1 //===--- TextDiagnostic.cpp - Text Diagnostic Pretty-Printing -------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8
9 #include "clang/Frontend/TextDiagnostic.h"
10 #include "clang/Basic/CharInfo.h"
11 #include "clang/Basic/DiagnosticOptions.h"
12 #include "clang/Basic/FileManager.h"
13 #include "clang/Basic/SourceManager.h"
14 #include "clang/Lex/Lexer.h"
15 #include "llvm/ADT/SmallString.h"
16 #include "llvm/ADT/StringExtras.h"
17 #include "llvm/Support/ConvertUTF.h"
18 #include "llvm/Support/ErrorHandling.h"
19 #include "llvm/Support/Locale.h"
20 #include "llvm/Support/Path.h"
21 #include "llvm/Support/raw_ostream.h"
22 #include <algorithm>
23 #include <optional>
24
25 using namespace clang;
26
27 static const enum raw_ostream::Colors noteColor =
28 raw_ostream::BLACK;
29 static const enum raw_ostream::Colors remarkColor =
30 raw_ostream::BLUE;
31 static const enum raw_ostream::Colors fixitColor =
32 raw_ostream::GREEN;
33 static const enum raw_ostream::Colors caretColor =
34 raw_ostream::GREEN;
35 static const enum raw_ostream::Colors warningColor =
36 raw_ostream::MAGENTA;
37 static const enum raw_ostream::Colors templateColor =
38 raw_ostream::CYAN;
39 static const enum raw_ostream::Colors errorColor = raw_ostream::RED;
40 static const enum raw_ostream::Colors fatalColor = raw_ostream::RED;
41 // Used for changing only the bold attribute.
42 static const enum raw_ostream::Colors savedColor =
43 raw_ostream::SAVEDCOLOR;
44
45 /// Add highlights to differences in template strings.
applyTemplateHighlighting(raw_ostream & OS,StringRef Str,bool & Normal,bool Bold)46 static void applyTemplateHighlighting(raw_ostream &OS, StringRef Str,
47 bool &Normal, bool Bold) {
48 while (true) {
49 size_t Pos = Str.find(ToggleHighlight);
50 OS << Str.slice(0, Pos);
51 if (Pos == StringRef::npos)
52 break;
53
54 Str = Str.substr(Pos + 1);
55 if (Normal)
56 OS.changeColor(templateColor, true);
57 else {
58 OS.resetColor();
59 if (Bold)
60 OS.changeColor(savedColor, true);
61 }
62 Normal = !Normal;
63 }
64 }
65
66 /// Number of spaces to indent when word-wrapping.
67 const unsigned WordWrapIndentation = 6;
68
bytesSincePreviousTabOrLineBegin(StringRef SourceLine,size_t i)69 static int bytesSincePreviousTabOrLineBegin(StringRef SourceLine, size_t i) {
70 int bytes = 0;
71 while (0<i) {
72 if (SourceLine[--i]=='\t')
73 break;
74 ++bytes;
75 }
76 return bytes;
77 }
78
79 /// returns a printable representation of first item from input range
80 ///
81 /// This function returns a printable representation of the next item in a line
82 /// of source. If the next byte begins a valid and printable character, that
83 /// character is returned along with 'true'.
84 ///
85 /// Otherwise, if the next byte begins a valid, but unprintable character, a
86 /// printable, escaped representation of the character is returned, along with
87 /// 'false'. Otherwise a printable, escaped representation of the next byte
88 /// is returned along with 'false'.
89 ///
90 /// \note The index is updated to be used with a subsequent call to
91 /// printableTextForNextCharacter.
92 ///
93 /// \param SourceLine The line of source
94 /// \param i Pointer to byte index,
95 /// \param TabStop used to expand tabs
96 /// \return pair(printable text, 'true' iff original text was printable)
97 ///
98 static std::pair<SmallString<16>, bool>
printableTextForNextCharacter(StringRef SourceLine,size_t * i,unsigned TabStop)99 printableTextForNextCharacter(StringRef SourceLine, size_t *i,
100 unsigned TabStop) {
101 assert(i && "i must not be null");
102 assert(*i<SourceLine.size() && "must point to a valid index");
103
104 if (SourceLine[*i]=='\t') {
105 assert(0 < TabStop && TabStop <= DiagnosticOptions::MaxTabStop &&
106 "Invalid -ftabstop value");
107 unsigned col = bytesSincePreviousTabOrLineBegin(SourceLine, *i);
108 unsigned NumSpaces = TabStop - col%TabStop;
109 assert(0 < NumSpaces && NumSpaces <= TabStop
110 && "Invalid computation of space amt");
111 ++(*i);
112
113 SmallString<16> expandedTab;
114 expandedTab.assign(NumSpaces, ' ');
115 return std::make_pair(expandedTab, true);
116 }
117
118 unsigned char const *begin, *end;
119 begin = reinterpret_cast<unsigned char const *>(&*(SourceLine.begin() + *i));
120 end = begin + (SourceLine.size() - *i);
121
122 if (llvm::isLegalUTF8Sequence(begin, end)) {
123 llvm::UTF32 c;
124 llvm::UTF32 *cptr = &c;
125 unsigned char const *original_begin = begin;
126 unsigned char const *cp_end =
127 begin + llvm::getNumBytesForUTF8(SourceLine[*i]);
128
129 llvm::ConversionResult res = llvm::ConvertUTF8toUTF32(
130 &begin, cp_end, &cptr, cptr + 1, llvm::strictConversion);
131 (void)res;
132 assert(llvm::conversionOK == res);
133 assert(0 < begin-original_begin
134 && "we must be further along in the string now");
135 *i += begin-original_begin;
136
137 if (!llvm::sys::locale::isPrint(c)) {
138 // If next character is valid UTF-8, but not printable
139 SmallString<16> expandedCP("<U+>");
140 while (c) {
141 expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(c%16));
142 c/=16;
143 }
144 while (expandedCP.size() < 8)
145 expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(0));
146 return std::make_pair(expandedCP, false);
147 }
148
149 // If next character is valid UTF-8, and printable
150 return std::make_pair(SmallString<16>(original_begin, cp_end), true);
151
152 }
153
154 // If next byte is not valid UTF-8 (and therefore not printable)
155 SmallString<16> expandedByte("<XX>");
156 unsigned char byte = SourceLine[*i];
157 expandedByte[1] = llvm::hexdigit(byte / 16);
158 expandedByte[2] = llvm::hexdigit(byte % 16);
159 ++(*i);
160 return std::make_pair(expandedByte, false);
161 }
162
expandTabs(std::string & SourceLine,unsigned TabStop)163 static void expandTabs(std::string &SourceLine, unsigned TabStop) {
164 size_t i = SourceLine.size();
165 while (i>0) {
166 i--;
167 if (SourceLine[i]!='\t')
168 continue;
169 size_t tmp_i = i;
170 std::pair<SmallString<16>,bool> res
171 = printableTextForNextCharacter(SourceLine, &tmp_i, TabStop);
172 SourceLine.replace(i, 1, res.first.c_str());
173 }
174 }
175
176 /// This function takes a raw source line and produces a mapping from the bytes
177 /// of the printable representation of the line to the columns those printable
178 /// characters will appear at (numbering the first column as 0).
179 ///
180 /// If a byte 'i' corresponds to multiple columns (e.g. the byte contains a tab
181 /// character) then the array will map that byte to the first column the
182 /// tab appears at and the next value in the map will have been incremented
183 /// more than once.
184 ///
185 /// If a byte is the first in a sequence of bytes that together map to a single
186 /// entity in the output, then the array will map that byte to the appropriate
187 /// column while the subsequent bytes will be -1.
188 ///
189 /// The last element in the array does not correspond to any byte in the input
190 /// and instead is the number of columns needed to display the source
191 ///
192 /// example: (given a tabstop of 8)
193 ///
194 /// "a \t \u3042" -> {0,1,2,8,9,-1,-1,11}
195 ///
196 /// (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
197 /// display)
byteToColumn(StringRef SourceLine,unsigned TabStop,SmallVectorImpl<int> & out)198 static void byteToColumn(StringRef SourceLine, unsigned TabStop,
199 SmallVectorImpl<int> &out) {
200 out.clear();
201
202 if (SourceLine.empty()) {
203 out.resize(1u,0);
204 return;
205 }
206
207 out.resize(SourceLine.size()+1, -1);
208
209 int columns = 0;
210 size_t i = 0;
211 while (i<SourceLine.size()) {
212 out[i] = columns;
213 std::pair<SmallString<16>,bool> res
214 = printableTextForNextCharacter(SourceLine, &i, TabStop);
215 columns += llvm::sys::locale::columnWidth(res.first);
216 }
217 out.back() = columns;
218 }
219
220 /// This function takes a raw source line and produces a mapping from columns
221 /// to the byte of the source line that produced the character displaying at
222 /// that column. This is the inverse of the mapping produced by byteToColumn()
223 ///
224 /// The last element in the array is the number of bytes in the source string
225 ///
226 /// example: (given a tabstop of 8)
227 ///
228 /// "a \t \u3042" -> {0,1,2,-1,-1,-1,-1,-1,3,4,-1,7}
229 ///
230 /// (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
231 /// display)
columnToByte(StringRef SourceLine,unsigned TabStop,SmallVectorImpl<int> & out)232 static void columnToByte(StringRef SourceLine, unsigned TabStop,
233 SmallVectorImpl<int> &out) {
234 out.clear();
235
236 if (SourceLine.empty()) {
237 out.resize(1u, 0);
238 return;
239 }
240
241 int columns = 0;
242 size_t i = 0;
243 while (i<SourceLine.size()) {
244 out.resize(columns+1, -1);
245 out.back() = i;
246 std::pair<SmallString<16>,bool> res
247 = printableTextForNextCharacter(SourceLine, &i, TabStop);
248 columns += llvm::sys::locale::columnWidth(res.first);
249 }
250 out.resize(columns+1, -1);
251 out.back() = i;
252 }
253
254 namespace {
255 struct SourceColumnMap {
SourceColumnMap__anon13c803a00111::SourceColumnMap256 SourceColumnMap(StringRef SourceLine, unsigned TabStop)
257 : m_SourceLine(SourceLine) {
258
259 ::byteToColumn(SourceLine, TabStop, m_byteToColumn);
260 ::columnToByte(SourceLine, TabStop, m_columnToByte);
261
262 assert(m_byteToColumn.size()==SourceLine.size()+1);
263 assert(0 < m_byteToColumn.size() && 0 < m_columnToByte.size());
264 assert(m_byteToColumn.size()
265 == static_cast<unsigned>(m_columnToByte.back()+1));
266 assert(static_cast<unsigned>(m_byteToColumn.back()+1)
267 == m_columnToByte.size());
268 }
columns__anon13c803a00111::SourceColumnMap269 int columns() const { return m_byteToColumn.back(); }
bytes__anon13c803a00111::SourceColumnMap270 int bytes() const { return m_columnToByte.back(); }
271
272 /// Map a byte to the column which it is at the start of, or return -1
273 /// if it is not at the start of a column (for a UTF-8 trailing byte).
byteToColumn__anon13c803a00111::SourceColumnMap274 int byteToColumn(int n) const {
275 assert(0<=n && n<static_cast<int>(m_byteToColumn.size()));
276 return m_byteToColumn[n];
277 }
278
279 /// Map a byte to the first column which contains it.
byteToContainingColumn__anon13c803a00111::SourceColumnMap280 int byteToContainingColumn(int N) const {
281 assert(0 <= N && N < static_cast<int>(m_byteToColumn.size()));
282 while (m_byteToColumn[N] == -1)
283 --N;
284 return m_byteToColumn[N];
285 }
286
287 /// Map a column to the byte which starts the column, or return -1 if
288 /// the column the second or subsequent column of an expanded tab or similar
289 /// multi-column entity.
columnToByte__anon13c803a00111::SourceColumnMap290 int columnToByte(int n) const {
291 assert(0<=n && n<static_cast<int>(m_columnToByte.size()));
292 return m_columnToByte[n];
293 }
294
295 /// Map from a byte index to the next byte which starts a column.
startOfNextColumn__anon13c803a00111::SourceColumnMap296 int startOfNextColumn(int N) const {
297 assert(0 <= N && N < static_cast<int>(m_byteToColumn.size() - 1));
298 while (byteToColumn(++N) == -1) {}
299 return N;
300 }
301
302 /// Map from a byte index to the previous byte which starts a column.
startOfPreviousColumn__anon13c803a00111::SourceColumnMap303 int startOfPreviousColumn(int N) const {
304 assert(0 < N && N < static_cast<int>(m_byteToColumn.size()));
305 while (byteToColumn(--N) == -1) {}
306 return N;
307 }
308
getSourceLine__anon13c803a00111::SourceColumnMap309 StringRef getSourceLine() const {
310 return m_SourceLine;
311 }
312
313 private:
314 const std::string m_SourceLine;
315 SmallVector<int,200> m_byteToColumn;
316 SmallVector<int,200> m_columnToByte;
317 };
318 } // end anonymous namespace
319
320 /// When the source code line we want to print is too long for
321 /// the terminal, select the "interesting" region.
selectInterestingSourceRegion(std::string & SourceLine,std::string & CaretLine,std::string & FixItInsertionLine,unsigned Columns,const SourceColumnMap & map)322 static void selectInterestingSourceRegion(std::string &SourceLine,
323 std::string &CaretLine,
324 std::string &FixItInsertionLine,
325 unsigned Columns,
326 const SourceColumnMap &map) {
327 unsigned CaretColumns = CaretLine.size();
328 unsigned FixItColumns = llvm::sys::locale::columnWidth(FixItInsertionLine);
329 unsigned MaxColumns = std::max(static_cast<unsigned>(map.columns()),
330 std::max(CaretColumns, FixItColumns));
331 // if the number of columns is less than the desired number we're done
332 if (MaxColumns <= Columns)
333 return;
334
335 // No special characters are allowed in CaretLine.
336 assert(llvm::none_of(CaretLine, [](char c) { return c < ' ' || '~' < c; }));
337
338 // Find the slice that we need to display the full caret line
339 // correctly.
340 unsigned CaretStart = 0, CaretEnd = CaretLine.size();
341 for (; CaretStart != CaretEnd; ++CaretStart)
342 if (!isWhitespace(CaretLine[CaretStart]))
343 break;
344
345 for (; CaretEnd != CaretStart; --CaretEnd)
346 if (!isWhitespace(CaretLine[CaretEnd - 1]))
347 break;
348
349 // caret has already been inserted into CaretLine so the above whitespace
350 // check is guaranteed to include the caret
351
352 // If we have a fix-it line, make sure the slice includes all of the
353 // fix-it information.
354 if (!FixItInsertionLine.empty()) {
355 unsigned FixItStart = 0, FixItEnd = FixItInsertionLine.size();
356 for (; FixItStart != FixItEnd; ++FixItStart)
357 if (!isWhitespace(FixItInsertionLine[FixItStart]))
358 break;
359
360 for (; FixItEnd != FixItStart; --FixItEnd)
361 if (!isWhitespace(FixItInsertionLine[FixItEnd - 1]))
362 break;
363
364 // We can safely use the byte offset FixItStart as the column offset
365 // because the characters up until FixItStart are all ASCII whitespace
366 // characters.
367 unsigned FixItStartCol = FixItStart;
368 unsigned FixItEndCol
369 = llvm::sys::locale::columnWidth(FixItInsertionLine.substr(0, FixItEnd));
370
371 CaretStart = std::min(FixItStartCol, CaretStart);
372 CaretEnd = std::max(FixItEndCol, CaretEnd);
373 }
374
375 // CaretEnd may have been set at the middle of a character
376 // If it's not at a character's first column then advance it past the current
377 // character.
378 while (static_cast<int>(CaretEnd) < map.columns() &&
379 -1 == map.columnToByte(CaretEnd))
380 ++CaretEnd;
381
382 assert((static_cast<int>(CaretStart) > map.columns() ||
383 -1!=map.columnToByte(CaretStart)) &&
384 "CaretStart must not point to a column in the middle of a source"
385 " line character");
386 assert((static_cast<int>(CaretEnd) > map.columns() ||
387 -1!=map.columnToByte(CaretEnd)) &&
388 "CaretEnd must not point to a column in the middle of a source line"
389 " character");
390
391 // CaretLine[CaretStart, CaretEnd) contains all of the interesting
392 // parts of the caret line. While this slice is smaller than the
393 // number of columns we have, try to grow the slice to encompass
394 // more context.
395
396 unsigned SourceStart = map.columnToByte(std::min<unsigned>(CaretStart,
397 map.columns()));
398 unsigned SourceEnd = map.columnToByte(std::min<unsigned>(CaretEnd,
399 map.columns()));
400
401 unsigned CaretColumnsOutsideSource = CaretEnd-CaretStart
402 - (map.byteToColumn(SourceEnd)-map.byteToColumn(SourceStart));
403
404 char const *front_ellipse = " ...";
405 char const *front_space = " ";
406 char const *back_ellipse = "...";
407 unsigned ellipses_space = strlen(front_ellipse) + strlen(back_ellipse);
408
409 unsigned TargetColumns = Columns;
410 // Give us extra room for the ellipses
411 // and any of the caret line that extends past the source
412 if (TargetColumns > ellipses_space+CaretColumnsOutsideSource)
413 TargetColumns -= ellipses_space+CaretColumnsOutsideSource;
414
415 while (SourceStart>0 || SourceEnd<SourceLine.size()) {
416 bool ExpandedRegion = false;
417
418 if (SourceStart>0) {
419 unsigned NewStart = map.startOfPreviousColumn(SourceStart);
420
421 // Skip over any whitespace we see here; we're looking for
422 // another bit of interesting text.
423 // FIXME: Detect non-ASCII whitespace characters too.
424 while (NewStart && isWhitespace(SourceLine[NewStart]))
425 NewStart = map.startOfPreviousColumn(NewStart);
426
427 // Skip over this bit of "interesting" text.
428 while (NewStart) {
429 unsigned Prev = map.startOfPreviousColumn(NewStart);
430 if (isWhitespace(SourceLine[Prev]))
431 break;
432 NewStart = Prev;
433 }
434
435 assert(map.byteToColumn(NewStart) != -1);
436 unsigned NewColumns = map.byteToColumn(SourceEnd) -
437 map.byteToColumn(NewStart);
438 if (NewColumns <= TargetColumns) {
439 SourceStart = NewStart;
440 ExpandedRegion = true;
441 }
442 }
443
444 if (SourceEnd<SourceLine.size()) {
445 unsigned NewEnd = map.startOfNextColumn(SourceEnd);
446
447 // Skip over any whitespace we see here; we're looking for
448 // another bit of interesting text.
449 // FIXME: Detect non-ASCII whitespace characters too.
450 while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
451 NewEnd = map.startOfNextColumn(NewEnd);
452
453 // Skip over this bit of "interesting" text.
454 while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
455 NewEnd = map.startOfNextColumn(NewEnd);
456
457 assert(map.byteToColumn(NewEnd) != -1);
458 unsigned NewColumns = map.byteToColumn(NewEnd) -
459 map.byteToColumn(SourceStart);
460 if (NewColumns <= TargetColumns) {
461 SourceEnd = NewEnd;
462 ExpandedRegion = true;
463 }
464 }
465
466 if (!ExpandedRegion)
467 break;
468 }
469
470 CaretStart = map.byteToColumn(SourceStart);
471 CaretEnd = map.byteToColumn(SourceEnd) + CaretColumnsOutsideSource;
472
473 // [CaretStart, CaretEnd) is the slice we want. Update the various
474 // output lines to show only this slice, with two-space padding
475 // before the lines so that it looks nicer.
476
477 assert(CaretStart!=(unsigned)-1 && CaretEnd!=(unsigned)-1 &&
478 SourceStart!=(unsigned)-1 && SourceEnd!=(unsigned)-1);
479 assert(SourceStart <= SourceEnd);
480 assert(CaretStart <= CaretEnd);
481
482 unsigned BackColumnsRemoved
483 = map.byteToColumn(SourceLine.size())-map.byteToColumn(SourceEnd);
484 unsigned FrontColumnsRemoved = CaretStart;
485 unsigned ColumnsKept = CaretEnd-CaretStart;
486
487 // We checked up front that the line needed truncation
488 assert(FrontColumnsRemoved+ColumnsKept+BackColumnsRemoved > Columns);
489
490 // The line needs some truncation, and we'd prefer to keep the front
491 // if possible, so remove the back
492 if (BackColumnsRemoved > strlen(back_ellipse))
493 SourceLine.replace(SourceEnd, std::string::npos, back_ellipse);
494
495 // If that's enough then we're done
496 if (FrontColumnsRemoved+ColumnsKept <= Columns)
497 return;
498
499 // Otherwise remove the front as well
500 if (FrontColumnsRemoved > strlen(front_ellipse)) {
501 SourceLine.replace(0, SourceStart, front_ellipse);
502 CaretLine.replace(0, CaretStart, front_space);
503 if (!FixItInsertionLine.empty())
504 FixItInsertionLine.replace(0, CaretStart, front_space);
505 }
506 }
507
508 /// Skip over whitespace in the string, starting at the given
509 /// index.
510 ///
511 /// \returns The index of the first non-whitespace character that is
512 /// greater than or equal to Idx or, if no such character exists,
513 /// returns the end of the string.
skipWhitespace(unsigned Idx,StringRef Str,unsigned Length)514 static unsigned skipWhitespace(unsigned Idx, StringRef Str, unsigned Length) {
515 while (Idx < Length && isWhitespace(Str[Idx]))
516 ++Idx;
517 return Idx;
518 }
519
520 /// If the given character is the start of some kind of
521 /// balanced punctuation (e.g., quotes or parentheses), return the
522 /// character that will terminate the punctuation.
523 ///
524 /// \returns The ending punctuation character, if any, or the NULL
525 /// character if the input character does not start any punctuation.
findMatchingPunctuation(char c)526 static inline char findMatchingPunctuation(char c) {
527 switch (c) {
528 case '\'': return '\'';
529 case '`': return '\'';
530 case '"': return '"';
531 case '(': return ')';
532 case '[': return ']';
533 case '{': return '}';
534 default: break;
535 }
536
537 return 0;
538 }
539
540 /// Find the end of the word starting at the given offset
541 /// within a string.
542 ///
543 /// \returns the index pointing one character past the end of the
544 /// word.
findEndOfWord(unsigned Start,StringRef Str,unsigned Length,unsigned Column,unsigned Columns)545 static unsigned findEndOfWord(unsigned Start, StringRef Str,
546 unsigned Length, unsigned Column,
547 unsigned Columns) {
548 assert(Start < Str.size() && "Invalid start position!");
549 unsigned End = Start + 1;
550
551 // If we are already at the end of the string, take that as the word.
552 if (End == Str.size())
553 return End;
554
555 // Determine if the start of the string is actually opening
556 // punctuation, e.g., a quote or parentheses.
557 char EndPunct = findMatchingPunctuation(Str[Start]);
558 if (!EndPunct) {
559 // This is a normal word. Just find the first space character.
560 while (End < Length && !isWhitespace(Str[End]))
561 ++End;
562 return End;
563 }
564
565 // We have the start of a balanced punctuation sequence (quotes,
566 // parentheses, etc.). Determine the full sequence is.
567 SmallString<16> PunctuationEndStack;
568 PunctuationEndStack.push_back(EndPunct);
569 while (End < Length && !PunctuationEndStack.empty()) {
570 if (Str[End] == PunctuationEndStack.back())
571 PunctuationEndStack.pop_back();
572 else if (char SubEndPunct = findMatchingPunctuation(Str[End]))
573 PunctuationEndStack.push_back(SubEndPunct);
574
575 ++End;
576 }
577
578 // Find the first space character after the punctuation ended.
579 while (End < Length && !isWhitespace(Str[End]))
580 ++End;
581
582 unsigned PunctWordLength = End - Start;
583 if (// If the word fits on this line
584 Column + PunctWordLength <= Columns ||
585 // ... or the word is "short enough" to take up the next line
586 // without too much ugly white space
587 PunctWordLength < Columns/3)
588 return End; // Take the whole thing as a single "word".
589
590 // The whole quoted/parenthesized string is too long to print as a
591 // single "word". Instead, find the "word" that starts just after
592 // the punctuation and use that end-point instead. This will recurse
593 // until it finds something small enough to consider a word.
594 return findEndOfWord(Start + 1, Str, Length, Column + 1, Columns);
595 }
596
597 /// Print the given string to a stream, word-wrapping it to
598 /// some number of columns in the process.
599 ///
600 /// \param OS the stream to which the word-wrapping string will be
601 /// emitted.
602 /// \param Str the string to word-wrap and output.
603 /// \param Columns the number of columns to word-wrap to.
604 /// \param Column the column number at which the first character of \p
605 /// Str will be printed. This will be non-zero when part of the first
606 /// line has already been printed.
607 /// \param Bold if the current text should be bold
608 /// \param Indentation the number of spaces to indent any lines beyond
609 /// the first line.
610 /// \returns true if word-wrapping was required, or false if the
611 /// string fit on the first line.
printWordWrapped(raw_ostream & OS,StringRef Str,unsigned Columns,unsigned Column=0,bool Bold=false,unsigned Indentation=WordWrapIndentation)612 static bool printWordWrapped(raw_ostream &OS, StringRef Str,
613 unsigned Columns,
614 unsigned Column = 0,
615 bool Bold = false,
616 unsigned Indentation = WordWrapIndentation) {
617 const unsigned Length = std::min(Str.find('\n'), Str.size());
618 bool TextNormal = true;
619
620 // The string used to indent each line.
621 SmallString<16> IndentStr;
622 IndentStr.assign(Indentation, ' ');
623 bool Wrapped = false;
624 for (unsigned WordStart = 0, WordEnd; WordStart < Length;
625 WordStart = WordEnd) {
626 // Find the beginning of the next word.
627 WordStart = skipWhitespace(WordStart, Str, Length);
628 if (WordStart == Length)
629 break;
630
631 // Find the end of this word.
632 WordEnd = findEndOfWord(WordStart, Str, Length, Column, Columns);
633
634 // Does this word fit on the current line?
635 unsigned WordLength = WordEnd - WordStart;
636 if (Column + WordLength < Columns) {
637 // This word fits on the current line; print it there.
638 if (WordStart) {
639 OS << ' ';
640 Column += 1;
641 }
642 applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
643 TextNormal, Bold);
644 Column += WordLength;
645 continue;
646 }
647
648 // This word does not fit on the current line, so wrap to the next
649 // line.
650 OS << '\n';
651 OS.write(&IndentStr[0], Indentation);
652 applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
653 TextNormal, Bold);
654 Column = Indentation + WordLength;
655 Wrapped = true;
656 }
657
658 // Append any remaning text from the message with its existing formatting.
659 applyTemplateHighlighting(OS, Str.substr(Length), TextNormal, Bold);
660
661 assert(TextNormal && "Text highlighted at end of diagnostic message.");
662
663 return Wrapped;
664 }
665
TextDiagnostic(raw_ostream & OS,const LangOptions & LangOpts,DiagnosticOptions * DiagOpts)666 TextDiagnostic::TextDiagnostic(raw_ostream &OS,
667 const LangOptions &LangOpts,
668 DiagnosticOptions *DiagOpts)
669 : DiagnosticRenderer(LangOpts, DiagOpts), OS(OS) {}
670
~TextDiagnostic()671 TextDiagnostic::~TextDiagnostic() {}
672
emitDiagnosticMessage(FullSourceLoc Loc,PresumedLoc PLoc,DiagnosticsEngine::Level Level,StringRef Message,ArrayRef<clang::CharSourceRange> Ranges,DiagOrStoredDiag D)673 void TextDiagnostic::emitDiagnosticMessage(
674 FullSourceLoc Loc, PresumedLoc PLoc, DiagnosticsEngine::Level Level,
675 StringRef Message, ArrayRef<clang::CharSourceRange> Ranges,
676 DiagOrStoredDiag D) {
677 uint64_t StartOfLocationInfo = OS.tell();
678
679 // Emit the location of this particular diagnostic.
680 if (Loc.isValid())
681 emitDiagnosticLoc(Loc, PLoc, Level, Ranges);
682
683 if (DiagOpts->ShowColors)
684 OS.resetColor();
685
686 if (DiagOpts->ShowLevel)
687 printDiagnosticLevel(OS, Level, DiagOpts->ShowColors);
688 printDiagnosticMessage(OS,
689 /*IsSupplemental*/ Level == DiagnosticsEngine::Note,
690 Message, OS.tell() - StartOfLocationInfo,
691 DiagOpts->MessageLength, DiagOpts->ShowColors);
692 }
693
694 /*static*/ void
printDiagnosticLevel(raw_ostream & OS,DiagnosticsEngine::Level Level,bool ShowColors)695 TextDiagnostic::printDiagnosticLevel(raw_ostream &OS,
696 DiagnosticsEngine::Level Level,
697 bool ShowColors) {
698 if (ShowColors) {
699 // Print diagnostic category in bold and color
700 switch (Level) {
701 case DiagnosticsEngine::Ignored:
702 llvm_unreachable("Invalid diagnostic type");
703 case DiagnosticsEngine::Note: OS.changeColor(noteColor, true); break;
704 case DiagnosticsEngine::Remark: OS.changeColor(remarkColor, true); break;
705 case DiagnosticsEngine::Warning: OS.changeColor(warningColor, true); break;
706 case DiagnosticsEngine::Error: OS.changeColor(errorColor, true); break;
707 case DiagnosticsEngine::Fatal: OS.changeColor(fatalColor, true); break;
708 }
709 }
710
711 switch (Level) {
712 case DiagnosticsEngine::Ignored:
713 llvm_unreachable("Invalid diagnostic type");
714 case DiagnosticsEngine::Note: OS << "note: "; break;
715 case DiagnosticsEngine::Remark: OS << "remark: "; break;
716 case DiagnosticsEngine::Warning: OS << "warning: "; break;
717 case DiagnosticsEngine::Error: OS << "error: "; break;
718 case DiagnosticsEngine::Fatal: OS << "fatal error: "; break;
719 }
720
721 if (ShowColors)
722 OS.resetColor();
723 }
724
725 /*static*/
printDiagnosticMessage(raw_ostream & OS,bool IsSupplemental,StringRef Message,unsigned CurrentColumn,unsigned Columns,bool ShowColors)726 void TextDiagnostic::printDiagnosticMessage(raw_ostream &OS,
727 bool IsSupplemental,
728 StringRef Message,
729 unsigned CurrentColumn,
730 unsigned Columns, bool ShowColors) {
731 bool Bold = false;
732 if (ShowColors && !IsSupplemental) {
733 // Print primary diagnostic messages in bold and without color, to visually
734 // indicate the transition from continuation notes and other output.
735 OS.changeColor(savedColor, true);
736 Bold = true;
737 }
738
739 if (Columns)
740 printWordWrapped(OS, Message, Columns, CurrentColumn, Bold);
741 else {
742 bool Normal = true;
743 applyTemplateHighlighting(OS, Message, Normal, Bold);
744 assert(Normal && "Formatting should have returned to normal");
745 }
746
747 if (ShowColors)
748 OS.resetColor();
749 OS << '\n';
750 }
751
emitFilename(StringRef Filename,const SourceManager & SM)752 void TextDiagnostic::emitFilename(StringRef Filename, const SourceManager &SM) {
753 #ifdef _WIN32
754 SmallString<4096> TmpFilename;
755 #endif
756 if (DiagOpts->AbsolutePath) {
757 auto File = SM.getFileManager().getFile(Filename);
758 if (File) {
759 // We want to print a simplified absolute path, i. e. without "dots".
760 //
761 // The hardest part here are the paths like "<part1>/<link>/../<part2>".
762 // On Unix-like systems, we cannot just collapse "<link>/..", because
763 // paths are resolved sequentially, and, thereby, the path
764 // "<part1>/<part2>" may point to a different location. That is why
765 // we use FileManager::getCanonicalName(), which expands all indirections
766 // with llvm::sys::fs::real_path() and caches the result.
767 //
768 // On the other hand, it would be better to preserve as much of the
769 // original path as possible, because that helps a user to recognize it.
770 // real_path() expands all links, which sometimes too much. Luckily,
771 // on Windows we can just use llvm::sys::path::remove_dots(), because,
772 // on that system, both aforementioned paths point to the same place.
773 #ifdef _WIN32
774 TmpFilename = (*File)->getName();
775 llvm::sys::fs::make_absolute(TmpFilename);
776 llvm::sys::path::native(TmpFilename);
777 llvm::sys::path::remove_dots(TmpFilename, /* remove_dot_dot */ true);
778 Filename = StringRef(TmpFilename.data(), TmpFilename.size());
779 #else
780 Filename = SM.getFileManager().getCanonicalName(*File);
781 #endif
782 }
783 }
784
785 OS << Filename;
786 }
787
788 /// Print out the file/line/column information and include trace.
789 ///
790 /// This method handlen the emission of the diagnostic location information.
791 /// This includes extracting as much location information as is present for
792 /// the diagnostic and printing it, as well as any include stack or source
793 /// ranges necessary.
emitDiagnosticLoc(FullSourceLoc Loc,PresumedLoc PLoc,DiagnosticsEngine::Level Level,ArrayRef<CharSourceRange> Ranges)794 void TextDiagnostic::emitDiagnosticLoc(FullSourceLoc Loc, PresumedLoc PLoc,
795 DiagnosticsEngine::Level Level,
796 ArrayRef<CharSourceRange> Ranges) {
797 if (PLoc.isInvalid()) {
798 // At least print the file name if available:
799 FileID FID = Loc.getFileID();
800 if (FID.isValid()) {
801 if (const FileEntry *FE = Loc.getFileEntry()) {
802 emitFilename(FE->getName(), Loc.getManager());
803 OS << ": ";
804 }
805 }
806 return;
807 }
808 unsigned LineNo = PLoc.getLine();
809
810 if (!DiagOpts->ShowLocation)
811 return;
812
813 if (DiagOpts->ShowColors)
814 OS.changeColor(savedColor, true);
815
816 emitFilename(PLoc.getFilename(), Loc.getManager());
817 switch (DiagOpts->getFormat()) {
818 case DiagnosticOptions::SARIF:
819 case DiagnosticOptions::Clang:
820 if (DiagOpts->ShowLine)
821 OS << ':' << LineNo;
822 break;
823 case DiagnosticOptions::MSVC: OS << '(' << LineNo; break;
824 case DiagnosticOptions::Vi: OS << " +" << LineNo; break;
825 }
826
827 if (DiagOpts->ShowColumn)
828 // Compute the column number.
829 if (unsigned ColNo = PLoc.getColumn()) {
830 if (DiagOpts->getFormat() == DiagnosticOptions::MSVC) {
831 OS << ',';
832 // Visual Studio 2010 or earlier expects column number to be off by one
833 if (LangOpts.MSCompatibilityVersion &&
834 !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2012))
835 ColNo--;
836 } else
837 OS << ':';
838 OS << ColNo;
839 }
840 switch (DiagOpts->getFormat()) {
841 case DiagnosticOptions::SARIF:
842 case DiagnosticOptions::Clang:
843 case DiagnosticOptions::Vi: OS << ':'; break;
844 case DiagnosticOptions::MSVC:
845 // MSVC2013 and before print 'file(4) : error'. MSVC2015 gets rid of the
846 // space and prints 'file(4): error'.
847 OS << ')';
848 if (LangOpts.MSCompatibilityVersion &&
849 !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2015))
850 OS << ' ';
851 OS << ':';
852 break;
853 }
854
855 if (DiagOpts->ShowSourceRanges && !Ranges.empty()) {
856 FileID CaretFileID = Loc.getExpansionLoc().getFileID();
857 bool PrintedRange = false;
858
859 for (ArrayRef<CharSourceRange>::const_iterator RI = Ranges.begin(),
860 RE = Ranges.end();
861 RI != RE; ++RI) {
862 // Ignore invalid ranges.
863 if (!RI->isValid()) continue;
864
865 auto &SM = Loc.getManager();
866 SourceLocation B = SM.getExpansionLoc(RI->getBegin());
867 CharSourceRange ERange = SM.getExpansionRange(RI->getEnd());
868 SourceLocation E = ERange.getEnd();
869 bool IsTokenRange = ERange.isTokenRange();
870
871 std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(B);
872 std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(E);
873
874 // If the start or end of the range is in another file, just discard
875 // it.
876 if (BInfo.first != CaretFileID || EInfo.first != CaretFileID)
877 continue;
878
879 // Add in the length of the token, so that we cover multi-char
880 // tokens.
881 unsigned TokSize = 0;
882 if (IsTokenRange)
883 TokSize = Lexer::MeasureTokenLength(E, SM, LangOpts);
884
885 FullSourceLoc BF(B, SM), EF(E, SM);
886 OS << '{'
887 << BF.getLineNumber() << ':' << BF.getColumnNumber() << '-'
888 << EF.getLineNumber() << ':' << (EF.getColumnNumber() + TokSize)
889 << '}';
890 PrintedRange = true;
891 }
892
893 if (PrintedRange)
894 OS << ':';
895 }
896 OS << ' ';
897 }
898
emitIncludeLocation(FullSourceLoc Loc,PresumedLoc PLoc)899 void TextDiagnostic::emitIncludeLocation(FullSourceLoc Loc, PresumedLoc PLoc) {
900 if (DiagOpts->ShowLocation && PLoc.isValid())
901 OS << "In file included from " << PLoc.getFilename() << ':'
902 << PLoc.getLine() << ":\n";
903 else
904 OS << "In included file:\n";
905 }
906
emitImportLocation(FullSourceLoc Loc,PresumedLoc PLoc,StringRef ModuleName)907 void TextDiagnostic::emitImportLocation(FullSourceLoc Loc, PresumedLoc PLoc,
908 StringRef ModuleName) {
909 if (DiagOpts->ShowLocation && PLoc.isValid())
910 OS << "In module '" << ModuleName << "' imported from "
911 << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
912 else
913 OS << "In module '" << ModuleName << "':\n";
914 }
915
emitBuildingModuleLocation(FullSourceLoc Loc,PresumedLoc PLoc,StringRef ModuleName)916 void TextDiagnostic::emitBuildingModuleLocation(FullSourceLoc Loc,
917 PresumedLoc PLoc,
918 StringRef ModuleName) {
919 if (DiagOpts->ShowLocation && PLoc.isValid())
920 OS << "While building module '" << ModuleName << "' imported from "
921 << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
922 else
923 OS << "While building module '" << ModuleName << "':\n";
924 }
925
926 /// Find the suitable set of lines to show to include a set of ranges.
927 static std::optional<std::pair<unsigned, unsigned>>
findLinesForRange(const CharSourceRange & R,FileID FID,const SourceManager & SM)928 findLinesForRange(const CharSourceRange &R, FileID FID,
929 const SourceManager &SM) {
930 if (!R.isValid())
931 return std::nullopt;
932
933 SourceLocation Begin = R.getBegin();
934 SourceLocation End = R.getEnd();
935 if (SM.getFileID(Begin) != FID || SM.getFileID(End) != FID)
936 return std::nullopt;
937
938 return std::make_pair(SM.getExpansionLineNumber(Begin),
939 SM.getExpansionLineNumber(End));
940 }
941
942 /// Add as much of range B into range A as possible without exceeding a maximum
943 /// size of MaxRange. Ranges are inclusive.
944 static std::pair<unsigned, unsigned>
maybeAddRange(std::pair<unsigned,unsigned> A,std::pair<unsigned,unsigned> B,unsigned MaxRange)945 maybeAddRange(std::pair<unsigned, unsigned> A, std::pair<unsigned, unsigned> B,
946 unsigned MaxRange) {
947 // If A is already the maximum size, we're done.
948 unsigned Slack = MaxRange - (A.second - A.first + 1);
949 if (Slack == 0)
950 return A;
951
952 // Easy case: merge succeeds within MaxRange.
953 unsigned Min = std::min(A.first, B.first);
954 unsigned Max = std::max(A.second, B.second);
955 if (Max - Min + 1 <= MaxRange)
956 return {Min, Max};
957
958 // If we can't reach B from A within MaxRange, there's nothing to do.
959 // Don't add lines to the range that contain nothing interesting.
960 if ((B.first > A.first && B.first - A.first + 1 > MaxRange) ||
961 (B.second < A.second && A.second - B.second + 1 > MaxRange))
962 return A;
963
964 // Otherwise, expand A towards B to produce a range of size MaxRange. We
965 // attempt to expand by the same amount in both directions if B strictly
966 // contains A.
967
968 // Expand downwards by up to half the available amount, then upwards as
969 // much as possible, then downwards as much as possible.
970 A.second = std::min(A.second + (Slack + 1) / 2, Max);
971 Slack = MaxRange - (A.second - A.first + 1);
972 A.first = std::max(Min + Slack, A.first) - Slack;
973 A.second = std::min(A.first + MaxRange - 1, Max);
974 return A;
975 }
976
977 /// Highlight a SourceRange (with ~'s) for any characters on LineNo.
highlightRange(const CharSourceRange & R,unsigned LineNo,FileID FID,const SourceColumnMap & map,std::string & CaretLine,const SourceManager & SM,const LangOptions & LangOpts)978 static void highlightRange(const CharSourceRange &R,
979 unsigned LineNo, FileID FID,
980 const SourceColumnMap &map,
981 std::string &CaretLine,
982 const SourceManager &SM,
983 const LangOptions &LangOpts) {
984 if (!R.isValid()) return;
985
986 SourceLocation Begin = R.getBegin();
987 SourceLocation End = R.getEnd();
988
989 unsigned StartLineNo = SM.getExpansionLineNumber(Begin);
990 if (StartLineNo > LineNo || SM.getFileID(Begin) != FID)
991 return; // No intersection.
992
993 unsigned EndLineNo = SM.getExpansionLineNumber(End);
994 if (EndLineNo < LineNo || SM.getFileID(End) != FID)
995 return; // No intersection.
996
997 // Compute the column number of the start.
998 unsigned StartColNo = 0;
999 if (StartLineNo == LineNo) {
1000 StartColNo = SM.getExpansionColumnNumber(Begin);
1001 if (StartColNo) --StartColNo; // Zero base the col #.
1002 }
1003
1004 // Compute the column number of the end.
1005 unsigned EndColNo = map.getSourceLine().size();
1006 if (EndLineNo == LineNo) {
1007 EndColNo = SM.getExpansionColumnNumber(End);
1008 if (EndColNo) {
1009 --EndColNo; // Zero base the col #.
1010
1011 // Add in the length of the token, so that we cover multi-char tokens if
1012 // this is a token range.
1013 if (R.isTokenRange())
1014 EndColNo += Lexer::MeasureTokenLength(End, SM, LangOpts);
1015 } else {
1016 EndColNo = CaretLine.size();
1017 }
1018 }
1019
1020 assert(StartColNo <= EndColNo && "Invalid range!");
1021
1022 // Check that a token range does not highlight only whitespace.
1023 if (R.isTokenRange()) {
1024 // Pick the first non-whitespace column.
1025 while (StartColNo < map.getSourceLine().size() &&
1026 (map.getSourceLine()[StartColNo] == ' ' ||
1027 map.getSourceLine()[StartColNo] == '\t'))
1028 StartColNo = map.startOfNextColumn(StartColNo);
1029
1030 // Pick the last non-whitespace column.
1031 if (EndColNo > map.getSourceLine().size())
1032 EndColNo = map.getSourceLine().size();
1033 while (EndColNo &&
1034 (map.getSourceLine()[EndColNo-1] == ' ' ||
1035 map.getSourceLine()[EndColNo-1] == '\t'))
1036 EndColNo = map.startOfPreviousColumn(EndColNo);
1037
1038 // If the start/end passed each other, then we are trying to highlight a
1039 // range that just exists in whitespace. That most likely means we have
1040 // a multi-line highlighting range that covers a blank line.
1041 if (StartColNo > EndColNo) {
1042 assert(StartLineNo != EndLineNo && "trying to highlight whitespace");
1043 StartColNo = EndColNo;
1044 }
1045 }
1046
1047 assert(StartColNo <= map.getSourceLine().size() && "Invalid range!");
1048 assert(EndColNo <= map.getSourceLine().size() && "Invalid range!");
1049
1050 // Fill the range with ~'s.
1051 StartColNo = map.byteToContainingColumn(StartColNo);
1052 EndColNo = map.byteToContainingColumn(EndColNo);
1053
1054 assert(StartColNo <= EndColNo && "Invalid range!");
1055 if (CaretLine.size() < EndColNo)
1056 CaretLine.resize(EndColNo,' ');
1057 std::fill(CaretLine.begin()+StartColNo,CaretLine.begin()+EndColNo,'~');
1058 }
1059
buildFixItInsertionLine(FileID FID,unsigned LineNo,const SourceColumnMap & map,ArrayRef<FixItHint> Hints,const SourceManager & SM,const DiagnosticOptions * DiagOpts)1060 static std::string buildFixItInsertionLine(FileID FID,
1061 unsigned LineNo,
1062 const SourceColumnMap &map,
1063 ArrayRef<FixItHint> Hints,
1064 const SourceManager &SM,
1065 const DiagnosticOptions *DiagOpts) {
1066 std::string FixItInsertionLine;
1067 if (Hints.empty() || !DiagOpts->ShowFixits)
1068 return FixItInsertionLine;
1069 unsigned PrevHintEndCol = 0;
1070
1071 for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1072 I != E; ++I) {
1073 if (!I->CodeToInsert.empty()) {
1074 // We have an insertion hint. Determine whether the inserted
1075 // code contains no newlines and is on the same line as the caret.
1076 std::pair<FileID, unsigned> HintLocInfo
1077 = SM.getDecomposedExpansionLoc(I->RemoveRange.getBegin());
1078 if (FID == HintLocInfo.first &&
1079 LineNo == SM.getLineNumber(HintLocInfo.first, HintLocInfo.second) &&
1080 StringRef(I->CodeToInsert).find_first_of("\n\r") == StringRef::npos) {
1081 // Insert the new code into the line just below the code
1082 // that the user wrote.
1083 // Note: When modifying this function, be very careful about what is a
1084 // "column" (printed width, platform-dependent) and what is a
1085 // "byte offset" (SourceManager "column").
1086 unsigned HintByteOffset
1087 = SM.getColumnNumber(HintLocInfo.first, HintLocInfo.second) - 1;
1088
1089 // The hint must start inside the source or right at the end
1090 assert(HintByteOffset < static_cast<unsigned>(map.bytes())+1);
1091 unsigned HintCol = map.byteToContainingColumn(HintByteOffset);
1092
1093 // If we inserted a long previous hint, push this one forwards, and add
1094 // an extra space to show that this is not part of the previous
1095 // completion. This is sort of the best we can do when two hints appear
1096 // to overlap.
1097 //
1098 // Note that if this hint is located immediately after the previous
1099 // hint, no space will be added, since the location is more important.
1100 if (HintCol < PrevHintEndCol)
1101 HintCol = PrevHintEndCol + 1;
1102
1103 // This should NOT use HintByteOffset, because the source might have
1104 // Unicode characters in earlier columns.
1105 unsigned NewFixItLineSize = FixItInsertionLine.size() +
1106 (HintCol - PrevHintEndCol) + I->CodeToInsert.size();
1107 if (NewFixItLineSize > FixItInsertionLine.size())
1108 FixItInsertionLine.resize(NewFixItLineSize, ' ');
1109
1110 std::copy(I->CodeToInsert.begin(), I->CodeToInsert.end(),
1111 FixItInsertionLine.end() - I->CodeToInsert.size());
1112
1113 PrevHintEndCol =
1114 HintCol + llvm::sys::locale::columnWidth(I->CodeToInsert);
1115 }
1116 }
1117 }
1118
1119 expandTabs(FixItInsertionLine, DiagOpts->TabStop);
1120
1121 return FixItInsertionLine;
1122 }
1123
1124 /// Emit a code snippet and caret line.
1125 ///
1126 /// This routine emits a single line's code snippet and caret line..
1127 ///
1128 /// \param Loc The location for the caret.
1129 /// \param Ranges The underlined ranges for this code snippet.
1130 /// \param Hints The FixIt hints active for this diagnostic.
emitSnippetAndCaret(FullSourceLoc Loc,DiagnosticsEngine::Level Level,SmallVectorImpl<CharSourceRange> & Ranges,ArrayRef<FixItHint> Hints)1131 void TextDiagnostic::emitSnippetAndCaret(
1132 FullSourceLoc Loc, DiagnosticsEngine::Level Level,
1133 SmallVectorImpl<CharSourceRange> &Ranges, ArrayRef<FixItHint> Hints) {
1134 assert(Loc.isValid() && "must have a valid source location here");
1135 assert(Loc.isFileID() && "must have a file location here");
1136
1137 // If caret diagnostics are enabled and we have location, we want to
1138 // emit the caret. However, we only do this if the location moved
1139 // from the last diagnostic, if the last diagnostic was a note that
1140 // was part of a different warning or error diagnostic, or if the
1141 // diagnostic has ranges. We don't want to emit the same caret
1142 // multiple times if one loc has multiple diagnostics.
1143 if (!DiagOpts->ShowCarets)
1144 return;
1145 if (Loc == LastLoc && Ranges.empty() && Hints.empty() &&
1146 (LastLevel != DiagnosticsEngine::Note || Level == LastLevel))
1147 return;
1148
1149 // Decompose the location into a FID/Offset pair.
1150 std::pair<FileID, unsigned> LocInfo = Loc.getDecomposedLoc();
1151 FileID FID = LocInfo.first;
1152 const SourceManager &SM = Loc.getManager();
1153
1154 // Get information about the buffer it points into.
1155 bool Invalid = false;
1156 StringRef BufData = Loc.getBufferData(&Invalid);
1157 if (Invalid)
1158 return;
1159
1160 unsigned CaretLineNo = Loc.getLineNumber();
1161 unsigned CaretColNo = Loc.getColumnNumber();
1162
1163 // Arbitrarily stop showing snippets when the line is too long.
1164 static const size_t MaxLineLengthToPrint = 4096;
1165 if (CaretColNo > MaxLineLengthToPrint)
1166 return;
1167
1168 // Find the set of lines to include.
1169 const unsigned MaxLines = DiagOpts->SnippetLineLimit;
1170 std::pair<unsigned, unsigned> Lines = {CaretLineNo, CaretLineNo};
1171 for (SmallVectorImpl<CharSourceRange>::iterator I = Ranges.begin(),
1172 E = Ranges.end();
1173 I != E; ++I)
1174 if (auto OptionalRange = findLinesForRange(*I, FID, SM))
1175 Lines = maybeAddRange(Lines, *OptionalRange, MaxLines);
1176
1177 for (unsigned LineNo = Lines.first; LineNo != Lines.second + 1; ++LineNo) {
1178 const char *BufStart = BufData.data();
1179 const char *BufEnd = BufStart + BufData.size();
1180
1181 // Rewind from the current position to the start of the line.
1182 const char *LineStart =
1183 BufStart +
1184 SM.getDecomposedLoc(SM.translateLineCol(FID, LineNo, 1)).second;
1185 if (LineStart == BufEnd)
1186 break;
1187
1188 // Compute the line end.
1189 const char *LineEnd = LineStart;
1190 while (*LineEnd != '\n' && *LineEnd != '\r' && LineEnd != BufEnd)
1191 ++LineEnd;
1192
1193 // Arbitrarily stop showing snippets when the line is too long.
1194 // FIXME: Don't print any lines in this case.
1195 if (size_t(LineEnd - LineStart) > MaxLineLengthToPrint)
1196 return;
1197
1198 // Trim trailing null-bytes.
1199 StringRef Line(LineStart, LineEnd - LineStart);
1200 while (!Line.empty() && Line.back() == '\0' &&
1201 (LineNo != CaretLineNo || Line.size() > CaretColNo))
1202 Line = Line.drop_back();
1203
1204 // Copy the line of code into an std::string for ease of manipulation.
1205 std::string SourceLine(Line.begin(), Line.end());
1206
1207 // Build the byte to column map.
1208 const SourceColumnMap sourceColMap(SourceLine, DiagOpts->TabStop);
1209
1210 // Create a line for the caret that is filled with spaces that is the same
1211 // number of columns as the line of source code.
1212 std::string CaretLine(sourceColMap.columns(), ' ');
1213
1214 // Highlight all of the characters covered by Ranges with ~ characters.
1215 for (SmallVectorImpl<CharSourceRange>::iterator I = Ranges.begin(),
1216 E = Ranges.end();
1217 I != E; ++I)
1218 highlightRange(*I, LineNo, FID, sourceColMap, CaretLine, SM, LangOpts);
1219
1220 // Next, insert the caret itself.
1221 if (CaretLineNo == LineNo) {
1222 CaretColNo = sourceColMap.byteToContainingColumn(CaretColNo - 1);
1223 if (CaretLine.size() < CaretColNo + 1)
1224 CaretLine.resize(CaretColNo + 1, ' ');
1225 CaretLine[CaretColNo] = '^';
1226 }
1227
1228 std::string FixItInsertionLine = buildFixItInsertionLine(
1229 FID, LineNo, sourceColMap, Hints, SM, DiagOpts.get());
1230
1231 // If the source line is too long for our terminal, select only the
1232 // "interesting" source region within that line.
1233 unsigned Columns = DiagOpts->MessageLength;
1234 if (Columns)
1235 selectInterestingSourceRegion(SourceLine, CaretLine, FixItInsertionLine,
1236 Columns, sourceColMap);
1237
1238 // If we are in -fdiagnostics-print-source-range-info mode, we are trying
1239 // to produce easily machine parsable output. Add a space before the
1240 // source line and the caret to make it trivial to tell the main diagnostic
1241 // line from what the user is intended to see.
1242 if (DiagOpts->ShowSourceRanges) {
1243 SourceLine = ' ' + SourceLine;
1244 CaretLine = ' ' + CaretLine;
1245 }
1246
1247 // Finally, remove any blank spaces from the end of CaretLine.
1248 while (!CaretLine.empty() && CaretLine[CaretLine.size() - 1] == ' ')
1249 CaretLine.erase(CaretLine.end() - 1);
1250
1251 // Emit what we have computed.
1252 emitSnippet(SourceLine);
1253
1254 if (!CaretLine.empty()) {
1255 if (DiagOpts->ShowColors)
1256 OS.changeColor(caretColor, true);
1257 OS << CaretLine << '\n';
1258 if (DiagOpts->ShowColors)
1259 OS.resetColor();
1260 }
1261
1262 if (!FixItInsertionLine.empty()) {
1263 if (DiagOpts->ShowColors)
1264 // Print fixit line in color
1265 OS.changeColor(fixitColor, false);
1266 if (DiagOpts->ShowSourceRanges)
1267 OS << ' ';
1268 OS << FixItInsertionLine << '\n';
1269 if (DiagOpts->ShowColors)
1270 OS.resetColor();
1271 }
1272 }
1273
1274 // Print out any parseable fixit information requested by the options.
1275 emitParseableFixits(Hints, SM);
1276 }
1277
emitSnippet(StringRef line)1278 void TextDiagnostic::emitSnippet(StringRef line) {
1279 if (line.empty())
1280 return;
1281
1282 size_t i = 0;
1283
1284 std::string to_print;
1285 bool print_reversed = false;
1286
1287 while (i<line.size()) {
1288 std::pair<SmallString<16>,bool> res
1289 = printableTextForNextCharacter(line, &i, DiagOpts->TabStop);
1290 bool was_printable = res.second;
1291
1292 if (DiagOpts->ShowColors && was_printable == print_reversed) {
1293 if (print_reversed)
1294 OS.reverseColor();
1295 OS << to_print;
1296 to_print.clear();
1297 if (DiagOpts->ShowColors)
1298 OS.resetColor();
1299 }
1300
1301 print_reversed = !was_printable;
1302 to_print += res.first.str();
1303 }
1304
1305 if (print_reversed && DiagOpts->ShowColors)
1306 OS.reverseColor();
1307 OS << to_print;
1308 if (print_reversed && DiagOpts->ShowColors)
1309 OS.resetColor();
1310
1311 OS << '\n';
1312 }
1313
emitParseableFixits(ArrayRef<FixItHint> Hints,const SourceManager & SM)1314 void TextDiagnostic::emitParseableFixits(ArrayRef<FixItHint> Hints,
1315 const SourceManager &SM) {
1316 if (!DiagOpts->ShowParseableFixits)
1317 return;
1318
1319 // We follow FixItRewriter's example in not (yet) handling
1320 // fix-its in macros.
1321 for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1322 I != E; ++I) {
1323 if (I->RemoveRange.isInvalid() ||
1324 I->RemoveRange.getBegin().isMacroID() ||
1325 I->RemoveRange.getEnd().isMacroID())
1326 return;
1327 }
1328
1329 for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1330 I != E; ++I) {
1331 SourceLocation BLoc = I->RemoveRange.getBegin();
1332 SourceLocation ELoc = I->RemoveRange.getEnd();
1333
1334 std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(BLoc);
1335 std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(ELoc);
1336
1337 // Adjust for token ranges.
1338 if (I->RemoveRange.isTokenRange())
1339 EInfo.second += Lexer::MeasureTokenLength(ELoc, SM, LangOpts);
1340
1341 // We specifically do not do word-wrapping or tab-expansion here,
1342 // because this is supposed to be easy to parse.
1343 PresumedLoc PLoc = SM.getPresumedLoc(BLoc);
1344 if (PLoc.isInvalid())
1345 break;
1346
1347 OS << "fix-it:\"";
1348 OS.write_escaped(PLoc.getFilename());
1349 OS << "\":{" << SM.getLineNumber(BInfo.first, BInfo.second)
1350 << ':' << SM.getColumnNumber(BInfo.first, BInfo.second)
1351 << '-' << SM.getLineNumber(EInfo.first, EInfo.second)
1352 << ':' << SM.getColumnNumber(EInfo.first, EInfo.second)
1353 << "}:\"";
1354 OS.write_escaped(I->CodeToInsert);
1355 OS << "\"\n";
1356 }
1357 }
1358