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/Support/CommandLine.h"
20 #include "llvm/Support/MemoryBuffer.h"
21 #include "llvm/Support/PrettyStackTrace.h"
22 #include "llvm/Support/Regex.h"
23 #include "llvm/Support/SourceMgr.h"
24 #include "llvm/Support/raw_ostream.h"
25 #include "llvm/System/Signals.h"
28 static cl::opt<std::string>
29 CheckFilename(cl::Positional, cl::desc("<check-file>"), cl::Required);
31 static cl::opt<std::string>
32 InputFilename("input-file", cl::desc("File to check (defaults to stdin)"),
33 cl::init("-"), cl::value_desc("filename"));
35 static cl::opt<std::string>
36 CheckPrefix("check-prefix", cl::init("CHECK"),
37 cl::desc("Prefix to use from check file (defaults to 'CHECK')"));
40 NoCanonicalizeWhiteSpace("strict-whitespace",
41 cl::desc("Do not treat all horizontal whitespace as equivalent"));
43 //===----------------------------------------------------------------------===//
44 // Pattern Handling Code.
45 //===----------------------------------------------------------------------===//
51 PatternChunk(StringRef S, bool isRE) : Str(S), isRegEx(isRE) {}
54 size_t Match(StringRef Buffer, size_t &MatchLen) const {
56 // Fixed string match.
57 MatchLen = Str.size();
58 return Buffer.find(Str);
62 SmallVector<StringRef, 4> MatchInfo;
63 if (!Regex(Str, Regex::Sub).match(Buffer, &MatchInfo))
64 return StringRef::npos;
66 // Successful regex match.
67 assert(!MatchInfo.empty() && "Didn't get any match");
68 StringRef FullMatch = MatchInfo[0];
70 MatchLen = FullMatch.size();
71 return FullMatch.data()-Buffer.data();
77 /// Chunks - The pattern chunks to match. If the bool is false, it is a fixed
78 /// string match, if it is true, it is a regex match.
79 SmallVector<PatternChunk, 4> Chunks;
84 bool ParsePattern(StringRef PatternStr, SourceMgr &SM);
86 /// Match - Match the pattern string against the input buffer Buffer. This
87 /// returns the position that is matched or npos if there is no match. If
88 /// there is a match, the size of the matched string is returned in MatchLen.
89 size_t Match(StringRef Buffer, size_t &MatchLen) const;
92 bool Pattern::ParsePattern(StringRef PatternStr, SourceMgr &SM) {
93 // Ignore trailing whitespace.
94 while (!PatternStr.empty() &&
95 (PatternStr.back() == ' ' || PatternStr.back() == '\t'))
96 PatternStr = PatternStr.substr(0, PatternStr.size()-1);
98 // Check that there is something on the line.
99 if (PatternStr.empty()) {
100 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
101 "found empty check string with prefix '"+CheckPrefix+":'",
106 // Scan the pattern to break it into regex and non-regex pieces.
107 while (!PatternStr.empty()) {
108 // Handle fixed string matches.
109 if (PatternStr.size() < 2 ||
110 PatternStr[0] != '{' || PatternStr[1] != '{') {
111 // Find the end, which is the start of the next regex.
112 size_t FixedMatchEnd = PatternStr.find("{{");
114 Chunks.push_back(PatternChunk(PatternStr.substr(0, FixedMatchEnd),false));
115 PatternStr = PatternStr.substr(FixedMatchEnd);
119 // Otherwise, this is the start of a regex match. Scan for the }}.
120 size_t End = PatternStr.find("}}");
121 if (End == StringRef::npos) {
122 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()),
123 "found start of regex string with no end '}}'", "error");
127 Regex R(PatternStr.substr(2, End-2));
129 if (!R.isValid(Error)) {
130 SM.PrintMessage(SMLoc::getFromPointer(PatternStr.data()+2),
131 "invalid regex: " + Error, "error");
135 Chunks.push_back(PatternChunk(PatternStr.substr(2, End-2), true));
136 PatternStr = PatternStr.substr(End+2);
142 /// Match - Match the pattern string against the input buffer Buffer. This
143 /// returns the position that is matched or npos if there is no match. If
144 /// there is a match, the size of the matched string is returned in MatchLen.
145 size_t Pattern::Match(StringRef Buffer, size_t &MatchLen) const {
146 size_t FirstMatch = StringRef::npos;
149 while (!Buffer.empty()) {
150 StringRef MatchAttempt = Buffer;
152 unsigned ChunkNo = 0, e = Chunks.size();
153 for (; ChunkNo != e; ++ChunkNo) {
154 size_t ThisMatch, ThisLength = StringRef::npos;
155 ThisMatch = Chunks[ChunkNo].Match(MatchAttempt, ThisLength);
157 // Otherwise, what we do depends on if this is the first match or not. If
158 // this is the first match, it doesn't match to match at the start of
161 // If the first match fails then this pattern will never match in
163 if (ThisMatch == StringRef::npos)
166 FirstMatch = ThisMatch;
167 MatchAttempt = MatchAttempt.substr(FirstMatch);
171 // If this chunk didn't match, then the entire pattern didn't match from
172 // FirstMatch, try later in the buffer.
173 if (ThisMatch == StringRef::npos)
176 // Ok, if the match didn't match at the beginning of MatchAttempt, then we
177 // have something like "ABC{{DEF}} and something was in-between. Reject
182 // Otherwise, match the string and move to the next chunk.
183 MatchLen += ThisLength;
184 MatchAttempt = MatchAttempt.substr(ThisLength);
187 // If the whole thing matched, we win.
191 // Otherwise, try matching again after FirstMatch to see if this pattern
192 // matches later in the buffer.
193 Buffer = Buffer.substr(FirstMatch+1);
196 // If we ran out of stuff to scan, then we didn't match.
197 return StringRef::npos;
201 //===----------------------------------------------------------------------===//
203 //===----------------------------------------------------------------------===//
205 /// CheckString - This is a check that we found in the input file.
207 /// Pat - The pattern to match.
210 /// Loc - The location in the match file that the check string was specified.
213 /// IsCheckNext - This is true if this is a CHECK-NEXT: directive (as opposed
214 /// to a CHECK: directive.
217 /// NotStrings - These are all of the strings that are disallowed from
218 /// occurring between this match string and the previous one (or start of
220 std::vector<std::pair<SMLoc, Pattern> > NotStrings;
222 CheckString(const Pattern &P, SMLoc L, bool isCheckNext)
223 : Pat(P), Loc(L), IsCheckNext(isCheckNext) {}
226 /// CanonicalizeInputFile - Remove duplicate horizontal space from the specified
227 /// memory buffer, free it, and return a new one.
228 static MemoryBuffer *CanonicalizeInputFile(MemoryBuffer *MB) {
229 SmallVector<char, 16> NewFile;
230 NewFile.reserve(MB->getBufferSize());
232 for (const char *Ptr = MB->getBufferStart(), *End = MB->getBufferEnd();
234 // If C is not a horizontal whitespace, skip it.
235 if (*Ptr != ' ' && *Ptr != '\t') {
236 NewFile.push_back(*Ptr);
240 // Otherwise, add one space and advance over neighboring space.
241 NewFile.push_back(' ');
242 while (Ptr+1 != End &&
243 (Ptr[1] == ' ' || Ptr[1] == '\t'))
247 // Free the old buffer and return a new one.
249 MemoryBuffer::getMemBufferCopy(NewFile.data(),
250 NewFile.data() + NewFile.size(),
251 MB->getBufferIdentifier());
258 /// ReadCheckFile - Read the check file, which specifies the sequence of
259 /// expected strings. The strings are added to the CheckStrings vector.
260 static bool ReadCheckFile(SourceMgr &SM,
261 std::vector<CheckString> &CheckStrings) {
262 // Open the check file, and tell SourceMgr about it.
263 std::string ErrorStr;
265 MemoryBuffer::getFileOrSTDIN(CheckFilename.c_str(), &ErrorStr);
267 errs() << "Could not open check file '" << CheckFilename << "': "
272 // If we want to canonicalize whitespace, strip excess whitespace from the
273 // buffer containing the CHECK lines.
274 if (!NoCanonicalizeWhiteSpace)
275 F = CanonicalizeInputFile(F);
277 SM.AddNewSourceBuffer(F, SMLoc());
279 // Find all instances of CheckPrefix followed by : in the file.
280 StringRef Buffer = F->getBuffer();
282 std::vector<std::pair<SMLoc, Pattern> > NotMatches;
285 // See if Prefix occurs in the memory buffer.
286 Buffer = Buffer.substr(Buffer.find(CheckPrefix));
288 // If we didn't find a match, we're done.
292 const char *CheckPrefixStart = Buffer.data();
294 // When we find a check prefix, keep track of whether we find CHECK: or
296 bool IsCheckNext = false, IsCheckNot = false;
298 // Verify that the : is present after the prefix.
299 if (Buffer[CheckPrefix.size()] == ':') {
300 Buffer = Buffer.substr(CheckPrefix.size()+1);
301 } else if (Buffer.size() > CheckPrefix.size()+6 &&
302 memcmp(Buffer.data()+CheckPrefix.size(), "-NEXT:", 6) == 0) {
303 Buffer = Buffer.substr(CheckPrefix.size()+7);
305 } else if (Buffer.size() > CheckPrefix.size()+5 &&
306 memcmp(Buffer.data()+CheckPrefix.size(), "-NOT:", 5) == 0) {
307 Buffer = Buffer.substr(CheckPrefix.size()+6);
310 Buffer = Buffer.substr(1);
314 // Okay, we found the prefix, yay. Remember the rest of the line, but
315 // ignore leading and trailing whitespace.
316 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t"));
318 // Scan ahead to the end of line.
319 size_t EOL = Buffer.find_first_of("\n\r");
321 // Parse the pattern.
323 if (P.ParsePattern(Buffer.substr(0, EOL), SM))
326 Buffer = Buffer.substr(EOL);
329 // Verify that CHECK-NEXT lines have at least one CHECK line before them.
330 if (IsCheckNext && CheckStrings.empty()) {
331 SM.PrintMessage(SMLoc::getFromPointer(CheckPrefixStart),
332 "found '"+CheckPrefix+"-NEXT:' without previous '"+
333 CheckPrefix+ ": line", "error");
339 NotMatches.push_back(std::make_pair(SMLoc::getFromPointer(Buffer.data()),
345 // Okay, add the string we captured to the output vector and move on.
346 CheckStrings.push_back(CheckString(P,
347 SMLoc::getFromPointer(Buffer.data()),
349 std::swap(NotMatches, CheckStrings.back().NotStrings);
352 if (CheckStrings.empty()) {
353 errs() << "error: no check strings found with prefix '" << CheckPrefix
358 if (!NotMatches.empty()) {
359 errs() << "error: '" << CheckPrefix
360 << "-NOT:' not supported after last check line.\n";
367 static void PrintCheckFailed(const SourceMgr &SM, const CheckString &CheckStr,
369 // Otherwise, we have an error, emit an error message.
370 SM.PrintMessage(CheckStr.Loc, "expected string not found in input",
373 // Print the "scanning from here" line. If the current position is at the
374 // end of a line, advance to the start of the next line.
375 Buffer = Buffer.substr(Buffer.find_first_not_of(" \t\n\r"));
377 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()), "scanning from here",
381 /// CountNumNewlinesBetween - Count the number of newlines in the specified
383 static unsigned CountNumNewlinesBetween(StringRef Range) {
384 unsigned NumNewLines = 0;
387 Range = Range.substr(Range.find_first_of("\n\r"));
388 if (Range.empty()) return NumNewLines;
392 // Handle \n\r and \r\n as a single newline.
393 if (Range.size() > 1 &&
394 (Range[1] == '\n' || Range[1] == '\r') &&
395 (Range[0] != Range[1]))
396 Range = Range.substr(1);
397 Range = Range.substr(1);
401 int main(int argc, char **argv) {
402 sys::PrintStackTraceOnErrorSignal();
403 PrettyStackTraceProgram X(argc, argv);
404 cl::ParseCommandLineOptions(argc, argv);
408 // Read the expected strings from the check file.
409 std::vector<CheckString> CheckStrings;
410 if (ReadCheckFile(SM, CheckStrings))
413 // Open the file to check and add it to SourceMgr.
414 std::string ErrorStr;
416 MemoryBuffer::getFileOrSTDIN(InputFilename.c_str(), &ErrorStr);
418 errs() << "Could not open input file '" << InputFilename << "': "
423 // Remove duplicate spaces in the input file if requested.
424 if (!NoCanonicalizeWhiteSpace)
425 F = CanonicalizeInputFile(F);
427 SM.AddNewSourceBuffer(F, SMLoc());
429 // Check that we have all of the expected strings, in order, in the input
431 StringRef Buffer = F->getBuffer();
433 const char *LastMatch = Buffer.data();
435 for (unsigned StrNo = 0, e = CheckStrings.size(); StrNo != e; ++StrNo) {
436 const CheckString &CheckStr = CheckStrings[StrNo];
438 StringRef SearchFrom = Buffer;
440 // Find StrNo in the file.
442 Buffer = Buffer.substr(CheckStr.Pat.Match(Buffer, MatchLen));
444 // If we didn't find a match, reject the input.
445 if (Buffer.empty()) {
446 PrintCheckFailed(SM, CheckStr, SearchFrom);
450 StringRef SkippedRegion(LastMatch, Buffer.data()-LastMatch);
452 // If this check is a "CHECK-NEXT", verify that the previous match was on
453 // the previous line (i.e. that there is one newline between them).
454 if (CheckStr.IsCheckNext) {
455 // Count the number of newlines between the previous match and this one.
456 assert(LastMatch != F->getBufferStart() &&
457 "CHECK-NEXT can't be the first check in a file");
459 unsigned NumNewLines = CountNumNewlinesBetween(SkippedRegion);
460 if (NumNewLines == 0) {
461 SM.PrintMessage(CheckStr.Loc,
462 CheckPrefix+"-NEXT: is on the same line as previous match",
464 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()),
465 "'next' match was here", "note");
466 SM.PrintMessage(SMLoc::getFromPointer(LastMatch),
467 "previous match was here", "note");
471 if (NumNewLines != 1) {
472 SM.PrintMessage(CheckStr.Loc,
474 "-NEXT: is not on the line after the previous match",
476 SM.PrintMessage(SMLoc::getFromPointer(Buffer.data()),
477 "'next' match was here", "note");
478 SM.PrintMessage(SMLoc::getFromPointer(LastMatch),
479 "previous match was here", "note");
484 // If this match had "not strings", verify that they don't exist in the
486 for (unsigned ChunkNo = 0, e = CheckStr.NotStrings.size(); ChunkNo != e; ++ChunkNo) {
488 size_t Pos = CheckStr.NotStrings[ChunkNo].second.Match(SkippedRegion, MatchLen);
489 if (Pos == StringRef::npos) continue;
491 SM.PrintMessage(SMLoc::getFromPointer(LastMatch+Pos),
492 CheckPrefix+"-NOT: string occurred!", "error");
493 SM.PrintMessage(CheckStr.NotStrings[ChunkNo].first,
494 CheckPrefix+"-NOT: pattern specified here", "note");
499 // Otherwise, everything is good. Step over the matched text and remember
500 // the position after the match as the end of the last match.
501 Buffer = Buffer.substr(MatchLen);
502 LastMatch = Buffer.data();