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/ADT/StringSet.h"
24 #include "llvm/Support/CommandLine.h"
25 #include "llvm/Support/MemoryBuffer.h"
26 #include "llvm/Support/PrettyStackTrace.h"
27 #include "llvm/Support/Regex.h"
28 #include "llvm/Support/Signals.h"
29 #include "llvm/Support/SourceMgr.h"
30 #include "llvm/Support/raw_ostream.h"
31 #include "llvm/Support/system_error.h"
39 static cl::opt<std::string>
40 CheckFilename(cl::Positional, cl::desc("<check-file>"), cl::Required);
42 static cl::opt<std::string>
43 InputFilename("input-file", cl::desc("File to check (defaults to stdin)"),
44 cl::init("-"), cl::value_desc("filename"));
46 static cl::list<std::string>
47 CheckPrefixes("check-prefix",
48 cl::desc("Prefix to use from check file (defaults to 'CHECK')"));
51 NoCanonicalizeWhiteSpace("strict-whitespace",
52 cl::desc("Do not treat all horizontal whitespace as equivalent"));
54 typedef cl::list<std::string>::const_iterator prefix_iterator;
56 //===----------------------------------------------------------------------===//
57 // Pattern Handling Code.
58 //===----------------------------------------------------------------------===//
69 /// MatchEOF - When set, this pattern only matches the end of file. This is
70 /// used for trailing CHECK-NOTs.
78 Check::CheckType CheckTy;
80 /// FixedStr - If non-empty, this pattern is a fixed string match with the
81 /// specified fixed string.
84 /// RegEx - If non-empty, this is a regex pattern.
87 /// \brief Contains the number of line this pattern is in.
90 /// VariableUses - Entries in this vector map to uses of a variable in the
91 /// pattern, e.g. "foo[[bar]]baz". In this case, the RegExStr will contain
92 /// "foobaz" and we'll get an entry in this vector that tells us to insert the
93 /// value of bar at offset 3.
94 std::vector<std::pair<StringRef, unsigned> > VariableUses;
96 /// VariableDefs - Maps definitions of variables to their parenthesized
98 /// E.g. for the pattern "foo[[bar:.*]]baz", VariableDefs will map "bar" to 1.
99 std::map<StringRef, unsigned> VariableDefs;
103 Pattern(Check::CheckType Ty)
106 /// getLoc - Return the location in source code.
107 SMLoc getLoc() const { return PatternLoc; }
109 /// ParsePattern - Parse the given string into the Pattern. Prefix provides
110 /// which prefix is being matched, SM provides the SourceMgr used for error
111 /// reports, and LineNumber is the line number in the input file from which
112 /// the pattern string was read. Returns true in case of an error, false
114 bool ParsePattern(StringRef PatternStr,
117 unsigned LineNumber);
119 /// Match - Match the pattern string against the input buffer Buffer. This
120 /// returns the position that is matched or npos if there is no match. If
121 /// there is a match, the size of the matched string is returned in MatchLen.
123 /// The VariableTable StringMap provides the current values of filecheck
124 /// variables and is updated if this match defines new values.
125 size_t Match(StringRef Buffer, size_t &MatchLen,
126 StringMap<StringRef> &VariableTable) const;
128 /// PrintFailureInfo - Print additional information about a failure to match
129 /// involving this pattern.
130 void PrintFailureInfo(const SourceMgr &SM, StringRef Buffer,
131 const StringMap<StringRef> &VariableTable) const;
133 bool hasVariable() const { return !(VariableUses.empty() &&
134 VariableDefs.empty()); }
136 Check::CheckType getCheckTy() const { return CheckTy; }
139 bool AddRegExToRegEx(StringRef RS, unsigned &CurParen, SourceMgr &SM);
140 void AddBackrefToRegEx(unsigned BackrefNum);
142 /// ComputeMatchDistance - Compute an arbitrary estimate for the quality of
143 /// matching this pattern at the start of \arg Buffer; a distance of zero
144 /// should correspond to a perfect match.
145 unsigned ComputeMatchDistance(StringRef Buffer,
146 const StringMap<StringRef> &VariableTable) const;
148 /// \brief Evaluates expression and stores the result to \p Value.
149 /// \return true on success. false when the expression has invalid syntax.
150 bool EvaluateExpression(StringRef Expr, std::string &Value) const;
152 /// \brief Finds the closing sequence of a regex variable usage or
153 /// definition. Str has to point in the beginning of the definition
154 /// (right after the opening sequence).
155 /// \return offset of the closing sequence within Str, or npos if it was not
157 size_t FindRegexVarEnd(StringRef Str, SourceMgr &SM);
161 bool Pattern::ParsePattern(StringRef PatternStr,
164 unsigned LineNumber) {
165 this->LineNumber = LineNumber;
166 PatternLoc = SMLoc::getFromPointer(PatternStr.data());
168 // Ignore trailing whitespace.
169 while (!PatternStr.empty() &&
170 (PatternStr.back() == ' ' || PatternStr.back() == '\t'))
171 PatternStr = PatternStr.substr(0, PatternStr.size()-1);
173 // Check that there is something on the line.
174 if (PatternStr.empty()) {
175 SM.PrintMessage(PatternLoc, SourceMgr::DK_Error,
176 "found empty check string with prefix '" +
181 // Check to see if this is a fixed string, or if it has regex pieces.
182 if (PatternStr.size() < 2 ||
183 (PatternStr.find("{{") == StringRef::npos &&
184 PatternStr.find("[[") == StringRef::npos)) {
185 FixedStr = PatternStr;
189 // Paren value #0 is for the fully matched string. Any new parenthesized
190 // values add from there.
191 unsigned CurParen = 1;
193 // Otherwise, there is at least one regex piece. Build up the regex pattern
194 // by escaping scary characters in fixed strings, building up one big regex.
195 while (!PatternStr.empty()) {
197 if (PatternStr.startswith("{{")) {
198 // This is the start of a regex match. Scan for the }}.
199 size_t End = PatternStr.find("}}");
200 if (End == StringRef::npos) {
201 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
203 "found start of regex string with no end '}}'");
207 // Enclose {{}} patterns in parens just like [[]] even though we're not
208 // capturing the result for any purpose. This is required in case the
209 // expression contains an alternation like: CHECK: abc{{x|z}}def. We
210 // want this to turn into: "abc(x|z)def" not "abcx|zdef".
214 if (AddRegExToRegEx(PatternStr.substr(2, End-2), CurParen, SM))
218 PatternStr = PatternStr.substr(End+2);
222 // Named RegEx matches. These are of two forms: [[foo:.*]] which matches .*
223 // (or some other regex) and assigns it to the FileCheck variable 'foo'. The
224 // second form is [[foo]] which is a reference to foo. The variable name
225 // itself must be of the form "[a-zA-Z_][0-9a-zA-Z_]*", otherwise we reject
226 // it. This is to catch some common errors.
227 if (PatternStr.startswith("[[")) {
228 // Find the closing bracket pair ending the match. End is going to be an
229 // offset relative to the beginning of the match string.
230 size_t End = FindRegexVarEnd(PatternStr.substr(2), SM);
232 if (End == StringRef::npos) {
233 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
235 "invalid named regex reference, no ]] found");
239 StringRef MatchStr = PatternStr.substr(2, End);
240 PatternStr = PatternStr.substr(End+4);
242 // Get the regex name (e.g. "foo").
243 size_t NameEnd = MatchStr.find(':');
244 StringRef Name = MatchStr.substr(0, NameEnd);
247 SM.PrintMessage(SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
248 "invalid name in named regex: empty name");
252 // Verify that the name/expression is well formed. FileCheck currently
253 // supports @LINE, @LINE+number, @LINE-number expressions. The check here
254 // is relaxed, more strict check is performed in \c EvaluateExpression.
255 bool IsExpression = false;
256 for (unsigned i = 0, e = Name.size(); i != e; ++i) {
257 if (i == 0 && Name[i] == '@') {
258 if (NameEnd != StringRef::npos) {
259 SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
261 "invalid name in named regex definition");
267 if (Name[i] != '_' && !isalnum(Name[i]) &&
268 (!IsExpression || (Name[i] != '+' && Name[i] != '-'))) {
269 SM.PrintMessage(SMLoc::getFromPointer(Name.data()+i),
270 SourceMgr::DK_Error, "invalid name in named regex");
275 // Name can't start with a digit.
276 if (isdigit(static_cast<unsigned char>(Name[0]))) {
277 SM.PrintMessage(SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
278 "invalid name in named regex");
283 if (NameEnd == StringRef::npos) {
284 // Handle variables that were defined earlier on the same line by
285 // emitting a backreference.
286 if (VariableDefs.find(Name) != VariableDefs.end()) {
287 unsigned VarParenNum = VariableDefs[Name];
288 if (VarParenNum < 1 || VarParenNum > 9) {
289 SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
291 "Can't back-reference more than 9 variables");
294 AddBackrefToRegEx(VarParenNum);
296 VariableUses.push_back(std::make_pair(Name, RegExStr.size()));
301 // Handle [[foo:.*]].
302 VariableDefs[Name] = CurParen;
306 if (AddRegExToRegEx(MatchStr.substr(NameEnd+1), CurParen, SM))
312 // Handle fixed string matches.
313 // Find the end, which is the start of the next regex.
314 size_t FixedMatchEnd = PatternStr.find("{{");
315 FixedMatchEnd = std::min(FixedMatchEnd, PatternStr.find("[["));
316 RegExStr += Regex::escape(PatternStr.substr(0, FixedMatchEnd));
317 PatternStr = PatternStr.substr(FixedMatchEnd);
323 bool Pattern::AddRegExToRegEx(StringRef RS, unsigned &CurParen,
327 if (!R.isValid(Error)) {
328 SM.PrintMessage(SMLoc::getFromPointer(RS.data()), SourceMgr::DK_Error,
329 "invalid regex: " + Error);
333 RegExStr += RS.str();
334 CurParen += R.getNumMatches();
338 void Pattern::AddBackrefToRegEx(unsigned BackrefNum) {
339 assert(BackrefNum >= 1 && BackrefNum <= 9 && "Invalid backref number");
340 std::string Backref = std::string("\\") +
341 std::string(1, '0' + BackrefNum);
345 bool Pattern::EvaluateExpression(StringRef Expr, std::string &Value) const {
346 // The only supported expression is @LINE([\+-]\d+)?
347 if (!Expr.startswith("@LINE"))
349 Expr = Expr.substr(StringRef("@LINE").size());
353 Expr = Expr.substr(1);
354 else if (Expr[0] != '-')
356 if (Expr.getAsInteger(10, Offset))
359 Value = llvm::itostr(LineNumber + Offset);
363 /// Match - Match the pattern string against the input buffer Buffer. This
364 /// returns the position that is matched or npos if there is no match. If
365 /// there is a match, the size of the matched string is returned in MatchLen.
366 size_t Pattern::Match(StringRef Buffer, size_t &MatchLen,
367 StringMap<StringRef> &VariableTable) const {
368 // If this is the EOF pattern, match it immediately.
369 if (CheckTy == Check::CheckEOF) {
371 return Buffer.size();
374 // If this is a fixed string pattern, just match it now.
375 if (!FixedStr.empty()) {
376 MatchLen = FixedStr.size();
377 return Buffer.find(FixedStr);
382 // If there are variable uses, we need to create a temporary string with the
384 StringRef RegExToMatch = RegExStr;
386 if (!VariableUses.empty()) {
389 unsigned InsertOffset = 0;
390 for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
393 if (VariableUses[i].first[0] == '@') {
394 if (!EvaluateExpression(VariableUses[i].first, Value))
395 return StringRef::npos;
397 StringMap<StringRef>::iterator it =
398 VariableTable.find(VariableUses[i].first);
399 // If the variable is undefined, return an error.
400 if (it == VariableTable.end())
401 return StringRef::npos;
403 // Look up the value and escape it so that we can put it into the regex.
404 Value += Regex::escape(it->second);
407 // Plop it into the regex at the adjusted offset.
408 TmpStr.insert(TmpStr.begin()+VariableUses[i].second+InsertOffset,
409 Value.begin(), Value.end());
410 InsertOffset += Value.size();
413 // Match the newly constructed regex.
414 RegExToMatch = TmpStr;
418 SmallVector<StringRef, 4> MatchInfo;
419 if (!Regex(RegExToMatch, Regex::Newline).match(Buffer, &MatchInfo))
420 return StringRef::npos;
422 // Successful regex match.
423 assert(!MatchInfo.empty() && "Didn't get any match");
424 StringRef FullMatch = MatchInfo[0];
426 // If this defines any variables, remember their values.
427 for (std::map<StringRef, unsigned>::const_iterator I = VariableDefs.begin(),
428 E = VariableDefs.end();
430 assert(I->second < MatchInfo.size() && "Internal paren error");
431 VariableTable[I->first] = MatchInfo[I->second];
434 MatchLen = FullMatch.size();
435 return FullMatch.data()-Buffer.data();
438 unsigned Pattern::ComputeMatchDistance(StringRef Buffer,
439 const StringMap<StringRef> &VariableTable) const {
440 // Just compute the number of matching characters. For regular expressions, we
441 // just compare against the regex itself and hope for the best.
443 // FIXME: One easy improvement here is have the regex lib generate a single
444 // example regular expression which matches, and use that as the example
446 StringRef ExampleString(FixedStr);
447 if (ExampleString.empty())
448 ExampleString = RegExStr;
450 // Only compare up to the first line in the buffer, or the string size.
451 StringRef BufferPrefix = Buffer.substr(0, ExampleString.size());
452 BufferPrefix = BufferPrefix.split('\n').first;
453 return BufferPrefix.edit_distance(ExampleString);
456 void Pattern::PrintFailureInfo(const SourceMgr &SM, StringRef Buffer,
457 const StringMap<StringRef> &VariableTable) const{
458 // If this was a regular expression using variables, print the current
460 if (!VariableUses.empty()) {
461 for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
462 SmallString<256> Msg;
463 raw_svector_ostream OS(Msg);
464 StringRef Var = VariableUses[i].first;
467 if (EvaluateExpression(Var, Value)) {
468 OS << "with expression \"";
469 OS.write_escaped(Var) << "\" equal to \"";
470 OS.write_escaped(Value) << "\"";
472 OS << "uses incorrect expression \"";
473 OS.write_escaped(Var) << "\"";
476 StringMap<StringRef>::const_iterator it = VariableTable.find(Var);
478 // Check for undefined variable references.
479 if (it == VariableTable.end()) {
480 OS << "uses undefined variable \"";
481 OS.write_escaped(Var) << "\"";
483 OS << "with variable \"";
484 OS.write_escaped(Var) << "\" equal to \"";
485 OS.write_escaped(it->second) << "\"";
489 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
494 // Attempt to find the closest/best fuzzy match. Usually an error happens
495 // because some string in the output didn't exactly match. In these cases, we
496 // would like to show the user a best guess at what "should have" matched, to
497 // save them having to actually check the input manually.
498 size_t NumLinesForward = 0;
499 size_t Best = StringRef::npos;
500 double BestQuality = 0;
502 // Use an arbitrary 4k limit on how far we will search.
503 for (size_t i = 0, e = std::min(size_t(4096), Buffer.size()); i != e; ++i) {
504 if (Buffer[i] == '\n')
507 // Patterns have leading whitespace stripped, so skip whitespace when
508 // looking for something which looks like a pattern.
509 if (Buffer[i] == ' ' || Buffer[i] == '\t')
512 // Compute the "quality" of this match as an arbitrary combination of the
513 // match distance and the number of lines skipped to get to this match.
514 unsigned Distance = ComputeMatchDistance(Buffer.substr(i), VariableTable);
515 double Quality = Distance + (NumLinesForward / 100.);
517 if (Quality < BestQuality || Best == StringRef::npos) {
519 BestQuality = Quality;
523 // Print the "possible intended match here" line if we found something
524 // reasonable and not equal to what we showed in the "scanning from here"
526 if (Best && Best != StringRef::npos && BestQuality < 50) {
527 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + Best),
528 SourceMgr::DK_Note, "possible intended match here");
530 // FIXME: If we wanted to be really friendly we would show why the match
531 // failed, as it can be hard to spot simple one character differences.
535 size_t Pattern::FindRegexVarEnd(StringRef Str, SourceMgr &SM) {
536 // Offset keeps track of the current offset within the input Str
538 // [...] Nesting depth
539 size_t BracketDepth = 0;
541 while (!Str.empty()) {
542 if (Str.startswith("]]") && BracketDepth == 0)
544 if (Str[0] == '\\') {
545 // Backslash escapes the next char within regexes, so skip them both.
556 if (BracketDepth == 0) {
557 SM.PrintMessage(SMLoc::getFromPointer(Str.data()),
559 "missing closing \"]\" for regex variable");
570 return StringRef::npos;
574 //===----------------------------------------------------------------------===//
576 //===----------------------------------------------------------------------===//
578 /// CheckString - This is a check that we found in the input file.
580 /// Pat - The pattern to match.
583 /// Prefix - Which prefix name this check matched.
586 /// Loc - The location in the match file that the check string was specified.
589 /// CheckTy - Specify what kind of check this is. e.g. CHECK-NEXT: directive,
590 /// as opposed to a CHECK: directive.
591 Check::CheckType CheckTy;
593 /// DagNotStrings - These are all of the strings that are disallowed from
594 /// occurring between this match string and the previous one (or start of
596 std::vector<Pattern> DagNotStrings;
599 CheckString(const Pattern &P,
603 : Pat(P), Prefix(S), Loc(L), CheckTy(Ty) {}
605 /// Check - Match check string and its "not strings" and/or "dag strings".
606 size_t Check(const SourceMgr &SM, StringRef Buffer, bool IsLabelScanMode,
607 size_t &MatchLen, StringMap<StringRef> &VariableTable) const;
609 /// CheckNext - Verify there is a single line in the given buffer.
610 bool CheckNext(const SourceMgr &SM, StringRef Buffer) const;
612 /// CheckNot - Verify there's no "not strings" in the given buffer.
613 bool CheckNot(const SourceMgr &SM, StringRef Buffer,
614 const std::vector<const Pattern *> &NotStrings,
615 StringMap<StringRef> &VariableTable) const;
617 /// CheckDag - Match "dag strings" and their mixed "not strings".
618 size_t CheckDag(const SourceMgr &SM, StringRef Buffer,
619 std::vector<const Pattern *> &NotStrings,
620 StringMap<StringRef> &VariableTable) const;
623 /// Canonicalize whitespaces in the input file. Line endings are replaced
624 /// with UNIX-style '\n'.
626 /// \param PreserveHorizontal Don't squash consecutive horizontal whitespace
627 /// characters to a single space.
628 static MemoryBuffer *CanonicalizeInputFile(MemoryBuffer *MB,
629 bool PreserveHorizontal) {
630 SmallString<128> NewFile;
631 NewFile.reserve(MB->getBufferSize());
633 for (const char *Ptr = MB->getBufferStart(), *End = MB->getBufferEnd();
635 // Eliminate trailing dosish \r.
636 if (Ptr <= End - 2 && Ptr[0] == '\r' && Ptr[1] == '\n') {
640 // If current char is not a horizontal whitespace or if horizontal
641 // whitespace canonicalization is disabled, dump it to output as is.
642 if (PreserveHorizontal || (*Ptr != ' ' && *Ptr != '\t')) {
643 NewFile.push_back(*Ptr);
647 // Otherwise, add one space and advance over neighboring space.
648 NewFile.push_back(' ');
649 while (Ptr+1 != End &&
650 (Ptr[1] == ' ' || Ptr[1] == '\t'))
654 // Free the old buffer and return a new one.
656 MemoryBuffer::getMemBufferCopy(NewFile.str(), MB->getBufferIdentifier());
662 static bool IsPartOfWord(char c) {
663 return (isalnum(c) || c == '-' || c == '_');
666 // Get the size of the prefix extension.
667 static size_t CheckTypeSize(Check::CheckType Ty) {
669 case Check::CheckNone:
672 case Check::CheckPlain:
673 return sizeof(":") - 1;
675 case Check::CheckNext:
676 return sizeof("-NEXT:") - 1;
678 case Check::CheckNot:
679 return sizeof("-NOT:") - 1;
681 case Check::CheckDAG:
682 return sizeof("-DAG:") - 1;
684 case Check::CheckLabel:
685 return sizeof("-LABEL:") - 1;
687 case Check::CheckEOF:
688 llvm_unreachable("Should not be using EOF size");
691 llvm_unreachable("Bad check type");
694 static Check::CheckType FindCheckType(StringRef Buffer, StringRef Prefix) {
695 char NextChar = Buffer[Prefix.size()];
697 // Verify that the : is present after the prefix.
699 return Check::CheckPlain;
702 return Check::CheckNone;
704 StringRef Rest = Buffer.drop_front(Prefix.size() + 1);
705 if (Rest.startswith("NEXT:"))
706 return Check::CheckNext;
708 if (Rest.startswith("NOT:"))
709 return Check::CheckNot;
711 if (Rest.startswith("DAG:"))
712 return Check::CheckDAG;
714 if (Rest.startswith("LABEL:"))
715 return Check::CheckLabel;
717 return Check::CheckNone;
720 // From the given position, find the next character after the word.
721 static size_t SkipWord(StringRef Str, size_t Loc) {
722 while (Loc < Str.size() && IsPartOfWord(Str[Loc]))
727 // Try to find the first match in buffer for any prefix. If a valid match is
728 // found, return that prefix and set its type and location. If there are almost
729 // matches (e.g. the actual prefix string is found, but is not an actual check
730 // string), but no valid match, return an empty string and set the position to
731 // resume searching from. If no partial matches are found, return an empty
732 // string and the location will be StringRef::npos. If one prefix is a substring
733 // of another, the maximal match should be found. e.g. if "A" and "AA" are
734 // prefixes then AA-CHECK: should match the second one.
735 static StringRef FindFirstCandidateMatch(StringRef &Buffer,
736 Check::CheckType &CheckTy,
738 StringRef FirstPrefix;
739 size_t FirstLoc = StringRef::npos;
740 size_t SearchLoc = StringRef::npos;
741 Check::CheckType FirstTy = Check::CheckNone;
743 CheckTy = Check::CheckNone;
744 CheckLoc = StringRef::npos;
746 for (prefix_iterator I = CheckPrefixes.begin(), E = CheckPrefixes.end();
748 StringRef Prefix(*I);
749 size_t PrefixLoc = Buffer.find(Prefix);
751 if (PrefixLoc == StringRef::npos)
754 // Track where we are searching for invalid prefixes that look almost right.
755 // We need to only advance to the first partial match on the next attempt
756 // since a partial match could be a substring of a later, valid prefix.
757 // Need to skip to the end of the word, otherwise we could end up
758 // matching a prefix in a substring later.
759 if (PrefixLoc < SearchLoc)
760 SearchLoc = SkipWord(Buffer, PrefixLoc);
762 // We only want to find the first match to avoid skipping some.
763 if (PrefixLoc > FirstLoc)
765 // If one matching check-prefix is a prefix of another, choose the
767 if (PrefixLoc == FirstLoc && Prefix.size() < FirstPrefix.size())
770 StringRef Rest = Buffer.drop_front(PrefixLoc);
771 // Make sure we have actually found the prefix, and not a word containing
772 // it. This should also prevent matching the wrong prefix when one is a
773 // substring of another.
774 if (PrefixLoc != 0 && IsPartOfWord(Buffer[PrefixLoc - 1]))
775 FirstTy = Check::CheckNone;
777 FirstTy = FindCheckType(Rest, Prefix);
779 FirstLoc = PrefixLoc;
780 FirstPrefix = Prefix;
783 // If the first prefix is invalid, we should continue the search after it.
784 if (FirstTy == Check::CheckNone) {
785 CheckLoc = SearchLoc;
794 static StringRef FindFirstMatchingPrefix(StringRef &Buffer,
795 unsigned &LineNumber,
796 Check::CheckType &CheckTy,
798 while (!Buffer.empty()) {
799 StringRef Prefix = FindFirstCandidateMatch(Buffer, CheckTy, CheckLoc);
800 // If we found a real match, we are done.
801 if (!Prefix.empty()) {
802 LineNumber += Buffer.substr(0, CheckLoc).count('\n');
806 // We didn't find any almost matches either, we are also done.
807 if (CheckLoc == StringRef::npos)
810 LineNumber += Buffer.substr(0, CheckLoc + 1).count('\n');
812 // Advance to the last possible match we found and try again.
813 Buffer = Buffer.drop_front(CheckLoc + 1);
819 /// ReadCheckFile - Read the check file, which specifies the sequence of
820 /// expected strings. The strings are added to the CheckStrings vector.
821 /// Returns true in case of an error, false otherwise.
822 static bool ReadCheckFile(SourceMgr &SM,
823 std::vector<CheckString> &CheckStrings) {
824 OwningPtr<MemoryBuffer> File;
826 MemoryBuffer::getFileOrSTDIN(CheckFilename, File)) {
827 errs() << "Could not open check file '" << CheckFilename << "': "
828 << ec.message() << '\n';
832 // If we want to canonicalize whitespace, strip excess whitespace from the
833 // buffer containing the CHECK lines. Remove DOS style line endings.
835 CanonicalizeInputFile(File.take(), NoCanonicalizeWhiteSpace);
837 SM.AddNewSourceBuffer(F, SMLoc());
839 // Find all instances of CheckPrefix followed by : in the file.
840 StringRef Buffer = F->getBuffer();
841 std::vector<Pattern> DagNotMatches;
843 // LineNumber keeps track of the line on which CheckPrefix instances are
845 unsigned LineNumber = 1;
848 Check::CheckType CheckTy;
851 // See if a prefix occurs in the memory buffer.
852 StringRef UsedPrefix = FindFirstMatchingPrefix(Buffer,
856 if (UsedPrefix.empty())
859 Buffer = Buffer.drop_front(PrefixLoc);
861 // Location to use for error messages.
862 const char *UsedPrefixStart = Buffer.data() + (PrefixLoc == 0 ? 0 : 1);
864 // PrefixLoc is to the start of the prefix. Skip to the end.
865 Buffer = Buffer.drop_front(UsedPrefix.size() + CheckTypeSize(CheckTy));
867 // Okay, we found the prefix, yay. Remember the rest of the line, but ignore
868 // leading and trailing whitespace.
869 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t"));
871 // Scan ahead to the end of line.
872 size_t EOL = Buffer.find_first_of("\n\r");
874 // Remember the location of the start of the pattern, for diagnostics.
875 SMLoc PatternLoc = SMLoc::getFromPointer(Buffer.data());
877 // Parse the pattern.
879 if (P.ParsePattern(Buffer.substr(0, EOL), UsedPrefix, SM, LineNumber))
882 // Verify that CHECK-LABEL lines do not define or use variables
883 if ((CheckTy == Check::CheckLabel) && P.hasVariable()) {
884 SM.PrintMessage(SMLoc::getFromPointer(UsedPrefixStart),
886 "found '" + UsedPrefix + "-LABEL:'"
887 " with variable definition or use");
891 Buffer = Buffer.substr(EOL);
893 // Verify that CHECK-NEXT lines have at least one CHECK line before them.
894 if ((CheckTy == Check::CheckNext) && CheckStrings.empty()) {
895 SM.PrintMessage(SMLoc::getFromPointer(UsedPrefixStart),
897 "found '" + UsedPrefix + "-NEXT:' without previous '"
898 + UsedPrefix + ": line");
902 // Handle CHECK-DAG/-NOT.
903 if (CheckTy == Check::CheckDAG || CheckTy == Check::CheckNot) {
904 DagNotMatches.push_back(P);
908 // Okay, add the string we captured to the output vector and move on.
909 CheckStrings.push_back(CheckString(P,
913 std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
916 // Add an EOF pattern for any trailing CHECK-DAG/-NOTs, and use the first
917 // prefix as a filler for the error message.
918 if (!DagNotMatches.empty()) {
919 CheckStrings.push_back(CheckString(Pattern(Check::CheckEOF),
921 SMLoc::getFromPointer(Buffer.data()),
923 std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
926 if (CheckStrings.empty()) {
927 errs() << "error: no check strings found with prefix"
928 << (CheckPrefixes.size() > 1 ? "es " : " ");
929 for (size_t I = 0, N = CheckPrefixes.size(); I != N; ++I) {
930 StringRef Prefix(CheckPrefixes[I]);
931 errs() << '\'' << Prefix << ":'";
943 static void PrintCheckFailed(const SourceMgr &SM, const SMLoc &Loc,
944 const Pattern &Pat, StringRef Buffer,
945 StringMap<StringRef> &VariableTable) {
946 // Otherwise, we have an error, emit an error message.
947 SM.PrintMessage(Loc, SourceMgr::DK_Error,
948 "expected string not found in input");
950 // Print the "scanning from here" line. If the current position is at the
951 // end of a line, advance to the start of the next line.
952 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t\n\r"));
954 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
955 "scanning from here");
957 // Allow the pattern to print additional information if desired.
958 Pat.PrintFailureInfo(SM, Buffer, VariableTable);
961 static void PrintCheckFailed(const SourceMgr &SM, const CheckString &CheckStr,
963 StringMap<StringRef> &VariableTable) {
964 PrintCheckFailed(SM, CheckStr.Loc, CheckStr.Pat, Buffer, VariableTable);
967 /// CountNumNewlinesBetween - Count the number of newlines in the specified
969 static unsigned CountNumNewlinesBetween(StringRef Range) {
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);
987 size_t CheckString::Check(const SourceMgr &SM, StringRef Buffer,
988 bool IsLabelScanMode, size_t &MatchLen,
989 StringMap<StringRef> &VariableTable) const {
991 std::vector<const Pattern *> NotStrings;
993 // IsLabelScanMode is true when we are scanning forward to find CHECK-LABEL
994 // bounds; we have not processed variable definitions within the bounded block
995 // yet so cannot handle any final CHECK-DAG yet; this is handled when going
996 // over the block again (including the last CHECK-LABEL) in normal mode.
997 if (!IsLabelScanMode) {
998 // Match "dag strings" (with mixed "not strings" if any).
999 LastPos = CheckDag(SM, Buffer, NotStrings, VariableTable);
1000 if (LastPos == StringRef::npos)
1001 return StringRef::npos;
1004 // Match itself from the last position after matching CHECK-DAG.
1005 StringRef MatchBuffer = Buffer.substr(LastPos);
1006 size_t MatchPos = Pat.Match(MatchBuffer, MatchLen, VariableTable);
1007 if (MatchPos == StringRef::npos) {
1008 PrintCheckFailed(SM, *this, MatchBuffer, VariableTable);
1009 return StringRef::npos;
1011 MatchPos += LastPos;
1013 // Similar to the above, in "label-scan mode" we can't yet handle CHECK-NEXT
1015 if (!IsLabelScanMode) {
1016 StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
1018 // If this check is a "CHECK-NEXT", verify that the previous match was on
1019 // the previous line (i.e. that there is one newline between them).
1020 if (CheckNext(SM, SkippedRegion))
1021 return StringRef::npos;
1023 // If this match had "not strings", verify that they don't exist in the
1025 if (CheckNot(SM, SkippedRegion, NotStrings, VariableTable))
1026 return StringRef::npos;
1032 bool CheckString::CheckNext(const SourceMgr &SM, StringRef Buffer) const {
1033 if (CheckTy != Check::CheckNext)
1036 // Count the number of newlines between the previous match and this one.
1037 assert(Buffer.data() !=
1039 SM.FindBufferContainingLoc(
1040 SMLoc::getFromPointer(Buffer.data())))->getBufferStart() &&
1041 "CHECK-NEXT can't be the first check in a file");
1043 unsigned NumNewLines = CountNumNewlinesBetween(Buffer);
1045 if (NumNewLines == 0) {
1046 SM.PrintMessage(Loc, SourceMgr::DK_Error, Prefix +
1047 "-NEXT: is on the same line as previous match");
1048 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()),
1049 SourceMgr::DK_Note, "'next' match was here");
1050 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
1051 "previous match ended here");
1055 if (NumNewLines != 1) {
1056 SM.PrintMessage(Loc, SourceMgr::DK_Error, Prefix +
1057 "-NEXT: is not on the line after the previous match");
1058 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()),
1059 SourceMgr::DK_Note, "'next' match was here");
1060 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
1061 "previous match ended here");
1068 bool CheckString::CheckNot(const SourceMgr &SM, StringRef Buffer,
1069 const std::vector<const Pattern *> &NotStrings,
1070 StringMap<StringRef> &VariableTable) const {
1071 for (unsigned ChunkNo = 0, e = NotStrings.size();
1072 ChunkNo != e; ++ChunkNo) {
1073 const Pattern *Pat = NotStrings[ChunkNo];
1074 assert((Pat->getCheckTy() == Check::CheckNot) && "Expect CHECK-NOT!");
1076 size_t MatchLen = 0;
1077 size_t Pos = Pat->Match(Buffer, MatchLen, VariableTable);
1079 if (Pos == StringRef::npos) continue;
1081 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()+Pos),
1082 SourceMgr::DK_Error,
1083 Prefix + "-NOT: string occurred!");
1084 SM.PrintMessage(Pat->getLoc(), SourceMgr::DK_Note,
1085 Prefix + "-NOT: pattern specified here");
1092 size_t CheckString::CheckDag(const SourceMgr &SM, StringRef Buffer,
1093 std::vector<const Pattern *> &NotStrings,
1094 StringMap<StringRef> &VariableTable) const {
1095 if (DagNotStrings.empty())
1099 size_t StartPos = LastPos;
1101 for (unsigned ChunkNo = 0, e = DagNotStrings.size();
1102 ChunkNo != e; ++ChunkNo) {
1103 const Pattern &Pat = DagNotStrings[ChunkNo];
1105 assert((Pat.getCheckTy() == Check::CheckDAG ||
1106 Pat.getCheckTy() == Check::CheckNot) &&
1107 "Invalid CHECK-DAG or CHECK-NOT!");
1109 if (Pat.getCheckTy() == Check::CheckNot) {
1110 NotStrings.push_back(&Pat);
1114 assert((Pat.getCheckTy() == Check::CheckDAG) && "Expect CHECK-DAG!");
1116 size_t MatchLen = 0, MatchPos;
1118 // CHECK-DAG always matches from the start.
1119 StringRef MatchBuffer = Buffer.substr(StartPos);
1120 MatchPos = Pat.Match(MatchBuffer, MatchLen, VariableTable);
1121 // With a group of CHECK-DAGs, a single mismatching means the match on
1122 // that group of CHECK-DAGs fails immediately.
1123 if (MatchPos == StringRef::npos) {
1124 PrintCheckFailed(SM, Pat.getLoc(), Pat, MatchBuffer, VariableTable);
1125 return StringRef::npos;
1127 // Re-calc it as the offset relative to the start of the original string.
1128 MatchPos += StartPos;
1130 if (!NotStrings.empty()) {
1131 if (MatchPos < LastPos) {
1133 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + MatchPos),
1134 SourceMgr::DK_Error,
1135 Prefix + "-DAG: found a match of CHECK-DAG"
1136 " reordering across a CHECK-NOT");
1137 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + LastPos),
1139 Prefix + "-DAG: the farthest match of CHECK-DAG"
1141 SM.PrintMessage(NotStrings[0]->getLoc(), SourceMgr::DK_Note,
1142 Prefix + "-NOT: the crossed pattern specified"
1144 SM.PrintMessage(Pat.getLoc(), SourceMgr::DK_Note,
1145 Prefix + "-DAG: the reordered pattern specified"
1147 return StringRef::npos;
1149 // All subsequent CHECK-DAGs should be matched from the farthest
1150 // position of all precedent CHECK-DAGs (including this one.)
1152 // If there's CHECK-NOTs between two CHECK-DAGs or from CHECK to
1153 // CHECK-DAG, verify that there's no 'not' strings occurred in that
1155 StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
1156 if (CheckNot(SM, SkippedRegion, NotStrings, VariableTable))
1157 return StringRef::npos;
1158 // Clear "not strings".
1162 // Update the last position with CHECK-DAG matches.
1163 LastPos = std::max(MatchPos + MatchLen, LastPos);
1169 // A check prefix must contain only alphanumeric, hyphens and underscores.
1170 static bool ValidateCheckPrefix(StringRef CheckPrefix) {
1171 Regex Validator("^[a-zA-Z0-9_-]*$");
1172 return Validator.match(CheckPrefix);
1175 static bool ValidateCheckPrefixes() {
1176 StringSet<> PrefixSet;
1178 for (prefix_iterator I = CheckPrefixes.begin(), E = CheckPrefixes.end();
1180 StringRef Prefix(*I);
1182 if (!PrefixSet.insert(Prefix))
1185 if (!ValidateCheckPrefix(Prefix))
1192 // I don't think there's a way to specify an initial value for cl::list,
1193 // so if nothing was specified, add the default
1194 static void AddCheckPrefixIfNeeded() {
1195 if (CheckPrefixes.empty())
1196 CheckPrefixes.push_back("CHECK");
1199 int main(int argc, char **argv) {
1200 sys::PrintStackTraceOnErrorSignal();
1201 PrettyStackTraceProgram X(argc, argv);
1202 cl::ParseCommandLineOptions(argc, argv);
1204 if (!ValidateCheckPrefixes()) {
1205 errs() << "Supplied check-prefix is invalid! Prefixes must be unique and "
1206 "start with a letter and contain only alphanumeric characters, "
1207 "hyphens and underscores\n";
1211 AddCheckPrefixIfNeeded();
1215 // Read the expected strings from the check file.
1216 std::vector<CheckString> CheckStrings;
1217 if (ReadCheckFile(SM, CheckStrings))
1220 // Open the file to check and add it to SourceMgr.
1221 OwningPtr<MemoryBuffer> File;
1223 MemoryBuffer::getFileOrSTDIN(InputFilename, File)) {
1224 errs() << "Could not open input file '" << InputFilename << "': "
1225 << ec.message() << '\n';
1229 if (File->getBufferSize() == 0) {
1230 errs() << "FileCheck error: '" << InputFilename << "' is empty.\n";
1234 // Remove duplicate spaces in the input file if requested.
1235 // Remove DOS style line endings.
1237 CanonicalizeInputFile(File.take(), NoCanonicalizeWhiteSpace);
1239 SM.AddNewSourceBuffer(F, SMLoc());
1241 /// VariableTable - This holds all the current filecheck variables.
1242 StringMap<StringRef> VariableTable;
1244 // Check that we have all of the expected strings, in order, in the input
1246 StringRef Buffer = F->getBuffer();
1248 bool hasError = false;
1250 unsigned i = 0, j = 0, e = CheckStrings.size();
1253 StringRef CheckRegion;
1255 CheckRegion = Buffer;
1257 const CheckString &CheckLabelStr = CheckStrings[j];
1258 if (CheckLabelStr.CheckTy != Check::CheckLabel) {
1263 // Scan to next CHECK-LABEL match, ignoring CHECK-NOT and CHECK-DAG
1264 size_t MatchLabelLen = 0;
1265 size_t MatchLabelPos = CheckLabelStr.Check(SM, Buffer, true,
1266 MatchLabelLen, VariableTable);
1267 if (MatchLabelPos == StringRef::npos) {
1272 CheckRegion = Buffer.substr(0, MatchLabelPos + MatchLabelLen);
1273 Buffer = Buffer.substr(MatchLabelPos + MatchLabelLen);
1277 for ( ; i != j; ++i) {
1278 const CheckString &CheckStr = CheckStrings[i];
1280 // Check each string within the scanned region, including a second check
1281 // of any final CHECK-LABEL (to verify CHECK-NOT and CHECK-DAG)
1282 size_t MatchLen = 0;
1283 size_t MatchPos = CheckStr.Check(SM, CheckRegion, false, MatchLen,
1286 if (MatchPos == StringRef::npos) {
1292 CheckRegion = CheckRegion.substr(MatchPos + MatchLen);
1299 return hasError ? 1 : 0;