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 static void AddFixedStringToRegEx(StringRef FixedStr, std::string &TheStr);
140 bool AddRegExToRegEx(StringRef RS, unsigned &CurParen, SourceMgr &SM);
141 void AddBackrefToRegEx(unsigned BackrefNum);
143 /// ComputeMatchDistance - Compute an arbitrary estimate for the quality of
144 /// matching this pattern at the start of \arg Buffer; a distance of zero
145 /// should correspond to a perfect match.
146 unsigned ComputeMatchDistance(StringRef Buffer,
147 const StringMap<StringRef> &VariableTable) const;
149 /// \brief Evaluates expression and stores the result to \p Value.
150 /// \return true on success. false when the expression has invalid syntax.
151 bool EvaluateExpression(StringRef Expr, std::string &Value) const;
153 /// \brief Finds the closing sequence of a regex variable usage or
154 /// definition. Str has to point in the beginning of the definition
155 /// (right after the opening sequence).
156 /// \return offset of the closing sequence within Str, or npos if it was not
158 size_t FindRegexVarEnd(StringRef Str);
162 bool Pattern::ParsePattern(StringRef PatternStr,
165 unsigned LineNumber) {
166 this->LineNumber = LineNumber;
167 PatternLoc = SMLoc::getFromPointer(PatternStr.data());
169 // Ignore trailing whitespace.
170 while (!PatternStr.empty() &&
171 (PatternStr.back() == ' ' || PatternStr.back() == '\t'))
172 PatternStr = PatternStr.substr(0, PatternStr.size()-1);
174 // Check that there is something on the line.
175 if (PatternStr.empty()) {
176 SM.PrintMessage(PatternLoc, SourceMgr::DK_Error,
177 "found empty check string with prefix '" +
182 // Check to see if this is a fixed string, or if it has regex pieces.
183 if (PatternStr.size() < 2 ||
184 (PatternStr.find("{{") == StringRef::npos &&
185 PatternStr.find("[[") == StringRef::npos)) {
186 FixedStr = PatternStr;
190 // Paren value #0 is for the fully matched string. Any new parenthesized
191 // values add from there.
192 unsigned CurParen = 1;
194 // Otherwise, there is at least one regex piece. Build up the regex pattern
195 // by escaping scary characters in fixed strings, building up one big regex.
196 while (!PatternStr.empty()) {
198 if (PatternStr.startswith("{{")) {
199 // This is the start of a regex match. Scan for the }}.
200 size_t End = PatternStr.find("}}");
201 if (End == StringRef::npos) {
202 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
204 "found start of regex string with no end '}}'");
208 // Enclose {{}} patterns in parens just like [[]] even though we're not
209 // capturing the result for any purpose. This is required in case the
210 // expression contains an alternation like: CHECK: abc{{x|z}}def. We
211 // want this to turn into: "abc(x|z)def" not "abcx|zdef".
215 if (AddRegExToRegEx(PatternStr.substr(2, End-2), CurParen, SM))
219 PatternStr = PatternStr.substr(End+2);
223 // Named RegEx matches. These are of two forms: [[foo:.*]] which matches .*
224 // (or some other regex) and assigns it to the FileCheck variable 'foo'. The
225 // second form is [[foo]] which is a reference to foo. The variable name
226 // itself must be of the form "[a-zA-Z_][0-9a-zA-Z_]*", otherwise we reject
227 // it. This is to catch some common errors.
228 if (PatternStr.startswith("[[")) {
229 // Find the closing bracket pair ending the match. End is going to be an
230 // offset relative to the beginning of the match string.
231 size_t End = FindRegexVarEnd(PatternStr.substr(2));
233 if (End == StringRef::npos) {
234 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
236 "invalid named regex reference, no ]] found");
240 StringRef MatchStr = PatternStr.substr(2, End);
241 PatternStr = PatternStr.substr(End+4);
243 // Get the regex name (e.g. "foo").
244 size_t NameEnd = MatchStr.find(':');
245 StringRef Name = MatchStr.substr(0, NameEnd);
248 SM.PrintMessage(SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
249 "invalid name in named regex: empty name");
253 // Verify that the name/expression is well formed. FileCheck currently
254 // supports @LINE, @LINE+number, @LINE-number expressions. The check here
255 // is relaxed, more strict check is performed in \c EvaluateExpression.
256 bool IsExpression = false;
257 for (unsigned i = 0, e = Name.size(); i != e; ++i) {
258 if (i == 0 && Name[i] == '@') {
259 if (NameEnd != StringRef::npos) {
260 SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
262 "invalid name in named regex definition");
268 if (Name[i] != '_' && !isalnum(Name[i]) &&
269 (!IsExpression || (Name[i] != '+' && Name[i] != '-'))) {
270 SM.PrintMessage(SMLoc::getFromPointer(Name.data()+i),
271 SourceMgr::DK_Error, "invalid name in named regex");
276 // Name can't start with a digit.
277 if (isdigit(static_cast<unsigned char>(Name[0]))) {
278 SM.PrintMessage(SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
279 "invalid name in named regex");
284 if (NameEnd == StringRef::npos) {
285 // Handle variables that were defined earlier on the same line by
286 // emitting a backreference.
287 if (VariableDefs.find(Name) != VariableDefs.end()) {
288 unsigned VarParenNum = VariableDefs[Name];
289 if (VarParenNum < 1 || VarParenNum > 9) {
290 SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
292 "Can't back-reference more than 9 variables");
295 AddBackrefToRegEx(VarParenNum);
297 VariableUses.push_back(std::make_pair(Name, RegExStr.size()));
302 // Handle [[foo:.*]].
303 VariableDefs[Name] = CurParen;
307 if (AddRegExToRegEx(MatchStr.substr(NameEnd+1), CurParen, SM))
313 // Handle fixed string matches.
314 // Find the end, which is the start of the next regex.
315 size_t FixedMatchEnd = PatternStr.find("{{");
316 FixedMatchEnd = std::min(FixedMatchEnd, PatternStr.find("[["));
317 AddFixedStringToRegEx(PatternStr.substr(0, FixedMatchEnd), RegExStr);
318 PatternStr = PatternStr.substr(FixedMatchEnd);
324 void Pattern::AddFixedStringToRegEx(StringRef FixedStr, std::string &TheStr) {
325 // Add the characters from FixedStr to the regex, escaping as needed. This
326 // avoids "leaning toothpicks" in common patterns.
327 for (unsigned i = 0, e = FixedStr.size(); i != e; ++i) {
328 switch (FixedStr[i]) {
329 // These are the special characters matched in "p_ere_exp".
345 TheStr += FixedStr[i];
351 bool Pattern::AddRegExToRegEx(StringRef RS, unsigned &CurParen,
355 if (!R.isValid(Error)) {
356 SM.PrintMessage(SMLoc::getFromPointer(RS.data()), SourceMgr::DK_Error,
357 "invalid regex: " + Error);
361 RegExStr += RS.str();
362 CurParen += R.getNumMatches();
366 void Pattern::AddBackrefToRegEx(unsigned BackrefNum) {
367 assert(BackrefNum >= 1 && BackrefNum <= 9 && "Invalid backref number");
368 std::string Backref = std::string("\\") +
369 std::string(1, '0' + BackrefNum);
373 bool Pattern::EvaluateExpression(StringRef Expr, std::string &Value) const {
374 // The only supported expression is @LINE([\+-]\d+)?
375 if (!Expr.startswith("@LINE"))
377 Expr = Expr.substr(StringRef("@LINE").size());
381 Expr = Expr.substr(1);
382 else if (Expr[0] != '-')
384 if (Expr.getAsInteger(10, Offset))
387 Value = llvm::itostr(LineNumber + Offset);
391 /// Match - Match the pattern string against the input buffer Buffer. This
392 /// returns the position that is matched or npos if there is no match. If
393 /// there is a match, the size of the matched string is returned in MatchLen.
394 size_t Pattern::Match(StringRef Buffer, size_t &MatchLen,
395 StringMap<StringRef> &VariableTable) const {
396 // If this is the EOF pattern, match it immediately.
397 if (CheckTy == Check::CheckEOF) {
399 return Buffer.size();
402 // If this is a fixed string pattern, just match it now.
403 if (!FixedStr.empty()) {
404 MatchLen = FixedStr.size();
405 return Buffer.find(FixedStr);
410 // If there are variable uses, we need to create a temporary string with the
412 StringRef RegExToMatch = RegExStr;
414 if (!VariableUses.empty()) {
417 unsigned InsertOffset = 0;
418 for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
421 if (VariableUses[i].first[0] == '@') {
422 if (!EvaluateExpression(VariableUses[i].first, Value))
423 return StringRef::npos;
425 StringMap<StringRef>::iterator it =
426 VariableTable.find(VariableUses[i].first);
427 // If the variable is undefined, return an error.
428 if (it == VariableTable.end())
429 return StringRef::npos;
431 // Look up the value and escape it so that we can plop it into the regex.
432 AddFixedStringToRegEx(it->second, Value);
435 // Plop it into the regex at the adjusted offset.
436 TmpStr.insert(TmpStr.begin()+VariableUses[i].second+InsertOffset,
437 Value.begin(), Value.end());
438 InsertOffset += Value.size();
441 // Match the newly constructed regex.
442 RegExToMatch = TmpStr;
446 SmallVector<StringRef, 4> MatchInfo;
447 if (!Regex(RegExToMatch, Regex::Newline).match(Buffer, &MatchInfo))
448 return StringRef::npos;
450 // Successful regex match.
451 assert(!MatchInfo.empty() && "Didn't get any match");
452 StringRef FullMatch = MatchInfo[0];
454 // If this defines any variables, remember their values.
455 for (std::map<StringRef, unsigned>::const_iterator I = VariableDefs.begin(),
456 E = VariableDefs.end();
458 assert(I->second < MatchInfo.size() && "Internal paren error");
459 VariableTable[I->first] = MatchInfo[I->second];
462 MatchLen = FullMatch.size();
463 return FullMatch.data()-Buffer.data();
466 unsigned Pattern::ComputeMatchDistance(StringRef Buffer,
467 const StringMap<StringRef> &VariableTable) const {
468 // Just compute the number of matching characters. For regular expressions, we
469 // just compare against the regex itself and hope for the best.
471 // FIXME: One easy improvement here is have the regex lib generate a single
472 // example regular expression which matches, and use that as the example
474 StringRef ExampleString(FixedStr);
475 if (ExampleString.empty())
476 ExampleString = RegExStr;
478 // Only compare up to the first line in the buffer, or the string size.
479 StringRef BufferPrefix = Buffer.substr(0, ExampleString.size());
480 BufferPrefix = BufferPrefix.split('\n').first;
481 return BufferPrefix.edit_distance(ExampleString);
484 void Pattern::PrintFailureInfo(const SourceMgr &SM, StringRef Buffer,
485 const StringMap<StringRef> &VariableTable) const{
486 // If this was a regular expression using variables, print the current
488 if (!VariableUses.empty()) {
489 for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
490 SmallString<256> Msg;
491 raw_svector_ostream OS(Msg);
492 StringRef Var = VariableUses[i].first;
495 if (EvaluateExpression(Var, Value)) {
496 OS << "with expression \"";
497 OS.write_escaped(Var) << "\" equal to \"";
498 OS.write_escaped(Value) << "\"";
500 OS << "uses incorrect expression \"";
501 OS.write_escaped(Var) << "\"";
504 StringMap<StringRef>::const_iterator it = VariableTable.find(Var);
506 // Check for undefined variable references.
507 if (it == VariableTable.end()) {
508 OS << "uses undefined variable \"";
509 OS.write_escaped(Var) << "\"";
511 OS << "with variable \"";
512 OS.write_escaped(Var) << "\" equal to \"";
513 OS.write_escaped(it->second) << "\"";
517 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
522 // Attempt to find the closest/best fuzzy match. Usually an error happens
523 // because some string in the output didn't exactly match. In these cases, we
524 // would like to show the user a best guess at what "should have" matched, to
525 // save them having to actually check the input manually.
526 size_t NumLinesForward = 0;
527 size_t Best = StringRef::npos;
528 double BestQuality = 0;
530 // Use an arbitrary 4k limit on how far we will search.
531 for (size_t i = 0, e = std::min(size_t(4096), Buffer.size()); i != e; ++i) {
532 if (Buffer[i] == '\n')
535 // Patterns have leading whitespace stripped, so skip whitespace when
536 // looking for something which looks like a pattern.
537 if (Buffer[i] == ' ' || Buffer[i] == '\t')
540 // Compute the "quality" of this match as an arbitrary combination of the
541 // match distance and the number of lines skipped to get to this match.
542 unsigned Distance = ComputeMatchDistance(Buffer.substr(i), VariableTable);
543 double Quality = Distance + (NumLinesForward / 100.);
545 if (Quality < BestQuality || Best == StringRef::npos) {
547 BestQuality = Quality;
551 // Print the "possible intended match here" line if we found something
552 // reasonable and not equal to what we showed in the "scanning from here"
554 if (Best && Best != StringRef::npos && BestQuality < 50) {
555 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + Best),
556 SourceMgr::DK_Note, "possible intended match here");
558 // FIXME: If we wanted to be really friendly we would show why the match
559 // failed, as it can be hard to spot simple one character differences.
563 size_t Pattern::FindRegexVarEnd(StringRef Str) {
564 // Offset keeps track of the current offset within the input Str
566 // [...] Nesting depth
567 size_t BracketDepth = 0;
569 while (!Str.empty()) {
570 if (Str.startswith("]]") && BracketDepth == 0)
572 if (Str[0] == '\\') {
573 // Backslash escapes the next char within regexes, so skip them both.
584 assert(BracketDepth > 0 && "Invalid regex");
593 return StringRef::npos;
597 //===----------------------------------------------------------------------===//
599 //===----------------------------------------------------------------------===//
601 /// CheckString - This is a check that we found in the input file.
603 /// Pat - The pattern to match.
606 /// Prefix - Which prefix name this check matched.
609 /// Loc - The location in the match file that the check string was specified.
612 /// CheckTy - Specify what kind of check this is. e.g. CHECK-NEXT: directive,
613 /// as opposed to a CHECK: directive.
614 Check::CheckType CheckTy;
616 /// DagNotStrings - These are all of the strings that are disallowed from
617 /// occurring between this match string and the previous one (or start of
619 std::vector<Pattern> DagNotStrings;
622 CheckString(const Pattern &P,
626 : Pat(P), Prefix(S), Loc(L), CheckTy(Ty) {}
628 /// Check - Match check string and its "not strings" and/or "dag strings".
629 size_t Check(const SourceMgr &SM, StringRef Buffer, bool IsLabelScanMode,
630 size_t &MatchLen, StringMap<StringRef> &VariableTable) const;
632 /// CheckNext - Verify there is a single line in the given buffer.
633 bool CheckNext(const SourceMgr &SM, StringRef Buffer) const;
635 /// CheckNot - Verify there's no "not strings" in the given buffer.
636 bool CheckNot(const SourceMgr &SM, StringRef Buffer,
637 const std::vector<const Pattern *> &NotStrings,
638 StringMap<StringRef> &VariableTable) const;
640 /// CheckDag - Match "dag strings" and their mixed "not strings".
641 size_t CheckDag(const SourceMgr &SM, StringRef Buffer,
642 std::vector<const Pattern *> &NotStrings,
643 StringMap<StringRef> &VariableTable) const;
646 /// Canonicalize whitespaces in the input file. Line endings are replaced
647 /// with UNIX-style '\n'.
649 /// \param PreserveHorizontal Don't squash consecutive horizontal whitespace
650 /// characters to a single space.
651 static MemoryBuffer *CanonicalizeInputFile(MemoryBuffer *MB,
652 bool PreserveHorizontal) {
653 SmallString<128> NewFile;
654 NewFile.reserve(MB->getBufferSize());
656 for (const char *Ptr = MB->getBufferStart(), *End = MB->getBufferEnd();
658 // Eliminate trailing dosish \r.
659 if (Ptr <= End - 2 && Ptr[0] == '\r' && Ptr[1] == '\n') {
663 // If current char is not a horizontal whitespace or if horizontal
664 // whitespace canonicalization is disabled, dump it to output as is.
665 if (PreserveHorizontal || (*Ptr != ' ' && *Ptr != '\t')) {
666 NewFile.push_back(*Ptr);
670 // Otherwise, add one space and advance over neighboring space.
671 NewFile.push_back(' ');
672 while (Ptr+1 != End &&
673 (Ptr[1] == ' ' || Ptr[1] == '\t'))
677 // Free the old buffer and return a new one.
679 MemoryBuffer::getMemBufferCopy(NewFile.str(), MB->getBufferIdentifier());
685 static bool IsPartOfWord(char c) {
686 return (isalnum(c) || c == '-' || c == '_');
689 // Get the size of the prefix extension.
690 static size_t CheckTypeSize(Check::CheckType Ty) {
692 case Check::CheckNone:
695 case Check::CheckPlain:
696 return sizeof(":") - 1;
698 case Check::CheckNext:
699 return sizeof("-NEXT:") - 1;
701 case Check::CheckNot:
702 return sizeof("-NOT:") - 1;
704 case Check::CheckDAG:
705 return sizeof("-DAG:") - 1;
707 case Check::CheckLabel:
708 return sizeof("-LABEL:") - 1;
710 case Check::CheckEOF:
711 llvm_unreachable("Should not be using EOF size");
714 llvm_unreachable("Bad check type");
717 static Check::CheckType FindCheckType(StringRef Buffer, StringRef Prefix) {
718 char NextChar = Buffer[Prefix.size()];
720 // Verify that the : is present after the prefix.
722 return Check::CheckPlain;
725 return Check::CheckNone;
727 StringRef Rest = Buffer.drop_front(Prefix.size() + 1);
728 if (Rest.startswith("NEXT:"))
729 return Check::CheckNext;
731 if (Rest.startswith("NOT:"))
732 return Check::CheckNot;
734 if (Rest.startswith("DAG:"))
735 return Check::CheckDAG;
737 if (Rest.startswith("LABEL:"))
738 return Check::CheckLabel;
740 return Check::CheckNone;
743 // From the given position, find the next character after the word.
744 static size_t SkipWord(StringRef Str, size_t Loc) {
745 while (Loc < Str.size() && IsPartOfWord(Str[Loc]))
750 // Try to find the first match in buffer for any prefix. If a valid match is
751 // found, return that prefix and set its type and location. If there are almost
752 // matches (e.g. the actual prefix string is found, but is not an actual check
753 // string), but no valid match, return an empty string and set the position to
754 // resume searching from. If no partial matches are found, return an empty
755 // string and the location will be StringRef::npos. If one prefix is a substring
756 // of another, the maximal match should be found. e.g. if "A" and "AA" are
757 // prefixes then AA-CHECK: should match the second one.
758 static StringRef FindFirstCandidateMatch(StringRef &Buffer,
759 Check::CheckType &CheckTy,
761 StringRef FirstPrefix;
762 size_t FirstLoc = StringRef::npos;
763 size_t SearchLoc = StringRef::npos;
764 Check::CheckType FirstTy = Check::CheckNone;
766 CheckTy = Check::CheckNone;
767 CheckLoc = StringRef::npos;
769 for (prefix_iterator I = CheckPrefixes.begin(), E = CheckPrefixes.end();
771 StringRef Prefix(*I);
772 size_t PrefixLoc = Buffer.find(Prefix);
774 if (PrefixLoc == StringRef::npos)
777 // Track where we are searching for invalid prefixes that look almost right.
778 // We need to only advance to the first partial match on the next attempt
779 // since a partial match could be a substring of a later, valid prefix.
780 // Need to skip to the end of the word, otherwise we could end up
781 // matching a prefix in a substring later.
782 if (PrefixLoc < SearchLoc)
783 SearchLoc = SkipWord(Buffer, PrefixLoc);
785 // We only want to find the first match to avoid skipping some.
786 if (PrefixLoc > FirstLoc)
788 // If one matching check-prefix is a prefix of another, choose the
790 if (PrefixLoc == FirstLoc && Prefix.size() < FirstPrefix.size())
793 StringRef Rest = Buffer.drop_front(PrefixLoc);
794 // Make sure we have actually found the prefix, and not a word containing
795 // it. This should also prevent matching the wrong prefix when one is a
796 // substring of another.
797 if (PrefixLoc != 0 && IsPartOfWord(Buffer[PrefixLoc - 1]))
800 FirstLoc = PrefixLoc;
801 FirstTy = FindCheckType(Rest, Prefix);
802 FirstPrefix = Prefix;
805 // If the first prefix is invalid, we should continue the search after it.
806 if (FirstTy == Check::CheckNone) {
807 CheckLoc = SearchLoc;
816 static StringRef FindFirstMatchingPrefix(StringRef &Buffer,
817 unsigned &LineNumber,
818 Check::CheckType &CheckTy,
820 while (!Buffer.empty()) {
821 StringRef Prefix = FindFirstCandidateMatch(Buffer, CheckTy, CheckLoc);
822 // If we found a real match, we are done.
823 if (!Prefix.empty()) {
824 LineNumber += Buffer.substr(0, CheckLoc).count('\n');
828 // We didn't find any almost matches either, we are also done.
829 if (CheckLoc == StringRef::npos)
832 LineNumber += Buffer.substr(0, CheckLoc + 1).count('\n');
834 // Advance to the last possible match we found and try again.
835 Buffer = Buffer.drop_front(CheckLoc + 1);
841 /// ReadCheckFile - Read the check file, which specifies the sequence of
842 /// expected strings. The strings are added to the CheckStrings vector.
843 /// Returns true in case of an error, false otherwise.
844 static bool ReadCheckFile(SourceMgr &SM,
845 std::vector<CheckString> &CheckStrings) {
846 OwningPtr<MemoryBuffer> File;
848 MemoryBuffer::getFileOrSTDIN(CheckFilename, File)) {
849 errs() << "Could not open check file '" << CheckFilename << "': "
850 << ec.message() << '\n';
854 // If we want to canonicalize whitespace, strip excess whitespace from the
855 // buffer containing the CHECK lines. Remove DOS style line endings.
857 CanonicalizeInputFile(File.take(), NoCanonicalizeWhiteSpace);
859 SM.AddNewSourceBuffer(F, SMLoc());
861 // Find all instances of CheckPrefix followed by : in the file.
862 StringRef Buffer = F->getBuffer();
863 std::vector<Pattern> DagNotMatches;
865 // LineNumber keeps track of the line on which CheckPrefix instances are
867 unsigned LineNumber = 1;
870 Check::CheckType CheckTy;
873 // See if a prefix occurs in the memory buffer.
874 StringRef UsedPrefix = FindFirstMatchingPrefix(Buffer,
878 if (UsedPrefix.empty())
881 Buffer = Buffer.drop_front(PrefixLoc);
883 // Location to use for error messages.
884 const char *UsedPrefixStart = Buffer.data() + (PrefixLoc == 0 ? 0 : 1);
886 // PrefixLoc is to the start of the prefix. Skip to the end.
887 Buffer = Buffer.drop_front(UsedPrefix.size() + CheckTypeSize(CheckTy));
889 // Okay, we found the prefix, yay. Remember the rest of the line, but ignore
890 // leading and trailing whitespace.
891 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t"));
893 // Scan ahead to the end of line.
894 size_t EOL = Buffer.find_first_of("\n\r");
896 // Remember the location of the start of the pattern, for diagnostics.
897 SMLoc PatternLoc = SMLoc::getFromPointer(Buffer.data());
899 // Parse the pattern.
901 if (P.ParsePattern(Buffer.substr(0, EOL), UsedPrefix, SM, LineNumber))
904 // Verify that CHECK-LABEL lines do not define or use variables
905 if ((CheckTy == Check::CheckLabel) && P.hasVariable()) {
906 SM.PrintMessage(SMLoc::getFromPointer(UsedPrefixStart),
908 "found '" + UsedPrefix + "-LABEL:'"
909 " with variable definition or use");
913 Buffer = Buffer.substr(EOL);
915 // Verify that CHECK-NEXT lines have at least one CHECK line before them.
916 if ((CheckTy == Check::CheckNext) && CheckStrings.empty()) {
917 SM.PrintMessage(SMLoc::getFromPointer(UsedPrefixStart),
919 "found '" + UsedPrefix + "-NEXT:' without previous '"
920 + UsedPrefix + ": line");
924 // Handle CHECK-DAG/-NOT.
925 if (CheckTy == Check::CheckDAG || CheckTy == Check::CheckNot) {
926 DagNotMatches.push_back(P);
930 // Okay, add the string we captured to the output vector and move on.
931 CheckStrings.push_back(CheckString(P,
935 std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
938 // Add an EOF pattern for any trailing CHECK-DAG/-NOTs, and use the first
939 // prefix as a filler for the error message.
940 if (!DagNotMatches.empty()) {
941 CheckStrings.push_back(CheckString(Pattern(Check::CheckEOF),
943 SMLoc::getFromPointer(Buffer.data()),
945 std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
948 if (CheckStrings.empty()) {
949 errs() << "error: no check strings found with prefix"
950 << (CheckPrefixes.size() > 1 ? "es " : " ");
951 for (size_t I = 0, N = CheckPrefixes.size(); I != N; ++I) {
952 StringRef Prefix(CheckPrefixes[I]);
953 errs() << '\'' << Prefix << ":'";
965 static void PrintCheckFailed(const SourceMgr &SM, const SMLoc &Loc,
966 const Pattern &Pat, StringRef Buffer,
967 StringMap<StringRef> &VariableTable) {
968 // Otherwise, we have an error, emit an error message.
969 SM.PrintMessage(Loc, SourceMgr::DK_Error,
970 "expected string not found in input");
972 // Print the "scanning from here" line. If the current position is at the
973 // end of a line, advance to the start of the next line.
974 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t\n\r"));
976 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
977 "scanning from here");
979 // Allow the pattern to print additional information if desired.
980 Pat.PrintFailureInfo(SM, Buffer, VariableTable);
983 static void PrintCheckFailed(const SourceMgr &SM, const CheckString &CheckStr,
985 StringMap<StringRef> &VariableTable) {
986 PrintCheckFailed(SM, CheckStr.Loc, CheckStr.Pat, Buffer, VariableTable);
989 /// CountNumNewlinesBetween - Count the number of newlines in the specified
991 static unsigned CountNumNewlinesBetween(StringRef Range) {
992 unsigned NumNewLines = 0;
995 Range = Range.substr(Range.find_first_of("\n\r"));
996 if (Range.empty()) return NumNewLines;
1000 // Handle \n\r and \r\n as a single newline.
1001 if (Range.size() > 1 &&
1002 (Range[1] == '\n' || Range[1] == '\r') &&
1003 (Range[0] != Range[1]))
1004 Range = Range.substr(1);
1005 Range = Range.substr(1);
1009 size_t CheckString::Check(const SourceMgr &SM, StringRef Buffer,
1010 bool IsLabelScanMode, size_t &MatchLen,
1011 StringMap<StringRef> &VariableTable) const {
1013 std::vector<const Pattern *> NotStrings;
1015 // IsLabelScanMode is true when we are scanning forward to find CHECK-LABEL
1016 // bounds; we have not processed variable definitions within the bounded block
1017 // yet so cannot handle any final CHECK-DAG yet; this is handled when going
1018 // over the block again (including the last CHECK-LABEL) in normal mode.
1019 if (!IsLabelScanMode) {
1020 // Match "dag strings" (with mixed "not strings" if any).
1021 LastPos = CheckDag(SM, Buffer, NotStrings, VariableTable);
1022 if (LastPos == StringRef::npos)
1023 return StringRef::npos;
1026 // Match itself from the last position after matching CHECK-DAG.
1027 StringRef MatchBuffer = Buffer.substr(LastPos);
1028 size_t MatchPos = Pat.Match(MatchBuffer, MatchLen, VariableTable);
1029 if (MatchPos == StringRef::npos) {
1030 PrintCheckFailed(SM, *this, MatchBuffer, VariableTable);
1031 return StringRef::npos;
1033 MatchPos += LastPos;
1035 // Similar to the above, in "label-scan mode" we can't yet handle CHECK-NEXT
1037 if (!IsLabelScanMode) {
1038 StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
1040 // If this check is a "CHECK-NEXT", verify that the previous match was on
1041 // the previous line (i.e. that there is one newline between them).
1042 if (CheckNext(SM, SkippedRegion))
1043 return StringRef::npos;
1045 // If this match had "not strings", verify that they don't exist in the
1047 if (CheckNot(SM, SkippedRegion, NotStrings, VariableTable))
1048 return StringRef::npos;
1054 bool CheckString::CheckNext(const SourceMgr &SM, StringRef Buffer) const {
1055 if (CheckTy != Check::CheckNext)
1058 // Count the number of newlines between the previous match and this one.
1059 assert(Buffer.data() !=
1061 SM.FindBufferContainingLoc(
1062 SMLoc::getFromPointer(Buffer.data())))->getBufferStart() &&
1063 "CHECK-NEXT can't be the first check in a file");
1065 unsigned NumNewLines = CountNumNewlinesBetween(Buffer);
1067 if (NumNewLines == 0) {
1068 SM.PrintMessage(Loc, SourceMgr::DK_Error, Prefix +
1069 "-NEXT: is on the same line as previous match");
1070 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()),
1071 SourceMgr::DK_Note, "'next' match was here");
1072 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
1073 "previous match ended here");
1077 if (NumNewLines != 1) {
1078 SM.PrintMessage(Loc, SourceMgr::DK_Error, Prefix +
1079 "-NEXT: is not on the line after the previous match");
1080 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()),
1081 SourceMgr::DK_Note, "'next' match was here");
1082 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
1083 "previous match ended here");
1090 bool CheckString::CheckNot(const SourceMgr &SM, StringRef Buffer,
1091 const std::vector<const Pattern *> &NotStrings,
1092 StringMap<StringRef> &VariableTable) const {
1093 for (unsigned ChunkNo = 0, e = NotStrings.size();
1094 ChunkNo != e; ++ChunkNo) {
1095 const Pattern *Pat = NotStrings[ChunkNo];
1096 assert((Pat->getCheckTy() == Check::CheckNot) && "Expect CHECK-NOT!");
1098 size_t MatchLen = 0;
1099 size_t Pos = Pat->Match(Buffer, MatchLen, VariableTable);
1101 if (Pos == StringRef::npos) continue;
1103 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()+Pos),
1104 SourceMgr::DK_Error,
1105 Prefix + "-NOT: string occurred!");
1106 SM.PrintMessage(Pat->getLoc(), SourceMgr::DK_Note,
1107 Prefix + "-NOT: pattern specified here");
1114 size_t CheckString::CheckDag(const SourceMgr &SM, StringRef Buffer,
1115 std::vector<const Pattern *> &NotStrings,
1116 StringMap<StringRef> &VariableTable) const {
1117 if (DagNotStrings.empty())
1121 size_t StartPos = LastPos;
1123 for (unsigned ChunkNo = 0, e = DagNotStrings.size();
1124 ChunkNo != e; ++ChunkNo) {
1125 const Pattern &Pat = DagNotStrings[ChunkNo];
1127 assert((Pat.getCheckTy() == Check::CheckDAG ||
1128 Pat.getCheckTy() == Check::CheckNot) &&
1129 "Invalid CHECK-DAG or CHECK-NOT!");
1131 if (Pat.getCheckTy() == Check::CheckNot) {
1132 NotStrings.push_back(&Pat);
1136 assert((Pat.getCheckTy() == Check::CheckDAG) && "Expect CHECK-DAG!");
1138 size_t MatchLen = 0, MatchPos;
1140 // CHECK-DAG always matches from the start.
1141 StringRef MatchBuffer = Buffer.substr(StartPos);
1142 MatchPos = Pat.Match(MatchBuffer, MatchLen, VariableTable);
1143 // With a group of CHECK-DAGs, a single mismatching means the match on
1144 // that group of CHECK-DAGs fails immediately.
1145 if (MatchPos == StringRef::npos) {
1146 PrintCheckFailed(SM, Pat.getLoc(), Pat, MatchBuffer, VariableTable);
1147 return StringRef::npos;
1149 // Re-calc it as the offset relative to the start of the original string.
1150 MatchPos += StartPos;
1152 if (!NotStrings.empty()) {
1153 if (MatchPos < LastPos) {
1155 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + MatchPos),
1156 SourceMgr::DK_Error,
1157 Prefix + "-DAG: found a match of CHECK-DAG"
1158 " reordering across a CHECK-NOT");
1159 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + LastPos),
1161 Prefix + "-DAG: the farthest match of CHECK-DAG"
1163 SM.PrintMessage(NotStrings[0]->getLoc(), SourceMgr::DK_Note,
1164 Prefix + "-NOT: the crossed pattern specified"
1166 SM.PrintMessage(Pat.getLoc(), SourceMgr::DK_Note,
1167 Prefix + "-DAG: the reordered pattern specified"
1169 return StringRef::npos;
1171 // All subsequent CHECK-DAGs should be matched from the farthest
1172 // position of all precedent CHECK-DAGs (including this one.)
1174 // If there's CHECK-NOTs between two CHECK-DAGs or from CHECK to
1175 // CHECK-DAG, verify that there's no 'not' strings occurred in that
1177 StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
1178 if (CheckNot(SM, SkippedRegion, NotStrings, VariableTable))
1179 return StringRef::npos;
1180 // Clear "not strings".
1184 // Update the last position with CHECK-DAG matches.
1185 LastPos = std::max(MatchPos + MatchLen, LastPos);
1191 // A check prefix must contain only alphanumeric, hyphens and underscores.
1192 static bool ValidateCheckPrefix(StringRef CheckPrefix) {
1193 Regex Validator("^[a-zA-Z0-9_-]*$");
1194 return Validator.match(CheckPrefix);
1197 static bool ValidateCheckPrefixes() {
1198 StringSet<> PrefixSet;
1200 for (prefix_iterator I = CheckPrefixes.begin(), E = CheckPrefixes.end();
1202 StringRef Prefix(*I);
1204 if (!PrefixSet.insert(Prefix))
1207 if (!ValidateCheckPrefix(Prefix))
1214 // I don't think there's a way to specify an initial value for cl::list,
1215 // so if nothing was specified, add the default
1216 static void AddCheckPrefixIfNeeded() {
1217 if (CheckPrefixes.empty())
1218 CheckPrefixes.push_back("CHECK");
1221 int main(int argc, char **argv) {
1222 sys::PrintStackTraceOnErrorSignal();
1223 PrettyStackTraceProgram X(argc, argv);
1224 cl::ParseCommandLineOptions(argc, argv);
1226 if (!ValidateCheckPrefixes()) {
1227 errs() << "Supplied check-prefix is invalid! Prefixes must be unique and "
1228 "start with a letter and contain only alphanumeric characters, "
1229 "hyphens and underscores\n";
1233 AddCheckPrefixIfNeeded();
1237 // Read the expected strings from the check file.
1238 std::vector<CheckString> CheckStrings;
1239 if (ReadCheckFile(SM, CheckStrings))
1242 // Open the file to check and add it to SourceMgr.
1243 OwningPtr<MemoryBuffer> File;
1245 MemoryBuffer::getFileOrSTDIN(InputFilename, File)) {
1246 errs() << "Could not open input file '" << InputFilename << "': "
1247 << ec.message() << '\n';
1251 if (File->getBufferSize() == 0) {
1252 errs() << "FileCheck error: '" << InputFilename << "' is empty.\n";
1256 // Remove duplicate spaces in the input file if requested.
1257 // Remove DOS style line endings.
1259 CanonicalizeInputFile(File.take(), NoCanonicalizeWhiteSpace);
1261 SM.AddNewSourceBuffer(F, SMLoc());
1263 /// VariableTable - This holds all the current filecheck variables.
1264 StringMap<StringRef> VariableTable;
1266 // Check that we have all of the expected strings, in order, in the input
1268 StringRef Buffer = F->getBuffer();
1270 bool hasError = false;
1272 unsigned i = 0, j = 0, e = CheckStrings.size();
1275 StringRef CheckRegion;
1277 CheckRegion = Buffer;
1279 const CheckString &CheckLabelStr = CheckStrings[j];
1280 if (CheckLabelStr.CheckTy != Check::CheckLabel) {
1285 // Scan to next CHECK-LABEL match, ignoring CHECK-NOT and CHECK-DAG
1286 size_t MatchLabelLen = 0;
1287 size_t MatchLabelPos = CheckLabelStr.Check(SM, Buffer, true,
1288 MatchLabelLen, VariableTable);
1289 if (MatchLabelPos == StringRef::npos) {
1294 CheckRegion = Buffer.substr(0, MatchLabelPos + MatchLabelLen);
1295 Buffer = Buffer.substr(MatchLabelPos + MatchLabelLen);
1299 for ( ; i != j; ++i) {
1300 const CheckString &CheckStr = CheckStrings[i];
1302 // Check each string within the scanned region, including a second check
1303 // of any final CHECK-LABEL (to verify CHECK-NOT and CHECK-DAG)
1304 size_t MatchLen = 0;
1305 size_t MatchPos = CheckStr.Check(SM, CheckRegion, false, MatchLen,
1308 if (MatchPos == StringRef::npos) {
1314 CheckRegion = CheckRegion.substr(MatchPos + MatchLen);
1321 return hasError ? 1 : 0;