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 "clang/Lex/Preprocessor.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/raw_ostream.h"
21#include <algorithm>
22#include <optional>
23
24using namespace clang;
25
26static const enum raw_ostream::Colors noteColor = raw_ostream::CYAN;
27static const enum raw_ostream::Colors remarkColor =
28 raw_ostream::BLUE;
29static const enum raw_ostream::Colors fixitColor =
30 raw_ostream::GREEN;
31static const enum raw_ostream::Colors caretColor =
32 raw_ostream::GREEN;
33static const enum raw_ostream::Colors warningColor =
34 raw_ostream::MAGENTA;
35static const enum raw_ostream::Colors templateColor =
36 raw_ostream::CYAN;
37static const enum raw_ostream::Colors errorColor = raw_ostream::RED;
38static const enum raw_ostream::Colors fatalColor = raw_ostream::RED;
39// Used for changing only the bold attribute.
40static const enum raw_ostream::Colors savedColor =
41 raw_ostream::SAVEDCOLOR;
42
43// Magenta is taken for 'warning'. Red is already 'error' and 'cyan'
44// is already taken for 'note'. Green is already used to underline
45// source ranges. White and black are bad because of the usual
46// terminal backgrounds. Which leaves us only with TWO options.
47static constexpr raw_ostream::Colors CommentColor = raw_ostream::YELLOW;
48static constexpr raw_ostream::Colors LiteralColor = raw_ostream::GREEN;
49static constexpr raw_ostream::Colors KeywordColor = raw_ostream::BLUE;
50
51/// Add highlights to differences in template strings.
52static void applyTemplateHighlighting(raw_ostream &OS, StringRef Str,
53 bool &Normal, bool Bold) {
54 while (true) {
55 size_t Pos = Str.find(C: ToggleHighlight);
56 OS << Str.slice(Start: 0, End: Pos);
57 if (Pos == StringRef::npos)
58 break;
59
60 Str = Str.substr(Start: Pos + 1);
61 if (Normal)
62 OS.changeColor(Color: templateColor, Bold: true);
63 else {
64 OS.resetColor();
65 if (Bold)
66 OS.changeColor(Color: savedColor, Bold: true);
67 }
68 Normal = !Normal;
69 }
70}
71
72/// Number of spaces to indent when word-wrapping.
73const unsigned WordWrapIndentation = 6;
74
75static int bytesSincePreviousTabOrLineBegin(StringRef SourceLine, size_t i) {
76 int bytes = 0;
77 while (0<i) {
78 if (SourceLine[--i]=='\t')
79 break;
80 ++bytes;
81 }
82 return bytes;
83}
84
85/// returns a printable representation of first item from input range
86///
87/// This function returns a printable representation of the next item in a line
88/// of source. If the next byte begins a valid and printable character, that
89/// character is returned along with 'true'.
90///
91/// Otherwise, if the next byte begins a valid, but unprintable character, a
92/// printable, escaped representation of the character is returned, along with
93/// 'false'. Otherwise a printable, escaped representation of the next byte
94/// is returned along with 'false'.
95///
96/// \note The index is updated to be used with a subsequent call to
97/// printableTextForNextCharacter.
98///
99/// \param SourceLine The line of source
100/// \param I Pointer to byte index,
101/// \param TabStop used to expand tabs
102/// \return pair(printable text, 'true' iff original text was printable)
103///
104static std::pair<SmallString<16>, bool>
105printableTextForNextCharacter(StringRef SourceLine, size_t *I,
106 unsigned TabStop) {
107 assert(I && "I must not be null");
108 assert(*I < SourceLine.size() && "must point to a valid index");
109
110 if (SourceLine[*I] == '\t') {
111 assert(0 < TabStop && TabStop <= DiagnosticOptions::MaxTabStop &&
112 "Invalid -ftabstop value");
113 unsigned Col = bytesSincePreviousTabOrLineBegin(SourceLine, i: *I);
114 unsigned NumSpaces = TabStop - (Col % TabStop);
115 assert(0 < NumSpaces && NumSpaces <= TabStop
116 && "Invalid computation of space amt");
117 ++(*I);
118
119 SmallString<16> ExpandedTab;
120 ExpandedTab.assign(NumElts: NumSpaces, Elt: ' ');
121 return std::make_pair(x&: ExpandedTab, y: true);
122 }
123
124 const unsigned char *Begin = SourceLine.bytes_begin() + *I;
125
126 // Fast path for the common ASCII case.
127 if (*Begin < 0x80 && llvm::sys::locale::isPrint(c: *Begin)) {
128 ++(*I);
129 return std::make_pair(x: SmallString<16>(Begin, Begin + 1), y: true);
130 }
131 unsigned CharSize = llvm::getNumBytesForUTF8(firstByte: *Begin);
132 const unsigned char *End = Begin + CharSize;
133
134 // Convert it to UTF32 and check if it's printable.
135 if (End <= SourceLine.bytes_end() && llvm::isLegalUTF8Sequence(source: Begin, sourceEnd: End)) {
136 llvm::UTF32 C;
137 llvm::UTF32 *CPtr = &C;
138
139 // Begin and end before conversion.
140 unsigned char const *OriginalBegin = Begin;
141 llvm::ConversionResult Res = llvm::ConvertUTF8toUTF32(
142 sourceStart: &Begin, sourceEnd: End, targetStart: &CPtr, targetEnd: CPtr + 1, flags: llvm::strictConversion);
143 (void)Res;
144 assert(Res == llvm::conversionOK);
145 assert(OriginalBegin < Begin);
146 assert(unsigned(Begin - OriginalBegin) == CharSize);
147
148 (*I) += (Begin - OriginalBegin);
149
150 // Valid, multi-byte, printable UTF8 character.
151 if (llvm::sys::locale::isPrint(c: C))
152 return std::make_pair(x: SmallString<16>(OriginalBegin, End), y: true);
153
154 // Valid but not printable.
155 SmallString<16> Str("<U+>");
156 while (C) {
157 Str.insert(I: Str.begin() + 3, Elt: llvm::hexdigit(X: C % 16));
158 C /= 16;
159 }
160 while (Str.size() < 8)
161 Str.insert(I: Str.begin() + 3, Elt: llvm::hexdigit(X: 0));
162 return std::make_pair(x&: Str, y: false);
163 }
164
165 // Otherwise, not printable since it's not valid UTF8.
166 SmallString<16> ExpandedByte("<XX>");
167 unsigned char Byte = SourceLine[*I];
168 ExpandedByte[1] = llvm::hexdigit(X: Byte / 16);
169 ExpandedByte[2] = llvm::hexdigit(X: Byte % 16);
170 ++(*I);
171 return std::make_pair(x&: ExpandedByte, y: false);
172}
173
174static void expandTabs(std::string &SourceLine, unsigned TabStop) {
175 size_t I = SourceLine.size();
176 while (I > 0) {
177 I--;
178 if (SourceLine[I] != '\t')
179 continue;
180 size_t TmpI = I;
181 auto [Str, Printable] =
182 printableTextForNextCharacter(SourceLine, I: &TmpI, TabStop);
183 SourceLine.replace(pos: I, n1: 1, s: Str.c_str());
184 }
185}
186
187/// \p BytesOut:
188/// A mapping from columns to the byte of the source line that produced the
189/// character displaying at that column. This is the inverse of \p ColumnsOut.
190///
191/// The last element in the array is the number of bytes in the source string.
192///
193/// example: (given a tabstop of 8)
194///
195/// "a \t \u3042" -> {0,1,2,-1,-1,-1,-1,-1,3,4,-1,7}
196///
197/// (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
198/// display)
199///
200/// \p ColumnsOut:
201/// A mapping from the bytes
202/// of the printable representation of the line to the columns those printable
203/// characters will appear at (numbering the first column as 0).
204///
205/// If a byte 'i' corresponds to multiple columns (e.g. the byte contains a tab
206/// character) then the array will map that byte to the first column the
207/// tab appears at and the next value in the map will have been incremented
208/// more than once.
209///
210/// If a byte is the first in a sequence of bytes that together map to a single
211/// entity in the output, then the array will map that byte to the appropriate
212/// column while the subsequent bytes will be -1.
213///
214/// The last element in the array does not correspond to any byte in the input
215/// and instead is the number of columns needed to display the source
216///
217/// example: (given a tabstop of 8)
218///
219/// "a \t \u3042" -> {0,1,2,8,9,-1,-1,11}
220///
221/// (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
222/// display)
223static void genColumnByteMapping(StringRef SourceLine, unsigned TabStop,
224 SmallVectorImpl<int> &BytesOut,
225 SmallVectorImpl<int> &ColumnsOut) {
226 assert(BytesOut.empty());
227 assert(ColumnsOut.empty());
228
229 if (SourceLine.empty()) {
230 BytesOut.resize(N: 1u, NV: 0);
231 ColumnsOut.resize(N: 1u, NV: 0);
232 return;
233 }
234
235 ColumnsOut.resize(N: SourceLine.size() + 1, NV: -1);
236
237 int Columns = 0;
238 size_t I = 0;
239 while (I < SourceLine.size()) {
240 ColumnsOut[I] = Columns;
241 BytesOut.resize(N: Columns + 1, NV: -1);
242 BytesOut.back() = I;
243 auto [Str, Printable] =
244 printableTextForNextCharacter(SourceLine, I: &I, TabStop);
245 Columns += llvm::sys::locale::columnWidth(s: Str);
246 }
247
248 ColumnsOut.back() = Columns;
249 BytesOut.resize(N: Columns + 1, NV: -1);
250 BytesOut.back() = I;
251}
252
253namespace {
254struct SourceColumnMap {
255 SourceColumnMap(StringRef SourceLine, unsigned TabStop)
256 : m_SourceLine(SourceLine) {
257
258 genColumnByteMapping(SourceLine, TabStop, BytesOut&: m_columnToByte, ColumnsOut&: m_byteToColumn);
259
260 assert(m_byteToColumn.size()==SourceLine.size()+1);
261 assert(0 < m_byteToColumn.size() && 0 < m_columnToByte.size());
262 assert(m_byteToColumn.size()
263 == static_cast<unsigned>(m_columnToByte.back()+1));
264 assert(static_cast<unsigned>(m_byteToColumn.back()+1)
265 == m_columnToByte.size());
266 }
267 int columns() const { return m_byteToColumn.back(); }
268 int bytes() const { return m_columnToByte.back(); }
269
270 /// Map a byte to the column which it is at the start of, or return -1
271 /// if it is not at the start of a column (for a UTF-8 trailing byte).
272 int byteToColumn(int n) const {
273 assert(0<=n && n<static_cast<int>(m_byteToColumn.size()));
274 return m_byteToColumn[n];
275 }
276
277 /// Map a byte to the first column which contains it.
278 int byteToContainingColumn(int N) const {
279 assert(0 <= N && N < static_cast<int>(m_byteToColumn.size()));
280 while (m_byteToColumn[N] == -1)
281 --N;
282 return m_byteToColumn[N];
283 }
284
285 /// Map a column to the byte which starts the column, or return -1 if
286 /// the column the second or subsequent column of an expanded tab or similar
287 /// multi-column entity.
288 int columnToByte(int n) const {
289 assert(0<=n && n<static_cast<int>(m_columnToByte.size()));
290 return m_columnToByte[n];
291 }
292
293 /// Map from a byte index to the next byte which starts a column.
294 int startOfNextColumn(int N) const {
295 assert(0 <= N && N < static_cast<int>(m_byteToColumn.size() - 1));
296 while (byteToColumn(n: ++N) == -1) {}
297 return N;
298 }
299
300 /// Map from a byte index to the previous byte which starts a column.
301 int startOfPreviousColumn(int N) const {
302 assert(0 < N && N < static_cast<int>(m_byteToColumn.size()));
303 while (byteToColumn(n: --N) == -1) {}
304 return N;
305 }
306
307 StringRef getSourceLine() const {
308 return m_SourceLine;
309 }
310
311private:
312 const std::string m_SourceLine;
313 SmallVector<int,200> m_byteToColumn;
314 SmallVector<int,200> m_columnToByte;
315};
316} // end anonymous namespace
317
318/// When the source code line we want to print is too long for
319/// the terminal, select the "interesting" region.
320static void selectInterestingSourceRegion(std::string &SourceLine,
321 std::string &CaretLine,
322 std::string &FixItInsertionLine,
323 unsigned Columns,
324 const SourceColumnMap &map) {
325 unsigned CaretColumns = CaretLine.size();
326 unsigned FixItColumns = llvm::sys::locale::columnWidth(s: FixItInsertionLine);
327 unsigned MaxColumns = std::max(a: static_cast<unsigned>(map.columns()),
328 b: std::max(a: CaretColumns, b: FixItColumns));
329 // if the number of columns is less than the desired number we're done
330 if (MaxColumns <= Columns)
331 return;
332
333 // No special characters are allowed in CaretLine.
334 assert(llvm::none_of(CaretLine, [](char c) { return c < ' ' || '~' < c; }));
335
336 // Find the slice that we need to display the full caret line
337 // correctly.
338 unsigned CaretStart = 0, CaretEnd = CaretLine.size();
339 for (; CaretStart != CaretEnd; ++CaretStart)
340 if (!isWhitespace(c: CaretLine[CaretStart]))
341 break;
342
343 for (; CaretEnd != CaretStart; --CaretEnd)
344 if (!isWhitespace(c: CaretLine[CaretEnd - 1]))
345 break;
346
347 // caret has already been inserted into CaretLine so the above whitespace
348 // check is guaranteed to include the caret
349
350 // If we have a fix-it line, make sure the slice includes all of the
351 // fix-it information.
352 if (!FixItInsertionLine.empty()) {
353 unsigned FixItStart = 0, FixItEnd = FixItInsertionLine.size();
354 for (; FixItStart != FixItEnd; ++FixItStart)
355 if (!isWhitespace(c: FixItInsertionLine[FixItStart]))
356 break;
357
358 for (; FixItEnd != FixItStart; --FixItEnd)
359 if (!isWhitespace(c: FixItInsertionLine[FixItEnd - 1]))
360 break;
361
362 // We can safely use the byte offset FixItStart as the column offset
363 // because the characters up until FixItStart are all ASCII whitespace
364 // characters.
365 unsigned FixItStartCol = FixItStart;
366 unsigned FixItEndCol
367 = llvm::sys::locale::columnWidth(s: FixItInsertionLine.substr(pos: 0, n: FixItEnd));
368
369 CaretStart = std::min(a: FixItStartCol, b: CaretStart);
370 CaretEnd = std::max(a: FixItEndCol, b: CaretEnd);
371 }
372
373 // CaretEnd may have been set at the middle of a character
374 // If it's not at a character's first column then advance it past the current
375 // character.
376 while (static_cast<int>(CaretEnd) < map.columns() &&
377 -1 == map.columnToByte(n: CaretEnd))
378 ++CaretEnd;
379
380 assert((static_cast<int>(CaretStart) > map.columns() ||
381 -1!=map.columnToByte(CaretStart)) &&
382 "CaretStart must not point to a column in the middle of a source"
383 " line character");
384 assert((static_cast<int>(CaretEnd) > map.columns() ||
385 -1!=map.columnToByte(CaretEnd)) &&
386 "CaretEnd must not point to a column in the middle of a source line"
387 " character");
388
389 // CaretLine[CaretStart, CaretEnd) contains all of the interesting
390 // parts of the caret line. While this slice is smaller than the
391 // number of columns we have, try to grow the slice to encompass
392 // more context.
393
394 unsigned SourceStart = map.columnToByte(n: std::min<unsigned>(a: CaretStart,
395 b: map.columns()));
396 unsigned SourceEnd = map.columnToByte(n: std::min<unsigned>(a: CaretEnd,
397 b: map.columns()));
398
399 unsigned CaretColumnsOutsideSource = CaretEnd-CaretStart
400 - (map.byteToColumn(n: SourceEnd)-map.byteToColumn(n: SourceStart));
401
402 char const *front_ellipse = " ...";
403 char const *front_space = " ";
404 char const *back_ellipse = "...";
405 unsigned ellipses_space = strlen(s: front_ellipse) + strlen(s: back_ellipse);
406
407 unsigned TargetColumns = Columns;
408 // Give us extra room for the ellipses
409 // and any of the caret line that extends past the source
410 if (TargetColumns > ellipses_space+CaretColumnsOutsideSource)
411 TargetColumns -= ellipses_space+CaretColumnsOutsideSource;
412
413 while (SourceStart>0 || SourceEnd<SourceLine.size()) {
414 bool ExpandedRegion = false;
415
416 if (SourceStart>0) {
417 unsigned NewStart = map.startOfPreviousColumn(N: SourceStart);
418
419 // Skip over any whitespace we see here; we're looking for
420 // another bit of interesting text.
421 // FIXME: Detect non-ASCII whitespace characters too.
422 while (NewStart && isWhitespace(c: SourceLine[NewStart]))
423 NewStart = map.startOfPreviousColumn(N: NewStart);
424
425 // Skip over this bit of "interesting" text.
426 while (NewStart) {
427 unsigned Prev = map.startOfPreviousColumn(N: NewStart);
428 if (isWhitespace(c: SourceLine[Prev]))
429 break;
430 NewStart = Prev;
431 }
432
433 assert(map.byteToColumn(NewStart) != -1);
434 unsigned NewColumns = map.byteToColumn(n: SourceEnd) -
435 map.byteToColumn(n: NewStart);
436 if (NewColumns <= TargetColumns) {
437 SourceStart = NewStart;
438 ExpandedRegion = true;
439 }
440 }
441
442 if (SourceEnd<SourceLine.size()) {
443 unsigned NewEnd = map.startOfNextColumn(N: SourceEnd);
444
445 // Skip over any whitespace we see here; we're looking for
446 // another bit of interesting text.
447 // FIXME: Detect non-ASCII whitespace characters too.
448 while (NewEnd < SourceLine.size() && isWhitespace(c: SourceLine[NewEnd]))
449 NewEnd = map.startOfNextColumn(N: NewEnd);
450
451 // Skip over this bit of "interesting" text.
452 while (NewEnd < SourceLine.size() && isWhitespace(c: SourceLine[NewEnd]))
453 NewEnd = map.startOfNextColumn(N: NewEnd);
454
455 assert(map.byteToColumn(NewEnd) != -1);
456 unsigned NewColumns = map.byteToColumn(n: NewEnd) -
457 map.byteToColumn(n: SourceStart);
458 if (NewColumns <= TargetColumns) {
459 SourceEnd = NewEnd;
460 ExpandedRegion = true;
461 }
462 }
463
464 if (!ExpandedRegion)
465 break;
466 }
467
468 CaretStart = map.byteToColumn(n: SourceStart);
469 CaretEnd = map.byteToColumn(n: SourceEnd) + CaretColumnsOutsideSource;
470
471 // [CaretStart, CaretEnd) is the slice we want. Update the various
472 // output lines to show only this slice.
473 assert(CaretStart!=(unsigned)-1 && CaretEnd!=(unsigned)-1 &&
474 SourceStart!=(unsigned)-1 && SourceEnd!=(unsigned)-1);
475 assert(SourceStart <= SourceEnd);
476 assert(CaretStart <= CaretEnd);
477
478 unsigned BackColumnsRemoved
479 = map.byteToColumn(n: SourceLine.size())-map.byteToColumn(n: SourceEnd);
480 unsigned FrontColumnsRemoved = CaretStart;
481 unsigned ColumnsKept = CaretEnd-CaretStart;
482
483 // We checked up front that the line needed truncation
484 assert(FrontColumnsRemoved+ColumnsKept+BackColumnsRemoved > Columns);
485
486 // The line needs some truncation, and we'd prefer to keep the front
487 // if possible, so remove the back
488 if (BackColumnsRemoved > strlen(s: back_ellipse))
489 SourceLine.replace(pos: SourceEnd, n1: std::string::npos, s: back_ellipse);
490
491 // If that's enough then we're done
492 if (FrontColumnsRemoved+ColumnsKept <= Columns)
493 return;
494
495 // Otherwise remove the front as well
496 if (FrontColumnsRemoved > strlen(s: front_ellipse)) {
497 SourceLine.replace(pos: 0, n1: SourceStart, s: front_ellipse);
498 CaretLine.replace(pos: 0, n1: CaretStart, s: front_space);
499 if (!FixItInsertionLine.empty())
500 FixItInsertionLine.replace(pos: 0, n1: CaretStart, s: front_space);
501 }
502}
503
504/// Skip over whitespace in the string, starting at the given
505/// index.
506///
507/// \returns The index of the first non-whitespace character that is
508/// greater than or equal to Idx or, if no such character exists,
509/// returns the end of the string.
510static unsigned skipWhitespace(unsigned Idx, StringRef Str, unsigned Length) {
511 while (Idx < Length && isWhitespace(c: Str[Idx]))
512 ++Idx;
513 return Idx;
514}
515
516/// If the given character is the start of some kind of
517/// balanced punctuation (e.g., quotes or parentheses), return the
518/// character that will terminate the punctuation.
519///
520/// \returns The ending punctuation character, if any, or the NULL
521/// character if the input character does not start any punctuation.
522static inline char findMatchingPunctuation(char c) {
523 switch (c) {
524 case '\'': return '\'';
525 case '`': return '\'';
526 case '"': return '"';
527 case '(': return ')';
528 case '[': return ']';
529 case '{': return '}';
530 default: break;
531 }
532
533 return 0;
534}
535
536/// Find the end of the word starting at the given offset
537/// within a string.
538///
539/// \returns the index pointing one character past the end of the
540/// word.
541static unsigned findEndOfWord(unsigned Start, StringRef Str,
542 unsigned Length, unsigned Column,
543 unsigned Columns) {
544 assert(Start < Str.size() && "Invalid start position!");
545 unsigned End = Start + 1;
546
547 // If we are already at the end of the string, take that as the word.
548 if (End == Str.size())
549 return End;
550
551 // Determine if the start of the string is actually opening
552 // punctuation, e.g., a quote or parentheses.
553 char EndPunct = findMatchingPunctuation(c: Str[Start]);
554 if (!EndPunct) {
555 // This is a normal word. Just find the first space character.
556 while (End < Length && !isWhitespace(c: Str[End]))
557 ++End;
558 return End;
559 }
560
561 // We have the start of a balanced punctuation sequence (quotes,
562 // parentheses, etc.). Determine the full sequence is.
563 SmallString<16> PunctuationEndStack;
564 PunctuationEndStack.push_back(Elt: EndPunct);
565 while (End < Length && !PunctuationEndStack.empty()) {
566 if (Str[End] == PunctuationEndStack.back())
567 PunctuationEndStack.pop_back();
568 else if (char SubEndPunct = findMatchingPunctuation(c: Str[End]))
569 PunctuationEndStack.push_back(Elt: SubEndPunct);
570
571 ++End;
572 }
573
574 // Find the first space character after the punctuation ended.
575 while (End < Length && !isWhitespace(c: Str[End]))
576 ++End;
577
578 unsigned PunctWordLength = End - Start;
579 if (// If the word fits on this line
580 Column + PunctWordLength <= Columns ||
581 // ... or the word is "short enough" to take up the next line
582 // without too much ugly white space
583 PunctWordLength < Columns/3)
584 return End; // Take the whole thing as a single "word".
585
586 // The whole quoted/parenthesized string is too long to print as a
587 // single "word". Instead, find the "word" that starts just after
588 // the punctuation and use that end-point instead. This will recurse
589 // until it finds something small enough to consider a word.
590 return findEndOfWord(Start: Start + 1, Str, Length, Column: Column + 1, Columns);
591}
592
593/// Print the given string to a stream, word-wrapping it to
594/// some number of columns in the process.
595///
596/// \param OS the stream to which the word-wrapping string will be
597/// emitted.
598/// \param Str the string to word-wrap and output.
599/// \param Columns the number of columns to word-wrap to.
600/// \param Column the column number at which the first character of \p
601/// Str will be printed. This will be non-zero when part of the first
602/// line has already been printed.
603/// \param Bold if the current text should be bold
604/// \returns true if word-wrapping was required, or false if the
605/// string fit on the first line.
606static bool printWordWrapped(raw_ostream &OS, StringRef Str, unsigned Columns,
607 unsigned Column, bool Bold) {
608 const unsigned Length = std::min(a: Str.find(C: '\n'), b: Str.size());
609 bool TextNormal = true;
610
611 bool Wrapped = false;
612 for (unsigned WordStart = 0, WordEnd; WordStart < Length;
613 WordStart = WordEnd) {
614 // Find the beginning of the next word.
615 WordStart = skipWhitespace(Idx: WordStart, Str, Length);
616 if (WordStart == Length)
617 break;
618
619 // Find the end of this word.
620 WordEnd = findEndOfWord(Start: WordStart, Str, Length, Column, Columns);
621
622 // Does this word fit on the current line?
623 unsigned WordLength = WordEnd - WordStart;
624 if (Column + WordLength < Columns) {
625 // This word fits on the current line; print it there.
626 if (WordStart) {
627 OS << ' ';
628 Column += 1;
629 }
630 applyTemplateHighlighting(OS, Str: Str.substr(Start: WordStart, N: WordLength),
631 Normal&: TextNormal, Bold);
632 Column += WordLength;
633 continue;
634 }
635
636 // This word does not fit on the current line, so wrap to the next
637 // line.
638 OS << '\n';
639 OS.indent(NumSpaces: WordWrapIndentation);
640 applyTemplateHighlighting(OS, Str: Str.substr(Start: WordStart, N: WordLength),
641 Normal&: TextNormal, Bold);
642 Column = WordWrapIndentation + WordLength;
643 Wrapped = true;
644 }
645
646 // Append any remaning text from the message with its existing formatting.
647 applyTemplateHighlighting(OS, Str: Str.substr(Start: Length), Normal&: TextNormal, Bold);
648
649 assert(TextNormal && "Text highlighted at end of diagnostic message.");
650
651 return Wrapped;
652}
653
654TextDiagnostic::TextDiagnostic(raw_ostream &OS, const LangOptions &LangOpts,
655 DiagnosticOptions &DiagOpts,
656 const Preprocessor *PP)
657 : DiagnosticRenderer(LangOpts, DiagOpts), OS(OS), PP(PP) {}
658
659TextDiagnostic::~TextDiagnostic() {}
660
661void TextDiagnostic::emitDiagnosticMessage(
662 FullSourceLoc Loc, PresumedLoc PLoc, DiagnosticsEngine::Level Level,
663 StringRef Message, ArrayRef<clang::CharSourceRange> Ranges,
664 DiagOrStoredDiag D) {
665 uint64_t StartOfLocationInfo = OS.tell();
666
667 // Emit the location of this particular diagnostic.
668 if (Loc.isValid())
669 emitDiagnosticLoc(Loc, PLoc, Level, Ranges);
670
671 if (DiagOpts.ShowColors)
672 OS.resetColor();
673
674 if (DiagOpts.ShowLevel)
675 printDiagnosticLevel(OS, Level, ShowColors: DiagOpts.ShowColors);
676 printDiagnosticMessage(OS,
677 /*IsSupplemental*/ Level == DiagnosticsEngine::Note,
678 Message, CurrentColumn: OS.tell() - StartOfLocationInfo,
679 Columns: DiagOpts.MessageLength, ShowColors: DiagOpts.ShowColors);
680}
681
682/*static*/ void
683TextDiagnostic::printDiagnosticLevel(raw_ostream &OS,
684 DiagnosticsEngine::Level Level,
685 bool ShowColors) {
686 if (ShowColors) {
687 // Print diagnostic category in bold and color
688 switch (Level) {
689 case DiagnosticsEngine::Ignored:
690 llvm_unreachable("Invalid diagnostic type");
691 case DiagnosticsEngine::Note: OS.changeColor(Color: noteColor, Bold: true); break;
692 case DiagnosticsEngine::Remark: OS.changeColor(Color: remarkColor, Bold: true); break;
693 case DiagnosticsEngine::Warning: OS.changeColor(Color: warningColor, Bold: true); break;
694 case DiagnosticsEngine::Error: OS.changeColor(Color: errorColor, Bold: true); break;
695 case DiagnosticsEngine::Fatal: OS.changeColor(Color: fatalColor, Bold: true); break;
696 }
697 }
698
699 switch (Level) {
700 case DiagnosticsEngine::Ignored:
701 llvm_unreachable("Invalid diagnostic type");
702 case DiagnosticsEngine::Note: OS << "note: "; break;
703 case DiagnosticsEngine::Remark: OS << "remark: "; break;
704 case DiagnosticsEngine::Warning: OS << "warning: "; break;
705 case DiagnosticsEngine::Error: OS << "error: "; break;
706 case DiagnosticsEngine::Fatal: OS << "fatal error: "; break;
707 }
708
709 if (ShowColors)
710 OS.resetColor();
711}
712
713/*static*/
714void TextDiagnostic::printDiagnosticMessage(raw_ostream &OS,
715 bool IsSupplemental,
716 StringRef Message,
717 unsigned CurrentColumn,
718 unsigned Columns, bool ShowColors) {
719 bool Bold = false;
720 if (ShowColors && !IsSupplemental) {
721 // Print primary diagnostic messages in bold and without color, to visually
722 // indicate the transition from continuation notes and other output.
723 OS.changeColor(Color: savedColor, Bold: true);
724 Bold = true;
725 }
726
727 if (Columns)
728 printWordWrapped(OS, Str: Message, Columns, Column: CurrentColumn, Bold);
729 else {
730 bool Normal = true;
731 applyTemplateHighlighting(OS, Str: Message, Normal, Bold);
732 assert(Normal && "Formatting should have returned to normal");
733 }
734
735 if (ShowColors)
736 OS.resetColor();
737 OS << '\n';
738}
739
740void TextDiagnostic::emitFilename(StringRef Filename, const SourceManager &SM) {
741#ifdef _WIN32
742 SmallString<4096> TmpFilename;
743#endif
744 if (DiagOpts.AbsolutePath) {
745 auto File = SM.getFileManager().getOptionalFileRef(Filename);
746 if (File) {
747 // We want to print a simplified absolute path, i. e. without "dots".
748 //
749 // The hardest part here are the paths like "<part1>/<link>/../<part2>".
750 // On Unix-like systems, we cannot just collapse "<link>/..", because
751 // paths are resolved sequentially, and, thereby, the path
752 // "<part1>/<part2>" may point to a different location. That is why
753 // we use FileManager::getCanonicalName(), which expands all indirections
754 // with llvm::sys::fs::real_path() and caches the result.
755 //
756 // On the other hand, it would be better to preserve as much of the
757 // original path as possible, because that helps a user to recognize it.
758 // real_path() expands all links, which sometimes too much. Luckily,
759 // on Windows we can just use llvm::sys::path::remove_dots(), because,
760 // on that system, both aforementioned paths point to the same place.
761#ifdef _WIN32
762 TmpFilename = File->getName();
763 llvm::sys::fs::make_absolute(TmpFilename);
764 llvm::sys::path::native(TmpFilename);
765 llvm::sys::path::remove_dots(TmpFilename, /* remove_dot_dot */ true);
766 Filename = StringRef(TmpFilename.data(), TmpFilename.size());
767#else
768 Filename = SM.getFileManager().getCanonicalName(File: *File);
769#endif
770 }
771 }
772
773 OS << Filename;
774}
775
776/// Print out the file/line/column information and include trace.
777///
778/// This method handles the emission of the diagnostic location information.
779/// This includes extracting as much location information as is present for
780/// the diagnostic and printing it, as well as any include stack or source
781/// ranges necessary.
782void TextDiagnostic::emitDiagnosticLoc(FullSourceLoc Loc, PresumedLoc PLoc,
783 DiagnosticsEngine::Level Level,
784 ArrayRef<CharSourceRange> Ranges) {
785 if (PLoc.isInvalid()) {
786 // At least print the file name if available:
787 if (FileID FID = Loc.getFileID(); FID.isValid()) {
788 if (OptionalFileEntryRef FE = Loc.getFileEntryRef()) {
789 emitFilename(Filename: FE->getName(), SM: Loc.getManager());
790 OS << ": ";
791 }
792 }
793 return;
794 }
795 unsigned LineNo = PLoc.getLine();
796
797 if (!DiagOpts.ShowLocation)
798 return;
799
800 if (DiagOpts.ShowColors)
801 OS.changeColor(Color: savedColor, Bold: true);
802
803 emitFilename(Filename: PLoc.getFilename(), SM: Loc.getManager());
804 switch (DiagOpts.getFormat()) {
805 case DiagnosticOptions::SARIF:
806 case DiagnosticOptions::Clang:
807 if (DiagOpts.ShowLine)
808 OS << ':' << LineNo;
809 break;
810 case DiagnosticOptions::MSVC: OS << '(' << LineNo; break;
811 case DiagnosticOptions::Vi: OS << " +" << LineNo; break;
812 }
813
814 if (DiagOpts.ShowColumn)
815 // Compute the column number.
816 if (unsigned ColNo = PLoc.getColumn()) {
817 if (DiagOpts.getFormat() == DiagnosticOptions::MSVC) {
818 OS << ',';
819 // Visual Studio 2010 or earlier expects column number to be off by one
820 if (LangOpts.MSCompatibilityVersion &&
821 !LangOpts.isCompatibleWithMSVC(MajorVersion: LangOptions::MSVC2012))
822 ColNo--;
823 } else
824 OS << ':';
825 OS << ColNo;
826 }
827 switch (DiagOpts.getFormat()) {
828 case DiagnosticOptions::SARIF:
829 case DiagnosticOptions::Clang:
830 case DiagnosticOptions::Vi: OS << ':'; break;
831 case DiagnosticOptions::MSVC:
832 // MSVC2013 and before print 'file(4) : error'. MSVC2015 gets rid of the
833 // space and prints 'file(4): error'.
834 OS << ')';
835 if (LangOpts.MSCompatibilityVersion &&
836 !LangOpts.isCompatibleWithMSVC(MajorVersion: LangOptions::MSVC2015))
837 OS << ' ';
838 OS << ':';
839 break;
840 }
841
842 if (DiagOpts.ShowSourceRanges && !Ranges.empty()) {
843 FileID CaretFileID = Loc.getExpansionLoc().getFileID();
844 bool PrintedRange = false;
845 const SourceManager &SM = Loc.getManager();
846
847 for (const auto &R : Ranges) {
848 // Ignore invalid ranges.
849 if (!R.isValid())
850 continue;
851
852 SourceLocation B = SM.getExpansionLoc(Loc: R.getBegin());
853 CharSourceRange ERange = SM.getExpansionRange(Loc: R.getEnd());
854 SourceLocation E = ERange.getEnd();
855
856 // If the start or end of the range is in another file, just
857 // discard it.
858 if (SM.getFileID(SpellingLoc: B) != CaretFileID || SM.getFileID(SpellingLoc: E) != CaretFileID)
859 continue;
860
861 // Add in the length of the token, so that we cover multi-char
862 // tokens.
863 unsigned TokSize = 0;
864 if (ERange.isTokenRange())
865 TokSize = Lexer::MeasureTokenLength(Loc: E, SM, LangOpts);
866
867 FullSourceLoc BF(B, SM), EF(E, SM);
868 OS << '{'
869 << BF.getLineNumber() << ':' << BF.getColumnNumber() << '-'
870 << EF.getLineNumber() << ':' << (EF.getColumnNumber() + TokSize)
871 << '}';
872 PrintedRange = true;
873 }
874
875 if (PrintedRange)
876 OS << ':';
877 }
878 OS << ' ';
879}
880
881void TextDiagnostic::emitIncludeLocation(FullSourceLoc Loc, PresumedLoc PLoc) {
882 if (DiagOpts.ShowLocation && PLoc.isValid()) {
883 OS << "In file included from ";
884 emitFilename(Filename: PLoc.getFilename(), SM: Loc.getManager());
885 OS << ':' << PLoc.getLine() << ":\n";
886 } else
887 OS << "In included file:\n";
888}
889
890void TextDiagnostic::emitImportLocation(FullSourceLoc Loc, PresumedLoc PLoc,
891 StringRef ModuleName) {
892 if (DiagOpts.ShowLocation && PLoc.isValid())
893 OS << "In module '" << ModuleName << "' imported from "
894 << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
895 else
896 OS << "In module '" << ModuleName << "':\n";
897}
898
899void TextDiagnostic::emitBuildingModuleLocation(FullSourceLoc Loc,
900 PresumedLoc PLoc,
901 StringRef ModuleName) {
902 if (DiagOpts.ShowLocation && PLoc.isValid())
903 OS << "While building module '" << ModuleName << "' imported from "
904 << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
905 else
906 OS << "While building module '" << ModuleName << "':\n";
907}
908
909/// Find the suitable set of lines to show to include a set of ranges.
910static std::optional<std::pair<unsigned, unsigned>>
911findLinesForRange(const CharSourceRange &R, FileID FID,
912 const SourceManager &SM) {
913 if (!R.isValid())
914 return std::nullopt;
915
916 SourceLocation Begin = R.getBegin();
917 SourceLocation End = R.getEnd();
918 if (SM.getFileID(SpellingLoc: Begin) != FID || SM.getFileID(SpellingLoc: End) != FID)
919 return std::nullopt;
920
921 return std::make_pair(x: SM.getExpansionLineNumber(Loc: Begin),
922 y: SM.getExpansionLineNumber(Loc: End));
923}
924
925/// Add as much of range B into range A as possible without exceeding a maximum
926/// size of MaxRange. Ranges are inclusive.
927static std::pair<unsigned, unsigned>
928maybeAddRange(std::pair<unsigned, unsigned> A, std::pair<unsigned, unsigned> B,
929 unsigned MaxRange) {
930 // If A is already the maximum size, we're done.
931 unsigned Slack = MaxRange - (A.second - A.first + 1);
932 if (Slack == 0)
933 return A;
934
935 // Easy case: merge succeeds within MaxRange.
936 unsigned Min = std::min(a: A.first, b: B.first);
937 unsigned Max = std::max(a: A.second, b: B.second);
938 if (Max - Min + 1 <= MaxRange)
939 return {Min, Max};
940
941 // If we can't reach B from A within MaxRange, there's nothing to do.
942 // Don't add lines to the range that contain nothing interesting.
943 if ((B.first > A.first && B.first - A.first + 1 > MaxRange) ||
944 (B.second < A.second && A.second - B.second + 1 > MaxRange))
945 return A;
946
947 // Otherwise, expand A towards B to produce a range of size MaxRange. We
948 // attempt to expand by the same amount in both directions if B strictly
949 // contains A.
950
951 // Expand downwards by up to half the available amount, then upwards as
952 // much as possible, then downwards as much as possible.
953 A.second = std::min(a: A.second + (Slack + 1) / 2, b: Max);
954 Slack = MaxRange - (A.second - A.first + 1);
955 A.first = std::max(a: Min + Slack, b: A.first) - Slack;
956 A.second = std::min(a: A.first + MaxRange - 1, b: Max);
957 return A;
958}
959
960struct LineRange {
961 unsigned LineNo;
962 unsigned StartCol;
963 unsigned EndCol;
964};
965
966/// Highlight \p R (with ~'s) on the current source line.
967static void highlightRange(const LineRange &R, const SourceColumnMap &Map,
968 std::string &CaretLine) {
969 // Pick the first non-whitespace column.
970 unsigned StartColNo = R.StartCol;
971 while (StartColNo < Map.getSourceLine().size() &&
972 (Map.getSourceLine()[StartColNo] == ' ' ||
973 Map.getSourceLine()[StartColNo] == '\t'))
974 StartColNo = Map.startOfNextColumn(N: StartColNo);
975
976 // Pick the last non-whitespace column.
977 unsigned EndColNo =
978 std::min(a: static_cast<size_t>(R.EndCol), b: Map.getSourceLine().size());
979 while (EndColNo && (Map.getSourceLine()[EndColNo - 1] == ' ' ||
980 Map.getSourceLine()[EndColNo - 1] == '\t'))
981 EndColNo = Map.startOfPreviousColumn(N: EndColNo);
982
983 // If the start/end passed each other, then we are trying to highlight a
984 // range that just exists in whitespace. That most likely means we have
985 // a multi-line highlighting range that covers a blank line.
986 if (StartColNo > EndColNo)
987 return;
988
989 // Fill the range with ~'s.
990 StartColNo = Map.byteToContainingColumn(N: StartColNo);
991 EndColNo = Map.byteToContainingColumn(N: EndColNo);
992
993 assert(StartColNo <= EndColNo && "Invalid range!");
994 if (CaretLine.size() < EndColNo)
995 CaretLine.resize(n: EndColNo, c: ' ');
996 std::fill(first: CaretLine.begin() + StartColNo, last: CaretLine.begin() + EndColNo, value: '~');
997}
998
999static std::string buildFixItInsertionLine(FileID FID, unsigned LineNo,
1000 const SourceColumnMap &map,
1001 ArrayRef<FixItHint> Hints,
1002 const SourceManager &SM,
1003 const DiagnosticOptions &DiagOpts) {
1004 std::string FixItInsertionLine;
1005 if (Hints.empty() || !DiagOpts.ShowFixits)
1006 return FixItInsertionLine;
1007 unsigned PrevHintEndCol = 0;
1008
1009 for (const auto &H : Hints) {
1010 if (H.CodeToInsert.empty())
1011 continue;
1012
1013 // We have an insertion hint. Determine whether the inserted
1014 // code contains no newlines and is on the same line as the caret.
1015 FileIDAndOffset HintLocInfo =
1016 SM.getDecomposedExpansionLoc(Loc: H.RemoveRange.getBegin());
1017 if (FID == HintLocInfo.first &&
1018 LineNo == SM.getLineNumber(FID: HintLocInfo.first, FilePos: HintLocInfo.second) &&
1019 StringRef(H.CodeToInsert).find_first_of(Chars: "\n\r") == StringRef::npos) {
1020 // Insert the new code into the line just below the code
1021 // that the user wrote.
1022 // Note: When modifying this function, be very careful about what is a
1023 // "column" (printed width, platform-dependent) and what is a
1024 // "byte offset" (SourceManager "column").
1025 unsigned HintByteOffset =
1026 SM.getColumnNumber(FID: HintLocInfo.first, FilePos: HintLocInfo.second) - 1;
1027
1028 // The hint must start inside the source or right at the end
1029 assert(HintByteOffset < static_cast<unsigned>(map.bytes()) + 1);
1030 unsigned HintCol = map.byteToContainingColumn(N: HintByteOffset);
1031
1032 // If we inserted a long previous hint, push this one forwards, and add
1033 // an extra space to show that this is not part of the previous
1034 // completion. This is sort of the best we can do when two hints appear
1035 // to overlap.
1036 //
1037 // Note that if this hint is located immediately after the previous
1038 // hint, no space will be added, since the location is more important.
1039 if (HintCol < PrevHintEndCol)
1040 HintCol = PrevHintEndCol + 1;
1041
1042 // This should NOT use HintByteOffset, because the source might have
1043 // Unicode characters in earlier columns.
1044 unsigned NewFixItLineSize = FixItInsertionLine.size() +
1045 (HintCol - PrevHintEndCol) +
1046 H.CodeToInsert.size();
1047 if (NewFixItLineSize > FixItInsertionLine.size())
1048 FixItInsertionLine.resize(n: NewFixItLineSize, c: ' ');
1049
1050 std::copy(first: H.CodeToInsert.begin(), last: H.CodeToInsert.end(),
1051 result: FixItInsertionLine.end() - H.CodeToInsert.size());
1052
1053 PrevHintEndCol = HintCol + llvm::sys::locale::columnWidth(s: H.CodeToInsert);
1054 }
1055 }
1056
1057 expandTabs(SourceLine&: FixItInsertionLine, TabStop: DiagOpts.TabStop);
1058
1059 return FixItInsertionLine;
1060}
1061
1062static unsigned getNumDisplayWidth(unsigned N) {
1063 unsigned L = 1u, M = 10u;
1064 while (M <= N && ++L != std::numeric_limits<unsigned>::digits10 + 1)
1065 M *= 10u;
1066
1067 return L;
1068}
1069
1070/// Filter out invalid ranges, ranges that don't fit into the window of
1071/// source lines we will print, and ranges from other files.
1072///
1073/// For the remaining ranges, convert them to simple LineRange structs,
1074/// which only cover one line at a time.
1075static SmallVector<LineRange>
1076prepareAndFilterRanges(const SmallVectorImpl<CharSourceRange> &Ranges,
1077 const SourceManager &SM,
1078 const std::pair<unsigned, unsigned> &Lines, FileID FID,
1079 const LangOptions &LangOpts) {
1080 SmallVector<LineRange> LineRanges;
1081
1082 for (const CharSourceRange &R : Ranges) {
1083 if (R.isInvalid())
1084 continue;
1085 SourceLocation Begin = R.getBegin();
1086 SourceLocation End = R.getEnd();
1087
1088 unsigned StartLineNo = SM.getExpansionLineNumber(Loc: Begin);
1089 if (StartLineNo > Lines.second || SM.getFileID(SpellingLoc: Begin) != FID)
1090 continue;
1091
1092 unsigned EndLineNo = SM.getExpansionLineNumber(Loc: End);
1093 if (EndLineNo < Lines.first || SM.getFileID(SpellingLoc: End) != FID)
1094 continue;
1095
1096 unsigned StartColumn = SM.getExpansionColumnNumber(Loc: Begin);
1097 unsigned EndColumn = SM.getExpansionColumnNumber(Loc: End);
1098 if (R.isTokenRange())
1099 EndColumn += Lexer::MeasureTokenLength(Loc: End, SM, LangOpts);
1100
1101 // Only a single line.
1102 if (StartLineNo == EndLineNo) {
1103 LineRanges.push_back(Elt: {.LineNo: StartLineNo, .StartCol: StartColumn - 1, .EndCol: EndColumn - 1});
1104 continue;
1105 }
1106
1107 // Start line.
1108 LineRanges.push_back(Elt: {.LineNo: StartLineNo, .StartCol: StartColumn - 1, .EndCol: ~0u});
1109
1110 // Middle lines.
1111 for (unsigned S = StartLineNo + 1; S != EndLineNo; ++S)
1112 LineRanges.push_back(Elt: {.LineNo: S, .StartCol: 0, .EndCol: ~0u});
1113
1114 // End line.
1115 LineRanges.push_back(Elt: {.LineNo: EndLineNo, .StartCol: 0, .EndCol: EndColumn - 1});
1116 }
1117
1118 return LineRanges;
1119}
1120
1121/// Creates syntax highlighting information in form of StyleRanges.
1122///
1123/// The returned unique ptr has always exactly size
1124/// (\p EndLineNumber - \p StartLineNumber + 1). Each SmallVector in there
1125/// corresponds to syntax highlighting information in one line. In each line,
1126/// the StyleRanges are non-overlapping and sorted from start to end of the
1127/// line.
1128static std::unique_ptr<llvm::SmallVector<TextDiagnostic::StyleRange>[]>
1129highlightLines(StringRef FileData, unsigned StartLineNumber,
1130 unsigned EndLineNumber, const Preprocessor *PP,
1131 const LangOptions &LangOpts, bool ShowColors, FileID FID,
1132 const SourceManager &SM) {
1133 assert(StartLineNumber <= EndLineNumber);
1134 auto SnippetRanges =
1135 std::make_unique<SmallVector<TextDiagnostic::StyleRange>[]>(
1136 num: EndLineNumber - StartLineNumber + 1);
1137
1138 if (!PP || !ShowColors)
1139 return SnippetRanges;
1140
1141 // Might cause emission of another diagnostic.
1142 if (PP->getIdentifierTable().getExternalIdentifierLookup())
1143 return SnippetRanges;
1144
1145 auto Buff = llvm::MemoryBuffer::getMemBuffer(InputData: FileData);
1146 Lexer L{FID, *Buff, SM, LangOpts};
1147 L.SetKeepWhitespaceMode(true);
1148
1149 const char *FirstLineStart =
1150 FileData.data() +
1151 SM.getDecomposedLoc(Loc: SM.translateLineCol(FID, Line: StartLineNumber, Col: 1)).second;
1152 if (const char *CheckPoint = PP->getCheckPoint(FID, Start: FirstLineStart)) {
1153 assert(CheckPoint >= Buff->getBufferStart() &&
1154 CheckPoint <= Buff->getBufferEnd());
1155 assert(CheckPoint <= FirstLineStart);
1156 size_t Offset = CheckPoint - Buff->getBufferStart();
1157 L.seek(Offset, /*IsAtStartOfLine=*/false);
1158 }
1159
1160 // Classify the given token and append it to the given vector.
1161 auto appendStyle =
1162 [PP, &LangOpts](SmallVector<TextDiagnostic::StyleRange> &Vec,
1163 const Token &T, unsigned Start, unsigned Length) -> void {
1164 if (T.is(K: tok::raw_identifier)) {
1165 StringRef RawIdent = T.getRawIdentifier();
1166 // Special case true/false/nullptr/... literals, since they will otherwise
1167 // be treated as keywords.
1168 // FIXME: It would be good to have a programmatic way of getting this
1169 // list.
1170 if (llvm::StringSwitch<bool>(RawIdent)
1171 .Case(S: "true", Value: true)
1172 .Case(S: "false", Value: true)
1173 .Case(S: "nullptr", Value: true)
1174 .Case(S: "__func__", Value: true)
1175 .Case(S: "__objc_yes__", Value: true)
1176 .Case(S: "__objc_no__", Value: true)
1177 .Case(S: "__null", Value: true)
1178 .Case(S: "__FUNCDNAME__", Value: true)
1179 .Case(S: "__FUNCSIG__", Value: true)
1180 .Case(S: "__FUNCTION__", Value: true)
1181 .Case(S: "__FUNCSIG__", Value: true)
1182 .Default(Value: false)) {
1183 Vec.emplace_back(Args&: Start, Args: Start + Length, Args: LiteralColor);
1184 } else {
1185 const IdentifierInfo *II = PP->getIdentifierInfo(Name: RawIdent);
1186 assert(II);
1187 if (II->isKeyword(LangOpts))
1188 Vec.emplace_back(Args&: Start, Args: Start + Length, Args: KeywordColor);
1189 }
1190 } else if (tok::isLiteral(K: T.getKind())) {
1191 Vec.emplace_back(Args&: Start, Args: Start + Length, Args: LiteralColor);
1192 } else {
1193 assert(T.is(tok::comment));
1194 Vec.emplace_back(Args&: Start, Args: Start + Length, Args: CommentColor);
1195 }
1196 };
1197
1198 bool Stop = false;
1199 while (!Stop) {
1200 Token T;
1201 Stop = L.LexFromRawLexer(Result&: T);
1202 if (T.is(K: tok::unknown))
1203 continue;
1204
1205 // We are only interested in identifiers, literals and comments.
1206 if (!T.is(K: tok::raw_identifier) && !T.is(K: tok::comment) &&
1207 !tok::isLiteral(K: T.getKind()))
1208 continue;
1209
1210 bool Invalid = false;
1211 unsigned TokenEndLine = SM.getSpellingLineNumber(Loc: T.getEndLoc(), Invalid: &Invalid);
1212 if (Invalid || TokenEndLine < StartLineNumber)
1213 continue;
1214
1215 assert(TokenEndLine >= StartLineNumber);
1216
1217 unsigned TokenStartLine =
1218 SM.getSpellingLineNumber(Loc: T.getLocation(), Invalid: &Invalid);
1219 if (Invalid)
1220 continue;
1221 // If this happens, we're done.
1222 if (TokenStartLine > EndLineNumber)
1223 break;
1224
1225 unsigned StartCol =
1226 SM.getSpellingColumnNumber(Loc: T.getLocation(), Invalid: &Invalid) - 1;
1227 if (Invalid)
1228 continue;
1229
1230 // Simple tokens.
1231 if (TokenStartLine == TokenEndLine) {
1232 SmallVector<TextDiagnostic::StyleRange> &LineRanges =
1233 SnippetRanges[TokenStartLine - StartLineNumber];
1234 appendStyle(LineRanges, T, StartCol, T.getLength());
1235 continue;
1236 }
1237 assert((TokenEndLine - TokenStartLine) >= 1);
1238
1239 // For tokens that span multiple lines (think multiline comments), we
1240 // divide them into multiple StyleRanges.
1241 unsigned EndCol = SM.getSpellingColumnNumber(Loc: T.getEndLoc(), Invalid: &Invalid) - 1;
1242 if (Invalid)
1243 continue;
1244
1245 std::string Spelling = Lexer::getSpelling(Tok: T, SourceMgr: SM, LangOpts);
1246
1247 unsigned L = TokenStartLine;
1248 unsigned LineLength = 0;
1249 for (unsigned I = 0; I <= Spelling.size(); ++I) {
1250 // This line is done.
1251 if (I == Spelling.size() || isVerticalWhitespace(c: Spelling[I])) {
1252 if (L >= StartLineNumber) {
1253 SmallVector<TextDiagnostic::StyleRange> &LineRanges =
1254 SnippetRanges[L - StartLineNumber];
1255
1256 if (L == TokenStartLine) // First line
1257 appendStyle(LineRanges, T, StartCol, LineLength);
1258 else if (L == TokenEndLine) // Last line
1259 appendStyle(LineRanges, T, 0, EndCol);
1260 else
1261 appendStyle(LineRanges, T, 0, LineLength);
1262 }
1263
1264 ++L;
1265 if (L > EndLineNumber)
1266 break;
1267 LineLength = 0;
1268 continue;
1269 }
1270 ++LineLength;
1271 }
1272 }
1273
1274 return SnippetRanges;
1275}
1276
1277/// Emit a code snippet and caret line.
1278///
1279/// This routine emits a single line's code snippet and caret line..
1280///
1281/// \param Loc The location for the caret.
1282/// \param Ranges The underlined ranges for this code snippet.
1283/// \param Hints The FixIt hints active for this diagnostic.
1284void TextDiagnostic::emitSnippetAndCaret(
1285 FullSourceLoc Loc, DiagnosticsEngine::Level Level,
1286 SmallVectorImpl<CharSourceRange> &Ranges, ArrayRef<FixItHint> Hints) {
1287 assert(Loc.isValid() && "must have a valid source location here");
1288 assert(Loc.isFileID() && "must have a file location here");
1289
1290 // If caret diagnostics are enabled and we have location, we want to
1291 // emit the caret. However, we only do this if the location moved
1292 // from the last diagnostic, if the last diagnostic was a note that
1293 // was part of a different warning or error diagnostic, or if the
1294 // diagnostic has ranges. We don't want to emit the same caret
1295 // multiple times if one loc has multiple diagnostics.
1296 if (!DiagOpts.ShowCarets)
1297 return;
1298 if (Loc == LastLoc && Ranges.empty() && Hints.empty() &&
1299 (LastLevel != DiagnosticsEngine::Note || Level == LastLevel))
1300 return;
1301
1302 FileID FID = Loc.getFileID();
1303 const SourceManager &SM = Loc.getManager();
1304
1305 // Get information about the buffer it points into.
1306 bool Invalid = false;
1307 StringRef BufData = Loc.getBufferData(Invalid: &Invalid);
1308 if (Invalid)
1309 return;
1310 const char *BufStart = BufData.data();
1311 const char *BufEnd = BufStart + BufData.size();
1312
1313 unsigned CaretLineNo = Loc.getLineNumber();
1314 unsigned CaretColNo = Loc.getColumnNumber();
1315
1316 // Arbitrarily stop showing snippets when the line is too long.
1317 static const size_t MaxLineLengthToPrint = 4096;
1318 if (CaretColNo > MaxLineLengthToPrint)
1319 return;
1320
1321 // Find the set of lines to include.
1322 const unsigned MaxLines = DiagOpts.SnippetLineLimit;
1323 std::pair<unsigned, unsigned> Lines = {CaretLineNo, CaretLineNo};
1324 unsigned DisplayLineNo = Loc.getPresumedLoc().getLine();
1325 for (const auto &I : Ranges) {
1326 if (auto OptionalRange = findLinesForRange(R: I, FID, SM))
1327 Lines = maybeAddRange(A: Lines, B: *OptionalRange, MaxRange: MaxLines);
1328
1329 DisplayLineNo =
1330 std::min(a: DisplayLineNo, b: SM.getPresumedLineNumber(Loc: I.getBegin()));
1331 }
1332
1333 // Our line numbers look like:
1334 // " [number] | "
1335 // Where [number] is MaxLineNoDisplayWidth columns
1336 // and the full thing is therefore MaxLineNoDisplayWidth + 4 columns.
1337 unsigned MaxLineNoDisplayWidth =
1338 DiagOpts.ShowLineNumbers
1339 ? std::max(a: 4u, b: getNumDisplayWidth(N: DisplayLineNo + MaxLines))
1340 : 0;
1341 auto indentForLineNumbers = [&] {
1342 if (MaxLineNoDisplayWidth > 0)
1343 OS.indent(NumSpaces: MaxLineNoDisplayWidth + 2) << "| ";
1344 };
1345
1346 // Prepare source highlighting information for the lines we're about to
1347 // emit, starting from the first line.
1348 std::unique_ptr<SmallVector<StyleRange>[]> SourceStyles =
1349 highlightLines(FileData: BufData, StartLineNumber: Lines.first, EndLineNumber: Lines.second, PP, LangOpts,
1350 ShowColors: DiagOpts.ShowColors, FID, SM);
1351
1352 SmallVector<LineRange> LineRanges =
1353 prepareAndFilterRanges(Ranges, SM, Lines, FID, LangOpts);
1354
1355 for (unsigned LineNo = Lines.first; LineNo != Lines.second + 1;
1356 ++LineNo, ++DisplayLineNo) {
1357 // Rewind from the current position to the start of the line.
1358 const char *LineStart =
1359 BufStart +
1360 SM.getDecomposedLoc(Loc: SM.translateLineCol(FID, Line: LineNo, Col: 1)).second;
1361 if (LineStart == BufEnd)
1362 break;
1363
1364 // Compute the line end.
1365 const char *LineEnd = LineStart;
1366 while (*LineEnd != '\n' && *LineEnd != '\r' && LineEnd != BufEnd)
1367 ++LineEnd;
1368
1369 // Arbitrarily stop showing snippets when the line is too long.
1370 // FIXME: Don't print any lines in this case.
1371 if (size_t(LineEnd - LineStart) > MaxLineLengthToPrint)
1372 return;
1373
1374 // Copy the line of code into an std::string for ease of manipulation.
1375 std::string SourceLine(LineStart, LineEnd);
1376 // Remove trailing null bytes.
1377 while (!SourceLine.empty() && SourceLine.back() == '\0' &&
1378 (LineNo != CaretLineNo || SourceLine.size() > CaretColNo))
1379 SourceLine.pop_back();
1380
1381 // Build the byte to column map.
1382 const SourceColumnMap sourceColMap(SourceLine, DiagOpts.TabStop);
1383
1384 std::string CaretLine;
1385 // Highlight all of the characters covered by Ranges with ~ characters.
1386 for (const auto &LR : LineRanges) {
1387 if (LR.LineNo == LineNo)
1388 highlightRange(R: LR, Map: sourceColMap, CaretLine);
1389 }
1390
1391 // Next, insert the caret itself.
1392 if (CaretLineNo == LineNo) {
1393 size_t Col = sourceColMap.byteToContainingColumn(N: CaretColNo - 1);
1394 CaretLine.resize(n: std::max(a: Col + 1, b: CaretLine.size()), c: ' ');
1395 CaretLine[Col] = '^';
1396 }
1397
1398 std::string FixItInsertionLine =
1399 buildFixItInsertionLine(FID, LineNo, map: sourceColMap, Hints, SM, DiagOpts);
1400
1401 // If the source line is too long for our terminal, select only the
1402 // "interesting" source region within that line.
1403 unsigned Columns = DiagOpts.MessageLength;
1404 if (Columns)
1405 selectInterestingSourceRegion(SourceLine, CaretLine, FixItInsertionLine,
1406 Columns, map: sourceColMap);
1407
1408 // If we are in -fdiagnostics-print-source-range-info mode, we are trying
1409 // to produce easily machine parsable output. Add a space before the
1410 // source line and the caret to make it trivial to tell the main diagnostic
1411 // line from what the user is intended to see.
1412 if (DiagOpts.ShowSourceRanges && !SourceLine.empty()) {
1413 SourceLine = ' ' + SourceLine;
1414 CaretLine = ' ' + CaretLine;
1415 }
1416
1417 // Emit what we have computed.
1418 emitSnippet(SourceLine, MaxLineNoDisplayWidth, LineNo, DisplayLineNo,
1419 Styles: SourceStyles[LineNo - Lines.first]);
1420
1421 if (!CaretLine.empty()) {
1422 indentForLineNumbers();
1423 if (DiagOpts.ShowColors)
1424 OS.changeColor(Color: caretColor, Bold: true);
1425 OS << CaretLine << '\n';
1426 if (DiagOpts.ShowColors)
1427 OS.resetColor();
1428 }
1429
1430 if (!FixItInsertionLine.empty()) {
1431 indentForLineNumbers();
1432 if (DiagOpts.ShowColors)
1433 // Print fixit line in color
1434 OS.changeColor(Color: fixitColor, Bold: false);
1435 if (DiagOpts.ShowSourceRanges)
1436 OS << ' ';
1437 OS << FixItInsertionLine << '\n';
1438 if (DiagOpts.ShowColors)
1439 OS.resetColor();
1440 }
1441 }
1442
1443 // Print out any parseable fixit information requested by the options.
1444 emitParseableFixits(Hints, SM);
1445}
1446
1447void TextDiagnostic::emitSnippet(StringRef SourceLine,
1448 unsigned MaxLineNoDisplayWidth,
1449 unsigned LineNo, unsigned DisplayLineNo,
1450 ArrayRef<StyleRange> Styles) {
1451 // Emit line number.
1452 if (MaxLineNoDisplayWidth > 0) {
1453 unsigned LineNoDisplayWidth = getNumDisplayWidth(N: DisplayLineNo);
1454 OS.indent(NumSpaces: MaxLineNoDisplayWidth - LineNoDisplayWidth + 1)
1455 << DisplayLineNo << " | ";
1456 }
1457
1458 // Print the source line one character at a time.
1459 bool PrintReversed = false;
1460 std::optional<llvm::raw_ostream::Colors> CurrentColor;
1461 size_t I = 0;
1462 while (I < SourceLine.size()) {
1463 auto [Str, WasPrintable] =
1464 printableTextForNextCharacter(SourceLine, I: &I, TabStop: DiagOpts.TabStop);
1465
1466 // Toggle inverted colors on or off for this character.
1467 if (DiagOpts.ShowColors) {
1468 if (WasPrintable == PrintReversed) {
1469 PrintReversed = !PrintReversed;
1470 if (PrintReversed)
1471 OS.reverseColor();
1472 else {
1473 OS.resetColor();
1474 CurrentColor = std::nullopt;
1475 }
1476 }
1477
1478 // Apply syntax highlighting information.
1479 const auto *CharStyle = llvm::find_if(Range&: Styles, P: [I](const StyleRange &R) {
1480 return (R.Start < I && R.End >= I);
1481 });
1482
1483 if (CharStyle != Styles.end()) {
1484 if (!CurrentColor ||
1485 (CurrentColor && *CurrentColor != CharStyle->Color)) {
1486 OS.changeColor(Color: CharStyle->Color, Bold: false);
1487 CurrentColor = CharStyle->Color;
1488 }
1489 } else if (CurrentColor) {
1490 OS.resetColor();
1491 CurrentColor = std::nullopt;
1492 }
1493 }
1494
1495 OS << Str;
1496 }
1497
1498 if (DiagOpts.ShowColors)
1499 OS.resetColor();
1500
1501 OS << '\n';
1502}
1503
1504void TextDiagnostic::emitParseableFixits(ArrayRef<FixItHint> Hints,
1505 const SourceManager &SM) {
1506 if (!DiagOpts.ShowParseableFixits)
1507 return;
1508
1509 // We follow FixItRewriter's example in not (yet) handling
1510 // fix-its in macros.
1511 for (const auto &H : Hints) {
1512 if (H.RemoveRange.isInvalid() || H.RemoveRange.getBegin().isMacroID() ||
1513 H.RemoveRange.getEnd().isMacroID())
1514 return;
1515 }
1516
1517 for (const auto &H : Hints) {
1518 SourceLocation BLoc = H.RemoveRange.getBegin();
1519 SourceLocation ELoc = H.RemoveRange.getEnd();
1520
1521 FileIDAndOffset BInfo = SM.getDecomposedLoc(Loc: BLoc);
1522 FileIDAndOffset EInfo = SM.getDecomposedLoc(Loc: ELoc);
1523
1524 // Adjust for token ranges.
1525 if (H.RemoveRange.isTokenRange())
1526 EInfo.second += Lexer::MeasureTokenLength(Loc: ELoc, SM, LangOpts);
1527
1528 // We specifically do not do word-wrapping or tab-expansion here,
1529 // because this is supposed to be easy to parse.
1530 PresumedLoc PLoc = SM.getPresumedLoc(Loc: BLoc);
1531 if (PLoc.isInvalid())
1532 break;
1533
1534 OS << "fix-it:\"";
1535 OS.write_escaped(Str: PLoc.getFilename());
1536 OS << "\":{" << SM.getLineNumber(FID: BInfo.first, FilePos: BInfo.second)
1537 << ':' << SM.getColumnNumber(FID: BInfo.first, FilePos: BInfo.second)
1538 << '-' << SM.getLineNumber(FID: EInfo.first, FilePos: EInfo.second)
1539 << ':' << SM.getColumnNumber(FID: EInfo.first, FilePos: EInfo.second)
1540 << "}:\"";
1541 OS.write_escaped(Str: H.CodeToInsert);
1542 OS << "\"\n";
1543 }
1544}
1545