1 //===- FileCheck.cpp - Check that File's Contents match what is expected --===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // FileCheck does a line-by line check of a file that validates whether it
11 // contains the expected content. This is useful for regression tests etc.
13 // This program exits with an error status of 2 on error, exit status of 0 if
14 // the file matched the expected contents, and exit status of 1 if it did not
15 // contain the expected contents.
17 //===----------------------------------------------------------------------===//
19 #include "llvm/ADT/OwningPtr.h"
20 #include "llvm/ADT/SmallString.h"
21 #include "llvm/ADT/StringExtras.h"
22 #include "llvm/ADT/StringMap.h"
23 #include "llvm/Support/CommandLine.h"
24 #include "llvm/Support/MemoryBuffer.h"
25 #include "llvm/Support/PrettyStackTrace.h"
26 #include "llvm/Support/Regex.h"
27 #include "llvm/Support/Signals.h"
28 #include "llvm/Support/SourceMgr.h"
29 #include "llvm/Support/raw_ostream.h"
30 #include "llvm/Support/system_error.h"
37 static cl::opt<std::string>
38 CheckFilename(cl::Positional, cl::desc("<check-file>"), cl::Required);
40 static cl::opt<std::string>
41 InputFilename("input-file", cl::desc("File to check (defaults to stdin)"),
42 cl::init("-"), cl::value_desc("filename"));
44 static cl::opt<std::string>
45 CheckPrefix("check-prefix", cl::init("CHECK"),
46 cl::desc("Prefix to use from check file (defaults to 'CHECK')"));
49 NoCanonicalizeWhiteSpace("strict-whitespace",
50 cl::desc("Do not treat all horizontal whitespace as equivalent"));
52 //===----------------------------------------------------------------------===//
53 // Pattern Handling Code.
54 //===----------------------------------------------------------------------===//
65 /// MatchEOF - When set, this pattern only matches the end of file. This is
66 /// used for trailing CHECK-NOTs.
74 Check::CheckType CheckTy;
76 /// FixedStr - If non-empty, this pattern is a fixed string match with the
77 /// specified fixed string.
80 /// RegEx - If non-empty, this is a regex pattern.
83 /// \brief Contains the number of line this pattern is in.
86 /// VariableUses - Entries in this vector map to uses of a variable in the
87 /// pattern, e.g. "foo[[bar]]baz". In this case, the RegExStr will contain
88 /// "foobaz" and we'll get an entry in this vector that tells us to insert the
89 /// value of bar at offset 3.
90 std::vector<std::pair<StringRef, unsigned> > VariableUses;
92 /// VariableDefs - Maps definitions of variables to their parenthesized
94 /// E.g. for the pattern "foo[[bar:.*]]baz", VariableDefs will map "bar" to 1.
95 std::map<StringRef, unsigned> VariableDefs;
99 Pattern(Check::CheckType Ty)
102 /// getLoc - Return the location in source code.
103 SMLoc getLoc() const { return PatternLoc; }
105 /// ParsePattern - Parse the given string into the Pattern. SM provides the
106 /// SourceMgr used for error reports, and LineNumber is the line number in
107 /// the input file from which the pattern string was read.
108 /// Returns true in case of an error, false otherwise.
109 bool ParsePattern(StringRef PatternStr, SourceMgr &SM, unsigned LineNumber);
111 /// Match - Match the pattern string against the input buffer Buffer. This
112 /// returns the position that is matched or npos if there is no match. If
113 /// there is a match, the size of the matched string is returned in MatchLen.
115 /// The VariableTable StringMap provides the current values of filecheck
116 /// variables and is updated if this match defines new values.
117 size_t Match(StringRef Buffer, size_t &MatchLen,
118 StringMap<StringRef> &VariableTable) const;
120 /// PrintFailureInfo - Print additional information about a failure to match
121 /// involving this pattern.
122 void PrintFailureInfo(const SourceMgr &SM, StringRef Buffer,
123 const StringMap<StringRef> &VariableTable) const;
125 bool hasVariable() const { return !(VariableUses.empty() &&
126 VariableDefs.empty()); }
128 Check::CheckType getCheckTy() const { return CheckTy; }
131 static void AddFixedStringToRegEx(StringRef FixedStr, std::string &TheStr);
132 bool AddRegExToRegEx(StringRef RS, unsigned &CurParen, SourceMgr &SM);
133 void AddBackrefToRegEx(unsigned BackrefNum);
135 /// ComputeMatchDistance - Compute an arbitrary estimate for the quality of
136 /// matching this pattern at the start of \arg Buffer; a distance of zero
137 /// should correspond to a perfect match.
138 unsigned ComputeMatchDistance(StringRef Buffer,
139 const StringMap<StringRef> &VariableTable) const;
141 /// \brief Evaluates expression and stores the result to \p Value.
142 /// \return true on success. false when the expression has invalid syntax.
143 bool EvaluateExpression(StringRef Expr, std::string &Value) const;
145 /// \brief Finds the closing sequence of a regex variable usage or
146 /// definition. Str has to point in the beginning of the definition
147 /// (right after the opening sequence).
148 /// \return offset of the closing sequence within Str, or npos if it was not
150 size_t FindRegexVarEnd(StringRef Str);
154 bool Pattern::ParsePattern(StringRef PatternStr, SourceMgr &SM,
155 unsigned LineNumber) {
156 this->LineNumber = LineNumber;
157 PatternLoc = SMLoc::getFromPointer(PatternStr.data());
159 // Ignore trailing whitespace.
160 while (!PatternStr.empty() &&
161 (PatternStr.back() == ' ' || PatternStr.back() == '\t'))
162 PatternStr = PatternStr.substr(0, PatternStr.size()-1);
164 // Check that there is something on the line.
165 if (PatternStr.empty()) {
166 SM.PrintMessage(PatternLoc, SourceMgr::DK_Error,
167 "found empty check string with prefix '" +
172 // Check to see if this is a fixed string, or if it has regex pieces.
173 if (PatternStr.size() < 2 ||
174 (PatternStr.find("{{") == StringRef::npos &&
175 PatternStr.find("[[") == StringRef::npos)) {
176 FixedStr = PatternStr;
180 // Paren value #0 is for the fully matched string. Any new parenthesized
181 // values add from there.
182 unsigned CurParen = 1;
184 // Otherwise, there is at least one regex piece. Build up the regex pattern
185 // by escaping scary characters in fixed strings, building up one big regex.
186 while (!PatternStr.empty()) {
188 if (PatternStr.startswith("{{")) {
189 // This is the start of a regex match. Scan for the }}.
190 size_t End = PatternStr.find("}}");
191 if (End == StringRef::npos) {
192 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
194 "found start of regex string with no end '}}'");
198 // Enclose {{}} patterns in parens just like [[]] even though we're not
199 // capturing the result for any purpose. This is required in case the
200 // expression contains an alternation like: CHECK: abc{{x|z}}def. We
201 // want this to turn into: "abc(x|z)def" not "abcx|zdef".
205 if (AddRegExToRegEx(PatternStr.substr(2, End-2), CurParen, SM))
209 PatternStr = PatternStr.substr(End+2);
213 // Named RegEx matches. These are of two forms: [[foo:.*]] which matches .*
214 // (or some other regex) and assigns it to the FileCheck variable 'foo'. The
215 // second form is [[foo]] which is a reference to foo. The variable name
216 // itself must be of the form "[a-zA-Z_][0-9a-zA-Z_]*", otherwise we reject
217 // it. This is to catch some common errors.
218 if (PatternStr.startswith("[[")) {
219 // Find the closing bracket pair ending the match. End is going to be an
220 // offset relative to the beginning of the match string.
221 size_t End = FindRegexVarEnd(PatternStr.substr(2));
223 if (End == StringRef::npos) {
224 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
226 "invalid named regex reference, no ]] found");
230 StringRef MatchStr = PatternStr.substr(2, End);
231 PatternStr = PatternStr.substr(End+4);
233 // Get the regex name (e.g. "foo").
234 size_t NameEnd = MatchStr.find(':');
235 StringRef Name = MatchStr.substr(0, NameEnd);
238 SM.PrintMessage(SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
239 "invalid name in named regex: empty name");
243 // Verify that the name/expression is well formed. FileCheck currently
244 // supports @LINE, @LINE+number, @LINE-number expressions. The check here
245 // is relaxed, more strict check is performed in \c EvaluateExpression.
246 bool IsExpression = false;
247 for (unsigned i = 0, e = Name.size(); i != e; ++i) {
248 if (i == 0 && Name[i] == '@') {
249 if (NameEnd != StringRef::npos) {
250 SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
252 "invalid name in named regex definition");
258 if (Name[i] != '_' && !isalnum(Name[i]) &&
259 (!IsExpression || (Name[i] != '+' && Name[i] != '-'))) {
260 SM.PrintMessage(SMLoc::getFromPointer(Name.data()+i),
261 SourceMgr::DK_Error, "invalid name in named regex");
266 // Name can't start with a digit.
267 if (isdigit(static_cast<unsigned char>(Name[0]))) {
268 SM.PrintMessage(SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
269 "invalid name in named regex");
274 if (NameEnd == StringRef::npos) {
275 // Handle variables that were defined earlier on the same line by
276 // emitting a backreference.
277 if (VariableDefs.find(Name) != VariableDefs.end()) {
278 unsigned VarParenNum = VariableDefs[Name];
279 if (VarParenNum < 1 || VarParenNum > 9) {
280 SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
282 "Can't back-reference more than 9 variables");
285 AddBackrefToRegEx(VarParenNum);
287 VariableUses.push_back(std::make_pair(Name, RegExStr.size()));
292 // Handle [[foo:.*]].
293 VariableDefs[Name] = CurParen;
297 if (AddRegExToRegEx(MatchStr.substr(NameEnd+1), CurParen, SM))
303 // Handle fixed string matches.
304 // Find the end, which is the start of the next regex.
305 size_t FixedMatchEnd = PatternStr.find("{{");
306 FixedMatchEnd = std::min(FixedMatchEnd, PatternStr.find("[["));
307 AddFixedStringToRegEx(PatternStr.substr(0, FixedMatchEnd), RegExStr);
308 PatternStr = PatternStr.substr(FixedMatchEnd);
314 void Pattern::AddFixedStringToRegEx(StringRef FixedStr, std::string &TheStr) {
315 // Add the characters from FixedStr to the regex, escaping as needed. This
316 // avoids "leaning toothpicks" in common patterns.
317 for (unsigned i = 0, e = FixedStr.size(); i != e; ++i) {
318 switch (FixedStr[i]) {
319 // These are the special characters matched in "p_ere_exp".
335 TheStr += FixedStr[i];
341 bool Pattern::AddRegExToRegEx(StringRef RS, unsigned &CurParen,
345 if (!R.isValid(Error)) {
346 SM.PrintMessage(SMLoc::getFromPointer(RS.data()), SourceMgr::DK_Error,
347 "invalid regex: " + Error);
351 RegExStr += RS.str();
352 CurParen += R.getNumMatches();
356 void Pattern::AddBackrefToRegEx(unsigned BackrefNum) {
357 assert(BackrefNum >= 1 && BackrefNum <= 9 && "Invalid backref number");
358 std::string Backref = std::string("\\") +
359 std::string(1, '0' + BackrefNum);
363 bool Pattern::EvaluateExpression(StringRef Expr, std::string &Value) const {
364 // The only supported expression is @LINE([\+-]\d+)?
365 if (!Expr.startswith("@LINE"))
367 Expr = Expr.substr(StringRef("@LINE").size());
371 Expr = Expr.substr(1);
372 else if (Expr[0] != '-')
374 if (Expr.getAsInteger(10, Offset))
377 Value = llvm::itostr(LineNumber + Offset);
381 /// Match - Match the pattern string against the input buffer Buffer. This
382 /// returns the position that is matched or npos if there is no match. If
383 /// there is a match, the size of the matched string is returned in MatchLen.
384 size_t Pattern::Match(StringRef Buffer, size_t &MatchLen,
385 StringMap<StringRef> &VariableTable) const {
386 // If this is the EOF pattern, match it immediately.
387 if (CheckTy == Check::CheckEOF) {
389 return Buffer.size();
392 // If this is a fixed string pattern, just match it now.
393 if (!FixedStr.empty()) {
394 MatchLen = FixedStr.size();
395 return Buffer.find(FixedStr);
400 // If there are variable uses, we need to create a temporary string with the
402 StringRef RegExToMatch = RegExStr;
404 if (!VariableUses.empty()) {
407 unsigned InsertOffset = 0;
408 for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
411 if (VariableUses[i].first[0] == '@') {
412 if (!EvaluateExpression(VariableUses[i].first, Value))
413 return StringRef::npos;
415 StringMap<StringRef>::iterator it =
416 VariableTable.find(VariableUses[i].first);
417 // If the variable is undefined, return an error.
418 if (it == VariableTable.end())
419 return StringRef::npos;
421 // Look up the value and escape it so that we can plop it into the regex.
422 AddFixedStringToRegEx(it->second, Value);
425 // Plop it into the regex at the adjusted offset.
426 TmpStr.insert(TmpStr.begin()+VariableUses[i].second+InsertOffset,
427 Value.begin(), Value.end());
428 InsertOffset += Value.size();
431 // Match the newly constructed regex.
432 RegExToMatch = TmpStr;
436 SmallVector<StringRef, 4> MatchInfo;
437 if (!Regex(RegExToMatch, Regex::Newline).match(Buffer, &MatchInfo))
438 return StringRef::npos;
440 // Successful regex match.
441 assert(!MatchInfo.empty() && "Didn't get any match");
442 StringRef FullMatch = MatchInfo[0];
444 // If this defines any variables, remember their values.
445 for (std::map<StringRef, unsigned>::const_iterator I = VariableDefs.begin(),
446 E = VariableDefs.end();
448 assert(I->second < MatchInfo.size() && "Internal paren error");
449 VariableTable[I->first] = MatchInfo[I->second];
452 MatchLen = FullMatch.size();
453 return FullMatch.data()-Buffer.data();
456 unsigned Pattern::ComputeMatchDistance(StringRef Buffer,
457 const StringMap<StringRef> &VariableTable) const {
458 // Just compute the number of matching characters. For regular expressions, we
459 // just compare against the regex itself and hope for the best.
461 // FIXME: One easy improvement here is have the regex lib generate a single
462 // example regular expression which matches, and use that as the example
464 StringRef ExampleString(FixedStr);
465 if (ExampleString.empty())
466 ExampleString = RegExStr;
468 // Only compare up to the first line in the buffer, or the string size.
469 StringRef BufferPrefix = Buffer.substr(0, ExampleString.size());
470 BufferPrefix = BufferPrefix.split('\n').first;
471 return BufferPrefix.edit_distance(ExampleString);
474 void Pattern::PrintFailureInfo(const SourceMgr &SM, StringRef Buffer,
475 const StringMap<StringRef> &VariableTable) const{
476 // If this was a regular expression using variables, print the current
478 if (!VariableUses.empty()) {
479 for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
480 SmallString<256> Msg;
481 raw_svector_ostream OS(Msg);
482 StringRef Var = VariableUses[i].first;
485 if (EvaluateExpression(Var, Value)) {
486 OS << "with expression \"";
487 OS.write_escaped(Var) << "\" equal to \"";
488 OS.write_escaped(Value) << "\"";
490 OS << "uses incorrect expression \"";
491 OS.write_escaped(Var) << "\"";
494 StringMap<StringRef>::const_iterator it = VariableTable.find(Var);
496 // Check for undefined variable references.
497 if (it == VariableTable.end()) {
498 OS << "uses undefined variable \"";
499 OS.write_escaped(Var) << "\"";
501 OS << "with variable \"";
502 OS.write_escaped(Var) << "\" equal to \"";
503 OS.write_escaped(it->second) << "\"";
507 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
512 // Attempt to find the closest/best fuzzy match. Usually an error happens
513 // because some string in the output didn't exactly match. In these cases, we
514 // would like to show the user a best guess at what "should have" matched, to
515 // save them having to actually check the input manually.
516 size_t NumLinesForward = 0;
517 size_t Best = StringRef::npos;
518 double BestQuality = 0;
520 // Use an arbitrary 4k limit on how far we will search.
521 for (size_t i = 0, e = std::min(size_t(4096), Buffer.size()); i != e; ++i) {
522 if (Buffer[i] == '\n')
525 // Patterns have leading whitespace stripped, so skip whitespace when
526 // looking for something which looks like a pattern.
527 if (Buffer[i] == ' ' || Buffer[i] == '\t')
530 // Compute the "quality" of this match as an arbitrary combination of the
531 // match distance and the number of lines skipped to get to this match.
532 unsigned Distance = ComputeMatchDistance(Buffer.substr(i), VariableTable);
533 double Quality = Distance + (NumLinesForward / 100.);
535 if (Quality < BestQuality || Best == StringRef::npos) {
537 BestQuality = Quality;
541 // Print the "possible intended match here" line if we found something
542 // reasonable and not equal to what we showed in the "scanning from here"
544 if (Best && Best != StringRef::npos && BestQuality < 50) {
545 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + Best),
546 SourceMgr::DK_Note, "possible intended match here");
548 // FIXME: If we wanted to be really friendly we would show why the match
549 // failed, as it can be hard to spot simple one character differences.
553 size_t Pattern::FindRegexVarEnd(StringRef Str) {
554 // Offset keeps track of the current offset within the input Str
556 // [...] Nesting depth
557 size_t BracketDepth = 0;
559 while (!Str.empty()) {
560 if (Str.startswith("]]") && BracketDepth == 0)
562 if (Str[0] == '\\') {
563 // Backslash escapes the next char within regexes, so skip them both.
574 assert(BracketDepth > 0 && "Invalid regex");
583 return StringRef::npos;
587 //===----------------------------------------------------------------------===//
589 //===----------------------------------------------------------------------===//
591 /// CheckString - This is a check that we found in the input file.
593 /// Pat - The pattern to match.
596 /// Loc - The location in the match file that the check string was specified.
599 /// CheckTy - Specify what kind of check this is. e.g. CHECK-NEXT: directive,
600 /// as opposed to a CHECK: directive.
601 Check::CheckType CheckTy;
603 /// DagNotStrings - These are all of the strings that are disallowed from
604 /// occurring between this match string and the previous one (or start of
606 std::vector<Pattern> DagNotStrings;
608 CheckString(const Pattern &P, SMLoc L, Check::CheckType Ty)
609 : Pat(P), Loc(L), CheckTy(Ty) {}
611 /// Check - Match check string and its "not strings" and/or "dag strings".
612 size_t Check(const SourceMgr &SM, StringRef Buffer,
613 size_t &MatchLen, StringMap<StringRef> &VariableTable) const;
615 /// CheckNext - Verify there is a single line in the given buffer.
616 bool CheckNext(const SourceMgr &SM, StringRef Buffer) const;
618 /// CheckNot - Verify there's no "not strings" in the given buffer.
619 bool CheckNot(const SourceMgr &SM, StringRef Buffer,
620 const std::vector<const Pattern *> &NotStrings,
621 StringMap<StringRef> &VariableTable) const;
623 /// CheckDag - Match "dag strings" and their mixed "not strings".
624 size_t CheckDag(const SourceMgr &SM, StringRef Buffer,
625 std::vector<const Pattern *> &NotStrings,
626 StringMap<StringRef> &VariableTable) const;
629 /// Canonicalize whitespaces in the input file. Line endings are replaced
630 /// with UNIX-style '\n'.
632 /// \param PreserveHorizontal Don't squash consecutive horizontal whitespace
633 /// characters to a single space.
634 static MemoryBuffer *CanonicalizeInputFile(MemoryBuffer *MB,
635 bool PreserveHorizontal) {
636 SmallString<128> NewFile;
637 NewFile.reserve(MB->getBufferSize());
639 for (const char *Ptr = MB->getBufferStart(), *End = MB->getBufferEnd();
641 // Eliminate trailing dosish \r.
642 if (Ptr <= End - 2 && Ptr[0] == '\r' && Ptr[1] == '\n') {
646 // If current char is not a horizontal whitespace or if horizontal
647 // whitespace canonicalization is disabled, dump it to output as is.
648 if (PreserveHorizontal || (*Ptr != ' ' && *Ptr != '\t')) {
649 NewFile.push_back(*Ptr);
653 // Otherwise, add one space and advance over neighboring space.
654 NewFile.push_back(' ');
655 while (Ptr+1 != End &&
656 (Ptr[1] == ' ' || Ptr[1] == '\t'))
660 // Free the old buffer and return a new one.
662 MemoryBuffer::getMemBufferCopy(NewFile.str(), MB->getBufferIdentifier());
668 static bool IsPartOfWord(char c) {
669 return (isalnum(c) || c == '-' || c == '_');
672 static Check::CheckType FindCheckType(StringRef &Buffer, StringRef Prefix) {
673 char NextChar = Buffer[Prefix.size()];
675 // Verify that the : is present after the prefix.
676 if (NextChar == ':') {
677 Buffer = Buffer.substr(Prefix.size() + 1);
678 return Check::CheckPlain;
681 if (NextChar != '-') {
682 Buffer = Buffer.drop_front(1);
683 return Check::CheckNone;
686 StringRef Rest = Buffer.drop_front(Prefix.size() + 1);
687 if (Rest.startswith("NEXT:")) {
688 Buffer = Rest.drop_front(sizeof("NEXT:") - 1);
689 return Check::CheckNext;
692 if (Rest.startswith("NOT:")) {
693 Buffer = Rest.drop_front(sizeof("NOT:") - 1);
694 return Check::CheckNot;
697 if (Rest.startswith("DAG:")) {
698 Buffer = Rest.drop_front(sizeof("DAG:") - 1);
699 return Check::CheckDAG;
702 if (Rest.startswith("LABEL:")) {
703 Buffer = Rest.drop_front(sizeof("LABEL:") - 1);
704 return Check::CheckLabel;
707 Buffer = Buffer.drop_front(1);
708 return Check::CheckNone;
711 /// ReadCheckFile - Read the check file, which specifies the sequence of
712 /// expected strings. The strings are added to the CheckStrings vector.
713 /// Returns true in case of an error, false otherwise.
714 static bool ReadCheckFile(SourceMgr &SM,
715 std::vector<CheckString> &CheckStrings) {
716 OwningPtr<MemoryBuffer> File;
718 MemoryBuffer::getFileOrSTDIN(CheckFilename, File)) {
719 errs() << "Could not open check file '" << CheckFilename << "': "
720 << ec.message() << '\n';
724 // If we want to canonicalize whitespace, strip excess whitespace from the
725 // buffer containing the CHECK lines. Remove DOS style line endings.
727 CanonicalizeInputFile(File.take(), NoCanonicalizeWhiteSpace);
729 SM.AddNewSourceBuffer(F, SMLoc());
731 // Find all instances of CheckPrefix followed by : in the file.
732 StringRef Buffer = F->getBuffer();
733 std::vector<Pattern> DagNotMatches;
735 // LineNumber keeps track of the line on which CheckPrefix instances are
737 unsigned LineNumber = 1;
740 // See if Prefix occurs in the memory buffer.
741 size_t PrefixLoc = Buffer.find(CheckPrefix);
742 // If we didn't find a match, we're done.
743 if (PrefixLoc == StringRef::npos)
746 LineNumber += Buffer.substr(0, PrefixLoc).count('\n');
748 // Keep the charcter before our prefix so we can validate that we have
749 // found our prefix, and account for cases when PrefixLoc is 0.
750 Buffer = Buffer.substr(std::min(PrefixLoc-1, PrefixLoc));
752 const char *CheckPrefixStart = Buffer.data() + (PrefixLoc == 0 ? 0 : 1);
754 // Make sure we have actually found our prefix, and not a word containing
756 if (PrefixLoc != 0 && IsPartOfWord(Buffer[0])) {
757 Buffer = Buffer.substr(CheckPrefix.size());
761 // When we find a check prefix, keep track of what kind of type of CHECK we
763 Check::CheckType CheckTy = FindCheckType(Buffer, CheckPrefix);
764 if (CheckTy == Check::CheckNone)
767 // Okay, we found the prefix, yay. Remember the rest of the line, but ignore
768 // leading and trailing whitespace.
769 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t"));
771 // Scan ahead to the end of line.
772 size_t EOL = Buffer.find_first_of("\n\r");
774 // Remember the location of the start of the pattern, for diagnostics.
775 SMLoc PatternLoc = SMLoc::getFromPointer(Buffer.data());
777 // Parse the pattern.
779 if (P.ParsePattern(Buffer.substr(0, EOL), SM, LineNumber))
782 // Verify that CHECK-LABEL lines do not define or use variables
783 if ((CheckTy == Check::CheckLabel) && P.hasVariable()) {
784 SM.PrintMessage(SMLoc::getFromPointer(CheckPrefixStart),
786 "found '"+CheckPrefix+"-LABEL:' with variable definition"
791 Buffer = Buffer.substr(EOL);
793 // Verify that CHECK-NEXT lines have at least one CHECK line before them.
794 if ((CheckTy == Check::CheckNext) && CheckStrings.empty()) {
795 SM.PrintMessage(SMLoc::getFromPointer(CheckPrefixStart),
797 "found '"+CheckPrefix+"-NEXT:' without previous '"+
798 CheckPrefix+ ": line");
802 // Handle CHECK-DAG/-NOT.
803 if (CheckTy == Check::CheckDAG || CheckTy == Check::CheckNot) {
804 DagNotMatches.push_back(P);
808 // Okay, add the string we captured to the output vector and move on.
809 CheckStrings.push_back(CheckString(P,
812 std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
815 // Add an EOF pattern for any trailing CHECK-DAG/-NOTs.
816 if (!DagNotMatches.empty()) {
817 CheckStrings.push_back(CheckString(Pattern(Check::CheckEOF),
818 SMLoc::getFromPointer(Buffer.data()),
820 std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
823 if (CheckStrings.empty()) {
824 errs() << "error: no check strings found with prefix '" << CheckPrefix
832 static void PrintCheckFailed(const SourceMgr &SM, const SMLoc &Loc,
833 const Pattern &Pat, StringRef Buffer,
834 StringMap<StringRef> &VariableTable) {
835 // Otherwise, we have an error, emit an error message.
836 SM.PrintMessage(Loc, SourceMgr::DK_Error,
837 "expected string not found in input");
839 // Print the "scanning from here" line. If the current position is at the
840 // end of a line, advance to the start of the next line.
841 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t\n\r"));
843 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
844 "scanning from here");
846 // Allow the pattern to print additional information if desired.
847 Pat.PrintFailureInfo(SM, Buffer, VariableTable);
850 static void PrintCheckFailed(const SourceMgr &SM, const CheckString &CheckStr,
852 StringMap<StringRef> &VariableTable) {
853 PrintCheckFailed(SM, CheckStr.Loc, CheckStr.Pat, Buffer, VariableTable);
856 /// CountNumNewlinesBetween - Count the number of newlines in the specified
858 static unsigned CountNumNewlinesBetween(StringRef Range) {
859 unsigned NumNewLines = 0;
862 Range = Range.substr(Range.find_first_of("\n\r"));
863 if (Range.empty()) return NumNewLines;
867 // Handle \n\r and \r\n as a single newline.
868 if (Range.size() > 1 &&
869 (Range[1] == '\n' || Range[1] == '\r') &&
870 (Range[0] != Range[1]))
871 Range = Range.substr(1);
872 Range = Range.substr(1);
876 size_t CheckString::Check(const SourceMgr &SM, StringRef Buffer,
878 StringMap<StringRef> &VariableTable) const {
880 std::vector<const Pattern *> NotStrings;
882 if (CheckTy != Check::CheckLabel) {
883 // Match "dag strings" (with mixed "not strings" if any).
884 LastPos = CheckDag(SM, Buffer, NotStrings, VariableTable);
885 if (LastPos == StringRef::npos)
886 return StringRef::npos;
889 // Match itself from the last position after matching CHECK-DAG.
890 StringRef MatchBuffer = Buffer.substr(LastPos);
891 size_t MatchPos = Pat.Match(MatchBuffer, MatchLen, VariableTable);
892 if (MatchPos == StringRef::npos) {
893 PrintCheckFailed(SM, *this, MatchBuffer, VariableTable);
894 return StringRef::npos;
898 if (CheckTy != Check::CheckLabel) {
899 StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
901 // If this check is a "CHECK-NEXT", verify that the previous match was on
902 // the previous line (i.e. that there is one newline between them).
903 if (CheckNext(SM, SkippedRegion))
904 return StringRef::npos;
906 // If this match had "not strings", verify that they don't exist in the
908 if (CheckNot(SM, SkippedRegion, NotStrings, VariableTable))
909 return StringRef::npos;
915 bool CheckString::CheckNext(const SourceMgr &SM, StringRef Buffer) const {
916 if (CheckTy != Check::CheckNext)
919 // Count the number of newlines between the previous match and this one.
920 assert(Buffer.data() !=
922 SM.FindBufferContainingLoc(
923 SMLoc::getFromPointer(Buffer.data())))->getBufferStart() &&
924 "CHECK-NEXT can't be the first check in a file");
926 unsigned NumNewLines = CountNumNewlinesBetween(Buffer);
928 if (NumNewLines == 0) {
929 SM.PrintMessage(Loc, SourceMgr::DK_Error, CheckPrefix+
930 "-NEXT: is on the same line as previous match");
931 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()),
932 SourceMgr::DK_Note, "'next' match was here");
933 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
934 "previous match ended here");
938 if (NumNewLines != 1) {
939 SM.PrintMessage(Loc, SourceMgr::DK_Error, CheckPrefix+
940 "-NEXT: is not on the line after the previous match");
941 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()),
942 SourceMgr::DK_Note, "'next' match was here");
943 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
944 "previous match ended here");
951 bool CheckString::CheckNot(const SourceMgr &SM, StringRef Buffer,
952 const std::vector<const Pattern *> &NotStrings,
953 StringMap<StringRef> &VariableTable) const {
954 for (unsigned ChunkNo = 0, e = NotStrings.size();
955 ChunkNo != e; ++ChunkNo) {
956 const Pattern *Pat = NotStrings[ChunkNo];
957 assert((Pat->getCheckTy() == Check::CheckNot) && "Expect CHECK-NOT!");
960 size_t Pos = Pat->Match(Buffer, MatchLen, VariableTable);
962 if (Pos == StringRef::npos) continue;
964 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()+Pos),
966 CheckPrefix+"-NOT: string occurred!");
967 SM.PrintMessage(Pat->getLoc(), SourceMgr::DK_Note,
968 CheckPrefix+"-NOT: pattern specified here");
975 size_t CheckString::CheckDag(const SourceMgr &SM, StringRef Buffer,
976 std::vector<const Pattern *> &NotStrings,
977 StringMap<StringRef> &VariableTable) const {
978 if (DagNotStrings.empty())
982 size_t StartPos = LastPos;
984 for (unsigned ChunkNo = 0, e = DagNotStrings.size();
985 ChunkNo != e; ++ChunkNo) {
986 const Pattern &Pat = DagNotStrings[ChunkNo];
988 assert((Pat.getCheckTy() == Check::CheckDAG ||
989 Pat.getCheckTy() == Check::CheckNot) &&
990 "Invalid CHECK-DAG or CHECK-NOT!");
992 if (Pat.getCheckTy() == Check::CheckNot) {
993 NotStrings.push_back(&Pat);
997 assert((Pat.getCheckTy() == Check::CheckDAG) && "Expect CHECK-DAG!");
999 size_t MatchLen = 0, MatchPos;
1001 // CHECK-DAG always matches from the start.
1002 StringRef MatchBuffer = Buffer.substr(StartPos);
1003 MatchPos = Pat.Match(MatchBuffer, MatchLen, VariableTable);
1004 // With a group of CHECK-DAGs, a single mismatching means the match on
1005 // that group of CHECK-DAGs fails immediately.
1006 if (MatchPos == StringRef::npos) {
1007 PrintCheckFailed(SM, Pat.getLoc(), Pat, MatchBuffer, VariableTable);
1008 return StringRef::npos;
1010 // Re-calc it as the offset relative to the start of the original string.
1011 MatchPos += StartPos;
1013 if (!NotStrings.empty()) {
1014 if (MatchPos < LastPos) {
1016 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + MatchPos),
1017 SourceMgr::DK_Error,
1018 CheckPrefix+"-DAG: found a match of CHECK-DAG"
1019 " reordering across a CHECK-NOT");
1020 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + LastPos),
1022 CheckPrefix+"-DAG: the farthest match of CHECK-DAG"
1024 SM.PrintMessage(NotStrings[0]->getLoc(), SourceMgr::DK_Note,
1025 CheckPrefix+"-NOT: the crossed pattern specified"
1027 SM.PrintMessage(Pat.getLoc(), SourceMgr::DK_Note,
1028 CheckPrefix+"-DAG: the reordered pattern specified"
1030 return StringRef::npos;
1032 // All subsequent CHECK-DAGs should be matched from the farthest
1033 // position of all precedent CHECK-DAGs (including this one.)
1035 // If there's CHECK-NOTs between two CHECK-DAGs or from CHECK to
1036 // CHECK-DAG, verify that there's no 'not' strings occurred in that
1038 StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
1039 if (CheckNot(SM, SkippedRegion, NotStrings, VariableTable))
1040 return StringRef::npos;
1041 // Clear "not strings".
1045 // Update the last position with CHECK-DAG matches.
1046 LastPos = std::max(MatchPos + MatchLen, LastPos);
1052 bool ValidateCheckPrefix() {
1053 // The check prefix must contain only alphanumeric, hyphens and underscores.
1054 Regex prefixValidator("^[a-zA-Z0-9_-]*$");
1055 return prefixValidator.match(CheckPrefix);
1058 int main(int argc, char **argv) {
1059 sys::PrintStackTraceOnErrorSignal();
1060 PrettyStackTraceProgram X(argc, argv);
1061 cl::ParseCommandLineOptions(argc, argv);
1063 if (!ValidateCheckPrefix()) {
1064 errs() << "Supplied check-prefix is invalid! Prefixes must start with a "
1065 "letter and contain only alphanumeric characters, hyphens and "
1072 // Read the expected strings from the check file.
1073 std::vector<CheckString> CheckStrings;
1074 if (ReadCheckFile(SM, CheckStrings))
1077 // Open the file to check and add it to SourceMgr.
1078 OwningPtr<MemoryBuffer> File;
1080 MemoryBuffer::getFileOrSTDIN(InputFilename, File)) {
1081 errs() << "Could not open input file '" << InputFilename << "': "
1082 << ec.message() << '\n';
1086 if (File->getBufferSize() == 0) {
1087 errs() << "FileCheck error: '" << InputFilename << "' is empty.\n";
1091 // Remove duplicate spaces in the input file if requested.
1092 // Remove DOS style line endings.
1094 CanonicalizeInputFile(File.take(), NoCanonicalizeWhiteSpace);
1096 SM.AddNewSourceBuffer(F, SMLoc());
1098 /// VariableTable - This holds all the current filecheck variables.
1099 StringMap<StringRef> VariableTable;
1101 // Check that we have all of the expected strings, in order, in the input
1103 StringRef Buffer = F->getBuffer();
1105 bool hasError = false;
1107 unsigned i = 0, j = 0, e = CheckStrings.size();
1110 StringRef CheckRegion;
1112 CheckRegion = Buffer;
1114 const CheckString &CheckLabelStr = CheckStrings[j];
1115 if (CheckLabelStr.CheckTy != Check::CheckLabel) {
1120 // Scan to next CHECK-LABEL match, ignoring CHECK-NOT and CHECK-DAG
1121 size_t MatchLabelLen = 0;
1122 size_t MatchLabelPos = CheckLabelStr.Check(SM, Buffer,
1123 MatchLabelLen, VariableTable);
1124 if (MatchLabelPos == StringRef::npos) {
1129 CheckRegion = Buffer.substr(0, MatchLabelPos + MatchLabelLen);
1130 Buffer = Buffer.substr(MatchLabelPos + MatchLabelLen);
1134 for ( ; i != j; ++i) {
1135 const CheckString &CheckStr = CheckStrings[i];
1137 // Check each string within the scanned region, including a second check
1138 // of any final CHECK-LABEL (to verify CHECK-NOT and CHECK-DAG)
1139 size_t MatchLen = 0;
1140 size_t MatchPos = CheckStr.Check(SM, CheckRegion, MatchLen,
1143 if (MatchPos == StringRef::npos) {
1149 CheckRegion = CheckRegion.substr(MatchPos + MatchLen);
1156 return hasError ? 1 : 0;