1 //===--- TextDiagnostic.cpp - Text Diagnostic Pretty-Printing -------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
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"
25 using namespace clang;
27 static const enum raw_ostream::Colors noteColor =
29 static const enum raw_ostream::Colors fixitColor =
31 static const enum raw_ostream::Colors caretColor =
33 static const enum raw_ostream::Colors warningColor =
35 static const enum raw_ostream::Colors templateColor =
37 static const enum raw_ostream::Colors errorColor = raw_ostream::RED;
38 static const enum raw_ostream::Colors fatalColor = raw_ostream::RED;
39 // Used for changing only the bold attribute.
40 static const enum raw_ostream::Colors savedColor =
41 raw_ostream::SAVEDCOLOR;
43 /// \brief Add highlights to differences in template strings.
44 static void applyTemplateHighlighting(raw_ostream &OS, StringRef Str,
45 bool &Normal, bool Bold) {
47 size_t Pos = Str.find(ToggleHighlight);
48 OS << Str.slice(0, Pos);
49 if (Pos == StringRef::npos)
52 Str = Str.substr(Pos + 1);
54 OS.changeColor(templateColor, true);
58 OS.changeColor(savedColor, true);
64 /// \brief Number of spaces to indent when word-wrapping.
65 const unsigned WordWrapIndentation = 6;
67 static int bytesSincePreviousTabOrLineBegin(StringRef SourceLine, size_t i) {
70 if (SourceLine[--i]=='\t')
77 /// \brief returns a printable representation of first item from input range
79 /// This function returns a printable representation of the next item in a line
80 /// of source. If the next byte begins a valid and printable character, that
81 /// character is returned along with 'true'.
83 /// Otherwise, if the next byte begins a valid, but unprintable character, a
84 /// printable, escaped representation of the character is returned, along with
85 /// 'false'. Otherwise a printable, escaped representation of the next byte
86 /// is returned along with 'false'.
88 /// \note The index is updated to be used with a subsequent call to
89 /// printableTextForNextCharacter.
91 /// \param SourceLine The line of source
92 /// \param i Pointer to byte index,
93 /// \param TabStop used to expand tabs
94 /// \return pair(printable text, 'true' iff original text was printable)
96 static std::pair<SmallString<16>, bool>
97 printableTextForNextCharacter(StringRef SourceLine, size_t *i,
99 assert(i && "i must not be null");
100 assert(*i<SourceLine.size() && "must point to a valid index");
102 if (SourceLine[*i]=='\t') {
103 assert(0 < TabStop && TabStop <= DiagnosticOptions::MaxTabStop &&
104 "Invalid -ftabstop value");
105 unsigned col = bytesSincePreviousTabOrLineBegin(SourceLine, *i);
106 unsigned NumSpaces = TabStop - col%TabStop;
107 assert(0 < NumSpaces && NumSpaces <= TabStop
108 && "Invalid computation of space amt");
111 SmallString<16> expandedTab;
112 expandedTab.assign(NumSpaces, ' ');
113 return std::make_pair(expandedTab, true);
116 unsigned char const *begin, *end;
117 begin = reinterpret_cast<unsigned char const *>(&*(SourceLine.begin() + *i));
118 end = begin + (SourceLine.size() - *i);
120 if (isLegalUTF8Sequence(begin, end)) {
123 unsigned char const *original_begin = begin;
124 unsigned char const *cp_end = begin+getNumBytesForUTF8(SourceLine[*i]);
126 ConversionResult res = ConvertUTF8toUTF32(&begin, cp_end, &cptr, cptr+1,
129 assert(conversionOK==res);
130 assert(0 < begin-original_begin
131 && "we must be further along in the string now");
132 *i += begin-original_begin;
134 if (!llvm::sys::locale::isPrint(c)) {
135 // If next character is valid UTF-8, but not printable
136 SmallString<16> expandedCP("<U+>");
138 expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(c%16));
141 while (expandedCP.size() < 8)
142 expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(0));
143 return std::make_pair(expandedCP, false);
146 // If next character is valid UTF-8, and printable
147 return std::make_pair(SmallString<16>(original_begin, cp_end), true);
151 // If next byte is not valid UTF-8 (and therefore not printable)
152 SmallString<16> expandedByte("<XX>");
153 unsigned char byte = SourceLine[*i];
154 expandedByte[1] = llvm::hexdigit(byte / 16);
155 expandedByte[2] = llvm::hexdigit(byte % 16);
157 return std::make_pair(expandedByte, false);
160 static void expandTabs(std::string &SourceLine, unsigned TabStop) {
161 size_t i = SourceLine.size();
164 if (SourceLine[i]!='\t')
167 std::pair<SmallString<16>,bool> res
168 = printableTextForNextCharacter(SourceLine, &tmp_i, TabStop);
169 SourceLine.replace(i, 1, res.first.c_str());
173 /// This function takes a raw source line and produces a mapping from the bytes
174 /// of the printable representation of the line to the columns those printable
175 /// characters will appear at (numbering the first column as 0).
177 /// If a byte 'i' corresponds to muliple columns (e.g. the byte contains a tab
178 /// character) then the array will map that byte to the first column the
179 /// tab appears at and the next value in the map will have been incremented
182 /// If a byte is the first in a sequence of bytes that together map to a single
183 /// entity in the output, then the array will map that byte to the appropriate
184 /// column while the subsequent bytes will be -1.
186 /// The last element in the array does not correspond to any byte in the input
187 /// and instead is the number of columns needed to display the source
189 /// example: (given a tabstop of 8)
191 /// "a \t \u3042" -> {0,1,2,8,9,-1,-1,11}
193 /// (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
195 static void byteToColumn(StringRef SourceLine, unsigned TabStop,
196 SmallVectorImpl<int> &out) {
199 if (SourceLine.empty()) {
204 out.resize(SourceLine.size()+1, -1);
208 while (i<SourceLine.size()) {
210 std::pair<SmallString<16>,bool> res
211 = printableTextForNextCharacter(SourceLine, &i, TabStop);
212 columns += llvm::sys::locale::columnWidth(res.first);
214 out.back() = columns;
217 /// This function takes a raw source line and produces a mapping from columns
218 /// to the byte of the source line that produced the character displaying at
219 /// that column. This is the inverse of the mapping produced by byteToColumn()
221 /// The last element in the array is the number of bytes in the source string
223 /// example: (given a tabstop of 8)
225 /// "a \t \u3042" -> {0,1,2,-1,-1,-1,-1,-1,3,4,-1,7}
227 /// (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
229 static void columnToByte(StringRef SourceLine, unsigned TabStop,
230 SmallVectorImpl<int> &out) {
233 if (SourceLine.empty()) {
240 while (i<SourceLine.size()) {
241 out.resize(columns+1, -1);
243 std::pair<SmallString<16>,bool> res
244 = printableTextForNextCharacter(SourceLine, &i, TabStop);
245 columns += llvm::sys::locale::columnWidth(res.first);
247 out.resize(columns+1, -1);
252 struct SourceColumnMap {
253 SourceColumnMap(StringRef SourceLine, unsigned TabStop)
254 : m_SourceLine(SourceLine) {
256 ::byteToColumn(SourceLine, TabStop, m_byteToColumn);
257 ::columnToByte(SourceLine, TabStop, m_columnToByte);
259 assert(m_byteToColumn.size()==SourceLine.size()+1);
260 assert(0 < m_byteToColumn.size() && 0 < m_columnToByte.size());
261 assert(m_byteToColumn.size()
262 == static_cast<unsigned>(m_columnToByte.back()+1));
263 assert(static_cast<unsigned>(m_byteToColumn.back()+1)
264 == m_columnToByte.size());
266 int columns() const { return m_byteToColumn.back(); }
267 int bytes() const { return m_columnToByte.back(); }
269 /// \brief Map a byte to the column which it is at the start of, or return -1
270 /// if it is not at the start of a column (for a UTF-8 trailing byte).
271 int byteToColumn(int n) const {
272 assert(0<=n && n<static_cast<int>(m_byteToColumn.size()));
273 return m_byteToColumn[n];
276 /// \brief Map a byte to the first column which contains it.
277 int byteToContainingColumn(int N) const {
278 assert(0 <= N && N < static_cast<int>(m_byteToColumn.size()));
279 while (m_byteToColumn[N] == -1)
281 return m_byteToColumn[N];
284 /// \brief Map a column to the byte which starts the column, or return -1 if
285 /// the column the second or subsequent column of an expanded tab or similar
286 /// multi-column entity.
287 int columnToByte(int n) const {
288 assert(0<=n && n<static_cast<int>(m_columnToByte.size()));
289 return m_columnToByte[n];
292 /// \brief Map from a byte index to the next byte which starts a column.
293 int startOfNextColumn(int N) const {
294 assert(0 <= N && N < static_cast<int>(m_columnToByte.size() - 1));
295 while (byteToColumn(++N) == -1) {}
299 /// \brief Map from a byte index to the previous byte which starts a column.
300 int startOfPreviousColumn(int N) const {
301 assert(0 < N && N < static_cast<int>(m_columnToByte.size()));
302 while (byteToColumn(--N) == -1) {}
306 StringRef getSourceLine() const {
311 const std::string m_SourceLine;
312 SmallVector<int,200> m_byteToColumn;
313 SmallVector<int,200> m_columnToByte;
316 // used in assert in selectInterestingSourceRegion()
317 struct char_out_of_range {
318 const char lower,upper;
319 char_out_of_range(char lower, char upper) :
320 lower(lower), upper(upper) {}
321 bool operator()(char c) { return c < lower || upper < c; }
323 } // end anonymous namespace
325 /// \brief When the source code line we want to print is too long for
326 /// the terminal, select the "interesting" region.
327 static void selectInterestingSourceRegion(std::string &SourceLine,
328 std::string &CaretLine,
329 std::string &FixItInsertionLine,
331 const SourceColumnMap &map) {
332 unsigned MaxColumns = std::max<unsigned>(map.columns(),
333 std::max(CaretLine.size(),
334 FixItInsertionLine.size()));
335 // if the number of columns is less than the desired number we're done
336 if (MaxColumns <= Columns)
339 // No special characters are allowed in CaretLine.
340 assert(CaretLine.end() ==
341 std::find_if(CaretLine.begin(), CaretLine.end(),
342 char_out_of_range(' ','~')));
344 // Find the slice that we need to display the full caret line
346 unsigned CaretStart = 0, CaretEnd = CaretLine.size();
347 for (; CaretStart != CaretEnd; ++CaretStart)
348 if (!isWhitespace(CaretLine[CaretStart]))
351 for (; CaretEnd != CaretStart; --CaretEnd)
352 if (!isWhitespace(CaretLine[CaretEnd - 1]))
355 // caret has already been inserted into CaretLine so the above whitespace
356 // check is guaranteed to include the caret
358 // If we have a fix-it line, make sure the slice includes all of the
359 // fix-it information.
360 if (!FixItInsertionLine.empty()) {
361 unsigned FixItStart = 0, FixItEnd = FixItInsertionLine.size();
362 for (; FixItStart != FixItEnd; ++FixItStart)
363 if (!isWhitespace(FixItInsertionLine[FixItStart]))
366 for (; FixItEnd != FixItStart; --FixItEnd)
367 if (!isWhitespace(FixItInsertionLine[FixItEnd - 1]))
370 // We can safely use the byte offset FixItStart as the column offset
371 // because the characters up until FixItStart are all ASCII whitespace
373 unsigned FixItStartCol = FixItStart;
375 = llvm::sys::locale::columnWidth(FixItInsertionLine.substr(0, FixItEnd));
377 CaretStart = std::min(FixItStartCol, CaretStart);
378 CaretEnd = std::max(FixItEndCol, CaretEnd);
381 // CaretEnd may have been set at the middle of a character
382 // If it's not at a character's first column then advance it past the current
384 while (static_cast<int>(CaretEnd) < map.columns() &&
385 -1 == map.columnToByte(CaretEnd))
388 assert((static_cast<int>(CaretStart) > map.columns() ||
389 -1!=map.columnToByte(CaretStart)) &&
390 "CaretStart must not point to a column in the middle of a source"
392 assert((static_cast<int>(CaretEnd) > map.columns() ||
393 -1!=map.columnToByte(CaretEnd)) &&
394 "CaretEnd must not point to a column in the middle of a source line"
397 // CaretLine[CaretStart, CaretEnd) contains all of the interesting
398 // parts of the caret line. While this slice is smaller than the
399 // number of columns we have, try to grow the slice to encompass
402 unsigned SourceStart = map.columnToByte(std::min<unsigned>(CaretStart,
404 unsigned SourceEnd = map.columnToByte(std::min<unsigned>(CaretEnd,
407 unsigned CaretColumnsOutsideSource = CaretEnd-CaretStart
408 - (map.byteToColumn(SourceEnd)-map.byteToColumn(SourceStart));
410 char const *front_ellipse = " ...";
411 char const *front_space = " ";
412 char const *back_ellipse = "...";
413 unsigned ellipses_space = strlen(front_ellipse) + strlen(back_ellipse);
415 unsigned TargetColumns = Columns;
416 // Give us extra room for the ellipses
417 // and any of the caret line that extends past the source
418 if (TargetColumns > ellipses_space+CaretColumnsOutsideSource)
419 TargetColumns -= ellipses_space+CaretColumnsOutsideSource;
421 while (SourceStart>0 || SourceEnd<SourceLine.size()) {
422 bool ExpandedRegion = false;
425 unsigned NewStart = map.startOfPreviousColumn(SourceStart);
427 // Skip over any whitespace we see here; we're looking for
428 // another bit of interesting text.
429 // FIXME: Detect non-ASCII whitespace characters too.
430 while (NewStart && isWhitespace(SourceLine[NewStart]))
431 NewStart = map.startOfPreviousColumn(NewStart);
433 // Skip over this bit of "interesting" text.
435 unsigned Prev = map.startOfPreviousColumn(NewStart);
436 if (isWhitespace(SourceLine[Prev]))
441 assert(map.byteToColumn(NewStart) != -1);
442 unsigned NewColumns = map.byteToColumn(SourceEnd) -
443 map.byteToColumn(NewStart);
444 if (NewColumns <= TargetColumns) {
445 SourceStart = NewStart;
446 ExpandedRegion = true;
450 if (SourceEnd<SourceLine.size()) {
451 unsigned NewEnd = map.startOfNextColumn(SourceEnd);
453 // Skip over any whitespace we see here; we're looking for
454 // another bit of interesting text.
455 // FIXME: Detect non-ASCII whitespace characters too.
456 while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
457 NewEnd = map.startOfNextColumn(NewEnd);
459 // Skip over this bit of "interesting" text.
460 while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
461 NewEnd = map.startOfNextColumn(NewEnd);
463 assert(map.byteToColumn(NewEnd) != -1);
464 unsigned NewColumns = map.byteToColumn(NewEnd) -
465 map.byteToColumn(SourceStart);
466 if (NewColumns <= TargetColumns) {
468 ExpandedRegion = true;
476 CaretStart = map.byteToColumn(SourceStart);
477 CaretEnd = map.byteToColumn(SourceEnd) + CaretColumnsOutsideSource;
479 // [CaretStart, CaretEnd) is the slice we want. Update the various
480 // output lines to show only this slice, with two-space padding
481 // before the lines so that it looks nicer.
483 assert(CaretStart!=(unsigned)-1 && CaretEnd!=(unsigned)-1 &&
484 SourceStart!=(unsigned)-1 && SourceEnd!=(unsigned)-1);
485 assert(SourceStart <= SourceEnd);
486 assert(CaretStart <= CaretEnd);
488 unsigned BackColumnsRemoved
489 = map.byteToColumn(SourceLine.size())-map.byteToColumn(SourceEnd);
490 unsigned FrontColumnsRemoved = CaretStart;
491 unsigned ColumnsKept = CaretEnd-CaretStart;
493 // We checked up front that the line needed truncation
494 assert(FrontColumnsRemoved+ColumnsKept+BackColumnsRemoved > Columns);
496 // The line needs some trunctiona, and we'd prefer to keep the front
497 // if possible, so remove the back
498 if (BackColumnsRemoved > strlen(back_ellipse))
499 SourceLine.replace(SourceEnd, std::string::npos, back_ellipse);
501 // If that's enough then we're done
502 if (FrontColumnsRemoved+ColumnsKept <= Columns)
505 // Otherwise remove the front as well
506 if (FrontColumnsRemoved > strlen(front_ellipse)) {
507 SourceLine.replace(0, SourceStart, front_ellipse);
508 CaretLine.replace(0, CaretStart, front_space);
509 if (!FixItInsertionLine.empty())
510 FixItInsertionLine.replace(0, CaretStart, front_space);
514 /// \brief Skip over whitespace in the string, starting at the given
517 /// \returns The index of the first non-whitespace character that is
518 /// greater than or equal to Idx or, if no such character exists,
519 /// returns the end of the string.
520 static unsigned skipWhitespace(unsigned Idx, StringRef Str, unsigned Length) {
521 while (Idx < Length && isWhitespace(Str[Idx]))
526 /// \brief If the given character is the start of some kind of
527 /// balanced punctuation (e.g., quotes or parentheses), return the
528 /// character that will terminate the punctuation.
530 /// \returns The ending punctuation character, if any, or the NULL
531 /// character if the input character does not start any punctuation.
532 static inline char findMatchingPunctuation(char c) {
534 case '\'': return '\'';
535 case '`': return '\'';
536 case '"': return '"';
537 case '(': return ')';
538 case '[': return ']';
539 case '{': return '}';
546 /// \brief Find the end of the word starting at the given offset
549 /// \returns the index pointing one character past the end of the
551 static unsigned findEndOfWord(unsigned Start, StringRef Str,
552 unsigned Length, unsigned Column,
554 assert(Start < Str.size() && "Invalid start position!");
555 unsigned End = Start + 1;
557 // If we are already at the end of the string, take that as the word.
558 if (End == Str.size())
561 // Determine if the start of the string is actually opening
562 // punctuation, e.g., a quote or parentheses.
563 char EndPunct = findMatchingPunctuation(Str[Start]);
565 // This is a normal word. Just find the first space character.
566 while (End < Length && !isWhitespace(Str[End]))
571 // We have the start of a balanced punctuation sequence (quotes,
572 // parentheses, etc.). Determine the full sequence is.
573 SmallString<16> PunctuationEndStack;
574 PunctuationEndStack.push_back(EndPunct);
575 while (End < Length && !PunctuationEndStack.empty()) {
576 if (Str[End] == PunctuationEndStack.back())
577 PunctuationEndStack.pop_back();
578 else if (char SubEndPunct = findMatchingPunctuation(Str[End]))
579 PunctuationEndStack.push_back(SubEndPunct);
584 // Find the first space character after the punctuation ended.
585 while (End < Length && !isWhitespace(Str[End]))
588 unsigned PunctWordLength = End - Start;
589 if (// If the word fits on this line
590 Column + PunctWordLength <= Columns ||
591 // ... or the word is "short enough" to take up the next line
592 // without too much ugly white space
593 PunctWordLength < Columns/3)
594 return End; // Take the whole thing as a single "word".
596 // The whole quoted/parenthesized string is too long to print as a
597 // single "word". Instead, find the "word" that starts just after
598 // the punctuation and use that end-point instead. This will recurse
599 // until it finds something small enough to consider a word.
600 return findEndOfWord(Start + 1, Str, Length, Column + 1, Columns);
603 /// \brief Print the given string to a stream, word-wrapping it to
604 /// some number of columns in the process.
606 /// \param OS the stream to which the word-wrapping string will be
608 /// \param Str the string to word-wrap and output.
609 /// \param Columns the number of columns to word-wrap to.
610 /// \param Column the column number at which the first character of \p
611 /// Str will be printed. This will be non-zero when part of the first
612 /// line has already been printed.
613 /// \param Bold if the current text should be bold
614 /// \param Indentation the number of spaces to indent any lines beyond
616 /// \returns true if word-wrapping was required, or false if the
617 /// string fit on the first line.
618 static bool printWordWrapped(raw_ostream &OS, StringRef Str,
622 unsigned Indentation = WordWrapIndentation) {
623 const unsigned Length = std::min(Str.find('\n'), Str.size());
624 bool TextNormal = true;
626 // The string used to indent each line.
627 SmallString<16> IndentStr;
628 IndentStr.assign(Indentation, ' ');
629 bool Wrapped = false;
630 for (unsigned WordStart = 0, WordEnd; WordStart < Length;
631 WordStart = WordEnd) {
632 // Find the beginning of the next word.
633 WordStart = skipWhitespace(WordStart, Str, Length);
634 if (WordStart == Length)
637 // Find the end of this word.
638 WordEnd = findEndOfWord(WordStart, Str, Length, Column, Columns);
640 // Does this word fit on the current line?
641 unsigned WordLength = WordEnd - WordStart;
642 if (Column + WordLength < Columns) {
643 // This word fits on the current line; print it there.
648 applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
650 Column += WordLength;
654 // This word does not fit on the current line, so wrap to the next
657 OS.write(&IndentStr[0], Indentation);
658 applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
660 Column = Indentation + WordLength;
664 // Append any remaning text from the message with its existing formatting.
665 applyTemplateHighlighting(OS, Str.substr(Length), TextNormal, Bold);
667 assert(TextNormal && "Text highlighted at end of diagnostic message.");
672 TextDiagnostic::TextDiagnostic(raw_ostream &OS,
673 const LangOptions &LangOpts,
674 DiagnosticOptions *DiagOpts)
675 : DiagnosticRenderer(LangOpts, DiagOpts), OS(OS) {}
677 TextDiagnostic::~TextDiagnostic() {}
680 TextDiagnostic::emitDiagnosticMessage(SourceLocation Loc,
682 DiagnosticsEngine::Level Level,
684 ArrayRef<clang::CharSourceRange> Ranges,
685 const SourceManager *SM,
686 DiagOrStoredDiag D) {
687 uint64_t StartOfLocationInfo = OS.tell();
689 // Emit the location of this particular diagnostic.
691 emitDiagnosticLoc(Loc, PLoc, Level, Ranges, *SM);
693 if (DiagOpts->ShowColors)
696 printDiagnosticLevel(OS, Level, DiagOpts->ShowColors,
697 DiagOpts->CLFallbackMode);
698 printDiagnosticMessage(OS, Level, Message,
699 OS.tell() - StartOfLocationInfo,
700 DiagOpts->MessageLength, DiagOpts->ShowColors);
704 TextDiagnostic::printDiagnosticLevel(raw_ostream &OS,
705 DiagnosticsEngine::Level Level,
707 bool CLFallbackMode) {
709 // Print diagnostic category in bold and color
711 case DiagnosticsEngine::Ignored:
712 llvm_unreachable("Invalid diagnostic type");
713 case DiagnosticsEngine::Note: OS.changeColor(noteColor, true); break;
714 case DiagnosticsEngine::Warning: OS.changeColor(warningColor, true); break;
715 case DiagnosticsEngine::Error: OS.changeColor(errorColor, true); break;
716 case DiagnosticsEngine::Fatal: OS.changeColor(fatalColor, true); break;
721 case DiagnosticsEngine::Ignored:
722 llvm_unreachable("Invalid diagnostic type");
723 case DiagnosticsEngine::Note: OS << "note"; break;
724 case DiagnosticsEngine::Warning: OS << "warning"; break;
725 case DiagnosticsEngine::Error: OS << "error"; break;
726 case DiagnosticsEngine::Fatal: OS << "fatal error"; break;
729 // In clang-cl /fallback mode, print diagnostics as "error(clang):". This
730 // makes it more clear whether a message is coming from clang or cl.exe,
731 // and it prevents MSBuild from concluding that the build failed just because
732 // there is an "error:" in the output.
743 TextDiagnostic::printDiagnosticMessage(raw_ostream &OS,
744 DiagnosticsEngine::Level Level,
746 unsigned CurrentColumn, unsigned Columns,
750 // Print warnings, errors and fatal errors in bold, no color
752 case DiagnosticsEngine::Warning:
753 case DiagnosticsEngine::Error:
754 case DiagnosticsEngine::Fatal:
755 OS.changeColor(savedColor, true);
758 default: break; //don't bold notes
763 printWordWrapped(OS, Message, Columns, CurrentColumn, Bold);
766 applyTemplateHighlighting(OS, Message, Normal, Bold);
767 assert(Normal && "Formatting should have returned to normal");
775 /// \brief Print out the file/line/column information and include trace.
777 /// This method handlen the emission of the diagnostic location information.
778 /// This includes extracting as much location information as is present for
779 /// the diagnostic and printing it, as well as any include stack or source
780 /// ranges necessary.
781 void TextDiagnostic::emitDiagnosticLoc(SourceLocation Loc, PresumedLoc PLoc,
782 DiagnosticsEngine::Level Level,
783 ArrayRef<CharSourceRange> Ranges,
784 const SourceManager &SM) {
785 if (PLoc.isInvalid()) {
786 // At least print the file name if available:
787 FileID FID = SM.getFileID(Loc);
788 if (!FID.isInvalid()) {
789 const FileEntry* FE = SM.getFileEntryForID(FID);
790 if (FE && FE->getName()) {
799 unsigned LineNo = PLoc.getLine();
801 if (!DiagOpts->ShowLocation)
804 if (DiagOpts->ShowColors)
805 OS.changeColor(savedColor, true);
807 OS << PLoc.getFilename();
808 switch (DiagOpts->getFormat()) {
809 case DiagnosticOptions::Clang: OS << ':' << LineNo; break;
810 case DiagnosticOptions::Msvc: OS << '(' << LineNo; break;
811 case DiagnosticOptions::Vi: OS << " +" << LineNo; break;
814 if (DiagOpts->ShowColumn)
815 // Compute the column number.
816 if (unsigned ColNo = PLoc.getColumn()) {
817 if (DiagOpts->getFormat() == DiagnosticOptions::Msvc) {
824 switch (DiagOpts->getFormat()) {
825 case DiagnosticOptions::Clang:
826 case DiagnosticOptions::Vi: OS << ':'; break;
827 case DiagnosticOptions::Msvc: OS << ") : "; break;
830 if (DiagOpts->ShowSourceRanges && !Ranges.empty()) {
832 SM.getFileID(SM.getExpansionLoc(Loc));
833 bool PrintedRange = false;
835 for (ArrayRef<CharSourceRange>::const_iterator RI = Ranges.begin(),
838 // Ignore invalid ranges.
839 if (!RI->isValid()) continue;
841 SourceLocation B = SM.getExpansionLoc(RI->getBegin());
842 SourceLocation E = SM.getExpansionLoc(RI->getEnd());
844 // If the End location and the start location are the same and are a
845 // macro location, then the range was something that came from a
846 // macro expansion or _Pragma. If this is an object-like macro, the
847 // best we can do is to highlight the range. If this is a
848 // function-like macro, we'd also like to highlight the arguments.
849 if (B == E && RI->getEnd().isMacroID())
850 E = SM.getExpansionRange(RI->getEnd()).second;
852 std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(B);
853 std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(E);
855 // If the start or end of the range is in another file, just discard
857 if (BInfo.first != CaretFileID || EInfo.first != CaretFileID)
860 // Add in the length of the token, so that we cover multi-char
862 unsigned TokSize = 0;
863 if (RI->isTokenRange())
864 TokSize = Lexer::MeasureTokenLength(E, SM, LangOpts);
866 OS << '{' << SM.getLineNumber(BInfo.first, BInfo.second) << ':'
867 << SM.getColumnNumber(BInfo.first, BInfo.second) << '-'
868 << SM.getLineNumber(EInfo.first, EInfo.second) << ':'
869 << (SM.getColumnNumber(EInfo.first, EInfo.second)+TokSize)
880 void TextDiagnostic::emitBasicNote(StringRef Message) {
881 // FIXME: Emit this as a real note diagnostic.
882 // FIXME: Format an actual diagnostic rather than a hard coded string.
883 OS << "note: " << Message << "\n";
886 void TextDiagnostic::emitIncludeLocation(SourceLocation Loc,
888 const SourceManager &SM) {
889 if (DiagOpts->ShowLocation)
890 OS << "In file included from " << PLoc.getFilename() << ':'
891 << PLoc.getLine() << ":\n";
893 OS << "In included file:\n";
896 void TextDiagnostic::emitImportLocation(SourceLocation Loc, PresumedLoc PLoc,
897 StringRef ModuleName,
898 const SourceManager &SM) {
899 if (DiagOpts->ShowLocation)
900 OS << "In module '" << ModuleName << "' imported from "
901 << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
903 OS << "In module " << ModuleName << "':\n";
906 void TextDiagnostic::emitBuildingModuleLocation(SourceLocation Loc,
908 StringRef ModuleName,
909 const SourceManager &SM) {
910 if (DiagOpts->ShowLocation && PLoc.getFilename())
911 OS << "While building module '" << ModuleName << "' imported from "
912 << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
914 OS << "While building module '" << ModuleName << "':\n";
917 /// \brief Highlight a SourceRange (with ~'s) for any characters on LineNo.
918 static void highlightRange(const CharSourceRange &R,
919 unsigned LineNo, FileID FID,
920 const SourceColumnMap &map,
921 std::string &CaretLine,
922 const SourceManager &SM,
923 const LangOptions &LangOpts) {
924 if (!R.isValid()) return;
926 SourceLocation Begin = R.getBegin();
927 SourceLocation End = R.getEnd();
929 unsigned StartLineNo = SM.getExpansionLineNumber(Begin);
930 if (StartLineNo > LineNo || SM.getFileID(Begin) != FID)
931 return; // No intersection.
933 unsigned EndLineNo = SM.getExpansionLineNumber(End);
934 if (EndLineNo < LineNo || SM.getFileID(End) != FID)
935 return; // No intersection.
937 // Compute the column number of the start.
938 unsigned StartColNo = 0;
939 if (StartLineNo == LineNo) {
940 StartColNo = SM.getExpansionColumnNumber(Begin);
941 if (StartColNo) --StartColNo; // Zero base the col #.
944 // Compute the column number of the end.
945 unsigned EndColNo = map.getSourceLine().size();
946 if (EndLineNo == LineNo) {
947 EndColNo = SM.getExpansionColumnNumber(End);
949 --EndColNo; // Zero base the col #.
951 // Add in the length of the token, so that we cover multi-char tokens if
952 // this is a token range.
953 if (R.isTokenRange())
954 EndColNo += Lexer::MeasureTokenLength(End, SM, LangOpts);
956 EndColNo = CaretLine.size();
960 assert(StartColNo <= EndColNo && "Invalid range!");
962 // Check that a token range does not highlight only whitespace.
963 if (R.isTokenRange()) {
964 // Pick the first non-whitespace column.
965 while (StartColNo < map.getSourceLine().size() &&
966 (map.getSourceLine()[StartColNo] == ' ' ||
967 map.getSourceLine()[StartColNo] == '\t'))
968 StartColNo = map.startOfNextColumn(StartColNo);
970 // Pick the last non-whitespace column.
971 if (EndColNo > map.getSourceLine().size())
972 EndColNo = map.getSourceLine().size();
974 (map.getSourceLine()[EndColNo-1] == ' ' ||
975 map.getSourceLine()[EndColNo-1] == '\t'))
976 EndColNo = map.startOfPreviousColumn(EndColNo);
978 // If the start/end passed each other, then we are trying to highlight a
979 // range that just exists in whitespace, which must be some sort of other
981 assert(StartColNo <= EndColNo && "Trying to highlight whitespace??");
984 assert(StartColNo <= map.getSourceLine().size() && "Invalid range!");
985 assert(EndColNo <= map.getSourceLine().size() && "Invalid range!");
987 // Fill the range with ~'s.
988 StartColNo = map.byteToContainingColumn(StartColNo);
989 EndColNo = map.byteToContainingColumn(EndColNo);
991 assert(StartColNo <= EndColNo && "Invalid range!");
992 if (CaretLine.size() < EndColNo)
993 CaretLine.resize(EndColNo,' ');
994 std::fill(CaretLine.begin()+StartColNo,CaretLine.begin()+EndColNo,'~');
997 static std::string buildFixItInsertionLine(unsigned LineNo,
998 const SourceColumnMap &map,
999 ArrayRef<FixItHint> Hints,
1000 const SourceManager &SM,
1001 const DiagnosticOptions *DiagOpts) {
1002 std::string FixItInsertionLine;
1003 if (Hints.empty() || !DiagOpts->ShowFixits)
1004 return FixItInsertionLine;
1005 unsigned PrevHintEndCol = 0;
1007 for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1009 if (!I->CodeToInsert.empty()) {
1010 // We have an insertion hint. Determine whether the inserted
1011 // code contains no newlines and is on the same line as the caret.
1012 std::pair<FileID, unsigned> HintLocInfo
1013 = SM.getDecomposedExpansionLoc(I->RemoveRange.getBegin());
1014 if (LineNo == SM.getLineNumber(HintLocInfo.first, HintLocInfo.second) &&
1015 StringRef(I->CodeToInsert).find_first_of("\n\r") == StringRef::npos) {
1016 // Insert the new code into the line just below the code
1017 // that the user wrote.
1018 // Note: When modifying this function, be very careful about what is a
1019 // "column" (printed width, platform-dependent) and what is a
1020 // "byte offset" (SourceManager "column").
1021 unsigned HintByteOffset
1022 = SM.getColumnNumber(HintLocInfo.first, HintLocInfo.second) - 1;
1024 // The hint must start inside the source or right at the end
1025 assert(HintByteOffset < static_cast<unsigned>(map.bytes())+1);
1026 unsigned HintCol = map.byteToContainingColumn(HintByteOffset);
1028 // If we inserted a long previous hint, push this one forwards, and add
1029 // an extra space to show that this is not part of the previous
1030 // completion. This is sort of the best we can do when two hints appear
1033 // Note that if this hint is located immediately after the previous
1034 // hint, no space will be added, since the location is more important.
1035 if (HintCol < PrevHintEndCol)
1036 HintCol = PrevHintEndCol + 1;
1038 // This should NOT use HintByteOffset, because the source might have
1039 // Unicode characters in earlier columns.
1040 unsigned NewFixItLineSize = FixItInsertionLine.size() +
1041 (HintCol - PrevHintEndCol) + I->CodeToInsert.size();
1042 if (NewFixItLineSize > FixItInsertionLine.size())
1043 FixItInsertionLine.resize(NewFixItLineSize, ' ');
1045 std::copy(I->CodeToInsert.begin(), I->CodeToInsert.end(),
1046 FixItInsertionLine.end() - I->CodeToInsert.size());
1049 HintCol + llvm::sys::locale::columnWidth(I->CodeToInsert);
1051 FixItInsertionLine.clear();
1057 expandTabs(FixItInsertionLine, DiagOpts->TabStop);
1059 return FixItInsertionLine;
1062 /// \brief Emit a code snippet and caret line.
1064 /// This routine emits a single line's code snippet and caret line..
1066 /// \param Loc The location for the caret.
1067 /// \param Ranges The underlined ranges for this code snippet.
1068 /// \param Hints The FixIt hints active for this diagnostic.
1069 void TextDiagnostic::emitSnippetAndCaret(
1070 SourceLocation Loc, DiagnosticsEngine::Level Level,
1071 SmallVectorImpl<CharSourceRange>& Ranges,
1072 ArrayRef<FixItHint> Hints,
1073 const SourceManager &SM) {
1074 assert(!Loc.isInvalid() && "must have a valid source location here");
1075 assert(Loc.isFileID() && "must have a file location here");
1077 // If caret diagnostics are enabled and we have location, we want to
1078 // emit the caret. However, we only do this if the location moved
1079 // from the last diagnostic, if the last diagnostic was a note that
1080 // was part of a different warning or error diagnostic, or if the
1081 // diagnostic has ranges. We don't want to emit the same caret
1082 // multiple times if one loc has multiple diagnostics.
1083 if (!DiagOpts->ShowCarets)
1085 if (Loc == LastLoc && Ranges.empty() && Hints.empty() &&
1086 (LastLevel != DiagnosticsEngine::Note || Level == LastLevel))
1089 // Decompose the location into a FID/Offset pair.
1090 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1091 FileID FID = LocInfo.first;
1092 unsigned FileOffset = LocInfo.second;
1094 // Get information about the buffer it points into.
1095 bool Invalid = false;
1096 const char *BufStart = SM.getBufferData(FID, &Invalid).data();
1100 unsigned LineNo = SM.getLineNumber(FID, FileOffset);
1101 unsigned ColNo = SM.getColumnNumber(FID, FileOffset);
1103 // Arbitrarily stop showing snippets when the line is too long.
1104 static const size_t MaxLineLengthToPrint = 4096;
1105 if (ColNo > MaxLineLengthToPrint)
1108 // Rewind from the current position to the start of the line.
1109 const char *TokPtr = BufStart+FileOffset;
1110 const char *LineStart = TokPtr-ColNo+1; // Column # is 1-based.
1112 // Compute the line end. Scan forward from the error position to the end of
1114 const char *LineEnd = TokPtr;
1115 while (*LineEnd != '\n' && *LineEnd != '\r' && *LineEnd != '\0')
1118 // Arbitrarily stop showing snippets when the line is too long.
1119 if (size_t(LineEnd - LineStart) > MaxLineLengthToPrint)
1122 // Copy the line of code into an std::string for ease of manipulation.
1123 std::string SourceLine(LineStart, LineEnd);
1125 // Create a line for the caret that is filled with spaces that is the same
1126 // length as the line of source code.
1127 std::string CaretLine(LineEnd-LineStart, ' ');
1129 const SourceColumnMap sourceColMap(SourceLine, DiagOpts->TabStop);
1131 // Highlight all of the characters covered by Ranges with ~ characters.
1132 for (SmallVectorImpl<CharSourceRange>::iterator I = Ranges.begin(),
1135 highlightRange(*I, LineNo, FID, sourceColMap, CaretLine, SM, LangOpts);
1137 // Next, insert the caret itself.
1138 ColNo = sourceColMap.byteToContainingColumn(ColNo-1);
1139 if (CaretLine.size()<ColNo+1)
1140 CaretLine.resize(ColNo+1, ' ');
1141 CaretLine[ColNo] = '^';
1143 std::string FixItInsertionLine = buildFixItInsertionLine(LineNo,
1148 // If the source line is too long for our terminal, select only the
1149 // "interesting" source region within that line.
1150 unsigned Columns = DiagOpts->MessageLength;
1152 selectInterestingSourceRegion(SourceLine, CaretLine, FixItInsertionLine,
1153 Columns, sourceColMap);
1155 // If we are in -fdiagnostics-print-source-range-info mode, we are trying
1156 // to produce easily machine parsable output. Add a space before the
1157 // source line and the caret to make it trivial to tell the main diagnostic
1158 // line from what the user is intended to see.
1159 if (DiagOpts->ShowSourceRanges) {
1160 SourceLine = ' ' + SourceLine;
1161 CaretLine = ' ' + CaretLine;
1164 // Finally, remove any blank spaces from the end of CaretLine.
1165 while (CaretLine[CaretLine.size()-1] == ' ')
1166 CaretLine.erase(CaretLine.end()-1);
1168 // Emit what we have computed.
1169 emitSnippet(SourceLine);
1171 if (DiagOpts->ShowColors)
1172 OS.changeColor(caretColor, true);
1173 OS << CaretLine << '\n';
1174 if (DiagOpts->ShowColors)
1177 if (!FixItInsertionLine.empty()) {
1178 if (DiagOpts->ShowColors)
1179 // Print fixit line in color
1180 OS.changeColor(fixitColor, false);
1181 if (DiagOpts->ShowSourceRanges)
1183 OS << FixItInsertionLine << '\n';
1184 if (DiagOpts->ShowColors)
1188 // Print out any parseable fixit information requested by the options.
1189 emitParseableFixits(Hints, SM);
1192 void TextDiagnostic::emitSnippet(StringRef line) {
1198 std::string to_print;
1199 bool print_reversed = false;
1201 while (i<line.size()) {
1202 std::pair<SmallString<16>,bool> res
1203 = printableTextForNextCharacter(line, &i, DiagOpts->TabStop);
1204 bool was_printable = res.second;
1206 if (DiagOpts->ShowColors && was_printable == print_reversed) {
1211 if (DiagOpts->ShowColors)
1215 print_reversed = !was_printable;
1216 to_print += res.first.str();
1219 if (print_reversed && DiagOpts->ShowColors)
1222 if (print_reversed && DiagOpts->ShowColors)
1228 void TextDiagnostic::emitParseableFixits(ArrayRef<FixItHint> Hints,
1229 const SourceManager &SM) {
1230 if (!DiagOpts->ShowParseableFixits)
1233 // We follow FixItRewriter's example in not (yet) handling
1234 // fix-its in macros.
1235 for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1237 if (I->RemoveRange.isInvalid() ||
1238 I->RemoveRange.getBegin().isMacroID() ||
1239 I->RemoveRange.getEnd().isMacroID())
1243 for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1245 SourceLocation BLoc = I->RemoveRange.getBegin();
1246 SourceLocation ELoc = I->RemoveRange.getEnd();
1248 std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(BLoc);
1249 std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(ELoc);
1251 // Adjust for token ranges.
1252 if (I->RemoveRange.isTokenRange())
1253 EInfo.second += Lexer::MeasureTokenLength(ELoc, SM, LangOpts);
1255 // We specifically do not do word-wrapping or tab-expansion here,
1256 // because this is supposed to be easy to parse.
1257 PresumedLoc PLoc = SM.getPresumedLoc(BLoc);
1258 if (PLoc.isInvalid())
1262 OS.write_escaped(PLoc.getFilename());
1263 OS << "\":{" << SM.getLineNumber(BInfo.first, BInfo.second)
1264 << ':' << SM.getColumnNumber(BInfo.first, BInfo.second)
1265 << '-' << SM.getLineNumber(EInfo.first, EInfo.second)
1266 << ':' << SM.getColumnNumber(EInfo.first, EInfo.second)
1268 OS.write_escaped(I->CodeToInsert);