1 //===- SourceMgr.cpp - Manager for Simple Source Buffers & Diagnostics ----===//
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 // This file implements the SourceMgr class. This class is used as a simple
11 // substrate for diagnostics, #include handling, and other low level things for
12 // simple parsers.
13 //
14 //===----------------------------------------------------------------------===//
15
16 #include "llvm/Support/SourceMgr.h"
17 #include "llvm/ADT/Twine.h"
18 #include "llvm/Support/MemoryBuffer.h"
19 #include "llvm/Support/Path.h"
20 #include "llvm/Support/raw_ostream.h"
21 using namespace llvm;
22
23 static const size_t TabStop = 8;
24
25 namespace {
26 struct LineNoCacheTy {
27 unsigned LastQueryBufferID;
28 const char *LastQuery;
29 unsigned LineNoOfQuery;
30 };
31 }
32
getCache(void * Ptr)33 static LineNoCacheTy *getCache(void *Ptr) {
34 return (LineNoCacheTy*)Ptr;
35 }
36
37
~SourceMgr()38 SourceMgr::~SourceMgr() {
39 // Delete the line # cache if allocated.
40 if (LineNoCacheTy *Cache = getCache(LineNoCache))
41 delete Cache;
42 }
43
AddIncludeFile(const std::string & Filename,SMLoc IncludeLoc,std::string & IncludedFile)44 unsigned SourceMgr::AddIncludeFile(const std::string &Filename,
45 SMLoc IncludeLoc,
46 std::string &IncludedFile) {
47 IncludedFile = Filename;
48 ErrorOr<std::unique_ptr<MemoryBuffer>> NewBufOrErr =
49 MemoryBuffer::getFile(IncludedFile);
50
51 // If the file didn't exist directly, see if it's in an include path.
52 for (unsigned i = 0, e = IncludeDirectories.size(); i != e && !NewBufOrErr;
53 ++i) {
54 IncludedFile =
55 IncludeDirectories[i] + sys::path::get_separator().data() + Filename;
56 NewBufOrErr = MemoryBuffer::getFile(IncludedFile);
57 }
58
59 if (!NewBufOrErr)
60 return 0;
61
62 return AddNewSourceBuffer(std::move(*NewBufOrErr), IncludeLoc);
63 }
64
FindBufferContainingLoc(SMLoc Loc) const65 unsigned SourceMgr::FindBufferContainingLoc(SMLoc Loc) const {
66 for (unsigned i = 0, e = Buffers.size(); i != e; ++i)
67 if (Loc.getPointer() >= Buffers[i].Buffer->getBufferStart() &&
68 // Use <= here so that a pointer to the null at the end of the buffer
69 // is included as part of the buffer.
70 Loc.getPointer() <= Buffers[i].Buffer->getBufferEnd())
71 return i + 1;
72 return 0;
73 }
74
75 std::pair<unsigned, unsigned>
getLineAndColumn(SMLoc Loc,unsigned BufferID) const76 SourceMgr::getLineAndColumn(SMLoc Loc, unsigned BufferID) const {
77 if (!BufferID)
78 BufferID = FindBufferContainingLoc(Loc);
79 assert(BufferID && "Invalid Location!");
80
81 const MemoryBuffer *Buff = getMemoryBuffer(BufferID);
82
83 // Count the number of \n's between the start of the file and the specified
84 // location.
85 unsigned LineNo = 1;
86
87 const char *BufStart = Buff->getBufferStart();
88 const char *Ptr = BufStart;
89
90 // If we have a line number cache, and if the query is to a later point in the
91 // same file, start searching from the last query location. This optimizes
92 // for the case when multiple diagnostics come out of one file in order.
93 if (LineNoCacheTy *Cache = getCache(LineNoCache))
94 if (Cache->LastQueryBufferID == BufferID &&
95 Cache->LastQuery <= Loc.getPointer()) {
96 Ptr = Cache->LastQuery;
97 LineNo = Cache->LineNoOfQuery;
98 }
99
100 // Scan for the location being queried, keeping track of the number of lines
101 // we see.
102 for (; SMLoc::getFromPointer(Ptr) != Loc; ++Ptr)
103 if (*Ptr == '\n') ++LineNo;
104
105 // Allocate the line number cache if it doesn't exist.
106 if (!LineNoCache)
107 LineNoCache = new LineNoCacheTy();
108
109 // Update the line # cache.
110 LineNoCacheTy &Cache = *getCache(LineNoCache);
111 Cache.LastQueryBufferID = BufferID;
112 Cache.LastQuery = Ptr;
113 Cache.LineNoOfQuery = LineNo;
114
115 size_t NewlineOffs = StringRef(BufStart, Ptr-BufStart).find_last_of("\n\r");
116 if (NewlineOffs == StringRef::npos) NewlineOffs = ~(size_t)0;
117 return std::make_pair(LineNo, Ptr-BufStart-NewlineOffs);
118 }
119
PrintIncludeStack(SMLoc IncludeLoc,raw_ostream & OS) const120 void SourceMgr::PrintIncludeStack(SMLoc IncludeLoc, raw_ostream &OS) const {
121 if (IncludeLoc == SMLoc()) return; // Top of stack.
122
123 unsigned CurBuf = FindBufferContainingLoc(IncludeLoc);
124 assert(CurBuf && "Invalid or unspecified location!");
125
126 PrintIncludeStack(getBufferInfo(CurBuf).IncludeLoc, OS);
127
128 OS << "Included from "
129 << getBufferInfo(CurBuf).Buffer->getBufferIdentifier()
130 << ":" << FindLineNumber(IncludeLoc, CurBuf) << ":\n";
131 }
132
133
GetMessage(SMLoc Loc,SourceMgr::DiagKind Kind,const Twine & Msg,ArrayRef<SMRange> Ranges,ArrayRef<SMFixIt> FixIts) const134 SMDiagnostic SourceMgr::GetMessage(SMLoc Loc, SourceMgr::DiagKind Kind,
135 const Twine &Msg,
136 ArrayRef<SMRange> Ranges,
137 ArrayRef<SMFixIt> FixIts) const {
138
139 // First thing to do: find the current buffer containing the specified
140 // location to pull out the source line.
141 SmallVector<std::pair<unsigned, unsigned>, 4> ColRanges;
142 std::pair<unsigned, unsigned> LineAndCol;
143 const char *BufferID = "<unknown>";
144 std::string LineStr;
145
146 if (Loc.isValid()) {
147 unsigned CurBuf = FindBufferContainingLoc(Loc);
148 assert(CurBuf && "Invalid or unspecified location!");
149
150 const MemoryBuffer *CurMB = getMemoryBuffer(CurBuf);
151 BufferID = CurMB->getBufferIdentifier();
152
153 // Scan backward to find the start of the line.
154 const char *LineStart = Loc.getPointer();
155 const char *BufStart = CurMB->getBufferStart();
156 while (LineStart != BufStart && LineStart[-1] != '\n' &&
157 LineStart[-1] != '\r')
158 --LineStart;
159
160 // Get the end of the line.
161 const char *LineEnd = Loc.getPointer();
162 const char *BufEnd = CurMB->getBufferEnd();
163 while (LineEnd != BufEnd && LineEnd[0] != '\n' && LineEnd[0] != '\r')
164 ++LineEnd;
165 LineStr = std::string(LineStart, LineEnd);
166
167 // Convert any ranges to column ranges that only intersect the line of the
168 // location.
169 for (unsigned i = 0, e = Ranges.size(); i != e; ++i) {
170 SMRange R = Ranges[i];
171 if (!R.isValid()) continue;
172
173 // If the line doesn't contain any part of the range, then ignore it.
174 if (R.Start.getPointer() > LineEnd || R.End.getPointer() < LineStart)
175 continue;
176
177 // Ignore pieces of the range that go onto other lines.
178 if (R.Start.getPointer() < LineStart)
179 R.Start = SMLoc::getFromPointer(LineStart);
180 if (R.End.getPointer() > LineEnd)
181 R.End = SMLoc::getFromPointer(LineEnd);
182
183 // Translate from SMLoc ranges to column ranges.
184 // FIXME: Handle multibyte characters.
185 ColRanges.push_back(std::make_pair(R.Start.getPointer()-LineStart,
186 R.End.getPointer()-LineStart));
187 }
188
189 LineAndCol = getLineAndColumn(Loc, CurBuf);
190 }
191
192 return SMDiagnostic(*this, Loc, BufferID, LineAndCol.first,
193 LineAndCol.second-1, Kind, Msg.str(),
194 LineStr, ColRanges, FixIts);
195 }
196
PrintMessage(raw_ostream & OS,const SMDiagnostic & Diagnostic,bool ShowColors) const197 void SourceMgr::PrintMessage(raw_ostream &OS, const SMDiagnostic &Diagnostic,
198 bool ShowColors) const {
199 // Report the message with the diagnostic handler if present.
200 if (DiagHandler) {
201 DiagHandler(Diagnostic, DiagContext);
202 return;
203 }
204
205 if (Diagnostic.getLoc().isValid()) {
206 unsigned CurBuf = FindBufferContainingLoc(Diagnostic.getLoc());
207 assert(CurBuf && "Invalid or unspecified location!");
208 PrintIncludeStack(getBufferInfo(CurBuf).IncludeLoc, OS);
209 }
210
211 Diagnostic.print(nullptr, OS, ShowColors);
212 }
213
PrintMessage(raw_ostream & OS,SMLoc Loc,SourceMgr::DiagKind Kind,const Twine & Msg,ArrayRef<SMRange> Ranges,ArrayRef<SMFixIt> FixIts,bool ShowColors) const214 void SourceMgr::PrintMessage(raw_ostream &OS, SMLoc Loc,
215 SourceMgr::DiagKind Kind,
216 const Twine &Msg, ArrayRef<SMRange> Ranges,
217 ArrayRef<SMFixIt> FixIts, bool ShowColors) const {
218 PrintMessage(OS, GetMessage(Loc, Kind, Msg, Ranges, FixIts), ShowColors);
219 }
220
PrintMessage(SMLoc Loc,SourceMgr::DiagKind Kind,const Twine & Msg,ArrayRef<SMRange> Ranges,ArrayRef<SMFixIt> FixIts,bool ShowColors) const221 void SourceMgr::PrintMessage(SMLoc Loc, SourceMgr::DiagKind Kind,
222 const Twine &Msg, ArrayRef<SMRange> Ranges,
223 ArrayRef<SMFixIt> FixIts, bool ShowColors) const {
224 PrintMessage(llvm::errs(), Loc, Kind, Msg, Ranges, FixIts, ShowColors);
225 }
226
227 //===----------------------------------------------------------------------===//
228 // SMDiagnostic Implementation
229 //===----------------------------------------------------------------------===//
230
SMDiagnostic(const SourceMgr & sm,SMLoc L,StringRef FN,int Line,int Col,SourceMgr::DiagKind Kind,StringRef Msg,StringRef LineStr,ArrayRef<std::pair<unsigned,unsigned>> Ranges,ArrayRef<SMFixIt> Hints)231 SMDiagnostic::SMDiagnostic(const SourceMgr &sm, SMLoc L, StringRef FN,
232 int Line, int Col, SourceMgr::DiagKind Kind,
233 StringRef Msg, StringRef LineStr,
234 ArrayRef<std::pair<unsigned,unsigned> > Ranges,
235 ArrayRef<SMFixIt> Hints)
236 : SM(&sm), Loc(L), Filename(FN), LineNo(Line), ColumnNo(Col), Kind(Kind),
237 Message(Msg), LineContents(LineStr), Ranges(Ranges.vec()),
238 FixIts(Hints.begin(), Hints.end()) {
239 std::sort(FixIts.begin(), FixIts.end());
240 }
241
buildFixItLine(std::string & CaretLine,std::string & FixItLine,ArrayRef<SMFixIt> FixIts,ArrayRef<char> SourceLine)242 static void buildFixItLine(std::string &CaretLine, std::string &FixItLine,
243 ArrayRef<SMFixIt> FixIts, ArrayRef<char> SourceLine){
244 if (FixIts.empty())
245 return;
246
247 const char *LineStart = SourceLine.begin();
248 const char *LineEnd = SourceLine.end();
249
250 size_t PrevHintEndCol = 0;
251
252 for (ArrayRef<SMFixIt>::iterator I = FixIts.begin(), E = FixIts.end();
253 I != E; ++I) {
254 // If the fixit contains a newline or tab, ignore it.
255 if (I->getText().find_first_of("\n\r\t") != StringRef::npos)
256 continue;
257
258 SMRange R = I->getRange();
259
260 // If the line doesn't contain any part of the range, then ignore it.
261 if (R.Start.getPointer() > LineEnd || R.End.getPointer() < LineStart)
262 continue;
263
264 // Translate from SMLoc to column.
265 // Ignore pieces of the range that go onto other lines.
266 // FIXME: Handle multibyte characters in the source line.
267 unsigned FirstCol;
268 if (R.Start.getPointer() < LineStart)
269 FirstCol = 0;
270 else
271 FirstCol = R.Start.getPointer() - LineStart;
272
273 // If we inserted a long previous hint, push this one forwards, and add
274 // an extra space to show that this is not part of the previous
275 // completion. This is sort of the best we can do when two hints appear
276 // to overlap.
277 //
278 // Note that if this hint is located immediately after the previous
279 // hint, no space will be added, since the location is more important.
280 unsigned HintCol = FirstCol;
281 if (HintCol < PrevHintEndCol)
282 HintCol = PrevHintEndCol + 1;
283
284 // This relies on one byte per column in our fixit hints.
285 unsigned LastColumnModified = HintCol + I->getText().size();
286 if (LastColumnModified > FixItLine.size())
287 FixItLine.resize(LastColumnModified, ' ');
288
289 std::copy(I->getText().begin(), I->getText().end(),
290 FixItLine.begin() + HintCol);
291
292 PrevHintEndCol = LastColumnModified;
293
294 // For replacements, mark the removal range with '~'.
295 // FIXME: Handle multibyte characters in the source line.
296 unsigned LastCol;
297 if (R.End.getPointer() >= LineEnd)
298 LastCol = LineEnd - LineStart;
299 else
300 LastCol = R.End.getPointer() - LineStart;
301
302 std::fill(&CaretLine[FirstCol], &CaretLine[LastCol], '~');
303 }
304 }
305
printSourceLine(raw_ostream & S,StringRef LineContents)306 static void printSourceLine(raw_ostream &S, StringRef LineContents) {
307 // Print out the source line one character at a time, so we can expand tabs.
308 for (unsigned i = 0, e = LineContents.size(), OutCol = 0; i != e; ++i) {
309 if (LineContents[i] != '\t') {
310 S << LineContents[i];
311 ++OutCol;
312 continue;
313 }
314
315 // If we have a tab, emit at least one space, then round up to 8 columns.
316 do {
317 S << ' ';
318 ++OutCol;
319 } while ((OutCol % TabStop) != 0);
320 }
321 S << '\n';
322 }
323
isNonASCII(char c)324 static bool isNonASCII(char c) {
325 return c & 0x80;
326 }
327
print(const char * ProgName,raw_ostream & S,bool ShowColors,bool ShowKindLabel) const328 void SMDiagnostic::print(const char *ProgName, raw_ostream &S, bool ShowColors,
329 bool ShowKindLabel) const {
330 // Display colors only if OS supports colors.
331 ShowColors &= S.has_colors();
332
333 if (ShowColors)
334 S.changeColor(raw_ostream::SAVEDCOLOR, true);
335
336 if (ProgName && ProgName[0])
337 S << ProgName << ": ";
338
339 if (!Filename.empty()) {
340 if (Filename == "-")
341 S << "<stdin>";
342 else
343 S << Filename;
344
345 if (LineNo != -1) {
346 S << ':' << LineNo;
347 if (ColumnNo != -1)
348 S << ':' << (ColumnNo+1);
349 }
350 S << ": ";
351 }
352
353 if (ShowKindLabel) {
354 switch (Kind) {
355 case SourceMgr::DK_Error:
356 if (ShowColors)
357 S.changeColor(raw_ostream::RED, true);
358 S << "error: ";
359 break;
360 case SourceMgr::DK_Warning:
361 if (ShowColors)
362 S.changeColor(raw_ostream::MAGENTA, true);
363 S << "warning: ";
364 break;
365 case SourceMgr::DK_Note:
366 if (ShowColors)
367 S.changeColor(raw_ostream::BLACK, true);
368 S << "note: ";
369 break;
370 }
371
372 if (ShowColors) {
373 S.resetColor();
374 S.changeColor(raw_ostream::SAVEDCOLOR, true);
375 }
376 }
377
378 S << Message << '\n';
379
380 if (ShowColors)
381 S.resetColor();
382
383 if (LineNo == -1 || ColumnNo == -1)
384 return;
385
386 // FIXME: If there are multibyte or multi-column characters in the source, all
387 // our ranges will be wrong. To do this properly, we'll need a byte-to-column
388 // map like Clang's TextDiagnostic. For now, we'll just handle tabs by
389 // expanding them later, and bail out rather than show incorrect ranges and
390 // misaligned fixits for any other odd characters.
391 if (std::find_if(LineContents.begin(), LineContents.end(), isNonASCII) !=
392 LineContents.end()) {
393 printSourceLine(S, LineContents);
394 return;
395 }
396 size_t NumColumns = LineContents.size();
397
398 // Build the line with the caret and ranges.
399 std::string CaretLine(NumColumns+1, ' ');
400
401 // Expand any ranges.
402 for (unsigned r = 0, e = Ranges.size(); r != e; ++r) {
403 std::pair<unsigned, unsigned> R = Ranges[r];
404 std::fill(&CaretLine[R.first],
405 &CaretLine[std::min((size_t)R.second, CaretLine.size())],
406 '~');
407 }
408
409 // Add any fix-its.
410 // FIXME: Find the beginning of the line properly for multibyte characters.
411 std::string FixItInsertionLine;
412 buildFixItLine(CaretLine, FixItInsertionLine, FixIts,
413 makeArrayRef(Loc.getPointer() - ColumnNo,
414 LineContents.size()));
415
416 // Finally, plop on the caret.
417 if (unsigned(ColumnNo) <= NumColumns)
418 CaretLine[ColumnNo] = '^';
419 else
420 CaretLine[NumColumns] = '^';
421
422 // ... and remove trailing whitespace so the output doesn't wrap for it. We
423 // know that the line isn't completely empty because it has the caret in it at
424 // least.
425 CaretLine.erase(CaretLine.find_last_not_of(' ')+1);
426
427 printSourceLine(S, LineContents);
428
429 if (ShowColors)
430 S.changeColor(raw_ostream::GREEN, true);
431
432 // Print out the caret line, matching tabs in the source line.
433 for (unsigned i = 0, e = CaretLine.size(), OutCol = 0; i != e; ++i) {
434 if (i >= LineContents.size() || LineContents[i] != '\t') {
435 S << CaretLine[i];
436 ++OutCol;
437 continue;
438 }
439
440 // Okay, we have a tab. Insert the appropriate number of characters.
441 do {
442 S << CaretLine[i];
443 ++OutCol;
444 } while ((OutCol % TabStop) != 0);
445 }
446 S << '\n';
447
448 if (ShowColors)
449 S.resetColor();
450
451 // Print out the replacement line, matching tabs in the source line.
452 if (FixItInsertionLine.empty())
453 return;
454
455 for (size_t i = 0, e = FixItInsertionLine.size(), OutCol = 0; i < e; ++i) {
456 if (i >= LineContents.size() || LineContents[i] != '\t') {
457 S << FixItInsertionLine[i];
458 ++OutCol;
459 continue;
460 }
461
462 // Okay, we have a tab. Insert the appropriate number of characters.
463 do {
464 S << FixItInsertionLine[i];
465 // FIXME: This is trying not to break up replacements, but then to re-sync
466 // with the tabs between replacements. This will fail, though, if two
467 // fix-it replacements are exactly adjacent, or if a fix-it contains a
468 // space. Really we should be precomputing column widths, which we'll
469 // need anyway for multibyte chars.
470 if (FixItInsertionLine[i] != ' ')
471 ++i;
472 ++OutCol;
473 } while (((OutCol % TabStop) != 0) && i != e);
474 }
475 S << '\n';
476 }
477