TmpStr.begin() + Substitution->getIndex() + InsertOffset,
641  Value->begin(), Value->end());
642  InsertOffset += Value->size();
643  }
644 
645  // Match the newly constructed regex.
646  RegExToMatch = TmpStr;
647  Context->LineVariable->clearValue();
648  }
649 
650  SmallVector<StringRef, 4> MatchInfo;
651  if (!Regex(RegExToMatch, Regex::Newline).match(Buffer, &MatchInfo))
652  return make_error<FileCheckNotFoundError>();
653 
654  // Successful regex match.
655  assert(!MatchInfo.empty() && "Didn't get any match");
656  StringRef FullMatch = MatchInfo[0];
657 
658  // If this defines any string variables, remember their values.
659  for (const auto &VariableDef : VariableDefs) {
660  assert(VariableDef.second < MatchInfo.size() && "Internal paren error");
661  Context->GlobalVariableTable[VariableDef.first] =
662  MatchInfo[VariableDef.second];
663  }
664 
665  // If this defines any numeric variables, remember their values.
666  for (const auto &NumericVariableDef : NumericVariableDefs) {
667  const FileCheckNumericVariableMatch &NumericVariableMatch =
668  NumericVariableDef.getValue();
669  unsigned CaptureParenGroup = NumericVariableMatch.CaptureParenGroup;
670  assert(CaptureParenGroup < MatchInfo.size() && "Internal paren error");
671  FileCheckNumericVariable *DefinedNumericVariable =
672  NumericVariableMatch.DefinedNumericVariable;
673 
674  StringRef MatchedValue = MatchInfo[CaptureParenGroup];
675  uint64_t Val;
676  if (MatchedValue.getAsInteger(10, Val))
677  return FileCheckErrorDiagnostic::get(SM, MatchedValue,
678  "Unable to represent numeric value");
679  DefinedNumericVariable->setValue(Val);
680  }
681 
682  // Like CHECK-NEXT, CHECK-EMPTY's match range is considered to start after
683  // the required preceding newline, which is consumed by the pattern in the
684  // case of CHECK-EMPTY but not CHECK-NEXT.
685  size_t MatchStartSkip = CheckTy == Check::CheckEmpty;
686  MatchLen = FullMatch.size() - MatchStartSkip;
687  return FullMatch.data() - Buffer.data() + MatchStartSkip;
688 }
689 
690 unsigned FileCheckPattern::computeMatchDistance(StringRef Buffer) const {
691  // Just compute the number of matching characters. For regular expressions, we
692  // just compare against the regex itself and hope for the best.
693  //
694  // FIXME: One easy improvement here is have the regex lib generate a single
695  // example regular expression which matches, and use that as the example
696  // string.
697  StringRef ExampleString(FixedStr);
698  if (ExampleString.empty())
699  ExampleString = RegExStr;
700 
701  // Only compare up to the first line in the buffer, or the string size.
702  StringRef BufferPrefix = Buffer.substr(0, ExampleString.size());
703  BufferPrefix = BufferPrefix.split('\n').first;
704  return BufferPrefix.edit_distance(ExampleString);
705 }
706 
708  SMRange MatchRange) const {
709  // Print what we know about substitutions.
710  if (!Substitutions.empty()) {
711  for (const auto &Substitution : Substitutions) {
712  SmallString<256> Msg;
713  raw_svector_ostream OS(Msg);
714  Expected<std::string> MatchedValue = Substitution->getResult();
715 
716  // Substitution failed or is not known at match time, print the undefined
717  // variables it uses.
718  if (!MatchedValue) {
719  bool UndefSeen = false;
720  handleAllErrors(MatchedValue.takeError(),
721  [](const FileCheckNotFoundError &E) {},
722  // Handled in PrintNoMatch().
723  [](const FileCheckErrorDiagnostic &E) {},
724  [&](const FileCheckUndefVarError &E) {
725  if (!UndefSeen) {
726  OS << "uses undefined variable(s):";
727  UndefSeen = true;
728  }
729  OS << " ";
730  E.log(OS);
731  });
732  } else {
733  // Substitution succeeded. Print substituted value.
734  OS << "with \"";
735  OS.write_escaped(Substitution->getFromString()) << "\" equal to \"";
736  OS.write_escaped(*MatchedValue) << "\"";
737  }
738 
739  if (MatchRange.isValid())
740  SM.PrintMessage(MatchRange.Start, SourceMgr::DK_Note, OS.str(),
741  {MatchRange});
742  else
744  SourceMgr::DK_Note, OS.str());
745  }
746  }
747 }
748 
750  const SourceMgr &SM, SMLoc Loc,
751  Check::FileCheckType CheckTy,
752  StringRef Buffer, size_t Pos, size_t Len,
753  std::vector<FileCheckDiag> *Diags,
754  bool AdjustPrevDiag = false) {
755  SMLoc Start = SMLoc::getFromPointer(Buffer.data() + Pos);
756  SMLoc End = SMLoc::getFromPointer(Buffer.data() + Pos + Len);
757  SMRange Range(Start, End);
758  if (Diags) {
759  if (AdjustPrevDiag)
760  Diags->rbegin()->MatchTy = MatchTy;
761  else
762  Diags->emplace_back(SM, CheckTy, Loc, MatchTy, Range);
763  }
764  return Range;
765 }
766 
768  const SourceMgr &SM, StringRef Buffer,
769  std::vector<FileCheckDiag> *Diags) const {
770  // Attempt to find the closest/best fuzzy match. Usually an error happens
771  // because some string in the output didn't exactly match. In these cases, we
772  // would like to show the user a best guess at what "should have" matched, to
773  // save them having to actually check the input manually.
774  size_t NumLinesForward = 0;
775  size_t Best = StringRef::npos;
776  double BestQuality = 0;
777 
778  // Use an arbitrary 4k limit on how far we will search.
779  for (size_t i = 0, e = std::min(size_t(4096), Buffer.size()); i != e; ++i) {
780  if (Buffer[i] == '\n')
781  ++NumLinesForward;
782 
783  // Patterns have leading whitespace stripped, so skip whitespace when
784  // looking for something which looks like a pattern.
785  if (Buffer[i] == ' ' || Buffer[i] == '\t')
786  continue;
787 
788  // Compute the "quality" of this match as an arbitrary combination of the
789  // match distance and the number of lines skipped to get to this match.
790  unsigned Distance = computeMatchDistance(Buffer.substr(i));
791  double Quality = Distance + (NumLinesForward / 100.);
792 
793  if (Quality < BestQuality || Best == StringRef::npos) {
794  Best = i;
795  BestQuality = Quality;
796  }
797  }
798 
799  // Print the "possible intended match here" line if we found something
800  // reasonable and not equal to what we showed in the "scanning from here"
801  // line.
802  if (Best && Best != StringRef::npos && BestQuality < 50) {
803  SMRange MatchRange =
805  getCheckTy(), Buffer, Best, 0, Diags);
806  SM.PrintMessage(MatchRange.Start, SourceMgr::DK_Note,
807  "possible intended match here");
808 
809  // FIXME: If we wanted to be really friendly we would show why the match
810  // failed, as it can be hard to spot simple one character differences.
811  }
812 }
813 
816  auto VarIter = GlobalVariableTable.find(VarName);
817  if (VarIter == GlobalVariableTable.end())
818  return make_error<FileCheckUndefVarError>(VarName);
819 
820  return VarIter->second;
821 }
822 
823 template <class... Types>
825 FileCheckPatternContext::makeNumericVariable(Types... args) {
826  NumericVariables.push_back(
827  llvm::make_unique<FileCheckNumericVariable>(args...));
828  return NumericVariables.back().get();
829 }
830 
832 FileCheckPatternContext::makeStringSubstitution(StringRef VarName,
833  size_t InsertIdx) {
834  Substitutions.push_back(
835  llvm::make_unique<FileCheckStringSubstitution>(this, VarName, InsertIdx));
836  return Substitutions.back().get();
837 }
838 
839 FileCheckSubstitution *FileCheckPatternContext::makeNumericSubstitution(
840  StringRef ExpressionStr,
841  std::unique_ptr<FileCheckExpressionAST> ExpressionAST, size_t InsertIdx) {
842  Substitutions.push_back(llvm::make_unique<FileCheckNumericSubstitution>(
843  this, ExpressionStr, std::move(ExpressionAST), InsertIdx));
844  return Substitutions.back().get();
845 }
846 
847 size_t FileCheckPattern::FindRegexVarEnd(StringRef Str, SourceMgr &SM) {
848  // Offset keeps track of the current offset within the input Str
849  size_t Offset = 0;
850  // [...] Nesting depth
851  size_t BracketDepth = 0;
852 
853  while (!Str.empty()) {
854  if (Str.startswith("]]") && BracketDepth == 0)
855  return Offset;
856  if (Str[0] == '\\') {
857  // Backslash escapes the next char within regexes, so skip them both.
858  Str = Str.substr(2);
859  Offset += 2;
860  } else {
861  switch (Str[0]) {
862  default:
863  break;
864  case '[':
865  BracketDepth++;
866  break;
867  case ']':
868  if (BracketDepth == 0) {
871  "missing closing \"]\" for regex variable");
872  exit(1);
873  }
874  BracketDepth--;
875  break;
876  }
877  Str = Str.substr(1);
878  Offset++;
879  }
880  }
881 
882  return StringRef::npos;
883 }
884 
886  SmallVectorImpl<char> &OutputBuffer) {
887  OutputBuffer.reserve(MB.getBufferSize());
888 
889  for (const char *Ptr = MB.getBufferStart(), *End = MB.getBufferEnd();
890  Ptr != End; ++Ptr) {
891  // Eliminate trailing dosish \r.
892  if (Ptr <= End - 2 && Ptr[0] == '\r' && Ptr[1] == '\n') {
893  continue;
894  }
895 
896  // If current char is not a horizontal whitespace or if horizontal
897  // whitespace canonicalization is disabled, dump it to output as is.
898  if (Req.NoCanonicalizeWhiteSpace || (*Ptr != ' ' && *Ptr != '\t')) {
899  OutputBuffer.push_back(*Ptr);
900  continue;
901  }
902 
903  // Otherwise, add one space and advance over neighboring space.
904  OutputBuffer.push_back(' ');
905  while (Ptr + 1 != End && (Ptr[1] == ' ' || Ptr[1] == '\t'))
906  ++Ptr;
907  }
908 
909  // Add a null byte and then return all but that byte.
910  OutputBuffer.push_back('\0');
911  return StringRef(OutputBuffer.data(), OutputBuffer.size() - 1);
912 }
913 
915  const Check::FileCheckType &CheckTy,
916  SMLoc CheckLoc, MatchType MatchTy,
917  SMRange InputRange)
918  : CheckTy(CheckTy), MatchTy(MatchTy) {
919  auto Start = SM.getLineAndColumn(InputRange.Start);
920  auto End = SM.getLineAndColumn(InputRange.End);
921  InputStartLine = Start.first;
922  InputStartCol = Start.second;
923  InputEndLine = End.first;
924  InputEndCol = End.second;
925  Start = SM.getLineAndColumn(CheckLoc);
926  CheckLine = Start.first;
927  CheckCol = Start.second;
928 }
929 
930 static bool IsPartOfWord(char c) {
931  return (isalnum(c) || c == '-' || c == '_');
932 }
933 
935  assert(Count > 0 && "zero and negative counts are not supported");
936  assert((C == 1 || Kind == CheckPlain) &&
937  "count supported only for plain CHECK directives");
938  Count = C;
939  return *this;
940 }
941 
943  switch (Kind) {
944  case Check::CheckNone:
945  return "invalid";
946  case Check::CheckPlain:
947  if (Count > 1)
948  return Prefix.str() + "-COUNT";
949  return Prefix;
950  case Check::CheckNext:
951  return Prefix.str() + "-NEXT";
952  case Check::CheckSame:
953  return Prefix.str() + "-SAME";
954  case Check::CheckNot:
955  return Prefix.str() + "-NOT";
956  case Check::CheckDAG:
957  return Prefix.str() + "-DAG";
958  case Check::CheckLabel:
959  return Prefix.str() + "-LABEL";
960  case Check::CheckEmpty:
961  return Prefix.str() + "-EMPTY";
962  case Check::CheckEOF:
963  return "implicit EOF";
964  case Check::CheckBadNot:
965  return "bad NOT";
967  return "bad COUNT";
968  }
969  llvm_unreachable("unknown FileCheckType");
970 }
971 
972 static std::pair<Check::FileCheckType, StringRef>
974  if (Buffer.size() <= Prefix.size())
975  return {Check::CheckNone, StringRef()};
976 
977  char NextChar = Buffer[Prefix.size()];
978 
979  StringRef Rest = Buffer.drop_front(Prefix.size() + 1);
980  // Verify that the : is present after the prefix.
981  if (NextChar == ':')
982  return {Check::CheckPlain, Rest};
983 
984  if (NextChar != '-')
985  return {Check::CheckNone, StringRef()};
986 
987  if (Rest.consume_front("COUNT-")) {
988  int64_t Count;
989  if (Rest.consumeInteger(10, Count))
990  // Error happened in parsing integer.
991  return {Check::CheckBadCount, Rest};
992  if (Count <= 0 || Count > INT32_MAX)
993  return {Check::CheckBadCount, Rest};
994  if (!Rest.consume_front(":"))
995  return {Check::CheckBadCount, Rest};
996  return {Check::FileCheckType(Check::CheckPlain).setCount(Count), Rest};
997  }
998 
999  if (Rest.consume_front("NEXT:"))
1000  return {Check::CheckNext, Rest};
1001 
1002  if (Rest.consume_front("SAME:"))
1003  return {Check::CheckSame, Rest};
1004 
1005  if (Rest.consume_front("NOT:"))
1006  return {Check::CheckNot, Rest};
1007 
1008  if (Rest.consume_front("DAG:"))
1009  return {Check::CheckDAG, Rest};
1010 
1011  if (Rest.consume_front("LABEL:"))
1012  return {Check::CheckLabel, Rest};
1013 
1014  if (Rest.consume_front("EMPTY:"))
1015  return {Check::CheckEmpty, Rest};
1016 
1017  // You can't combine -NOT with another suffix.
1018  if (Rest.startswith("DAG-NOT:") || Rest.startswith("NOT-DAG:") ||
1019  Rest.startswith("NEXT-NOT:") || Rest.startswith("NOT-NEXT:") ||
1020  Rest.startswith("SAME-NOT:") || Rest.startswith("NOT-SAME:") ||
1021  Rest.startswith("EMPTY-NOT:") || Rest.startswith("NOT-EMPTY:"))
1022  return {Check::CheckBadNot, Rest};
1023 
1024  return {Check::CheckNone, Rest};
1025 }
1026 
1027 // From the given position, find the next character after the word.
1028 static size_t SkipWord(StringRef Str, size_t Loc) {
1029  while (Loc < Str.size() && IsPartOfWord(Str[Loc]))
1030  ++Loc;
1031  return Loc;
1032 }
1033 
1034 /// Searches the buffer for the first prefix in the prefix regular expression.
1035 ///
1036 /// This searches the buffer using the provided regular expression, however it
1037 /// enforces constraints beyond that:
1038 /// 1) The found prefix must not be a suffix of something that looks like
1039 /// a valid prefix.
1040 /// 2) The found prefix must be followed by a valid check type suffix using \c
1041 /// FindCheckType above.
1042 ///
1043 /// \returns a pair of StringRefs into the Buffer, which combines:
1044 /// - the first match of the regular expression to satisfy these two is
1045 /// returned,
1046 /// otherwise an empty StringRef is returned to indicate failure.
1047 /// - buffer rewound to the location right after parsed suffix, for parsing
1048 /// to continue from
1049 ///
1050 /// If this routine returns a valid prefix, it will also shrink \p Buffer to
1051 /// start at the beginning of the returned prefix, increment \p LineNumber for
1052 /// each new line consumed from \p Buffer, and set \p CheckTy to the type of
1053 /// check found by examining the suffix.
1054 ///
1055 /// If no valid prefix is found, the state of Buffer, LineNumber, and CheckTy
1056 /// is unspecified.
1057 static std::pair<StringRef, StringRef>
1059  unsigned &LineNumber, Check::FileCheckType &CheckTy) {
1060  SmallVector<StringRef, 2> Matches;
1061 
1062  while (!Buffer.empty()) {
1063  // Find the first (longest) match using the RE.
1064  if (!PrefixRE.match(Buffer, &Matches))
1065  // No match at all, bail.
1066  return {StringRef(), StringRef()};
1067 
1068  StringRef Prefix = Matches[0];
1069  Matches.clear();
1070 
1071  assert(Prefix.data() >= Buffer.data() &&
1072  Prefix.data() < Buffer.data() + Buffer.size() &&
1073  "Prefix doesn't start inside of buffer!");
1074  size_t Loc = Prefix.data() - Buffer.data();
1075  StringRef Skipped = Buffer.substr(0, Loc);
1076  Buffer = Buffer.drop_front(Loc);
1077  LineNumber += Skipped.count('\n');
1078 
1079  // Check that the matched prefix isn't a suffix of some other check-like
1080  // word.
1081  // FIXME: This is a very ad-hoc check. it would be better handled in some
1082  // other way. Among other things it seems hard to distinguish between
1083  // intentional and unintentional uses of this feature.
1084  if (Skipped.empty() || !IsPartOfWord(Skipped.back())) {
1085  // Now extract the type.
1086  StringRef AfterSuffix;
1087  std::tie(CheckTy, AfterSuffix) = FindCheckType(Buffer, Prefix);
1088 
1089  // If we've found a valid check type for this prefix, we're done.
1090  if (CheckTy != Check::CheckNone)
1091  return {Prefix, AfterSuffix};
1092  }
1093 
1094  // If we didn't successfully find a prefix, we need to skip this invalid
1095  // prefix and continue scanning. We directly skip the prefix that was
1096  // matched and any additional parts of that check-like word.
1097  Buffer = Buffer.drop_front(SkipWord(Buffer, Prefix.size()));
1098  }
1099 
1100  // We ran out of buffer while skipping partial matches so give up.
1101  return {StringRef(), StringRef()};
1102 }
1103 
1105  assert(!LineVariable && "@LINE pseudo numeric variable already created");
1106  StringRef LineName = "@LINE";
1107  LineVariable = makeNumericVariable(LineName);
1108  GlobalNumericVariableTable[LineName] = LineVariable;
1109 }
1110 
1112  std::vector<FileCheckString> &CheckStrings) {
1113  Error DefineError =
1114  PatternContext.defineCmdlineVariables(Req.GlobalDefines, SM);
1115  if (DefineError) {
1116  logAllUnhandledErrors(std::move(DefineError), errs());
1117  return true;
1118  }
1119 
1120  PatternContext.createLineVariable();
1121 
1122  std::vector<FileCheckPattern> ImplicitNegativeChecks;
1123  for (const auto &PatternString : Req.ImplicitCheckNot) {
1124  // Create a buffer with fake command line content in order to display the
1125  // command line option responsible for the specific implicit CHECK-NOT.
1126  std::string Prefix = "-implicit-check-not='";
1127  std::string Suffix = "'";
1128  std::unique_ptr<MemoryBuffer> CmdLine = MemoryBuffer::getMemBufferCopy(
1129  Prefix + PatternString + Suffix, "command line");
1130 
1131  StringRef PatternInBuffer =
1132  CmdLine->getBuffer().substr(Prefix.size(), PatternString.size());
1133  SM.AddNewSourceBuffer(std::move(CmdLine), SMLoc());
1134 
1135  ImplicitNegativeChecks.push_back(
1136  FileCheckPattern(Check::CheckNot, &PatternContext));
1137  ImplicitNegativeChecks.back().parsePattern(PatternInBuffer,
1138  "IMPLICIT-CHECK", SM, Req);
1139  }
1140 
1141  std::vector<FileCheckPattern> DagNotMatches = ImplicitNegativeChecks;
1142 
1143  // LineNumber keeps track of the line on which CheckPrefix instances are
1144  // found.
1145  unsigned LineNumber = 1;
1146 
1147  while (1) {
1149 
1150  // See if a prefix occurs in the memory buffer.
1151  StringRef UsedPrefix;
1152  StringRef AfterSuffix;
1153  std::tie(UsedPrefix, AfterSuffix) =
1154  FindFirstMatchingPrefix(PrefixRE, Buffer, LineNumber, CheckTy);
1155  if (UsedPrefix.empty())
1156  break;
1157  assert(UsedPrefix.data() == Buffer.data() &&
1158  "Failed to move Buffer's start forward, or pointed prefix outside "
1159  "of the buffer!");
1160  assert(AfterSuffix.data() >= Buffer.data() &&
1161  AfterSuffix.data() < Buffer.data() + Buffer.size() &&
1162  "Parsing after suffix doesn't start inside of buffer!");
1163 
1164  // Location to use for error messages.
1165  const char *UsedPrefixStart = UsedPrefix.data();
1166 
1167  // Skip the buffer to the end of parsed suffix (or just prefix, if no good
1168  // suffix was processed).
1169  Buffer = AfterSuffix.empty() ? Buffer.drop_front(UsedPrefix.size())
1170  : AfterSuffix;
1171 
1172  // Complain about useful-looking but unsupported suffixes.
1173  if (CheckTy == Check::CheckBadNot) {
1175  "unsupported -NOT combo on prefix '" + UsedPrefix + "'");
1176  return true;
1177  }
1178 
1179  // Complain about invalid count specification.
1180  if (CheckTy == Check::CheckBadCount) {
1182  "invalid count in -COUNT specification on prefix '" +
1183  UsedPrefix + "'");
1184  return true;
1185  }
1186 
1187  // Okay, we found the prefix, yay. Remember the rest of the line, but ignore
1188  // leading whitespace.
1189  if (!(Req.NoCanonicalizeWhiteSpace && Req.MatchFullLines))
1190  Buffer = Buffer.substr(Buffer.find_first_not_of(" \t"));
1191 
1192  // Scan ahead to the end of line.
1193  size_t EOL = Buffer.find_first_of("\n\r");
1194 
1195  // Remember the location of the start of the pattern, for diagnostics.
1196  SMLoc PatternLoc = SMLoc::getFromPointer(Buffer.data());
1197 
1198  // Parse the pattern.
1199  FileCheckPattern P(CheckTy, &PatternContext, LineNumber);
1200  if (P.parsePattern(Buffer.substr(0, EOL), UsedPrefix, SM, Req))
1201  return true;
1202 
1203  // Verify that CHECK-LABEL lines do not define or use variables
1204  if ((CheckTy == Check::CheckLabel) && P.hasVariable()) {
1205  SM.PrintMessage(
1206  SMLoc::getFromPointer(UsedPrefixStart), SourceMgr::DK_Error,
1207  "found '" + UsedPrefix + "-LABEL:'"
1208  " with variable definition or use");
1209  return true;
1210  }
1211 
1212  Buffer = Buffer.substr(EOL);
1213 
1214  // Verify that CHECK-NEXT/SAME/EMPTY lines have at least one CHECK line before them.
1215  if ((CheckTy == Check::CheckNext || CheckTy == Check::CheckSame ||
1216  CheckTy == Check::CheckEmpty) &&
1217  CheckStrings.empty()) {
1218  StringRef Type = CheckTy == Check::CheckNext
1219  ? "NEXT"
1220  : CheckTy == Check::CheckEmpty ? "EMPTY" : "SAME";
1221  SM.PrintMessage(SMLoc::getFromPointer(UsedPrefixStart),
1223  "found '" + UsedPrefix + "-" + Type +
1224  "' without previous '" + UsedPrefix + ": line");
1225  return true;
1226  }
1227 
1228  // Handle CHECK-DAG/-NOT.
1229  if (CheckTy == Check::CheckDAG || CheckTy == Check::CheckNot) {
1230  DagNotMatches.push_back(P);
1231  continue;
1232  }
1233 
1234  // Okay, add the string we captured to the output vector and move on.
1235  CheckStrings.emplace_back(P, UsedPrefix, PatternLoc);
1236  std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
1237  DagNotMatches = ImplicitNegativeChecks;
1238  }
1239 
1240  // Add an EOF pattern for any trailing CHECK-DAG/-NOTs, and use the first
1241  // prefix as a filler for the error message.
1242  if (!DagNotMatches.empty()) {
1243  CheckStrings.emplace_back(
1244  FileCheckPattern(Check::CheckEOF, &PatternContext, LineNumber + 1),
1245  *Req.CheckPrefixes.begin(), SMLoc::getFromPointer(Buffer.data()));
1246  std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
1247  }
1248 
1249  if (CheckStrings.empty()) {
1250  errs() << "error: no check strings found with prefix"
1251  << (Req.CheckPrefixes.size() > 1 ? "es " : " ");
1252  auto I = Req.CheckPrefixes.begin();
1253  auto E = Req.CheckPrefixes.end();
1254  if (I != E) {
1255  errs() << "\'" << *I << ":'";
1256  ++I;
1257  }
1258  for (; I != E; ++I)
1259  errs() << ", \'" << *I << ":'";
1260 
1261  errs() << '\n';
1262  return true;
1263  }
1264 
1265  return false;
1266 }
1267 
1268 static void PrintMatch(bool ExpectedMatch, const SourceMgr &SM,
1269  StringRef Prefix, SMLoc Loc, const FileCheckPattern &Pat,
1270  int MatchedCount, StringRef Buffer, size_t MatchPos,
1271  size_t MatchLen, const FileCheckRequest &Req,
1272  std::vector<FileCheckDiag> *Diags) {
1273  bool PrintDiag = true;
1274  if (ExpectedMatch) {
1275  if (!Req.Verbose)
1276  return;
1277  if (!Req.VerboseVerbose && Pat.getCheckTy() == Check::CheckEOF)
1278  return;
1279  // Due to their verbosity, we don't print verbose diagnostics here if we're
1280  // gathering them for a different rendering, but we always print other
1281  // diagnostics.
1282  PrintDiag = !Diags;
1283  }
1284  SMRange MatchRange = ProcessMatchResult(
1285  ExpectedMatch ? FileCheckDiag::MatchFoundAndExpected
1287  SM, Loc, Pat.getCheckTy(), Buffer, MatchPos, MatchLen, Diags);
1288  if (!PrintDiag)
1289  return;
1290 
1291  std::string Message = formatv("{0}: {1} string found in input",
1292  Pat.getCheckTy().getDescription(Prefix),
1293  (ExpectedMatch ? "expected" : "excluded"))
1294  .str();
1295  if (Pat.getCount() > 1)
1296  Message += formatv(" ({0} out of {1})", MatchedCount, Pat.getCount()).str();
1297 
1298  SM.PrintMessage(
1299  Loc, ExpectedMatch ? SourceMgr::DK_Remark : SourceMgr::DK_Error, Message);
1300  SM.PrintMessage(MatchRange.Start, SourceMgr::DK_Note, "found here",
1301  {MatchRange});
1302  Pat.printSubstitutions(SM, Buffer, MatchRange);
1303 }
1304 
1305 static void PrintMatch(bool ExpectedMatch, const SourceMgr &SM,
1306  const FileCheckString &CheckStr, int MatchedCount,
1307  StringRef Buffer, size_t MatchPos, size_t MatchLen,
1308  FileCheckRequest &Req,
1309  std::vector<FileCheckDiag> *Diags) {
1310  PrintMatch(ExpectedMatch, SM, CheckStr.Prefix, CheckStr.Loc, CheckStr.Pat,
1311  MatchedCount, Buffer, MatchPos, MatchLen, Req, Diags);
1312 }
1313 
1314 static void PrintNoMatch(bool ExpectedMatch, const SourceMgr &SM,
1315  StringRef Prefix, SMLoc Loc,
1316  const FileCheckPattern &Pat, int MatchedCount,
1317  StringRef Buffer, bool VerboseVerbose,
1318  std::vector<FileCheckDiag> *Diags, Error MatchErrors) {
1319  assert(MatchErrors && "Called on successful match");
1320  bool PrintDiag = true;
1321  if (!ExpectedMatch) {
1322  if (!VerboseVerbose) {
1323  consumeError(std::move(MatchErrors));
1324  return;
1325  }
1326  // Due to their verbosity, we don't print verbose diagnostics here if we're
1327  // gathering them for a different rendering, but we always print other
1328  // diagnostics.
1329  PrintDiag = !Diags;
1330  }
1331 
1332  // If the current position is at the end of a line, advance to the start of
1333  // the next line.
1334  Buffer = Buffer.substr(Buffer.find_first_not_of(" \t\n\r"));
1335  SMRange SearchRange = ProcessMatchResult(
1336  ExpectedMatch ? FileCheckDiag::MatchNoneButExpected
1338  SM, Loc, Pat.getCheckTy(), Buffer, 0, Buffer.size(), Diags);
1339  if (!PrintDiag) {
1340  consumeError(std::move(MatchErrors));
1341  return;
1342  }
1343 
1344  MatchErrors =
1345  handleErrors(std::move(MatchErrors),
1346  [](const FileCheckErrorDiagnostic &E) { E.log(errs()); });
1347 
1348  // No problem matching the string per se.
1349  if (!MatchErrors)
1350  return;
1351  consumeError(std::move(MatchErrors));
1352 
1353  // Print "not found" diagnostic.
1354  std::string Message = formatv("{0}: {1} string not found in input",
1355  Pat.getCheckTy().getDescription(Prefix),
1356  (ExpectedMatch ? "expected" : "excluded"))
1357  .str();
1358  if (Pat.getCount() > 1)
1359  Message += formatv(" ({0} out of {1})", MatchedCount, Pat.getCount()).str();
1360  SM.PrintMessage(
1361  Loc, ExpectedMatch ? SourceMgr::DK_Error : SourceMgr::DK_Remark, Message);
1362 
1363  // Print the "scanning from here" line.
1364  SM.PrintMessage(SearchRange.Start, SourceMgr::DK_Note, "scanning from here");
1365 
1366  // Allow the pattern to print additional information if desired.
1367  Pat.printSubstitutions(SM, Buffer);
1368 
1369  if (ExpectedMatch)
1370  Pat.printFuzzyMatch(SM, Buffer, Diags);
1371 }
1372 
1373 static void PrintNoMatch(bool ExpectedMatch, const SourceMgr &SM,
1374  const FileCheckString &CheckStr, int MatchedCount,
1375  StringRef Buffer, bool VerboseVerbose,
1376  std::vector<FileCheckDiag> *Diags, Error MatchErrors) {
1377  PrintNoMatch(ExpectedMatch, SM, CheckStr.Prefix, CheckStr.Loc, CheckStr.Pat,
1378  MatchedCount, Buffer, VerboseVerbose, Diags,
1379  std::move(MatchErrors));
1380 }
1381 
1382 /// Counts the number of newlines in the specified range.
1383 static unsigned CountNumNewlinesBetween(StringRef Range,
1384  const char *&FirstNewLine) {
1385  unsigned NumNewLines = 0;
1386  while (1) {
1387  // Scan for newline.
1388  Range = Range.substr(Range.find_first_of("\n\r"));
1389  if (Range.empty())
1390  return NumNewLines;
1391 
1392  ++NumNewLines;
1393 
1394  // Handle \n\r and \r\n as a single newline.
1395  if (Range.size() > 1 && (Range[1] == '\n' || Range[1] == '\r') &&
1396  (Range[0] != Range[1]))
1397  Range = Range.substr(1);
1398  Range = Range.substr(1);
1399 
1400  if (NumNewLines == 1)
1401  FirstNewLine = Range.begin();
1402  }
1403 }
1404 
1405 size_t FileCheckString::Check(const SourceMgr &SM, StringRef Buffer,
1406  bool IsLabelScanMode, size_t &MatchLen,
1407  FileCheckRequest &Req,
1408  std::vector<FileCheckDiag> *Diags) const {
1409  size_t LastPos = 0;
1410  std::vector<const FileCheckPattern *> NotStrings;
1411 
1412  // IsLabelScanMode is true when we are scanning forward to find CHECK-LABEL
1413  // bounds; we have not processed variable definitions within the bounded block
1414  // yet so cannot handle any final CHECK-DAG yet; this is handled when going
1415  // over the block again (including the last CHECK-LABEL) in normal mode.
1416  if (!IsLabelScanMode) {
1417  // Match "dag strings" (with mixed "not strings" if any).
1418  LastPos = CheckDag(SM, Buffer, NotStrings, Req, Diags);
1419  if (LastPos == StringRef::npos)
1420  return StringRef::npos;
1421  }
1422 
1423  // Match itself from the last position after matching CHECK-DAG.
1424  size_t LastMatchEnd = LastPos;
1425  size_t FirstMatchPos = 0;
1426  // Go match the pattern Count times. Majority of patterns only match with
1427  // count 1 though.
1428  assert(Pat.getCount() != 0 && "pattern count can not be zero");
1429  for (int i = 1; i <= Pat.getCount(); i++) {
1430  StringRef MatchBuffer = Buffer.substr(LastMatchEnd);
1431  size_t CurrentMatchLen;
1432  // get a match at current start point
1433  Expected<size_t> MatchResult = Pat.match(MatchBuffer, CurrentMatchLen, SM);
1434 
1435  // report
1436  if (!MatchResult) {
1437  PrintNoMatch(true, SM, *this, i, MatchBuffer, Req.VerboseVerbose, Diags,
1438  MatchResult.takeError());
1439  return StringRef::npos;
1440  }
1441  size_t MatchPos = *MatchResult;
1442  PrintMatch(true, SM, *this, i, MatchBuffer, MatchPos, CurrentMatchLen, Req,
1443  Diags);
1444  if (i == 1)
1445  FirstMatchPos = LastPos + MatchPos;
1446 
1447  // move start point after the match
1448  LastMatchEnd += MatchPos + CurrentMatchLen;
1449  }
1450  // Full match len counts from first match pos.
1451  MatchLen = LastMatchEnd - FirstMatchPos;
1452 
1453  // Similar to the above, in "label-scan mode" we can't yet handle CHECK-NEXT
1454  // or CHECK-NOT
1455  if (!IsLabelScanMode) {
1456  size_t MatchPos = FirstMatchPos - LastPos;
1457  StringRef MatchBuffer = Buffer.substr(LastPos);
1458  StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
1459 
1460  // If this check is a "CHECK-NEXT", verify that the previous match was on
1461  // the previous line (i.e. that there is one newline between them).
1462  if (CheckNext(SM, SkippedRegion)) {
1464  Pat.getCheckTy(), MatchBuffer, MatchPos, MatchLen,
1465  Diags, Req.Verbose);
1466  return StringRef::npos;
1467  }
1468 
1469  // If this check is a "CHECK-SAME", verify that the previous match was on
1470  // the same line (i.e. that there is no newline between them).
1471  if (CheckSame(SM, SkippedRegion)) {
1473  Pat.getCheckTy(), MatchBuffer, MatchPos, MatchLen,
1474  Diags, Req.Verbose);
1475  return StringRef::npos;
1476  }
1477 
1478  // If this match had "not strings", verify that they don't exist in the
1479  // skipped region.
1480  if (CheckNot(SM, SkippedRegion, NotStrings, Req, Diags))
1481  return StringRef::npos;
1482  }
1483 
1484  return FirstMatchPos;
1485 }
1486 
1487 bool FileCheckString::CheckNext(const SourceMgr &SM, StringRef Buffer) const {
1488  if (Pat.getCheckTy() != Check::CheckNext &&
1489  Pat.getCheckTy() != Check::CheckEmpty)
1490  return false;
1491 
1492  Twine CheckName =
1493  Prefix +
1494  Twine(Pat.getCheckTy() == Check::CheckEmpty ? "-EMPTY" : "-NEXT");
1495 
1496  // Count the number of newlines between the previous match and this one.
1497  const char *FirstNewLine = nullptr;
1498  unsigned NumNewLines = CountNumNewlinesBetween(Buffer, FirstNewLine);
1499 
1500  if (NumNewLines == 0) {
1502  CheckName + ": is on the same line as previous match");
1504  "'next' match was here");
1506  "previous match ended here");
1507  return true;
1508  }
1509 
1510  if (NumNewLines != 1) {
1512  CheckName +
1513  ": is not on the line after the previous match");
1515  "'next' match was here");
1517  "previous match ended here");
1519  "non-matching line after previous match is here");
1520  return true;
1521  }
1522 
1523  return false;
1524 }
1525 
1526 bool FileCheckString::CheckSame(const SourceMgr &SM, StringRef Buffer) const {
1527  if (Pat.getCheckTy() != Check::CheckSame)
1528  return false;
1529 
1530  // Count the number of newlines between the previous match and this one.
1531  const char *FirstNewLine = nullptr;
1532  unsigned NumNewLines = CountNumNewlinesBetween(Buffer, FirstNewLine);
1533 
1534  if (NumNewLines != 0) {
1536  Prefix +
1537  "-SAME: is not on the same line as the previous match");
1539  "'next' match was here");
1541  "previous match ended here");
1542  return true;
1543  }
1544 
1545  return false;
1546 }
1547 
1549  const SourceMgr &SM, StringRef Buffer,
1550  const std::vector<const FileCheckPattern *> &NotStrings,
1551  const FileCheckRequest &Req, std::vector<FileCheckDiag> *Diags) const {
1552  for (const FileCheckPattern *Pat : NotStrings) {
1553  assert((Pat->getCheckTy() == Check::CheckNot) && "Expect CHECK-NOT!");
1554 
1555  size_t MatchLen = 0;
1556  Expected<size_t> MatchResult = Pat->match(Buffer, MatchLen, SM);
1557 
1558  if (!MatchResult) {
1559  PrintNoMatch(false, SM, Prefix, Pat->getLoc(), *Pat, 1, Buffer,
1560  Req.VerboseVerbose, Diags, MatchResult.takeError());
1561  continue;
1562  }
1563  size_t Pos = *MatchResult;
1564 
1565  PrintMatch(false, SM, Prefix, Pat->getLoc(), *Pat, 1, Buffer, Pos, MatchLen,
1566  Req, Diags);
1567 
1568  return true;
1569  }
1570 
1571  return false;
1572 }
1573 
1574 size_t
1576  std::vector<const FileCheckPattern *> &NotStrings,
1577  const FileCheckRequest &Req,
1578  std::vector<FileCheckDiag> *Diags) const {
1579  if (DagNotStrings.empty())
1580  return 0;
1581 
1582  // The start of the search range.
1583  size_t StartPos = 0;
1584 
1585  struct MatchRange {
1586  size_t Pos;
1587  size_t End;
1588  };
1589  // A sorted list of ranges for non-overlapping CHECK-DAG matches. Match
1590  // ranges are erased from this list once they are no longer in the search
1591  // range.
1592  std::list<MatchRange> MatchRanges;
1593 
1594  // We need PatItr and PatEnd later for detecting the end of a CHECK-DAG
1595  // group, so we don't use a range-based for loop here.
1596  for (auto PatItr = DagNotStrings.begin(), PatEnd = DagNotStrings.end();
1597  PatItr != PatEnd; ++PatItr) {
1598  const FileCheckPattern &Pat = *PatItr;
1599  assert((Pat.getCheckTy() == Check::CheckDAG ||
1600  Pat.getCheckTy() == Check::CheckNot) &&
1601  "Invalid CHECK-DAG or CHECK-NOT!");
1602 
1603  if (Pat.getCheckTy() == Check::CheckNot) {
1604  NotStrings.push_back(&Pat);
1605  continue;
1606  }
1607 
1608  assert((Pat.getCheckTy() == Check::CheckDAG) && "Expect CHECK-DAG!");
1609 
1610  // CHECK-DAG always matches from the start.
1611  size_t MatchLen = 0, MatchPos = StartPos;
1612 
1613  // Search for a match that doesn't overlap a previous match in this
1614  // CHECK-DAG group.
1615  for (auto MI = MatchRanges.begin(), ME = MatchRanges.end(); true; ++MI) {
1616  StringRef MatchBuffer = Buffer.substr(MatchPos);
1617  Expected<size_t> MatchResult = Pat.match(MatchBuffer, MatchLen, SM);
1618  // With a group of CHECK-DAGs, a single mismatching means the match on
1619  // that group of CHECK-DAGs fails immediately.
1620  if (!MatchResult) {
1621  PrintNoMatch(true, SM, Prefix, Pat.getLoc(), Pat, 1, MatchBuffer,
1622  Req.VerboseVerbose, Diags, MatchResult.takeError());
1623  return StringRef::npos;
1624  }
1625  size_t MatchPosBuf = *MatchResult;
1626  // Re-calc it as the offset relative to the start of the original string.
1627  MatchPos += MatchPosBuf;
1628  if (Req.VerboseVerbose)
1629  PrintMatch(true, SM, Prefix, Pat.getLoc(), Pat, 1, Buffer, MatchPos,
1630  MatchLen, Req, Diags);
1631  MatchRange M{MatchPos, MatchPos + MatchLen};
1632  if (Req.AllowDeprecatedDagOverlap) {
1633  // We don't need to track all matches in this mode, so we just maintain
1634  // one match range that encompasses the current CHECK-DAG group's
1635  // matches.
1636  if (MatchRanges.empty())
1637  MatchRanges.insert(MatchRanges.end(), M);
1638  else {
1639  auto Block = MatchRanges.begin();
1640  Block->Pos = std::min(Block->Pos, M.Pos);
1641  Block->End = std::max(Block->End, M.End);
1642  }
1643  break;
1644  }
1645  // Iterate previous matches until overlapping match or insertion point.
1646  bool Overlap = false;
1647  for (; MI != ME; ++MI) {
1648  if (M.Pos < MI->End) {
1649  // !Overlap => New match has no overlap and is before this old match.
1650  // Overlap => New match overlaps this old match.
1651  Overlap = MI->Pos < M.End;
1652  break;
1653  }
1654  }
1655  if (!Overlap) {
1656  // Insert non-overlapping match into list.
1657  MatchRanges.insert(MI, M);
1658  break;
1659  }
1660  if (Req.VerboseVerbose) {
1661  // Due to their verbosity, we don't print verbose diagnostics here if
1662  // we're gathering them for a different rendering, but we always print
1663  // other diagnostics.
1664  if (!Diags) {
1665  SMLoc OldStart = SMLoc::getFromPointer(Buffer.data() + MI->Pos);
1666  SMLoc OldEnd = SMLoc::getFromPointer(Buffer.data() + MI->End);
1667  SMRange OldRange(OldStart, OldEnd);
1668  SM.PrintMessage(OldStart, SourceMgr::DK_Note,
1669  "match discarded, overlaps earlier DAG match here",
1670  {OldRange});
1671  } else
1672  Diags->rbegin()->MatchTy = FileCheckDiag::MatchFoundButDiscarded;
1673  }
1674  MatchPos = MI->End;
1675  }
1676  if (!Req.VerboseVerbose)
1677  PrintMatch(true, SM, Prefix, Pat.getLoc(), Pat, 1, Buffer, MatchPos,
1678  MatchLen, Req, Diags);
1679 
1680  // Handle the end of a CHECK-DAG group.
1681  if (std::next(PatItr) == PatEnd ||
1682  std::next(PatItr)->getCheckTy() == Check::CheckNot) {
1683  if (!NotStrings.empty()) {
1684  // If there are CHECK-NOTs between two CHECK-DAGs or from CHECK to
1685  // CHECK-DAG, verify that there are no 'not' strings occurred in that
1686  // region.
1687  StringRef SkippedRegion =
1688  Buffer.slice(StartPos, MatchRanges.begin()->Pos);
1689  if (CheckNot(SM, SkippedRegion, NotStrings, Req, Diags))
1690  return StringRef::npos;
1691  // Clear "not strings".
1692  NotStrings.clear();
1693  }
1694  // All subsequent CHECK-DAGs and CHECK-NOTs should be matched from the
1695  // end of this CHECK-DAG group's match range.
1696  StartPos = MatchRanges.rbegin()->End;
1697  // Don't waste time checking for (impossible) overlaps before that.
1698  MatchRanges.clear();
1699  }
1700  }
1701 
1702  return StartPos;
1703 }
1704 
1705 // A check prefix must contain only alphanumeric, hyphens and underscores.
1706 static bool ValidateCheckPrefix(StringRef CheckPrefix) {
1707  Regex Validator("^[a-zA-Z0-9_-]*$");
1708  return Validator.match(CheckPrefix);
1709 }
1710 
1712  StringSet<> PrefixSet;
1713 
1714  for (StringRef Prefix : Req.CheckPrefixes) {
1715  // Reject empty prefixes.
1716  if (Prefix == "")
1717  return false;
1718 
1719  if (!PrefixSet.insert(Prefix).second)
1720  return false;
1721 
1722  if (!ValidateCheckPrefix(Prefix))
1723  return false;
1724  }
1725 
1726  return true;
1727 }
1728 
1730  // I don't think there's a way to specify an initial value for cl::list,
1731  // so if nothing was specified, add the default
1732  if (Req.CheckPrefixes.empty())
1733  Req.CheckPrefixes.push_back("CHECK");
1734 
1735  // We already validated the contents of CheckPrefixes so just concatenate
1736  // them as alternatives.
1737  SmallString<32> PrefixRegexStr;
1738  for (StringRef Prefix : Req.CheckPrefixes) {
1739  if (Prefix != Req.CheckPrefixes.front())
1740  PrefixRegexStr.push_back('|');
1741 
1742  PrefixRegexStr.append(Prefix);
1743  }
1744 
1745  return Regex(PrefixRegexStr);
1746 }
1747 
1749  std::vector<std::string> &CmdlineDefines, SourceMgr &SM) {
1750  assert(GlobalVariableTable.empty() && GlobalNumericVariableTable.empty() &&
1751  "Overriding defined variable with command-line variable definitions");
1752 
1753  if (CmdlineDefines.empty())
1754  return Error::success();
1755 
1756  // Create a string representing the vector of command-line definitions. Each
1757  // definition is on its own line and prefixed with a definition number to
1758  // clarify which definition a given diagnostic corresponds to.
1759  unsigned I = 0;
1760  Error Errs = Error::success();
1761  std::string CmdlineDefsDiag;
1762  StringRef Prefix1 = "Global define #";
1763  StringRef Prefix2 = ": ";
1764  for (StringRef CmdlineDef : CmdlineDefines)
1765  CmdlineDefsDiag +=
1766  (Prefix1 + Twine(++I) + Prefix2 + CmdlineDef + "\n").str();
1767 
1768  // Create a buffer with fake command line content in order to display
1769  // parsing diagnostic with location information and point to the
1770  // global definition with invalid syntax.
1771  std::unique_ptr<MemoryBuffer> CmdLineDefsDiagBuffer =
1772  MemoryBuffer::getMemBufferCopy(CmdlineDefsDiag, "Global defines");
1773  StringRef CmdlineDefsDiagRef = CmdLineDefsDiagBuffer->getBuffer();
1774  SM.AddNewSourceBuffer(std::move(CmdLineDefsDiagBuffer), SMLoc());
1775 
1776  SmallVector<StringRef, 4> CmdlineDefsDiagVec;
1777  CmdlineDefsDiagRef.split(CmdlineDefsDiagVec, '\n', -1 /*MaxSplit*/,
1778  false /*KeepEmpty*/);
1779  for (StringRef CmdlineDefDiag : CmdlineDefsDiagVec) {
1780  unsigned DefStart = CmdlineDefDiag.find(Prefix2) + Prefix2.size();
1781  StringRef CmdlineDef = CmdlineDefDiag.substr(DefStart);
1782  size_t EqIdx = CmdlineDef.find('=');
1783  if (EqIdx == StringRef::npos) {
1784  Errs = joinErrors(
1785  std::move(Errs),
1787  SM, CmdlineDef, "missing equal sign in global definition"));
1788  continue;
1789  }
1790 
1791  // Numeric variable definition.
1792  if (CmdlineDef[0] == '#') {
1793  StringRef CmdlineName = CmdlineDef.substr(1, EqIdx - 1);
1796  None, SM);
1797  if (!ParseResult) {
1798  Errs = joinErrors(std::move(Errs), ParseResult.takeError());
1799  continue;
1800  }
1801 
1802  StringRef CmdlineVal = CmdlineDef.substr(EqIdx + 1);
1803  uint64_t Val;
1804  if (CmdlineVal.getAsInteger(10, Val)) {
1805  Errs = joinErrors(std::move(Errs),
1807  SM, CmdlineVal,
1808  "invalid value in numeric variable definition '" +
1809  CmdlineVal + "'"));
1810  continue;
1811  }
1812  FileCheckNumericVariable *DefinedNumericVariable = *ParseResult;
1813  DefinedNumericVariable->setValue(Val);
1814 
1815  // Record this variable definition.
1816  GlobalNumericVariableTable[DefinedNumericVariable->getName()] =
1817  DefinedNumericVariable;
1818  } else {
1819  // String variable definition.
1820  std::pair<StringRef, StringRef> CmdlineNameVal = CmdlineDef.split('=');
1821  StringRef CmdlineName = CmdlineNameVal.first;
1822  StringRef OrigCmdlineName = CmdlineName;
1824  FileCheckPattern::parseVariable(CmdlineName, SM);
1825  if (!ParseVarResult) {
1826  Errs = joinErrors(std::move(Errs), ParseVarResult.takeError());
1827  continue;
1828  }
1829  // Check that CmdlineName does not denote a pseudo variable is only
1830  // composed of the parsed numeric variable. This catches cases like
1831  // "FOO+2" in a "FOO+2=10" definition.
1832  if (ParseVarResult->IsPseudo || !CmdlineName.empty()) {
1833  Errs = joinErrors(std::move(Errs),
1835  SM, OrigCmdlineName,
1836  "invalid name in string variable definition '" +
1837  OrigCmdlineName + "'"));
1838  continue;
1839  }
1840  StringRef Name = ParseVarResult->Name;
1841 
1842  // Detect collisions between string and numeric variables when the former
1843  // is created later than the latter.
1844  if (GlobalNumericVariableTable.find(Name) !=
1845  GlobalNumericVariableTable.end()) {
1846  Errs = joinErrors(std::move(Errs), FileCheckErrorDiagnostic::get(
1847  SM, Name,
1848  "numeric variable with name '" +
1849  Name + "' already exists"));
1850  continue;
1851  }
1852  GlobalVariableTable.insert(CmdlineNameVal);
1853  // Mark the string variable as defined to detect collisions between
1854  // string and numeric variables in DefineCmdlineVariables when the latter
1855  // is created later than the former. We cannot reuse GlobalVariableTable
1856  // for this by populating it with an empty string since we would then
1857  // lose the ability to detect the use of an undefined variable in
1858  // match().
1859  DefinedVariableTable[Name] = true;
1860  }
1861  }
1862 
1863  return Errs;
1864 }
1865 
1867  SmallVector<StringRef, 16> LocalPatternVars, LocalNumericVars;
1868  for (const StringMapEntry<StringRef> &Var : GlobalVariableTable)
1869  if (Var.first()[0] != '$')
1870  LocalPatternVars.push_back(Var.first());
1871 
1872  // Numeric substitution reads the value of a variable directly, not via
1873  // GlobalNumericVariableTable. Therefore, we clear local variables by
1874  // clearing their value which will lead to a numeric substitution failure. We
1875  // also mark the variable for removal from GlobalNumericVariableTable since
1876  // this is what defineCmdlineVariables checks to decide that no global
1877  // variable has been defined.
1878  for (const auto &Var : GlobalNumericVariableTable)
1879  if (Var.first()[0] != '$') {
1880  Var.getValue()->clearValue();
1881  LocalNumericVars.push_back(Var.first());
1882  }
1883 
1884  for (const auto &Var : LocalPatternVars)
1885  GlobalVariableTable.erase(Var);
1886  for (const auto &Var : LocalNumericVars)
1887  GlobalNumericVariableTable.erase(Var);
1888 }
1889 
1891  ArrayRef<FileCheckString> CheckStrings,
1892  std::vector<FileCheckDiag> *Diags) {
1893  bool ChecksFailed = false;
1894 
1895  unsigned i = 0, j = 0, e = CheckStrings.size();
1896  while (true) {
1897  StringRef CheckRegion;
1898  if (j == e) {
1899  CheckRegion = Buffer;
1900  } else {
1901  const FileCheckString &CheckLabelStr = CheckStrings[j];
1902  if (CheckLabelStr.Pat.getCheckTy() != Check::CheckLabel) {
1903  ++j;
1904  continue;
1905  }
1906 
1907  // Scan to next CHECK-LABEL match, ignoring CHECK-NOT and CHECK-DAG
1908  size_t MatchLabelLen = 0;
1909  size_t MatchLabelPos =
1910  CheckLabelStr.Check(SM, Buffer, true, MatchLabelLen, Req, Diags);
1911  if (MatchLabelPos == StringRef::npos)
1912  // Immediately bail if CHECK-LABEL fails, nothing else we can do.
1913  return false;
1914 
1915  CheckRegion = Buffer.substr(0, MatchLabelPos + MatchLabelLen);
1916  Buffer = Buffer.substr(MatchLabelPos + MatchLabelLen);
1917  ++j;
1918  }
1919 
1920  // Do not clear the first region as it's the one before the first
1921  // CHECK-LABEL and it would clear variables defined on the command-line
1922  // before they get used.
1923  if (i != 0 && Req.EnableVarScope)
1924  PatternContext.clearLocalVars();
1925 
1926  for (; i != j; ++i) {
1927  const FileCheckString &CheckStr = CheckStrings[i];
1928 
1929  // Check each string within the scanned region, including a second check
1930  // of any final CHECK-LABEL (to verify CHECK-NOT and CHECK-DAG)
1931  size_t MatchLen = 0;
1932  size_t MatchPos =
1933  CheckStr.Check(SM, CheckRegion, false, MatchLen, Req, Diags);
1934 
1935  if (MatchPos == StringRef::npos) {
1936  ChecksFailed = true;
1937  i = j;
1938  break;
1939  }
1940 
1941  CheckRegion = CheckRegion.substr(MatchPos + MatchLen);
1942  }
1943 
1944  if (j == e)
1945  break;
1946  }
1947 
1948  // Success if no checks failed.
1949  return !ChecksFailed;
1950 }
uint64_t CallInst * C
static void PrintNoMatch(bool ExpectedMatch, const SourceMgr &SM, StringRef Prefix, SMLoc Loc, const FileCheckPattern &Pat, int MatchedCount, StringRef Buffer, bool VerboseVerbose, std::vector< FileCheckDiag > *Diags, Error MatchErrors)
Definition: FileCheck.cpp:1314
static Error get(const SourceMgr &SM, SMLoc Loc, const Twine &ErrMsg)
Definition: FileCheck.h:392
Represents a range in source code.
Definition: SMLoc.h:48
LLVM_NODISCARD StringRef take_front(size_t N=1) const
Return a StringRef equal to &#39;this&#39; but with only the first N elements remaining.
Definition: StringRef.h:587
static Expected< FileCheckNumericVariable * > parseNumericVariableDefinition(StringRef &Expr, FileCheckPatternContext *Context, Optional< size_t > LineNumber, const SourceMgr &SM)
Parses Expr for the name of a numeric variable to be defined at line LineNumber or before input is pa...
Definition: FileCheck.cpp:126
Indicates a good match for an expected pattern.
Definition: FileCheck.h:634
raw_ostream & errs()
This returns a reference to a raw_ostream for standard error.
Indicates no match for an excluded pattern.
Definition: FileCheck.h:643
Parsing information about a variable.
Definition: FileCheck.h:499
GCNRegPressure max(const GCNRegPressure &P1, const GCNRegPressure &P2)
LLVM_NODISCARD std::string str() const
str - Get the contents as an std::string.
Definition: StringRef.h:218
LLVMContext & Context
This class represents lattice values for constants.
Definition: AllocatorList.h:23
StringMapEntry - This is used to represent one value that is inserted into a StringMap.
Definition: StringMap.h:125
Indicates a discarded match for an expected pattern.
Definition: FileCheck.h:641
amdgpu Simplify well known AMD library false FunctionCallee Value const Twine & Name
LLVM_NODISCARD bool startswith(StringRef Prefix) const
Check if this string starts with the given Prefix.
Definition: StringRef.h:256
void push_back(const T &Elt)
Definition: SmallVector.h:211
bool consume_front(StringRef Prefix)
Returns true if this StringRef has the given prefix and removes that prefix.
Definition: StringRef.h:649
bool CheckSame(const SourceMgr &SM, StringRef Buffer) const
Verifies that there is no newline in the given Buffer.
Definition: FileCheck.cpp:1526
static std::string escape(StringRef String)
Turn String into a regex by escaping its special characters.
Definition: Regex.cpp:200
iterator find(StringRef Key)
Definition: StringMap.h:332
A raw_ostream that writes to an SmallVector or SmallString.
Definition: raw_ostream.h:509
auto formatv(const char *Fmt, Ts &&... Vals) -> formatv_object< decltype(std::make_tuple(detail::build_format_adapter(std::forward< Ts >(Vals))...))>
std::string getDescription(StringRef Prefix) const
Definition: FileCheck.cpp:942
Marks when parsing found a -NOT check combined with another CHECK suffix.
Definition: FileCheck.h:270
size_t getBufferSize() const
Definition: MemoryBuffer.h:61
Regex buildCheckPrefixRegex()
Definition: FileCheck.cpp:1729
LLVM_NODISCARD StringRef ltrim(char Char) const
Return string with consecutive Char characters starting from the the left removed.
Definition: StringRef.h:789
Error takeError()
Take ownership of the stored error.
Definition: Error.h:552
void printSubstitutions(const SourceMgr &SM, StringRef Buffer, SMRange MatchRange=None) const
Prints the value of successful substitutions or the name of the undefined string or numeric variables...
Definition: FileCheck.cpp:707
void reserve(size_type N)
Definition: SmallVector.h:369
std::pair< unsigned, unsigned > getLineAndColumn(SMLoc Loc, unsigned BufferID=0) const
Find the line and column number for the specified location in the specified file. ...
Definition: SourceMgr.cpp:125
LLVM_NODISCARD unsigned edit_distance(StringRef Other, bool AllowReplacements=true, unsigned MaxEditDistance=0) const
Determine the edit distance between this string and another string.
Definition: StringRef.cpp:94
size_t CheckDag(const SourceMgr &SM, StringRef Buffer, std::vector< const FileCheckPattern *> &NotStrings, const FileCheckRequest &Req, std::vector< FileCheckDiag > *Diags) const
Matches "dag strings" and their mixed "not strings".
Definition: FileCheck.cpp:1575
bool CheckNext(const SourceMgr &SM, StringRef Buffer) const
Verifies that there is a single line in the given Buffer.
Definition: FileCheck.cpp:1487
SMLoc Start
Definition: SMLoc.h:50
Class representing a numeric variable and its associated current value.
Definition: FileCheck.h:92
static LLVM_ATTRIBUTE_ALWAYS_INLINE bool CheckSame(const unsigned char *MatcherTable, unsigned &MatcherIndex, SDValue N, const SmallVectorImpl< std::pair< SDValue, SDNode *>> &RecordedNodes)
CheckSame - Implements OP_CheckSame.
StringRef Prefix
Which prefix name this check matched.
Definition: FileCheck.h:672
bool match(Val *V, const Pattern &P)
Definition: PatternMatch.h:47
FileCheckPattern Pat
The pattern to match.
Definition: FileCheck.h:669
void PrintMessage(raw_ostream &OS, SMLoc Loc, DiagKind Kind, const Twine &Msg, ArrayRef< SMRange > Ranges=None, ArrayRef< SMFixIt > FixIts=None, bool ShowColors=true) const
Emit a message about the specified location with the specified string.
Definition: SourceMgr.cpp:242
Class holding the FileCheckPattern global state, shared by all patterns: tables holding values of var...
Definition: FileCheck.h:299
Twine - A lightweight data structure for efficiently representing the concatenation of temporary valu...
Definition: Twine.h:80
LLVM_NODISCARD StringRef drop_front(size_t N=1) const
Return a StringRef equal to &#39;this&#39; but with the first N elements dropped.
Definition: StringRef.h:620
LLVM_NODISCARD size_t count(char C) const
Return the number of occurrences of C in the string.
Definition: StringRef.h:457
SMLoc Loc
The location in the match file that the check string was specified.
Definition: FileCheck.h:675
Check::FileCheckType CheckTy
What is the FileCheck directive for this diagnostic?
Definition: FileCheck.h:615
Indicates a match for an expected pattern, but the match is on the wrong line.
Definition: FileCheck.h:639
bool parsePattern(StringRef PatternStr, StringRef Prefix, SourceMgr &SM, const FileCheckRequest &Req)
Parses the pattern in PatternStr and initializes this FileCheckPattern instance accordingly.
Definition: FileCheck.cpp:319
LLVM_NODISCARD StringRef slice(size_t Start, size_t End) const
Return a reference to the substring from [Start, End).
Definition: StringRef.h:679
Expected< std::string > getResult() const override
Definition: FileCheck.cpp:63
Class representing a substitution to perform in the RegExStr string.
Definition: FileCheck.h:184
Tagged union holding either a T or a Error.
Definition: CachePruning.h:22
LLVM_NODISCARD StringRef substr(size_t Start, size_t N=npos) const
Return a reference to the substring from [Start, Start + N).
Definition: StringRef.h:578
LLVM_NODISCARD bool empty() const
empty - Check if the string is empty.
Definition: StringRef.h:126
Marks when parsing found a -COUNT directive with invalid count value.
Definition: FileCheck.h:273
Expected< std::string > getResult() const override
Definition: FileCheck.cpp:70
Class to represent an error holding a diagnostic with location information used when printing it...
Definition: FileCheck.h:376
static std::pair< Check::FileCheckType, StringRef > FindCheckType(StringRef Buffer, StringRef Prefix)
Definition: FileCheck.cpp:973
std::enable_if< std::numeric_limits< T >::is_signed, bool >::type consumeInteger(unsigned Radix, T &Result)
Parse the current string as an integer of the specified radix.
Definition: StringRef.h:512
static unsigned CountNumNewlinesBetween(StringRef Range, const char *&FirstNewLine)
Counts the number of newlines in the specified range.
Definition: FileCheck.cpp:1383
unsigned AddNewSourceBuffer(std::unique_ptr< MemoryBuffer > F, SMLoc IncludeLoc)
Add a new source buffer to this source manager.
Definition: SourceMgr.h:153
ArrayRef - Represent a constant reference to an array (0 or more elements consecutively in memory)...
Definition: APInt.h:32
Contains info about various FileCheck options.
Definition: FileCheck.h:26
static bool ValidateCheckPrefix(StringRef CheckPrefix)
Definition: FileCheck.cpp:1706
LLVM_NODISCARD size_t size() const
size - Get the string size.
Definition: StringRef.h:130
Expected< uint64_t > eval() const
Definition: FileCheck.cpp:38
size_t Check(const SourceMgr &SM, StringRef Buffer, bool IsLabelScanMode, size_t &MatchLen, FileCheckRequest &Req, std::vector< FileCheckDiag > *Diags) const
Matches check string and its "not strings" and/or "dag strings".
Definition: FileCheck.cpp:1405
void append(in_iter S, in_iter E)
Append from an iterator pair.
Definition: SmallString.h:74
bool ValidateCheckPrefixes()
Definition: FileCheck.cpp:1711
unsigned InputEndLine
Definition: FileCheck.h:656
#define P(N)
A check that we found in the input file.
Definition: FileCheck.h:667
Expected< std::unique_ptr< FileCheckExpressionAST > > parseNumericSubstitutionBlock(StringRef Expr, Optional< FileCheckNumericVariable *> &DefinedNumericVariable, bool IsLegacyLineExpr, const SourceMgr &SM) const
Parses Expr for a numeric substitution block.
Definition: FileCheck.cpp:270
Compile for newline-sensitive matching.
Definition: Regex.h:38
void clearLocalVars()
Undefines local variables (variables whose name does not start with a &#39;$&#39; sign), i.e.
Definition: FileCheck.cpp:1866
FileCheckDiag(const SourceMgr &SM, const Check::FileCheckType &CheckTy, SMLoc CheckLoc, MatchType MatchTy, SMRange InputRange)
Definition: FileCheck.cpp:914
The instances of the Type class are immutable: once they are created, they are never changed...
Definition: Type.h:45
StringRef CanonicalizeFile(MemoryBuffer &MB, SmallVectorImpl< char > &OutputBuffer)
Canonicalizes whitespaces in the file.
Definition: FileCheck.cpp:885
size_t size() const
size - Get the array size.
Definition: ArrayRef.h:148
static GCRegistry::Add< CoreCLRGC > E("coreclr", "CoreCLR-compatible GC")
LLVM_NODISCARD size_t find_first_not_of(char C, size_t From=0) const
Find the first character in the string that is not C or npos if not found.
Definition: StringRef.cpp:249
LLVM_NODISCARD size_t find(char C, size_t From=0) const
Search for the first character C in the string.
Definition: StringRef.h:285
unsigned CheckLine
Where is the FileCheck directive for this diagnostic?
Definition: FileCheck.h:617
SMLoc getLoc() const
Definition: FileCheck.h:489
void createLineVariable()
Create pseudo variable.
Definition: FileCheck.cpp:1104
void clearValue()
Clears value of this numeric variable, regardless of whether it is currently defined or not...
Definition: FileCheck.cpp:32
This owns the files read by a parser, handles include stacks, and handles diagnostic wrangling...
Definition: SourceMgr.h:41
Check::FileCheckType getCheckTy() const
Definition: FileCheck.h:566
void consumeError(Error Err)
Consume a Error without doing anything.
Definition: Error.h:981
size_t size() const
Definition: SmallVector.h:52
LLVM_NODISCARD char back() const
back - Get the last character in the string.
Definition: StringRef.h:141
void logAllUnhandledErrors(Error E, raw_ostream &OS, Twine ErrorBanner={})
Log all errors (if any) in E to OS.
Definition: Error.cpp:61
Class to represent an undefined variable error, which quotes that variable&#39;s name when printed...
Definition: FileCheck.h:69
#define llvm_unreachable(msg)
Marks that the current location is not supposed to be reachable.
std::pair< typename base::iterator, bool > insert(StringRef Key)
Definition: StringSet.h:38
bool ReadCheckFile(SourceMgr &SM, StringRef Buffer, Regex &PrefixRE, std::vector< FileCheckString > &CheckStrings)
Reads the check file from Buffer and records the expected strings it contains in the CheckStrings vec...
Definition: FileCheck.cpp:1111
StringRef getName() const
Definition: FileCheck.h:113
std::enable_if< std::numeric_limits< T >::is_signed, bool >::type getAsInteger(unsigned Radix, T &Result) const
Parse the current string as an integer of the specified radix.
Definition: StringRef.h:478
Indicates the pattern only matches the end of file.
Definition: FileCheck.h:267
bool isValid() const
Definition: SMLoc.h:59
static size_t SkipWord(StringRef Str, size_t Loc)
Definition: FileCheck.cpp:1028
void handleAllErrors(Error E, HandlerTs &&... Handlers)
Behaves the same as handleErrors, except that by contract all errors must be handled by the given han...
Definition: Error.h:904
raw_ostream & write_escaped(StringRef Str, bool UseHexEscapes=false)
Output Str, turning &#39;\&#39;, &#39;&#39;, &#39; &#39;, &#39;"&#39;, and anything that doesn&#39;t satisfy llvm::isPrint into an escape...
static ErrorSuccess success()
Create a success value.
Definition: Error.h:326
Indicates a fuzzy match that serves as a suggestion for the next intended match for an expected patte...
Definition: FileCheck.h:650
static std::pair< StringRef, StringRef > FindFirstMatchingPrefix(Regex &PrefixRE, StringRef &Buffer, unsigned &LineNumber, Check::FileCheckType &CheckTy)
Searches the buffer for the first prefix in the prefix regular expression.
Definition: FileCheck.cpp:1058
void log(raw_ostream &OS) const override
Print diagnostic associated with this error when printing the error.
Definition: FileCheck.h:390
static uint64_t add(uint64_t LeftOp, uint64_t RightOp)
Definition: FileCheck.cpp:218
This is a &#39;vector&#39; (really, a variable-sized array), optimized for the case when the array is small...
Definition: SmallVector.h:837
This is a utility class that provides an abstraction for the common functionality between Instruction...
Definition: Operator.h:30
Indicates no match for an expected pattern, but this might follow good matches when multiple matches ...
Definition: FileCheck.h:647
LLVM_NODISCARD std::pair< StringRef, StringRef > split(char Separator) const
Split into two substrings around the first occurrence of a separator character.
Definition: StringRef.h:696
static Expected< VariableProperties > parseVariable(StringRef &Str, const SourceMgr &SM)
Parses the string at the start of Str for a variable name.
Definition: FileCheck.cpp:83
bool isValid(std::string &Error) const
isValid - returns the error encountered during regex compilation, or matching, if any...
Definition: Regex.cpp:55
StringRef str()
Return a StringRef for the vector contents.
Definition: raw_ostream.h:534
unsigned InputEndCol
Definition: FileCheck.h:657
uint64_t(*)(uint64_t, uint64_t) binop_eval_t
Type of functions evaluating a given binary operation.
Definition: FileCheck.h:151
std::string utostr(uint64_t X, bool isNeg=false)
Definition: StringExtras.h:223
Expected< uint64_t > eval() const
Evaluates the value of the binary operation represented by this AST, using EvalBinop on the result of...
Definition: FileCheck.cpp:45
This interface provides simple read-only access to a block of memory, and provides simple methods for...
Definition: MemoryBuffer.h:41
static void PrintMatch(bool ExpectedMatch, const SourceMgr &SM, StringRef Prefix, SMLoc Loc, const FileCheckPattern &Pat, int MatchedCount, StringRef Buffer, size_t MatchPos, size_t MatchLen, const FileCheckRequest &Req, std::vector< FileCheckDiag > *Diags)
Definition: FileCheck.cpp:1268
void swap(llvm::BitVector &LHS, llvm::BitVector &RHS)
Implement std::swap in terms of BitVector swap.
Definition: BitVector.h:940
StringRef SpaceChars
Definition: FileCheck.cpp:112
static std::unique_ptr< MemoryBuffer > getMemBufferCopy(StringRef InputData, const Twine &BufferName="")
Open the specified memory range as a MemoryBuffer, copying the contents and taking ownership of it...
static SMRange 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 AdjustPrevDiag=false)
Definition: FileCheck.cpp:749
bool CheckNot(const SourceMgr &SM, StringRef Buffer, const std::vector< const FileCheckPattern *> &NotStrings, const FileCheckRequest &Req, std::vector< FileCheckDiag > *Diags) const
Verifies that none of the strings in NotStrings are found in the given Buffer.
Definition: FileCheck.cpp:1548
SMLoc End
Definition: SMLoc.h:50
void printFuzzyMatch(const SourceMgr &SM, StringRef Buffer, std::vector< FileCheckDiag > *Diags) const
Definition: FileCheck.cpp:767
unsigned getNumMatches() const
getNumMatches - In a valid regex, return the number of parenthesized matches it contains.
Definition: Regex.cpp:68
Expected< StringRef > getPatternVarValue(StringRef VarName)
Definition: FileCheck.cpp:815
iterator begin() const
Definition: StringRef.h:101
static SMLoc getFromPointer(const char *Ptr)
Definition: SMLoc.h:36
unsigned InputStartLine
The search range if MatchTy is MatchNoneAndExcluded or MatchNoneButExpected, or the match range other...
Definition: FileCheck.h:654
LLVM_NODISCARD bool equals(StringRef RHS) const
equals - Check for string equality, this is more efficient than compare() when the relative ordering ...
Definition: StringRef.h:160
bool CheckInput(SourceMgr &SM, StringRef Buffer, ArrayRef< FileCheckString > CheckStrings, std::vector< FileCheckDiag > *Diags=nullptr)
Checks the input to FileCheck provided in the Buffer against the CheckStrings read from the check fil...
Definition: FileCheck.cpp:1890
Indicates a match for an excluded pattern.
Definition: FileCheck.h:636
const char * getBufferEnd() const
Definition: MemoryBuffer.h:60
FileCheckType & setCount(int C)
Definition: FileCheck.cpp:934
pointer data()
Return a pointer to the vector&#39;s buffer, even if empty().
Definition: SmallVector.h:144
static const size_t npos
Definition: StringRef.h:50
LLVM_NODISCARD bool empty() const
Definition: SmallVector.h:55
Error joinErrors(Error E1, Error E2)
Concatenate errors.
Definition: Error.h:423
LLVM_NODISCARD size_t find_first_of(char C, size_t From=0) const
Find the first character in the string that is C, or npos if not found.
Definition: StringRef.h:380
#define I(x, y, z)
Definition: MD5.cpp:58
const char * getBufferStart() const
Definition: MemoryBuffer.h:59
static bool IsPartOfWord(char c)
Definition: FileCheck.cpp:930
LLVM_NODISCARD const char * data() const
data - Get a pointer to the start of the string (which may not be null terminated).
Definition: StringRef.h:122
unsigned InputStartCol
Definition: FileCheck.h:655
Expected< size_t > match(StringRef Buffer, size_t &MatchLen, const SourceMgr &SM) const
Matches the pattern string against the input buffer Buffer.
Definition: FileCheck.cpp:599
assert(ImpDefSCC.getReg()==AMDGPU::SCC &&ImpDefSCC.isDef())
LLVM_NODISCARD char front() const
front - Get the first character in the string.
Definition: StringRef.h:134
Error defineCmdlineVariables(std::vector< std::string > &CmdlineDefines, SourceMgr &SM)
Defines string and numeric variables from definitions given on the command line, passed as a vector o...
Definition: FileCheck.cpp:1748
LLVM Value Representation.
Definition: Value.h:72
static char popFront(StringRef &S)
Definition: FileCheck.cpp:115
Lightweight error class with error context and mandatory checking.
Definition: Error.h:157
void setValue(uint64_t Value)
Sets value of this numeric variable, if undefined.
Definition: FileCheck.cpp:27
StringSet - A wrapper for StringMap that provides set-like functionality.
Definition: StringSet.h:27
Error handleErrors(Error E, HandlerTs &&... Hs)
Pass the ErrorInfo(s) contained in E to their respective handlers.
Definition: Error.h:881
static uint64_t sub(uint64_t LeftOp, uint64_t RightOp)
Definition: FileCheck.cpp:222
bool match(StringRef String, SmallVectorImpl< StringRef > *Matches=nullptr)
matches - Match the regex against a given String.
Definition: Regex.cpp:72
IRTranslator LLVM IR MI
StringRef - Represent a constant reference to a string, i.e.
Definition: StringRef.h:48
MatchType
What type of match result does this diagnostic describe?
Definition: FileCheck.h:632
int getCount() const
Definition: FileCheck.h:568
Represents a location in source code.
Definition: SMLoc.h:23
iterator end() const
Definition: StringRef.h:103
static bool isValidVarNameStart(char C)
Definition: FileCheck.cpp:78
iterator end()
Definition: StringMap.h:317