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/SmallString.h"
17#include "llvm/ADT/StringExtras.h"
18#include "llvm/Support/ConvertUTF.h"
19#include "llvm/Support/ErrorHandling.h"
20#include "llvm/Support/Locale.h"
21#include "llvm/Support/Path.h"
22#include "llvm/Support/raw_ostream.h"
23#include <algorithm>
24#include <optional>
25
26using namespace clang;
27
28static const enum raw_ostream::Colors noteColor = raw_ostream::CYAN;
29static const enum raw_ostream::Colors remarkColor =
30 raw_ostream::BLUE;
31static const enum raw_ostream::Colors fixitColor =
32 raw_ostream::GREEN;
33static const enum raw_ostream::Colors caretColor =
34 raw_ostream::GREEN;
35static const enum raw_ostream::Colors warningColor =
36 raw_ostream::MAGENTA;
37static const enum raw_ostream::Colors templateColor =
38 raw_ostream::CYAN;
39static const enum raw_ostream::Colors errorColor = raw_ostream::RED;
40static const enum raw_ostream::Colors fatalColor = raw_ostream::RED;
41// Used for changing only the bold attribute.
42static const enum raw_ostream::Colors savedColor =
43 raw_ostream::SAVEDCOLOR;
44
45// Magenta is taken for 'warning'. Red is already 'error' and 'cyan'
46// is already taken for 'note'. Green is already used to underline
47// source ranges. White and black are bad because of the usual
48// terminal backgrounds. Which leaves us only with TWO options.
49static constexpr raw_ostream::Colors CommentColor = raw_ostream::YELLOW;
50static constexpr raw_ostream::Colors LiteralColor = raw_ostream::GREEN;
51static constexpr raw_ostream::Colors KeywordColor = raw_ostream::BLUE;
52
53/// Add highlights to differences in template strings.
54static void applyTemplateHighlighting(raw_ostream &OS, StringRef Str,
55 bool &Normal, bool Bold) {
56 while (true) {
57 size_t Pos = Str.find(C: ToggleHighlight);
58 OS << Str.slice(Start: 0, End: Pos);
59 if (Pos == StringRef::npos)
60 break;
61
62 Str = Str.substr(Start: Pos + 1);
63 if (Normal)
64 OS.changeColor(Color: templateColor, Bold: true);
65 else {
66 OS.resetColor();
67 if (Bold)
68 OS.changeColor(Color: savedColor, Bold: true);
69 }
70 Normal = !Normal;
71 }
72}
73
74/// Number of spaces to indent when word-wrapping.
75const unsigned WordWrapIndentation = 6;
76
77static int bytesSincePreviousTabOrLineBegin(StringRef SourceLine, size_t i) {
78 int bytes = 0;
79 while (0<i) {
80 if (SourceLine[--i]=='\t')
81 break;
82 ++bytes;
83 }
84 return bytes;
85}
86
87/// returns a printable representation of first item from input range
88///
89/// This function returns a printable representation of the next item in a line
90/// of source. If the next byte begins a valid and printable character, that
91/// character is returned along with 'true'.
92///
93/// Otherwise, if the next byte begins a valid, but unprintable character, a
94/// printable, escaped representation of the character is returned, along with
95/// 'false'. Otherwise a printable, escaped representation of the next byte
96/// is returned along with 'false'.
97///
98/// \note The index is updated to be used with a subsequent call to
99/// printableTextForNextCharacter.
100///
101/// \param SourceLine The line of source
102/// \param I Pointer to byte index,
103/// \param TabStop used to expand tabs
104/// \return pair(printable text, 'true' iff original text was printable)
105///
106static std::pair<SmallString<16>, bool>
107printableTextForNextCharacter(StringRef SourceLine, size_t *I,
108 unsigned TabStop) {
109 assert(I && "I must not be null");
110 assert(*I < SourceLine.size() && "must point to a valid index");
111
112 if (SourceLine[*I] == '\t') {
113 assert(0 < TabStop && TabStop <= DiagnosticOptions::MaxTabStop &&
114 "Invalid -ftabstop value");
115 unsigned Col = bytesSincePreviousTabOrLineBegin(SourceLine, i: *I);
116 unsigned NumSpaces = TabStop - (Col % TabStop);
117 assert(0 < NumSpaces && NumSpaces <= TabStop
118 && "Invalid computation of space amt");
119 ++(*I);
120
121 SmallString<16> ExpandedTab;
122 ExpandedTab.assign(NumElts: NumSpaces, Elt: ' ');
123 return std::make_pair(x&: ExpandedTab, y: true);
124 }
125
126 const unsigned char *Begin = SourceLine.bytes_begin() + *I;
127
128 // Fast path for the common ASCII case.
129 if (*Begin < 0x80 && llvm::sys::locale::isPrint(c: *Begin)) {
130 ++(*I);
131 return std::make_pair(x: SmallString<16>(Begin, Begin + 1), y: true);
132 }
133 unsigned CharSize = llvm::getNumBytesForUTF8(firstByte: *Begin);
134 const unsigned char *End = Begin + CharSize;
135
136 // Convert it to UTF32 and check if it's printable.
137 if (End <= SourceLine.bytes_end() && llvm::isLegalUTF8Sequence(source: Begin, sourceEnd: End)) {
138 llvm::UTF32 C;
139 llvm::UTF32 *CPtr = &C;
140
141 // Begin and end before conversion.
142 unsigned char const *OriginalBegin = Begin;
143 llvm::ConversionResult Res = llvm::ConvertUTF8toUTF32(
144 sourceStart: &Begin, sourceEnd: End, targetStart: &CPtr, targetEnd: CPtr + 1, flags: llvm::strictConversion);
145 (void)Res;
146 assert(Res == llvm::conversionOK);
147 assert(OriginalBegin < Begin);
148 assert(unsigned(Begin - OriginalBegin) == CharSize);
149
150 (*I) += (Begin - OriginalBegin);
151
152 // Valid, multi-byte, printable UTF8 character.
153 if (llvm::sys::locale::isPrint(c: C))
154 return std::make_pair(x: SmallString<16>(OriginalBegin, End), y: true);
155
156 // Valid but not printable.
157 SmallString<16> Str("<U+>");
158 while (C) {
159 Str.insert(I: Str.begin() + 3, Elt: llvm::hexdigit(X: C % 16));
160 C /= 16;
161 }
162 while (Str.size() < 8)
163 Str.insert(I: Str.begin() + 3, Elt: llvm::hexdigit(X: 0));
164 return std::make_pair(x&: Str, y: false);
165 }
166
167 // Otherwise, not printable since it's not valid UTF8.
168 SmallString<16> ExpandedByte("<XX>");
169 unsigned char Byte = SourceLine[*I];
170 ExpandedByte[1] = llvm::hexdigit(X: Byte / 16);
171 ExpandedByte[2] = llvm::hexdigit(X: Byte % 16);
172 ++(*I);
173 return std::make_pair(x&: ExpandedByte, y: false);
174}
175
176static void expandTabs(std::string &SourceLine, unsigned TabStop) {
177 size_t I = SourceLine.size();
178 while (I > 0) {
179 I--;
180 if (SourceLine[I] != '\t')
181 continue;
182 size_t TmpI = I;
183 auto [Str, Printable] =
184 printableTextForNextCharacter(SourceLine, I: &TmpI, TabStop);
185 SourceLine.replace(pos: I, n1: 1, s: Str.c_str());
186 }
187}
188
189/// \p BytesOut:
190/// A mapping from columns to the byte of the source line that produced the
191/// character displaying at that column. This is the inverse of \p ColumnsOut.
192///
193/// The last element in the array is the number of bytes in the source string.
194///
195/// example: (given a tabstop of 8)
196///
197/// "a \t \u3042" -> {0,1,2,-1,-1,-1,-1,-1,3,4,-1,7}
198///
199/// (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
200/// display)
201///
202/// \p ColumnsOut:
203/// A mapping from the bytes
204/// of the printable representation of the line to the columns those printable
205/// characters will appear at (numbering the first column as 0).
206///
207/// If a byte 'i' corresponds to multiple columns (e.g. the byte contains a tab
208/// character) then the array will map that byte to the first column the
209/// tab appears at and the next value in the map will have been incremented
210/// more than once.
211///
212/// If a byte is the first in a sequence of bytes that together map to a single
213/// entity in the output, then the array will map that byte to the appropriate
214/// column while the subsequent bytes will be -1.
215///
216/// The last element in the array does not correspond to any byte in the input
217/// and instead is the number of columns needed to display the source
218///
219/// example: (given a tabstop of 8)
220///
221/// "a \t \u3042" -> {0,1,2,8,9,-1,-1,11}
222///
223/// (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
224/// display)
225static void genColumnByteMapping(StringRef SourceLine, unsigned TabStop,
226 SmallVectorImpl<int> &BytesOut,
227 SmallVectorImpl<int> &ColumnsOut) {
228 assert(BytesOut.empty());
229 assert(ColumnsOut.empty());
230
231 if (SourceLine.empty()) {
232 BytesOut.resize(N: 1u, NV: 0);
233 ColumnsOut.resize(N: 1u, NV: 0);
234 return;
235 }
236
237 ColumnsOut.resize(N: SourceLine.size() + 1, NV: -1);
238
239 int Columns = 0;
240 size_t I = 0;
241 while (I < SourceLine.size()) {
242 ColumnsOut[I] = Columns;
243 BytesOut.resize(N: Columns + 1, NV: -1);
244 BytesOut.back() = I;
245 auto [Str, Printable] =
246 printableTextForNextCharacter(SourceLine, I: &I, TabStop);
247 Columns += llvm::sys::locale::columnWidth(s: Str);
248 }
249
250 ColumnsOut.back() = Columns;
251 BytesOut.resize(N: Columns + 1, NV: -1);
252 BytesOut.back() = I;
253}
254
255namespace {
256struct SourceColumnMap {
257 SourceColumnMap(StringRef SourceLine, unsigned TabStop)
258 : m_SourceLine(SourceLine) {
259
260 genColumnByteMapping(SourceLine, TabStop, BytesOut&: m_columnToByte, ColumnsOut&: m_byteToColumn);
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 }
269 int columns() const { return m_byteToColumn.back(); }
270 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).
274 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.
280 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.
290 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.
296 int startOfNextColumn(int N) const {
297 assert(0 <= N && N < static_cast<int>(m_byteToColumn.size() - 1));
298 while (byteToColumn(n: ++N) == -1) {}
299 return N;
300 }
301
302 /// Map from a byte index to the previous byte which starts a column.
303 int startOfPreviousColumn(int N) const {
304 assert(0 < N && N < static_cast<int>(m_byteToColumn.size()));
305 while (byteToColumn(n: --N) == -1) {}
306 return N;
307 }
308
309 StringRef getSourceLine() const {
310 return m_SourceLine;
311 }
312
313private:
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.
322static 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(s: FixItInsertionLine);
329 unsigned MaxColumns = std::max(a: static_cast<unsigned>(map.columns()),
330 b: std::max(a: CaretColumns, b: 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(c: CaretLine[CaretStart]))
343 break;
344
345 for (; CaretEnd != CaretStart; --CaretEnd)
346 if (!isWhitespace(c: 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(c: FixItInsertionLine[FixItStart]))
358 break;
359
360 for (; FixItEnd != FixItStart; --FixItEnd)
361 if (!isWhitespace(c: 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(s: FixItInsertionLine.substr(pos: 0, n: FixItEnd));
370
371 CaretStart = std::min(a: FixItStartCol, b: CaretStart);
372 CaretEnd = std::max(a: FixItEndCol, b: 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(n: 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(n: std::min<unsigned>(a: CaretStart,
397 b: map.columns()));
398 unsigned SourceEnd = map.columnToByte(n: std::min<unsigned>(a: CaretEnd,
399 b: map.columns()));
400
401 unsigned CaretColumnsOutsideSource = CaretEnd-CaretStart
402 - (map.byteToColumn(n: SourceEnd)-map.byteToColumn(n: SourceStart));
403
404 char const *front_ellipse = " ...";
405 char const *front_space = " ";
406 char const *back_ellipse = "...";
407 unsigned ellipses_space = strlen(s: front_ellipse) + strlen(s: 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(N: 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(c: SourceLine[NewStart]))
425 NewStart = map.startOfPreviousColumn(N: NewStart);
426
427 // Skip over this bit of "interesting" text.
428 while (NewStart) {
429 unsigned Prev = map.startOfPreviousColumn(N: NewStart);
430 if (isWhitespace(c: SourceLine[Prev]))
431 break;
432 NewStart = Prev;
433 }
434
435 assert(map.byteToColumn(NewStart) != -1);
436 unsigned NewColumns = map.byteToColumn(n: SourceEnd) -
437 map.byteToColumn(n: NewStart);
438 if (NewColumns <= TargetColumns) {
439 SourceStart = NewStart;
440 ExpandedRegion = true;
441 }
442 }
443
444 if (SourceEnd<SourceLine.size()) {
445 unsigned NewEnd = map.startOfNextColumn(N: 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(c: SourceLine[NewEnd]))
451 NewEnd = map.startOfNextColumn(N: NewEnd);
452
453 // Skip over this bit of "interesting" text.
454 while (NewEnd < SourceLine.size() && isWhitespace(c: SourceLine[NewEnd]))
455 NewEnd = map.startOfNextColumn(N: NewEnd);
456
457 assert(map.byteToColumn(NewEnd) != -1);
458 unsigned NewColumns = map.byteToColumn(n: NewEnd) -
459 map.byteToColumn(n: 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(n: SourceStart);
471 CaretEnd = map.byteToColumn(n: SourceEnd) + CaretColumnsOutsideSource;
472
473 // [CaretStart, CaretEnd) is the slice we want. Update the various
474 // output lines to show only this slice.
475 assert(CaretStart!=(unsigned)-1 && CaretEnd!=(unsigned)-1 &&
476 SourceStart!=(unsigned)-1 && SourceEnd!=(unsigned)-1);
477 assert(SourceStart <= SourceEnd);
478 assert(CaretStart <= CaretEnd);
479
480 unsigned BackColumnsRemoved
481 = map.byteToColumn(n: SourceLine.size())-map.byteToColumn(n: SourceEnd);
482 unsigned FrontColumnsRemoved = CaretStart;
483 unsigned ColumnsKept = CaretEnd-CaretStart;
484
485 // We checked up front that the line needed truncation
486 assert(FrontColumnsRemoved+ColumnsKept+BackColumnsRemoved > Columns);
487
488 // The line needs some truncation, and we'd prefer to keep the front
489 // if possible, so remove the back
490 if (BackColumnsRemoved > strlen(s: back_ellipse))
491 SourceLine.replace(pos: SourceEnd, n1: std::string::npos, s: back_ellipse);
492
493 // If that's enough then we're done
494 if (FrontColumnsRemoved+ColumnsKept <= Columns)
495 return;
496
497 // Otherwise remove the front as well
498 if (FrontColumnsRemoved > strlen(s: front_ellipse)) {
499 SourceLine.replace(pos: 0, n1: SourceStart, s: front_ellipse);
500 CaretLine.replace(pos: 0, n1: CaretStart, s: front_space);
501 if (!FixItInsertionLine.empty())
502 FixItInsertionLine.replace(pos: 0, n1: CaretStart, s: front_space);
503 }
504}
505
506/// Skip over whitespace in the string, starting at the given
507/// index.
508///
509/// \returns The index of the first non-whitespace character that is
510/// greater than or equal to Idx or, if no such character exists,
511/// returns the end of the string.
512static unsigned skipWhitespace(unsigned Idx, StringRef Str, unsigned Length) {
513 while (Idx < Length && isWhitespace(c: Str[Idx]))
514 ++Idx;
515 return Idx;
516}
517
518/// If the given character is the start of some kind of
519/// balanced punctuation (e.g., quotes or parentheses), return the
520/// character that will terminate the punctuation.
521///
522/// \returns The ending punctuation character, if any, or the NULL
523/// character if the input character does not start any punctuation.
524static inline char findMatchingPunctuation(char c) {
525 switch (c) {
526 case '\'': return '\'';
527 case '`': return '\'';
528 case '"': return '"';
529 case '(': return ')';
530 case '[': return ']';
531 case '{': return '}';
532 default: break;
533 }
534
535 return 0;
536}
537
538/// Find the end of the word starting at the given offset
539/// within a string.
540///
541/// \returns the index pointing one character past the end of the
542/// word.
543static unsigned findEndOfWord(unsigned Start, StringRef Str,
544 unsigned Length, unsigned Column,
545 unsigned Columns) {
546 assert(Start < Str.size() && "Invalid start position!");
547 unsigned End = Start + 1;
548
549 // If we are already at the end of the string, take that as the word.
550 if (End == Str.size())
551 return End;
552
553 // Determine if the start of the string is actually opening
554 // punctuation, e.g., a quote or parentheses.
555 char EndPunct = findMatchingPunctuation(c: Str[Start]);
556 if (!EndPunct) {
557 // This is a normal word. Just find the first space character.
558 while (End < Length && !isWhitespace(c: Str[End]))
559 ++End;
560 return End;
561 }
562
563 // We have the start of a balanced punctuation sequence (quotes,
564 // parentheses, etc.). Determine the full sequence is.
565 SmallString<16> PunctuationEndStack;
566 PunctuationEndStack.push_back(Elt: EndPunct);
567 while (End < Length && !PunctuationEndStack.empty()) {
568 if (Str[End] == PunctuationEndStack.back())
569 PunctuationEndStack.pop_back();
570 else if (char SubEndPunct = findMatchingPunctuation(c: Str[End]))
571 PunctuationEndStack.push_back(Elt: SubEndPunct);
572
573 ++End;
574 }
575
576 // Find the first space character after the punctuation ended.
577 while (End < Length && !isWhitespace(c: Str[End]))
578 ++End;
579
580 unsigned PunctWordLength = End - Start;
581 if (// If the word fits on this line
582 Column + PunctWordLength <= Columns ||
583 // ... or the word is "short enough" to take up the next line
584 // without too much ugly white space
585 PunctWordLength < Columns/3)
586 return End; // Take the whole thing as a single "word".
587
588 // The whole quoted/parenthesized string is too long to print as a
589 // single "word". Instead, find the "word" that starts just after
590 // the punctuation and use that end-point instead. This will recurse
591 // until it finds something small enough to consider a word.
592 return findEndOfWord(Start: Start + 1, Str, Length, Column: Column + 1, Columns);
593}
594
595/// Print the given string to a stream, word-wrapping it to
596/// some number of columns in the process.
597///
598/// \param OS the stream to which the word-wrapping string will be
599/// emitted.
600/// \param Str the string to word-wrap and output.
601/// \param Columns the number of columns to word-wrap to.
602/// \param Column the column number at which the first character of \p
603/// Str will be printed. This will be non-zero when part of the first
604/// line has already been printed.
605/// \param Bold if the current text should be bold
606/// \returns true if word-wrapping was required, or false if the
607/// string fit on the first line.
608static bool printWordWrapped(raw_ostream &OS, StringRef Str, unsigned Columns,
609 unsigned Column, bool Bold) {
610 const unsigned Length = std::min(a: Str.find(C: '\n'), b: Str.size());
611 bool TextNormal = true;
612
613 bool Wrapped = false;
614 for (unsigned WordStart = 0, WordEnd; WordStart < Length;
615 WordStart = WordEnd) {
616 // Find the beginning of the next word.
617 WordStart = skipWhitespace(Idx: WordStart, Str, Length);
618 if (WordStart == Length)
619 break;
620
621 // Find the end of this word.
622 WordEnd = findEndOfWord(Start: WordStart, Str, Length, Column, Columns);
623
624 // Does this word fit on the current line?
625 unsigned WordLength = WordEnd - WordStart;
626 if (Column + WordLength < Columns) {
627 // This word fits on the current line; print it there.
628 if (WordStart) {
629 OS << ' ';
630 Column += 1;
631 }
632 applyTemplateHighlighting(OS, Str: Str.substr(Start: WordStart, N: WordLength),
633 Normal&: TextNormal, Bold);
634 Column += WordLength;
635 continue;
636 }
637
638 // This word does not fit on the current line, so wrap to the next
639 // line.
640 OS << '\n';
641 OS.indent(NumSpaces: WordWrapIndentation);
642 applyTemplateHighlighting(OS, Str: Str.substr(Start: WordStart, N: WordLength),
643 Normal&: TextNormal, Bold);
644 Column = WordWrapIndentation + WordLength;
645 Wrapped = true;
646 }
647
648 // Append any remaning text from the message with its existing formatting.
649 applyTemplateHighlighting(OS, Str: Str.substr(Start: Length), Normal&: TextNormal, Bold);
650
651 assert(TextNormal && "Text highlighted at end of diagnostic message.");
652
653 return Wrapped;
654}
655
656TextDiagnostic::TextDiagnostic(raw_ostream &OS, const LangOptions &LangOpts,
657 DiagnosticOptions *DiagOpts,
658 const Preprocessor *PP)
659 : DiagnosticRenderer(LangOpts, DiagOpts), OS(OS), PP(PP) {}
660
661TextDiagnostic::~TextDiagnostic() {}
662
663void TextDiagnostic::emitDiagnosticMessage(
664 FullSourceLoc Loc, PresumedLoc PLoc, DiagnosticsEngine::Level Level,
665 StringRef Message, ArrayRef<clang::CharSourceRange> Ranges,
666 DiagOrStoredDiag D) {
667 uint64_t StartOfLocationInfo = OS.tell();
668
669 // Emit the location of this particular diagnostic.
670 if (Loc.isValid())
671 emitDiagnosticLoc(Loc, PLoc, Level, Ranges);
672
673 if (DiagOpts->ShowColors)
674 OS.resetColor();
675
676 if (DiagOpts->ShowLevel)
677 printDiagnosticLevel(OS, Level, ShowColors: DiagOpts->ShowColors);
678 printDiagnosticMessage(OS,
679 /*IsSupplemental*/ Level == DiagnosticsEngine::Note,
680 Message, CurrentColumn: OS.tell() - StartOfLocationInfo,
681 Columns: DiagOpts->MessageLength, ShowColors: DiagOpts->ShowColors);
682}
683
684/*static*/ void
685TextDiagnostic::printDiagnosticLevel(raw_ostream &OS,
686 DiagnosticsEngine::Level Level,
687 bool ShowColors) {
688 if (ShowColors) {
689 // Print diagnostic category in bold and color
690 switch (Level) {
691 case DiagnosticsEngine::Ignored:
692 llvm_unreachable("Invalid diagnostic type");
693 case DiagnosticsEngine::Note: OS.changeColor(Color: noteColor, Bold: true); break;
694 case DiagnosticsEngine::Remark: OS.changeColor(Color: remarkColor, Bold: true); break;
695 case DiagnosticsEngine::Warning: OS.changeColor(Color: warningColor, Bold: true); break;
696 case DiagnosticsEngine::Error: OS.changeColor(Color: errorColor, Bold: true); break;
697 case DiagnosticsEngine::Fatal: OS.changeColor(Color: fatalColor, Bold: true); break;
698 }
699 }
700
701 switch (Level) {
702 case DiagnosticsEngine::Ignored:
703 llvm_unreachable("Invalid diagnostic type");
704 case DiagnosticsEngine::Note: OS << "note: "; break;
705 case DiagnosticsEngine::Remark: OS << "remark: "; break;
706 case DiagnosticsEngine::Warning: OS << "warning: "; break;
707 case DiagnosticsEngine::Error: OS << "error: "; break;
708 case DiagnosticsEngine::Fatal: OS << "fatal error: "; break;
709 }
710
711 if (ShowColors)
712 OS.resetColor();
713}
714
715/*static*/
716void TextDiagnostic::printDiagnosticMessage(raw_ostream &OS,
717 bool IsSupplemental,
718 StringRef Message,
719 unsigned CurrentColumn,
720 unsigned Columns, bool ShowColors) {
721 bool Bold = false;
722 if (ShowColors && !IsSupplemental) {
723 // Print primary diagnostic messages in bold and without color, to visually
724 // indicate the transition from continuation notes and other output.
725 OS.changeColor(Color: savedColor, Bold: true);
726 Bold = true;
727 }
728
729 if (Columns)
730 printWordWrapped(OS, Str: Message, Columns, Column: CurrentColumn, Bold);
731 else {
732 bool Normal = true;
733 applyTemplateHighlighting(OS, Str: Message, Normal, Bold);
734 assert(Normal && "Formatting should have returned to normal");
735 }
736
737 if (ShowColors)
738 OS.resetColor();
739 OS << '\n';
740}
741
742void TextDiagnostic::emitFilename(StringRef Filename, const SourceManager &SM) {
743#ifdef _WIN32
744 SmallString<4096> TmpFilename;
745#endif
746 if (DiagOpts->AbsolutePath) {
747 auto File = SM.getFileManager().getOptionalFileRef(Filename);
748 if (File) {
749 // We want to print a simplified absolute path, i. e. without "dots".
750 //
751 // The hardest part here are the paths like "<part1>/<link>/../<part2>".
752 // On Unix-like systems, we cannot just collapse "<link>/..", because
753 // paths are resolved sequentially, and, thereby, the path
754 // "<part1>/<part2>" may point to a different location. That is why
755 // we use FileManager::getCanonicalName(), which expands all indirections
756 // with llvm::sys::fs::real_path() and caches the result.
757 //
758 // On the other hand, it would be better to preserve as much of the
759 // original path as possible, because that helps a user to recognize it.
760 // real_path() expands all links, which sometimes too much. Luckily,
761 // on Windows we can just use llvm::sys::path::remove_dots(), because,
762 // on that system, both aforementioned paths point to the same place.
763#ifdef _WIN32
764 TmpFilename = File->getName();
765 llvm::sys::fs::make_absolute(TmpFilename);
766 llvm::sys::path::native(TmpFilename);
767 llvm::sys::path::remove_dots(TmpFilename, /* remove_dot_dot */ true);
768 Filename = StringRef(TmpFilename.data(), TmpFilename.size());
769#else
770 Filename = SM.getFileManager().getCanonicalName(File: *File);
771#endif
772 }
773 }
774
775 OS << Filename;
776}
777
778/// Print out the file/line/column information and include trace.
779///
780/// This method handles the emission of the diagnostic location information.
781/// This includes extracting as much location information as is present for
782/// the diagnostic and printing it, as well as any include stack or source
783/// ranges necessary.
784void TextDiagnostic::emitDiagnosticLoc(FullSourceLoc Loc, PresumedLoc PLoc,
785 DiagnosticsEngine::Level Level,
786 ArrayRef<CharSourceRange> Ranges) {
787 if (PLoc.isInvalid()) {
788 // At least print the file name if available:
789 if (FileID FID = Loc.getFileID(); FID.isValid()) {
790 if (OptionalFileEntryRef FE = Loc.getFileEntryRef()) {
791 emitFilename(Filename: FE->getName(), SM: Loc.getManager());
792 OS << ": ";
793 }
794 }
795 return;
796 }
797 unsigned LineNo = PLoc.getLine();
798
799 if (!DiagOpts->ShowLocation)
800 return;
801
802 if (DiagOpts->ShowColors)
803 OS.changeColor(Color: savedColor, Bold: true);
804
805 emitFilename(Filename: PLoc.getFilename(), SM: Loc.getManager());
806 switch (DiagOpts->getFormat()) {
807 case DiagnosticOptions::SARIF:
808 case DiagnosticOptions::Clang:
809 if (DiagOpts->ShowLine)
810 OS << ':' << LineNo;
811 break;
812 case DiagnosticOptions::MSVC: OS << '(' << LineNo; break;
813 case DiagnosticOptions::Vi: OS << " +" << LineNo; break;
814 }
815
816 if (DiagOpts->ShowColumn)
817 // Compute the column number.
818 if (unsigned ColNo = PLoc.getColumn()) {
819 if (DiagOpts->getFormat() == DiagnosticOptions::MSVC) {
820 OS << ',';
821 // Visual Studio 2010 or earlier expects column number to be off by one
822 if (LangOpts.MSCompatibilityVersion &&
823 !LangOpts.isCompatibleWithMSVC(MajorVersion: LangOptions::MSVC2012))
824 ColNo--;
825 } else
826 OS << ':';
827 OS << ColNo;
828 }
829 switch (DiagOpts->getFormat()) {
830 case DiagnosticOptions::SARIF:
831 case DiagnosticOptions::Clang:
832 case DiagnosticOptions::Vi: OS << ':'; break;
833 case DiagnosticOptions::MSVC:
834 // MSVC2013 and before print 'file(4) : error'. MSVC2015 gets rid of the
835 // space and prints 'file(4): error'.
836 OS << ')';
837 if (LangOpts.MSCompatibilityVersion &&
838 !LangOpts.isCompatibleWithMSVC(MajorVersion: LangOptions::MSVC2015))
839 OS << ' ';
840 OS << ':';
841 break;
842 }
843
844 if (DiagOpts->ShowSourceRanges && !Ranges.empty()) {
845 FileID CaretFileID = Loc.getExpansionLoc().getFileID();
846 bool PrintedRange = false;
847 const SourceManager &SM = Loc.getManager();
848
849 for (const auto &R : Ranges) {
850 // Ignore invalid ranges.
851 if (!R.isValid())
852 continue;
853
854 SourceLocation B = SM.getExpansionLoc(Loc: R.getBegin());
855 CharSourceRange ERange = SM.getExpansionRange(Loc: R.getEnd());
856 SourceLocation E = ERange.getEnd();
857
858 // If the start or end of the range is in another file, just
859 // discard it.
860 if (SM.getFileID(SpellingLoc: B) != CaretFileID || SM.getFileID(SpellingLoc: E) != CaretFileID)
861 continue;
862
863 // Add in the length of the token, so that we cover multi-char
864 // tokens.
865 unsigned TokSize = 0;
866 if (ERange.isTokenRange())
867 TokSize = Lexer::MeasureTokenLength(Loc: E, SM, LangOpts);
868
869 FullSourceLoc BF(B, SM), EF(E, SM);
870 OS << '{'
871 << BF.getLineNumber() << ':' << BF.getColumnNumber() << '-'
872 << EF.getLineNumber() << ':' << (EF.getColumnNumber() + TokSize)
873 << '}';
874 PrintedRange = true;
875 }
876
877 if (PrintedRange)
878 OS << ':';
879 }
880 OS << ' ';
881}
882
883void TextDiagnostic::emitIncludeLocation(FullSourceLoc Loc, PresumedLoc PLoc) {
884 if (DiagOpts->ShowLocation && PLoc.isValid()) {
885 OS << "In file included from ";
886 emitFilename(Filename: PLoc.getFilename(), SM: Loc.getManager());
887 OS << ':' << PLoc.getLine() << ":\n";
888 } else
889 OS << "In included file:\n";
890}
891
892void TextDiagnostic::emitImportLocation(FullSourceLoc Loc, PresumedLoc PLoc,
893 StringRef ModuleName) {
894 if (DiagOpts->ShowLocation && PLoc.isValid())
895 OS << "In module '" << ModuleName << "' imported from "
896 << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
897 else
898 OS << "In module '" << ModuleName << "':\n";
899}
900
901void TextDiagnostic::emitBuildingModuleLocation(FullSourceLoc Loc,
902 PresumedLoc PLoc,
903 StringRef ModuleName) {
904 if (DiagOpts->ShowLocation && PLoc.isValid())
905 OS << "While building module '" << ModuleName << "' imported from "
906 << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
907 else
908 OS << "While building module '" << ModuleName << "':\n";
909}
910
911/// Find the suitable set of lines to show to include a set of ranges.
912static std::optional<std::pair<unsigned, unsigned>>
913findLinesForRange(const CharSourceRange &R, FileID FID,
914 const SourceManager &SM) {
915 if (!R.isValid())
916 return std::nullopt;
917
918 SourceLocation Begin = R.getBegin();
919 SourceLocation End = R.getEnd();
920 if (SM.getFileID(SpellingLoc: Begin) != FID || SM.getFileID(SpellingLoc: End) != FID)
921 return std::nullopt;
922
923 return std::make_pair(x: SM.getExpansionLineNumber(Loc: Begin),
924 y: SM.getExpansionLineNumber(Loc: End));
925}
926
927/// Add as much of range B into range A as possible without exceeding a maximum
928/// size of MaxRange. Ranges are inclusive.
929static std::pair<unsigned, unsigned>
930maybeAddRange(std::pair<unsigned, unsigned> A, std::pair<unsigned, unsigned> B,
931 unsigned MaxRange) {
932 // If A is already the maximum size, we're done.
933 unsigned Slack = MaxRange - (A.second - A.first + 1);
934 if (Slack == 0)
935 return A;
936
937 // Easy case: merge succeeds within MaxRange.
938 unsigned Min = std::min(a: A.first, b: B.first);
939 unsigned Max = std::max(a: A.second, b: B.second);
940 if (Max - Min + 1 <= MaxRange)
941 return {Min, Max};
942
943 // If we can't reach B from A within MaxRange, there's nothing to do.
944 // Don't add lines to the range that contain nothing interesting.
945 if ((B.first > A.first && B.first - A.first + 1 > MaxRange) ||
946 (B.second < A.second && A.second - B.second + 1 > MaxRange))
947 return A;
948
949 // Otherwise, expand A towards B to produce a range of size MaxRange. We
950 // attempt to expand by the same amount in both directions if B strictly
951 // contains A.
952
953 // Expand downwards by up to half the available amount, then upwards as
954 // much as possible, then downwards as much as possible.
955 A.second = std::min(a: A.second + (Slack + 1) / 2, b: Max);
956 Slack = MaxRange - (A.second - A.first + 1);
957 A.first = std::max(a: Min + Slack, b: A.first) - Slack;
958 A.second = std::min(a: A.first + MaxRange - 1, b: Max);
959 return A;
960}
961
962struct LineRange {
963 unsigned LineNo;
964 unsigned StartCol;
965 unsigned EndCol;
966};
967
968/// Highlight \p R (with ~'s) on the current source line.
969static void highlightRange(const LineRange &R, const SourceColumnMap &Map,
970 std::string &CaretLine) {
971 // Pick the first non-whitespace column.
972 unsigned StartColNo = R.StartCol;
973 while (StartColNo < Map.getSourceLine().size() &&
974 (Map.getSourceLine()[StartColNo] == ' ' ||
975 Map.getSourceLine()[StartColNo] == '\t'))
976 StartColNo = Map.startOfNextColumn(N: StartColNo);
977
978 // Pick the last non-whitespace column.
979 unsigned EndColNo =
980 std::min(a: static_cast<size_t>(R.EndCol), b: Map.getSourceLine().size());
981 while (EndColNo && (Map.getSourceLine()[EndColNo - 1] == ' ' ||
982 Map.getSourceLine()[EndColNo - 1] == '\t'))
983 EndColNo = Map.startOfPreviousColumn(N: EndColNo);
984
985 // If the start/end passed each other, then we are trying to highlight a
986 // range that just exists in whitespace. That most likely means we have
987 // a multi-line highlighting range that covers a blank line.
988 if (StartColNo > EndColNo)
989 return;
990
991 // Fill the range with ~'s.
992 StartColNo = Map.byteToContainingColumn(N: StartColNo);
993 EndColNo = Map.byteToContainingColumn(N: EndColNo);
994
995 assert(StartColNo <= EndColNo && "Invalid range!");
996 if (CaretLine.size() < EndColNo)
997 CaretLine.resize(n: EndColNo, c: ' ');
998 std::fill(CaretLine.begin() + StartColNo, CaretLine.begin() + EndColNo, '~');
999}
1000
1001static std::string buildFixItInsertionLine(FileID FID,
1002 unsigned LineNo,
1003 const SourceColumnMap &map,
1004 ArrayRef<FixItHint> Hints,
1005 const SourceManager &SM,
1006 const DiagnosticOptions *DiagOpts) {
1007 std::string FixItInsertionLine;
1008 if (Hints.empty() || !DiagOpts->ShowFixits)
1009 return FixItInsertionLine;
1010 unsigned PrevHintEndCol = 0;
1011
1012 for (const auto &H : Hints) {
1013 if (H.CodeToInsert.empty())
1014 continue;
1015
1016 // We have an insertion hint. Determine whether the inserted
1017 // code contains no newlines and is on the same line as the caret.
1018 std::pair<FileID, unsigned> HintLocInfo =
1019 SM.getDecomposedExpansionLoc(Loc: H.RemoveRange.getBegin());
1020 if (FID == HintLocInfo.first &&
1021 LineNo == SM.getLineNumber(FID: HintLocInfo.first, FilePos: HintLocInfo.second) &&
1022 StringRef(H.CodeToInsert).find_first_of(Chars: "\n\r") == StringRef::npos) {
1023 // Insert the new code into the line just below the code
1024 // that the user wrote.
1025 // Note: When modifying this function, be very careful about what is a
1026 // "column" (printed width, platform-dependent) and what is a
1027 // "byte offset" (SourceManager "column").
1028 unsigned HintByteOffset =
1029 SM.getColumnNumber(FID: HintLocInfo.first, FilePos: HintLocInfo.second) - 1;
1030
1031 // The hint must start inside the source or right at the end
1032 assert(HintByteOffset < static_cast<unsigned>(map.bytes()) + 1);
1033 unsigned HintCol = map.byteToContainingColumn(N: HintByteOffset);
1034
1035 // If we inserted a long previous hint, push this one forwards, and add
1036 // an extra space to show that this is not part of the previous
1037 // completion. This is sort of the best we can do when two hints appear
1038 // to overlap.
1039 //
1040 // Note that if this hint is located immediately after the previous
1041 // hint, no space will be added, since the location is more important.
1042 if (HintCol < PrevHintEndCol)
1043 HintCol = PrevHintEndCol + 1;
1044
1045 // This should NOT use HintByteOffset, because the source might have
1046 // Unicode characters in earlier columns.
1047 unsigned NewFixItLineSize = FixItInsertionLine.size() +
1048 (HintCol - PrevHintEndCol) +
1049 H.CodeToInsert.size();
1050 if (NewFixItLineSize > FixItInsertionLine.size())
1051 FixItInsertionLine.resize(n: NewFixItLineSize, c: ' ');
1052
1053 std::copy(H.CodeToInsert.begin(), H.CodeToInsert.end(),
1054 FixItInsertionLine.end() - H.CodeToInsert.size());
1055
1056 PrevHintEndCol = HintCol + llvm::sys::locale::columnWidth(s: H.CodeToInsert);
1057 }
1058 }
1059
1060 expandTabs(SourceLine&: FixItInsertionLine, TabStop: DiagOpts->TabStop);
1061
1062 return FixItInsertionLine;
1063}
1064
1065static unsigned getNumDisplayWidth(unsigned N) {
1066 unsigned L = 1u, M = 10u;
1067 while (M <= N && ++L != std::numeric_limits<unsigned>::digits10 + 1)
1068 M *= 10u;
1069
1070 return L;
1071}
1072
1073/// Filter out invalid ranges, ranges that don't fit into the window of
1074/// source lines we will print, and ranges from other files.
1075///
1076/// For the remaining ranges, convert them to simple LineRange structs,
1077/// which only cover one line at a time.
1078static SmallVector<LineRange>
1079prepareAndFilterRanges(const SmallVectorImpl<CharSourceRange> &Ranges,
1080 const SourceManager &SM,
1081 const std::pair<unsigned, unsigned> &Lines, FileID FID,
1082 const LangOptions &LangOpts) {
1083 SmallVector<LineRange> LineRanges;
1084
1085 for (const CharSourceRange &R : Ranges) {
1086 if (R.isInvalid())
1087 continue;
1088 SourceLocation Begin = R.getBegin();
1089 SourceLocation End = R.getEnd();
1090
1091 unsigned StartLineNo = SM.getExpansionLineNumber(Loc: Begin);
1092 if (StartLineNo > Lines.second || SM.getFileID(SpellingLoc: Begin) != FID)
1093 continue;
1094
1095 unsigned EndLineNo = SM.getExpansionLineNumber(Loc: End);
1096 if (EndLineNo < Lines.first || SM.getFileID(SpellingLoc: End) != FID)
1097 continue;
1098
1099 unsigned StartColumn = SM.getExpansionColumnNumber(Loc: Begin);
1100 unsigned EndColumn = SM.getExpansionColumnNumber(Loc: End);
1101 if (R.isTokenRange())
1102 EndColumn += Lexer::MeasureTokenLength(Loc: End, SM, LangOpts);
1103
1104 // Only a single line.
1105 if (StartLineNo == EndLineNo) {
1106 LineRanges.push_back(Elt: {.LineNo: StartLineNo, .StartCol: StartColumn - 1, .EndCol: EndColumn - 1});
1107 continue;
1108 }
1109
1110 // Start line.
1111 LineRanges.push_back(Elt: {.LineNo: StartLineNo, .StartCol: StartColumn - 1, .EndCol: ~0u});
1112
1113 // Middle lines.
1114 for (unsigned S = StartLineNo + 1; S != EndLineNo; ++S)
1115 LineRanges.push_back(Elt: {.LineNo: S, .StartCol: 0, .EndCol: ~0u});
1116
1117 // End line.
1118 LineRanges.push_back(Elt: {.LineNo: EndLineNo, .StartCol: 0, .EndCol: EndColumn - 1});
1119 }
1120
1121 return LineRanges;
1122}
1123
1124/// Creates syntax highlighting information in form of StyleRanges.
1125///
1126/// The returned unique ptr has always exactly size
1127/// (\p EndLineNumber - \p StartLineNumber + 1). Each SmallVector in there
1128/// corresponds to syntax highlighting information in one line. In each line,
1129/// the StyleRanges are non-overlapping and sorted from start to end of the
1130/// line.
1131static std::unique_ptr<llvm::SmallVector<TextDiagnostic::StyleRange>[]>
1132highlightLines(StringRef FileData, unsigned StartLineNumber,
1133 unsigned EndLineNumber, const Preprocessor *PP,
1134 const LangOptions &LangOpts, bool ShowColors, FileID FID,
1135 const SourceManager &SM) {
1136 assert(StartLineNumber <= EndLineNumber);
1137 auto SnippetRanges =
1138 std::make_unique<SmallVector<TextDiagnostic::StyleRange>[]>(
1139 num: EndLineNumber - StartLineNumber + 1);
1140
1141 if (!PP || !ShowColors)
1142 return SnippetRanges;
1143
1144 // Might cause emission of another diagnostic.
1145 if (PP->getIdentifierTable().getExternalIdentifierLookup())
1146 return SnippetRanges;
1147
1148 auto Buff = llvm::MemoryBuffer::getMemBuffer(InputData: FileData);
1149 Lexer L{FID, *Buff, SM, LangOpts};
1150 L.SetKeepWhitespaceMode(true);
1151
1152 const char *FirstLineStart =
1153 FileData.data() +
1154 SM.getDecomposedLoc(Loc: SM.translateLineCol(FID, Line: StartLineNumber, Col: 1)).second;
1155 if (const char *CheckPoint = PP->getCheckPoint(FID, Start: FirstLineStart)) {
1156 assert(CheckPoint >= Buff->getBufferStart() &&
1157 CheckPoint <= Buff->getBufferEnd());
1158 assert(CheckPoint <= FirstLineStart);
1159 size_t Offset = CheckPoint - Buff->getBufferStart();
1160 L.seek(Offset, /*IsAtStartOfLine=*/false);
1161 }
1162
1163 // Classify the given token and append it to the given vector.
1164 auto appendStyle =
1165 [PP, &LangOpts](SmallVector<TextDiagnostic::StyleRange> &Vec,
1166 const Token &T, unsigned Start, unsigned Length) -> void {
1167 if (T.is(K: tok::raw_identifier)) {
1168 StringRef RawIdent = T.getRawIdentifier();
1169 // Special case true/false/nullptr/... literals, since they will otherwise
1170 // be treated as keywords.
1171 // FIXME: It would be good to have a programmatic way of getting this
1172 // list.
1173 if (llvm::StringSwitch<bool>(RawIdent)
1174 .Case(S: "true", Value: true)
1175 .Case(S: "false", Value: true)
1176 .Case(S: "nullptr", Value: true)
1177 .Case(S: "__func__", Value: true)
1178 .Case(S: "__objc_yes__", Value: true)
1179 .Case(S: "__objc_no__", Value: true)
1180 .Case(S: "__null", Value: true)
1181 .Case(S: "__FUNCDNAME__", Value: true)
1182 .Case(S: "__FUNCSIG__", Value: true)
1183 .Case(S: "__FUNCTION__", Value: true)
1184 .Case(S: "__FUNCSIG__", Value: true)
1185 .Default(Value: false)) {
1186 Vec.emplace_back(Args&: Start, Args: Start + Length, Args: LiteralColor);
1187 } else {
1188 const IdentifierInfo *II = PP->getIdentifierInfo(Name: RawIdent);
1189 assert(II);
1190 if (II->isKeyword(LangOpts))
1191 Vec.emplace_back(Args&: Start, Args: Start + Length, Args: KeywordColor);
1192 }
1193 } else if (tok::isLiteral(K: T.getKind())) {
1194 Vec.emplace_back(Args&: Start, Args: Start + Length, Args: LiteralColor);
1195 } else {
1196 assert(T.is(tok::comment));
1197 Vec.emplace_back(Args&: Start, Args: Start + Length, Args: CommentColor);
1198 }
1199 };
1200
1201 bool Stop = false;
1202 while (!Stop) {
1203 Token T;
1204 Stop = L.LexFromRawLexer(Result&: T);
1205 if (T.is(K: tok::unknown))
1206 continue;
1207
1208 // We are only interested in identifiers, literals and comments.
1209 if (!T.is(K: tok::raw_identifier) && !T.is(K: tok::comment) &&
1210 !tok::isLiteral(K: T.getKind()))
1211 continue;
1212
1213 bool Invalid = false;
1214 unsigned TokenEndLine = SM.getSpellingLineNumber(Loc: T.getEndLoc(), Invalid: &Invalid);
1215 if (Invalid || TokenEndLine < StartLineNumber)
1216 continue;
1217
1218 assert(TokenEndLine >= StartLineNumber);
1219
1220 unsigned TokenStartLine =
1221 SM.getSpellingLineNumber(Loc: T.getLocation(), Invalid: &Invalid);
1222 if (Invalid)
1223 continue;
1224 // If this happens, we're done.
1225 if (TokenStartLine > EndLineNumber)
1226 break;
1227
1228 unsigned StartCol =
1229 SM.getSpellingColumnNumber(Loc: T.getLocation(), Invalid: &Invalid) - 1;
1230 if (Invalid)
1231 continue;
1232
1233 // Simple tokens.
1234 if (TokenStartLine == TokenEndLine) {
1235 SmallVector<TextDiagnostic::StyleRange> &LineRanges =
1236 SnippetRanges[TokenStartLine - StartLineNumber];
1237 appendStyle(LineRanges, T, StartCol, T.getLength());
1238 continue;
1239 }
1240 assert((TokenEndLine - TokenStartLine) >= 1);
1241
1242 // For tokens that span multiple lines (think multiline comments), we
1243 // divide them into multiple StyleRanges.
1244 unsigned EndCol = SM.getSpellingColumnNumber(Loc: T.getEndLoc(), Invalid: &Invalid) - 1;
1245 if (Invalid)
1246 continue;
1247
1248 std::string Spelling = Lexer::getSpelling(Tok: T, SourceMgr: SM, LangOpts);
1249
1250 unsigned L = TokenStartLine;
1251 unsigned LineLength = 0;
1252 for (unsigned I = 0; I <= Spelling.size(); ++I) {
1253 // This line is done.
1254 if (I == Spelling.size() || isVerticalWhitespace(c: Spelling[I])) {
1255 SmallVector<TextDiagnostic::StyleRange> &LineRanges =
1256 SnippetRanges[L - StartLineNumber];
1257
1258 if (L >= StartLineNumber) {
1259 if (L == TokenStartLine) // First line
1260 appendStyle(LineRanges, T, StartCol, LineLength);
1261 else if (L == TokenEndLine) // Last line
1262 appendStyle(LineRanges, T, 0, EndCol);
1263 else
1264 appendStyle(LineRanges, T, 0, LineLength);
1265 }
1266
1267 ++L;
1268 if (L > EndLineNumber)
1269 break;
1270 LineLength = 0;
1271 continue;
1272 }
1273 ++LineLength;
1274 }
1275 }
1276
1277 return SnippetRanges;
1278}
1279
1280/// Emit a code snippet and caret line.
1281///
1282/// This routine emits a single line's code snippet and caret line..
1283///
1284/// \param Loc The location for the caret.
1285/// \param Ranges The underlined ranges for this code snippet.
1286/// \param Hints The FixIt hints active for this diagnostic.
1287void TextDiagnostic::emitSnippetAndCaret(
1288 FullSourceLoc Loc, DiagnosticsEngine::Level Level,
1289 SmallVectorImpl<CharSourceRange> &Ranges, ArrayRef<FixItHint> Hints) {
1290 assert(Loc.isValid() && "must have a valid source location here");
1291 assert(Loc.isFileID() && "must have a file location here");
1292
1293 // If caret diagnostics are enabled and we have location, we want to
1294 // emit the caret. However, we only do this if the location moved
1295 // from the last diagnostic, if the last diagnostic was a note that
1296 // was part of a different warning or error diagnostic, or if the
1297 // diagnostic has ranges. We don't want to emit the same caret
1298 // multiple times if one loc has multiple diagnostics.
1299 if (!DiagOpts->ShowCarets)
1300 return;
1301 if (Loc == LastLoc && Ranges.empty() && Hints.empty() &&
1302 (LastLevel != DiagnosticsEngine::Note || Level == LastLevel))
1303 return;
1304
1305 FileID FID = Loc.getFileID();
1306 const SourceManager &SM = Loc.getManager();
1307
1308 // Get information about the buffer it points into.
1309 bool Invalid = false;
1310 StringRef BufData = Loc.getBufferData(Invalid: &Invalid);
1311 if (Invalid)
1312 return;
1313 const char *BufStart = BufData.data();
1314 const char *BufEnd = BufStart + BufData.size();
1315
1316 unsigned CaretLineNo = Loc.getLineNumber();
1317 unsigned CaretColNo = Loc.getColumnNumber();
1318
1319 // Arbitrarily stop showing snippets when the line is too long.
1320 static const size_t MaxLineLengthToPrint = 4096;
1321 if (CaretColNo > MaxLineLengthToPrint)
1322 return;
1323
1324 // Find the set of lines to include.
1325 const unsigned MaxLines = DiagOpts->SnippetLineLimit;
1326 std::pair<unsigned, unsigned> Lines = {CaretLineNo, CaretLineNo};
1327 unsigned DisplayLineNo = Loc.getPresumedLoc().getLine();
1328 for (const auto &I : Ranges) {
1329 if (auto OptionalRange = findLinesForRange(R: I, FID, SM))
1330 Lines = maybeAddRange(A: Lines, B: *OptionalRange, MaxRange: MaxLines);
1331
1332 DisplayLineNo =
1333 std::min(a: DisplayLineNo, b: SM.getPresumedLineNumber(Loc: I.getBegin()));
1334 }
1335
1336 // Our line numbers look like:
1337 // " [number] | "
1338 // Where [number] is MaxLineNoDisplayWidth columns
1339 // and the full thing is therefore MaxLineNoDisplayWidth + 4 columns.
1340 unsigned MaxLineNoDisplayWidth =
1341 DiagOpts->ShowLineNumbers
1342 ? std::max(a: 4u, b: getNumDisplayWidth(N: DisplayLineNo + MaxLines))
1343 : 0;
1344 auto indentForLineNumbers = [&] {
1345 if (MaxLineNoDisplayWidth > 0)
1346 OS.indent(NumSpaces: MaxLineNoDisplayWidth + 2) << "| ";
1347 };
1348
1349 // Prepare source highlighting information for the lines we're about to
1350 // emit, starting from the first line.
1351 std::unique_ptr<SmallVector<StyleRange>[]> SourceStyles =
1352 highlightLines(FileData: BufData, StartLineNumber: Lines.first, EndLineNumber: Lines.second, PP, LangOpts,
1353 ShowColors: DiagOpts->ShowColors, FID, SM);
1354
1355 SmallVector<LineRange> LineRanges =
1356 prepareAndFilterRanges(Ranges, SM, Lines, FID, LangOpts);
1357
1358 for (unsigned LineNo = Lines.first; LineNo != Lines.second + 1;
1359 ++LineNo, ++DisplayLineNo) {
1360 // Rewind from the current position to the start of the line.
1361 const char *LineStart =
1362 BufStart +
1363 SM.getDecomposedLoc(Loc: SM.translateLineCol(FID, Line: LineNo, Col: 1)).second;
1364 if (LineStart == BufEnd)
1365 break;
1366
1367 // Compute the line end.
1368 const char *LineEnd = LineStart;
1369 while (*LineEnd != '\n' && *LineEnd != '\r' && LineEnd != BufEnd)
1370 ++LineEnd;
1371
1372 // Arbitrarily stop showing snippets when the line is too long.
1373 // FIXME: Don't print any lines in this case.
1374 if (size_t(LineEnd - LineStart) > MaxLineLengthToPrint)
1375 return;
1376
1377 // Copy the line of code into an std::string for ease of manipulation.
1378 std::string SourceLine(LineStart, LineEnd);
1379 // Remove trailing null bytes.
1380 while (!SourceLine.empty() && SourceLine.back() == '\0' &&
1381 (LineNo != CaretLineNo || SourceLine.size() > CaretColNo))
1382 SourceLine.pop_back();
1383
1384 // Build the byte to column map.
1385 const SourceColumnMap sourceColMap(SourceLine, DiagOpts->TabStop);
1386
1387 std::string CaretLine;
1388 // Highlight all of the characters covered by Ranges with ~ characters.
1389 for (const auto &LR : LineRanges) {
1390 if (LR.LineNo == LineNo)
1391 highlightRange(R: LR, Map: sourceColMap, CaretLine);
1392 }
1393
1394 // Next, insert the caret itself.
1395 if (CaretLineNo == LineNo) {
1396 size_t Col = sourceColMap.byteToContainingColumn(N: CaretColNo - 1);
1397 CaretLine.resize(n: std::max(a: Col + 1, b: CaretLine.size()), c: ' ');
1398 CaretLine[Col] = '^';
1399 }
1400
1401 std::string FixItInsertionLine = buildFixItInsertionLine(
1402 FID, LineNo, map: sourceColMap, Hints, SM, DiagOpts: DiagOpts.get());
1403
1404 // If the source line is too long for our terminal, select only the
1405 // "interesting" source region within that line.
1406 unsigned Columns = DiagOpts->MessageLength;
1407 if (Columns)
1408 selectInterestingSourceRegion(SourceLine, CaretLine, FixItInsertionLine,
1409 Columns, map: sourceColMap);
1410
1411 // If we are in -fdiagnostics-print-source-range-info mode, we are trying
1412 // to produce easily machine parsable output. Add a space before the
1413 // source line and the caret to make it trivial to tell the main diagnostic
1414 // line from what the user is intended to see.
1415 if (DiagOpts->ShowSourceRanges && !SourceLine.empty()) {
1416 SourceLine = ' ' + SourceLine;
1417 CaretLine = ' ' + CaretLine;
1418 }
1419
1420 // Emit what we have computed.
1421 emitSnippet(SourceLine, MaxLineNoDisplayWidth, LineNo, DisplayLineNo,
1422 Styles: SourceStyles[LineNo - Lines.first]);
1423
1424 if (!CaretLine.empty()) {
1425 indentForLineNumbers();
1426 if (DiagOpts->ShowColors)
1427 OS.changeColor(Color: caretColor, Bold: true);
1428 OS << CaretLine << '\n';
1429 if (DiagOpts->ShowColors)
1430 OS.resetColor();
1431 }
1432
1433 if (!FixItInsertionLine.empty()) {
1434 indentForLineNumbers();
1435 if (DiagOpts->ShowColors)
1436 // Print fixit line in color
1437 OS.changeColor(Color: fixitColor, Bold: false);
1438 if (DiagOpts->ShowSourceRanges)
1439 OS << ' ';
1440 OS << FixItInsertionLine << '\n';
1441 if (DiagOpts->ShowColors)
1442 OS.resetColor();
1443 }
1444 }
1445
1446 // Print out any parseable fixit information requested by the options.
1447 emitParseableFixits(Hints, SM);
1448}
1449
1450void TextDiagnostic::emitSnippet(StringRef SourceLine,
1451 unsigned MaxLineNoDisplayWidth,
1452 unsigned LineNo, unsigned DisplayLineNo,
1453 ArrayRef<StyleRange> Styles) {
1454 // Emit line number.
1455 if (MaxLineNoDisplayWidth > 0) {
1456 unsigned LineNoDisplayWidth = getNumDisplayWidth(N: DisplayLineNo);
1457 OS.indent(NumSpaces: MaxLineNoDisplayWidth - LineNoDisplayWidth + 1)
1458 << DisplayLineNo << " | ";
1459 }
1460
1461 // Print the source line one character at a time.
1462 bool PrintReversed = false;
1463 std::optional<llvm::raw_ostream::Colors> CurrentColor;
1464 size_t I = 0;
1465 while (I < SourceLine.size()) {
1466 auto [Str, WasPrintable] =
1467 printableTextForNextCharacter(SourceLine, I: &I, TabStop: DiagOpts->TabStop);
1468
1469 // Toggle inverted colors on or off for this character.
1470 if (DiagOpts->ShowColors) {
1471 if (WasPrintable == PrintReversed) {
1472 PrintReversed = !PrintReversed;
1473 if (PrintReversed)
1474 OS.reverseColor();
1475 else {
1476 OS.resetColor();
1477 CurrentColor = std::nullopt;
1478 }
1479 }
1480
1481 // Apply syntax highlighting information.
1482 const auto *CharStyle = llvm::find_if(Range&: Styles, P: [I](const StyleRange &R) {
1483 return (R.Start < I && R.End >= I);
1484 });
1485
1486 if (CharStyle != Styles.end()) {
1487 if (!CurrentColor ||
1488 (CurrentColor && *CurrentColor != CharStyle->Color)) {
1489 OS.changeColor(Color: CharStyle->Color, Bold: false);
1490 CurrentColor = CharStyle->Color;
1491 }
1492 } else if (CurrentColor) {
1493 OS.resetColor();
1494 CurrentColor = std::nullopt;
1495 }
1496 }
1497
1498 OS << Str;
1499 }
1500
1501 if (DiagOpts->ShowColors)
1502 OS.resetColor();
1503
1504 OS << '\n';
1505}
1506
1507void TextDiagnostic::emitParseableFixits(ArrayRef<FixItHint> Hints,
1508 const SourceManager &SM) {
1509 if (!DiagOpts->ShowParseableFixits)
1510 return;
1511
1512 // We follow FixItRewriter's example in not (yet) handling
1513 // fix-its in macros.
1514 for (const auto &H : Hints) {
1515 if (H.RemoveRange.isInvalid() || H.RemoveRange.getBegin().isMacroID() ||
1516 H.RemoveRange.getEnd().isMacroID())
1517 return;
1518 }
1519
1520 for (const auto &H : Hints) {
1521 SourceLocation BLoc = H.RemoveRange.getBegin();
1522 SourceLocation ELoc = H.RemoveRange.getEnd();
1523
1524 std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(Loc: BLoc);
1525 std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(Loc: ELoc);
1526
1527 // Adjust for token ranges.
1528 if (H.RemoveRange.isTokenRange())
1529 EInfo.second += Lexer::MeasureTokenLength(Loc: ELoc, SM, LangOpts);
1530
1531 // We specifically do not do word-wrapping or tab-expansion here,
1532 // because this is supposed to be easy to parse.
1533 PresumedLoc PLoc = SM.getPresumedLoc(Loc: BLoc);
1534 if (PLoc.isInvalid())
1535 break;
1536
1537 OS << "fix-it:\"";
1538 OS.write_escaped(Str: PLoc.getFilename());
1539 OS << "\":{" << SM.getLineNumber(FID: BInfo.first, FilePos: BInfo.second)
1540 << ':' << SM.getColumnNumber(FID: BInfo.first, FilePos: BInfo.second)
1541 << '-' << SM.getLineNumber(FID: EInfo.first, FilePos: EInfo.second)
1542 << ':' << SM.getColumnNumber(FID: EInfo.first, FilePos: EInfo.second)
1543 << "}:\"";
1544 OS.write_escaped(Str: H.CodeToInsert);
1545 OS << "\"\n";
1546 }
1547}
1548