Home | History | Annotate | Download | only in FileCheck
      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/OwningPtr.h"
     20 #include "llvm/Support/CommandLine.h"
     21 #include "llvm/Support/MemoryBuffer.h"
     22 #include "llvm/Support/PrettyStackTrace.h"
     23 #include "llvm/Support/Regex.h"
     24 #include "llvm/Support/SourceMgr.h"
     25 #include "llvm/Support/raw_ostream.h"
     26 #include "llvm/Support/Signals.h"
     27 #include "llvm/Support/system_error.h"
     28 #include "llvm/ADT/SmallString.h"
     29 #include "llvm/ADT/StringMap.h"
     30 #include <algorithm>
     31 using namespace llvm;
     32 
     33 static cl::opt<std::string>
     34 CheckFilename(cl::Positional, cl::desc("<check-file>"), cl::Required);
     35 
     36 static cl::opt<std::string>
     37 InputFilename("input-file", cl::desc("File to check (defaults to stdin)"),
     38               cl::init("-"), cl::value_desc("filename"));
     39 
     40 static cl::opt<std::string>
     41 CheckPrefix("check-prefix", cl::init("CHECK"),
     42             cl::desc("Prefix to use from check file (defaults to 'CHECK')"));
     43 
     44 static cl::opt<bool>
     45 NoCanonicalizeWhiteSpace("strict-whitespace",
     46               cl::desc("Do not treat all horizontal whitespace as equivalent"));
     47 
     48 //===----------------------------------------------------------------------===//
     49 // Pattern Handling Code.
     50 //===----------------------------------------------------------------------===//
     51 
     52 class Pattern {
     53   SMLoc PatternLoc;
     54 
     55   /// MatchEOF - When set, this pattern only matches the end of file. This is
     56   /// used for trailing CHECK-NOTs.
     57   bool MatchEOF;
     58 
     59   /// FixedStr - If non-empty, this pattern is a fixed string match with the
     60   /// specified fixed string.
     61   StringRef FixedStr;
     62 
     63   /// RegEx - If non-empty, this is a regex pattern.
     64   std::string RegExStr;
     65 
     66   /// VariableUses - Entries in this vector map to uses of a variable in the
     67   /// pattern, e.g. "foo[[bar]]baz".  In this case, the RegExStr will contain
     68   /// "foobaz" and we'll get an entry in this vector that tells us to insert the
     69   /// value of bar at offset 3.
     70   std::vector<std::pair<StringRef, unsigned> > VariableUses;
     71 
     72   /// VariableDefs - Entries in this vector map to definitions of a variable in
     73   /// the pattern, e.g. "foo[[bar:.*]]baz".  In this case, the RegExStr will
     74   /// contain "foo(.*)baz" and VariableDefs will contain the pair "bar",1.  The
     75   /// index indicates what parenthesized value captures the variable value.
     76   std::vector<std::pair<StringRef, unsigned> > VariableDefs;
     77 
     78 public:
     79 
     80   Pattern(bool matchEOF = false) : MatchEOF(matchEOF) { }
     81 
     82   bool ParsePattern(StringRef PatternStr, SourceMgr &SM);
     83 
     84   /// Match - Match the pattern string against the input buffer Buffer.  This
     85   /// returns the position that is matched or npos if there is no match.  If
     86   /// there is a match, the size of the matched string is returned in MatchLen.
     87   ///
     88   /// The VariableTable StringMap provides the current values of filecheck
     89   /// variables and is updated if this match defines new values.
     90   size_t Match(StringRef Buffer, size_t &MatchLen,
     91                StringMap<StringRef> &VariableTable) const;
     92 
     93   /// PrintFailureInfo - Print additional information about a failure to match
     94   /// involving this pattern.
     95   void PrintFailureInfo(const SourceMgr &SM, StringRef Buffer,
     96                         const StringMap<StringRef> &VariableTable) const;
     97 
     98 private:
     99   static void AddFixedStringToRegEx(StringRef FixedStr, std::string &TheStr);
    100   bool AddRegExToRegEx(StringRef RegExStr, unsigned &CurParen, SourceMgr &SM);
    101 
    102   /// ComputeMatchDistance - Compute an arbitrary estimate for the quality of
    103   /// matching this pattern at the start of \arg Buffer; a distance of zero
    104   /// should correspond to a perfect match.
    105   unsigned ComputeMatchDistance(StringRef Buffer,
    106                                const StringMap<StringRef> &VariableTable) const;
    107 };
    108 
    109 
    110 bool Pattern::ParsePattern(StringRef PatternStr, SourceMgr &SM) {
    111   PatternLoc = SMLoc::getFromPointer(PatternStr.data());
    112 
    113   // Ignore trailing whitespace.
    114   while (!PatternStr.empty() &&
    115          (PatternStr.back() == ' ' || PatternStr.back() == '\t'))
    116     PatternStr = PatternStr.substr(0, PatternStr.size()-1);
    117 
    118   // Check that there is something on the line.
    119   if (PatternStr.empty()) {
    120     SM.PrintMessage(PatternLoc, "found empty check string with prefix '" +
    121                     CheckPrefix+":'", "error");
    122     return true;
    123   }
    124 
    125   // Check to see if this is a fixed string, or if it has regex pieces.
    126   if (PatternStr.size() < 2 ||
    127       (PatternStr.find("{{") == StringRef::npos &&
    128        PatternStr.find("[[") == StringRef::npos)) {
    129     FixedStr = PatternStr;
    130     return false;
    131   }
    132 
    133   // Paren value #0 is for the fully matched string.  Any new parenthesized
    134   // values add from there.
    135   unsigned CurParen = 1;
    136 
    137   // Otherwise, there is at least one regex piece.  Build up the regex pattern
    138   // by escaping scary characters in fixed strings, building up one big regex.
    139   while (!PatternStr.empty()) {
    140     // RegEx matches.
    141     if (PatternStr.startswith("{{")) {
    142 
    143       // Otherwise, this is the start of a regex match.  Scan for the }}.
    144       size_t End = PatternStr.find("}}");
    145       if (End == StringRef::npos) {
    146         SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
    147                         "found start of regex string with no end '}}'","error");
    148         return true;
    149       }
    150 
    151       // Enclose {{}} patterns in parens just like [[]] even though we're not
    152       // capturing the result for any purpose.  This is required in case the
    153       // expression contains an alternation like: CHECK:  abc{{x|z}}def.  We
    154       // want this to turn into: "abc(x|z)def" not "abcx|zdef".
    155       RegExStr += '(';
    156       ++CurParen;
    157 
    158       if (AddRegExToRegEx(PatternStr.substr(2, End-2), CurParen, SM))
    159         return true;
    160       RegExStr += ')';
    161 
    162       PatternStr = PatternStr.substr(End+2);
    163       continue;
    164     }
    165 
    166     // Named RegEx matches.  These are of two forms: [[foo:.*]] which matches .*
    167     // (or some other regex) and assigns it to the FileCheck variable 'foo'. The
    168     // second form is [[foo]] which is a reference to foo.  The variable name
    169     // itself must be of the form "[a-zA-Z_][0-9a-zA-Z_]*", otherwise we reject
    170     // it.  This is to catch some common errors.
    171     if (PatternStr.startswith("[[")) {
    172       // Verify that it is terminated properly.
    173       size_t End = PatternStr.find("]]");
    174       if (End == StringRef::npos) {
    175         SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
    176                         "invalid named regex reference, no ]] found", "error");
    177         return true;
    178       }
    179 
    180       StringRef MatchStr = PatternStr.substr(2, End-2);
    181       PatternStr = PatternStr.substr(End+2);
    182 
    183       // Get the regex name (e.g. "foo").
    184       size_t NameEnd = MatchStr.find(':');
    185       StringRef Name = MatchStr.substr(0, NameEnd);
    186 
    187       if (Name.empty()) {
    188         SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
    189                         "invalid name in named regex: empty name", "error");
    190         return true;
    191       }
    192 
    193       // Verify that the name is well formed.
    194       for (unsigned i = 0, e = Name.size(); i != e; ++i)
    195         if (Name[i] != '_' && !isalnum(Name[i])) {
    196           SM.PrintMessage(SMLoc::getFromPointer(Name.data()+i),
    197                           "invalid name in named regex", "error");
    198           return true;
    199         }
    200 
    201       // Name can't start with a digit.
    202       if (isdigit(Name[0])) {
    203         SM.PrintMessage(SMLoc::getFromPointer(Name.data()),
    204                         "invalid name in named regex", "error");
    205         return true;
    206       }
    207 
    208       // Handle [[foo]].
    209       if (NameEnd == StringRef::npos) {
    210         VariableUses.push_back(std::make_pair(Name, RegExStr.size()));
    211         continue;
    212       }
    213 
    214       // Handle [[foo:.*]].
    215       VariableDefs.push_back(std::make_pair(Name, CurParen));
    216       RegExStr += '(';
    217       ++CurParen;
    218 
    219       if (AddRegExToRegEx(MatchStr.substr(NameEnd+1), CurParen, SM))
    220         return true;
    221 
    222       RegExStr += ')';
    223     }
    224 
    225     // Handle fixed string matches.
    226     // Find the end, which is the start of the next regex.
    227     size_t FixedMatchEnd = PatternStr.find("{{");
    228     FixedMatchEnd = std::min(FixedMatchEnd, PatternStr.find("[["));
    229     AddFixedStringToRegEx(PatternStr.substr(0, FixedMatchEnd), RegExStr);
    230     PatternStr = PatternStr.substr(FixedMatchEnd);
    231     continue;
    232   }
    233 
    234   return false;
    235 }
    236 
    237 void Pattern::AddFixedStringToRegEx(StringRef FixedStr, std::string &TheStr) {
    238   // Add the characters from FixedStr to the regex, escaping as needed.  This
    239   // avoids "leaning toothpicks" in common patterns.
    240   for (unsigned i = 0, e = FixedStr.size(); i != e; ++i) {
    241     switch (FixedStr[i]) {
    242     // These are the special characters matched in "p_ere_exp".
    243     case '(':
    244     case ')':
    245     case '^':
    246     case '$':
    247     case '|':
    248     case '*':
    249     case '+':
    250     case '?':
    251     case '.':
    252     case '[':
    253     case '\\':
    254     case '{':
    255       TheStr += '\\';
    256       // FALL THROUGH.
    257     default:
    258       TheStr += FixedStr[i];
    259       break;
    260     }
    261   }
    262 }
    263 
    264 bool Pattern::AddRegExToRegEx(StringRef RegexStr, unsigned &CurParen,
    265                               SourceMgr &SM) {
    266   Regex R(RegexStr);
    267   std::string Error;
    268   if (!R.isValid(Error)) {
    269     SM.PrintMessage(SMLoc::getFromPointer(RegexStr.data()),
    270                     "invalid regex: " + Error, "error");
    271     return true;
    272   }
    273 
    274   RegExStr += RegexStr.str();
    275   CurParen += R.getNumMatches();
    276   return false;
    277 }
    278 
    279 /// Match - Match the pattern string against the input buffer Buffer.  This
    280 /// returns the position that is matched or npos if there is no match.  If
    281 /// there is a match, the size of the matched string is returned in MatchLen.
    282 size_t Pattern::Match(StringRef Buffer, size_t &MatchLen,
    283                       StringMap<StringRef> &VariableTable) const {
    284   // If this is the EOF pattern, match it immediately.
    285   if (MatchEOF) {
    286     MatchLen = 0;
    287     return Buffer.size();
    288   }
    289 
    290   // If this is a fixed string pattern, just match it now.
    291   if (!FixedStr.empty()) {
    292     MatchLen = FixedStr.size();
    293     return Buffer.find(FixedStr);
    294   }
    295 
    296   // Regex match.
    297 
    298   // If there are variable uses, we need to create a temporary string with the
    299   // actual value.
    300   StringRef RegExToMatch = RegExStr;
    301   std::string TmpStr;
    302   if (!VariableUses.empty()) {
    303     TmpStr = RegExStr;
    304 
    305     unsigned InsertOffset = 0;
    306     for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
    307       StringMap<StringRef>::iterator it =
    308         VariableTable.find(VariableUses[i].first);
    309       // If the variable is undefined, return an error.
    310       if (it == VariableTable.end())
    311         return StringRef::npos;
    312 
    313       // Look up the value and escape it so that we can plop it into the regex.
    314       std::string Value;
    315       AddFixedStringToRegEx(it->second, Value);
    316 
    317       // Plop it into the regex at the adjusted offset.
    318       TmpStr.insert(TmpStr.begin()+VariableUses[i].second+InsertOffset,
    319                     Value.begin(), Value.end());
    320       InsertOffset += Value.size();
    321     }
    322 
    323     // Match the newly constructed regex.
    324     RegExToMatch = TmpStr;
    325   }
    326 
    327 
    328   SmallVector<StringRef, 4> MatchInfo;
    329   if (!Regex(RegExToMatch, Regex::Newline).match(Buffer, &MatchInfo))
    330     return StringRef::npos;
    331 
    332   // Successful regex match.
    333   assert(!MatchInfo.empty() && "Didn't get any match");
    334   StringRef FullMatch = MatchInfo[0];
    335 
    336   // If this defines any variables, remember their values.
    337   for (unsigned i = 0, e = VariableDefs.size(); i != e; ++i) {
    338     assert(VariableDefs[i].second < MatchInfo.size() &&
    339            "Internal paren error");
    340     VariableTable[VariableDefs[i].first] = MatchInfo[VariableDefs[i].second];
    341   }
    342 
    343   MatchLen = FullMatch.size();
    344   return FullMatch.data()-Buffer.data();
    345 }
    346 
    347 unsigned Pattern::ComputeMatchDistance(StringRef Buffer,
    348                               const StringMap<StringRef> &VariableTable) const {
    349   // Just compute the number of matching characters. For regular expressions, we
    350   // just compare against the regex itself and hope for the best.
    351   //
    352   // FIXME: One easy improvement here is have the regex lib generate a single
    353   // example regular expression which matches, and use that as the example
    354   // string.
    355   StringRef ExampleString(FixedStr);
    356   if (ExampleString.empty())
    357     ExampleString = RegExStr;
    358 
    359   // Only compare up to the first line in the buffer, or the string size.
    360   StringRef BufferPrefix = Buffer.substr(0, ExampleString.size());
    361   BufferPrefix = BufferPrefix.split('\n').first;
    362   return BufferPrefix.edit_distance(ExampleString);
    363 }
    364 
    365 void Pattern::PrintFailureInfo(const SourceMgr &SM, StringRef Buffer,
    366                                const StringMap<StringRef> &VariableTable) const{
    367   // If this was a regular expression using variables, print the current
    368   // variable values.
    369   if (!VariableUses.empty()) {
    370     for (unsigned i = 0, e = VariableUses.size(); i != e; ++i) {
    371       StringRef Var = VariableUses[i].first;
    372       StringMap<StringRef>::const_iterator it = VariableTable.find(Var);
    373       SmallString<256> Msg;
    374       raw_svector_ostream OS(Msg);
    375 
    376       // Check for undefined variable references.
    377       if (it == VariableTable.end()) {
    378         OS << "uses undefined variable \"";
    379         OS.write_escaped(Var) << "\"";;
    380       } else {
    381         OS << "with variable \"";
    382         OS.write_escaped(Var) << "\" equal to \"";
    383         OS.write_escaped(it->second) << "\"";
    384       }
    385 
    386       SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), OS.str(), "note",
    387                       /*ShowLine=*/false);
    388     }
    389   }
    390 
    391   // Attempt to find the closest/best fuzzy match.  Usually an error happens
    392   // because some string in the output didn't exactly match. In these cases, we
    393   // would like to show the user a best guess at what "should have" matched, to
    394   // save them having to actually check the input manually.
    395   size_t NumLinesForward = 0;
    396   size_t Best = StringRef::npos;
    397   double BestQuality = 0;
    398 
    399   // Use an arbitrary 4k limit on how far we will search.
    400   for (size_t i = 0, e = std::min(size_t(4096), Buffer.size()); i != e; ++i) {
    401     if (Buffer[i] == '\n')
    402       ++NumLinesForward;
    403 
    404     // Patterns have leading whitespace stripped, so skip whitespace when
    405     // looking for something which looks like a pattern.
    406     if (Buffer[i] == ' ' || Buffer[i] == '\t')
    407       continue;
    408 
    409     // Compute the "quality" of this match as an arbitrary combination of the
    410     // match distance and the number of lines skipped to get to this match.
    411     unsigned Distance = ComputeMatchDistance(Buffer.substr(i), VariableTable);
    412     double Quality = Distance + (NumLinesForward / 100.);
    413 
    414     if (Quality < BestQuality || Best == StringRef::npos) {
    415       Best = i;
    416       BestQuality = Quality;
    417     }
    418   }
    419 
    420   // Print the "possible intended match here" line if we found something
    421   // reasonable and not equal to what we showed in the "scanning from here"
    422   // line.
    423   if (Best && Best != StringRef::npos && BestQuality < 50) {
    424       SM.PrintMessage(SMLoc::getFromPointer(Buffer.data() + Best),
    425                       "possible intended match here", "note");
    426 
    427     // FIXME: If we wanted to be really friendly we would show why the match
    428     // failed, as it can be hard to spot simple one character differences.
    429   }
    430 }
    431 
    432 //===----------------------------------------------------------------------===//
    433 // Check Strings.
    434 //===----------------------------------------------------------------------===//
    435 
    436 /// CheckString - This is a check that we found in the input file.
    437 struct CheckString {
    438   /// Pat - The pattern to match.
    439   Pattern Pat;
    440 
    441   /// Loc - The location in the match file that the check string was specified.
    442   SMLoc Loc;
    443 
    444   /// IsCheckNext - This is true if this is a CHECK-NEXT: directive (as opposed
    445   /// to a CHECK: directive.
    446   bool IsCheckNext;
    447 
    448   /// NotStrings - These are all of the strings that are disallowed from
    449   /// occurring between this match string and the previous one (or start of
    450   /// file).
    451   std::vector<std::pair<SMLoc, Pattern> > NotStrings;
    452 
    453   CheckString(const Pattern &P, SMLoc L, bool isCheckNext)
    454     : Pat(P), Loc(L), IsCheckNext(isCheckNext) {}
    455 };
    456 
    457 /// CanonicalizeInputFile - Remove duplicate horizontal space from the specified
    458 /// memory buffer, free it, and return a new one.
    459 static MemoryBuffer *CanonicalizeInputFile(MemoryBuffer *MB) {
    460   SmallString<128> NewFile;
    461   NewFile.reserve(MB->getBufferSize());
    462 
    463   for (const char *Ptr = MB->getBufferStart(), *End = MB->getBufferEnd();
    464        Ptr != End; ++Ptr) {
    465     // Eliminate trailing dosish \r.
    466     if (Ptr <= End - 2 && Ptr[0] == '\r' && Ptr[1] == '\n') {
    467       continue;
    468     }
    469 
    470     // If C is not a horizontal whitespace, skip it.
    471     if (*Ptr != ' ' && *Ptr != '\t') {
    472       NewFile.push_back(*Ptr);
    473       continue;
    474     }
    475 
    476     // Otherwise, add one space and advance over neighboring space.
    477     NewFile.push_back(' ');
    478     while (Ptr+1 != End &&
    479            (Ptr[1] == ' ' || Ptr[1] == '\t'))
    480       ++Ptr;
    481   }
    482 
    483   // Free the old buffer and return a new one.
    484   MemoryBuffer *MB2 =
    485     MemoryBuffer::getMemBufferCopy(NewFile.str(), MB->getBufferIdentifier());
    486 
    487   delete MB;
    488   return MB2;
    489 }
    490 
    491 
    492 /// ReadCheckFile - Read the check file, which specifies the sequence of
    493 /// expected strings.  The strings are added to the CheckStrings vector.
    494 static bool ReadCheckFile(SourceMgr &SM,
    495                           std::vector<CheckString> &CheckStrings) {
    496   // Open the check file, and tell SourceMgr about it.
    497   OwningPtr<MemoryBuffer> File;
    498   if (error_code ec =
    499         MemoryBuffer::getFileOrSTDIN(CheckFilename.c_str(), File)) {
    500     errs() << "Could not open check file '" << CheckFilename << "': "
    501            << ec.message() << '\n';
    502     return true;
    503   }
    504   MemoryBuffer *F = File.take();
    505 
    506   // If we want to canonicalize whitespace, strip excess whitespace from the
    507   // buffer containing the CHECK lines.
    508   if (!NoCanonicalizeWhiteSpace)
    509     F = CanonicalizeInputFile(F);
    510 
    511   SM.AddNewSourceBuffer(F, SMLoc());
    512 
    513   // Find all instances of CheckPrefix followed by : in the file.
    514   StringRef Buffer = F->getBuffer();
    515 
    516   std::vector<std::pair<SMLoc, Pattern> > NotMatches;
    517 
    518   while (1) {
    519     // See if Prefix occurs in the memory buffer.
    520     Buffer = Buffer.substr(Buffer.find(CheckPrefix));
    521 
    522     // If we didn't find a match, we're done.
    523     if (Buffer.empty())
    524       break;
    525 
    526     const char *CheckPrefixStart = Buffer.data();
    527 
    528     // When we find a check prefix, keep track of whether we find CHECK: or
    529     // CHECK-NEXT:
    530     bool IsCheckNext = false, IsCheckNot = false;
    531 
    532     // Verify that the : is present after the prefix.
    533     if (Buffer[CheckPrefix.size()] == ':') {
    534       Buffer = Buffer.substr(CheckPrefix.size()+1);
    535     } else if (Buffer.size() > CheckPrefix.size()+6 &&
    536                memcmp(Buffer.data()+CheckPrefix.size(), "-NEXT:", 6) == 0) {
    537       Buffer = Buffer.substr(CheckPrefix.size()+7);
    538       IsCheckNext = true;
    539     } else if (Buffer.size() > CheckPrefix.size()+5 &&
    540                memcmp(Buffer.data()+CheckPrefix.size(), "-NOT:", 5) == 0) {
    541       Buffer = Buffer.substr(CheckPrefix.size()+6);
    542       IsCheckNot = true;
    543     } else {
    544       Buffer = Buffer.substr(1);
    545       continue;
    546     }
    547 
    548     // Okay, we found the prefix, yay.  Remember the rest of the line, but
    549     // ignore leading and trailing whitespace.
    550     Buffer = Buffer.substr(Buffer.find_first_not_of(" \t"));
    551 
    552     // Scan ahead to the end of line.
    553     size_t EOL = Buffer.find_first_of("\n\r");
    554 
    555     // Remember the location of the start of the pattern, for diagnostics.
    556     SMLoc PatternLoc = SMLoc::getFromPointer(Buffer.data());
    557 
    558     // Parse the pattern.
    559     Pattern P;
    560     if (P.ParsePattern(Buffer.substr(0, EOL), SM))
    561       return true;
    562 
    563     Buffer = Buffer.substr(EOL);
    564 
    565 
    566     // Verify that CHECK-NEXT lines have at least one CHECK line before them.
    567     if (IsCheckNext && CheckStrings.empty()) {
    568       SM.PrintMessage(SMLoc::getFromPointer(CheckPrefixStart),
    569                       "found '"+CheckPrefix+"-NEXT:' without previous '"+
    570                       CheckPrefix+ ": line", "error");
    571       return true;
    572     }
    573 
    574     // Handle CHECK-NOT.
    575     if (IsCheckNot) {
    576       NotMatches.push_back(std::make_pair(SMLoc::getFromPointer(Buffer.data()),
    577                                           P));
    578       continue;
    579     }
    580 
    581 
    582     // Okay, add the string we captured to the output vector and move on.
    583     CheckStrings.push_back(CheckString(P,
    584                                        PatternLoc,
    585                                        IsCheckNext));
    586     std::swap(NotMatches, CheckStrings.back().NotStrings);
    587   }
    588 
    589   // Add an EOF pattern for any trailing CHECK-NOTs.
    590   if (!NotMatches.empty()) {
    591     CheckStrings.push_back(CheckString(Pattern(true),
    592                                        SMLoc::getFromPointer(Buffer.data()),
    593                                        false));
    594     std::swap(NotMatches, CheckStrings.back().NotStrings);
    595   }
    596 
    597   if (CheckStrings.empty()) {
    598     errs() << "error: no check strings found with prefix '" << CheckPrefix
    599            << ":'\n";
    600     return true;
    601   }
    602 
    603   return false;
    604 }
    605 
    606 static void PrintCheckFailed(const SourceMgr &SM, const CheckString &CheckStr,
    607                              StringRef Buffer,
    608                              StringMap<StringRef> &VariableTable) {
    609   // Otherwise, we have an error, emit an error message.
    610   SM.PrintMessage(CheckStr.Loc, "expected string not found in input",
    611                   "error");
    612 
    613   // Print the "scanning from here" line.  If the current position is at the
    614   // end of a line, advance to the start of the next line.
    615   Buffer = Buffer.substr(Buffer.find_first_not_of(" \t\n\r"));
    616 
    617   SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), "scanning from here",
    618                   "note");
    619 
    620   // Allow the pattern to print additional information if desired.
    621   CheckStr.Pat.PrintFailureInfo(SM, Buffer, VariableTable);
    622 }
    623 
    624 /// CountNumNewlinesBetween - Count the number of newlines in the specified
    625 /// range.
    626 static unsigned CountNumNewlinesBetween(StringRef Range) {
    627   unsigned NumNewLines = 0;
    628   while (1) {
    629     // Scan for newline.
    630     Range = Range.substr(Range.find_first_of("\n\r"));
    631     if (Range.empty()) return NumNewLines;
    632 
    633     ++NumNewLines;
    634 
    635     // Handle \n\r and \r\n as a single newline.
    636     if (Range.size() > 1 &&
    637         (Range[1] == '\n' || Range[1] == '\r') &&
    638         (Range[0] != Range[1]))
    639       Range = Range.substr(1);
    640     Range = Range.substr(1);
    641   }
    642 }
    643 
    644 int main(int argc, char **argv) {
    645   sys::PrintStackTraceOnErrorSignal();
    646   PrettyStackTraceProgram X(argc, argv);
    647   cl::ParseCommandLineOptions(argc, argv);
    648 
    649   SourceMgr SM;
    650 
    651   // Read the expected strings from the check file.
    652   std::vector<CheckString> CheckStrings;
    653   if (ReadCheckFile(SM, CheckStrings))
    654     return 2;
    655 
    656   // Open the file to check and add it to SourceMgr.
    657   OwningPtr<MemoryBuffer> File;
    658   if (error_code ec =
    659         MemoryBuffer::getFileOrSTDIN(InputFilename.c_str(), File)) {
    660     errs() << "Could not open input file '" << InputFilename << "': "
    661            << ec.message() << '\n';
    662     return true;
    663   }
    664   MemoryBuffer *F = File.take();
    665 
    666   if (F->getBufferSize() == 0) {
    667     errs() << "FileCheck error: '" << InputFilename << "' is empty.\n";
    668     return 1;
    669   }
    670 
    671   // Remove duplicate spaces in the input file if requested.
    672   if (!NoCanonicalizeWhiteSpace)
    673     F = CanonicalizeInputFile(F);
    674 
    675   SM.AddNewSourceBuffer(F, SMLoc());
    676 
    677   /// VariableTable - This holds all the current filecheck variables.
    678   StringMap<StringRef> VariableTable;
    679 
    680   // Check that we have all of the expected strings, in order, in the input
    681   // file.
    682   StringRef Buffer = F->getBuffer();
    683 
    684   const char *LastMatch = Buffer.data();
    685 
    686   for (unsigned StrNo = 0, e = CheckStrings.size(); StrNo != e; ++StrNo) {
    687     const CheckString &CheckStr = CheckStrings[StrNo];
    688 
    689     StringRef SearchFrom = Buffer;
    690 
    691     // Find StrNo in the file.
    692     size_t MatchLen = 0;
    693     size_t MatchPos = CheckStr.Pat.Match(Buffer, MatchLen, VariableTable);
    694     Buffer = Buffer.substr(MatchPos);
    695 
    696     // If we didn't find a match, reject the input.
    697     if (MatchPos == StringRef::npos) {
    698       PrintCheckFailed(SM, CheckStr, SearchFrom, VariableTable);
    699       return 1;
    700     }
    701 
    702     StringRef SkippedRegion(LastMatch, Buffer.data()-LastMatch);
    703 
    704     // If this check is a "CHECK-NEXT", verify that the previous match was on
    705     // the previous line (i.e. that there is one newline between them).
    706     if (CheckStr.IsCheckNext) {
    707       // Count the number of newlines between the previous match and this one.
    708       assert(LastMatch != F->getBufferStart() &&
    709              "CHECK-NEXT can't be the first check in a file");
    710 
    711       unsigned NumNewLines = CountNumNewlinesBetween(SkippedRegion);
    712       if (NumNewLines == 0) {
    713         SM.PrintMessage(CheckStr.Loc,
    714                     CheckPrefix+"-NEXT: is on the same line as previous match",
    715                         "error");
    716         SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()),
    717                         "'next' match was here", "note");
    718         SM.PrintMessage(SMLoc::getFromPointer(LastMatch),
    719                         "previous match was here", "note");
    720         return 1;
    721       }
    722 
    723       if (NumNewLines != 1) {
    724         SM.PrintMessage(CheckStr.Loc,
    725                         CheckPrefix+
    726                         "-NEXT: is not on the line after the previous match",
    727                         "error");
    728         SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()),
    729                         "'next' match was here", "note");
    730         SM.PrintMessage(SMLoc::getFromPointer(LastMatch),
    731                         "previous match was here", "note");
    732         return 1;
    733       }
    734     }
    735 
    736     // If this match had "not strings", verify that they don't exist in the
    737     // skipped region.
    738     for (unsigned ChunkNo = 0, e = CheckStr.NotStrings.size();
    739          ChunkNo != e; ++ChunkNo) {
    740       size_t MatchLen = 0;
    741       size_t Pos = CheckStr.NotStrings[ChunkNo].second.Match(SkippedRegion,
    742                                                              MatchLen,
    743                                                              VariableTable);
    744       if (Pos == StringRef::npos) continue;
    745 
    746       SM.PrintMessage(SMLoc::getFromPointer(LastMatch+Pos),
    747                       CheckPrefix+"-NOT: string occurred!", "error");
    748       SM.PrintMessage(CheckStr.NotStrings[ChunkNo].first,
    749                       CheckPrefix+"-NOT: pattern specified here", "note");
    750       return 1;
    751     }
    752 
    753 
    754     // Otherwise, everything is good.  Step over the matched text and remember
    755     // the position after the match as the end of the last match.
    756     Buffer = Buffer.substr(MatchLen);
    757     LastMatch = Buffer.data();
    758   }
    759 
    760   return 0;
    761 }
    762