1 //===- FileCheck.cpp - Check that File's Contents match what is expected --===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // FileCheck does a line-by line check of a file that validates whether it
11 // contains the expected content. This is useful for regression tests etc.
13 // This program exits with an error status of 2 on error, exit status of 0 if
14 // the file matched the expected contents, and exit status of 1 if it did not
15 // contain the expected contents.
17 //===----------------------------------------------------------------------===//
19 #include "llvm/ADT/OwningPtr.h"
20 #include "llvm/ADT/SmallString.h"
21 #include "llvm/ADT/StringExtras.h"
22 #include "llvm/ADT/StringMap.h"
23 #include "llvm/Support/CommandLine.h"
24 #include "llvm/Support/MemoryBuffer.h"
25 #include "llvm/Support/PrettyStackTrace.h"
26 #include "llvm/Support/Regex.h"
27 #include "llvm/Support/Signals.h"
28 #include "llvm/Support/SourceMgr.h"
29 #include "llvm/Support/raw_ostream.h"
30 #include "llvm/Support/system_error.h"
37 static cl::opt<std::string>
38 CheckFilename(cl::Positional, cl::desc("<check-file>"), cl::Required);
40 static cl::opt<std::string>
41 InputFilename("input-file", cl::desc("File to check (defaults to stdin)"),
42 cl::init("-"), cl::value_desc("filename"));
44 static cl::opt<std::string>
45 CheckPrefix("check-prefix", cl::init("CHECK"),
46 cl::desc("Prefix to use from check file (defaults to 'CHECK')"));
49 NoCanonicalizeWhiteSpace("strict-whitespace",
50 cl::desc("Do not treat all horizontal whitespace as equivalent"));
52 //===----------------------------------------------------------------------===//
53 // Pattern Handling Code.
54 //===----------------------------------------------------------------------===//
59 /// MatchEOF - When set, this pattern only matches the end of file. This is
60 /// used for trailing CHECK-NOTs.
69 /// FixedStr - If non-empty, this pattern is a fixed string match with the
70 /// specified fixed string.
73 /// RegEx - If non-empty, this is a regex pattern.
76 /// \brief Contains the number of line this pattern is in.
79 /// VariableUses - Entries in this vector map to uses of a variable in the
80 /// pattern, e.g. "foo[[bar]]baz". In this case, the RegExStr will contain
81 /// "foobaz" and we'll get an entry in this vector that tells us to insert the
82 /// value of bar at offset 3.
83 std::vector<std::pair<StringRef, unsigned> > VariableUses;
85 /// VariableDefs - Maps definitions of variables to their parenthesized
87 /// E.g. for the pattern "foo[[bar:.*]]baz", VariableDefs will map "bar" to 1.
88 std::map<StringRef, unsigned> VariableDefs;
92 Pattern(bool matchEOF = false)
93 : MatchEOF(matchEOF), MatchNot(false), MatchDag(false) { }
95 /// getLoc - Return the location in source code.
96 SMLoc getLoc() const { return PatternLoc; }
98 /// ParsePattern - Parse the given string into the Pattern. SM provides the
99 /// SourceMgr used for error reports, and LineNumber is the line number in
100 /// the input file from which the pattern string was read.
101 /// Returns true in case of an error, false otherwise.
102 bool ParsePattern(StringRef PatternStr, SourceMgr &SM, unsigned LineNumber);
104 /// Match - Match the pattern string against the input buffer Buffer. This
105 /// returns the position that is matched or npos if there is no match. If
106 /// there is a match, the size of the matched string is returned in MatchLen.
108 /// The VariableTable StringMap provides the current values of filecheck
109 /// variables and is updated if this match defines new values.
110 size_t Match(StringRef Buffer, size_t &MatchLen,
111 StringMap<StringRef> &VariableTable) const;
113 /// PrintFailureInfo - Print additional information about a failure to match
114 /// involving this pattern.
115 void PrintFailureInfo(const SourceMgr &SM, StringRef Buffer,
116 const StringMap<StringRef> &VariableTable) const;
118 void setMatchNot(bool Not) { MatchNot = Not; }
119 bool getMatchNot() const { return MatchNot; }
121 void setMatchDag(bool Dag) { MatchDag = Dag; }
122 bool getMatchDag() const { return MatchDag; }
125 static void AddFixedStringToRegEx(StringRef FixedStr, std::string &TheStr);
126 bool AddRegExToRegEx(StringRef RS, unsigned &CurParen, SourceMgr &SM);
127 void AddBackrefToRegEx(unsigned BackrefNum);
129 /// ComputeMatchDistance - Compute an arbitrary estimate for the quality of
130 /// matching this pattern at the start of \arg Buffer; a distance of zero
131 /// should correspond to a perfect match.
132 unsigned ComputeMatchDistance(StringRef Buffer,
133 const StringMap<StringRef> &VariableTable) const;
135 /// \brief Evaluates expression and stores the result to \p Value.
136 /// \return true on success. false when the expression has invalid syntax.
137 bool EvaluateExpression(StringRef Expr, std::string &Value) const;
139 /// \brief Finds the closing sequence of a regex variable usage or
140 /// definition. Str has to point in the beginning of the definition
141 /// (right after the opening sequence).
142 /// \return offset of the closing sequence within Str, or npos if it was not
144 size_t FindRegexVarEnd(StringRef Str);
148 bool Pattern::ParsePattern(StringRef PatternStr, SourceMgr &SM,
149 unsigned LineNumber) {
150 this->LineNumber = LineNumber;
151 PatternLoc = SMLoc::getFromPointer(PatternStr.data());
153 // Ignore trailing whitespace.
154 while (!PatternStr.empty() &&
155 (PatternStr.back() == ' ' || PatternStr.back() == '\t'))
156 PatternStr = PatternStr.substr(0, PatternStr.size()-1);
158 // Check that there is something on the line.
159 if (PatternStr.empty()) {
160 SM.PrintMessage(PatternLoc, SourceMgr::DK_Error,
161 "found empty check string with prefix '" +
166 // Check to see if this is a fixed string, or if it has regex pieces.
167 if (PatternStr.size() < 2 ||
168 (PatternStr.find("{{") == StringRef::npos &&
169 PatternStr.find("[[") == StringRef::npos)) {
170 FixedStr = PatternStr;
174 // Paren value #0 is for the fully matched string. Any new parenthesized
175 // values add from there.
176 unsigned CurParen = 1;
178 // Otherwise, there is at least one regex piece. Build up the regex pattern
179 // by escaping scary characters in fixed strings, building up one big regex.
180 while (!PatternStr.empty()) {
182 if (PatternStr.startswith("{{")) {
183 // This is the start of a regex match. Scan for the }}.
184 size_t End = PatternStr.find("}}");
185 if (End == StringRef::npos) {
186 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
188 "found start of regex string with no end '}}'");
192 // Enclose {{}} patterns in parens just like [[]] even though we're not
193 // capturing the result for any purpose. This is required in case the
194 // expression contains an alternation like: CHECK: abc{{x|z}}def. We
195 // want this to turn into: "abc(x|z)def" not "abcx|zdef".
199 if (AddRegExToRegEx(PatternStr.substr(2, End-2), CurParen, SM))
203 PatternStr = PatternStr.substr(End+2);
207 // Named RegEx matches. These are of two forms: [[foo:.*]] which matches .*
208 // (or some other regex) and assigns it to the FileCheck variable 'foo'. The
209 // second form is [[foo]] which is a reference to foo. The variable name
210 // itself must be of the form "[a-zA-Z_][0-9a-zA-Z_]*", otherwise we reject
211 // it. This is to catch some common errors.
212 if (PatternStr.startswith("[[")) {
213 // Find the closing bracket pair ending the match. End is going to be an
214 // offset relative to the beginning of the match string.
215 size_t End = FindRegexVarEnd(PatternStr.substr(2));
217 if (End == StringRef::npos) {
218 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
220 "invalid named regex reference, no ]] found");
224 StringRef MatchStr = PatternStr.substr(2, End);
225 PatternStr = PatternStr.substr(End+4);
227 // Get the regex name (e.g. "foo").
228 size_t NameEnd = MatchStr.find(':');
229 StringRef Name = MatchStr.substr(0, NameEnd);
232 SM.PrintMessage(SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
233 "invalid name in named regex: empty name");
237 // Verify that the name/expression is well formed. FileCheck currently
238 // supports @LINE, @LINE+number, @LINE-number expressions. The check here
239 // is relaxed, more strict check is performed in \c EvaluateExpression.
240 bool IsExpression = false;
241 for (unsigned i = 0, e = Name.size(); i != e; ++i) {
242 if (i == 0 && Name[i] == '@') {
243 if (NameEnd != StringRef::npos) {
244 SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
246 "invalid name in named regex definition");
252 if (Name[i] != '_' && !isalnum(Name[i]) &&
253 (!IsExpression || (Name[i] != '+' && Name[i] != '-'))) {
254 SM.PrintMessage(SMLoc::getFromPointer(Name.data()+i),
255 SourceMgr::DK_Error, "invalid name in named regex");
260 // Name can't start with a digit.
261 if (isdigit(static_cast<unsigned char>(Name[0]))) {
262 SM.PrintMessage(SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
263 "invalid name in named regex");
268 if (NameEnd == StringRef::npos) {
269 // Handle variables that were defined earlier on the same line by
270 // emitting a backreference.
271 if (VariableDefs.find(Name) != VariableDefs.end()) {
272 unsigned VarParenNum = VariableDefs[Name];
273 if (VarParenNum < 1 || VarParenNum > 9) {
274 SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
276 "Can't back-reference more than 9 variables");
279 AddBackrefToRegEx(VarParenNum);
281 VariableUses.push_back(std::make_pair(Name, RegExStr.size()));
286 // Handle [[foo:.*]].
287 VariableDefs[Name] = CurParen;
291 if (AddRegExToRegEx(MatchStr.substr(NameEnd+1), CurParen, SM))
297 // Handle fixed string matches.
298 // Find the end, which is the start of the next regex.
299 size_t FixedMatchEnd = PatternStr.find("{{");
300 FixedMatchEnd = std::min(FixedMatchEnd, PatternStr.find("[["));
301 AddFixedStringToRegEx(PatternStr.substr(0, FixedMatchEnd), RegExStr);
302 PatternStr = PatternStr.substr(FixedMatchEnd);
308 void Pattern::AddFixedStringToRegEx(StringRef FixedStr, std::string &TheStr) {
309 // Add the characters from FixedStr to the regex, escaping as needed. This
310 // avoids "leaning toothpicks" in common patterns.
311 for (unsigned i = 0, e = FixedStr.size(); i != e; ++i) {
312 switch (FixedStr[i]) {
313 // These are the special characters matched in "p_ere_exp".
329 TheStr += FixedStr[i];
335 bool Pattern::AddRegExToRegEx(StringRef RS, unsigned &CurParen,
339 if (!R.isValid(Error)) {
340 SM.PrintMessage(SMLoc::getFromPointer(RS.data()), SourceMgr::DK_Error,
341 "invalid regex: " + Error);
345 RegExStr += RS.str();
346 CurParen += R.getNumMatches();
350 void Pattern::AddBackrefToRegEx(unsigned BackrefNum) {
351 assert(BackrefNum >= 1 && BackrefNum <= 9 && "Invalid backref number");
352 std::string Backref = std::string("\\") +
353 std::string(1, '0' + BackrefNum);
357 bool Pattern::EvaluateExpression(StringRef Expr, std::string &Value) const {
358 // The only supported expression is @LINE([\+-]\d+)?
359 if (!Expr.startswith("@LINE"))
361 Expr = Expr.substr(StringRef("@LINE").size());
365 Expr = Expr.substr(1);
366 else if (Expr[0] != '-')
368 if (Expr.getAsInteger(10, Offset))
371 Value = llvm::itostr(LineNumber + Offset);
375 /// Match - Match the pattern string against the input buffer Buffer. This
376 /// returns the position that is matched or npos if there is no match. If
377 /// there is a match, the size of the matched string is returned in MatchLen.
378 size_t Pattern::Match(StringRef Buffer, size_t &MatchLen,
379 StringMap<StringRef> &VariableTable) const {
380 // If this is the EOF pattern, match it immediately.
383 return Buffer.size();
386 // If this is a fixed string pattern, just match it now.
387 if (!FixedStr.empty()) {
388 MatchLen = FixedStr.size();
389 return Buffer.find(FixedStr);
394 // If there are variable uses, we need to create a temporary string with the
396 StringRef RegExToMatch = RegExStr;
398 if (!VariableUses.empty()) {
401 unsigned InsertOffset = 0;
402 for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
405 if (VariableUses[i].first[0] == '@') {
406 if (!EvaluateExpression(VariableUses[i].first, Value))
407 return StringRef::npos;
409 StringMap<StringRef>::iterator it =
410 VariableTable.find(VariableUses[i].first);
411 // If the variable is undefined, return an error.
412 if (it == VariableTable.end())
413 return StringRef::npos;
415 // Look up the value and escape it so that we can plop it into the regex.
416 AddFixedStringToRegEx(it->second, Value);
419 // Plop it into the regex at the adjusted offset.
420 TmpStr.insert(TmpStr.begin()+VariableUses[i].second+InsertOffset,
421 Value.begin(), Value.end());
422 InsertOffset += Value.size();
425 // Match the newly constructed regex.
426 RegExToMatch = TmpStr;
430 SmallVector<StringRef, 4> MatchInfo;
431 if (!Regex(RegExToMatch, Regex::Newline).match(Buffer, &MatchInfo))
432 return StringRef::npos;
434 // Successful regex match.
435 assert(!MatchInfo.empty() && "Didn't get any match");
436 StringRef FullMatch = MatchInfo[0];
438 // If this defines any variables, remember their values.
439 for (std::map<StringRef, unsigned>::const_iterator I = VariableDefs.begin(),
440 E = VariableDefs.end();
442 assert(I->second < MatchInfo.size() && "Internal paren error");
443 VariableTable[I->first] = MatchInfo[I->second];
446 MatchLen = FullMatch.size();
447 return FullMatch.data()-Buffer.data();
450 unsigned Pattern::ComputeMatchDistance(StringRef Buffer,
451 const StringMap<StringRef> &VariableTable) const {
452 // Just compute the number of matching characters. For regular expressions, we
453 // just compare against the regex itself and hope for the best.
455 // FIXME: One easy improvement here is have the regex lib generate a single
456 // example regular expression which matches, and use that as the example
458 StringRef ExampleString(FixedStr);
459 if (ExampleString.empty())
460 ExampleString = RegExStr;
462 // Only compare up to the first line in the buffer, or the string size.
463 StringRef BufferPrefix = Buffer.substr(0, ExampleString.size());
464 BufferPrefix = BufferPrefix.split('\n').first;
465 return BufferPrefix.edit_distance(ExampleString);
468 void Pattern::PrintFailureInfo(const SourceMgr &SM, StringRef Buffer,
469 const StringMap<StringRef> &VariableTable) const{
470 // If this was a regular expression using variables, print the current
472 if (!VariableUses.empty()) {
473 for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
474 SmallString<256> Msg;
475 raw_svector_ostream OS(Msg);
476 StringRef Var = VariableUses[i].first;
479 if (EvaluateExpression(Var, Value)) {
480 OS << "with expression \"";
481 OS.write_escaped(Var) << "\" equal to \"";
482 OS.write_escaped(Value) << "\"";
484 OS << "uses incorrect expression \"";
485 OS.write_escaped(Var) << "\"";
488 StringMap<StringRef>::const_iterator it = VariableTable.find(Var);
490 // Check for undefined variable references.
491 if (it == VariableTable.end()) {
492 OS << "uses undefined variable \"";
493 OS.write_escaped(Var) << "\"";
495 OS << "with variable \"";
496 OS.write_escaped(Var) << "\" equal to \"";
497 OS.write_escaped(it->second) << "\"";
501 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
506 // Attempt to find the closest/best fuzzy match. Usually an error happens
507 // because some string in the output didn't exactly match. In these cases, we
508 // would like to show the user a best guess at what "should have" matched, to
509 // save them having to actually check the input manually.
510 size_t NumLinesForward = 0;
511 size_t Best = StringRef::npos;
512 double BestQuality = 0;
514 // Use an arbitrary 4k limit on how far we will search.
515 for (size_t i = 0, e = std::min(size_t(4096), Buffer.size()); i != e; ++i) {
516 if (Buffer[i] == '\n')
519 // Patterns have leading whitespace stripped, so skip whitespace when
520 // looking for something which looks like a pattern.
521 if (Buffer[i] == ' ' || Buffer[i] == '\t')
524 // Compute the "quality" of this match as an arbitrary combination of the
525 // match distance and the number of lines skipped to get to this match.
526 unsigned Distance = ComputeMatchDistance(Buffer.substr(i), VariableTable);
527 double Quality = Distance + (NumLinesForward / 100.);
529 if (Quality < BestQuality || Best == StringRef::npos) {
531 BestQuality = Quality;
535 // Print the "possible intended match here" line if we found something
536 // reasonable and not equal to what we showed in the "scanning from here"
538 if (Best && Best != StringRef::npos && BestQuality < 50) {
539 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + Best),
540 SourceMgr::DK_Note, "possible intended match here");
542 // FIXME: If we wanted to be really friendly we would show why the match
543 // failed, as it can be hard to spot simple one character differences.
547 size_t Pattern::FindRegexVarEnd(StringRef Str) {
548 // Offset keeps track of the current offset within the input Str
550 // [...] Nesting depth
551 size_t BracketDepth = 0;
553 while (!Str.empty()) {
554 if (Str.startswith("]]") && BracketDepth == 0)
556 if (Str[0] == '\\') {
557 // Backslash escapes the next char within regexes, so skip them both.
568 assert(BracketDepth > 0 && "Invalid regex");
577 return StringRef::npos;
581 //===----------------------------------------------------------------------===//
583 //===----------------------------------------------------------------------===//
585 /// CheckString - This is a check that we found in the input file.
587 /// Pat - The pattern to match.
590 /// Loc - The location in the match file that the check string was specified.
593 /// IsCheckNext - This is true if this is a CHECK-NEXT: directive (as opposed
594 /// to a CHECK: directive.
597 /// DagNotStrings - These are all of the strings that are disallowed from
598 /// occurring between this match string and the previous one (or start of
600 std::vector<Pattern> DagNotStrings;
602 CheckString(const Pattern &P, SMLoc L, bool isCheckNext)
603 : Pat(P), Loc(L), IsCheckNext(isCheckNext) {}
605 /// Check - Match check string and its "not strings" and/or "dag strings".
606 size_t Check(const SourceMgr &SM, StringRef Buffer, size_t &MatchLen,
607 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());
663 /// ReadCheckFile - Read the check file, which specifies the sequence of
664 /// expected strings. The strings are added to the CheckStrings vector.
665 /// Returns true in case of an error, false otherwise.
666 static bool ReadCheckFile(SourceMgr &SM,
667 std::vector<CheckString> &CheckStrings) {
668 OwningPtr<MemoryBuffer> File;
670 MemoryBuffer::getFileOrSTDIN(CheckFilename, File)) {
671 errs() << "Could not open check file '" << CheckFilename << "': "
672 << ec.message() << '\n';
676 // If we want to canonicalize whitespace, strip excess whitespace from the
677 // buffer containing the CHECK lines. Remove DOS style line endings.
679 CanonicalizeInputFile(File.take(), NoCanonicalizeWhiteSpace);
681 SM.AddNewSourceBuffer(F, SMLoc());
683 // Find all instances of CheckPrefix followed by : in the file.
684 StringRef Buffer = F->getBuffer();
685 std::vector<Pattern> DagNotMatches;
687 // LineNumber keeps track of the line on which CheckPrefix instances are
689 unsigned LineNumber = 1;
692 // See if Prefix occurs in the memory buffer.
693 size_t PrefixLoc = Buffer.find(CheckPrefix);
694 // If we didn't find a match, we're done.
695 if (PrefixLoc == StringRef::npos)
698 LineNumber += Buffer.substr(0, PrefixLoc).count('\n');
700 Buffer = Buffer.substr(PrefixLoc);
702 const char *CheckPrefixStart = Buffer.data();
704 // When we find a check prefix, keep track of whether we find CHECK: or
706 bool IsCheckNext = false, IsCheckNot = false, IsCheckDag = false;
708 // Verify that the : is present after the prefix.
709 if (Buffer[CheckPrefix.size()] == ':') {
710 Buffer = Buffer.substr(CheckPrefix.size()+1);
711 } else if (Buffer.size() > CheckPrefix.size()+6 &&
712 memcmp(Buffer.data()+CheckPrefix.size(), "-NEXT:", 6) == 0) {
713 Buffer = Buffer.substr(CheckPrefix.size()+6);
715 } else if (Buffer.size() > CheckPrefix.size()+5 &&
716 memcmp(Buffer.data()+CheckPrefix.size(), "-NOT:", 5) == 0) {
717 Buffer = Buffer.substr(CheckPrefix.size()+5);
719 } else if (Buffer.size() > CheckPrefix.size()+5 &&
720 memcmp(Buffer.data()+CheckPrefix.size(), "-DAG:", 5) == 0) {
721 Buffer = Buffer.substr(CheckPrefix.size()+5);
724 Buffer = Buffer.substr(1);
728 // Okay, we found the prefix, yay. Remember the rest of the line, but
729 // ignore leading and trailing whitespace.
730 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t"));
732 // Scan ahead to the end of line.
733 size_t EOL = Buffer.find_first_of("\n\r");
735 // Remember the location of the start of the pattern, for diagnostics.
736 SMLoc PatternLoc = SMLoc::getFromPointer(Buffer.data());
738 // Parse the pattern.
740 if (P.ParsePattern(Buffer.substr(0, EOL), SM, LineNumber))
743 P.setMatchNot(IsCheckNot);
744 P.setMatchDag(IsCheckDag);
746 Buffer = Buffer.substr(EOL);
748 // Verify that CHECK-NEXT lines have at least one CHECK line before them.
749 if (IsCheckNext && CheckStrings.empty()) {
750 SM.PrintMessage(SMLoc::getFromPointer(CheckPrefixStart),
752 "found '"+CheckPrefix+"-NEXT:' without previous '"+
753 CheckPrefix+ ": line");
757 // Handle CHECK-DAG/-NOT.
758 if (IsCheckDag || IsCheckNot) {
759 DagNotMatches.push_back(P);
763 // Okay, add the string we captured to the output vector and move on.
764 CheckStrings.push_back(CheckString(P,
767 std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
770 // Add an EOF pattern for any trailing CHECK-DAG/-NOTs.
771 if (!DagNotMatches.empty()) {
772 CheckStrings.push_back(CheckString(Pattern(true),
773 SMLoc::getFromPointer(Buffer.data()),
775 std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
778 if (CheckStrings.empty()) {
779 errs() << "error: no check strings found with prefix '" << CheckPrefix
787 static void PrintCheckFailed(const SourceMgr &SM, const SMLoc &Loc,
788 const Pattern &Pat, StringRef Buffer,
789 StringMap<StringRef> &VariableTable) {
790 // Otherwise, we have an error, emit an error message.
791 SM.PrintMessage(Loc, SourceMgr::DK_Error,
792 "expected string not found in input");
794 // Print the "scanning from here" line. If the current position is at the
795 // end of a line, advance to the start of the next line.
796 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t\n\r"));
798 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
799 "scanning from here");
801 // Allow the pattern to print additional information if desired.
802 Pat.PrintFailureInfo(SM, Buffer, VariableTable);
805 static void PrintCheckFailed(const SourceMgr &SM, const CheckString &CheckStr,
807 StringMap<StringRef> &VariableTable) {
808 PrintCheckFailed(SM, CheckStr.Loc, CheckStr.Pat, Buffer, VariableTable);
811 /// CountNumNewlinesBetween - Count the number of newlines in the specified
813 static unsigned CountNumNewlinesBetween(StringRef Range) {
814 unsigned NumNewLines = 0;
817 Range = Range.substr(Range.find_first_of("\n\r"));
818 if (Range.empty()) return NumNewLines;
822 // Handle \n\r and \r\n as a single newline.
823 if (Range.size() > 1 &&
824 (Range[1] == '\n' || Range[1] == '\r') &&
825 (Range[0] != Range[1]))
826 Range = Range.substr(1);
827 Range = Range.substr(1);
831 size_t CheckString::Check(const SourceMgr &SM, StringRef Buffer,
833 StringMap<StringRef> &VariableTable) const {
835 std::vector<const Pattern *> NotStrings;
837 // Match "dag strings" (with mixed "not strings" if any).
838 LastPos = CheckDag(SM, Buffer, NotStrings, VariableTable);
839 if (LastPos == StringRef::npos)
840 return StringRef::npos;
842 // Match itself from the last position after matching CHECK-DAG.
843 StringRef MatchBuffer = Buffer.substr(LastPos);
844 size_t MatchPos = Pat.Match(MatchBuffer, MatchLen, VariableTable);
845 if (MatchPos == StringRef::npos) {
846 PrintCheckFailed(SM, *this, MatchBuffer, VariableTable);
847 return StringRef::npos;
851 StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
853 // If this check is a "CHECK-NEXT", verify that the previous match was on
854 // the previous line (i.e. that there is one newline between them).
855 if (CheckNext(SM, SkippedRegion))
856 return StringRef::npos;
858 // If this match had "not strings", verify that they don't exist in the
860 if (CheckNot(SM, SkippedRegion, NotStrings, VariableTable))
861 return StringRef::npos;
866 bool CheckString::CheckNext(const SourceMgr &SM, StringRef Buffer) const {
870 // Count the number of newlines between the previous match and this one.
871 assert(Buffer.data() !=
873 SM.FindBufferContainingLoc(
874 SMLoc::getFromPointer(Buffer.data())))->getBufferStart() &&
875 "CHECK-NEXT can't be the first check in a file");
877 unsigned NumNewLines = CountNumNewlinesBetween(Buffer);
879 if (NumNewLines == 0) {
880 SM.PrintMessage(Loc, SourceMgr::DK_Error, CheckPrefix+
881 "-NEXT: is on the same line as previous match");
882 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()),
883 SourceMgr::DK_Note, "'next' match was here");
884 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
885 "previous match ended here");
889 if (NumNewLines != 1) {
890 SM.PrintMessage(Loc, SourceMgr::DK_Error, CheckPrefix+
891 "-NEXT: is not on the line after the previous match");
892 SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()),
893 SourceMgr::DK_Note, "'next' match was here");
894 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
895 "previous match ended here");
902 bool CheckString::CheckNot(const SourceMgr &SM, StringRef Buffer,
903 const std::vector<const Pattern *> &NotStrings,
904 StringMap<StringRef> &VariableTable) const {
905 for (unsigned ChunkNo = 0, e = NotStrings.size();
906 ChunkNo != e; ++ChunkNo) {
907 const Pattern *Pat = NotStrings[ChunkNo];
908 assert(Pat->getMatchNot() && "Expect CHECK-NOT!");
911 size_t Pos = Pat->Match(Buffer, MatchLen, VariableTable);
913 if (Pos == StringRef::npos) continue;
915 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()+Pos),
917 CheckPrefix+"-NOT: string occurred!");
918 SM.PrintMessage(Pat->getLoc(), SourceMgr::DK_Note,
919 CheckPrefix+"-NOT: pattern specified here");
926 size_t CheckString::CheckDag(const SourceMgr &SM, StringRef Buffer,
927 std::vector<const Pattern *> &NotStrings,
928 StringMap<StringRef> &VariableTable) const {
929 if (DagNotStrings.empty())
933 size_t StartPos = LastPos;
935 for (unsigned ChunkNo = 0, e = DagNotStrings.size();
936 ChunkNo != e; ++ChunkNo) {
937 const Pattern &Pat = DagNotStrings[ChunkNo];
939 assert((Pat.getMatchDag() ^ Pat.getMatchNot()) &&
940 "Invalid CHECK-DAG or CHECK-NOT!");
942 if (Pat.getMatchNot()) {
943 NotStrings.push_back(&Pat);
947 assert(Pat.getMatchDag() && "Expect CHECK-DAG!");
949 size_t MatchLen = 0, MatchPos;
951 // CHECK-DAG always matches from the start.
952 StringRef MatchBuffer = Buffer.substr(StartPos);
953 MatchPos = Pat.Match(MatchBuffer, MatchLen, VariableTable);
954 // With a group of CHECK-DAGs, a single mismatching means the match on
955 // that group of CHECK-DAGs fails immediately.
956 if (MatchPos == StringRef::npos) {
957 PrintCheckFailed(SM, Pat.getLoc(), Pat, MatchBuffer, VariableTable);
958 return StringRef::npos;
960 // Re-calc it as the offset relative to the start of the original string.
961 MatchPos += StartPos;
963 if (!NotStrings.empty()) {
964 if (MatchPos < LastPos) {
966 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + MatchPos),
968 CheckPrefix+"-DAG: found a match of CHECK-DAG"
969 " reordering across a CHECK-NOT");
970 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + LastPos),
972 CheckPrefix+"-DAG: the farthest match of CHECK-DAG"
974 SM.PrintMessage(NotStrings[0]->getLoc(), SourceMgr::DK_Note,
975 CheckPrefix+"-NOT: the crossed pattern specified"
977 SM.PrintMessage(Pat.getLoc(), SourceMgr::DK_Note,
978 CheckPrefix+"-DAG: the reordered pattern specified"
980 return StringRef::npos;
982 // All subsequent CHECK-DAGs should be matched from the farthest
983 // position of all precedent CHECK-DAGs (including this one.)
985 // If there's CHECK-NOTs between two CHECK-DAGs or from CHECK to
986 // CHECK-DAG, verify that there's no 'not' strings occurred in that
988 StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
989 size_t Pos = CheckNot(SM, SkippedRegion, NotStrings, VariableTable);
990 if (Pos != StringRef::npos)
991 return StringRef::npos;
992 // Clear "not strings".
996 // Update the last position with CHECK-DAG matches.
997 LastPos = std::max(MatchPos + MatchLen, LastPos);
1003 int main(int argc, char **argv) {
1004 sys::PrintStackTraceOnErrorSignal();
1005 PrettyStackTraceProgram X(argc, argv);
1006 cl::ParseCommandLineOptions(argc, argv);
1010 // Read the expected strings from the check file.
1011 std::vector<CheckString> CheckStrings;
1012 if (ReadCheckFile(SM, CheckStrings))
1015 // Open the file to check and add it to SourceMgr.
1016 OwningPtr<MemoryBuffer> File;
1018 MemoryBuffer::getFileOrSTDIN(InputFilename, File)) {
1019 errs() << "Could not open input file '" << InputFilename << "': "
1020 << ec.message() << '\n';
1024 if (File->getBufferSize() == 0) {
1025 errs() << "FileCheck error: '" << InputFilename << "' is empty.\n";
1029 // Remove duplicate spaces in the input file if requested.
1030 // Remove DOS style line endings.
1032 CanonicalizeInputFile(File.take(), NoCanonicalizeWhiteSpace);
1034 SM.AddNewSourceBuffer(F, SMLoc());
1036 /// VariableTable - This holds all the current filecheck variables.
1037 StringMap<StringRef> VariableTable;
1039 // Check that we have all of the expected strings, in order, in the input
1041 StringRef Buffer = F->getBuffer();
1043 for (unsigned StrNo = 0, e = CheckStrings.size(); StrNo != e; ++StrNo) {
1044 const CheckString &CheckStr = CheckStrings[StrNo];
1046 // Find StrNo in the file.
1047 size_t MatchLen = 0;
1048 size_t MatchPos = CheckStr.Check(SM, Buffer, MatchLen, VariableTable);
1050 if (MatchPos == StringRef::npos)
1053 Buffer = Buffer.substr(MatchPos + MatchLen);