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                                       bool &Normal, bool Bold) {
47   while (1) {
48     size_t Pos = Str.find(ToggleHighlight);
49     OS << Str.slice(0, Pos);
50     if (Pos == StringRef::npos)
51       break;
52 
53     Str = Str.substr(Pos + 1);
54     if (Normal)
55       OS.changeColor(templateColor, true);
56     else {
57       OS.resetColor();
58       if (Bold)
59         OS.changeColor(savedColor, true);
60     }
61     Normal = !Normal;
62   }
63 }
64 
65 /// Number of spaces to indent when word-wrapping.
66 const unsigned WordWrapIndentation = 6;
67 
68 static int bytesSincePreviousTabOrLineBegin(StringRef SourceLine, size_t i) {
69   int bytes = 0;
70   while (0<i) {
71     if (SourceLine[--i]=='\t')
72       break;
73     ++bytes;
74   }
75   return bytes;
76 }
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                               unsigned TabStop) {
100   assert(i && "i must not be null");
101   assert(*i<SourceLine.size() && "must point to a valid index");
102 
103   if (SourceLine[*i]=='\t') {
104     assert(0 < TabStop && TabStop <= DiagnosticOptions::MaxTabStop &&
105            "Invalid -ftabstop value");
106     unsigned col = bytesSincePreviousTabOrLineBegin(SourceLine, *i);
107     unsigned NumSpaces = TabStop - col%TabStop;
108     assert(0 < NumSpaces && NumSpaces <= TabStop
109            && "Invalid computation of space amt");
110     ++(*i);
111 
112     SmallString<16> expandedTab;
113     expandedTab.assign(NumSpaces, ' ');
114     return std::make_pair(expandedTab, true);
115   }
116 
117   unsigned char const *begin, *end;
118   begin = reinterpret_cast<unsigned char const *>(&*(SourceLine.begin() + *i));
119   end = begin + (SourceLine.size() - *i);
120 
121   if (llvm::isLegalUTF8Sequence(begin, end)) {
122     llvm::UTF32 c;
123     llvm::UTF32 *cptr = &c;
124     unsigned char const *original_begin = begin;
125     unsigned char const *cp_end =
126         begin + llvm::getNumBytesForUTF8(SourceLine[*i]);
127 
128     llvm::ConversionResult res = llvm::ConvertUTF8toUTF32(
129         &begin, cp_end, &cptr, cptr + 1, llvm::strictConversion);
130     (void)res;
131     assert(llvm::conversionOK == res);
132     assert(0 < begin-original_begin
133            && "we must be further along in the string now");
134     *i += begin-original_begin;
135 
136     if (!llvm::sys::locale::isPrint(c)) {
137       // If next character is valid UTF-8, but not printable
138       SmallString<16> expandedCP("<U+>");
139       while (c) {
140         expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(c%16));
141         c/=16;
142       }
143       while (expandedCP.size() < 8)
144         expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(0));
145       return std::make_pair(expandedCP, false);
146     }
147 
148     // If next character is valid UTF-8, and printable
149     return std::make_pair(SmallString<16>(original_begin, cp_end), true);
150 
151   }
152 
153   // If next byte is not valid UTF-8 (and therefore not printable)
154   SmallString<16> expandedByte("<XX>");
155   unsigned char byte = SourceLine[*i];
156   expandedByte[1] = llvm::hexdigit(byte / 16);
157   expandedByte[2] = llvm::hexdigit(byte % 16);
158   ++(*i);
159   return std::make_pair(expandedByte, false);
160 }
161 
162 static void expandTabs(std::string &SourceLine, unsigned TabStop) {
163   size_t i = SourceLine.size();
164   while (i>0) {
165     i--;
166     if (SourceLine[i]!='\t')
167       continue;
168     size_t tmp_i = i;
169     std::pair<SmallString<16>,bool> res
170       = printableTextForNextCharacter(SourceLine, &tmp_i, TabStop);
171     SourceLine.replace(i, 1, res.first.c_str());
172   }
173 }
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                          SmallVectorImpl<int> &out) {
199   out.clear();
200 
201   if (SourceLine.empty()) {
202     out.resize(1u,0);
203     return;
204   }
205 
206   out.resize(SourceLine.size()+1, -1);
207 
208   int columns = 0;
209   size_t i = 0;
210   while (i<SourceLine.size()) {
211     out[i] = columns;
212     std::pair<SmallString<16>,bool> res
213       = printableTextForNextCharacter(SourceLine, &i, TabStop);
214     columns += llvm::sys::locale::columnWidth(res.first);
215   }
216   out.back() = columns;
217 }
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                          SmallVectorImpl<int> &out) {
233   out.clear();
234 
235   if (SourceLine.empty()) {
236     out.resize(1u, 0);
237     return;
238   }
239 
240   int columns = 0;
241   size_t i = 0;
242   while (i<SourceLine.size()) {
243     out.resize(columns+1, -1);
244     out.back() = i;
245     std::pair<SmallString<16>,bool> res
246       = printableTextForNextCharacter(SourceLine, &i, TabStop);
247     columns += llvm::sys::locale::columnWidth(res.first);
248   }
249   out.resize(columns+1, -1);
250   out.back() = i;
251 }
252 
253 namespace {
254 struct SourceColumnMap {
255   SourceColumnMap(StringRef SourceLine, unsigned TabStop)
256   : m_SourceLine(SourceLine) {
257 
258     ::byteToColumn(SourceLine, TabStop, m_byteToColumn);
259     ::columnToByte(SourceLine, TabStop, m_columnToByte);
260 
261     assert(m_byteToColumn.size()==SourceLine.size()+1);
262     assert(0 < m_byteToColumn.size() && 0 < m_columnToByte.size());
263     assert(m_byteToColumn.size()
264            == static_cast<unsigned>(m_columnToByte.back()+1));
265     assert(static_cast<unsigned>(m_byteToColumn.back()+1)
266            == m_columnToByte.size());
267   }
268   int columns() const { return m_byteToColumn.back(); }
269   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   int byteToColumn(int n) const {
274     assert(0<=n && n<static_cast<int>(m_byteToColumn.size()));
275     return m_byteToColumn[n];
276   }
277 
278   /// Map a byte to the first column which contains it.
279   int byteToContainingColumn(int N) const {
280     assert(0 <= N && N < static_cast<int>(m_byteToColumn.size()));
281     while (m_byteToColumn[N] == -1)
282       --N;
283     return m_byteToColumn[N];
284   }
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   int columnToByte(int n) const {
290     assert(0<=n && n<static_cast<int>(m_columnToByte.size()));
291     return m_columnToByte[n];
292   }
293 
294   /// Map from a byte index to the next byte which starts a column.
295   int startOfNextColumn(int N) const {
296     assert(0 <= N && N < static_cast<int>(m_byteToColumn.size() - 1));
297     while (byteToColumn(++N) == -1) {}
298     return N;
299   }
300 
301   /// Map from a byte index to the previous byte which starts a column.
302   int startOfPreviousColumn(int N) const {
303     assert(0 < N && N < static_cast<int>(m_byteToColumn.size()));
304     while (byteToColumn(--N) == -1) {}
305     return N;
306   }
307 
308   StringRef getSourceLine() const {
309     return m_SourceLine;
310   }
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                                           const SourceColumnMap &map) {
326   unsigned CaretColumns = CaretLine.size();
327   unsigned FixItColumns = llvm::sys::locale::columnWidth(FixItInsertionLine);
328   unsigned MaxColumns = std::max(static_cast<unsigned>(map.columns()),
329                                  std::max(CaretColumns, FixItColumns));
330   // if the number of columns is less than the desired number we're done
331   if (MaxColumns <= Columns)
332     return;
333 
334   // No special characters are allowed in CaretLine.
335   assert(CaretLine.end() ==
336          llvm::find_if(CaretLine, [](char c) { return c < ' ' || '~' < c; }));
337 
338   // Find the slice that we need to display the full caret line
339   // correctly.
340   unsigned CaretStart = 0, CaretEnd = CaretLine.size();
341   for (; CaretStart != CaretEnd; ++CaretStart)
342     if (!isWhitespace(CaretLine[CaretStart]))
343       break;
344 
345   for (; CaretEnd != CaretStart; --CaretEnd)
346     if (!isWhitespace(CaretLine[CaretEnd - 1]))
347       break;
348 
349   // caret has already been inserted into CaretLine so the above whitespace
350   // check is guaranteed to include the caret
351 
352   // If we have a fix-it line, make sure the slice includes all of the
353   // fix-it information.
354   if (!FixItInsertionLine.empty()) {
355     unsigned FixItStart = 0, FixItEnd = FixItInsertionLine.size();
356     for (; FixItStart != FixItEnd; ++FixItStart)
357       if (!isWhitespace(FixItInsertionLine[FixItStart]))
358         break;
359 
360     for (; FixItEnd != FixItStart; --FixItEnd)
361       if (!isWhitespace(FixItInsertionLine[FixItEnd - 1]))
362         break;
363 
364     // We can safely use the byte offset FixItStart as the column offset
365     // because the characters up until FixItStart are all ASCII whitespace
366     // characters.
367     unsigned FixItStartCol = FixItStart;
368     unsigned FixItEndCol
369       = llvm::sys::locale::columnWidth(FixItInsertionLine.substr(0, FixItEnd));
370 
371     CaretStart = std::min(FixItStartCol, CaretStart);
372     CaretEnd = std::max(FixItEndCol, CaretEnd);
373   }
374 
375   // CaretEnd may have been set at the middle of a character
376   // If it's not at a character's first column then advance it past the current
377   //   character.
378   while (static_cast<int>(CaretEnd) < map.columns() &&
379          -1 == map.columnToByte(CaretEnd))
380     ++CaretEnd;
381 
382   assert((static_cast<int>(CaretStart) > map.columns() ||
383           -1!=map.columnToByte(CaretStart)) &&
384          "CaretStart must not point to a column in the middle of a source"
385          " line character");
386   assert((static_cast<int>(CaretEnd) > map.columns() ||
387           -1!=map.columnToByte(CaretEnd)) &&
388          "CaretEnd must not point to a column in the middle of a source line"
389          " character");
390 
391   // CaretLine[CaretStart, CaretEnd) contains all of the interesting
392   // parts of the caret line. While this slice is smaller than the
393   // number of columns we have, try to grow the slice to encompass
394   // more context.
395 
396   unsigned SourceStart = map.columnToByte(std::min<unsigned>(CaretStart,
397                                                              map.columns()));
398   unsigned SourceEnd = map.columnToByte(std::min<unsigned>(CaretEnd,
399                                                            map.columns()));
400 
401   unsigned CaretColumnsOutsideSource = CaretEnd-CaretStart
402     - (map.byteToColumn(SourceEnd)-map.byteToColumn(SourceStart));
403 
404   char const *front_ellipse = "  ...";
405   char const *front_space   = "     ";
406   char const *back_ellipse = "...";
407   unsigned ellipses_space = strlen(front_ellipse) + strlen(back_ellipse);
408 
409   unsigned TargetColumns = Columns;
410   // Give us extra room for the ellipses
411   //  and any of the caret line that extends past the source
412   if (TargetColumns > ellipses_space+CaretColumnsOutsideSource)
413     TargetColumns -= ellipses_space+CaretColumnsOutsideSource;
414 
415   while (SourceStart>0 || SourceEnd<SourceLine.size()) {
416     bool ExpandedRegion = false;
417 
418     if (SourceStart>0) {
419       unsigned NewStart = map.startOfPreviousColumn(SourceStart);
420 
421       // Skip over any whitespace we see here; we're looking for
422       // another bit of interesting text.
423       // FIXME: Detect non-ASCII whitespace characters too.
424       while (NewStart && isWhitespace(SourceLine[NewStart]))
425         NewStart = map.startOfPreviousColumn(NewStart);
426 
427       // Skip over this bit of "interesting" text.
428       while (NewStart) {
429         unsigned Prev = map.startOfPreviousColumn(NewStart);
430         if (isWhitespace(SourceLine[Prev]))
431           break;
432         NewStart = Prev;
433       }
434 
435       assert(map.byteToColumn(NewStart) != -1);
436       unsigned NewColumns = map.byteToColumn(SourceEnd) -
437                               map.byteToColumn(NewStart);
438       if (NewColumns <= TargetColumns) {
439         SourceStart = NewStart;
440         ExpandedRegion = true;
441       }
442     }
443 
444     if (SourceEnd<SourceLine.size()) {
445       unsigned NewEnd = map.startOfNextColumn(SourceEnd);
446 
447       // Skip over any whitespace we see here; we're looking for
448       // another bit of interesting text.
449       // FIXME: Detect non-ASCII whitespace characters too.
450       while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
451         NewEnd = map.startOfNextColumn(NewEnd);
452 
453       // Skip over this bit of "interesting" text.
454       while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
455         NewEnd = map.startOfNextColumn(NewEnd);
456 
457       assert(map.byteToColumn(NewEnd) != -1);
458       unsigned NewColumns = map.byteToColumn(NewEnd) -
459                               map.byteToColumn(SourceStart);
460       if (NewColumns <= TargetColumns) {
461         SourceEnd = NewEnd;
462         ExpandedRegion = true;
463       }
464     }
465 
466     if (!ExpandedRegion)
467       break;
468   }
469 
470   CaretStart = map.byteToColumn(SourceStart);
471   CaretEnd = map.byteToColumn(SourceEnd) + CaretColumnsOutsideSource;
472 
473   // [CaretStart, CaretEnd) is the slice we want. Update the various
474   // output lines to show only this slice, with two-space padding
475   // before the lines so that it looks nicer.
476 
477   assert(CaretStart!=(unsigned)-1 && CaretEnd!=(unsigned)-1 &&
478          SourceStart!=(unsigned)-1 && SourceEnd!=(unsigned)-1);
479   assert(SourceStart <= SourceEnd);
480   assert(CaretStart <= CaretEnd);
481 
482   unsigned BackColumnsRemoved
483     = map.byteToColumn(SourceLine.size())-map.byteToColumn(SourceEnd);
484   unsigned FrontColumnsRemoved = CaretStart;
485   unsigned ColumnsKept = CaretEnd-CaretStart;
486 
487   // We checked up front that the line needed truncation
488   assert(FrontColumnsRemoved+ColumnsKept+BackColumnsRemoved > Columns);
489 
490   // The line needs some truncation, and we'd prefer to keep the front
491   //  if possible, so remove the back
492   if (BackColumnsRemoved > strlen(back_ellipse))
493     SourceLine.replace(SourceEnd, std::string::npos, back_ellipse);
494 
495   // If that's enough then we're done
496   if (FrontColumnsRemoved+ColumnsKept <= Columns)
497     return;
498 
499   // Otherwise remove the front as well
500   if (FrontColumnsRemoved > strlen(front_ellipse)) {
501     SourceLine.replace(0, SourceStart, front_ellipse);
502     CaretLine.replace(0, CaretStart, front_space);
503     if (!FixItInsertionLine.empty())
504       FixItInsertionLine.replace(0, CaretStart, front_space);
505   }
506 }
507 
508 /// Skip over whitespace in the string, starting at the given
509 /// index.
510 ///
511 /// \returns The index of the first non-whitespace character that is
512 /// greater than or equal to Idx or, if no such character exists,
513 /// returns the end of the string.
514 static unsigned skipWhitespace(unsigned Idx, StringRef Str, unsigned Length) {
515   while (Idx < Length && isWhitespace(Str[Idx]))
516     ++Idx;
517   return Idx;
518 }
519 
520 /// If the given character is the start of some kind of
521 /// balanced punctuation (e.g., quotes or parentheses), return the
522 /// character that will terminate the punctuation.
523 ///
524 /// \returns The ending punctuation character, if any, or the NULL
525 /// character if the input character does not start any punctuation.
526 static inline char findMatchingPunctuation(char c) {
527   switch (c) {
528   case '\'': return '\'';
529   case '`': return '\'';
530   case '"':  return '"';
531   case '(':  return ')';
532   case '[': return ']';
533   case '{': return '}';
534   default: break;
535   }
536 
537   return 0;
538 }
539 
540 /// Find the end of the word starting at the given offset
541 /// within a string.
542 ///
543 /// \returns the index pointing one character past the end of the
544 /// word.
545 static unsigned findEndOfWord(unsigned Start, StringRef Str,
546                               unsigned Length, unsigned Column,
547                               unsigned Columns) {
548   assert(Start < Str.size() && "Invalid start position!");
549   unsigned End = Start + 1;
550 
551   // If we are already at the end of the string, take that as the word.
552   if (End == Str.size())
553     return End;
554 
555   // Determine if the start of the string is actually opening
556   // punctuation, e.g., a quote or parentheses.
557   char EndPunct = findMatchingPunctuation(Str[Start]);
558   if (!EndPunct) {
559     // This is a normal word. Just find the first space character.
560     while (End < Length && !isWhitespace(Str[End]))
561       ++End;
562     return End;
563   }
564 
565   // We have the start of a balanced punctuation sequence (quotes,
566   // parentheses, etc.). Determine the full sequence is.
567   SmallString<16> PunctuationEndStack;
568   PunctuationEndStack.push_back(EndPunct);
569   while (End < Length && !PunctuationEndStack.empty()) {
570     if (Str[End] == PunctuationEndStack.back())
571       PunctuationEndStack.pop_back();
572     else if (char SubEndPunct = findMatchingPunctuation(Str[End]))
573       PunctuationEndStack.push_back(SubEndPunct);
574 
575     ++End;
576   }
577 
578   // Find the first space character after the punctuation ended.
579   while (End < Length && !isWhitespace(Str[End]))
580     ++End;
581 
582   unsigned PunctWordLength = End - Start;
583   if (// If the word fits on this line
584       Column + PunctWordLength <= Columns ||
585       // ... or the word is "short enough" to take up the next line
586       // without too much ugly white space
587       PunctWordLength < Columns/3)
588     return End; // Take the whole thing as a single "word".
589 
590   // The whole quoted/parenthesized string is too long to print as a
591   // single "word". Instead, find the "word" that starts just after
592   // the punctuation and use that end-point instead. This will recurse
593   // until it finds something small enough to consider a word.
594   return findEndOfWord(Start + 1, Str, Length, Column + 1, Columns);
595 }
596 
597 /// Print the given string to a stream, word-wrapping it to
598 /// some number of columns in the process.
599 ///
600 /// \param OS the stream to which the word-wrapping string will be
601 /// emitted.
602 /// \param Str the string to word-wrap and output.
603 /// \param Columns the number of columns to word-wrap to.
604 /// \param Column the column number at which the first character of \p
605 /// Str will be printed. This will be non-zero when part of the first
606 /// line has already been printed.
607 /// \param Bold if the current text should be bold
608 /// \param Indentation the number of spaces to indent any lines beyond
609 /// the first line.
610 /// \returns true if word-wrapping was required, or false if the
611 /// string fit on the first line.
612 static bool printWordWrapped(raw_ostream &OS, StringRef Str,
613                              unsigned Columns,
614                              unsigned Column = 0,
615                              bool Bold = false,
616                              unsigned Indentation = WordWrapIndentation) {
617   const unsigned Length = std::min(Str.find('\n'), Str.size());
618   bool TextNormal = true;
619 
620   // The string used to indent each line.
621   SmallString<16> IndentStr;
622   IndentStr.assign(Indentation, ' ');
623   bool Wrapped = false;
624   for (unsigned WordStart = 0, WordEnd; WordStart < Length;
625        WordStart = WordEnd) {
626     // Find the beginning of the next word.
627     WordStart = skipWhitespace(WordStart, Str, Length);
628     if (WordStart == Length)
629       break;
630 
631     // Find the end of this word.
632     WordEnd = findEndOfWord(WordStart, Str, Length, Column, Columns);
633 
634     // Does this word fit on the current line?
635     unsigned WordLength = WordEnd - WordStart;
636     if (Column + WordLength < Columns) {
637       // This word fits on the current line; print it there.
638       if (WordStart) {
639         OS << ' ';
640         Column += 1;
641       }
642       applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
643                                 TextNormal, Bold);
644       Column += WordLength;
645       continue;
646     }
647 
648     // This word does not fit on the current line, so wrap to the next
649     // line.
650     OS << '\n';
651     OS.write(&IndentStr[0], Indentation);
652     applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
653                               TextNormal, Bold);
654     Column = Indentation + WordLength;
655     Wrapped = true;
656   }
657 
658   // Append any remaning text from the message with its existing formatting.
659   applyTemplateHighlighting(OS, Str.substr(Length), TextNormal, Bold);
660 
661   assert(TextNormal && "Text highlighted at end of diagnostic message.");
662 
663   return Wrapped;
664 }
665 
666 TextDiagnostic::TextDiagnostic(raw_ostream &OS,
667                                const LangOptions &LangOpts,
668                                DiagnosticOptions *DiagOpts)
669   : DiagnosticRenderer(LangOpts, DiagOpts), OS(OS) {}
670 
671 TextDiagnostic::~TextDiagnostic() {}
672 
673 void TextDiagnostic::emitDiagnosticMessage(
674     FullSourceLoc Loc, PresumedLoc PLoc, DiagnosticsEngine::Level Level,
675     StringRef Message, ArrayRef<clang::CharSourceRange> Ranges,
676     DiagOrStoredDiag D) {
677   uint64_t StartOfLocationInfo = OS.tell();
678 
679   // Emit the location of this particular diagnostic.
680   if (Loc.isValid())
681     emitDiagnosticLoc(Loc, PLoc, Level, Ranges);
682 
683   if (DiagOpts->ShowColors)
684     OS.resetColor();
685 
686   if (DiagOpts->ShowLevel)
687     printDiagnosticLevel(OS, Level, DiagOpts->ShowColors,
688                          DiagOpts->CLFallbackMode);
689   printDiagnosticMessage(OS,
690                          /*IsSupplemental*/ Level == DiagnosticsEngine::Note,
691                          Message, OS.tell() - StartOfLocationInfo,
692                          DiagOpts->MessageLength, DiagOpts->ShowColors);
693 }
694 
695 /*static*/ void
696 TextDiagnostic::printDiagnosticLevel(raw_ostream &OS,
697                                      DiagnosticsEngine::Level Level,
698                                      bool ShowColors,
699                                      bool CLFallbackMode) {
700   if (ShowColors) {
701     // Print diagnostic category in bold and color
702     switch (Level) {
703     case DiagnosticsEngine::Ignored:
704       llvm_unreachable("Invalid diagnostic type");
705     case DiagnosticsEngine::Note:    OS.changeColor(noteColor, true); break;
706     case DiagnosticsEngine::Remark:  OS.changeColor(remarkColor, true); break;
707     case DiagnosticsEngine::Warning: OS.changeColor(warningColor, true); break;
708     case DiagnosticsEngine::Error:   OS.changeColor(errorColor, true); break;
709     case DiagnosticsEngine::Fatal:   OS.changeColor(fatalColor, true); break;
710     }
711   }
712 
713   switch (Level) {
714   case DiagnosticsEngine::Ignored:
715     llvm_unreachable("Invalid diagnostic type");
716   case DiagnosticsEngine::Note:    OS << "note"; break;
717   case DiagnosticsEngine::Remark:  OS << "remark"; break;
718   case DiagnosticsEngine::Warning: OS << "warning"; break;
719   case DiagnosticsEngine::Error:   OS << "error"; break;
720   case DiagnosticsEngine::Fatal:   OS << "fatal error"; break;
721   }
722 
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   if (CLFallbackMode)
728     OS << "(clang)";
729 
730   OS << ": ";
731 
732   if (ShowColors)
733     OS.resetColor();
734 }
735 
736 /*static*/
737 void TextDiagnostic::printDiagnosticMessage(raw_ostream &OS,
738                                             bool IsSupplemental,
739                                             StringRef Message,
740                                             unsigned CurrentColumn,
741                                             unsigned Columns, bool ShowColors) {
742   bool Bold = false;
743   if (ShowColors && !IsSupplemental) {
744     // Print primary diagnostic messages in bold and without color, to visually
745     // indicate the transition from continuation notes and other output.
746     OS.changeColor(savedColor, true);
747     Bold = true;
748   }
749 
750   if (Columns)
751     printWordWrapped(OS, Message, Columns, CurrentColumn, Bold);
752   else {
753     bool Normal = true;
754     applyTemplateHighlighting(OS, Message, Normal, Bold);
755     assert(Normal && "Formatting should have returned to normal");
756   }
757 
758   if (ShowColors)
759     OS.resetColor();
760   OS << '\n';
761 }
762 
763 void TextDiagnostic::emitFilename(StringRef Filename, const SourceManager &SM) {
764 #ifdef _WIN32
765   SmallString<4096> TmpFilename;
766 #endif
767   if (DiagOpts->AbsolutePath) {
768     auto File = SM.getFileManager().getFile(Filename);
769     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       Filename = SM.getFileManager().getCanonicalName(*File);
792 #endif
793     }
794   }
795 
796   OS << Filename;
797 }
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                                        ArrayRef<CharSourceRange> Ranges) {
808   if (PLoc.isInvalid()) {
809     // At least print the file name if available:
810     FileID FID = Loc.getFileID();
811     if (FID.isValid()) {
812       const FileEntry *FE = Loc.getFileEntry();
813       if (FE && FE->isValid()) {
814         emitFilename(FE->getName(), Loc.getManager());
815         OS << ": ";
816       }
817     }
818     return;
819   }
820   unsigned LineNo = PLoc.getLine();
821 
822   if (!DiagOpts->ShowLocation)
823     return;
824 
825   if (DiagOpts->ShowColors)
826     OS.changeColor(savedColor, true);
827 
828   emitFilename(PLoc.getFilename(), Loc.getManager());
829   switch (DiagOpts->getFormat()) {
830   case DiagnosticOptions::Clang:
831     if (DiagOpts->ShowLine)
832       OS << ':' << LineNo;
833     break;
834   case DiagnosticOptions::MSVC:  OS << '('  << LineNo; break;
835   case DiagnosticOptions::Vi:    OS << " +" << LineNo; break;
836   }
837 
838   if (DiagOpts->ShowColumn)
839     // Compute the column number.
840     if (unsigned ColNo = PLoc.getColumn()) {
841       if (DiagOpts->getFormat() == DiagnosticOptions::MSVC) {
842         OS << ',';
843         // Visual Studio 2010 or earlier expects column number to be off by one
844         if (LangOpts.MSCompatibilityVersion &&
845             !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2012))
846           ColNo--;
847       } else
848         OS << ':';
849       OS << ColNo;
850     }
851   switch (DiagOpts->getFormat()) {
852   case DiagnosticOptions::Clang:
853   case DiagnosticOptions::Vi:    OS << ':';    break;
854   case DiagnosticOptions::MSVC:
855     // MSVC2013 and before print 'file(4) : error'. MSVC2015 gets rid of the
856     // space and prints 'file(4): error'.
857     OS << ')';
858     if (LangOpts.MSCompatibilityVersion &&
859         !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2015))
860       OS << ' ';
861     OS << ':';
862     break;
863   }
864 
865   if (DiagOpts->ShowSourceRanges && !Ranges.empty()) {
866     FileID CaretFileID = Loc.getExpansionLoc().getFileID();
867     bool PrintedRange = false;
868 
869     for (ArrayRef<CharSourceRange>::const_iterator RI = Ranges.begin(),
870          RE = Ranges.end();
871          RI != RE; ++RI) {
872       // Ignore invalid ranges.
873       if (!RI->isValid()) continue;
874 
875       auto &SM = Loc.getManager();
876       SourceLocation B = SM.getExpansionLoc(RI->getBegin());
877       CharSourceRange ERange = SM.getExpansionRange(RI->getEnd());
878       SourceLocation E = ERange.getEnd();
879       bool IsTokenRange = ERange.isTokenRange();
880 
881       std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(B);
882       std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(E);
883 
884       // If the start or end of the range is in another file, just discard
885       // it.
886       if (BInfo.first != CaretFileID || EInfo.first != CaretFileID)
887         continue;
888 
889       // Add in the length of the token, so that we cover multi-char
890       // tokens.
891       unsigned TokSize = 0;
892       if (IsTokenRange)
893         TokSize = Lexer::MeasureTokenLength(E, SM, LangOpts);
894 
895       FullSourceLoc BF(B, SM), EF(E, SM);
896       OS << '{'
897          << BF.getLineNumber() << ':' << BF.getColumnNumber() << '-'
898          << EF.getLineNumber() << ':' << (EF.getColumnNumber() + TokSize)
899          << '}';
900       PrintedRange = true;
901     }
902 
903     if (PrintedRange)
904       OS << ':';
905   }
906   OS << ' ';
907 }
908 
909 void TextDiagnostic::emitIncludeLocation(FullSourceLoc Loc, PresumedLoc PLoc) {
910   if (DiagOpts->ShowLocation && PLoc.isValid())
911     OS << "In file included from " << PLoc.getFilename() << ':'
912        << PLoc.getLine() << ":\n";
913   else
914     OS << "In included file:\n";
915 }
916 
917 void TextDiagnostic::emitImportLocation(FullSourceLoc Loc, PresumedLoc PLoc,
918                                         StringRef ModuleName) {
919   if (DiagOpts->ShowLocation && PLoc.isValid())
920     OS << "In module '" << ModuleName << "' imported from "
921        << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
922   else
923     OS << "In module '" << ModuleName << "':\n";
924 }
925 
926 void TextDiagnostic::emitBuildingModuleLocation(FullSourceLoc Loc,
927                                                 PresumedLoc PLoc,
928                                                 StringRef ModuleName) {
929   if (DiagOpts->ShowLocation && PLoc.isValid())
930     OS << "While building module '" << ModuleName << "' imported from "
931       << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
932   else
933     OS << "While building module '" << ModuleName << "':\n";
934 }
935 
936 /// Find the suitable set of lines to show to include a set of ranges.
937 static llvm::Optional<std::pair<unsigned, unsigned>>
938 findLinesForRange(const CharSourceRange &R, FileID FID,
939                   const SourceManager &SM) {
940   if (!R.isValid()) return None;
941 
942   SourceLocation Begin = R.getBegin();
943   SourceLocation End = R.getEnd();
944   if (SM.getFileID(Begin) != FID || SM.getFileID(End) != FID)
945     return None;
946 
947   return std::make_pair(SM.getExpansionLineNumber(Begin),
948                         SM.getExpansionLineNumber(End));
949 }
950 
951 /// Add as much of range B into range A as possible without exceeding a maximum
952 /// size of MaxRange. Ranges are inclusive.
953 static std::pair<unsigned, unsigned>
954 maybeAddRange(std::pair<unsigned, unsigned> A, std::pair<unsigned, unsigned> B,
955               unsigned MaxRange) {
956   // If A is already the maximum size, we're done.
957   unsigned Slack = MaxRange - (A.second - A.first + 1);
958   if (Slack == 0)
959     return A;
960 
961   // Easy case: merge succeeds within MaxRange.
962   unsigned Min = std::min(A.first, B.first);
963   unsigned Max = std::max(A.second, B.second);
964   if (Max - Min + 1 <= MaxRange)
965     return {Min, Max};
966 
967   // If we can't reach B from A within MaxRange, there's nothing to do.
968   // Don't add lines to the range that contain nothing interesting.
969   if ((B.first > A.first && B.first - A.first + 1 > MaxRange) ||
970       (B.second < A.second && A.second - B.second + 1 > MaxRange))
971     return A;
972 
973   // Otherwise, expand A towards B to produce a range of size MaxRange. We
974   // attempt to expand by the same amount in both directions if B strictly
975   // contains A.
976 
977   // Expand downwards by up to half the available amount, then upwards as
978   // much as possible, then downwards as much as possible.
979   A.second = std::min(A.second + (Slack + 1) / 2, Max);
980   Slack = MaxRange - (A.second - A.first + 1);
981   A.first = std::max(Min + Slack, A.first) - Slack;
982   A.second = std::min(A.first + MaxRange - 1, Max);
983   return A;
984 }
985 
986 /// Highlight a SourceRange (with ~'s) for any characters on LineNo.
987 static void highlightRange(const CharSourceRange &R,
988                            unsigned LineNo, FileID FID,
989                            const SourceColumnMap &map,
990                            std::string &CaretLine,
991                            const SourceManager &SM,
992                            const LangOptions &LangOpts) {
993   if (!R.isValid()) return;
994 
995   SourceLocation Begin = R.getBegin();
996   SourceLocation End = R.getEnd();
997 
998   unsigned StartLineNo = SM.getExpansionLineNumber(Begin);
999   if (StartLineNo > LineNo || SM.getFileID(Begin) != FID)
1000     return;  // No intersection.
1001 
1002   unsigned EndLineNo = SM.getExpansionLineNumber(End);
1003   if (EndLineNo < LineNo || SM.getFileID(End) != FID)
1004     return;  // No intersection.
1005 
1006   // Compute the column number of the start.
1007   unsigned StartColNo = 0;
1008   if (StartLineNo == LineNo) {
1009     StartColNo = SM.getExpansionColumnNumber(Begin);
1010     if (StartColNo) --StartColNo;  // Zero base the col #.
1011   }
1012 
1013   // Compute the column number of the end.
1014   unsigned EndColNo = map.getSourceLine().size();
1015   if (EndLineNo == LineNo) {
1016     EndColNo = SM.getExpansionColumnNumber(End);
1017     if (EndColNo) {
1018       --EndColNo;  // Zero base the col #.
1019 
1020       // Add in the length of the token, so that we cover multi-char tokens if
1021       // this is a token range.
1022       if (R.isTokenRange())
1023         EndColNo += Lexer::MeasureTokenLength(End, SM, LangOpts);
1024     } else {
1025       EndColNo = CaretLine.size();
1026     }
1027   }
1028 
1029   assert(StartColNo <= EndColNo && "Invalid range!");
1030 
1031   // Check that a token range does not highlight only whitespace.
1032   if (R.isTokenRange()) {
1033     // Pick the first non-whitespace column.
1034     while (StartColNo < map.getSourceLine().size() &&
1035            (map.getSourceLine()[StartColNo] == ' ' ||
1036             map.getSourceLine()[StartColNo] == '\t'))
1037       StartColNo = map.startOfNextColumn(StartColNo);
1038 
1039     // Pick the last non-whitespace column.
1040     if (EndColNo > map.getSourceLine().size())
1041       EndColNo = map.getSourceLine().size();
1042     while (EndColNo &&
1043            (map.getSourceLine()[EndColNo-1] == ' ' ||
1044             map.getSourceLine()[EndColNo-1] == '\t'))
1045       EndColNo = map.startOfPreviousColumn(EndColNo);
1046 
1047     // If the start/end passed each other, then we are trying to highlight a
1048     // range that just exists in whitespace. That most likely means we have
1049     // a multi-line highlighting range that covers a blank line.
1050     if (StartColNo > EndColNo) {
1051       assert(StartLineNo != EndLineNo && "trying to highlight whitespace");
1052       StartColNo = EndColNo;
1053     }
1054   }
1055 
1056   assert(StartColNo <= map.getSourceLine().size() && "Invalid range!");
1057   assert(EndColNo <= map.getSourceLine().size() && "Invalid range!");
1058 
1059   // Fill the range with ~'s.
1060   StartColNo = map.byteToContainingColumn(StartColNo);
1061   EndColNo = map.byteToContainingColumn(EndColNo);
1062 
1063   assert(StartColNo <= EndColNo && "Invalid range!");
1064   if (CaretLine.size() < EndColNo)
1065     CaretLine.resize(EndColNo,' ');
1066   std::fill(CaretLine.begin()+StartColNo,CaretLine.begin()+EndColNo,'~');
1067 }
1068 
1069 static std::string buildFixItInsertionLine(FileID FID,
1070                                            unsigned LineNo,
1071                                            const SourceColumnMap &map,
1072                                            ArrayRef<FixItHint> Hints,
1073                                            const SourceManager &SM,
1074                                            const DiagnosticOptions *DiagOpts) {
1075   std::string FixItInsertionLine;
1076   if (Hints.empty() || !DiagOpts->ShowFixits)
1077     return FixItInsertionLine;
1078   unsigned PrevHintEndCol = 0;
1079 
1080   for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1081        I != E; ++I) {
1082     if (!I->CodeToInsert.empty()) {
1083       // We have an insertion hint. Determine whether the inserted
1084       // code contains no newlines and is on the same line as the caret.
1085       std::pair<FileID, unsigned> HintLocInfo
1086         = SM.getDecomposedExpansionLoc(I->RemoveRange.getBegin());
1087       if (FID == HintLocInfo.first &&
1088           LineNo == SM.getLineNumber(HintLocInfo.first, HintLocInfo.second) &&
1089           StringRef(I->CodeToInsert).find_first_of("\n\r") == StringRef::npos) {
1090         // Insert the new code into the line just below the code
1091         // that the user wrote.
1092         // Note: When modifying this function, be very careful about what is a
1093         // "column" (printed width, platform-dependent) and what is a
1094         // "byte offset" (SourceManager "column").
1095         unsigned HintByteOffset
1096           = SM.getColumnNumber(HintLocInfo.first, HintLocInfo.second) - 1;
1097 
1098         // The hint must start inside the source or right at the end
1099         assert(HintByteOffset < static_cast<unsigned>(map.bytes())+1);
1100         unsigned HintCol = map.byteToContainingColumn(HintByteOffset);
1101 
1102         // If we inserted a long previous hint, push this one forwards, and add
1103         // an extra space to show that this is not part of the previous
1104         // completion. This is sort of the best we can do when two hints appear
1105         // to overlap.
1106         //
1107         // Note that if this hint is located immediately after the previous
1108         // hint, no space will be added, since the location is more important.
1109         if (HintCol < PrevHintEndCol)
1110           HintCol = PrevHintEndCol + 1;
1111 
1112         // This should NOT use HintByteOffset, because the source might have
1113         // Unicode characters in earlier columns.
1114         unsigned NewFixItLineSize = FixItInsertionLine.size() +
1115           (HintCol - PrevHintEndCol) + I->CodeToInsert.size();
1116         if (NewFixItLineSize > FixItInsertionLine.size())
1117           FixItInsertionLine.resize(NewFixItLineSize, ' ');
1118 
1119         std::copy(I->CodeToInsert.begin(), I->CodeToInsert.end(),
1120                   FixItInsertionLine.end() - I->CodeToInsert.size());
1121 
1122         PrevHintEndCol =
1123           HintCol + llvm::sys::locale::columnWidth(I->CodeToInsert);
1124       }
1125     }
1126   }
1127 
1128   expandTabs(FixItInsertionLine, DiagOpts->TabStop);
1129 
1130   return FixItInsertionLine;
1131 }
1132 
1133 /// Emit a code snippet and caret line.
1134 ///
1135 /// This routine emits a single line's code snippet and caret line..
1136 ///
1137 /// \param Loc The location for the caret.
1138 /// \param Ranges The underlined ranges for this code snippet.
1139 /// \param Hints The FixIt hints active for this diagnostic.
1140 void TextDiagnostic::emitSnippetAndCaret(
1141     FullSourceLoc Loc, DiagnosticsEngine::Level Level,
1142     SmallVectorImpl<CharSourceRange> &Ranges, ArrayRef<FixItHint> Hints) {
1143   assert(Loc.isValid() && "must have a valid source location here");
1144   assert(Loc.isFileID() && "must have a file location here");
1145 
1146   // If caret diagnostics are enabled and we have location, we want to
1147   // emit the caret.  However, we only do this if the location moved
1148   // from the last diagnostic, if the last diagnostic was a note that
1149   // was part of a different warning or error diagnostic, or if the
1150   // diagnostic has ranges.  We don't want to emit the same caret
1151   // multiple times if one loc has multiple diagnostics.
1152   if (!DiagOpts->ShowCarets)
1153     return;
1154   if (Loc == LastLoc && Ranges.empty() && Hints.empty() &&
1155       (LastLevel != DiagnosticsEngine::Note || Level == LastLevel))
1156     return;
1157 
1158   // Decompose the location into a FID/Offset pair.
1159   std::pair<FileID, unsigned> LocInfo = Loc.getDecomposedLoc();
1160   FileID FID = LocInfo.first;
1161   const SourceManager &SM = Loc.getManager();
1162 
1163   // Get information about the buffer it points into.
1164   bool Invalid = false;
1165   StringRef BufData = Loc.getBufferData(&Invalid);
1166   if (Invalid)
1167     return;
1168 
1169   unsigned CaretLineNo = Loc.getLineNumber();
1170   unsigned CaretColNo = Loc.getColumnNumber();
1171 
1172   // Arbitrarily stop showing snippets when the line is too long.
1173   static const size_t MaxLineLengthToPrint = 4096;
1174   if (CaretColNo > MaxLineLengthToPrint)
1175     return;
1176 
1177   // Find the set of lines to include.
1178   const unsigned MaxLines = DiagOpts->SnippetLineLimit;
1179   std::pair<unsigned, unsigned> Lines = {CaretLineNo, CaretLineNo};
1180   for (SmallVectorImpl<CharSourceRange>::iterator I = Ranges.begin(),
1181                                                   E = Ranges.end();
1182        I != E; ++I)
1183     if (auto OptionalRange = findLinesForRange(*I, FID, SM))
1184       Lines = maybeAddRange(Lines, *OptionalRange, MaxLines);
1185 
1186   for (unsigned LineNo = Lines.first; LineNo != Lines.second + 1; ++LineNo) {
1187     const char *BufStart = BufData.data();
1188     const char *BufEnd = BufStart + BufData.size();
1189 
1190     // Rewind from the current position to the start of the line.
1191     const char *LineStart =
1192         BufStart +
1193         SM.getDecomposedLoc(SM.translateLineCol(FID, LineNo, 1)).second;
1194     if (LineStart == BufEnd)
1195       break;
1196 
1197     // Compute the line end.
1198     const char *LineEnd = LineStart;
1199     while (*LineEnd != '\n' && *LineEnd != '\r' && LineEnd != BufEnd)
1200       ++LineEnd;
1201 
1202     // Arbitrarily stop showing snippets when the line is too long.
1203     // FIXME: Don't print any lines in this case.
1204     if (size_t(LineEnd - LineStart) > MaxLineLengthToPrint)
1205       return;
1206 
1207     // Trim trailing null-bytes.
1208     StringRef Line(LineStart, LineEnd - LineStart);
1209     while (!Line.empty() && Line.back() == '\0' &&
1210            (LineNo != CaretLineNo || Line.size() > CaretColNo))
1211       Line = Line.drop_back();
1212 
1213     // Copy the line of code into an std::string for ease of manipulation.
1214     std::string SourceLine(Line.begin(), Line.end());
1215 
1216     // Build the byte to column map.
1217     const SourceColumnMap sourceColMap(SourceLine, DiagOpts->TabStop);
1218 
1219     // Create a line for the caret that is filled with spaces that is the same
1220     // number of columns as the line of source code.
1221     std::string CaretLine(sourceColMap.columns(), ' ');
1222 
1223     // Highlight all of the characters covered by Ranges with ~ characters.
1224     for (SmallVectorImpl<CharSourceRange>::iterator I = Ranges.begin(),
1225                                                     E = Ranges.end();
1226          I != E; ++I)
1227       highlightRange(*I, LineNo, FID, sourceColMap, CaretLine, SM, LangOpts);
1228 
1229     // Next, insert the caret itself.
1230     if (CaretLineNo == LineNo) {
1231       CaretColNo = sourceColMap.byteToContainingColumn(CaretColNo - 1);
1232       if (CaretLine.size() < CaretColNo + 1)
1233         CaretLine.resize(CaretColNo + 1, ' ');
1234       CaretLine[CaretColNo] = '^';
1235     }
1236 
1237     std::string FixItInsertionLine = buildFixItInsertionLine(
1238         FID, LineNo, sourceColMap, Hints, SM, DiagOpts.get());
1239 
1240     // If the source line is too long for our terminal, select only the
1241     // "interesting" source region within that line.
1242     unsigned Columns = DiagOpts->MessageLength;
1243     if (Columns)
1244       selectInterestingSourceRegion(SourceLine, CaretLine, FixItInsertionLine,
1245                                     Columns, sourceColMap);
1246 
1247     // If we are in -fdiagnostics-print-source-range-info mode, we are trying
1248     // to produce easily machine parsable output.  Add a space before the
1249     // source line and the caret to make it trivial to tell the main diagnostic
1250     // line from what the user is intended to see.
1251     if (DiagOpts->ShowSourceRanges) {
1252       SourceLine = ' ' + SourceLine;
1253       CaretLine = ' ' + CaretLine;
1254     }
1255 
1256     // Finally, remove any blank spaces from the end of CaretLine.
1257     while (!CaretLine.empty() && CaretLine[CaretLine.size() - 1] == ' ')
1258       CaretLine.erase(CaretLine.end() - 1);
1259 
1260     // Emit what we have computed.
1261     emitSnippet(SourceLine);
1262 
1263     if (!CaretLine.empty()) {
1264       if (DiagOpts->ShowColors)
1265         OS.changeColor(caretColor, true);
1266       OS << CaretLine << '\n';
1267       if (DiagOpts->ShowColors)
1268         OS.resetColor();
1269     }
1270 
1271     if (!FixItInsertionLine.empty()) {
1272       if (DiagOpts->ShowColors)
1273         // Print fixit line in color
1274         OS.changeColor(fixitColor, false);
1275       if (DiagOpts->ShowSourceRanges)
1276         OS << ' ';
1277       OS << FixItInsertionLine << '\n';
1278       if (DiagOpts->ShowColors)
1279         OS.resetColor();
1280     }
1281   }
1282 
1283   // Print out any parseable fixit information requested by the options.
1284   emitParseableFixits(Hints, SM);
1285 }
1286 
1287 void TextDiagnostic::emitSnippet(StringRef line) {
1288   if (line.empty())
1289     return;
1290 
1291   size_t i = 0;
1292 
1293   std::string to_print;
1294   bool print_reversed = false;
1295 
1296   while (i<line.size()) {
1297     std::pair<SmallString<16>,bool> res
1298         = printableTextForNextCharacter(line, &i, DiagOpts->TabStop);
1299     bool was_printable = res.second;
1300 
1301     if (DiagOpts->ShowColors && was_printable == print_reversed) {
1302       if (print_reversed)
1303         OS.reverseColor();
1304       OS << to_print;
1305       to_print.clear();
1306       if (DiagOpts->ShowColors)
1307         OS.resetColor();
1308     }
1309 
1310     print_reversed = !was_printable;
1311     to_print += res.first.str();
1312   }
1313 
1314   if (print_reversed && DiagOpts->ShowColors)
1315     OS.reverseColor();
1316   OS << to_print;
1317   if (print_reversed && DiagOpts->ShowColors)
1318     OS.resetColor();
1319 
1320   OS << '\n';
1321 }
1322 
1323 void TextDiagnostic::emitParseableFixits(ArrayRef<FixItHint> Hints,
1324                                          const SourceManager &SM) {
1325   if (!DiagOpts->ShowParseableFixits)
1326     return;
1327 
1328   // We follow FixItRewriter's example in not (yet) handling
1329   // fix-its in macros.
1330   for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1331        I != E; ++I) {
1332     if (I->RemoveRange.isInvalid() ||
1333         I->RemoveRange.getBegin().isMacroID() ||
1334         I->RemoveRange.getEnd().isMacroID())
1335       return;
1336   }
1337 
1338   for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1339        I != E; ++I) {
1340     SourceLocation BLoc = I->RemoveRange.getBegin();
1341     SourceLocation ELoc = I->RemoveRange.getEnd();
1342 
1343     std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(BLoc);
1344     std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(ELoc);
1345 
1346     // Adjust for token ranges.
1347     if (I->RemoveRange.isTokenRange())
1348       EInfo.second += Lexer::MeasureTokenLength(ELoc, SM, LangOpts);
1349 
1350     // We specifically do not do word-wrapping or tab-expansion here,
1351     // because this is supposed to be easy to parse.
1352     PresumedLoc PLoc = SM.getPresumedLoc(BLoc);
1353     if (PLoc.isInvalid())
1354       break;
1355 
1356     OS << "fix-it:\"";
1357     OS.write_escaped(PLoc.getFilename());
1358     OS << "\":{" << SM.getLineNumber(BInfo.first, BInfo.second)
1359       << ':' << SM.getColumnNumber(BInfo.first, BInfo.second)
1360       << '-' << SM.getLineNumber(EInfo.first, EInfo.second)
1361       << ':' << SM.getColumnNumber(EInfo.first, EInfo.second)
1362       << "}:\"";
1363     OS.write_escaped(I->CodeToInsert);
1364     OS << "\"\n";
1365   }
1366 }
1367