• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 //===- FileCheck.cpp - Check that File's Contents match what is expected --===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
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.
12 //
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.
16 //
17 //===----------------------------------------------------------------------===//
18 
19 #include "llvm/ADT/SmallString.h"
20 #include "llvm/ADT/StringExtras.h"
21 #include "llvm/ADT/StringMap.h"
22 #include "llvm/ADT/StringSet.h"
23 #include "llvm/Support/CommandLine.h"
24 #include "llvm/Support/MemoryBuffer.h"
25 #include "llvm/Support/PrettyStackTrace.h"
26 #include "llvm/Support/Regex.h"
27 #include "llvm/Support/Signals.h"
28 #include "llvm/Support/SourceMgr.h"
29 #include "llvm/Support/raw_ostream.h"
30 #include <algorithm>
31 #include <cctype>
32 #include <map>
33 #include <string>
34 #include <system_error>
35 #include <vector>
36 using namespace llvm;
37 
38 static cl::opt<std::string>
39 CheckFilename(cl::Positional, cl::desc("<check-file>"), cl::Required);
40 
41 static cl::opt<std::string>
42 InputFilename("input-file", cl::desc("File to check (defaults to stdin)"),
43               cl::init("-"), cl::value_desc("filename"));
44 
45 static cl::list<std::string>
46 CheckPrefixes("check-prefix",
47               cl::desc("Prefix to use from check file (defaults to 'CHECK')"));
48 
49 static cl::opt<bool>
50 NoCanonicalizeWhiteSpace("strict-whitespace",
51               cl::desc("Do not treat all horizontal whitespace as equivalent"));
52 
53 typedef cl::list<std::string>::const_iterator prefix_iterator;
54 
55 //===----------------------------------------------------------------------===//
56 // Pattern Handling Code.
57 //===----------------------------------------------------------------------===//
58 
59 namespace Check {
60   enum CheckType {
61     CheckNone = 0,
62     CheckPlain,
63     CheckNext,
64     CheckNot,
65     CheckDAG,
66     CheckLabel,
67 
68     /// MatchEOF - When set, this pattern only matches the end of file. This is
69     /// used for trailing CHECK-NOTs.
70     CheckEOF
71   };
72 }
73 
74 class Pattern {
75   SMLoc PatternLoc;
76 
77   Check::CheckType CheckTy;
78 
79   /// FixedStr - If non-empty, this pattern is a fixed string match with the
80   /// specified fixed string.
81   StringRef FixedStr;
82 
83   /// RegEx - If non-empty, this is a regex pattern.
84   std::string RegExStr;
85 
86   /// \brief Contains the number of line this pattern is in.
87   unsigned LineNumber;
88 
89   /// VariableUses - Entries in this vector map to uses of a variable in the
90   /// pattern, e.g. "foo[[bar]]baz".  In this case, the RegExStr will contain
91   /// "foobaz" and we'll get an entry in this vector that tells us to insert the
92   /// value of bar at offset 3.
93   std::vector<std::pair<StringRef, unsigned> > VariableUses;
94 
95   /// VariableDefs - Maps definitions of variables to their parenthesized
96   /// capture numbers.
97   /// E.g. for the pattern "foo[[bar:.*]]baz", VariableDefs will map "bar" to 1.
98   std::map<StringRef, unsigned> VariableDefs;
99 
100 public:
101 
Pattern(Check::CheckType Ty)102   Pattern(Check::CheckType Ty)
103     : CheckTy(Ty) { }
104 
105   /// getLoc - Return the location in source code.
getLoc() const106   SMLoc getLoc() const { return PatternLoc; }
107 
108   /// ParsePattern - Parse the given string into the Pattern. Prefix provides
109   /// which prefix is being matched, SM provides the SourceMgr used for error
110   /// reports, and LineNumber is the line number in the input file from which
111   /// the pattern string was read.  Returns true in case of an error, false
112   /// otherwise.
113   bool ParsePattern(StringRef PatternStr,
114                     StringRef Prefix,
115                     SourceMgr &SM,
116                     unsigned LineNumber);
117 
118   /// Match - Match the pattern string against the input buffer Buffer.  This
119   /// returns the position that is matched or npos if there is no match.  If
120   /// there is a match, the size of the matched string is returned in MatchLen.
121   ///
122   /// The VariableTable StringMap provides the current values of filecheck
123   /// variables and is updated if this match defines new values.
124   size_t Match(StringRef Buffer, size_t &MatchLen,
125                StringMap<StringRef> &VariableTable) const;
126 
127   /// PrintFailureInfo - Print additional information about a failure to match
128   /// involving this pattern.
129   void PrintFailureInfo(const SourceMgr &SM, StringRef Buffer,
130                         const StringMap<StringRef> &VariableTable) const;
131 
hasVariable() const132   bool hasVariable() const { return !(VariableUses.empty() &&
133                                       VariableDefs.empty()); }
134 
getCheckTy() const135   Check::CheckType getCheckTy() const { return CheckTy; }
136 
137 private:
138   bool AddRegExToRegEx(StringRef RS, unsigned &CurParen, SourceMgr &SM);
139   void AddBackrefToRegEx(unsigned BackrefNum);
140 
141   /// ComputeMatchDistance - Compute an arbitrary estimate for the quality of
142   /// matching this pattern at the start of \arg Buffer; a distance of zero
143   /// should correspond to a perfect match.
144   unsigned ComputeMatchDistance(StringRef Buffer,
145                                const StringMap<StringRef> &VariableTable) const;
146 
147   /// \brief Evaluates expression and stores the result to \p Value.
148   /// \return true on success. false when the expression has invalid syntax.
149   bool EvaluateExpression(StringRef Expr, std::string &Value) const;
150 
151   /// \brief Finds the closing sequence of a regex variable usage or
152   /// definition. Str has to point in the beginning of the definition
153   /// (right after the opening sequence).
154   /// \return offset of the closing sequence within Str, or npos if it was not
155   /// found.
156   size_t FindRegexVarEnd(StringRef Str, SourceMgr &SM);
157 };
158 
159 
ParsePattern(StringRef PatternStr,StringRef Prefix,SourceMgr & SM,unsigned LineNumber)160 bool Pattern::ParsePattern(StringRef PatternStr,
161                            StringRef Prefix,
162                            SourceMgr &SM,
163                            unsigned LineNumber) {
164   this->LineNumber = LineNumber;
165   PatternLoc = SMLoc::getFromPointer(PatternStr.data());
166 
167   // Ignore trailing whitespace.
168   while (!PatternStr.empty() &&
169          (PatternStr.back() == ' ' || PatternStr.back() == '\t'))
170     PatternStr = PatternStr.substr(0, PatternStr.size()-1);
171 
172   // Check that there is something on the line.
173   if (PatternStr.empty()) {
174     SM.PrintMessage(PatternLoc, SourceMgr::DK_Error,
175                     "found empty check string with prefix '" +
176                     Prefix + ":'");
177     return true;
178   }
179 
180   // Check to see if this is a fixed string, or if it has regex pieces.
181   if (PatternStr.size() < 2 ||
182       (PatternStr.find("{{") == StringRef::npos &&
183        PatternStr.find("[[") == StringRef::npos)) {
184     FixedStr = PatternStr;
185     return false;
186   }
187 
188   // Paren value #0 is for the fully matched string.  Any new parenthesized
189   // values add from there.
190   unsigned CurParen = 1;
191 
192   // Otherwise, there is at least one regex piece.  Build up the regex pattern
193   // by escaping scary characters in fixed strings, building up one big regex.
194   while (!PatternStr.empty()) {
195     // RegEx matches.
196     if (PatternStr.startswith("{{")) {
197       // This is the start of a regex match.  Scan for the }}.
198       size_t End = PatternStr.find("}}");
199       if (End == StringRef::npos) {
200         SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
201                         SourceMgr::DK_Error,
202                         "found start of regex string with no end '}}'");
203         return true;
204       }
205 
206       // Enclose {{}} patterns in parens just like [[]] even though we're not
207       // capturing the result for any purpose.  This is required in case the
208       // expression contains an alternation like: CHECK:  abc{{x|z}}def.  We
209       // want this to turn into: "abc(x|z)def" not "abcx|zdef".
210       RegExStr += '(';
211       ++CurParen;
212 
213       if (AddRegExToRegEx(PatternStr.substr(2, End-2), CurParen, SM))
214         return true;
215       RegExStr += ')';
216 
217       PatternStr = PatternStr.substr(End+2);
218       continue;
219     }
220 
221     // Named RegEx matches.  These are of two forms: [[foo:.*]] which matches .*
222     // (or some other regex) and assigns it to the FileCheck variable 'foo'. The
223     // second form is [[foo]] which is a reference to foo.  The variable name
224     // itself must be of the form "[a-zA-Z_][0-9a-zA-Z_]*", otherwise we reject
225     // it.  This is to catch some common errors.
226     if (PatternStr.startswith("[[")) {
227       // Find the closing bracket pair ending the match.  End is going to be an
228       // offset relative to the beginning of the match string.
229       size_t End = FindRegexVarEnd(PatternStr.substr(2), SM);
230 
231       if (End == StringRef::npos) {
232         SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
233                         SourceMgr::DK_Error,
234                         "invalid named regex reference, no ]] found");
235         return true;
236       }
237 
238       StringRef MatchStr = PatternStr.substr(2, End);
239       PatternStr = PatternStr.substr(End+4);
240 
241       // Get the regex name (e.g. "foo").
242       size_t NameEnd = MatchStr.find(':');
243       StringRef Name = MatchStr.substr(0, NameEnd);
244 
245       if (Name.empty()) {
246         SM.PrintMessage(SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
247                         "invalid name in named regex: empty name");
248         return true;
249       }
250 
251       // Verify that the name/expression is well formed. FileCheck currently
252       // supports @LINE, @LINE+number, @LINE-number expressions. The check here
253       // is relaxed, more strict check is performed in \c EvaluateExpression.
254       bool IsExpression = false;
255       for (unsigned i = 0, e = Name.size(); i != e; ++i) {
256         if (i == 0 && Name[i] == '@') {
257           if (NameEnd != StringRef::npos) {
258             SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
259                             SourceMgr::DK_Error,
260                             "invalid name in named regex definition");
261             return true;
262           }
263           IsExpression = true;
264           continue;
265         }
266         if (Name[i] != '_' && !isalnum(Name[i]) &&
267             (!IsExpression || (Name[i] != '+' && Name[i] != '-'))) {
268           SM.PrintMessage(SMLoc::getFromPointer(Name.data()+i),
269                           SourceMgr::DK_Error, "invalid name in named regex");
270           return true;
271         }
272       }
273 
274       // Name can't start with a digit.
275       if (isdigit(static_cast<unsigned char>(Name[0]))) {
276         SM.PrintMessage(SMLoc::getFromPointer(Name.data()), SourceMgr::DK_Error,
277                         "invalid name in named regex");
278         return true;
279       }
280 
281       // Handle [[foo]].
282       if (NameEnd == StringRef::npos) {
283         // Handle variables that were defined earlier on the same line by
284         // emitting a backreference.
285         if (VariableDefs.find(Name) != VariableDefs.end()) {
286           unsigned VarParenNum = VariableDefs[Name];
287           if (VarParenNum < 1 || VarParenNum > 9) {
288             SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
289                             SourceMgr::DK_Error,
290                             "Can't back-reference more than 9 variables");
291             return true;
292           }
293           AddBackrefToRegEx(VarParenNum);
294         } else {
295           VariableUses.push_back(std::make_pair(Name, RegExStr.size()));
296         }
297         continue;
298       }
299 
300       // Handle [[foo:.*]].
301       VariableDefs[Name] = CurParen;
302       RegExStr += '(';
303       ++CurParen;
304 
305       if (AddRegExToRegEx(MatchStr.substr(NameEnd+1), CurParen, SM))
306         return true;
307 
308       RegExStr += ')';
309     }
310 
311     // Handle fixed string matches.
312     // Find the end, which is the start of the next regex.
313     size_t FixedMatchEnd = PatternStr.find("{{");
314     FixedMatchEnd = std::min(FixedMatchEnd, PatternStr.find("[["));
315     RegExStr += Regex::escape(PatternStr.substr(0, FixedMatchEnd));
316     PatternStr = PatternStr.substr(FixedMatchEnd);
317   }
318 
319   return false;
320 }
321 
AddRegExToRegEx(StringRef RS,unsigned & CurParen,SourceMgr & SM)322 bool Pattern::AddRegExToRegEx(StringRef RS, unsigned &CurParen,
323                               SourceMgr &SM) {
324   Regex R(RS);
325   std::string Error;
326   if (!R.isValid(Error)) {
327     SM.PrintMessage(SMLoc::getFromPointer(RS.data()), SourceMgr::DK_Error,
328                     "invalid regex: " + Error);
329     return true;
330   }
331 
332   RegExStr += RS.str();
333   CurParen += R.getNumMatches();
334   return false;
335 }
336 
AddBackrefToRegEx(unsigned BackrefNum)337 void Pattern::AddBackrefToRegEx(unsigned BackrefNum) {
338   assert(BackrefNum >= 1 && BackrefNum <= 9 && "Invalid backref number");
339   std::string Backref = std::string("\\") +
340                         std::string(1, '0' + BackrefNum);
341   RegExStr += Backref;
342 }
343 
EvaluateExpression(StringRef Expr,std::string & Value) const344 bool Pattern::EvaluateExpression(StringRef Expr, std::string &Value) const {
345   // The only supported expression is @LINE([\+-]\d+)?
346   if (!Expr.startswith("@LINE"))
347     return false;
348   Expr = Expr.substr(StringRef("@LINE").size());
349   int Offset = 0;
350   if (!Expr.empty()) {
351     if (Expr[0] == '+')
352       Expr = Expr.substr(1);
353     else if (Expr[0] != '-')
354       return false;
355     if (Expr.getAsInteger(10, Offset))
356       return false;
357   }
358   Value = llvm::itostr(LineNumber + Offset);
359   return true;
360 }
361 
362 /// Match - Match the pattern string against the input buffer Buffer.  This
363 /// returns the position that is matched or npos if there is no match.  If
364 /// there is a match, the size of the matched string is returned in MatchLen.
Match(StringRef Buffer,size_t & MatchLen,StringMap<StringRef> & VariableTable) const365 size_t Pattern::Match(StringRef Buffer, size_t &MatchLen,
366                       StringMap<StringRef> &VariableTable) const {
367   // If this is the EOF pattern, match it immediately.
368   if (CheckTy == Check::CheckEOF) {
369     MatchLen = 0;
370     return Buffer.size();
371   }
372 
373   // If this is a fixed string pattern, just match it now.
374   if (!FixedStr.empty()) {
375     MatchLen = FixedStr.size();
376     return Buffer.find(FixedStr);
377   }
378 
379   // Regex match.
380 
381   // If there are variable uses, we need to create a temporary string with the
382   // actual value.
383   StringRef RegExToMatch = RegExStr;
384   std::string TmpStr;
385   if (!VariableUses.empty()) {
386     TmpStr = RegExStr;
387 
388     unsigned InsertOffset = 0;
389     for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
390       std::string Value;
391 
392       if (VariableUses[i].first[0] == '@') {
393         if (!EvaluateExpression(VariableUses[i].first, Value))
394           return StringRef::npos;
395       } else {
396         StringMap<StringRef>::iterator it =
397           VariableTable.find(VariableUses[i].first);
398         // If the variable is undefined, return an error.
399         if (it == VariableTable.end())
400           return StringRef::npos;
401 
402         // Look up the value and escape it so that we can put it into the regex.
403         Value += Regex::escape(it->second);
404       }
405 
406       // Plop it into the regex at the adjusted offset.
407       TmpStr.insert(TmpStr.begin()+VariableUses[i].second+InsertOffset,
408                     Value.begin(), Value.end());
409       InsertOffset += Value.size();
410     }
411 
412     // Match the newly constructed regex.
413     RegExToMatch = TmpStr;
414   }
415 
416 
417   SmallVector<StringRef, 4> MatchInfo;
418   if (!Regex(RegExToMatch, Regex::Newline).match(Buffer, &MatchInfo))
419     return StringRef::npos;
420 
421   // Successful regex match.
422   assert(!MatchInfo.empty() && "Didn't get any match");
423   StringRef FullMatch = MatchInfo[0];
424 
425   // If this defines any variables, remember their values.
426   for (std::map<StringRef, unsigned>::const_iterator I = VariableDefs.begin(),
427                                                      E = VariableDefs.end();
428        I != E; ++I) {
429     assert(I->second < MatchInfo.size() && "Internal paren error");
430     VariableTable[I->first] = MatchInfo[I->second];
431   }
432 
433   MatchLen = FullMatch.size();
434   return FullMatch.data()-Buffer.data();
435 }
436 
ComputeMatchDistance(StringRef Buffer,const StringMap<StringRef> & VariableTable) const437 unsigned Pattern::ComputeMatchDistance(StringRef Buffer,
438                               const StringMap<StringRef> &VariableTable) const {
439   // Just compute the number of matching characters. For regular expressions, we
440   // just compare against the regex itself and hope for the best.
441   //
442   // FIXME: One easy improvement here is have the regex lib generate a single
443   // example regular expression which matches, and use that as the example
444   // string.
445   StringRef ExampleString(FixedStr);
446   if (ExampleString.empty())
447     ExampleString = RegExStr;
448 
449   // Only compare up to the first line in the buffer, or the string size.
450   StringRef BufferPrefix = Buffer.substr(0, ExampleString.size());
451   BufferPrefix = BufferPrefix.split('\n').first;
452   return BufferPrefix.edit_distance(ExampleString);
453 }
454 
PrintFailureInfo(const SourceMgr & SM,StringRef Buffer,const StringMap<StringRef> & VariableTable) const455 void Pattern::PrintFailureInfo(const SourceMgr &SM, StringRef Buffer,
456                                const StringMap<StringRef> &VariableTable) const{
457   // If this was a regular expression using variables, print the current
458   // variable values.
459   if (!VariableUses.empty()) {
460     for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
461       SmallString<256> Msg;
462       raw_svector_ostream OS(Msg);
463       StringRef Var = VariableUses[i].first;
464       if (Var[0] == '@') {
465         std::string Value;
466         if (EvaluateExpression(Var, Value)) {
467           OS << "with expression \"";
468           OS.write_escaped(Var) << "\" equal to \"";
469           OS.write_escaped(Value) << "\"";
470         } else {
471           OS << "uses incorrect expression \"";
472           OS.write_escaped(Var) << "\"";
473         }
474       } else {
475         StringMap<StringRef>::const_iterator it = VariableTable.find(Var);
476 
477         // Check for undefined variable references.
478         if (it == VariableTable.end()) {
479           OS << "uses undefined variable \"";
480           OS.write_escaped(Var) << "\"";
481         } else {
482           OS << "with variable \"";
483           OS.write_escaped(Var) << "\" equal to \"";
484           OS.write_escaped(it->second) << "\"";
485         }
486       }
487 
488       SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
489                       OS.str());
490     }
491   }
492 
493   // Attempt to find the closest/best fuzzy match.  Usually an error happens
494   // because some string in the output didn't exactly match. In these cases, we
495   // would like to show the user a best guess at what "should have" matched, to
496   // save them having to actually check the input manually.
497   size_t NumLinesForward = 0;
498   size_t Best = StringRef::npos;
499   double BestQuality = 0;
500 
501   // Use an arbitrary 4k limit on how far we will search.
502   for (size_t i = 0, e = std::min(size_t(4096), Buffer.size()); i != e; ++i) {
503     if (Buffer[i] == '\n')
504       ++NumLinesForward;
505 
506     // Patterns have leading whitespace stripped, so skip whitespace when
507     // looking for something which looks like a pattern.
508     if (Buffer[i] == ' ' || Buffer[i] == '\t')
509       continue;
510 
511     // Compute the "quality" of this match as an arbitrary combination of the
512     // match distance and the number of lines skipped to get to this match.
513     unsigned Distance = ComputeMatchDistance(Buffer.substr(i), VariableTable);
514     double Quality = Distance + (NumLinesForward / 100.);
515 
516     if (Quality < BestQuality || Best == StringRef::npos) {
517       Best = i;
518       BestQuality = Quality;
519     }
520   }
521 
522   // Print the "possible intended match here" line if we found something
523   // reasonable and not equal to what we showed in the "scanning from here"
524   // line.
525   if (Best && Best != StringRef::npos && BestQuality < 50) {
526       SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + Best),
527                       SourceMgr::DK_Note, "possible intended match here");
528 
529     // FIXME: If we wanted to be really friendly we would show why the match
530     // failed, as it can be hard to spot simple one character differences.
531   }
532 }
533 
FindRegexVarEnd(StringRef Str,SourceMgr & SM)534 size_t Pattern::FindRegexVarEnd(StringRef Str, SourceMgr &SM) {
535   // Offset keeps track of the current offset within the input Str
536   size_t Offset = 0;
537   // [...] Nesting depth
538   size_t BracketDepth = 0;
539 
540   while (!Str.empty()) {
541     if (Str.startswith("]]") && BracketDepth == 0)
542       return Offset;
543     if (Str[0] == '\\') {
544       // Backslash escapes the next char within regexes, so skip them both.
545       Str = Str.substr(2);
546       Offset += 2;
547     } else {
548       switch (Str[0]) {
549         default:
550           break;
551         case '[':
552           BracketDepth++;
553           break;
554         case ']':
555           if (BracketDepth == 0) {
556             SM.PrintMessage(SMLoc::getFromPointer(Str.data()),
557                             SourceMgr::DK_Error,
558                             "missing closing \"]\" for regex variable");
559             exit(1);
560           }
561           BracketDepth--;
562           break;
563       }
564       Str = Str.substr(1);
565       Offset++;
566     }
567   }
568 
569   return StringRef::npos;
570 }
571 
572 
573 //===----------------------------------------------------------------------===//
574 // Check Strings.
575 //===----------------------------------------------------------------------===//
576 
577 /// CheckString - This is a check that we found in the input file.
578 struct CheckString {
579   /// Pat - The pattern to match.
580   Pattern Pat;
581 
582   /// Prefix - Which prefix name this check matched.
583   StringRef Prefix;
584 
585   /// Loc - The location in the match file that the check string was specified.
586   SMLoc Loc;
587 
588   /// CheckTy - Specify what kind of check this is. e.g. CHECK-NEXT: directive,
589   /// as opposed to a CHECK: directive.
590   Check::CheckType CheckTy;
591 
592   /// DagNotStrings - These are all of the strings that are disallowed from
593   /// occurring between this match string and the previous one (or start of
594   /// file).
595   std::vector<Pattern> DagNotStrings;
596 
597 
CheckStringCheckString598   CheckString(const Pattern &P,
599               StringRef S,
600               SMLoc L,
601               Check::CheckType Ty)
602     : Pat(P), Prefix(S), Loc(L), CheckTy(Ty) {}
603 
604   /// Check - Match check string and its "not strings" and/or "dag strings".
605   size_t Check(const SourceMgr &SM, StringRef Buffer, bool IsLabelScanMode,
606                size_t &MatchLen, StringMap<StringRef> &VariableTable) const;
607 
608   /// CheckNext - Verify there is a single line in the given buffer.
609   bool CheckNext(const SourceMgr &SM, StringRef Buffer) const;
610 
611   /// CheckNot - Verify there's no "not strings" in the given buffer.
612   bool CheckNot(const SourceMgr &SM, StringRef Buffer,
613                 const std::vector<const Pattern *> &NotStrings,
614                 StringMap<StringRef> &VariableTable) const;
615 
616   /// CheckDag - Match "dag strings" and their mixed "not strings".
617   size_t CheckDag(const SourceMgr &SM, StringRef Buffer,
618                   std::vector<const Pattern *> &NotStrings,
619                   StringMap<StringRef> &VariableTable) const;
620 };
621 
622 /// Canonicalize whitespaces in the input file. Line endings are replaced
623 /// with UNIX-style '\n'.
624 ///
625 /// \param PreserveHorizontal Don't squash consecutive horizontal whitespace
626 /// characters to a single space.
CanonicalizeInputFile(MemoryBuffer * MB,bool PreserveHorizontal)627 static MemoryBuffer *CanonicalizeInputFile(MemoryBuffer *MB,
628                                            bool PreserveHorizontal) {
629   SmallString<128> NewFile;
630   NewFile.reserve(MB->getBufferSize());
631 
632   for (const char *Ptr = MB->getBufferStart(), *End = MB->getBufferEnd();
633        Ptr != End; ++Ptr) {
634     // Eliminate trailing dosish \r.
635     if (Ptr <= End - 2 && Ptr[0] == '\r' && Ptr[1] == '\n') {
636       continue;
637     }
638 
639     // If current char is not a horizontal whitespace or if horizontal
640     // whitespace canonicalization is disabled, dump it to output as is.
641     if (PreserveHorizontal || (*Ptr != ' ' && *Ptr != '\t')) {
642       NewFile.push_back(*Ptr);
643       continue;
644     }
645 
646     // Otherwise, add one space and advance over neighboring space.
647     NewFile.push_back(' ');
648     while (Ptr+1 != End &&
649            (Ptr[1] == ' ' || Ptr[1] == '\t'))
650       ++Ptr;
651   }
652 
653   // Free the old buffer and return a new one.
654   MemoryBuffer *MB2 =
655     MemoryBuffer::getMemBufferCopy(NewFile.str(), MB->getBufferIdentifier());
656 
657   delete MB;
658   return MB2;
659 }
660 
IsPartOfWord(char c)661 static bool IsPartOfWord(char c) {
662   return (isalnum(c) || c == '-' || c == '_');
663 }
664 
665 // Get the size of the prefix extension.
CheckTypeSize(Check::CheckType Ty)666 static size_t CheckTypeSize(Check::CheckType Ty) {
667   switch (Ty) {
668   case Check::CheckNone:
669     return 0;
670 
671   case Check::CheckPlain:
672     return sizeof(":") - 1;
673 
674   case Check::CheckNext:
675     return sizeof("-NEXT:") - 1;
676 
677   case Check::CheckNot:
678     return sizeof("-NOT:") - 1;
679 
680   case Check::CheckDAG:
681     return sizeof("-DAG:") - 1;
682 
683   case Check::CheckLabel:
684     return sizeof("-LABEL:") - 1;
685 
686   case Check::CheckEOF:
687     llvm_unreachable("Should not be using EOF size");
688   }
689 
690   llvm_unreachable("Bad check type");
691 }
692 
FindCheckType(StringRef Buffer,StringRef Prefix)693 static Check::CheckType FindCheckType(StringRef Buffer, StringRef Prefix) {
694   char NextChar = Buffer[Prefix.size()];
695 
696   // Verify that the : is present after the prefix.
697   if (NextChar == ':')
698     return Check::CheckPlain;
699 
700   if (NextChar != '-')
701     return Check::CheckNone;
702 
703   StringRef Rest = Buffer.drop_front(Prefix.size() + 1);
704   if (Rest.startswith("NEXT:"))
705     return Check::CheckNext;
706 
707   if (Rest.startswith("NOT:"))
708     return Check::CheckNot;
709 
710   if (Rest.startswith("DAG:"))
711     return Check::CheckDAG;
712 
713   if (Rest.startswith("LABEL:"))
714     return Check::CheckLabel;
715 
716   return Check::CheckNone;
717 }
718 
719 // From the given position, find the next character after the word.
SkipWord(StringRef Str,size_t Loc)720 static size_t SkipWord(StringRef Str, size_t Loc) {
721   while (Loc < Str.size() && IsPartOfWord(Str[Loc]))
722     ++Loc;
723   return Loc;
724 }
725 
726 // Try to find the first match in buffer for any prefix. If a valid match is
727 // found, return that prefix and set its type and location.  If there are almost
728 // matches (e.g. the actual prefix string is found, but is not an actual check
729 // string), but no valid match, return an empty string and set the position to
730 // resume searching from. If no partial matches are found, return an empty
731 // string and the location will be StringRef::npos. If one prefix is a substring
732 // of another, the maximal match should be found. e.g. if "A" and "AA" are
733 // prefixes then AA-CHECK: should match the second one.
FindFirstCandidateMatch(StringRef & Buffer,Check::CheckType & CheckTy,size_t & CheckLoc)734 static StringRef FindFirstCandidateMatch(StringRef &Buffer,
735                                          Check::CheckType &CheckTy,
736                                          size_t &CheckLoc) {
737   StringRef FirstPrefix;
738   size_t FirstLoc = StringRef::npos;
739   size_t SearchLoc = StringRef::npos;
740   Check::CheckType FirstTy = Check::CheckNone;
741 
742   CheckTy = Check::CheckNone;
743   CheckLoc = StringRef::npos;
744 
745   for (prefix_iterator I = CheckPrefixes.begin(), E = CheckPrefixes.end();
746        I != E; ++I) {
747     StringRef Prefix(*I);
748     size_t PrefixLoc = Buffer.find(Prefix);
749 
750     if (PrefixLoc == StringRef::npos)
751       continue;
752 
753     // Track where we are searching for invalid prefixes that look almost right.
754     // We need to only advance to the first partial match on the next attempt
755     // since a partial match could be a substring of a later, valid prefix.
756     // Need to skip to the end of the word, otherwise we could end up
757     // matching a prefix in a substring later.
758     if (PrefixLoc < SearchLoc)
759       SearchLoc = SkipWord(Buffer, PrefixLoc);
760 
761     // We only want to find the first match to avoid skipping some.
762     if (PrefixLoc > FirstLoc)
763       continue;
764     // If one matching check-prefix is a prefix of another, choose the
765     // longer one.
766     if (PrefixLoc == FirstLoc && Prefix.size() < FirstPrefix.size())
767       continue;
768 
769     StringRef Rest = Buffer.drop_front(PrefixLoc);
770     // Make sure we have actually found the prefix, and not a word containing
771     // it. This should also prevent matching the wrong prefix when one is a
772     // substring of another.
773     if (PrefixLoc != 0 && IsPartOfWord(Buffer[PrefixLoc - 1]))
774       FirstTy = Check::CheckNone;
775     else
776       FirstTy = FindCheckType(Rest, Prefix);
777 
778     FirstLoc = PrefixLoc;
779     FirstPrefix = Prefix;
780   }
781 
782   // If the first prefix is invalid, we should continue the search after it.
783   if (FirstTy == Check::CheckNone) {
784     CheckLoc = SearchLoc;
785     return "";
786   }
787 
788   CheckTy = FirstTy;
789   CheckLoc = FirstLoc;
790   return FirstPrefix;
791 }
792 
FindFirstMatchingPrefix(StringRef & Buffer,unsigned & LineNumber,Check::CheckType & CheckTy,size_t & CheckLoc)793 static StringRef FindFirstMatchingPrefix(StringRef &Buffer,
794                                          unsigned &LineNumber,
795                                          Check::CheckType &CheckTy,
796                                          size_t &CheckLoc) {
797   while (!Buffer.empty()) {
798     StringRef Prefix = FindFirstCandidateMatch(Buffer, CheckTy, CheckLoc);
799     // If we found a real match, we are done.
800     if (!Prefix.empty()) {
801       LineNumber += Buffer.substr(0, CheckLoc).count('\n');
802       return Prefix;
803     }
804 
805     // We didn't find any almost matches either, we are also done.
806     if (CheckLoc == StringRef::npos)
807       return StringRef();
808 
809     LineNumber += Buffer.substr(0, CheckLoc + 1).count('\n');
810 
811     // Advance to the last possible match we found and try again.
812     Buffer = Buffer.drop_front(CheckLoc + 1);
813   }
814 
815   return StringRef();
816 }
817 
818 /// ReadCheckFile - Read the check file, which specifies the sequence of
819 /// expected strings.  The strings are added to the CheckStrings vector.
820 /// Returns true in case of an error, false otherwise.
ReadCheckFile(SourceMgr & SM,std::vector<CheckString> & CheckStrings)821 static bool ReadCheckFile(SourceMgr &SM,
822                           std::vector<CheckString> &CheckStrings) {
823   ErrorOr<std::unique_ptr<MemoryBuffer>> FileOrErr =
824       MemoryBuffer::getFileOrSTDIN(CheckFilename);
825   if (std::error_code EC = FileOrErr.getError()) {
826     errs() << "Could not open check file '" << CheckFilename
827            << "': " << EC.message() << '\n';
828     return true;
829   }
830 
831   // If we want to canonicalize whitespace, strip excess whitespace from the
832   // buffer containing the CHECK lines. Remove DOS style line endings.
833   MemoryBuffer *F = CanonicalizeInputFile(FileOrErr.get().release(),
834                                           NoCanonicalizeWhiteSpace);
835 
836   SM.AddNewSourceBuffer(F, SMLoc());
837 
838   // Find all instances of CheckPrefix followed by : in the file.
839   StringRef Buffer = F->getBuffer();
840   std::vector<Pattern> DagNotMatches;
841 
842   // LineNumber keeps track of the line on which CheckPrefix instances are
843   // found.
844   unsigned LineNumber = 1;
845 
846   while (1) {
847     Check::CheckType CheckTy;
848     size_t PrefixLoc;
849 
850     // See if a prefix occurs in the memory buffer.
851     StringRef UsedPrefix = FindFirstMatchingPrefix(Buffer,
852                                                    LineNumber,
853                                                    CheckTy,
854                                                    PrefixLoc);
855     if (UsedPrefix.empty())
856       break;
857 
858     Buffer = Buffer.drop_front(PrefixLoc);
859 
860     // Location to use for error messages.
861     const char *UsedPrefixStart = Buffer.data() + (PrefixLoc == 0 ? 0 : 1);
862 
863     // PrefixLoc is to the start of the prefix. Skip to the end.
864     Buffer = Buffer.drop_front(UsedPrefix.size() + CheckTypeSize(CheckTy));
865 
866     // Okay, we found the prefix, yay. Remember the rest of the line, but ignore
867     // leading and trailing whitespace.
868     Buffer = Buffer.substr(Buffer.find_first_not_of(" \t"));
869 
870     // Scan ahead to the end of line.
871     size_t EOL = Buffer.find_first_of("\n\r");
872 
873     // Remember the location of the start of the pattern, for diagnostics.
874     SMLoc PatternLoc = SMLoc::getFromPointer(Buffer.data());
875 
876     // Parse the pattern.
877     Pattern P(CheckTy);
878     if (P.ParsePattern(Buffer.substr(0, EOL), UsedPrefix, SM, LineNumber))
879       return true;
880 
881     // Verify that CHECK-LABEL lines do not define or use variables
882     if ((CheckTy == Check::CheckLabel) && P.hasVariable()) {
883       SM.PrintMessage(SMLoc::getFromPointer(UsedPrefixStart),
884                       SourceMgr::DK_Error,
885                       "found '" + UsedPrefix + "-LABEL:'"
886                       " with variable definition or use");
887       return true;
888     }
889 
890     Buffer = Buffer.substr(EOL);
891 
892     // Verify that CHECK-NEXT lines have at least one CHECK line before them.
893     if ((CheckTy == Check::CheckNext) && CheckStrings.empty()) {
894       SM.PrintMessage(SMLoc::getFromPointer(UsedPrefixStart),
895                       SourceMgr::DK_Error,
896                       "found '" + UsedPrefix + "-NEXT:' without previous '"
897                       + UsedPrefix + ": line");
898       return true;
899     }
900 
901     // Handle CHECK-DAG/-NOT.
902     if (CheckTy == Check::CheckDAG || CheckTy == Check::CheckNot) {
903       DagNotMatches.push_back(P);
904       continue;
905     }
906 
907     // Okay, add the string we captured to the output vector and move on.
908     CheckStrings.push_back(CheckString(P,
909                                        UsedPrefix,
910                                        PatternLoc,
911                                        CheckTy));
912     std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
913   }
914 
915   // Add an EOF pattern for any trailing CHECK-DAG/-NOTs, and use the first
916   // prefix as a filler for the error message.
917   if (!DagNotMatches.empty()) {
918     CheckStrings.push_back(CheckString(Pattern(Check::CheckEOF),
919                                        CheckPrefixes[0],
920                                        SMLoc::getFromPointer(Buffer.data()),
921                                        Check::CheckEOF));
922     std::swap(DagNotMatches, CheckStrings.back().DagNotStrings);
923   }
924 
925   if (CheckStrings.empty()) {
926     errs() << "error: no check strings found with prefix"
927            << (CheckPrefixes.size() > 1 ? "es " : " ");
928     for (size_t I = 0, N = CheckPrefixes.size(); I != N; ++I) {
929       StringRef Prefix(CheckPrefixes[I]);
930       errs() << '\'' << Prefix << ":'";
931       if (I != N - 1)
932         errs() << ", ";
933     }
934 
935     errs() << '\n';
936     return true;
937   }
938 
939   return false;
940 }
941 
PrintCheckFailed(const SourceMgr & SM,const SMLoc & Loc,const Pattern & Pat,StringRef Buffer,StringMap<StringRef> & VariableTable)942 static void PrintCheckFailed(const SourceMgr &SM, const SMLoc &Loc,
943                              const Pattern &Pat, StringRef Buffer,
944                              StringMap<StringRef> &VariableTable) {
945   // Otherwise, we have an error, emit an error message.
946   SM.PrintMessage(Loc, SourceMgr::DK_Error,
947                   "expected string not found in input");
948 
949   // Print the "scanning from here" line.  If the current position is at the
950   // end of a line, advance to the start of the next line.
951   Buffer = Buffer.substr(Buffer.find_first_not_of(" \t\n\r"));
952 
953   SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
954                   "scanning from here");
955 
956   // Allow the pattern to print additional information if desired.
957   Pat.PrintFailureInfo(SM, Buffer, VariableTable);
958 }
959 
PrintCheckFailed(const SourceMgr & SM,const CheckString & CheckStr,StringRef Buffer,StringMap<StringRef> & VariableTable)960 static void PrintCheckFailed(const SourceMgr &SM, const CheckString &CheckStr,
961                              StringRef Buffer,
962                              StringMap<StringRef> &VariableTable) {
963   PrintCheckFailed(SM, CheckStr.Loc, CheckStr.Pat, Buffer, VariableTable);
964 }
965 
966 /// CountNumNewlinesBetween - Count the number of newlines in the specified
967 /// range.
CountNumNewlinesBetween(StringRef Range,const char * & FirstNewLine)968 static unsigned CountNumNewlinesBetween(StringRef Range,
969                                         const char *&FirstNewLine) {
970   unsigned NumNewLines = 0;
971   while (1) {
972     // Scan for newline.
973     Range = Range.substr(Range.find_first_of("\n\r"));
974     if (Range.empty()) return NumNewLines;
975 
976     ++NumNewLines;
977 
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);
984 
985     if (NumNewLines == 1)
986       FirstNewLine = Range.begin();
987   }
988 }
989 
Check(const SourceMgr & SM,StringRef Buffer,bool IsLabelScanMode,size_t & MatchLen,StringMap<StringRef> & VariableTable) const990 size_t CheckString::Check(const SourceMgr &SM, StringRef Buffer,
991                           bool IsLabelScanMode, size_t &MatchLen,
992                           StringMap<StringRef> &VariableTable) const {
993   size_t LastPos = 0;
994   std::vector<const Pattern *> NotStrings;
995 
996   // IsLabelScanMode is true when we are scanning forward to find CHECK-LABEL
997   // bounds; we have not processed variable definitions within the bounded block
998   // yet so cannot handle any final CHECK-DAG yet; this is handled when going
999   // over the block again (including the last CHECK-LABEL) in normal mode.
1000   if (!IsLabelScanMode) {
1001     // Match "dag strings" (with mixed "not strings" if any).
1002     LastPos = CheckDag(SM, Buffer, NotStrings, VariableTable);
1003     if (LastPos == StringRef::npos)
1004       return StringRef::npos;
1005   }
1006 
1007   // Match itself from the last position after matching CHECK-DAG.
1008   StringRef MatchBuffer = Buffer.substr(LastPos);
1009   size_t MatchPos = Pat.Match(MatchBuffer, MatchLen, VariableTable);
1010   if (MatchPos == StringRef::npos) {
1011     PrintCheckFailed(SM, *this, MatchBuffer, VariableTable);
1012     return StringRef::npos;
1013   }
1014   MatchPos += LastPos;
1015 
1016   // Similar to the above, in "label-scan mode" we can't yet handle CHECK-NEXT
1017   // or CHECK-NOT
1018   if (!IsLabelScanMode) {
1019     StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
1020 
1021     // If this check is a "CHECK-NEXT", verify that the previous match was on
1022     // the previous line (i.e. that there is one newline between them).
1023     if (CheckNext(SM, SkippedRegion))
1024       return StringRef::npos;
1025 
1026     // If this match had "not strings", verify that they don't exist in the
1027     // skipped region.
1028     if (CheckNot(SM, SkippedRegion, NotStrings, VariableTable))
1029       return StringRef::npos;
1030   }
1031 
1032   return MatchPos;
1033 }
1034 
CheckNext(const SourceMgr & SM,StringRef Buffer) const1035 bool CheckString::CheckNext(const SourceMgr &SM, StringRef Buffer) const {
1036   if (CheckTy != Check::CheckNext)
1037     return false;
1038 
1039   // Count the number of newlines between the previous match and this one.
1040   assert(Buffer.data() !=
1041          SM.getMemoryBuffer(
1042            SM.FindBufferContainingLoc(
1043              SMLoc::getFromPointer(Buffer.data())))->getBufferStart() &&
1044          "CHECK-NEXT can't be the first check in a file");
1045 
1046   const char *FirstNewLine = nullptr;
1047   unsigned NumNewLines = CountNumNewlinesBetween(Buffer, FirstNewLine);
1048 
1049   if (NumNewLines == 0) {
1050     SM.PrintMessage(Loc, SourceMgr::DK_Error, Prefix +
1051                     "-NEXT: is on the same line as previous match");
1052     SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()),
1053                     SourceMgr::DK_Note, "'next' match was here");
1054     SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
1055                     "previous match ended here");
1056     return true;
1057   }
1058 
1059   if (NumNewLines != 1) {
1060     SM.PrintMessage(Loc, SourceMgr::DK_Error, Prefix +
1061                     "-NEXT: is not on the line after the previous match");
1062     SM.PrintMessage(SMLoc::getFromPointer(Buffer.end()),
1063                     SourceMgr::DK_Note, "'next' match was here");
1064     SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), SourceMgr::DK_Note,
1065                     "previous match ended here");
1066     SM.PrintMessage(SMLoc::getFromPointer(FirstNewLine), SourceMgr::DK_Note,
1067                     "non-matching line after previous match is here");
1068     return true;
1069   }
1070 
1071   return false;
1072 }
1073 
CheckNot(const SourceMgr & SM,StringRef Buffer,const std::vector<const Pattern * > & NotStrings,StringMap<StringRef> & VariableTable) const1074 bool CheckString::CheckNot(const SourceMgr &SM, StringRef Buffer,
1075                            const std::vector<const Pattern *> &NotStrings,
1076                            StringMap<StringRef> &VariableTable) const {
1077   for (unsigned ChunkNo = 0, e = NotStrings.size();
1078        ChunkNo != e; ++ChunkNo) {
1079     const Pattern *Pat = NotStrings[ChunkNo];
1080     assert((Pat->getCheckTy() == Check::CheckNot) && "Expect CHECK-NOT!");
1081 
1082     size_t MatchLen = 0;
1083     size_t Pos = Pat->Match(Buffer, MatchLen, VariableTable);
1084 
1085     if (Pos == StringRef::npos) continue;
1086 
1087     SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()+Pos),
1088                     SourceMgr::DK_Error,
1089                     Prefix + "-NOT: string occurred!");
1090     SM.PrintMessage(Pat->getLoc(), SourceMgr::DK_Note,
1091                     Prefix + "-NOT: pattern specified here");
1092     return true;
1093   }
1094 
1095   return false;
1096 }
1097 
CheckDag(const SourceMgr & SM,StringRef Buffer,std::vector<const Pattern * > & NotStrings,StringMap<StringRef> & VariableTable) const1098 size_t CheckString::CheckDag(const SourceMgr &SM, StringRef Buffer,
1099                              std::vector<const Pattern *> &NotStrings,
1100                              StringMap<StringRef> &VariableTable) const {
1101   if (DagNotStrings.empty())
1102     return 0;
1103 
1104   size_t LastPos = 0;
1105   size_t StartPos = LastPos;
1106 
1107   for (unsigned ChunkNo = 0, e = DagNotStrings.size();
1108        ChunkNo != e; ++ChunkNo) {
1109     const Pattern &Pat = DagNotStrings[ChunkNo];
1110 
1111     assert((Pat.getCheckTy() == Check::CheckDAG ||
1112             Pat.getCheckTy() == Check::CheckNot) &&
1113            "Invalid CHECK-DAG or CHECK-NOT!");
1114 
1115     if (Pat.getCheckTy() == Check::CheckNot) {
1116       NotStrings.push_back(&Pat);
1117       continue;
1118     }
1119 
1120     assert((Pat.getCheckTy() == Check::CheckDAG) && "Expect CHECK-DAG!");
1121 
1122     size_t MatchLen = 0, MatchPos;
1123 
1124     // CHECK-DAG always matches from the start.
1125     StringRef MatchBuffer = Buffer.substr(StartPos);
1126     MatchPos = Pat.Match(MatchBuffer, MatchLen, VariableTable);
1127     // With a group of CHECK-DAGs, a single mismatching means the match on
1128     // that group of CHECK-DAGs fails immediately.
1129     if (MatchPos == StringRef::npos) {
1130       PrintCheckFailed(SM, Pat.getLoc(), Pat, MatchBuffer, VariableTable);
1131       return StringRef::npos;
1132     }
1133     // Re-calc it as the offset relative to the start of the original string.
1134     MatchPos += StartPos;
1135 
1136     if (!NotStrings.empty()) {
1137       if (MatchPos < LastPos) {
1138         // Reordered?
1139         SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + MatchPos),
1140                         SourceMgr::DK_Error,
1141                         Prefix + "-DAG: found a match of CHECK-DAG"
1142                         " reordering across a CHECK-NOT");
1143         SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + LastPos),
1144                         SourceMgr::DK_Note,
1145                         Prefix + "-DAG: the farthest match of CHECK-DAG"
1146                         " is found here");
1147         SM.PrintMessage(NotStrings[0]->getLoc(), SourceMgr::DK_Note,
1148                         Prefix + "-NOT: the crossed pattern specified"
1149                         " here");
1150         SM.PrintMessage(Pat.getLoc(), SourceMgr::DK_Note,
1151                         Prefix + "-DAG: the reordered pattern specified"
1152                         " here");
1153         return StringRef::npos;
1154       }
1155       // All subsequent CHECK-DAGs should be matched from the farthest
1156       // position of all precedent CHECK-DAGs (including this one.)
1157       StartPos = LastPos;
1158       // If there's CHECK-NOTs between two CHECK-DAGs or from CHECK to
1159       // CHECK-DAG, verify that there's no 'not' strings occurred in that
1160       // region.
1161       StringRef SkippedRegion = Buffer.substr(LastPos, MatchPos);
1162       if (CheckNot(SM, SkippedRegion, NotStrings, VariableTable))
1163         return StringRef::npos;
1164       // Clear "not strings".
1165       NotStrings.clear();
1166     }
1167 
1168     // Update the last position with CHECK-DAG matches.
1169     LastPos = std::max(MatchPos + MatchLen, LastPos);
1170   }
1171 
1172   return LastPos;
1173 }
1174 
1175 // A check prefix must contain only alphanumeric, hyphens and underscores.
ValidateCheckPrefix(StringRef CheckPrefix)1176 static bool ValidateCheckPrefix(StringRef CheckPrefix) {
1177   Regex Validator("^[a-zA-Z0-9_-]*$");
1178   return Validator.match(CheckPrefix);
1179 }
1180 
ValidateCheckPrefixes()1181 static bool ValidateCheckPrefixes() {
1182   StringSet<> PrefixSet;
1183 
1184   for (prefix_iterator I = CheckPrefixes.begin(), E = CheckPrefixes.end();
1185        I != E; ++I) {
1186     StringRef Prefix(*I);
1187 
1188     if (!PrefixSet.insert(Prefix))
1189       return false;
1190 
1191     if (!ValidateCheckPrefix(Prefix))
1192       return false;
1193   }
1194 
1195   return true;
1196 }
1197 
1198 // I don't think there's a way to specify an initial value for cl::list,
1199 // so if nothing was specified, add the default
AddCheckPrefixIfNeeded()1200 static void AddCheckPrefixIfNeeded() {
1201   if (CheckPrefixes.empty())
1202     CheckPrefixes.push_back("CHECK");
1203 }
1204 
main(int argc,char ** argv)1205 int main(int argc, char **argv) {
1206   sys::PrintStackTraceOnErrorSignal();
1207   PrettyStackTraceProgram X(argc, argv);
1208   cl::ParseCommandLineOptions(argc, argv);
1209 
1210   if (!ValidateCheckPrefixes()) {
1211     errs() << "Supplied check-prefix is invalid! Prefixes must be unique and "
1212               "start with a letter and contain only alphanumeric characters, "
1213               "hyphens and underscores\n";
1214     return 2;
1215   }
1216 
1217   AddCheckPrefixIfNeeded();
1218 
1219   SourceMgr SM;
1220 
1221   // Read the expected strings from the check file.
1222   std::vector<CheckString> CheckStrings;
1223   if (ReadCheckFile(SM, CheckStrings))
1224     return 2;
1225 
1226   // Open the file to check and add it to SourceMgr.
1227   ErrorOr<std::unique_ptr<MemoryBuffer>> FileOrErr =
1228       MemoryBuffer::getFileOrSTDIN(InputFilename);
1229   if (std::error_code EC = FileOrErr.getError()) {
1230     errs() << "Could not open input file '" << InputFilename
1231            << "': " << EC.message() << '\n';
1232     return 2;
1233   }
1234   std::unique_ptr<MemoryBuffer> File = std::move(FileOrErr.get());
1235 
1236   if (File->getBufferSize() == 0) {
1237     errs() << "FileCheck error: '" << InputFilename << "' is empty.\n";
1238     return 2;
1239   }
1240 
1241   // Remove duplicate spaces in the input file if requested.
1242   // Remove DOS style line endings.
1243   MemoryBuffer *F =
1244     CanonicalizeInputFile(File.release(), NoCanonicalizeWhiteSpace);
1245 
1246   SM.AddNewSourceBuffer(F, SMLoc());
1247 
1248   /// VariableTable - This holds all the current filecheck variables.
1249   StringMap<StringRef> VariableTable;
1250 
1251   // Check that we have all of the expected strings, in order, in the input
1252   // file.
1253   StringRef Buffer = F->getBuffer();
1254 
1255   bool hasError = false;
1256 
1257   unsigned i = 0, j = 0, e = CheckStrings.size();
1258 
1259   while (true) {
1260     StringRef CheckRegion;
1261     if (j == e) {
1262       CheckRegion = Buffer;
1263     } else {
1264       const CheckString &CheckLabelStr = CheckStrings[j];
1265       if (CheckLabelStr.CheckTy != Check::CheckLabel) {
1266         ++j;
1267         continue;
1268       }
1269 
1270       // Scan to next CHECK-LABEL match, ignoring CHECK-NOT and CHECK-DAG
1271       size_t MatchLabelLen = 0;
1272       size_t MatchLabelPos = CheckLabelStr.Check(SM, Buffer, true,
1273                                                  MatchLabelLen, VariableTable);
1274       if (MatchLabelPos == StringRef::npos) {
1275         hasError = true;
1276         break;
1277       }
1278 
1279       CheckRegion = Buffer.substr(0, MatchLabelPos + MatchLabelLen);
1280       Buffer = Buffer.substr(MatchLabelPos + MatchLabelLen);
1281       ++j;
1282     }
1283 
1284     for ( ; i != j; ++i) {
1285       const CheckString &CheckStr = CheckStrings[i];
1286 
1287       // Check each string within the scanned region, including a second check
1288       // of any final CHECK-LABEL (to verify CHECK-NOT and CHECK-DAG)
1289       size_t MatchLen = 0;
1290       size_t MatchPos = CheckStr.Check(SM, CheckRegion, false, MatchLen,
1291                                        VariableTable);
1292 
1293       if (MatchPos == StringRef::npos) {
1294         hasError = true;
1295         i = j;
1296         break;
1297       }
1298 
1299       CheckRegion = CheckRegion.substr(MatchPos + MatchLen);
1300     }
1301 
1302     if (j == e)
1303       break;
1304   }
1305 
1306   return hasError ? 1 : 0;
1307 }
1308