1 //===--- TextDiagnostic.cpp - Text Diagnostic Pretty-Printing -------------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 
10 #include "clang/Frontend/TextDiagnostic.h"
11 #include "clang/Basic/CharInfo.h"
12 #include "clang/Basic/DiagnosticOptions.h"
13 #include "clang/Basic/FileManager.h"
14 #include "clang/Basic/SourceManager.h"
15 #include "clang/Lex/Lexer.h"
16 #include "llvm/ADT/SmallString.h"
17 #include "llvm/ADT/StringExtras.h"
18 #include "llvm/Support/ConvertUTF.h"
19 #include "llvm/Support/ErrorHandling.h"
20 #include "llvm/Support/Locale.h"
21 #include "llvm/Support/MemoryBuffer.h"
22 #include "llvm/Support/raw_ostream.h"
23 #include <algorithm>
24 
25 using namespace clang;
26 
27 static const enum raw_ostream::Colors noteColor =
28   raw_ostream::BLACK;
29 static const enum raw_ostream::Colors remarkColor =
30   raw_ostream::BLUE;
31 static const enum raw_ostream::Colors fixitColor =
32   raw_ostream::GREEN;
33 static const enum raw_ostream::Colors caretColor =
34   raw_ostream::GREEN;
35 static const enum raw_ostream::Colors warningColor =
36   raw_ostream::MAGENTA;
37 static const enum raw_ostream::Colors templateColor =
38   raw_ostream::CYAN;
39 static const enum raw_ostream::Colors errorColor = raw_ostream::RED;
40 static const enum raw_ostream::Colors fatalColor = raw_ostream::RED;
41 // Used for changing only the bold attribute.
42 static const enum raw_ostream::Colors savedColor =
43   raw_ostream::SAVEDCOLOR;
44 
45 /// \brief Add highlights to differences in template strings.
applyTemplateHighlighting(raw_ostream & OS,StringRef Str,bool & Normal,bool Bold)46 static void applyTemplateHighlighting(raw_ostream &OS, StringRef Str,
47                                       bool &Normal, bool Bold) {
48   while (1) {
49     size_t Pos = Str.find(ToggleHighlight);
50     OS << Str.slice(0, Pos);
51     if (Pos == StringRef::npos)
52       break;
53 
54     Str = Str.substr(Pos + 1);
55     if (Normal)
56       OS.changeColor(templateColor, true);
57     else {
58       OS.resetColor();
59       if (Bold)
60         OS.changeColor(savedColor, true);
61     }
62     Normal = !Normal;
63   }
64 }
65 
66 /// \brief Number of spaces to indent when word-wrapping.
67 const unsigned WordWrapIndentation = 6;
68 
bytesSincePreviousTabOrLineBegin(StringRef SourceLine,size_t i)69 static int bytesSincePreviousTabOrLineBegin(StringRef SourceLine, size_t i) {
70   int bytes = 0;
71   while (0<i) {
72     if (SourceLine[--i]=='\t')
73       break;
74     ++bytes;
75   }
76   return bytes;
77 }
78 
79 /// \brief returns a printable representation of first item from input range
80 ///
81 /// This function returns a printable representation of the next item in a line
82 ///  of source. If the next byte begins a valid and printable character, that
83 ///  character is returned along with 'true'.
84 ///
85 /// Otherwise, if the next byte begins a valid, but unprintable character, a
86 ///  printable, escaped representation of the character is returned, along with
87 ///  'false'. Otherwise a printable, escaped representation of the next byte
88 ///  is returned along with 'false'.
89 ///
90 /// \note The index is updated to be used with a subsequent call to
91 ///        printableTextForNextCharacter.
92 ///
93 /// \param SourceLine The line of source
94 /// \param i Pointer to byte index,
95 /// \param TabStop used to expand tabs
96 /// \return pair(printable text, 'true' iff original text was printable)
97 ///
98 static std::pair<SmallString<16>, bool>
printableTextForNextCharacter(StringRef SourceLine,size_t * i,unsigned TabStop)99 printableTextForNextCharacter(StringRef SourceLine, size_t *i,
100                               unsigned TabStop) {
101   assert(i && "i must not be null");
102   assert(*i<SourceLine.size() && "must point to a valid index");
103 
104   if (SourceLine[*i]=='\t') {
105     assert(0 < TabStop && TabStop <= DiagnosticOptions::MaxTabStop &&
106            "Invalid -ftabstop value");
107     unsigned col = bytesSincePreviousTabOrLineBegin(SourceLine, *i);
108     unsigned NumSpaces = TabStop - col%TabStop;
109     assert(0 < NumSpaces && NumSpaces <= TabStop
110            && "Invalid computation of space amt");
111     ++(*i);
112 
113     SmallString<16> expandedTab;
114     expandedTab.assign(NumSpaces, ' ');
115     return std::make_pair(expandedTab, true);
116   }
117 
118   unsigned char const *begin, *end;
119   begin = reinterpret_cast<unsigned char const *>(&*(SourceLine.begin() + *i));
120   end = begin + (SourceLine.size() - *i);
121 
122   if (isLegalUTF8Sequence(begin, end)) {
123     UTF32 c;
124     UTF32 *cptr = &c;
125     unsigned char const *original_begin = begin;
126     unsigned char const *cp_end = begin+getNumBytesForUTF8(SourceLine[*i]);
127 
128     ConversionResult res = ConvertUTF8toUTF32(&begin, cp_end, &cptr, cptr+1,
129                                               strictConversion);
130     (void)res;
131     assert(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 
expandTabs(std::string & SourceLine,unsigned TabStop)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)
byteToColumn(StringRef SourceLine,unsigned TabStop,SmallVectorImpl<int> & out)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)
columnToByte(StringRef SourceLine,unsigned TabStop,SmallVectorImpl<int> & out)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 {
SourceColumnMap__anon7709ae750111::SourceColumnMap255   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   }
columns__anon7709ae750111::SourceColumnMap268   int columns() const { return m_byteToColumn.back(); }
bytes__anon7709ae750111::SourceColumnMap269   int bytes() const { return m_columnToByte.back(); }
270 
271   /// \brief 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).
byteToColumn__anon7709ae750111::SourceColumnMap273   int byteToColumn(int n) const {
274     assert(0<=n && n<static_cast<int>(m_byteToColumn.size()));
275     return m_byteToColumn[n];
276   }
277 
278   /// \brief Map a byte to the first column which contains it.
byteToContainingColumn__anon7709ae750111::SourceColumnMap279   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   /// \brief 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.
columnToByte__anon7709ae750111::SourceColumnMap289   int columnToByte(int n) const {
290     assert(0<=n && n<static_cast<int>(m_columnToByte.size()));
291     return m_columnToByte[n];
292   }
293 
294   /// \brief Map from a byte index to the next byte which starts a column.
startOfNextColumn__anon7709ae750111::SourceColumnMap295   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   /// \brief Map from a byte index to the previous byte which starts a column.
startOfPreviousColumn__anon7709ae750111::SourceColumnMap302   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 
getSourceLine__anon7709ae750111::SourceColumnMap308   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 /// \brief When the source code line we want to print is too long for
320 /// the terminal, select the "interesting" region.
selectInterestingSourceRegion(std::string & SourceLine,std::string & CaretLine,std::string & FixItInsertionLine,unsigned Columns,const SourceColumnMap & map)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          std::find_if(CaretLine.begin(), CaretLine.end(),
337                       [](char c) { return c < ' ' || '~' < c; }));
338 
339   // Find the slice that we need to display the full caret line
340   // correctly.
341   unsigned CaretStart = 0, CaretEnd = CaretLine.size();
342   for (; CaretStart != CaretEnd; ++CaretStart)
343     if (!isWhitespace(CaretLine[CaretStart]))
344       break;
345 
346   for (; CaretEnd != CaretStart; --CaretEnd)
347     if (!isWhitespace(CaretLine[CaretEnd - 1]))
348       break;
349 
350   // caret has already been inserted into CaretLine so the above whitespace
351   // check is guaranteed to include the caret
352 
353   // If we have a fix-it line, make sure the slice includes all of the
354   // fix-it information.
355   if (!FixItInsertionLine.empty()) {
356     unsigned FixItStart = 0, FixItEnd = FixItInsertionLine.size();
357     for (; FixItStart != FixItEnd; ++FixItStart)
358       if (!isWhitespace(FixItInsertionLine[FixItStart]))
359         break;
360 
361     for (; FixItEnd != FixItStart; --FixItEnd)
362       if (!isWhitespace(FixItInsertionLine[FixItEnd - 1]))
363         break;
364 
365     // We can safely use the byte offset FixItStart as the column offset
366     // because the characters up until FixItStart are all ASCII whitespace
367     // characters.
368     unsigned FixItStartCol = FixItStart;
369     unsigned FixItEndCol
370       = llvm::sys::locale::columnWidth(FixItInsertionLine.substr(0, FixItEnd));
371 
372     CaretStart = std::min(FixItStartCol, CaretStart);
373     CaretEnd = std::max(FixItEndCol, CaretEnd);
374   }
375 
376   // CaretEnd may have been set at the middle of a character
377   // If it's not at a character's first column then advance it past the current
378   //   character.
379   while (static_cast<int>(CaretEnd) < map.columns() &&
380          -1 == map.columnToByte(CaretEnd))
381     ++CaretEnd;
382 
383   assert((static_cast<int>(CaretStart) > map.columns() ||
384           -1!=map.columnToByte(CaretStart)) &&
385          "CaretStart must not point to a column in the middle of a source"
386          " line character");
387   assert((static_cast<int>(CaretEnd) > map.columns() ||
388           -1!=map.columnToByte(CaretEnd)) &&
389          "CaretEnd must not point to a column in the middle of a source line"
390          " character");
391 
392   // CaretLine[CaretStart, CaretEnd) contains all of the interesting
393   // parts of the caret line. While this slice is smaller than the
394   // number of columns we have, try to grow the slice to encompass
395   // more context.
396 
397   unsigned SourceStart = map.columnToByte(std::min<unsigned>(CaretStart,
398                                                              map.columns()));
399   unsigned SourceEnd = map.columnToByte(std::min<unsigned>(CaretEnd,
400                                                            map.columns()));
401 
402   unsigned CaretColumnsOutsideSource = CaretEnd-CaretStart
403     - (map.byteToColumn(SourceEnd)-map.byteToColumn(SourceStart));
404 
405   char const *front_ellipse = "  ...";
406   char const *front_space   = "     ";
407   char const *back_ellipse = "...";
408   unsigned ellipses_space = strlen(front_ellipse) + strlen(back_ellipse);
409 
410   unsigned TargetColumns = Columns;
411   // Give us extra room for the ellipses
412   //  and any of the caret line that extends past the source
413   if (TargetColumns > ellipses_space+CaretColumnsOutsideSource)
414     TargetColumns -= ellipses_space+CaretColumnsOutsideSource;
415 
416   while (SourceStart>0 || SourceEnd<SourceLine.size()) {
417     bool ExpandedRegion = false;
418 
419     if (SourceStart>0) {
420       unsigned NewStart = map.startOfPreviousColumn(SourceStart);
421 
422       // Skip over any whitespace we see here; we're looking for
423       // another bit of interesting text.
424       // FIXME: Detect non-ASCII whitespace characters too.
425       while (NewStart && isWhitespace(SourceLine[NewStart]))
426         NewStart = map.startOfPreviousColumn(NewStart);
427 
428       // Skip over this bit of "interesting" text.
429       while (NewStart) {
430         unsigned Prev = map.startOfPreviousColumn(NewStart);
431         if (isWhitespace(SourceLine[Prev]))
432           break;
433         NewStart = Prev;
434       }
435 
436       assert(map.byteToColumn(NewStart) != -1);
437       unsigned NewColumns = map.byteToColumn(SourceEnd) -
438                               map.byteToColumn(NewStart);
439       if (NewColumns <= TargetColumns) {
440         SourceStart = NewStart;
441         ExpandedRegion = true;
442       }
443     }
444 
445     if (SourceEnd<SourceLine.size()) {
446       unsigned NewEnd = map.startOfNextColumn(SourceEnd);
447 
448       // Skip over any whitespace we see here; we're looking for
449       // another bit of interesting text.
450       // FIXME: Detect non-ASCII whitespace characters too.
451       while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
452         NewEnd = map.startOfNextColumn(NewEnd);
453 
454       // Skip over this bit of "interesting" text.
455       while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
456         NewEnd = map.startOfNextColumn(NewEnd);
457 
458       assert(map.byteToColumn(NewEnd) != -1);
459       unsigned NewColumns = map.byteToColumn(NewEnd) -
460                               map.byteToColumn(SourceStart);
461       if (NewColumns <= TargetColumns) {
462         SourceEnd = NewEnd;
463         ExpandedRegion = true;
464       }
465     }
466 
467     if (!ExpandedRegion)
468       break;
469   }
470 
471   CaretStart = map.byteToColumn(SourceStart);
472   CaretEnd = map.byteToColumn(SourceEnd) + CaretColumnsOutsideSource;
473 
474   // [CaretStart, CaretEnd) is the slice we want. Update the various
475   // output lines to show only this slice, with two-space padding
476   // before the lines so that it looks nicer.
477 
478   assert(CaretStart!=(unsigned)-1 && CaretEnd!=(unsigned)-1 &&
479          SourceStart!=(unsigned)-1 && SourceEnd!=(unsigned)-1);
480   assert(SourceStart <= SourceEnd);
481   assert(CaretStart <= CaretEnd);
482 
483   unsigned BackColumnsRemoved
484     = map.byteToColumn(SourceLine.size())-map.byteToColumn(SourceEnd);
485   unsigned FrontColumnsRemoved = CaretStart;
486   unsigned ColumnsKept = CaretEnd-CaretStart;
487 
488   // We checked up front that the line needed truncation
489   assert(FrontColumnsRemoved+ColumnsKept+BackColumnsRemoved > Columns);
490 
491   // The line needs some truncation, and we'd prefer to keep the front
492   //  if possible, so remove the back
493   if (BackColumnsRemoved > strlen(back_ellipse))
494     SourceLine.replace(SourceEnd, std::string::npos, back_ellipse);
495 
496   // If that's enough then we're done
497   if (FrontColumnsRemoved+ColumnsKept <= Columns)
498     return;
499 
500   // Otherwise remove the front as well
501   if (FrontColumnsRemoved > strlen(front_ellipse)) {
502     SourceLine.replace(0, SourceStart, front_ellipse);
503     CaretLine.replace(0, CaretStart, front_space);
504     if (!FixItInsertionLine.empty())
505       FixItInsertionLine.replace(0, CaretStart, front_space);
506   }
507 }
508 
509 /// \brief Skip over whitespace in the string, starting at the given
510 /// index.
511 ///
512 /// \returns The index of the first non-whitespace character that is
513 /// greater than or equal to Idx or, if no such character exists,
514 /// returns the end of the string.
skipWhitespace(unsigned Idx,StringRef Str,unsigned Length)515 static unsigned skipWhitespace(unsigned Idx, StringRef Str, unsigned Length) {
516   while (Idx < Length && isWhitespace(Str[Idx]))
517     ++Idx;
518   return Idx;
519 }
520 
521 /// \brief If the given character is the start of some kind of
522 /// balanced punctuation (e.g., quotes or parentheses), return the
523 /// character that will terminate the punctuation.
524 ///
525 /// \returns The ending punctuation character, if any, or the NULL
526 /// character if the input character does not start any punctuation.
findMatchingPunctuation(char c)527 static inline char findMatchingPunctuation(char c) {
528   switch (c) {
529   case '\'': return '\'';
530   case '`': return '\'';
531   case '"':  return '"';
532   case '(':  return ')';
533   case '[': return ']';
534   case '{': return '}';
535   default: break;
536   }
537 
538   return 0;
539 }
540 
541 /// \brief Find the end of the word starting at the given offset
542 /// within a string.
543 ///
544 /// \returns the index pointing one character past the end of the
545 /// word.
findEndOfWord(unsigned Start,StringRef Str,unsigned Length,unsigned Column,unsigned Columns)546 static unsigned findEndOfWord(unsigned Start, StringRef Str,
547                               unsigned Length, unsigned Column,
548                               unsigned Columns) {
549   assert(Start < Str.size() && "Invalid start position!");
550   unsigned End = Start + 1;
551 
552   // If we are already at the end of the string, take that as the word.
553   if (End == Str.size())
554     return End;
555 
556   // Determine if the start of the string is actually opening
557   // punctuation, e.g., a quote or parentheses.
558   char EndPunct = findMatchingPunctuation(Str[Start]);
559   if (!EndPunct) {
560     // This is a normal word. Just find the first space character.
561     while (End < Length && !isWhitespace(Str[End]))
562       ++End;
563     return End;
564   }
565 
566   // We have the start of a balanced punctuation sequence (quotes,
567   // parentheses, etc.). Determine the full sequence is.
568   SmallString<16> PunctuationEndStack;
569   PunctuationEndStack.push_back(EndPunct);
570   while (End < Length && !PunctuationEndStack.empty()) {
571     if (Str[End] == PunctuationEndStack.back())
572       PunctuationEndStack.pop_back();
573     else if (char SubEndPunct = findMatchingPunctuation(Str[End]))
574       PunctuationEndStack.push_back(SubEndPunct);
575 
576     ++End;
577   }
578 
579   // Find the first space character after the punctuation ended.
580   while (End < Length && !isWhitespace(Str[End]))
581     ++End;
582 
583   unsigned PunctWordLength = End - Start;
584   if (// If the word fits on this line
585       Column + PunctWordLength <= Columns ||
586       // ... or the word is "short enough" to take up the next line
587       // without too much ugly white space
588       PunctWordLength < Columns/3)
589     return End; // Take the whole thing as a single "word".
590 
591   // The whole quoted/parenthesized string is too long to print as a
592   // single "word". Instead, find the "word" that starts just after
593   // the punctuation and use that end-point instead. This will recurse
594   // until it finds something small enough to consider a word.
595   return findEndOfWord(Start + 1, Str, Length, Column + 1, Columns);
596 }
597 
598 /// \brief Print the given string to a stream, word-wrapping it to
599 /// some number of columns in the process.
600 ///
601 /// \param OS the stream to which the word-wrapping string will be
602 /// emitted.
603 /// \param Str the string to word-wrap and output.
604 /// \param Columns the number of columns to word-wrap to.
605 /// \param Column the column number at which the first character of \p
606 /// Str will be printed. This will be non-zero when part of the first
607 /// line has already been printed.
608 /// \param Bold if the current text should be bold
609 /// \param Indentation the number of spaces to indent any lines beyond
610 /// the first line.
611 /// \returns true if word-wrapping was required, or false if the
612 /// string fit on the first line.
printWordWrapped(raw_ostream & OS,StringRef Str,unsigned Columns,unsigned Column=0,bool Bold=false,unsigned Indentation=WordWrapIndentation)613 static bool printWordWrapped(raw_ostream &OS, StringRef Str,
614                              unsigned Columns,
615                              unsigned Column = 0,
616                              bool Bold = false,
617                              unsigned Indentation = WordWrapIndentation) {
618   const unsigned Length = std::min(Str.find('\n'), Str.size());
619   bool TextNormal = true;
620 
621   // The string used to indent each line.
622   SmallString<16> IndentStr;
623   IndentStr.assign(Indentation, ' ');
624   bool Wrapped = false;
625   for (unsigned WordStart = 0, WordEnd; WordStart < Length;
626        WordStart = WordEnd) {
627     // Find the beginning of the next word.
628     WordStart = skipWhitespace(WordStart, Str, Length);
629     if (WordStart == Length)
630       break;
631 
632     // Find the end of this word.
633     WordEnd = findEndOfWord(WordStart, Str, Length, Column, Columns);
634 
635     // Does this word fit on the current line?
636     unsigned WordLength = WordEnd - WordStart;
637     if (Column + WordLength < Columns) {
638       // This word fits on the current line; print it there.
639       if (WordStart) {
640         OS << ' ';
641         Column += 1;
642       }
643       applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
644                                 TextNormal, Bold);
645       Column += WordLength;
646       continue;
647     }
648 
649     // This word does not fit on the current line, so wrap to the next
650     // line.
651     OS << '\n';
652     OS.write(&IndentStr[0], Indentation);
653     applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
654                               TextNormal, Bold);
655     Column = Indentation + WordLength;
656     Wrapped = true;
657   }
658 
659   // Append any remaning text from the message with its existing formatting.
660   applyTemplateHighlighting(OS, Str.substr(Length), TextNormal, Bold);
661 
662   assert(TextNormal && "Text highlighted at end of diagnostic message.");
663 
664   return Wrapped;
665 }
666 
TextDiagnostic(raw_ostream & OS,const LangOptions & LangOpts,DiagnosticOptions * DiagOpts)667 TextDiagnostic::TextDiagnostic(raw_ostream &OS,
668                                const LangOptions &LangOpts,
669                                DiagnosticOptions *DiagOpts)
670   : DiagnosticRenderer(LangOpts, DiagOpts), OS(OS) {}
671 
~TextDiagnostic()672 TextDiagnostic::~TextDiagnostic() {}
673 
674 void
emitDiagnosticMessage(SourceLocation Loc,PresumedLoc PLoc,DiagnosticsEngine::Level Level,StringRef Message,ArrayRef<clang::CharSourceRange> Ranges,const SourceManager * SM,DiagOrStoredDiag D)675 TextDiagnostic::emitDiagnosticMessage(SourceLocation Loc,
676                                       PresumedLoc PLoc,
677                                       DiagnosticsEngine::Level Level,
678                                       StringRef Message,
679                                       ArrayRef<clang::CharSourceRange> Ranges,
680                                       const SourceManager *SM,
681                                       DiagOrStoredDiag D) {
682   uint64_t StartOfLocationInfo = OS.tell();
683 
684   // Emit the location of this particular diagnostic.
685   if (Loc.isValid())
686     emitDiagnosticLoc(Loc, PLoc, Level, Ranges, *SM);
687 
688   if (DiagOpts->ShowColors)
689     OS.resetColor();
690 
691   printDiagnosticLevel(OS, Level, DiagOpts->ShowColors,
692                        DiagOpts->CLFallbackMode);
693   printDiagnosticMessage(OS,
694                          /*IsSupplemental*/ Level == DiagnosticsEngine::Note,
695                          Message, OS.tell() - StartOfLocationInfo,
696                          DiagOpts->MessageLength, DiagOpts->ShowColors);
697 }
698 
699 /*static*/ void
printDiagnosticLevel(raw_ostream & OS,DiagnosticsEngine::Level Level,bool ShowColors,bool CLFallbackMode)700 TextDiagnostic::printDiagnosticLevel(raw_ostream &OS,
701                                      DiagnosticsEngine::Level Level,
702                                      bool ShowColors,
703                                      bool CLFallbackMode) {
704   if (ShowColors) {
705     // Print diagnostic category in bold and color
706     switch (Level) {
707     case DiagnosticsEngine::Ignored:
708       llvm_unreachable("Invalid diagnostic type");
709     case DiagnosticsEngine::Note:    OS.changeColor(noteColor, true); break;
710     case DiagnosticsEngine::Remark:  OS.changeColor(remarkColor, true); break;
711     case DiagnosticsEngine::Warning: OS.changeColor(warningColor, true); break;
712     case DiagnosticsEngine::Error:   OS.changeColor(errorColor, true); break;
713     case DiagnosticsEngine::Fatal:   OS.changeColor(fatalColor, true); break;
714     }
715   }
716 
717   switch (Level) {
718   case DiagnosticsEngine::Ignored:
719     llvm_unreachable("Invalid diagnostic type");
720   case DiagnosticsEngine::Note:    OS << "note"; break;
721   case DiagnosticsEngine::Remark:  OS << "remark"; break;
722   case DiagnosticsEngine::Warning: OS << "warning"; break;
723   case DiagnosticsEngine::Error:   OS << "error"; break;
724   case DiagnosticsEngine::Fatal:   OS << "fatal error"; break;
725   }
726 
727   // In clang-cl /fallback mode, print diagnostics as "error(clang):". This
728   // makes it more clear whether a message is coming from clang or cl.exe,
729   // and it prevents MSBuild from concluding that the build failed just because
730   // there is an "error:" in the output.
731   if (CLFallbackMode)
732     OS << "(clang)";
733 
734   OS << ": ";
735 
736   if (ShowColors)
737     OS.resetColor();
738 }
739 
740 /*static*/
printDiagnosticMessage(raw_ostream & OS,bool IsSupplemental,StringRef Message,unsigned CurrentColumn,unsigned Columns,bool ShowColors)741 void TextDiagnostic::printDiagnosticMessage(raw_ostream &OS,
742                                             bool IsSupplemental,
743                                             StringRef Message,
744                                             unsigned CurrentColumn,
745                                             unsigned Columns, bool ShowColors) {
746   bool Bold = false;
747   if (ShowColors && !IsSupplemental) {
748     // Print primary diagnostic messages in bold and without color, to visually
749     // indicate the transition from continuation notes and other output.
750     OS.changeColor(savedColor, true);
751     Bold = true;
752   }
753 
754   if (Columns)
755     printWordWrapped(OS, Message, Columns, CurrentColumn, Bold);
756   else {
757     bool Normal = true;
758     applyTemplateHighlighting(OS, Message, Normal, Bold);
759     assert(Normal && "Formatting should have returned to normal");
760   }
761 
762   if (ShowColors)
763     OS.resetColor();
764   OS << '\n';
765 }
766 
767 /// \brief Print out the file/line/column information and include trace.
768 ///
769 /// This method handlen the emission of the diagnostic location information.
770 /// This includes extracting as much location information as is present for
771 /// the diagnostic and printing it, as well as any include stack or source
772 /// ranges necessary.
emitDiagnosticLoc(SourceLocation Loc,PresumedLoc PLoc,DiagnosticsEngine::Level Level,ArrayRef<CharSourceRange> Ranges,const SourceManager & SM)773 void TextDiagnostic::emitDiagnosticLoc(SourceLocation Loc, PresumedLoc PLoc,
774                                        DiagnosticsEngine::Level Level,
775                                        ArrayRef<CharSourceRange> Ranges,
776                                        const SourceManager &SM) {
777   if (PLoc.isInvalid()) {
778     // At least print the file name if available:
779     FileID FID = SM.getFileID(Loc);
780     if (FID.isValid()) {
781       const FileEntry* FE = SM.getFileEntryForID(FID);
782       if (FE && FE->isValid()) {
783         OS << FE->getName();
784         if (FE->isInPCH())
785           OS << " (in PCH)";
786         OS << ": ";
787       }
788     }
789     return;
790   }
791   unsigned LineNo = PLoc.getLine();
792 
793   if (!DiagOpts->ShowLocation)
794     return;
795 
796   if (DiagOpts->ShowColors)
797     OS.changeColor(savedColor, true);
798 
799   OS << PLoc.getFilename();
800   switch (DiagOpts->getFormat()) {
801   case DiagnosticOptions::Clang: OS << ':'  << LineNo; break;
802   case DiagnosticOptions::MSVC:  OS << '('  << LineNo; break;
803   case DiagnosticOptions::Vi:    OS << " +" << LineNo; break;
804   }
805 
806   if (DiagOpts->ShowColumn)
807     // Compute the column number.
808     if (unsigned ColNo = PLoc.getColumn()) {
809       if (DiagOpts->getFormat() == DiagnosticOptions::MSVC) {
810         OS << ',';
811         // Visual Studio 2010 or earlier expects column number to be off by one
812         if (LangOpts.MSCompatibilityVersion &&
813             !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2012))
814           ColNo--;
815       } else
816         OS << ':';
817       OS << ColNo;
818     }
819   switch (DiagOpts->getFormat()) {
820   case DiagnosticOptions::Clang:
821   case DiagnosticOptions::Vi:    OS << ':';    break;
822   case DiagnosticOptions::MSVC:  OS << ") : "; break;
823   }
824 
825   if (DiagOpts->ShowSourceRanges && !Ranges.empty()) {
826     FileID CaretFileID =
827       SM.getFileID(SM.getExpansionLoc(Loc));
828     bool PrintedRange = false;
829 
830     for (ArrayRef<CharSourceRange>::const_iterator RI = Ranges.begin(),
831          RE = Ranges.end();
832          RI != RE; ++RI) {
833       // Ignore invalid ranges.
834       if (!RI->isValid()) continue;
835 
836       SourceLocation B = SM.getExpansionLoc(RI->getBegin());
837       SourceLocation E = SM.getExpansionLoc(RI->getEnd());
838 
839       // If the End location and the start location are the same and are a
840       // macro location, then the range was something that came from a
841       // macro expansion or _Pragma.  If this is an object-like macro, the
842       // best we can do is to highlight the range.  If this is a
843       // function-like macro, we'd also like to highlight the arguments.
844       if (B == E && RI->getEnd().isMacroID())
845         E = SM.getExpansionRange(RI->getEnd()).second;
846 
847       std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(B);
848       std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(E);
849 
850       // If the start or end of the range is in another file, just discard
851       // it.
852       if (BInfo.first != CaretFileID || EInfo.first != CaretFileID)
853         continue;
854 
855       // Add in the length of the token, so that we cover multi-char
856       // tokens.
857       unsigned TokSize = 0;
858       if (RI->isTokenRange())
859         TokSize = Lexer::MeasureTokenLength(E, SM, LangOpts);
860 
861       OS << '{' << SM.getLineNumber(BInfo.first, BInfo.second) << ':'
862         << SM.getColumnNumber(BInfo.first, BInfo.second) << '-'
863         << SM.getLineNumber(EInfo.first, EInfo.second) << ':'
864         << (SM.getColumnNumber(EInfo.first, EInfo.second)+TokSize)
865         << '}';
866       PrintedRange = true;
867     }
868 
869     if (PrintedRange)
870       OS << ':';
871   }
872   OS << ' ';
873 }
874 
emitIncludeLocation(SourceLocation Loc,PresumedLoc PLoc,const SourceManager & SM)875 void TextDiagnostic::emitIncludeLocation(SourceLocation Loc,
876                                          PresumedLoc PLoc,
877                                          const SourceManager &SM) {
878   if (DiagOpts->ShowLocation && PLoc.getFilename())
879     OS << "In file included from " << PLoc.getFilename() << ':'
880        << PLoc.getLine() << ":\n";
881   else
882     OS << "In included file:\n";
883 }
884 
emitImportLocation(SourceLocation Loc,PresumedLoc PLoc,StringRef ModuleName,const SourceManager & SM)885 void TextDiagnostic::emitImportLocation(SourceLocation Loc, PresumedLoc PLoc,
886                                         StringRef ModuleName,
887                                         const SourceManager &SM) {
888   if (DiagOpts->ShowLocation && PLoc.getFilename())
889     OS << "In module '" << ModuleName << "' imported from "
890        << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
891   else
892     OS << "In module '" << ModuleName << "':\n";
893 }
894 
emitBuildingModuleLocation(SourceLocation Loc,PresumedLoc PLoc,StringRef ModuleName,const SourceManager & SM)895 void TextDiagnostic::emitBuildingModuleLocation(SourceLocation Loc,
896                                                 PresumedLoc PLoc,
897                                                 StringRef ModuleName,
898                                                 const SourceManager &SM) {
899   if (DiagOpts->ShowLocation && PLoc.getFilename())
900     OS << "While building module '" << ModuleName << "' imported from "
901       << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
902   else
903     OS << "While building module '" << ModuleName << "':\n";
904 }
905 
906 /// \brief Highlight a SourceRange (with ~'s) for any characters on LineNo.
highlightRange(const CharSourceRange & R,unsigned LineNo,FileID FID,const SourceColumnMap & map,std::string & CaretLine,const SourceManager & SM,const LangOptions & LangOpts)907 static void highlightRange(const CharSourceRange &R,
908                            unsigned LineNo, FileID FID,
909                            const SourceColumnMap &map,
910                            std::string &CaretLine,
911                            const SourceManager &SM,
912                            const LangOptions &LangOpts) {
913   if (!R.isValid()) return;
914 
915   SourceLocation Begin = R.getBegin();
916   SourceLocation End = R.getEnd();
917 
918   unsigned StartLineNo = SM.getExpansionLineNumber(Begin);
919   if (StartLineNo > LineNo || SM.getFileID(Begin) != FID)
920     return;  // No intersection.
921 
922   unsigned EndLineNo = SM.getExpansionLineNumber(End);
923   if (EndLineNo < LineNo || SM.getFileID(End) != FID)
924     return;  // No intersection.
925 
926   // Compute the column number of the start.
927   unsigned StartColNo = 0;
928   if (StartLineNo == LineNo) {
929     StartColNo = SM.getExpansionColumnNumber(Begin);
930     if (StartColNo) --StartColNo;  // Zero base the col #.
931   }
932 
933   // Compute the column number of the end.
934   unsigned EndColNo = map.getSourceLine().size();
935   if (EndLineNo == LineNo) {
936     EndColNo = SM.getExpansionColumnNumber(End);
937     if (EndColNo) {
938       --EndColNo;  // Zero base the col #.
939 
940       // Add in the length of the token, so that we cover multi-char tokens if
941       // this is a token range.
942       if (R.isTokenRange())
943         EndColNo += Lexer::MeasureTokenLength(End, SM, LangOpts);
944     } else {
945       EndColNo = CaretLine.size();
946     }
947   }
948 
949   assert(StartColNo <= EndColNo && "Invalid range!");
950 
951   // Check that a token range does not highlight only whitespace.
952   if (R.isTokenRange()) {
953     // Pick the first non-whitespace column.
954     while (StartColNo < map.getSourceLine().size() &&
955            (map.getSourceLine()[StartColNo] == ' ' ||
956             map.getSourceLine()[StartColNo] == '\t'))
957       StartColNo = map.startOfNextColumn(StartColNo);
958 
959     // Pick the last non-whitespace column.
960     if (EndColNo > map.getSourceLine().size())
961       EndColNo = map.getSourceLine().size();
962     while (EndColNo &&
963            (map.getSourceLine()[EndColNo-1] == ' ' ||
964             map.getSourceLine()[EndColNo-1] == '\t'))
965       EndColNo = map.startOfPreviousColumn(EndColNo);
966 
967     // If the start/end passed each other, then we are trying to highlight a
968     // range that just exists in whitespace, which must be some sort of other
969     // bug.
970     assert(StartColNo <= EndColNo && "Trying to highlight whitespace??");
971   }
972 
973   assert(StartColNo <= map.getSourceLine().size() && "Invalid range!");
974   assert(EndColNo <= map.getSourceLine().size() && "Invalid range!");
975 
976   // Fill the range with ~'s.
977   StartColNo = map.byteToContainingColumn(StartColNo);
978   EndColNo = map.byteToContainingColumn(EndColNo);
979 
980   assert(StartColNo <= EndColNo && "Invalid range!");
981   if (CaretLine.size() < EndColNo)
982     CaretLine.resize(EndColNo,' ');
983   std::fill(CaretLine.begin()+StartColNo,CaretLine.begin()+EndColNo,'~');
984 }
985 
buildFixItInsertionLine(unsigned LineNo,const SourceColumnMap & map,ArrayRef<FixItHint> Hints,const SourceManager & SM,const DiagnosticOptions * DiagOpts)986 static std::string buildFixItInsertionLine(unsigned LineNo,
987                                            const SourceColumnMap &map,
988                                            ArrayRef<FixItHint> Hints,
989                                            const SourceManager &SM,
990                                            const DiagnosticOptions *DiagOpts) {
991   std::string FixItInsertionLine;
992   if (Hints.empty() || !DiagOpts->ShowFixits)
993     return FixItInsertionLine;
994   unsigned PrevHintEndCol = 0;
995 
996   for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
997        I != E; ++I) {
998     if (!I->CodeToInsert.empty()) {
999       // We have an insertion hint. Determine whether the inserted
1000       // code contains no newlines and is on the same line as the caret.
1001       std::pair<FileID, unsigned> HintLocInfo
1002         = SM.getDecomposedExpansionLoc(I->RemoveRange.getBegin());
1003       if (LineNo == SM.getLineNumber(HintLocInfo.first, HintLocInfo.second) &&
1004           StringRef(I->CodeToInsert).find_first_of("\n\r") == StringRef::npos) {
1005         // Insert the new code into the line just below the code
1006         // that the user wrote.
1007         // Note: When modifying this function, be very careful about what is a
1008         // "column" (printed width, platform-dependent) and what is a
1009         // "byte offset" (SourceManager "column").
1010         unsigned HintByteOffset
1011           = SM.getColumnNumber(HintLocInfo.first, HintLocInfo.second) - 1;
1012 
1013         // The hint must start inside the source or right at the end
1014         assert(HintByteOffset < static_cast<unsigned>(map.bytes())+1);
1015         unsigned HintCol = map.byteToContainingColumn(HintByteOffset);
1016 
1017         // If we inserted a long previous hint, push this one forwards, and add
1018         // an extra space to show that this is not part of the previous
1019         // completion. This is sort of the best we can do when two hints appear
1020         // to overlap.
1021         //
1022         // Note that if this hint is located immediately after the previous
1023         // hint, no space will be added, since the location is more important.
1024         if (HintCol < PrevHintEndCol)
1025           HintCol = PrevHintEndCol + 1;
1026 
1027         // This should NOT use HintByteOffset, because the source might have
1028         // Unicode characters in earlier columns.
1029         unsigned NewFixItLineSize = FixItInsertionLine.size() +
1030           (HintCol - PrevHintEndCol) + I->CodeToInsert.size();
1031         if (NewFixItLineSize > FixItInsertionLine.size())
1032           FixItInsertionLine.resize(NewFixItLineSize, ' ');
1033 
1034         std::copy(I->CodeToInsert.begin(), I->CodeToInsert.end(),
1035                   FixItInsertionLine.end() - I->CodeToInsert.size());
1036 
1037         PrevHintEndCol =
1038           HintCol + llvm::sys::locale::columnWidth(I->CodeToInsert);
1039       } else {
1040         FixItInsertionLine.clear();
1041         break;
1042       }
1043     }
1044   }
1045 
1046   expandTabs(FixItInsertionLine, DiagOpts->TabStop);
1047 
1048   return FixItInsertionLine;
1049 }
1050 
1051 /// \brief Emit a code snippet and caret line.
1052 ///
1053 /// This routine emits a single line's code snippet and caret line..
1054 ///
1055 /// \param Loc The location for the caret.
1056 /// \param Ranges The underlined ranges for this code snippet.
1057 /// \param Hints The FixIt hints active for this diagnostic.
emitSnippetAndCaret(SourceLocation Loc,DiagnosticsEngine::Level Level,SmallVectorImpl<CharSourceRange> & Ranges,ArrayRef<FixItHint> Hints,const SourceManager & SM)1058 void TextDiagnostic::emitSnippetAndCaret(
1059     SourceLocation Loc, DiagnosticsEngine::Level Level,
1060     SmallVectorImpl<CharSourceRange>& Ranges,
1061     ArrayRef<FixItHint> Hints,
1062     const SourceManager &SM) {
1063   assert(Loc.isValid() && "must have a valid source location here");
1064   assert(Loc.isFileID() && "must have a file location here");
1065 
1066   // If caret diagnostics are enabled and we have location, we want to
1067   // emit the caret.  However, we only do this if the location moved
1068   // from the last diagnostic, if the last diagnostic was a note that
1069   // was part of a different warning or error diagnostic, or if the
1070   // diagnostic has ranges.  We don't want to emit the same caret
1071   // multiple times if one loc has multiple diagnostics.
1072   if (!DiagOpts->ShowCarets)
1073     return;
1074   if (Loc == LastLoc && Ranges.empty() && Hints.empty() &&
1075       (LastLevel != DiagnosticsEngine::Note || Level == LastLevel))
1076     return;
1077 
1078   // Decompose the location into a FID/Offset pair.
1079   std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1080   FileID FID = LocInfo.first;
1081   unsigned FileOffset = LocInfo.second;
1082 
1083   // Get information about the buffer it points into.
1084   bool Invalid = false;
1085   const char *BufStart = SM.getBufferData(FID, &Invalid).data();
1086   if (Invalid)
1087     return;
1088 
1089   unsigned LineNo = SM.getLineNumber(FID, FileOffset);
1090   unsigned ColNo = SM.getColumnNumber(FID, FileOffset);
1091 
1092   // Arbitrarily stop showing snippets when the line is too long.
1093   static const size_t MaxLineLengthToPrint = 4096;
1094   if (ColNo > MaxLineLengthToPrint)
1095     return;
1096 
1097   // Rewind from the current position to the start of the line.
1098   const char *TokPtr = BufStart+FileOffset;
1099   const char *LineStart = TokPtr-ColNo+1; // Column # is 1-based.
1100 
1101   // Compute the line end.  Scan forward from the error position to the end of
1102   // the line.
1103   const char *LineEnd = TokPtr;
1104   while (*LineEnd != '\n' && *LineEnd != '\r' && *LineEnd != '\0')
1105     ++LineEnd;
1106 
1107   // Arbitrarily stop showing snippets when the line is too long.
1108   if (size_t(LineEnd - LineStart) > MaxLineLengthToPrint)
1109     return;
1110 
1111   // Copy the line of code into an std::string for ease of manipulation.
1112   std::string SourceLine(LineStart, LineEnd);
1113 
1114   // Build the byte to column map.
1115   const SourceColumnMap sourceColMap(SourceLine, DiagOpts->TabStop);
1116 
1117   // Create a line for the caret that is filled with spaces that is the same
1118   // number of columns as the line of source code.
1119   std::string CaretLine(sourceColMap.columns(), ' ');
1120 
1121   // Highlight all of the characters covered by Ranges with ~ characters.
1122   for (SmallVectorImpl<CharSourceRange>::iterator I = Ranges.begin(),
1123                                                   E = Ranges.end();
1124        I != E; ++I)
1125     highlightRange(*I, LineNo, FID, sourceColMap, CaretLine, SM, LangOpts);
1126 
1127   // Next, insert the caret itself.
1128   ColNo = sourceColMap.byteToContainingColumn(ColNo-1);
1129   if (CaretLine.size()<ColNo+1)
1130     CaretLine.resize(ColNo+1, ' ');
1131   CaretLine[ColNo] = '^';
1132 
1133   std::string FixItInsertionLine = buildFixItInsertionLine(LineNo,
1134                                                            sourceColMap,
1135                                                            Hints, SM,
1136                                                            DiagOpts.get());
1137 
1138   // If the source line is too long for our terminal, select only the
1139   // "interesting" source region within that line.
1140   unsigned Columns = DiagOpts->MessageLength;
1141   if (Columns)
1142     selectInterestingSourceRegion(SourceLine, CaretLine, FixItInsertionLine,
1143                                   Columns, sourceColMap);
1144 
1145   // If we are in -fdiagnostics-print-source-range-info mode, we are trying
1146   // to produce easily machine parsable output.  Add a space before the
1147   // source line and the caret to make it trivial to tell the main diagnostic
1148   // line from what the user is intended to see.
1149   if (DiagOpts->ShowSourceRanges) {
1150     SourceLine = ' ' + SourceLine;
1151     CaretLine = ' ' + CaretLine;
1152   }
1153 
1154   // Finally, remove any blank spaces from the end of CaretLine.
1155   while (CaretLine[CaretLine.size()-1] == ' ')
1156     CaretLine.erase(CaretLine.end()-1);
1157 
1158   // Emit what we have computed.
1159   emitSnippet(SourceLine);
1160 
1161   if (DiagOpts->ShowColors)
1162     OS.changeColor(caretColor, true);
1163   OS << CaretLine << '\n';
1164   if (DiagOpts->ShowColors)
1165     OS.resetColor();
1166 
1167   if (!FixItInsertionLine.empty()) {
1168     if (DiagOpts->ShowColors)
1169       // Print fixit line in color
1170       OS.changeColor(fixitColor, false);
1171     if (DiagOpts->ShowSourceRanges)
1172       OS << ' ';
1173     OS << FixItInsertionLine << '\n';
1174     if (DiagOpts->ShowColors)
1175       OS.resetColor();
1176   }
1177 
1178   // Print out any parseable fixit information requested by the options.
1179   emitParseableFixits(Hints, SM);
1180 }
1181 
emitSnippet(StringRef line)1182 void TextDiagnostic::emitSnippet(StringRef line) {
1183   if (line.empty())
1184     return;
1185 
1186   size_t i = 0;
1187 
1188   std::string to_print;
1189   bool print_reversed = false;
1190 
1191   while (i<line.size()) {
1192     std::pair<SmallString<16>,bool> res
1193         = printableTextForNextCharacter(line, &i, DiagOpts->TabStop);
1194     bool was_printable = res.second;
1195 
1196     if (DiagOpts->ShowColors && was_printable == print_reversed) {
1197       if (print_reversed)
1198         OS.reverseColor();
1199       OS << to_print;
1200       to_print.clear();
1201       if (DiagOpts->ShowColors)
1202         OS.resetColor();
1203     }
1204 
1205     print_reversed = !was_printable;
1206     to_print += res.first.str();
1207   }
1208 
1209   if (print_reversed && DiagOpts->ShowColors)
1210     OS.reverseColor();
1211   OS << to_print;
1212   if (print_reversed && DiagOpts->ShowColors)
1213     OS.resetColor();
1214 
1215   OS << '\n';
1216 }
1217 
emitParseableFixits(ArrayRef<FixItHint> Hints,const SourceManager & SM)1218 void TextDiagnostic::emitParseableFixits(ArrayRef<FixItHint> Hints,
1219                                          const SourceManager &SM) {
1220   if (!DiagOpts->ShowParseableFixits)
1221     return;
1222 
1223   // We follow FixItRewriter's example in not (yet) handling
1224   // fix-its in macros.
1225   for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1226        I != E; ++I) {
1227     if (I->RemoveRange.isInvalid() ||
1228         I->RemoveRange.getBegin().isMacroID() ||
1229         I->RemoveRange.getEnd().isMacroID())
1230       return;
1231   }
1232 
1233   for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1234        I != E; ++I) {
1235     SourceLocation BLoc = I->RemoveRange.getBegin();
1236     SourceLocation ELoc = I->RemoveRange.getEnd();
1237 
1238     std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(BLoc);
1239     std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(ELoc);
1240 
1241     // Adjust for token ranges.
1242     if (I->RemoveRange.isTokenRange())
1243       EInfo.second += Lexer::MeasureTokenLength(ELoc, SM, LangOpts);
1244 
1245     // We specifically do not do word-wrapping or tab-expansion here,
1246     // because this is supposed to be easy to parse.
1247     PresumedLoc PLoc = SM.getPresumedLoc(BLoc);
1248     if (PLoc.isInvalid())
1249       break;
1250 
1251     OS << "fix-it:\"";
1252     OS.write_escaped(PLoc.getFilename());
1253     OS << "\":{" << SM.getLineNumber(BInfo.first, BInfo.second)
1254       << ':' << SM.getColumnNumber(BInfo.first, BInfo.second)
1255       << '-' << SM.getLineNumber(EInfo.first, EInfo.second)
1256       << ':' << SM.getColumnNumber(EInfo.first, EInfo.second)
1257       << "}:\"";
1258     OS.write_escaped(I->CodeToInsert);
1259     OS << "\"\n";
1260   }
1261 }
1262