Coverage Report

Created: 2020-09-19 12:23

/Users/buildslave/jenkins/workspace/coverage/llvm-project/clang/lib/Frontend/TextDiagnostic.cpp
Line
Count
Source (jump to first uncovered line)
1
//===--- TextDiagnostic.cpp - Text Diagnostic Pretty-Printing -------------===//
2
//
3
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4
// See https://llvm.org/LICENSE.txt for license information.
5
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6
//
7
//===----------------------------------------------------------------------===//
8
9
#include "clang/Frontend/TextDiagnostic.h"
10
#include "clang/Basic/CharInfo.h"
11
#include "clang/Basic/DiagnosticOptions.h"
12
#include "clang/Basic/FileManager.h"
13
#include "clang/Basic/SourceManager.h"
14
#include "clang/Lex/Lexer.h"
15
#include "llvm/ADT/SmallString.h"
16
#include "llvm/ADT/StringExtras.h"
17
#include "llvm/Support/ConvertUTF.h"
18
#include "llvm/Support/ErrorHandling.h"
19
#include "llvm/Support/Locale.h"
20
#include "llvm/Support/Path.h"
21
#include "llvm/Support/raw_ostream.h"
22
#include <algorithm>
23
24
using namespace clang;
25
26
static const enum raw_ostream::Colors noteColor =
27
  raw_ostream::BLACK;
28
static const enum raw_ostream::Colors remarkColor =
29
  raw_ostream::BLUE;
30
static const enum raw_ostream::Colors fixitColor =
31
  raw_ostream::GREEN;
32
static const enum raw_ostream::Colors caretColor =
33
  raw_ostream::GREEN;
34
static const enum raw_ostream::Colors warningColor =
35
  raw_ostream::MAGENTA;
36
static const enum raw_ostream::Colors templateColor =
37
  raw_ostream::CYAN;
38
static const enum raw_ostream::Colors errorColor = raw_ostream::RED;
39
static const enum raw_ostream::Colors fatalColor = raw_ostream::RED;
40
// Used for changing only the bold attribute.
41
static const enum raw_ostream::Colors savedColor =
42
  raw_ostream::SAVEDCOLOR;
43
44
/// Add highlights to differences in template strings.
45
static void applyTemplateHighlighting(raw_ostream &OS, StringRef Str,
46
35.1k
                                      bool &Normal, bool Bold) {
47
35.2k
  while (1) {
48
35.2k
    size_t Pos = Str.find(ToggleHighlight);
49
35.2k
    OS << Str.slice(0, Pos);
50
35.2k
    if (Pos == StringRef::npos)
51
35.1k
      break;
52
110
53
110
    Str = Str.substr(Pos + 1);
54
110
    if (Normal)
55
55
      OS.changeColor(templateColor, true);
56
55
    else {
57
55
      OS.resetColor();
58
55
      if (Bold)
59
36
        OS.changeColor(savedColor, true);
60
55
    }
61
110
    Normal = !Normal;
62
110
  }
63
35.1k
}
64
65
/// Number of spaces to indent when word-wrapping.
66
const unsigned WordWrapIndentation = 6;
67
68
1.01k
static int bytesSincePreviousTabOrLineBegin(StringRef SourceLine, size_t i) {
69
1.01k
  int bytes = 0;
70
2.24k
  while (0<i) {
71
1.35k
    if (SourceLine[--i]=='\t')
72
120
      break;
73
1.23k
    ++bytes;
74
1.23k
  }
75
1.01k
  return bytes;
76
1.01k
}
77
78
/// returns a printable representation of first item from input range
79
///
80
/// This function returns a printable representation of the next item in a line
81
///  of source. If the next byte begins a valid and printable character, that
82
///  character is returned along with 'true'.
83
///
84
/// Otherwise, if the next byte begins a valid, but unprintable character, a
85
///  printable, escaped representation of the character is returned, along with
86
///  'false'. Otherwise a printable, escaped representation of the next byte
87
///  is returned along with 'false'.
88
///
89
/// \note The index is updated to be used with a subsequent call to
90
///        printableTextForNextCharacter.
91
///
92
/// \param SourceLine The line of source
93
/// \param i Pointer to byte index,
94
/// \param TabStop used to expand tabs
95
/// \return pair(printable text, 'true' iff original text was printable)
96
///
97
static std::pair<SmallString<16>, bool>
98
printableTextForNextCharacter(StringRef SourceLine, size_t *i,
99
3.33M
                              unsigned TabStop) {
100
3.33M
  assert(i && "i must not be null");
101
3.33M
  assert(*i<SourceLine.size() && "must point to a valid index");
102
3.33M
103
3.33M
  if (SourceLine[*i]=='\t') {
104
1.01k
    assert(0 < TabStop && TabStop <= DiagnosticOptions::MaxTabStop &&
105
1.01k
           "Invalid -ftabstop value");
106
1.01k
    unsigned col = bytesSincePreviousTabOrLineBegin(SourceLine, *i);
107
1.01k
    unsigned NumSpaces = TabStop - col%TabStop;
108
1.01k
    assert(0 < NumSpaces && NumSpaces <= TabStop
109
1.01k
           && "Invalid computation of space amt");
110
1.01k
    ++(*i);
111
1.01k
112
1.01k
    SmallString<16> expandedTab;
113
1.01k
    expandedTab.assign(NumSpaces, ' ');
114
1.01k
    return std::make_pair(expandedTab, true);
115
1.01k
  }
116
3.33M
117
3.33M
  unsigned char const *begin, *end;
118
3.33M
  begin = reinterpret_cast<unsigned char const *>(&*(SourceLine.begin() + *i));
119
3.33M
  end = begin + (SourceLine.size() - *i);
120
3.33M
121
3.33M
  if (llvm::isLegalUTF8Sequence(begin, end)) {
122
3.33M
    llvm::UTF32 c;
123
3.33M
    llvm::UTF32 *cptr = &c;
124
3.33M
    unsigned char const *original_begin = begin;
125
3.33M
    unsigned char const *cp_end =
126
3.33M
        begin + llvm::getNumBytesForUTF8(SourceLine[*i]);
127
3.33M
128
3.33M
    llvm::ConversionResult res = llvm::ConvertUTF8toUTF32(
129
3.33M
        &begin, cp_end, &cptr, cptr + 1, llvm::strictConversion);
130
3.33M
    (void)res;
131
3.33M
    assert(llvm::conversionOK == res);
132
3.33M
    assert(0 < begin-original_begin
133
3.33M
           && "we must be further along in the string now");
134
3.33M
    *i += begin-original_begin;
135
3.33M
136
3.33M
    if (!llvm::sys::locale::isPrint(c)) {
137
      // If next character is valid UTF-8, but not printable
138
509
      SmallString<16> expandedCP("<U+>");
139
878
      while (c) {
140
369
        expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(c%16));
141
369
        c/=16;
142
369
      }
143
2.17k
      while (expandedCP.size() < 8)
144
1.66k
        expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(0));
145
509
      return std::make_pair(expandedCP, false);
146
509
    }
147
3.33M
148
    // If next character is valid UTF-8, and printable
149
3.33M
    return std::make_pair(SmallString<16>(original_begin, cp_end), true);
150
3.33M
151
3.33M
  }
152
486
153
  // If next byte is not valid UTF-8 (and therefore not printable)
154
486
  SmallString<16> expandedByte("<XX>");
155
486
  unsigned char byte = SourceLine[*i];
156
486
  expandedByte[1] = llvm::hexdigit(byte / 16);
157
486
  expandedByte[2] = llvm::hexdigit(byte % 16);
158
486
  ++(*i);
159
486
  return std::make_pair(expandedByte, false);
160
486
}
161
162
2.92k
static void expandTabs(std::string &SourceLine, unsigned TabStop) {
163
2.92k
  size_t i = SourceLine.size();
164
53.4k
  while (i>0) {
165
50.5k
    i--;
166
50.5k
    if (SourceLine[i]!='\t')
167
50.5k
      continue;
168
0
    size_t tmp_i = i;
169
0
    std::pair<SmallString<16>,bool> res
170
0
      = printableTextForNextCharacter(SourceLine, &tmp_i, TabStop);
171
0
    SourceLine.replace(i, 1, res.first.c_str());
172
0
  }
173
2.92k
}
174
175
/// This function takes a raw source line and produces a mapping from the bytes
176
///  of the printable representation of the line to the columns those printable
177
///  characters will appear at (numbering the first column as 0).
178
///
179
/// If a byte 'i' corresponds to multiple columns (e.g. the byte contains a tab
180
///  character) then the array will map that byte to the first column the
181
///  tab appears at and the next value in the map will have been incremented
182
///  more than once.
183
///
184
/// If a byte is the first in a sequence of bytes that together map to a single
185
///  entity in the output, then the array will map that byte to the appropriate
186
///  column while the subsequent bytes will be -1.
187
///
188
/// The last element in the array does not correspond to any byte in the input
189
///  and instead is the number of columns needed to display the source
190
///
191
/// example: (given a tabstop of 8)
192
///
193
///    "a \t \u3042" -> {0,1,2,8,9,-1,-1,11}
194
///
195
///  (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
196
///   display)
197
static void byteToColumn(StringRef SourceLine, unsigned TabStop,
198
23.6k
                         SmallVectorImpl<int> &out) {
199
23.6k
  out.clear();
200
23.6k
201
23.6k
  if (SourceLine.empty()) {
202
1
    out.resize(1u,0);
203
1
    return;
204
1
  }
205
23.6k
206
23.6k
  out.resize(SourceLine.size()+1, -1);
207
23.6k
208
23.6k
  int columns = 0;
209
23.6k
  size_t i = 0;
210
1.13M
  while (i<SourceLine.size()) {
211
1.11M
    out[i] = columns;
212
1.11M
    std::pair<SmallString<16>,bool> res
213
1.11M
      = printableTextForNextCharacter(SourceLine, &i, TabStop);
214
1.11M
    columns += llvm::sys::locale::columnWidth(res.first);
215
1.11M
  }
216
23.6k
  out.back() = columns;
217
23.6k
}
218
219
/// This function takes a raw source line and produces a mapping from columns
220
///  to the byte of the source line that produced the character displaying at
221
///  that column. This is the inverse of the mapping produced by byteToColumn()
222
///
223
/// The last element in the array is the number of bytes in the source string
224
///
225
/// example: (given a tabstop of 8)
226
///
227
///    "a \t \u3042" -> {0,1,2,-1,-1,-1,-1,-1,3,4,-1,7}
228
///
229
///  (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
230
///   display)
231
static void columnToByte(StringRef SourceLine, unsigned TabStop,
232
23.6k
                         SmallVectorImpl<int> &out) {
233
23.6k
  out.clear();
234
23.6k
235
23.6k
  if (SourceLine.empty()) {
236
1
    out.resize(1u, 0);
237
1
    return;
238
1
  }
239
23.6k
240
23.6k
  int columns = 0;
241
23.6k
  size_t i = 0;
242
1.13M
  while (i<SourceLine.size()) {
243
1.11M
    out.resize(columns+1, -1);
244
1.11M
    out.back() = i;
245
1.11M
    std::pair<SmallString<16>,bool> res
246
1.11M
      = printableTextForNextCharacter(SourceLine, &i, TabStop);
247
1.11M
    columns += llvm::sys::locale::columnWidth(res.first);
248
1.11M
  }
249
23.6k
  out.resize(columns+1, -1);
250
23.6k
  out.back() = i;
251
23.6k
}
252
253
namespace {
254
struct SourceColumnMap {
255
  SourceColumnMap(StringRef SourceLine, unsigned TabStop)
256
23.6k
  : m_SourceLine(SourceLine) {
257
23.6k
258
23.6k
    ::byteToColumn(SourceLine, TabStop, m_byteToColumn);
259
23.6k
    ::columnToByte(SourceLine, TabStop, m_columnToByte);
260
23.6k
261
23.6k
    assert(m_byteToColumn.size()==SourceLine.size()+1);
262
23.6k
    assert(0 < m_byteToColumn.size() && 0 < m_columnToByte.size());
263
23.6k
    assert(m_byteToColumn.size()
264
23.6k
           == static_cast<unsigned>(m_columnToByte.back()+1));
265
23.6k
    assert(static_cast<unsigned>(m_byteToColumn.back()+1)
266
23.6k
           == m_columnToByte.size());
267
23.6k
  }
268
23.8k
  int columns() const { return m_byteToColumn.back(); }
269
2.89k
  int bytes() const { return m_columnToByte.back(); }
270
271
  /// Map a byte to the column which it is at the start of, or return -1
272
  /// if it is not at the start of a column (for a UTF-8 trailing byte).
273
2.18k
  int byteToColumn(int n) const {
274
2.18k
    assert(0<=n && n<static_cast<int>(m_byteToColumn.size()));
275
2.18k
    return m_byteToColumn[n];
276
2.18k
  }
277
278
  /// Map a byte to the first column which contains it.
279
52.6k
  int byteToContainingColumn(int N) const {
280
52.6k
    assert(0 <= N && N < static_cast<int>(m_byteToColumn.size()));
281
52.6k
    while (m_byteToColumn[N] == -1)
282
0
      --N;
283
52.6k
    return m_byteToColumn[N];
284
52.6k
  }
285
286
  /// Map a column to the byte which starts the column, or return -1 if
287
  /// the column the second or subsequent column of an expanded tab or similar
288
  /// multi-column entity.
289
132
  int columnToByte(int n) const {
290
132
    assert(0<=n && n<static_cast<int>(m_columnToByte.size()));
291
132
    return m_columnToByte[n];
292
132
  }
293
294
  /// Map from a byte index to the next byte which starts a column.
295
179
  int startOfNextColumn(int N) const {
296
179
    assert(0 <= N && N < static_cast<int>(m_byteToColumn.size() - 1));
297
211
    while (byteToColumn(++N) == -1) 
{}32
298
179
    return N;
299
179
  }
300
301
  /// Map from a byte index to the previous byte which starts a column.
302
1.15k
  int startOfPreviousColumn(int N) const {
303
1.15k
    assert(0 < N && N < static_cast<int>(m_byteToColumn.size()));
304
1.21k
    while (byteToColumn(--N) == -1) 
{}66
305
1.15k
    return N;
306
1.15k
  }
307
308
93.7k
  StringRef getSourceLine() const {
309
93.7k
    return m_SourceLine;
310
93.7k
  }
311
312
private:
313
  const std::string m_SourceLine;
314
  SmallVector<int,200> m_byteToColumn;
315
  SmallVector<int,200> m_columnToByte;
316
};
317
} // end anonymous namespace
318
319
/// When the source code line we want to print is too long for
320
/// the terminal, select the "interesting" region.
321
static void selectInterestingSourceRegion(std::string &SourceLine,
322
                                          std::string &CaretLine,
323
                                          std::string &FixItInsertionLine,
324
                                          unsigned Columns,
325
42
                                          const SourceColumnMap &map) {
326
42
  unsigned CaretColumns = CaretLine.size();
327
42
  unsigned FixItColumns = llvm::sys::locale::columnWidth(FixItInsertionLine);
328
42
  unsigned MaxColumns = std::max(static_cast<unsigned>(map.columns()),
329
42
                                 std::max(CaretColumns, FixItColumns));
330
  // if the number of columns is less than the desired number we're done
331
42
  if (MaxColumns <= Columns)
332
13
    return;
333
29
334
  // No special characters are allowed in CaretLine.
335
29
  assert(CaretLine.end() ==
336
29
         llvm::find_if(CaretLine, [](char c) { return c < ' ' || '~' < c; }));
337
29
338
  // Find the slice that we need to display the full caret line
339
  // correctly.
340
29
  unsigned CaretStart = 0, CaretEnd = CaretLine.size();
341
1.55k
  for (; CaretStart != CaretEnd; 
++CaretStart1.52k
)
342
1.55k
    if (!isWhitespace(CaretLine[CaretStart]))
343
29
      break;
344
29
345
1.12k
  for (; CaretEnd != CaretStart; 
--CaretEnd1.09k
)
346
1.12k
    if (!isWhitespace(CaretLine[CaretEnd - 1]))
347
29
      break;
348
29
349
  // caret has already been inserted into CaretLine so the above whitespace
350
  // check is guaranteed to include the caret
351
29
352
  // If we have a fix-it line, make sure the slice includes all of the
353
  // fix-it information.
354
29
  if (!FixItInsertionLine.empty()) {
355
7
    unsigned FixItStart = 0, FixItEnd = FixItInsertionLine.size();
356
622
    for (; FixItStart != FixItEnd; 
++FixItStart615
)
357
622
      if (!isWhitespace(FixItInsertionLine[FixItStart]))
358
7
        break;
359
7
360
10
    for (; FixItEnd != FixItStart; 
--FixItEnd3
)
361
10
      if (!isWhitespace(FixItInsertionLine[FixItEnd - 1]))
362
7
        break;
363
7
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
7
    unsigned FixItStartCol = FixItStart;
368
7
    unsigned FixItEndCol
369
7
      = llvm::sys::locale::columnWidth(FixItInsertionLine.substr(0, FixItEnd));
370
7
371
7
    CaretStart = std::min(FixItStartCol, CaretStart);
372
7
    CaretEnd = std::max(FixItEndCol, CaretEnd);
373
7
  }
374
29
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
29
  while (static_cast<int>(CaretEnd) < map.columns() &&
379
22
         -1 == map.columnToByte(CaretEnd))
380
0
    ++CaretEnd;
381
29
382
29
  assert((static_cast<int>(CaretStart) > map.columns() ||
383
29
          -1!=map.columnToByte(CaretStart)) &&
384
29
         "CaretStart must not point to a column in the middle of a source"
385
29
         " line character");
386
29
  assert((static_cast<int>(CaretEnd) > map.columns() ||
387
29
          -1!=map.columnToByte(CaretEnd)) &&
388
29
         "CaretEnd must not point to a column in the middle of a source line"
389
29
         " character");
390
29
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
29
396
29
  unsigned SourceStart = map.columnToByte(std::min<unsigned>(CaretStart,
397
29
                                                             map.columns()));
398
29
  unsigned SourceEnd = map.columnToByte(std::min<unsigned>(CaretEnd,
399
29
                                                           map.columns()));
400
29
401
29
  unsigned CaretColumnsOutsideSource = CaretEnd-CaretStart
402
29
    - (map.byteToColumn(SourceEnd)-map.byteToColumn(SourceStart));
403
29
404
29
  char const *front_ellipse = "  ...";
405
29
  char const *front_space   = "     ";
406
29
  char const *back_ellipse = "...";
407
29
  unsigned ellipses_space = strlen(front_ellipse) + strlen(back_ellipse);
408
29
409
29
  unsigned TargetColumns = Columns;
410
  // Give us extra room for the ellipses
411
  //  and any of the caret line that extends past the source
412
29
  if (TargetColumns > ellipses_space+CaretColumnsOutsideSource)
413
17
    TargetColumns -= ellipses_space+CaretColumnsOutsideSource;
414
29
415
155
  while (SourceStart>0 || 
SourceEnd<SourceLine.size()85
) {
416
155
    bool ExpandedRegion = false;
417
155
418
155
    if (SourceStart>0) {
419
70
      unsigned NewStart = map.startOfPreviousColumn(SourceStart);
420
70
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
346
      while (NewStart && 
isWhitespace(SourceLine[NewStart])336
)
425
276
        NewStart = map.startOfPreviousColumn(NewStart);
426
70
427
      // Skip over this bit of "interesting" text.
428
817
      while (NewStart) {
429
806
        unsigned Prev = map.startOfPreviousColumn(NewStart);
430
806
        if (isWhitespace(SourceLine[Prev]))
431
59
          break;
432
747
        NewStart = Prev;
433
747
      }
434
70
435
70
      assert(map.byteToColumn(NewStart) != -1);
436
70
      unsigned NewColumns = map.byteToColumn(SourceEnd) -
437
70
                              map.byteToColumn(NewStart);
438
70
      if (NewColumns <= TargetColumns) {
439
38
        SourceStart = NewStart;
440
38
        ExpandedRegion = true;
441
38
      }
442
70
    }
443
155
444
155
    if (SourceEnd<SourceLine.size()) {
445
125
      unsigned NewEnd = map.startOfNextColumn(SourceEnd);
446
125
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
171
      while (NewEnd < SourceLine.size() && 
isWhitespace(SourceLine[NewEnd])160
)
451
46
        NewEnd = map.startOfNextColumn(NewEnd);
452
125
453
      // Skip over this bit of "interesting" text.
454
125
      while (NewEnd < SourceLine.size() && 
isWhitespace(SourceLine[NewEnd])114
)
455
0
        NewEnd = map.startOfNextColumn(NewEnd);
456
125
457
125
      assert(map.byteToColumn(NewEnd) != -1);
458
125
      unsigned NewColumns = map.byteToColumn(NewEnd) -
459
125
                              map.byteToColumn(SourceStart);
460
125
      if (NewColumns <= TargetColumns) {
461
104
        SourceEnd = NewEnd;
462
104
        ExpandedRegion = true;
463
104
      }
464
125
    }
465
155
466
155
    if (!ExpandedRegion)
467
29
      break;
468
155
  }
469
29
470
29
  CaretStart = map.byteToColumn(SourceStart);
471
29
  CaretEnd = map.byteToColumn(SourceEnd) + CaretColumnsOutsideSource;
472
29
473
  // [CaretStart, CaretEnd) is the slice we want. Update the various
474
  // output lines to show only this slice, with two-space padding
475
  // before the lines so that it looks nicer.
476
29
477
29
  assert(CaretStart!=(unsigned)-1 && CaretEnd!=(unsigned)-1 &&
478
29
         SourceStart!=(unsigned)-1 && SourceEnd!=(unsigned)-1);
479
29
  assert(SourceStart <= SourceEnd);
480
29
  assert(CaretStart <= CaretEnd);
481
29
482
29
  unsigned BackColumnsRemoved
483
29
    = map.byteToColumn(SourceLine.size())-map.byteToColumn(SourceEnd);
484
29
  unsigned FrontColumnsRemoved = CaretStart;
485
29
  unsigned ColumnsKept = CaretEnd-CaretStart;
486
29
487
  // We checked up front that the line needed truncation
488
29
  assert(FrontColumnsRemoved+ColumnsKept+BackColumnsRemoved > Columns);
489
29
490
  // The line needs some truncation, and we'd prefer to keep the front
491
  //  if possible, so remove the back
492
29
  if (BackColumnsRemoved > strlen(back_ellipse))
493
11
    SourceLine.replace(SourceEnd, std::string::npos, back_ellipse);
494
29
495
  // If that's enough then we're done
496
29
  if (FrontColumnsRemoved+ColumnsKept <= Columns)
497
4
    return;
498
25
499
  // Otherwise remove the front as well
500
25
  if (FrontColumnsRemoved > strlen(front_ellipse)) {
501
20
    SourceLine.replace(0, SourceStart, front_ellipse);
502
20
    CaretLine.replace(0, CaretStart, front_space);
503
20
    if (!FixItInsertionLine.empty())
504
6
      FixItInsertionLine.replace(0, CaretStart, front_space);
505
20
  }
506
25
}
507
508
/// Skip over whitespace in the string, starting at the given
509
/// index.
510
///
511
/// \returns The index of the first non-whitespace character that is
512
/// greater than or equal to Idx or, if no such character exists,
513
/// returns the end of the string.
514
2.75k
static unsigned skipWhitespace(unsigned Idx, StringRef Str, unsigned Length) {
515
5.20k
  while (Idx < Length && isWhitespace(Str[Idx]))
516
2.45k
    ++Idx;
517
2.75k
  return Idx;
518
2.75k
}
519
520
/// If the given character is the start of some kind of
521
/// balanced punctuation (e.g., quotes or parentheses), return the
522
/// character that will terminate the punctuation.
523
///
524
/// \returns The ending punctuation character, if any, or the NULL
525
/// character if the input character does not start any punctuation.
526
15.7k
static inline char findMatchingPunctuation(char c) {
527
15.7k
  switch (c) {
528
573
  case '\'': return '\'';
529
0
  case '`': return '\'';
530
0
  case '"':  return '"';
531
33
  case '(':  return ')';
532
294
  case '[': return ']';
533
0
  case '{': return '}';
534
14.8k
  default: break;
535
14.8k
  }
536
14.8k
537
14.8k
  return 0;
538
14.8k
}
539
540
/// Find the end of the word starting at the given offset
541
/// within a string.
542
///
543
/// \returns the index pointing one character past the end of the
544
/// word.
545
static unsigned findEndOfWord(unsigned Start, StringRef Str,
546
                              unsigned Length, unsigned Column,
547
2.81k
                              unsigned Columns) {
548
2.81k
  assert(Start < Str.size() && "Invalid start position!");
549
2.81k
  unsigned End = Start + 1;
550
2.81k
551
  // If we are already at the end of the string, take that as the word.
552
2.81k
  if (End == Str.size())
553
0
    return End;
554
2.81k
555
  // Determine if the start of the string is actually opening
556
  // punctuation, e.g., a quote or parentheses.
557
2.81k
  char EndPunct = findMatchingPunctuation(Str[Start]);
558
2.81k
  if (!EndPunct) {
559
    // This is a normal word. Just find the first space character.
560
15.3k
    while (End < Length && 
!isWhitespace(Str[End])15.2k
)
561
13.3k
      ++End;
562
1.93k
    return End;
563
1.93k
  }
564
876
565
  // We have the start of a balanced punctuation sequence (quotes,
566
  // parentheses, etc.). Determine the full sequence is.
567
876
  SmallString<16> PunctuationEndStack;
568
876
  PunctuationEndStack.push_back(EndPunct);
569
14.7k
  while (End < Length && 
!PunctuationEndStack.empty()14.4k
) {
570
13.8k
    if (Str[End] == PunctuationEndStack.back())
571
900
      PunctuationEndStack.pop_back();
572
12.9k
    else if (char SubEndPunct = findMatchingPunctuation(Str[End]))
573
24
      PunctuationEndStack.push_back(SubEndPunct);
574
13.8k
575
13.8k
    ++End;
576
13.8k
  }
577
876
578
  // Find the first space character after the punctuation ended.
579
918
  while (End < Length && 
!isWhitespace(Str[End])618
)
580
42
    ++End;
581
876
582
876
  unsigned PunctWordLength = End - Start;
583
876
  if (// If the word fits on this line
584
876
      Column + PunctWordLength <= Columns ||
585
      // ... or the word is "short enough" to take up the next line
586
      // without too much ugly white space
587
323
      PunctWordLength < Columns/3)
588
818
    return End; // Take the whole thing as a single "word".
589
58
590
  // The whole quoted/parenthesized string is too long to print as a
591
  // single "word". Instead, find the "word" that starts just after
592
  // the punctuation and use that end-point instead. This will recurse
593
  // until it finds something small enough to consider a word.
594
58
  return findEndOfWord(Start + 1, Str, Length, Column + 1, Columns);
595
58
}
596
597
/// Print the given string to a stream, word-wrapping it to
598
/// some number of columns in the process.
599
///
600
/// \param OS the stream to which the word-wrapping string will be
601
/// emitted.
602
/// \param Str the string to word-wrap and output.
603
/// \param Columns the number of columns to word-wrap to.
604
/// \param Column the column number at which the first character of \p
605
/// Str will be printed. This will be non-zero when part of the first
606
/// line has already been printed.
607
/// \param Bold if the current text should be bold
608
/// \param Indentation the number of spaces to indent any lines beyond
609
/// the first line.
610
/// \returns true if word-wrapping was required, or false if the
611
/// string fit on the first line.
612
static bool printWordWrapped(raw_ostream &OS, StringRef Str,
613
                             unsigned Columns,
614
                             unsigned Column = 0,
615
                             bool Bold = false,
616
305
                             unsigned Indentation = WordWrapIndentation) {
617
305
  const unsigned Length = std::min(Str.find('\n'), Str.size());
618
305
  bool TextNormal = true;
619
305
620
  // The string used to indent each line.
621
305
  SmallString<16> IndentStr;
622
305
  IndentStr.assign(Indentation, ' ');
623
305
  bool Wrapped = false;
624
3.06k
  for (unsigned WordStart = 0, WordEnd; WordStart < Length;
625
2.75k
       WordStart = WordEnd) {
626
    // Find the beginning of the next word.
627
2.75k
    WordStart = skipWhitespace(WordStart, Str, Length);
628
2.75k
    if (WordStart == Length)
629
0
      break;
630
2.75k
631
    // Find the end of this word.
632
2.75k
    WordEnd = findEndOfWord(WordStart, Str, Length, Column, Columns);
633
2.75k
634
    // Does this word fit on the current line?
635
2.75k
    unsigned WordLength = WordEnd - WordStart;
636
2.75k
    if (Column + WordLength < Columns) {
637
      // This word fits on the current line; print it there.
638
1.74k
      if (WordStart) {
639
1.74k
        OS << ' ';
640
1.74k
        Column += 1;
641
1.74k
      }
642
1.74k
      applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
643
1.74k
                                TextNormal, Bold);
644
1.74k
      Column += WordLength;
645
1.74k
      continue;
646
1.74k
    }
647
1.00k
648
    // This word does not fit on the current line, so wrap to the next
649
    // line.
650
1.00k
    OS << '\n';
651
1.00k
    OS.write(&IndentStr[0], Indentation);
652
1.00k
    applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
653
1.00k
                              TextNormal, Bold);
654
1.00k
    Column = Indentation + WordLength;
655
1.00k
    Wrapped = true;
656
1.00k
  }
657
305
658
  // Append any remaning text from the message with its existing formatting.
659
305
  applyTemplateHighlighting(OS, Str.substr(Length), TextNormal, Bold);
660
305
661
305
  assert(TextNormal && "Text highlighted at end of diagnostic message.");
662
305
663
305
  return Wrapped;
664
305
}
665
666
TextDiagnostic::TextDiagnostic(raw_ostream &OS,
667
                               const LangOptions &LangOpts,
668
                               DiagnosticOptions *DiagOpts)
669
80.8k
  : DiagnosticRenderer(LangOpts, DiagOpts), OS(OS) {}
670
671
80.7k
TextDiagnostic::~TextDiagnostic() {}
672
673
void TextDiagnostic::emitDiagnosticMessage(
674
    FullSourceLoc Loc, PresumedLoc PLoc, DiagnosticsEngine::Level Level,
675
    StringRef Message, ArrayRef<clang::CharSourceRange> Ranges,
676
25.6k
    DiagOrStoredDiag D) {
677
25.6k
  uint64_t StartOfLocationInfo = OS.tell();
678
25.6k
679
  // Emit the location of this particular diagnostic.
680
25.6k
  if (Loc.isValid())
681
25.6k
    emitDiagnosticLoc(Loc, PLoc, Level, Ranges);
682
25.6k
683
25.6k
  if (DiagOpts->ShowColors)
684
51
    OS.resetColor();
685
25.6k
686
25.6k
  if (DiagOpts->ShowLevel)
687
24.7k
    printDiagnosticLevel(OS, Level, DiagOpts->ShowColors,
688
24.7k
                         DiagOpts->CLFallbackMode);
689
25.6k
  printDiagnosticMessage(OS,
690
25.6k
                         /*IsSupplemental*/ Level == DiagnosticsEngine::Note,
691
25.6k
                         Message, OS.tell() - StartOfLocationInfo,
692
25.6k
                         DiagOpts->MessageLength, DiagOpts->ShowColors);
693
25.6k
}
694
695
/*static*/ void
696
TextDiagnostic::printDiagnosticLevel(raw_ostream &OS,
697
                                     DiagnosticsEngine::Level Level,
698
                                     bool ShowColors,
699
31.5k
                                     bool CLFallbackMode) {
700
31.5k
  if (ShowColors) {
701
    // Print diagnostic category in bold and color
702
52
    switch (Level) {
703
0
    case DiagnosticsEngine::Ignored:
704
0
      llvm_unreachable("Invalid diagnostic type");
705
28
    case DiagnosticsEngine::Note:    OS.changeColor(noteColor, true); break;
706
0
    case DiagnosticsEngine::Remark:  OS.changeColor(remarkColor, true); break;
707
0
    case DiagnosticsEngine::Warning: OS.changeColor(warningColor, true); break;
708
23
    case DiagnosticsEngine::Error:   OS.changeColor(errorColor, true); break;
709
1
    case DiagnosticsEngine::Fatal:   OS.changeColor(fatalColor, true); break;
710
31.5k
    }
711
31.5k
  }
712
31.5k
713
31.5k
  switch (Level) {
714
0
  case DiagnosticsEngine::Ignored:
715
0
    llvm_unreachable("Invalid diagnostic type");
716
10.1k
  case DiagnosticsEngine::Note:    OS << "note"; break;
717
219
  case DiagnosticsEngine::Remark:  OS << "remark"; break;
718
14.7k
  case DiagnosticsEngine::Warning: OS << "warning"; break;
719
6.32k
  case DiagnosticsEngine::Error:   OS << "error"; break;
720
118
  case DiagnosticsEngine::Fatal:   OS << "fatal error"; break;
721
31.5k
  }
722
31.5k
723
  // In clang-cl /fallback mode, print diagnostics as "error(clang):". This
724
  // makes it more clear whether a message is coming from clang or cl.exe,
725
  // and it prevents MSBuild from concluding that the build failed just because
726
  // there is an "error:" in the output.
727
31.5k
  if (CLFallbackMode)
728
5
    OS << "(clang)";
729
31.5k
730
31.5k
  OS << ": ";
731
31.5k
732
31.5k
  if (ShowColors)
733
52
    OS.resetColor();
734
31.5k
}
735
736
/*static*/
737
void TextDiagnostic::printDiagnosticMessage(raw_ostream &OS,
738
                                            bool IsSupplemental,
739
                                            StringRef Message,
740
                                            unsigned CurrentColumn,
741
32.4k
                                            unsigned Columns, bool ShowColors) {
742
32.4k
  bool Bold = false;
743
32.4k
  if (ShowColors && 
!IsSupplemental52
) {
744
    // Print primary diagnostic messages in bold and without color, to visually
745
    // indicate the transition from continuation notes and other output.
746
23
    OS.changeColor(savedColor, true);
747
23
    Bold = true;
748
23
  }
749
32.4k
750
32.4k
  if (Columns)
751
305
    printWordWrapped(OS, Message, Columns, CurrentColumn, Bold);
752
32.1k
  else {
753
32.1k
    bool Normal = true;
754
32.1k
    applyTemplateHighlighting(OS, Message, Normal, Bold);
755
32.1k
    assert(Normal && "Formatting should have returned to normal");
756
32.1k
  }
757
32.4k
758
32.4k
  if (ShowColors)
759
52
    OS.resetColor();
760
32.4k
  OS << '\n';
761
32.4k
}
762
763
25.6k
void TextDiagnostic::emitFilename(StringRef Filename, const SourceManager &SM) {
764
#ifdef _WIN32
765
  SmallString<4096> TmpFilename;
766
#endif
767
25.6k
  if (DiagOpts->AbsolutePath) {
768
4
    auto File = SM.getFileManager().getFile(Filename);
769
4
    if (File) {
770
      // We want to print a simplified absolute path, i. e. without "dots".
771
      //
772
      // The hardest part here are the paths like "<part1>/<link>/../<part2>".
773
      // On Unix-like systems, we cannot just collapse "<link>/..", because
774
      // paths are resolved sequentially, and, thereby, the path
775
      // "<part1>/<part2>" may point to a different location. That is why
776
      // we use FileManager::getCanonicalName(), which expands all indirections
777
      // with llvm::sys::fs::real_path() and caches the result.
778
      //
779
      // On the other hand, it would be better to preserve as much of the
780
      // original path as possible, because that helps a user to recognize it.
781
      // real_path() expands all links, which sometimes too much. Luckily,
782
      // on Windows we can just use llvm::sys::path::remove_dots(), because,
783
      // on that system, both aforementioned paths point to the same place.
784
#ifdef _WIN32
785
      TmpFilename = (*File)->getName();
786
      llvm::sys::fs::make_absolute(TmpFilename);
787
      llvm::sys::path::native(TmpFilename);
788
      llvm::sys::path::remove_dots(TmpFilename, /* remove_dot_dot */ true);
789
      Filename = StringRef(TmpFilename.data(), TmpFilename.size());
790
#else
791
3
      Filename = SM.getFileManager().getCanonicalName(*File);
792
3
#endif
793
3
    }
794
4
  }
795
25.6k
796
25.6k
  OS << Filename;
797
25.6k
}
798
799
/// Print out the file/line/column information and include trace.
800
///
801
/// This method handlen the emission of the diagnostic location information.
802
/// This includes extracting as much location information as is present for
803
/// the diagnostic and printing it, as well as any include stack or source
804
/// ranges necessary.
805
void TextDiagnostic::emitDiagnosticLoc(FullSourceLoc Loc, PresumedLoc PLoc,
806
                                       DiagnosticsEngine::Level Level,
807
25.6k
                                       ArrayRef<CharSourceRange> Ranges) {
808
25.6k
  if (PLoc.isInvalid()) {
809
    // At least print the file name if available:
810
2
    FileID FID = Loc.getFileID();
811
2
    if (FID.isValid()) {
812
2
      const FileEntry *FE = Loc.getFileEntry();
813
2
      if (FE && 
FE->isValid()0
) {
814
0
        emitFilename(FE->getName(), Loc.getManager());
815
0
        OS << ": ";
816
0
      }
817
2
    }
818
2
    return;
819
2
  }
820
25.6k
  unsigned LineNo = PLoc.getLine();
821
25.6k
822
25.6k
  if (!DiagOpts->ShowLocation)
823
0
    return;
824
25.6k
825
25.6k
  if (DiagOpts->ShowColors)
826
51
    OS.changeColor(savedColor, true);
827
25.6k
828
25.6k
  emitFilename(PLoc.getFilename(), Loc.getManager());
829
25.6k
  switch (DiagOpts->getFormat()) {
830
25.6k
  case DiagnosticOptions::Clang: OS << ':'  << LineNo; break;
831
25
  case DiagnosticOptions::MSVC:  OS << '('  << LineNo; break;
832
8
  case DiagnosticOptions::Vi:    OS << " +" << LineNo; break;
833
25.6k
  }
834
25.6k
835
25.6k
  if (DiagOpts->ShowColumn)
836
    // Compute the column number.
837
25.6k
    if (unsigned ColNo = PLoc.getColumn()) {
838
25.6k
      if (DiagOpts->getFormat() == DiagnosticOptions::MSVC) {
839
23
        OS << ',';
840
        // Visual Studio 2010 or earlier expects column number to be off by one
841
23
        if (LangOpts.MSCompatibilityVersion &&
842
22
            !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2012))
843
8
          ColNo--;
844
23
      } else
845
25.6k
        OS << ':';
846
25.6k
      OS << ColNo;
847
25.6k
    }
848
25.6k
  switch (DiagOpts->getFormat()) {
849
25.6k
  case DiagnosticOptions::Clang:
850
25.6k
  case DiagnosticOptions::Vi:    OS << ':';    break;
851
25
  case DiagnosticOptions::MSVC:
852
    // MSVC2013 and before print 'file(4) : error'. MSVC2015 gets rid of the
853
    // space and prints 'file(4): error'.
854
25
    OS << ')';
855
25
    if (LangOpts.MSCompatibilityVersion &&
856
24
        !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2015))
857
11
      OS << ' ';
858
25
    OS << ':';
859
25
    break;
860
25.6k
  }
861
25.6k
862
25.6k
  if (DiagOpts->ShowSourceRanges && 
!Ranges.empty()17
) {
863
9
    FileID CaretFileID = Loc.getExpansionLoc().getFileID();
864
9
    bool PrintedRange = false;
865
9
866
9
    for (ArrayRef<CharSourceRange>::const_iterator RI = Ranges.begin(),
867
9
         RE = Ranges.end();
868
24
         RI != RE; 
++RI15
) {
869
      // Ignore invalid ranges.
870
15
      if (!RI->isValid()) 
continue2
;
871
13
872
13
      auto &SM = Loc.getManager();
873
13
      SourceLocation B = SM.getExpansionLoc(RI->getBegin());
874
13
      CharSourceRange ERange = SM.getExpansionRange(RI->getEnd());
875
13
      SourceLocation E = ERange.getEnd();
876
13
      bool IsTokenRange = ERange.isTokenRange();
877
13
878
13
      std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(B);
879
13
      std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(E);
880
13
881
      // If the start or end of the range is in another file, just discard
882
      // it.
883
13
      if (BInfo.first != CaretFileID || EInfo.first != CaretFileID)
884
0
        continue;
885
13
886
      // Add in the length of the token, so that we cover multi-char
887
      // tokens.
888
13
      unsigned TokSize = 0;
889
13
      if (IsTokenRange)
890
9
        TokSize = Lexer::MeasureTokenLength(E, SM, LangOpts);
891
13
892
13
      FullSourceLoc BF(B, SM), EF(E, SM);
893
13
      OS << '{'
894
13
         << BF.getLineNumber() << ':' << BF.getColumnNumber() << '-'
895
13
         << EF.getLineNumber() << ':' << (EF.getColumnNumber() + TokSize)
896
13
         << '}';
897
13
      PrintedRange = true;
898
13
    }
899
9
900
9
    if (PrintedRange)
901
9
      OS << ':';
902
9
  }
903
25.6k
  OS << ' ';
904
25.6k
}
905
906
502
void TextDiagnostic::emitIncludeLocation(FullSourceLoc Loc, PresumedLoc PLoc) {
907
502
  if (DiagOpts->ShowLocation && PLoc.isValid())
908
502
    OS << "In file included from " << PLoc.getFilename() << ':'
909
502
       << PLoc.getLine() << ":\n";
910
0
  else
911
0
    OS << "In included file:\n";
912
502
}
913
914
void TextDiagnostic::emitImportLocation(FullSourceLoc Loc, PresumedLoc PLoc,
915
26
                                        StringRef ModuleName) {
916
26
  if (DiagOpts->ShowLocation && PLoc.isValid())
917
22
    OS << "In module '" << ModuleName << "' imported from "
918
22
       << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
919
4
  else
920
4
    OS << "In module '" << ModuleName << "':\n";
921
26
}
922
923
void TextDiagnostic::emitBuildingModuleLocation(FullSourceLoc Loc,
924
                                                PresumedLoc PLoc,
925
94
                                                StringRef ModuleName) {
926
94
  if (DiagOpts->ShowLocation && PLoc.isValid())
927
84
    OS << "While building module '" << ModuleName << "' imported from "
928
84
      << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
929
10
  else
930
10
    OS << "While building module '" << ModuleName << "':\n";
931
94
}
932
933
/// Find the suitable set of lines to show to include a set of ranges.
934
static llvm::Optional<std::pair<unsigned, unsigned>>
935
findLinesForRange(const CharSourceRange &R, FileID FID,
936
14.9k
                  const SourceManager &SM) {
937
14.9k
  if (!R.isValid()) 
return None0
;
938
14.9k
939
14.9k
  SourceLocation Begin = R.getBegin();
940
14.9k
  SourceLocation End = R.getEnd();
941
14.9k
  if (SM.getFileID(Begin) != FID || SM.getFileID(End) != FID)
942
0
    return None;
943
14.9k
944
14.9k
  return std::make_pair(SM.getExpansionLineNumber(Begin),
945
14.9k
                        SM.getExpansionLineNumber(End));
946
14.9k
}
947
948
/// Add as much of range B into range A as possible without exceeding a maximum
949
/// size of MaxRange. Ranges are inclusive.
950
static std::pair<unsigned, unsigned>
951
maybeAddRange(std::pair<unsigned, unsigned> A, std::pair<unsigned, unsigned> B,
952
14.9k
              unsigned MaxRange) {
953
  // If A is already the maximum size, we're done.
954
14.9k
  unsigned Slack = MaxRange - (A.second - A.first + 1);
955
14.9k
  if (Slack == 0)
956
14.9k
    return A;
957
29
958
  // Easy case: merge succeeds within MaxRange.
959
29
  unsigned Min = std::min(A.first, B.first);
960
29
  unsigned Max = std::max(A.second, B.second);
961
29
  if (Max - Min + 1 <= MaxRange)
962
24
    return {Min, Max};
963
5
964
  // If we can't reach B from A within MaxRange, there's nothing to do.
965
  // Don't add lines to the range that contain nothing interesting.
966
5
  if ((B.first > A.first && 
B.first - A.first + 1 > MaxRange2
) ||
967
5
      (B.second < A.second && 
A.second - B.second + 1 > MaxRange1
))
968
1
    return A;
969
4
970
  // Otherwise, expand A towards B to produce a range of size MaxRange. We
971
  // attempt to expand by the same amount in both directions if B strictly
972
  // contains A.
973
4
974
  // Expand downwards by up to half the available amount, then upwards as
975
  // much as possible, then downwards as much as possible.
976
4
  A.second = std::min(A.second + (Slack + 1) / 2, Max);
977
4
  Slack = MaxRange - (A.second - A.first + 1);
978
4
  A.first = std::max(Min + Slack, A.first) - Slack;
979
4
  A.second = std::min(A.first + MaxRange - 1, Max);
980
4
  return A;
981
4
}
982
983
/// Highlight a SourceRange (with ~'s) for any characters on LineNo.
984
static void highlightRange(const CharSourceRange &R,
985
                           unsigned LineNo, FileID FID,
986
                           const SourceColumnMap &map,
987
                           std::string &CaretLine,
988
                           const SourceManager &SM,
989
14.1k
                           const LangOptions &LangOpts) {
990
14.1k
  if (!R.isValid()) 
return0
;
991
14.1k
992
14.1k
  SourceLocation Begin = R.getBegin();
993
14.1k
  SourceLocation End = R.getEnd();
994
14.1k
995
14.1k
  unsigned StartLineNo = SM.getExpansionLineNumber(Begin);
996
14.1k
  if (StartLineNo > LineNo || 
SM.getFileID(Begin) != FID13.1k
)
997
1.02k
    return;  // No intersection.
998
13.1k
999
13.1k
  unsigned EndLineNo = SM.getExpansionLineNumber(End);
1000
13.1k
  if (EndLineNo < LineNo || 
SM.getFileID(End) != FID13.0k
)
1001
39
    return;  // No intersection.
1002
13.0k
1003
  // Compute the column number of the start.
1004
13.0k
  unsigned StartColNo = 0;
1005
13.0k
  if (StartLineNo == LineNo) {
1006
13.0k
    StartColNo = SM.getExpansionColumnNumber(Begin);
1007
13.0k
    if (StartColNo) --StartColNo;  // Zero base the col #.
1008
13.0k
  }
1009
13.0k
1010
  // Compute the column number of the end.
1011
13.0k
  unsigned EndColNo = map.getSourceLine().size();
1012
13.0k
  if (EndLineNo == LineNo) {
1013
13.0k
    EndColNo = SM.getExpansionColumnNumber(End);
1014
13.0k
    if (EndColNo) {
1015
13.0k
      --EndColNo;  // Zero base the col #.
1016
13.0k
1017
      // Add in the length of the token, so that we cover multi-char tokens if
1018
      // this is a token range.
1019
13.0k
      if (R.isTokenRange())
1020
9.05k
        EndColNo += Lexer::MeasureTokenLength(End, SM, LangOpts);
1021
0
    } else {
1022
0
      EndColNo = CaretLine.size();
1023
0
    }
1024
13.0k
  }
1025
13.0k
1026
13.0k
  assert(StartColNo <= EndColNo && "Invalid range!");
1027
13.0k
1028
  // Check that a token range does not highlight only whitespace.
1029
13.0k
  if (R.isTokenRange()) {
1030
    // Pick the first non-whitespace column.
1031
9.09k
    while (StartColNo < map.getSourceLine().size() &&
1032
9.09k
           (map.getSourceLine()[StartColNo] == ' ' ||
1033
9.09k
            map.getSourceLine()[StartColNo] == '\t'))
1034
8
      StartColNo = map.startOfNextColumn(StartColNo);
1035
9.09k
1036
    // Pick the last non-whitespace column.
1037
9.09k
    if (EndColNo > map.getSourceLine().size())
1038
1
      EndColNo = map.getSourceLine().size();
1039
9.09k
    while (EndColNo &&
1040
9.09k
           (map.getSourceLine()[EndColNo-1] == ' ' ||
1041
9.09k
            map.getSourceLine()[EndColNo-1] == '\t'))
1042
1
      EndColNo = map.startOfPreviousColumn(EndColNo);
1043
9.09k
1044
    // If the start/end passed each other, then we are trying to highlight a
1045
    // range that just exists in whitespace. That most likely means we have
1046
    // a multi-line highlighting range that covers a blank line.
1047
9.09k
    if (StartColNo > EndColNo) {
1048
0
      assert(StartLineNo != EndLineNo && "trying to highlight whitespace");
1049
0
      StartColNo = EndColNo;
1050
0
    }
1051
9.09k
  }
1052
13.0k
1053
13.0k
  assert(StartColNo <= map.getSourceLine().size() && "Invalid range!");
1054
13.0k
  assert(EndColNo <= map.getSourceLine().size() && "Invalid range!");
1055
13.0k
1056
  // Fill the range with ~'s.
1057
13.0k
  StartColNo = map.byteToContainingColumn(StartColNo);
1058
13.0k
  EndColNo = map.byteToContainingColumn(EndColNo);
1059
13.0k
1060
13.0k
  assert(StartColNo <= EndColNo && "Invalid range!");
1061
13.0k
  if (CaretLine.size() < EndColNo)
1062
0
    CaretLine.resize(EndColNo,' ');
1063
13.0k
  std::fill(CaretLine.begin()+StartColNo,CaretLine.begin()+EndColNo,'~');
1064
13.0k
}
1065
1066
static std::string buildFixItInsertionLine(FileID FID,
1067
                                           unsigned LineNo,
1068
                                           const SourceColumnMap &map,
1069
                                           ArrayRef<FixItHint> Hints,
1070
                                           const SourceManager &SM,
1071
23.6k
                                           const DiagnosticOptions *DiagOpts) {
1072
23.6k
  std::string FixItInsertionLine;
1073
23.6k
  if (Hints.empty() || 
!DiagOpts->ShowFixits2.92k
)
1074
20.6k
    return FixItInsertionLine;
1075
2.92k
  unsigned PrevHintEndCol = 0;
1076
2.92k
1077
2.92k
  for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1078
6.52k
       I != E; 
++I3.59k
) {
1079
3.59k
    if (!I->CodeToInsert.empty()) {
1080
      // We have an insertion hint. Determine whether the inserted
1081
      // code contains no newlines and is on the same line as the caret.
1082
3.02k
      std::pair<FileID, unsigned> HintLocInfo
1083
3.02k
        = SM.getDecomposedExpansionLoc(I->RemoveRange.getBegin());
1084
3.02k
      if (FID == HintLocInfo.first &&
1085
3.02k
          LineNo == SM.getLineNumber(HintLocInfo.first, HintLocInfo.second) &&
1086
2.94k
          StringRef(I->CodeToInsert).find_first_of("\n\r") == StringRef::npos) {
1087
        // Insert the new code into the line just below the code
1088
        // that the user wrote.
1089
        // Note: When modifying this function, be very careful about what is a
1090
        // "column" (printed width, platform-dependent) and what is a
1091
        // "byte offset" (SourceManager "column").
1092
2.89k
        unsigned HintByteOffset
1093
2.89k
          = SM.getColumnNumber(HintLocInfo.first, HintLocInfo.second) - 1;
1094
2.89k
1095
        // The hint must start inside the source or right at the end
1096
2.89k
        assert(HintByteOffset < static_cast<unsigned>(map.bytes())+1);
1097
2.89k
        unsigned HintCol = map.byteToContainingColumn(HintByteOffset);
1098
2.89k
1099
        // If we inserted a long previous hint, push this one forwards, and add
1100
        // an extra space to show that this is not part of the previous
1101
        // completion. This is sort of the best we can do when two hints appear
1102
        // to overlap.
1103
        //
1104
        // Note that if this hint is located immediately after the previous
1105
        // hint, no space will be added, since the location is more important.
1106
2.89k
        if (HintCol < PrevHintEndCol)
1107
85
          HintCol = PrevHintEndCol + 1;
1108
2.89k
1109
        // This should NOT use HintByteOffset, because the source might have
1110
        // Unicode characters in earlier columns.
1111
2.89k
        unsigned NewFixItLineSize = FixItInsertionLine.size() +
1112
2.89k
          (HintCol - PrevHintEndCol) + I->CodeToInsert.size();
1113
2.89k
        if (NewFixItLineSize > FixItInsertionLine.size())
1114
2.89k
          FixItInsertionLine.resize(NewFixItLineSize, ' ');
1115
2.89k
1116
2.89k
        std::copy(I->CodeToInsert.begin(), I->CodeToInsert.end(),
1117
2.89k
                  FixItInsertionLine.end() - I->CodeToInsert.size());
1118
2.89k
1119
2.89k
        PrevHintEndCol =
1120
2.89k
          HintCol + llvm::sys::locale::columnWidth(I->CodeToInsert);
1121
2.89k
      }
1122
3.02k
    }
1123
3.59k
  }
1124
2.92k
1125
2.92k
  expandTabs(FixItInsertionLine, DiagOpts->TabStop);
1126
2.92k
1127
2.92k
  return FixItInsertionLine;
1128
2.92k
}
1129
1130
/// Emit a code snippet and caret line.
1131
///
1132
/// This routine emits a single line's code snippet and caret line..
1133
///
1134
/// \param Loc The location for the caret.
1135
/// \param Ranges The underlined ranges for this code snippet.
1136
/// \param Hints The FixIt hints active for this diagnostic.
1137
void TextDiagnostic::emitSnippetAndCaret(
1138
    FullSourceLoc Loc, DiagnosticsEngine::Level Level,
1139
25.6k
    SmallVectorImpl<CharSourceRange> &Ranges, ArrayRef<FixItHint> Hints) {
1140
25.6k
  assert(Loc.isValid() && "must have a valid source location here");
1141
25.6k
  assert(Loc.isFileID() && "must have a file location here");
1142
25.6k
1143
  // If caret diagnostics are enabled and we have location, we want to
1144
  // emit the caret.  However, we only do this if the location moved
1145
  // from the last diagnostic, if the last diagnostic was a note that
1146
  // was part of a different warning or error diagnostic, or if the
1147
  // diagnostic has ranges.  We don't want to emit the same caret
1148
  // multiple times if one loc has multiple diagnostics.
1149
25.6k
  if (!DiagOpts->ShowCarets)
1150
52
    return;
1151
25.5k
  if (Loc == LastLoc && 
Ranges.empty()2.83k
&&
Hints.empty()1.47k
&&
1152
1.47k
      (LastLevel != DiagnosticsEngine::Note || 
Level == LastLevel60
))
1153
1.46k
    return;
1154
24.1k
1155
  // Decompose the location into a FID/Offset pair.
1156
24.1k
  std::pair<FileID, unsigned> LocInfo = Loc.getDecomposedLoc();
1157
24.1k
  FileID FID = LocInfo.first;
1158
24.1k
  const SourceManager &SM = Loc.getManager();
1159
24.1k
1160
  // Get information about the buffer it points into.
1161
24.1k
  bool Invalid = false;
1162
24.1k
  StringRef BufData = Loc.getBufferData(&Invalid);
1163
24.1k
  if (Invalid)
1164
2
    return;
1165
24.1k
1166
24.1k
  unsigned CaretLineNo = Loc.getLineNumber();
1167
24.1k
  unsigned CaretColNo = Loc.getColumnNumber();
1168
24.1k
1169
  // Arbitrarily stop showing snippets when the line is too long.
1170
24.1k
  static const size_t MaxLineLengthToPrint = 4096;
1171
24.1k
  if (CaretColNo > MaxLineLengthToPrint)
1172
96
    return;
1173
24.0k
1174
  // Find the set of lines to include.
1175
24.0k
  const unsigned MaxLines = DiagOpts->SnippetLineLimit;
1176
24.0k
  std::pair<unsigned, unsigned> Lines = {CaretLineNo, CaretLineNo};
1177
24.0k
  for (SmallVectorImpl<CharSourceRange>::iterator I = Ranges.begin(),
1178
24.0k
                                                  E = Ranges.end();
1179
38.9k
       I != E; 
++I14.9k
)
1180
14.9k
    if (auto OptionalRange = findLinesForRange(*I, FID, SM))
1181
14.9k
      Lines = maybeAddRange(Lines, *OptionalRange, MaxLines);
1182
24.0k
1183
47.6k
  for (unsigned LineNo = Lines.first; LineNo != Lines.second + 1; 
++LineNo23.6k
) {
1184
24.0k
    const char *BufStart = BufData.data();
1185
24.0k
    const char *BufEnd = BufStart + BufData.size();
1186
24.0k
1187
    // Rewind from the current position to the start of the line.
1188
24.0k
    const char *LineStart =
1189
24.0k
        BufStart +
1190
24.0k
        SM.getDecomposedLoc(SM.translateLineCol(FID, LineNo, 1)).second;
1191
24.0k
    if (LineStart == BufEnd)
1192
19
      break;
1193
24.0k
1194
    // Compute the line end.
1195
24.0k
    const char *LineEnd = LineStart;
1196
3.37M
    while (*LineEnd != '\n' && 
*LineEnd != '\r'3.34M
&&
LineEnd != BufEnd3.34M
)
1197
3.34M
      ++LineEnd;
1198
24.0k
1199
    // Arbitrarily stop showing snippets when the line is too long.
1200
    // FIXME: Don't print any lines in this case.
1201
24.0k
    if (size_t(LineEnd - LineStart) > MaxLineLengthToPrint)
1202
429
      return;
1203
23.6k
1204
    // Trim trailing null-bytes.
1205
23.6k
    StringRef Line(LineStart, LineEnd - LineStart);
1206
79.7k
    while (!Line.empty() && 
Line.back() == '\0'79.7k
&&
1207
56.1k
           (LineNo != CaretLineNo || Line.size() > CaretColNo))
1208
56.1k
      Line = Line.drop_back();
1209
23.6k
1210
    // Copy the line of code into an std::string for ease of manipulation.
1211
23.6k
    std::string SourceLine(Line.begin(), Line.end());
1212
23.6k
1213
    // Build the byte to column map.
1214
23.6k
    const SourceColumnMap sourceColMap(SourceLine, DiagOpts->TabStop);
1215
23.6k
1216
    // Create a line for the caret that is filled with spaces that is the same
1217
    // number of columns as the line of source code.
1218
23.6k
    std::string CaretLine(sourceColMap.columns(), ' ');
1219
23.6k
1220
    // Highlight all of the characters covered by Ranges with ~ characters.
1221
23.6k
    for (SmallVectorImpl<CharSourceRange>::iterator I = Ranges.begin(),
1222
23.6k
                                                    E = Ranges.end();
1223
37.7k
         I != E; 
++I14.1k
)
1224
14.1k
      highlightRange(*I, LineNo, FID, sourceColMap, CaretLine, SM, LangOpts);
1225
23.6k
1226
    // Next, insert the caret itself.
1227
23.6k
    if (CaretLineNo == LineNo) {
1228
23.5k
      CaretColNo = sourceColMap.byteToContainingColumn(CaretColNo - 1);
1229
23.5k
      if (CaretLine.size() < CaretColNo + 1)
1230
305
        CaretLine.resize(CaretColNo + 1, ' ');
1231
23.5k
      CaretLine[CaretColNo] = '^';
1232
23.5k
    }
1233
23.6k
1234
23.6k
    std::string FixItInsertionLine = buildFixItInsertionLine(
1235
23.6k
        FID, LineNo, sourceColMap, Hints, SM, DiagOpts.get());
1236
23.6k
1237
    // If the source line is too long for our terminal, select only the
1238
    // "interesting" source region within that line.
1239
23.6k
    unsigned Columns = DiagOpts->MessageLength;
1240
23.6k
    if (Columns)
1241
42
      selectInterestingSourceRegion(SourceLine, CaretLine, FixItInsertionLine,
1242
42
                                    Columns, sourceColMap);
1243
23.6k
1244
    // If we are in -fdiagnostics-print-source-range-info mode, we are trying
1245
    // to produce easily machine parsable output.  Add a space before the
1246
    // source line and the caret to make it trivial to tell the main diagnostic
1247
    // line from what the user is intended to see.
1248
23.6k
    if (DiagOpts->ShowSourceRanges) {
1249
16
      SourceLine = ' ' + SourceLine;
1250
16
      CaretLine = ' ' + CaretLine;
1251
16
    }
1252
23.6k
1253
    // Finally, remove any blank spaces from the end of CaretLine.
1254
618k
    while (!CaretLine.empty() && 
CaretLine[CaretLine.size() - 1] == ' '618k
)
1255
594k
      CaretLine.erase(CaretLine.end() - 1);
1256
23.6k
1257
    // Emit what we have computed.
1258
23.6k
    emitSnippet(SourceLine);
1259
23.6k
1260
23.6k
    if (!CaretLine.empty()) {
1261
23.6k
      if (DiagOpts->ShowColors)
1262
51
        OS.changeColor(caretColor, true);
1263
23.6k
      OS << CaretLine << '\n';
1264
23.6k
      if (DiagOpts->ShowColors)
1265
51
        OS.resetColor();
1266
23.6k
    }
1267
23.6k
1268
23.6k
    if (!FixItInsertionLine.empty()) {
1269
2.39k
      if (DiagOpts->ShowColors)
1270
        // Print fixit line in color
1271
0
        OS.changeColor(fixitColor, false);
1272
2.39k
      if (DiagOpts->ShowSourceRanges)
1273
4
        OS << ' ';
1274
2.39k
      OS << FixItInsertionLine << '\n';
1275
2.39k
      if (DiagOpts->ShowColors)
1276
0
        OS.resetColor();
1277
2.39k
    }
1278
23.6k
  }
1279
24.0k
1280
  // Print out any parseable fixit information requested by the options.
1281
23.6k
  emitParseableFixits(Hints, SM);
1282
23.6k
}
1283
1284
23.6k
void TextDiagnostic::emitSnippet(StringRef line) {
1285
23.6k
  if (line.empty())
1286
1
    return;
1287
23.6k
1288
23.6k
  size_t i = 0;
1289
23.6k
1290
23.6k
  std::string to_print;
1291
23.6k
  bool print_reversed = false;
1292
23.6k
1293
1.13M
  while (i<line.size()) {
1294
1.11M
    std::pair<SmallString<16>,bool> res
1295
1.11M
        = printableTextForNextCharacter(line, &i, DiagOpts->TabStop);
1296
1.11M
    bool was_printable = res.second;
1297
1.11M
1298
1.11M
    if (DiagOpts->ShowColors && 
was_printable == print_reversed1.62k
) {
1299
0
      if (print_reversed)
1300
0
        OS.reverseColor();
1301
0
      OS << to_print;
1302
0
      to_print.clear();
1303
0
      if (DiagOpts->ShowColors)
1304
0
        OS.resetColor();
1305
0
    }
1306
1.11M
1307
1.11M
    print_reversed = !was_printable;
1308
1.11M
    to_print += res.first.str();
1309
1.11M
  }
1310
23.6k
1311
23.6k
  if (print_reversed && 
DiagOpts->ShowColors1
)
1312
0
    OS.reverseColor();
1313
23.6k
  OS << to_print;
1314
23.6k
  if (print_reversed && 
DiagOpts->ShowColors1
)
1315
0
    OS.resetColor();
1316
23.6k
1317
23.6k
  OS << '\n';
1318
23.6k
}
1319
1320
void TextDiagnostic::emitParseableFixits(ArrayRef<FixItHint> Hints,
1321
23.6k
                                         const SourceManager &SM) {
1322
23.6k
  if (!DiagOpts->ShowParseableFixits)
1323
20.6k
    return;
1324
2.94k
1325
  // We follow FixItRewriter's example in not (yet) handling
1326
  // fix-its in macros.
1327
2.94k
  for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1328
4.74k
       I != E; 
++I1.80k
) {
1329
1.80k
    if (I->RemoveRange.isInvalid() ||
1330
1.80k
        I->RemoveRange.getBegin().isMacroID() ||
1331
1.80k
        I->RemoveRange.getEnd().isMacroID())
1332
0
      return;
1333
1.80k
  }
1334
2.94k
1335
2.94k
  for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1336
4.74k
       I != E; 
++I1.80k
) {
1337
1.80k
    SourceLocation BLoc = I->RemoveRange.getBegin();
1338
1.80k
    SourceLocation ELoc = I->RemoveRange.getEnd();
1339
1.80k
1340
1.80k
    std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(BLoc);
1341
1.80k
    std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(ELoc);
1342
1.80k
1343
    // Adjust for token ranges.
1344
1.80k
    if (I->RemoveRange.isTokenRange())
1345
0
      EInfo.second += Lexer::MeasureTokenLength(ELoc, SM, LangOpts);
1346
1.80k
1347
    // We specifically do not do word-wrapping or tab-expansion here,
1348
    // because this is supposed to be easy to parse.
1349
1.80k
    PresumedLoc PLoc = SM.getPresumedLoc(BLoc);
1350
1.80k
    if (PLoc.isInvalid())
1351
0
      break;
1352
1.80k
1353
1.80k
    OS << "fix-it:\"";
1354
1.80k
    OS.write_escaped(PLoc.getFilename());
1355
1.80k
    OS << "\":{" << SM.getLineNumber(BInfo.first, BInfo.second)
1356
1.80k
      << ':' << SM.getColumnNumber(BInfo.first, BInfo.second)
1357
1.80k
      << '-' << SM.getLineNumber(EInfo.first, EInfo.second)
1358
1.80k
      << ':' << SM.getColumnNumber(EInfo.first, EInfo.second)
1359
1.80k
      << "}:\"";
1360
1.80k
    OS.write_escaped(I->CodeToInsert);
1361
1.80k
    OS << "\"\n";
1362
1.80k
  }
1363
2.94k
}