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/SmallString.h"
20 #include "llvm/ADT/StringExtras.h"
21 #include "llvm/ADT/StringMap.h"
22 #include "llvm/ADT/StringSet.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"
38 static cl::opt<std::string>
39 CheckFilename(cl::Positional, cl::desc("<check-file>"), cl::Required);
41 static cl::opt<std::string>
42 InputFilename("input-file", cl::desc("File to check (defaults to stdin)"),
43 cl::init("-"), cl::value_desc("filename"));
45 static cl::list<std::string>
46 CheckPrefixes("check-prefix",
47 cl::desc("Prefix to use from check file (defaults to 'CHECK')"));
50 NoCanonicalizeWhiteSpace("strict-whitespace",
51 cl::desc("Do not treat all horizontal whitespace as equivalent"));
53 typedef cl::list<std::string>::const_iterator prefix_iterator;
55 //===----------------------------------------------------------------------===//
56 // Pattern Handling Code.
57 //===----------------------------------------------------------------------===//
68 /// MatchEOF - When set, this pattern only matches the end of file. This is
69 /// used for trailing CHECK-NOTs.
77 Check::CheckType CheckTy;
79 /// FixedStr - If non-empty, this pattern is a fixed string match with the
80 /// specified fixed string.
83 /// RegEx - If non-empty, this is a regex pattern.
86 /// \brief Contains the number of line this pattern is in.
89 /// VariableUses - Entries in this vector map to uses of a variable in the
90 /// pattern, e.g. "foo[[bar]]baz". In this case, the RegExStr will contain
91 /// "foobaz" and we'll get an entry in this vector that tells us to insert the
92 /// value of bar at offset 3.
93 std::vector<std::pair<StringRef, unsigned> > VariableUses;
95 /// VariableDefs - Maps definitions of variables to their parenthesized
97 /// E.g. for the pattern "foo[[bar:.*]]baz", VariableDefs will map "bar" to 1.
98 std::map<StringRef, unsigned> VariableDefs;
102 Pattern(Check::CheckType Ty)
105 /// getLoc - Return the location in source code.
106 SMLoc getLoc() const { return PatternLoc; }
108 /// ParsePattern - Parse the given string into the Pattern. Prefix provides
109 /// which prefix is being matched, SM provides the SourceMgr used for error
110 /// reports, and LineNumber is the line number in the input file from which
111 /// the pattern string was read. Returns true in case of an error, false
113 bool ParsePattern(StringRef PatternStr,
116 unsigned LineNumber);
118 /// Match - Match the pattern string against the input buffer Buffer. This
119 /// returns the position that is matched or npos if there is no match. If
120 /// there is a match, the size of the matched string is returned in MatchLen.
122 /// The VariableTable StringMap provides the current values of filecheck
123 /// variables and is updated if this match defines new values.
124 size_t Match(StringRef Buffer, size_t &MatchLen,
125 StringMap<StringRef> &VariableTable) const;
127 /// PrintFailureInfo - Print additional information about a failure to match
128 /// involving this pattern.
129 void PrintFailureInfo(const SourceMgr &SM, StringRef Buffer,
130 const StringMap<StringRef> &VariableTable) const;
132 bool hasVariable() const { return !(VariableUses.empty() &&
133 VariableDefs.empty()); }
135 Check::CheckType getCheckTy() const { return CheckTy; }
138 bool AddRegExToRegEx(StringRef RS, unsigned &CurParen, SourceMgr &SM);
139 void AddBackrefToRegEx(unsigned BackrefNum);
141 /// ComputeMatchDistance - Compute an arbitrary estimate for the quality of
142 /// matching this pattern at the start of \arg Buffer; a distance of zero
143 /// should correspond to a perfect match.
144 unsigned ComputeMatchDistance(StringRef Buffer,
145 const StringMap<StringRef> &VariableTable) const;
147 /// \brief Evaluates expression and stores the result to \p Value.
148 /// \return true on success. false when the expression has invalid syntax.
149 bool EvaluateExpression(StringRef Expr, std::string &Value) const;
151 /// \brief Finds the closing sequence of a regex variable usage or
152 /// definition. Str has to point in the beginning of the definition
153 /// (right after the opening sequence).
154 /// \return offset of the closing sequence within Str, or npos if it was not
156 size_t FindRegexVarEnd(StringRef Str, SourceMgr &SM);
160 bool Pattern::ParsePattern(StringRef PatternStr,
163 unsigned LineNumber) {
164 this->LineNumber = LineNumber;
165 PatternLoc = SMLoc::getFromPointer(PatternStr.data());
167 // Ignore trailing whitespace.
168 while (!PatternStr.empty() &&
169 (PatternStr.back() == ' ' || PatternStr.back() == '\t'))
170 PatternStr = PatternStr.substr(0, PatternStr.size()-1);
172 // Check that there is something on the line.
173 if (PatternStr.empty()) {
174 SM.PrintMessage(PatternLoc, SourceMgr::DK_Error,
175 "found empty check string with prefix '" +
180 // Check to see if this is a fixed string, or if it has regex pieces.
181 if (PatternStr.size() < 2 ||
182 (PatternStr.find("{{") == StringRef::npos &&
183 PatternStr.find("[[") == StringRef::npos)) {
184 FixedStr = PatternStr;
188 // Paren value #0 is for the fully matched string. Any new parenthesized
189 // values add from there.
190 unsigned CurParen = 1;
192 // Otherwise, there is at least one regex piece. Build up the regex pattern
193 // by escaping scary characters in fixed strings, building up one big regex.
194 while (!PatternStr.empty()) {
196 if (PatternStr.startswith("{{")) {
197 // This is the start of a regex match. Scan for the }}.
198 size_t End = PatternStr.find("}}");
199 if (End == StringRef::npos) {
200 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
202 "found start of regex string with no end '}}'");
206 // Enclose {{}} patterns in parens just like [[]] even though we're not
207 // capturing the result for any purpose. This is required in case the
208 // expression contains an alternation like: CHECK: abc{{x|z}}def. We
209 // want this to turn into: "abc(x|z)def" not "abcx|zdef".
213 if (AddRegExToRegEx(PatternStr.substr(2, End-2), CurParen, SM))
217 PatternStr = PatternStr.substr(End+2);
221 // Named RegEx matches. These are of two forms: [[foo:.*]] which matches .*
222 // (or some other regex) and assigns it to the FileCheck variable 'foo'. The
223 // second form is [[foo]] which is a reference to foo. The variable name
224 // itself must be of the form "[a-zA-Z_][0-9a-zA-Z_]*", otherwise we reject
225 // it. This is to catch some common errors.
226 if (PatternStr.startswith("[[")) {
227 // Find the closing bracket pair ending the match. End is going to be an
228 // offset relative to the beginning of the match string.
229 size_t End = FindRegexVarEnd(PatternStr.substr(2), SM);
231 if (End == StringRef::npos) {
232 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
234 "invalid named regex reference, no ]] found");
238 StringRef MatchStr = PatternStr.substr(2, End);
239 PatternStr = PatternStr.substr(End+4);
241 // Get the regex name (e.g. "foo").
242 size_t NameEnd = MatchStr.find(':');
243 StringRef Name = MatchStr.substr(0, NameEnd);
246 SM.PrintMessage(SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
247 "invalid name in named regex: empty name");
251 // Verify that the name/expression is well formed. FileCheck currently
252 // supports @LINE, @LINE+number, @LINE-number expressions. The check here
253 // is relaxed, more strict check is performed in \c EvaluateExpression.
254 bool IsExpression = false;
255 for (unsigned i = 0, e = Name.size(); i != e; ++i) {
256 if (i == 0 && Name[i] == '@') {
257 if (NameEnd != StringRef::npos) {
258 SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
260 "invalid name in named regex definition");
266 if (Name[i] != '_' && !isalnum(Name[i]) &&
267 (!IsExpression || (Name[i] != '+' && Name[i] != '-'))) {
268 SM.PrintMessage(SMLoc::getFromPointer(Name.data()+i),
269 SourceMgr::DK_Error, "invalid name in named regex");
274 // Name can't start with a digit.
275 if (isdigit(static_cast<unsigned char>(Name[0]))) {
276 SM.PrintMessage(SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
277 "invalid name in named regex");
282 if (NameEnd == StringRef::npos) {
283 // Handle variables that were defined earlier on the same line by
284 // emitting a backreference.
285 if (VariableDefs.find(Name) != VariableDefs.end()) {
286 unsigned VarParenNum = VariableDefs[Name];
287 if (VarParenNum < 1 || VarParenNum > 9) {
288 SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
290 "Can't back-reference more than 9 variables");
293 AddBackrefToRegEx(VarParenNum);
295 VariableUses.push_back(std::make_pair(Name, RegExStr.size()));
300 // Handle [[foo:.*]].
301 VariableDefs[Name] = CurParen;
305 if (AddRegExToRegEx(MatchStr.substr(NameEnd+1), CurParen, SM))
311 // Handle fixed string matches.
312 // Find the end, which is the start of the next regex.
313 size_t FixedMatchEnd = PatternStr.find("{{");
314 FixedMatchEnd = std::min(FixedMatchEnd, PatternStr.find("[["));
315 RegExStr += Regex::escape(PatternStr.substr(0, FixedMatchEnd));
316 PatternStr = PatternStr.substr(FixedMatchEnd);
322 bool Pattern::AddRegExToRegEx(StringRef RS, unsigned &CurParen,
326 if (!R.isValid(Error)) {
327 SM.PrintMessage(SMLoc::getFromPointer(RS.data()), SourceMgr::DK_Error,
328 "invalid regex: " + Error);
332 RegExStr += RS.str();
333 CurParen += R.getNumMatches();
337 void Pattern::AddBackrefToRegEx(unsigned BackrefNum) {
338 assert(BackrefNum >= 1 && BackrefNum <= 9 && "Invalid backref number");
339 std::string Backref = std::string("\\") +
340 std::string(1, '0' + BackrefNum);
344 bool Pattern::EvaluateExpression(StringRef Expr, std::string &Value) const {
345 // The only supported expression is @LINE([\+-]\d+)?
346 if (!Expr.startswith("@LINE"))
348 Expr = Expr.substr(StringRef("@LINE").size());
352 Expr = Expr.substr(1);
353 else if (Expr[0] != '-')
355 if (Expr.getAsInteger(10, Offset))
358 Value = llvm::itostr(LineNumber + Offset);
362 /// Match - Match the pattern string against the input buffer Buffer. This
363 /// returns the position that is matched or npos if there is no match. If
364 /// there is a match, the size of the matched string is returned in MatchLen.
365 size_t Pattern::Match(StringRef Buffer, size_t &MatchLen,
366 StringMap<StringRef> &VariableTable) const {
367 // If this is the EOF pattern, match it immediately.
368 if (CheckTy == Check::CheckEOF) {
370 return Buffer.size();
373 // If this is a fixed string pattern, just match it now.
374 if (!FixedStr.empty()) {
375 MatchLen = FixedStr.size();
376 return Buffer.find(FixedStr);
381 // If there are variable uses, we need to create a temporary string with the
383 StringRef RegExToMatch = RegExStr;
385 if (!VariableUses.empty()) {
388 unsigned InsertOffset = 0;
389 for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
392 if (VariableUses[i].first[0] == '@') {
393 if (!EvaluateExpression(VariableUses[i].first, Value))
394 return StringRef::npos;
396 StringMap<StringRef>::iterator it =
397 VariableTable.find(VariableUses[i].first);
398 // If the variable is undefined, return an error.
399 if (it == VariableTable.end())
400 return StringRef::npos;
402 // Look up the value and escape it so that we can put it into the regex.
403 Value += Regex::escape(it->second);
406 // Plop it into the regex at the adjusted offset.
407 TmpStr.insert(TmpStr.begin()+VariableUses[i].second+InsertOffset,
408 Value.begin(), Value.end());
409 InsertOffset += Value.size();
412 // Match the newly constructed regex.
413 RegExToMatch = TmpStr;
417 SmallVector<StringRef, 4> MatchInfo;
418 if (!Regex(RegExToMatch, Regex::Newline).match(Buffer, &MatchInfo))
419 return StringRef::npos;
421 // Successful regex match.
422 assert(!MatchInfo.empty() && "Didn't get any match");
423 StringRef FullMatch = MatchInfo[0];
425 // If this defines any variables, remember their values.
426 for (std::map<StringRef, unsigned>::const_iterator I = VariableDefs.begin(),
427 E = VariableDefs.end();
429 assert(I->second < MatchInfo.size() && "Internal paren error");
430 VariableTable[I->first] = MatchInfo[I->second];
433 MatchLen = FullMatch.size();
434 return FullMatch.data()-Buffer.data();
437 unsigned Pattern::ComputeMatchDistance(StringRef Buffer,
438 const StringMap<StringRef> &VariableTable) const {
439 // Just compute the number of matching characters. For regular expressions, we
440 // just compare against the regex itself and hope for the best.
442 // FIXME: One easy improvement here is have the regex lib generate a single
443 // example regular expression which matches, and use that as the example
445 StringRef ExampleString(FixedStr);
446 if (ExampleString.empty())
447 ExampleString = RegExStr;
449 // Only compare up to the first line in the buffer, or the string size.
450 StringRef BufferPrefix = Buffer.substr(0, ExampleString.size());
451 BufferPrefix = BufferPrefix.split('\n').first;
452 return BufferPrefix.edit_distance(ExampleString);
455 void Pattern::PrintFailureInfo(const SourceMgr &SM, StringRef Buffer,
456 const StringMap<StringRef> &VariableTable) const{
457 // If this was a regular expression using variables, print the current
459 if (!VariableUses.empty()) {
460 for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
461 SmallString<256> Msg;
462 raw_svector_ostream OS(Msg);
463 StringRef Var = VariableUses[i].first;
466 if (EvaluateExpression(Var, Value)) {
467 OS << "with expression \"";
468 OS.write_escaped(Var) << "\" equal to \"";
469 OS.write_escaped(Value) << "\"";
471 OS << "uses incorrect expression \"";
472 OS.write_escaped(Var) << "\"";
475 StringMap<StringRef>::const_iterator it = VariableTable.find(Var);
477 // Check for undefined variable references.
478 if (it == VariableTable.end()) {
479 OS << "uses undefined variable \"";
480 OS.write_escaped(Var) << "\"";
482 OS << "with variable \"";
483 OS.write_escaped(Var) << "\" equal to \"";
484 OS.write_escaped(it->second) << "\"";
488 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
493 // Attempt to find the closest/best fuzzy match. Usually an error happens
494 // because some string in the output didn't exactly match. In these cases, we
495 // would like to show the user a best guess at what "should have" matched, to
496 // save them having to actually check the input manually.
497 size_t NumLinesForward = 0;
498 size_t Best = StringRef::npos;
499 double BestQuality = 0;
501 // Use an arbitrary 4k limit on how far we will search.
502 for (size_t i = 0, e = std::min(size_t(4096), Buffer.size()); i != e; ++i) {
503 if (Buffer[i] == '\n')
506 // Patterns have leading whitespace stripped, so skip whitespace when
507 // looking for something which looks like a pattern.
508 if (Buffer[i] == ' ' || Buffer[i] == '\t')
511 // Compute the "quality" of this match as an arbitrary combination of the
512 // match distance and the number of lines skipped to get to this match.
513 unsigned Distance = ComputeMatchDistance(Buffer.substr(i), VariableTable);
514 double Quality = Distance + (NumLinesForward / 100.);
516 if (Quality < BestQuality || Best == StringRef::npos) {
518 BestQuality = Quality;
522 // Print the "possible intended match here" line if we found something
523 // reasonable and not equal to what we showed in the "scanning from here"
525 if (Best && Best != StringRef::npos && BestQuality < 50) {
526 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + Best),
527 SourceMgr::DK_Note, "possible intended match here");
529 // FIXME: If we wanted to be really friendly we would show why the match
530 // failed, as it can be hard to spot simple one character differences.
534 size_t Pattern::FindRegexVarEnd(StringRef Str, SourceMgr &SM) {
535 // Offset keeps track of the current offset within the input Str
537 // [...] Nesting depth
538 size_t BracketDepth = 0;
540 while (!Str.empty()) {
541 if (Str.startswith("]]") && BracketDepth == 0)
543 if (Str[0] == '\\') {
544 // Backslash escapes the next char within regexes, so skip them both.
555 if (BracketDepth == 0) {
556 SM.PrintMessage(SMLoc::getFromPointer(Str.data()),
558 "missing closing \"]\" for regex variable");
569 return StringRef::npos;
573 //===----------------------------------------------------------------------===//
575 //===----------------------------------------------------------------------===//
577 /// CheckString - This is a check that we found in the input file.
579 /// Pat - The pattern to match.
582 /// Prefix - Which prefix name this check matched.
585 /// Loc - The location in the match file that the check string was specified.
588 /// CheckTy - Specify what kind of check this is. e.g. CHECK-NEXT: directive,
589 /// as opposed to a CHECK: directive.
590 Check::CheckType CheckTy;
592 /// DagNotStrings - These are all of the strings that are disallowed from
593 /// occurring between this match string and the previous one (or start of
595 std::vector<Pattern> DagNotStrings;
598 CheckString(const Pattern &P,
602 : Pat(P), Prefix(S), Loc(L), CheckTy(Ty) {}
604 /// Check - Match check string and its "not strings" and/or "dag strings".
605 size_t Check(const SourceMgr &SM, StringRef Buffer, bool IsLabelScanMode,
606 size_t &MatchLen, StringMap<StringRef> &VariableTable) const;
608 /// CheckNext - Verify there is a single line in the given buffer.
609 bool CheckNext(const SourceMgr &SM, StringRef Buffer) const;
611 /// CheckNot - Verify there's no "not strings" in the given buffer.
612 bool CheckNot(const SourceMgr &SM, StringRef Buffer,
613 const std::vector<const Pattern *> &NotStrings,
614 StringMap<StringRef> &VariableTable) const;
616 /// CheckDag - Match "dag strings" and their mixed "not strings".
617 size_t CheckDag(const SourceMgr &SM, StringRef Buffer,
618 std::vector<const Pattern *> &NotStrings,
619 StringMap<StringRef> &VariableTable) const;
622 /// Canonicalize whitespaces in the input file. Line endings are replaced
623 /// with UNIX-style '\n'.
625 /// \param PreserveHorizontal Don't squash consecutive horizontal whitespace
626 /// characters to a single space.
627 static MemoryBuffer *CanonicalizeInputFile(MemoryBuffer *MB,
628 bool PreserveHorizontal) {
629 SmallString<128> NewFile;
630 NewFile.reserve(MB->getBufferSize());
632 for (const char *Ptr = MB->getBufferStart(), *End = MB->getBufferEnd();
634 // Eliminate trailing dosish \r.
635 if (Ptr <= End - 2 && Ptr[0] == '\r' && Ptr[1] == '\n') {
639 // If current char is not a horizontal whitespace or if horizontal
640 // whitespace canonicalization is disabled, dump it to output as is.
641 if (PreserveHorizontal || (*Ptr != ' ' && *Ptr != '\t')) {
642 NewFile.push_back(*Ptr);
646 // Otherwise, add one space and advance over neighboring space.
647 NewFile.push_back(' ');
648 while (Ptr+1 != End &&
649 (Ptr[1] == ' ' || Ptr[1] == '\t'))
653 // Free the old buffer and return a new one.
655 MemoryBuffer::getMemBufferCopy(NewFile.str(), MB->getBufferIdentifier());
661 static bool IsPartOfWord(char c) {
662 return (isalnum(c) || c == '-' || c == '_');
665 // Get the size of the prefix extension.
666 static size_t CheckTypeSize(Check::CheckType Ty) {
668 case Check::CheckNone:
671 case Check::CheckPlain:
672 return sizeof(":") - 1;
674 case Check::CheckNext:
675 return sizeof("-NEXT:") - 1;
677 case Check::CheckNot:
678 return sizeof("-NOT:") - 1;
680 case Check::CheckDAG:
681 return sizeof("-DAG:") - 1;
683 case Check::CheckLabel:
684 return sizeof("-LABEL:") - 1;
686 case Check::CheckEOF:
687 llvm_unreachable("Should not be using EOF size");
690 llvm_unreachable("Bad check type");
693 static Check::CheckType FindCheckType(StringRef Buffer, StringRef Prefix) {
694 char NextChar = Buffer[Prefix.size()];
696 // Verify that the : is present after the prefix.
698 return Check::CheckPlain;
701 return Check::CheckNone;
703 StringRef Rest = Buffer.drop_front(Prefix.size() + 1);
704 if (Rest.startswith("NEXT:"))
705 return Check::CheckNext;
707 if (Rest.startswith("NOT:"))
708 return Check::CheckNot;
710 if (Rest.startswith("DAG:"))
711 return Check::CheckDAG;
713 if (Rest.startswith("LABEL:"))
714 return Check::CheckLabel;
716 return Check::CheckNone;
719 // From the given position, find the next character after the word.
720 static size_t SkipWord(StringRef Str, size_t Loc) {
721 while (Loc < Str.size() && IsPartOfWord(Str[Loc]))
726 // Try to find the first match in buffer for any prefix. If a valid match is
727 // found, return that prefix and set its type and location. If there are almost
728 // matches (e.g. the actual prefix string is found, but is not an actual check
729 // string), but no valid match, return an empty string and set the position to
730 // resume searching from. If no partial matches are found, return an empty
731 // string and the location will be StringRef::npos. If one prefix is a substring
732 // of another, the maximal match should be found. e.g. if "A" and "AA" are
733 // prefixes then AA-CHECK: should match the second one.
734 static StringRef FindFirstCandidateMatch(StringRef &Buffer,
735 Check::CheckType &CheckTy,
737 StringRef FirstPrefix;
738 size_t FirstLoc = StringRef::npos;
739 size_t SearchLoc = StringRef::npos;
740 Check::CheckType FirstTy = Check::CheckNone;
742 CheckTy = Check::CheckNone;
743 CheckLoc = StringRef::npos;
745 for (prefix_iterator I = CheckPrefixes.begin(), E = CheckPrefixes.end();
747 StringRef Prefix(*I);
748 size_t PrefixLoc = Buffer.find(Prefix);
750 if (PrefixLoc == StringRef::npos)
753 // Track where we are searching for invalid prefixes that look almost right.
754 // We need to only advance to the first partial match on the next attempt
755 // since a partial match could be a substring of a later, valid prefix.
756 // Need to skip to the end of the word, otherwise we could end up
757 // matching a prefix in a substring later.
758 if (PrefixLoc < SearchLoc)
759 SearchLoc = SkipWord(Buffer, PrefixLoc);
761 // We only want to find the first match to avoid skipping some.
762 if (PrefixLoc > FirstLoc)
764 // If one matching check-prefix is a prefix of another, choose the
766 if (PrefixLoc == FirstLoc && Prefix.size() < FirstPrefix.size())
769 StringRef Rest = Buffer.drop_front(PrefixLoc);
770 // Make sure we have actually found the prefix, and not a word containing
771 // it. This should also prevent matching the wrong prefix when one is a
772 // substring of another.
773 if (PrefixLoc != 0 && IsPartOfWord(Buffer[PrefixLoc - 1]))
774 FirstTy = Check::CheckNone;
776 FirstTy = FindCheckType(Rest, Prefix);
778 FirstLoc = PrefixLoc;
779 FirstPrefix = Prefix;
782 // If the first prefix is invalid, we should continue the search after it.
783 if (FirstTy == Check::CheckNone) {
784 CheckLoc = SearchLoc;
793 static StringRef FindFirstMatchingPrefix(StringRef &Buffer,
794 unsigned &LineNumber,
795 Check::CheckType &CheckTy,
797 while (!Buffer.empty()) {
798 StringRef Prefix = FindFirstCandidateMatch(Buffer, CheckTy, CheckLoc);
799 // If we found a real match, we are done.
800 if (!Prefix.empty()) {
801 LineNumber += Buffer.substr(0, CheckLoc).count('\n');
805 // We didn't find any almost matches either, we are also done.
806 if (CheckLoc == StringRef::npos)
809 LineNumber += Buffer.substr(0, CheckLoc + 1).count('\n');
811 // Advance to the last possible match we found and try again.
812 Buffer = Buffer.drop_front(CheckLoc + 1);
818 /// ReadCheckFile - Read the check file, which specifies the sequence of
819 /// expected strings. The strings are added to the CheckStrings vector.
820 /// Returns true in case of an error, false otherwise.
821 static bool ReadCheckFile(SourceMgr &SM,
822 std::vector<CheckString> &CheckStrings) {
823 std::unique_ptr<MemoryBuffer> File;
825 MemoryBuffer::getFileOrSTDIN(CheckFilename, File)) {
826 errs() << "Could not open check file '" << CheckFilename << "': "
827 << ec.message() << '\n';
831 // If we want to canonicalize whitespace, strip excess whitespace from the
832 // buffer containing the CHECK lines. Remove DOS style line endings.
834 CanonicalizeInputFile(File.release(), NoCanonicalizeWhiteSpace);
836 SM.AddNewSourceBuffer(F, SMLoc());
838 // Find all instances of CheckPrefix followed by : in the file.
839 StringRef Buffer = F->getBuffer();
840 std::vector<Pattern> DagNotMatches;
842 // LineNumber keeps track of the line on which CheckPrefix instances are
844 unsigned LineNumber = 1;
847 Check::CheckType CheckTy;
850 // See if a prefix occurs in the memory buffer.
851 StringRef UsedPrefix = FindFirstMatchingPrefix(Buffer,
855 if (UsedPrefix.empty())
858 Buffer = Buffer.drop_front(PrefixLoc);
860 // Location to use for error messages.
861 const char *UsedPrefixStart = Buffer.data() + (PrefixLoc == 0 ? 0 : 1);
863 // PrefixLoc is to the start of the prefix. Skip to the end.
864 Buffer = Buffer.drop_front(UsedPrefix.size() + CheckTypeSize(CheckTy));
866 // Okay, we found the prefix, yay. Remember the rest of the line, but ignore
867 // leading and trailing whitespace.
868 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t"));
870 // Scan ahead to the end of line.
871 size_t EOL = Buffer.find_first_of("\n\r");
873 // Remember the location of the start of the pattern, for diagnostics.
874 SMLoc PatternLoc = SMLoc::getFromPointer(Buffer.data());
876 // Parse the pattern.
878 if (P.ParsePattern(Buffer.substr(0, EOL), UsedPrefix, SM, LineNumber))
881 // Verify that CHECK-LABEL lines do not define or use variables
882 if ((CheckTy == Check::CheckLabel) && P.hasVariable()) {
883 SM.PrintMessage(SMLoc::getFromPointer(UsedPrefixStart),
885 "found '" + UsedPrefix + "-LABEL:'"
886 " with variable definition or use");
890 Buffer = Buffer.substr(EOL);
892 // Verify that CHECK-NEXT lines have at least one CHECK line before them.
893 if ((CheckTy == Check::CheckNext) && CheckStrings.empty()) {
894 SM.PrintMessage(SMLoc::getFromPointer(UsedPrefixStart),
896 "found '" + UsedPrefix + "-NEXT:' without previous '"
897 + UsedPrefix + ": line");
901 // Handle CHECK-DAG/-NOT.
902 if (CheckTy == Check::CheckDAG || CheckTy == Check::CheckNot) {
903 DagNotMatches.push_back(P);
907 // Okay, add the string we captured to the output vector and move on.
908 CheckStrings.push_back(CheckString(P,
912 std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
915 // Add an EOF pattern for any trailing CHECK-DAG/-NOTs, and use the first
916 // prefix as a filler for the error message.
917 if (!DagNotMatches.empty()) {
918 CheckStrings.push_back(CheckString(Pattern(Check::CheckEOF),
920 SMLoc::getFromPointer(Buffer.data()),
922 std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
925 if (CheckStrings.empty()) {
926 errs() << "error: no check strings found with prefix"
927 << (CheckPrefixes.size() > 1 ? "es " : " ");
928 for (size_t I = 0, N = CheckPrefixes.size(); I != N; ++I) {
929 StringRef Prefix(CheckPrefixes[I]);
930 errs() << '\'' << Prefix << ":'";
942 static void PrintCheckFailed(const SourceMgr &SM, const SMLoc &Loc,
943 const Pattern &Pat, StringRef Buffer,
944 StringMap<StringRef> &VariableTable) {
945 // Otherwise, we have an error, emit an error message.
946 SM.PrintMessage(Loc, SourceMgr::DK_Error,
947 "expected string not found in input");
949 // Print the "scanning from here" line. If the current position is at the
950 // end of a line, advance to the start of the next line.
951 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t\n\r"));
953 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
954 "scanning from here");
956 // Allow the pattern to print additional information if desired.
957 Pat.PrintFailureInfo(SM, Buffer, VariableTable);
960 static void PrintCheckFailed(const SourceMgr &SM, const CheckString &CheckStr,
962 StringMap<StringRef> &VariableTable) {
963 PrintCheckFailed(SM, CheckStr.Loc, CheckStr.Pat, Buffer, VariableTable);
966 /// CountNumNewlinesBetween - Count the number of newlines in the specified
968 static unsigned CountNumNewlinesBetween(StringRef Range,
969 const char *&FirstNewLine) {
970 unsigned NumNewLines = 0;
973 Range = Range.substr(Range.find_first_of("\n\r"));
974 if (Range.empty()) return NumNewLines;
978 // Handle \n\r and \r\n as a single newline.
979 if (Range.size() > 1 &&
980 (Range[1] == '\n' || Range[1] == '\r') &&
981 (Range[0] != Range[1]))
982 Range = Range.substr(1);
983 Range = Range.substr(1);
985 if (NumNewLines == 1)
986 FirstNewLine = Range.begin();
990 size_t CheckString::Check(const SourceMgr &SM, StringRef Buffer,
991 bool IsLabelScanMode, size_t &MatchLen,
992 StringMap<StringRef> &VariableTable) const {
994 std::vector<const Pattern *> NotStrings;
996 // IsLabelScanMode is true when we are scanning forward to find CHECK-LABEL
997 // bounds; we have not processed variable definitions within the bounded block
998 // yet so cannot handle any final CHECK-DAG yet; this is handled when going
999 // over the block again (including the last CHECK-LABEL) in normal mode.
1000 if (!IsLabelScanMode) {
1001 // Match "dag strings" (with mixed "not strings" if any).
1002 LastPos = CheckDag(SM, Buffer, NotStrings, VariableTable);
1003 if (LastPos == StringRef::npos)
1004 return StringRef::npos;
1007 // Match itself from the last position after matching CHECK-DAG.
1008 StringRef MatchBuffer = Buffer.substr(LastPos);
1009 size_t MatchPos = Pat.Match(MatchBuffer, MatchLen, VariableTable);
1010 if (MatchPos == StringRef::npos) {
1011 PrintCheckFailed(SM, *this, MatchBuffer, VariableTable);
1012 return StringRef::npos;
1014 MatchPos += LastPos;
1016 // Similar to the above, in "label-scan mode" we can't yet handle CHECK-NEXT
1018 if (!IsLabelScanMode) {
1019 StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
1021 // If this check is a "CHECK-NEXT", verify that the previous match was on
1022 // the previous line (i.e. that there is one newline between them).
1023 if (CheckNext(SM, SkippedRegion))
1024 return StringRef::npos;
1026 // If this match had "not strings", verify that they don't exist in the
1028 if (CheckNot(SM, SkippedRegion, NotStrings, VariableTable))
1029 return StringRef::npos;
1035 bool CheckString::CheckNext(const SourceMgr &SM, StringRef Buffer) const {
1036 if (CheckTy != Check::CheckNext)
1039 // Count the number of newlines between the previous match and this one.
1040 assert(Buffer.data() !=
1042 SM.FindBufferContainingLoc(
1043 SMLoc::getFromPointer(Buffer.data())))->getBufferStart() &&
1044 "CHECK-NEXT can't be the first check in a file");
1046 const char *FirstNewLine = 0;
1047 unsigned NumNewLines = CountNumNewlinesBetween(Buffer, FirstNewLine);
1049 if (NumNewLines == 0) {
1050 SM.PrintMessage(Loc, SourceMgr::DK_Error, Prefix +
1051 "-NEXT: is on the same line as previous match");
1052 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()),
1053 SourceMgr::DK_Note, "'next' match was here");
1054 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
1055 "previous match ended here");
1059 if (NumNewLines != 1) {
1060 SM.PrintMessage(Loc, SourceMgr::DK_Error, Prefix +
1061 "-NEXT: is not on the line after the previous match");
1062 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()),
1063 SourceMgr::DK_Note, "'next' match was here");
1064 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
1065 "previous match ended here");
1066 SM.PrintMessage(SMLoc::getFromPointer(FirstNewLine), SourceMgr::DK_Note,
1067 "non-matching line after previous match is here");
1074 bool CheckString::CheckNot(const SourceMgr &SM, StringRef Buffer,
1075 const std::vector<const Pattern *> &NotStrings,
1076 StringMap<StringRef> &VariableTable) const {
1077 for (unsigned ChunkNo = 0, e = NotStrings.size();
1078 ChunkNo != e; ++ChunkNo) {
1079 const Pattern *Pat = NotStrings[ChunkNo];
1080 assert((Pat->getCheckTy() == Check::CheckNot) && "Expect CHECK-NOT!");
1082 size_t MatchLen = 0;
1083 size_t Pos = Pat->Match(Buffer, MatchLen, VariableTable);
1085 if (Pos == StringRef::npos) continue;
1087 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()+Pos),
1088 SourceMgr::DK_Error,
1089 Prefix + "-NOT: string occurred!");
1090 SM.PrintMessage(Pat->getLoc(), SourceMgr::DK_Note,
1091 Prefix + "-NOT: pattern specified here");
1098 size_t CheckString::CheckDag(const SourceMgr &SM, StringRef Buffer,
1099 std::vector<const Pattern *> &NotStrings,
1100 StringMap<StringRef> &VariableTable) const {
1101 if (DagNotStrings.empty())
1105 size_t StartPos = LastPos;
1107 for (unsigned ChunkNo = 0, e = DagNotStrings.size();
1108 ChunkNo != e; ++ChunkNo) {
1109 const Pattern &Pat = DagNotStrings[ChunkNo];
1111 assert((Pat.getCheckTy() == Check::CheckDAG ||
1112 Pat.getCheckTy() == Check::CheckNot) &&
1113 "Invalid CHECK-DAG or CHECK-NOT!");
1115 if (Pat.getCheckTy() == Check::CheckNot) {
1116 NotStrings.push_back(&Pat);
1120 assert((Pat.getCheckTy() == Check::CheckDAG) && "Expect CHECK-DAG!");
1122 size_t MatchLen = 0, MatchPos;
1124 // CHECK-DAG always matches from the start.
1125 StringRef MatchBuffer = Buffer.substr(StartPos);
1126 MatchPos = Pat.Match(MatchBuffer, MatchLen, VariableTable);
1127 // With a group of CHECK-DAGs, a single mismatching means the match on
1128 // that group of CHECK-DAGs fails immediately.
1129 if (MatchPos == StringRef::npos) {
1130 PrintCheckFailed(SM, Pat.getLoc(), Pat, MatchBuffer, VariableTable);
1131 return StringRef::npos;
1133 // Re-calc it as the offset relative to the start of the original string.
1134 MatchPos += StartPos;
1136 if (!NotStrings.empty()) {
1137 if (MatchPos < LastPos) {
1139 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + MatchPos),
1140 SourceMgr::DK_Error,
1141 Prefix + "-DAG: found a match of CHECK-DAG"
1142 " reordering across a CHECK-NOT");
1143 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + LastPos),
1145 Prefix + "-DAG: the farthest match of CHECK-DAG"
1147 SM.PrintMessage(NotStrings[0]->getLoc(), SourceMgr::DK_Note,
1148 Prefix + "-NOT: the crossed pattern specified"
1150 SM.PrintMessage(Pat.getLoc(), SourceMgr::DK_Note,
1151 Prefix + "-DAG: the reordered pattern specified"
1153 return StringRef::npos;
1155 // All subsequent CHECK-DAGs should be matched from the farthest
1156 // position of all precedent CHECK-DAGs (including this one.)
1158 // If there's CHECK-NOTs between two CHECK-DAGs or from CHECK to
1159 // CHECK-DAG, verify that there's no 'not' strings occurred in that
1161 StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
1162 if (CheckNot(SM, SkippedRegion, NotStrings, VariableTable))
1163 return StringRef::npos;
1164 // Clear "not strings".
1168 // Update the last position with CHECK-DAG matches.
1169 LastPos = std::max(MatchPos + MatchLen, LastPos);
1175 // A check prefix must contain only alphanumeric, hyphens and underscores.
1176 static bool ValidateCheckPrefix(StringRef CheckPrefix) {
1177 Regex Validator("^[a-zA-Z0-9_-]*$");
1178 return Validator.match(CheckPrefix);
1181 static bool ValidateCheckPrefixes() {
1182 StringSet<> PrefixSet;
1184 for (prefix_iterator I = CheckPrefixes.begin(), E = CheckPrefixes.end();
1186 StringRef Prefix(*I);
1188 if (!PrefixSet.insert(Prefix))
1191 if (!ValidateCheckPrefix(Prefix))
1198 // I don't think there's a way to specify an initial value for cl::list,
1199 // so if nothing was specified, add the default
1200 static void AddCheckPrefixIfNeeded() {
1201 if (CheckPrefixes.empty())
1202 CheckPrefixes.push_back("CHECK");
1205 int main(int argc, char **argv) {
1206 sys::PrintStackTraceOnErrorSignal();
1207 PrettyStackTraceProgram X(argc, argv);
1208 cl::ParseCommandLineOptions(argc, argv);
1210 if (!ValidateCheckPrefixes()) {
1211 errs() << "Supplied check-prefix is invalid! Prefixes must be unique and "
1212 "start with a letter and contain only alphanumeric characters, "
1213 "hyphens and underscores\n";
1217 AddCheckPrefixIfNeeded();
1221 // Read the expected strings from the check file.
1222 std::vector<CheckString> CheckStrings;
1223 if (ReadCheckFile(SM, CheckStrings))
1226 // Open the file to check and add it to SourceMgr.
1227 std::unique_ptr<MemoryBuffer> File;
1229 MemoryBuffer::getFileOrSTDIN(InputFilename, File)) {
1230 errs() << "Could not open input file '" << InputFilename << "': "
1231 << ec.message() << '\n';
1235 if (File->getBufferSize() == 0) {
1236 errs() << "FileCheck error: '" << InputFilename << "' is empty.\n";
1240 // Remove duplicate spaces in the input file if requested.
1241 // Remove DOS style line endings.
1243 CanonicalizeInputFile(File.release(), NoCanonicalizeWhiteSpace);
1245 SM.AddNewSourceBuffer(F, SMLoc());
1247 /// VariableTable - This holds all the current filecheck variables.
1248 StringMap<StringRef> VariableTable;
1250 // Check that we have all of the expected strings, in order, in the input
1252 StringRef Buffer = F->getBuffer();
1254 bool hasError = false;
1256 unsigned i = 0, j = 0, e = CheckStrings.size();
1259 StringRef CheckRegion;
1261 CheckRegion = Buffer;
1263 const CheckString &CheckLabelStr = CheckStrings[j];
1264 if (CheckLabelStr.CheckTy != Check::CheckLabel) {
1269 // Scan to next CHECK-LABEL match, ignoring CHECK-NOT and CHECK-DAG
1270 size_t MatchLabelLen = 0;
1271 size_t MatchLabelPos = CheckLabelStr.Check(SM, Buffer, true,
1272 MatchLabelLen, VariableTable);
1273 if (MatchLabelPos == StringRef::npos) {
1278 CheckRegion = Buffer.substr(0, MatchLabelPos + MatchLabelLen);
1279 Buffer = Buffer.substr(MatchLabelPos + MatchLabelLen);
1283 for ( ; i != j; ++i) {
1284 const CheckString &CheckStr = CheckStrings[i];
1286 // Check each string within the scanned region, including a second check
1287 // of any final CHECK-LABEL (to verify CHECK-NOT and CHECK-DAG)
1288 size_t MatchLen = 0;
1289 size_t MatchPos = CheckStr.Check(SM, CheckRegion, false, MatchLen,
1292 if (MatchPos == StringRef::npos) {
1298 CheckRegion = CheckRegion.substr(MatchPos + MatchLen);
1305 return hasError ? 1 : 0;