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/Path.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 remarkColor =
31 static const enum raw_ostream::Colors fixitColor =
33 static const enum raw_ostream::Colors caretColor =
35 static const enum raw_ostream::Colors warningColor =
37 static const enum raw_ostream::Colors templateColor =
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;
45 /// \brief Add highlights to differences in template strings.
46 static void applyTemplateHighlighting(raw_ostream &OS, StringRef Str,
47 bool &Normal, bool Bold) {
49 size_t Pos = Str.find(ToggleHighlight);
50 OS << Str.slice(0, Pos);
51 if (Pos == StringRef::npos)
54 Str = Str.substr(Pos + 1);
56 OS.changeColor(templateColor, true);
60 OS.changeColor(savedColor, true);
66 /// \brief Number of spaces to indent when word-wrapping.
67 const unsigned WordWrapIndentation = 6;
69 static int bytesSincePreviousTabOrLineBegin(StringRef SourceLine, size_t i) {
72 if (SourceLine[--i]=='\t')
79 /// \brief returns a printable representation of first item from input range
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'.
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'.
90 /// \note The index is updated to be used with a subsequent call to
91 /// printableTextForNextCharacter.
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)
98 static std::pair<SmallString<16>, bool>
99 printableTextForNextCharacter(StringRef SourceLine, size_t *i,
101 assert(i && "i must not be null");
102 assert(*i<SourceLine.size() && "must point to a valid index");
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");
113 SmallString<16> expandedTab;
114 expandedTab.assign(NumSpaces, ' ');
115 return std::make_pair(expandedTab, true);
118 unsigned char const *begin, *end;
119 begin = reinterpret_cast<unsigned char const *>(&*(SourceLine.begin() + *i));
120 end = begin + (SourceLine.size() - *i);
122 if (llvm::isLegalUTF8Sequence(begin, end)) {
124 llvm::UTF32 *cptr = &c;
125 unsigned char const *original_begin = begin;
126 unsigned char const *cp_end =
127 begin + llvm::getNumBytesForUTF8(SourceLine[*i]);
129 llvm::ConversionResult res = llvm::ConvertUTF8toUTF32(
130 &begin, cp_end, &cptr, cptr + 1, llvm::strictConversion);
132 assert(llvm::conversionOK == res);
133 assert(0 < begin-original_begin
134 && "we must be further along in the string now");
135 *i += begin-original_begin;
137 if (!llvm::sys::locale::isPrint(c)) {
138 // If next character is valid UTF-8, but not printable
139 SmallString<16> expandedCP("<U+>");
141 expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(c%16));
144 while (expandedCP.size() < 8)
145 expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(0));
146 return std::make_pair(expandedCP, false);
149 // If next character is valid UTF-8, and printable
150 return std::make_pair(SmallString<16>(original_begin, cp_end), true);
154 // If next byte is not valid UTF-8 (and therefore not printable)
155 SmallString<16> expandedByte("<XX>");
156 unsigned char byte = SourceLine[*i];
157 expandedByte[1] = llvm::hexdigit(byte / 16);
158 expandedByte[2] = llvm::hexdigit(byte % 16);
160 return std::make_pair(expandedByte, false);
163 static void expandTabs(std::string &SourceLine, unsigned TabStop) {
164 size_t i = SourceLine.size();
167 if (SourceLine[i]!='\t')
170 std::pair<SmallString<16>,bool> res
171 = printableTextForNextCharacter(SourceLine, &tmp_i, TabStop);
172 SourceLine.replace(i, 1, res.first.c_str());
176 /// This function takes a raw source line and produces a mapping from the bytes
177 /// of the printable representation of the line to the columns those printable
178 /// characters will appear at (numbering the first column as 0).
180 /// If a byte 'i' corresponds to multiple columns (e.g. the byte contains a tab
181 /// character) then the array will map that byte to the first column the
182 /// tab appears at and the next value in the map will have been incremented
185 /// If a byte is the first in a sequence of bytes that together map to a single
186 /// entity in the output, then the array will map that byte to the appropriate
187 /// column while the subsequent bytes will be -1.
189 /// The last element in the array does not correspond to any byte in the input
190 /// and instead is the number of columns needed to display the source
192 /// example: (given a tabstop of 8)
194 /// "a \t \u3042" -> {0,1,2,8,9,-1,-1,11}
196 /// (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
198 static void byteToColumn(StringRef SourceLine, unsigned TabStop,
199 SmallVectorImpl<int> &out) {
202 if (SourceLine.empty()) {
207 out.resize(SourceLine.size()+1, -1);
211 while (i<SourceLine.size()) {
213 std::pair<SmallString<16>,bool> res
214 = printableTextForNextCharacter(SourceLine, &i, TabStop);
215 columns += llvm::sys::locale::columnWidth(res.first);
217 out.back() = columns;
220 /// This function takes a raw source line and produces a mapping from columns
221 /// to the byte of the source line that produced the character displaying at
222 /// that column. This is the inverse of the mapping produced by byteToColumn()
224 /// The last element in the array is the number of bytes in the source string
226 /// example: (given a tabstop of 8)
228 /// "a \t \u3042" -> {0,1,2,-1,-1,-1,-1,-1,3,4,-1,7}
230 /// (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
232 static void columnToByte(StringRef SourceLine, unsigned TabStop,
233 SmallVectorImpl<int> &out) {
236 if (SourceLine.empty()) {
243 while (i<SourceLine.size()) {
244 out.resize(columns+1, -1);
246 std::pair<SmallString<16>,bool> res
247 = printableTextForNextCharacter(SourceLine, &i, TabStop);
248 columns += llvm::sys::locale::columnWidth(res.first);
250 out.resize(columns+1, -1);
255 struct SourceColumnMap {
256 SourceColumnMap(StringRef SourceLine, unsigned TabStop)
257 : m_SourceLine(SourceLine) {
259 ::byteToColumn(SourceLine, TabStop, m_byteToColumn);
260 ::columnToByte(SourceLine, TabStop, m_columnToByte);
262 assert(m_byteToColumn.size()==SourceLine.size()+1);
263 assert(0 < m_byteToColumn.size() && 0 < m_columnToByte.size());
264 assert(m_byteToColumn.size()
265 == static_cast<unsigned>(m_columnToByte.back()+1));
266 assert(static_cast<unsigned>(m_byteToColumn.back()+1)
267 == m_columnToByte.size());
269 int columns() const { return m_byteToColumn.back(); }
270 int bytes() const { return m_columnToByte.back(); }
272 /// \brief Map a byte to the column which it is at the start of, or return -1
273 /// if it is not at the start of a column (for a UTF-8 trailing byte).
274 int byteToColumn(int n) const {
275 assert(0<=n && n<static_cast<int>(m_byteToColumn.size()));
276 return m_byteToColumn[n];
279 /// \brief Map a byte to the first column which contains it.
280 int byteToContainingColumn(int N) const {
281 assert(0 <= N && N < static_cast<int>(m_byteToColumn.size()));
282 while (m_byteToColumn[N] == -1)
284 return m_byteToColumn[N];
287 /// \brief Map a column to the byte which starts the column, or return -1 if
288 /// the column the second or subsequent column of an expanded tab or similar
289 /// multi-column entity.
290 int columnToByte(int n) const {
291 assert(0<=n && n<static_cast<int>(m_columnToByte.size()));
292 return m_columnToByte[n];
295 /// \brief Map from a byte index to the next byte which starts a column.
296 int startOfNextColumn(int N) const {
297 assert(0 <= N && N < static_cast<int>(m_byteToColumn.size() - 1));
298 while (byteToColumn(++N) == -1) {}
302 /// \brief Map from a byte index to the previous byte which starts a column.
303 int startOfPreviousColumn(int N) const {
304 assert(0 < N && N < static_cast<int>(m_byteToColumn.size()));
305 while (byteToColumn(--N) == -1) {}
309 StringRef getSourceLine() const {
314 const std::string m_SourceLine;
315 SmallVector<int,200> m_byteToColumn;
316 SmallVector<int,200> m_columnToByte;
318 } // end anonymous namespace
320 /// \brief When the source code line we want to print is too long for
321 /// the terminal, select the "interesting" region.
322 static void selectInterestingSourceRegion(std::string &SourceLine,
323 std::string &CaretLine,
324 std::string &FixItInsertionLine,
326 const SourceColumnMap &map) {
327 unsigned CaretColumns = CaretLine.size();
328 unsigned FixItColumns = llvm::sys::locale::columnWidth(FixItInsertionLine);
329 unsigned MaxColumns = std::max(static_cast<unsigned>(map.columns()),
330 std::max(CaretColumns, FixItColumns));
331 // if the number of columns is less than the desired number we're done
332 if (MaxColumns <= Columns)
335 // No special characters are allowed in CaretLine.
336 assert(CaretLine.end() ==
337 std::find_if(CaretLine.begin(), CaretLine.end(),
338 [](char c) { return c < ' ' || '~' < c; }));
340 // Find the slice that we need to display the full caret line
342 unsigned CaretStart = 0, CaretEnd = CaretLine.size();
343 for (; CaretStart != CaretEnd; ++CaretStart)
344 if (!isWhitespace(CaretLine[CaretStart]))
347 for (; CaretEnd != CaretStart; --CaretEnd)
348 if (!isWhitespace(CaretLine[CaretEnd - 1]))
351 // caret has already been inserted into CaretLine so the above whitespace
352 // check is guaranteed to include the caret
354 // If we have a fix-it line, make sure the slice includes all of the
355 // fix-it information.
356 if (!FixItInsertionLine.empty()) {
357 unsigned FixItStart = 0, FixItEnd = FixItInsertionLine.size();
358 for (; FixItStart != FixItEnd; ++FixItStart)
359 if (!isWhitespace(FixItInsertionLine[FixItStart]))
362 for (; FixItEnd != FixItStart; --FixItEnd)
363 if (!isWhitespace(FixItInsertionLine[FixItEnd - 1]))
366 // We can safely use the byte offset FixItStart as the column offset
367 // because the characters up until FixItStart are all ASCII whitespace
369 unsigned FixItStartCol = FixItStart;
371 = llvm::sys::locale::columnWidth(FixItInsertionLine.substr(0, FixItEnd));
373 CaretStart = std::min(FixItStartCol, CaretStart);
374 CaretEnd = std::max(FixItEndCol, CaretEnd);
377 // CaretEnd may have been set at the middle of a character
378 // If it's not at a character's first column then advance it past the current
380 while (static_cast<int>(CaretEnd) < map.columns() &&
381 -1 == map.columnToByte(CaretEnd))
384 assert((static_cast<int>(CaretStart) > map.columns() ||
385 -1!=map.columnToByte(CaretStart)) &&
386 "CaretStart must not point to a column in the middle of a source"
388 assert((static_cast<int>(CaretEnd) > map.columns() ||
389 -1!=map.columnToByte(CaretEnd)) &&
390 "CaretEnd must not point to a column in the middle of a source line"
393 // CaretLine[CaretStart, CaretEnd) contains all of the interesting
394 // parts of the caret line. While this slice is smaller than the
395 // number of columns we have, try to grow the slice to encompass
398 unsigned SourceStart = map.columnToByte(std::min<unsigned>(CaretStart,
400 unsigned SourceEnd = map.columnToByte(std::min<unsigned>(CaretEnd,
403 unsigned CaretColumnsOutsideSource = CaretEnd-CaretStart
404 - (map.byteToColumn(SourceEnd)-map.byteToColumn(SourceStart));
406 char const *front_ellipse = " ...";
407 char const *front_space = " ";
408 char const *back_ellipse = "...";
409 unsigned ellipses_space = strlen(front_ellipse) + strlen(back_ellipse);
411 unsigned TargetColumns = Columns;
412 // Give us extra room for the ellipses
413 // and any of the caret line that extends past the source
414 if (TargetColumns > ellipses_space+CaretColumnsOutsideSource)
415 TargetColumns -= ellipses_space+CaretColumnsOutsideSource;
417 while (SourceStart>0 || SourceEnd<SourceLine.size()) {
418 bool ExpandedRegion = false;
421 unsigned NewStart = map.startOfPreviousColumn(SourceStart);
423 // Skip over any whitespace we see here; we're looking for
424 // another bit of interesting text.
425 // FIXME: Detect non-ASCII whitespace characters too.
426 while (NewStart && isWhitespace(SourceLine[NewStart]))
427 NewStart = map.startOfPreviousColumn(NewStart);
429 // Skip over this bit of "interesting" text.
431 unsigned Prev = map.startOfPreviousColumn(NewStart);
432 if (isWhitespace(SourceLine[Prev]))
437 assert(map.byteToColumn(NewStart) != -1);
438 unsigned NewColumns = map.byteToColumn(SourceEnd) -
439 map.byteToColumn(NewStart);
440 if (NewColumns <= TargetColumns) {
441 SourceStart = NewStart;
442 ExpandedRegion = true;
446 if (SourceEnd<SourceLine.size()) {
447 unsigned NewEnd = map.startOfNextColumn(SourceEnd);
449 // Skip over any whitespace we see here; we're looking for
450 // another bit of interesting text.
451 // FIXME: Detect non-ASCII whitespace characters too.
452 while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
453 NewEnd = map.startOfNextColumn(NewEnd);
455 // Skip over this bit of "interesting" text.
456 while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
457 NewEnd = map.startOfNextColumn(NewEnd);
459 assert(map.byteToColumn(NewEnd) != -1);
460 unsigned NewColumns = map.byteToColumn(NewEnd) -
461 map.byteToColumn(SourceStart);
462 if (NewColumns <= TargetColumns) {
464 ExpandedRegion = true;
472 CaretStart = map.byteToColumn(SourceStart);
473 CaretEnd = map.byteToColumn(SourceEnd) + CaretColumnsOutsideSource;
475 // [CaretStart, CaretEnd) is the slice we want. Update the various
476 // output lines to show only this slice, with two-space padding
477 // before the lines so that it looks nicer.
479 assert(CaretStart!=(unsigned)-1 && CaretEnd!=(unsigned)-1 &&
480 SourceStart!=(unsigned)-1 && SourceEnd!=(unsigned)-1);
481 assert(SourceStart <= SourceEnd);
482 assert(CaretStart <= CaretEnd);
484 unsigned BackColumnsRemoved
485 = map.byteToColumn(SourceLine.size())-map.byteToColumn(SourceEnd);
486 unsigned FrontColumnsRemoved = CaretStart;
487 unsigned ColumnsKept = CaretEnd-CaretStart;
489 // We checked up front that the line needed truncation
490 assert(FrontColumnsRemoved+ColumnsKept+BackColumnsRemoved > Columns);
492 // The line needs some truncation, and we'd prefer to keep the front
493 // if possible, so remove the back
494 if (BackColumnsRemoved > strlen(back_ellipse))
495 SourceLine.replace(SourceEnd, std::string::npos, back_ellipse);
497 // If that's enough then we're done
498 if (FrontColumnsRemoved+ColumnsKept <= Columns)
501 // Otherwise remove the front as well
502 if (FrontColumnsRemoved > strlen(front_ellipse)) {
503 SourceLine.replace(0, SourceStart, front_ellipse);
504 CaretLine.replace(0, CaretStart, front_space);
505 if (!FixItInsertionLine.empty())
506 FixItInsertionLine.replace(0, CaretStart, front_space);
510 /// \brief Skip over whitespace in the string, starting at the given
513 /// \returns The index of the first non-whitespace character that is
514 /// greater than or equal to Idx or, if no such character exists,
515 /// returns the end of the string.
516 static unsigned skipWhitespace(unsigned Idx, StringRef Str, unsigned Length) {
517 while (Idx < Length && isWhitespace(Str[Idx]))
522 /// \brief If the given character is the start of some kind of
523 /// balanced punctuation (e.g., quotes or parentheses), return the
524 /// character that will terminate the punctuation.
526 /// \returns The ending punctuation character, if any, or the NULL
527 /// character if the input character does not start any punctuation.
528 static inline char findMatchingPunctuation(char c) {
530 case '\'': return '\'';
531 case '`': return '\'';
532 case '"': return '"';
533 case '(': return ')';
534 case '[': return ']';
535 case '{': return '}';
542 /// \brief Find the end of the word starting at the given offset
545 /// \returns the index pointing one character past the end of the
547 static unsigned findEndOfWord(unsigned Start, StringRef Str,
548 unsigned Length, unsigned Column,
550 assert(Start < Str.size() && "Invalid start position!");
551 unsigned End = Start + 1;
553 // If we are already at the end of the string, take that as the word.
554 if (End == Str.size())
557 // Determine if the start of the string is actually opening
558 // punctuation, e.g., a quote or parentheses.
559 char EndPunct = findMatchingPunctuation(Str[Start]);
561 // This is a normal word. Just find the first space character.
562 while (End < Length && !isWhitespace(Str[End]))
567 // We have the start of a balanced punctuation sequence (quotes,
568 // parentheses, etc.). Determine the full sequence is.
569 SmallString<16> PunctuationEndStack;
570 PunctuationEndStack.push_back(EndPunct);
571 while (End < Length && !PunctuationEndStack.empty()) {
572 if (Str[End] == PunctuationEndStack.back())
573 PunctuationEndStack.pop_back();
574 else if (char SubEndPunct = findMatchingPunctuation(Str[End]))
575 PunctuationEndStack.push_back(SubEndPunct);
580 // Find the first space character after the punctuation ended.
581 while (End < Length && !isWhitespace(Str[End]))
584 unsigned PunctWordLength = End - Start;
585 if (// If the word fits on this line
586 Column + PunctWordLength <= Columns ||
587 // ... or the word is "short enough" to take up the next line
588 // without too much ugly white space
589 PunctWordLength < Columns/3)
590 return End; // Take the whole thing as a single "word".
592 // The whole quoted/parenthesized string is too long to print as a
593 // single "word". Instead, find the "word" that starts just after
594 // the punctuation and use that end-point instead. This will recurse
595 // until it finds something small enough to consider a word.
596 return findEndOfWord(Start + 1, Str, Length, Column + 1, Columns);
599 /// \brief Print the given string to a stream, word-wrapping it to
600 /// some number of columns in the process.
602 /// \param OS the stream to which the word-wrapping string will be
604 /// \param Str the string to word-wrap and output.
605 /// \param Columns the number of columns to word-wrap to.
606 /// \param Column the column number at which the first character of \p
607 /// Str will be printed. This will be non-zero when part of the first
608 /// line has already been printed.
609 /// \param Bold if the current text should be bold
610 /// \param Indentation the number of spaces to indent any lines beyond
612 /// \returns true if word-wrapping was required, or false if the
613 /// string fit on the first line.
614 static bool printWordWrapped(raw_ostream &OS, StringRef Str,
618 unsigned Indentation = WordWrapIndentation) {
619 const unsigned Length = std::min(Str.find('\n'), Str.size());
620 bool TextNormal = true;
622 // The string used to indent each line.
623 SmallString<16> IndentStr;
624 IndentStr.assign(Indentation, ' ');
625 bool Wrapped = false;
626 for (unsigned WordStart = 0, WordEnd; WordStart < Length;
627 WordStart = WordEnd) {
628 // Find the beginning of the next word.
629 WordStart = skipWhitespace(WordStart, Str, Length);
630 if (WordStart == Length)
633 // Find the end of this word.
634 WordEnd = findEndOfWord(WordStart, Str, Length, Column, Columns);
636 // Does this word fit on the current line?
637 unsigned WordLength = WordEnd - WordStart;
638 if (Column + WordLength < Columns) {
639 // This word fits on the current line; print it there.
644 applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
646 Column += WordLength;
650 // This word does not fit on the current line, so wrap to the next
653 OS.write(&IndentStr[0], Indentation);
654 applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
656 Column = Indentation + WordLength;
660 // Append any remaning text from the message with its existing formatting.
661 applyTemplateHighlighting(OS, Str.substr(Length), TextNormal, Bold);
663 assert(TextNormal && "Text highlighted at end of diagnostic message.");
668 TextDiagnostic::TextDiagnostic(raw_ostream &OS,
669 const LangOptions &LangOpts,
670 DiagnosticOptions *DiagOpts)
671 : DiagnosticRenderer(LangOpts, DiagOpts), OS(OS) {}
673 TextDiagnostic::~TextDiagnostic() {}
676 TextDiagnostic::emitDiagnosticMessage(SourceLocation Loc,
678 DiagnosticsEngine::Level Level,
680 ArrayRef<clang::CharSourceRange> Ranges,
681 const SourceManager *SM,
682 DiagOrStoredDiag D) {
683 uint64_t StartOfLocationInfo = OS.tell();
685 // Emit the location of this particular diagnostic.
687 emitDiagnosticLoc(Loc, PLoc, Level, Ranges, *SM);
689 if (DiagOpts->ShowColors)
692 printDiagnosticLevel(OS, Level, DiagOpts->ShowColors,
693 DiagOpts->CLFallbackMode);
694 printDiagnosticMessage(OS,
695 /*IsSupplemental*/ Level == DiagnosticsEngine::Note,
696 Message, OS.tell() - StartOfLocationInfo,
697 DiagOpts->MessageLength, DiagOpts->ShowColors);
701 TextDiagnostic::printDiagnosticLevel(raw_ostream &OS,
702 DiagnosticsEngine::Level Level,
704 bool CLFallbackMode) {
706 // Print diagnostic category in bold and color
708 case DiagnosticsEngine::Ignored:
709 llvm_unreachable("Invalid diagnostic type");
710 case DiagnosticsEngine::Note: OS.changeColor(noteColor, true); break;
711 case DiagnosticsEngine::Remark: OS.changeColor(remarkColor, true); break;
712 case DiagnosticsEngine::Warning: OS.changeColor(warningColor, true); break;
713 case DiagnosticsEngine::Error: OS.changeColor(errorColor, true); break;
714 case DiagnosticsEngine::Fatal: OS.changeColor(fatalColor, true); break;
719 case DiagnosticsEngine::Ignored:
720 llvm_unreachable("Invalid diagnostic type");
721 case DiagnosticsEngine::Note: OS << "note"; break;
722 case DiagnosticsEngine::Remark: OS << "remark"; break;
723 case DiagnosticsEngine::Warning: OS << "warning"; break;
724 case DiagnosticsEngine::Error: OS << "error"; break;
725 case DiagnosticsEngine::Fatal: OS << "fatal error"; break;
728 // In clang-cl /fallback mode, print diagnostics as "error(clang):". This
729 // makes it more clear whether a message is coming from clang or cl.exe,
730 // and it prevents MSBuild from concluding that the build failed just because
731 // there is an "error:" in the output.
742 void TextDiagnostic::printDiagnosticMessage(raw_ostream &OS,
745 unsigned CurrentColumn,
746 unsigned Columns, bool ShowColors) {
748 if (ShowColors && !IsSupplemental) {
749 // Print primary diagnostic messages in bold and without color, to visually
750 // indicate the transition from continuation notes and other output.
751 OS.changeColor(savedColor, true);
756 printWordWrapped(OS, Message, Columns, CurrentColumn, Bold);
759 applyTemplateHighlighting(OS, Message, Normal, Bold);
760 assert(Normal && "Formatting should have returned to normal");
768 void TextDiagnostic::emitFilename(StringRef Filename, const SourceManager &SM) {
769 SmallVector<char, 128> AbsoluteFilename;
770 if (DiagOpts->AbsolutePath) {
771 const DirectoryEntry *Dir = SM.getFileManager().getDirectory(
772 llvm::sys::path::parent_path(Filename));
774 StringRef DirName = SM.getFileManager().getCanonicalName(Dir);
775 llvm::sys::path::append(AbsoluteFilename, DirName,
776 llvm::sys::path::filename(Filename));
777 Filename = StringRef(AbsoluteFilename.data(), AbsoluteFilename.size());
784 /// \brief Print out the file/line/column information and include trace.
786 /// This method handlen the emission of the diagnostic location information.
787 /// This includes extracting as much location information as is present for
788 /// the diagnostic and printing it, as well as any include stack or source
789 /// ranges necessary.
790 void TextDiagnostic::emitDiagnosticLoc(SourceLocation Loc, PresumedLoc PLoc,
791 DiagnosticsEngine::Level Level,
792 ArrayRef<CharSourceRange> Ranges,
793 const SourceManager &SM) {
794 if (PLoc.isInvalid()) {
795 // At least print the file name if available:
796 FileID FID = SM.getFileID(Loc);
798 const FileEntry* FE = SM.getFileEntryForID(FID);
799 if (FE && FE->isValid()) {
800 emitFilename(FE->getName(), SM);
808 unsigned LineNo = PLoc.getLine();
810 if (!DiagOpts->ShowLocation)
813 if (DiagOpts->ShowColors)
814 OS.changeColor(savedColor, true);
816 emitFilename(PLoc.getFilename(), SM);
817 switch (DiagOpts->getFormat()) {
818 case DiagnosticOptions::Clang: OS << ':' << LineNo; break;
819 case DiagnosticOptions::MSVC: OS << '(' << LineNo; break;
820 case DiagnosticOptions::Vi: OS << " +" << LineNo; break;
823 if (DiagOpts->ShowColumn)
824 // Compute the column number.
825 if (unsigned ColNo = PLoc.getColumn()) {
826 if (DiagOpts->getFormat() == DiagnosticOptions::MSVC) {
828 // Visual Studio 2010 or earlier expects column number to be off by one
829 if (LangOpts.MSCompatibilityVersion &&
830 !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2012))
836 switch (DiagOpts->getFormat()) {
837 case DiagnosticOptions::Clang:
838 case DiagnosticOptions::Vi: OS << ':'; break;
839 case DiagnosticOptions::MSVC:
840 // MSVC2013 and before print 'file(4) : error'. MSVC2015 gets rid of the
841 // space and prints 'file(4): error'.
843 if (LangOpts.MSCompatibilityVersion &&
844 !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2015))
850 if (DiagOpts->ShowSourceRanges && !Ranges.empty()) {
852 SM.getFileID(SM.getExpansionLoc(Loc));
853 bool PrintedRange = false;
855 for (ArrayRef<CharSourceRange>::const_iterator RI = Ranges.begin(),
858 // Ignore invalid ranges.
859 if (!RI->isValid()) continue;
861 SourceLocation B = SM.getExpansionLoc(RI->getBegin());
862 SourceLocation E = SM.getExpansionLoc(RI->getEnd());
864 // If the End location and the start location are the same and are a
865 // macro location, then the range was something that came from a
866 // macro expansion or _Pragma. If this is an object-like macro, the
867 // best we can do is to highlight the range. If this is a
868 // function-like macro, we'd also like to highlight the arguments.
869 if (B == E && RI->getEnd().isMacroID())
870 E = SM.getExpansionRange(RI->getEnd()).second;
872 std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(B);
873 std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(E);
875 // If the start or end of the range is in another file, just discard
877 if (BInfo.first != CaretFileID || EInfo.first != CaretFileID)
880 // Add in the length of the token, so that we cover multi-char
882 unsigned TokSize = 0;
883 if (RI->isTokenRange())
884 TokSize = Lexer::MeasureTokenLength(E, SM, LangOpts);
886 OS << '{' << SM.getLineNumber(BInfo.first, BInfo.second) << ':'
887 << SM.getColumnNumber(BInfo.first, BInfo.second) << '-'
888 << SM.getLineNumber(EInfo.first, EInfo.second) << ':'
889 << (SM.getColumnNumber(EInfo.first, EInfo.second)+TokSize)
900 void TextDiagnostic::emitIncludeLocation(SourceLocation Loc,
902 const SourceManager &SM) {
903 if (DiagOpts->ShowLocation && PLoc.isValid())
904 OS << "In file included from " << PLoc.getFilename() << ':'
905 << PLoc.getLine() << ":\n";
907 OS << "In included file:\n";
910 void TextDiagnostic::emitImportLocation(SourceLocation Loc, PresumedLoc PLoc,
911 StringRef ModuleName,
912 const SourceManager &SM) {
913 if (DiagOpts->ShowLocation && PLoc.isValid())
914 OS << "In module '" << ModuleName << "' imported from "
915 << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
917 OS << "In module '" << ModuleName << "':\n";
920 void TextDiagnostic::emitBuildingModuleLocation(SourceLocation Loc,
922 StringRef ModuleName,
923 const SourceManager &SM) {
924 if (DiagOpts->ShowLocation && PLoc.isValid())
925 OS << "While building module '" << ModuleName << "' imported from "
926 << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
928 OS << "While building module '" << ModuleName << "':\n";
931 /// \brief Highlight a SourceRange (with ~'s) for any characters on LineNo.
932 static void highlightRange(const CharSourceRange &R,
933 unsigned LineNo, FileID FID,
934 const SourceColumnMap &map,
935 std::string &CaretLine,
936 const SourceManager &SM,
937 const LangOptions &LangOpts) {
938 if (!R.isValid()) return;
940 SourceLocation Begin = R.getBegin();
941 SourceLocation End = R.getEnd();
943 unsigned StartLineNo = SM.getExpansionLineNumber(Begin);
944 if (StartLineNo > LineNo || SM.getFileID(Begin) != FID)
945 return; // No intersection.
947 unsigned EndLineNo = SM.getExpansionLineNumber(End);
948 if (EndLineNo < LineNo || SM.getFileID(End) != FID)
949 return; // No intersection.
951 // Compute the column number of the start.
952 unsigned StartColNo = 0;
953 if (StartLineNo == LineNo) {
954 StartColNo = SM.getExpansionColumnNumber(Begin);
955 if (StartColNo) --StartColNo; // Zero base the col #.
958 // Compute the column number of the end.
959 unsigned EndColNo = map.getSourceLine().size();
960 if (EndLineNo == LineNo) {
961 EndColNo = SM.getExpansionColumnNumber(End);
963 --EndColNo; // Zero base the col #.
965 // Add in the length of the token, so that we cover multi-char tokens if
966 // this is a token range.
967 if (R.isTokenRange())
968 EndColNo += Lexer::MeasureTokenLength(End, SM, LangOpts);
970 EndColNo = CaretLine.size();
974 assert(StartColNo <= EndColNo && "Invalid range!");
976 // Check that a token range does not highlight only whitespace.
977 if (R.isTokenRange()) {
978 // Pick the first non-whitespace column.
979 while (StartColNo < map.getSourceLine().size() &&
980 (map.getSourceLine()[StartColNo] == ' ' ||
981 map.getSourceLine()[StartColNo] == '\t'))
982 StartColNo = map.startOfNextColumn(StartColNo);
984 // Pick the last non-whitespace column.
985 if (EndColNo > map.getSourceLine().size())
986 EndColNo = map.getSourceLine().size();
988 (map.getSourceLine()[EndColNo-1] == ' ' ||
989 map.getSourceLine()[EndColNo-1] == '\t'))
990 EndColNo = map.startOfPreviousColumn(EndColNo);
992 // If the start/end passed each other, then we are trying to highlight a
993 // range that just exists in whitespace, which must be some sort of other
995 assert(StartColNo <= EndColNo && "Trying to highlight whitespace??");
998 assert(StartColNo <= map.getSourceLine().size() && "Invalid range!");
999 assert(EndColNo <= map.getSourceLine().size() && "Invalid range!");
1001 // Fill the range with ~'s.
1002 StartColNo = map.byteToContainingColumn(StartColNo);
1003 EndColNo = map.byteToContainingColumn(EndColNo);
1005 assert(StartColNo <= EndColNo && "Invalid range!");
1006 if (CaretLine.size() < EndColNo)
1007 CaretLine.resize(EndColNo,' ');
1008 std::fill(CaretLine.begin()+StartColNo,CaretLine.begin()+EndColNo,'~');
1011 static std::string buildFixItInsertionLine(unsigned LineNo,
1012 const SourceColumnMap &map,
1013 ArrayRef<FixItHint> Hints,
1014 const SourceManager &SM,
1015 const DiagnosticOptions *DiagOpts) {
1016 std::string FixItInsertionLine;
1017 if (Hints.empty() || !DiagOpts->ShowFixits)
1018 return FixItInsertionLine;
1019 unsigned PrevHintEndCol = 0;
1021 for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1023 if (!I->CodeToInsert.empty()) {
1024 // We have an insertion hint. Determine whether the inserted
1025 // code contains no newlines and is on the same line as the caret.
1026 std::pair<FileID, unsigned> HintLocInfo
1027 = SM.getDecomposedExpansionLoc(I->RemoveRange.getBegin());
1028 if (LineNo == SM.getLineNumber(HintLocInfo.first, HintLocInfo.second) &&
1029 StringRef(I->CodeToInsert).find_first_of("\n\r") == StringRef::npos) {
1030 // Insert the new code into the line just below the code
1031 // that the user wrote.
1032 // Note: When modifying this function, be very careful about what is a
1033 // "column" (printed width, platform-dependent) and what is a
1034 // "byte offset" (SourceManager "column").
1035 unsigned HintByteOffset
1036 = SM.getColumnNumber(HintLocInfo.first, HintLocInfo.second) - 1;
1038 // The hint must start inside the source or right at the end
1039 assert(HintByteOffset < static_cast<unsigned>(map.bytes())+1);
1040 unsigned HintCol = map.byteToContainingColumn(HintByteOffset);
1042 // If we inserted a long previous hint, push this one forwards, and add
1043 // an extra space to show that this is not part of the previous
1044 // completion. This is sort of the best we can do when two hints appear
1047 // Note that if this hint is located immediately after the previous
1048 // hint, no space will be added, since the location is more important.
1049 if (HintCol < PrevHintEndCol)
1050 HintCol = PrevHintEndCol + 1;
1052 // This should NOT use HintByteOffset, because the source might have
1053 // Unicode characters in earlier columns.
1054 unsigned NewFixItLineSize = FixItInsertionLine.size() +
1055 (HintCol - PrevHintEndCol) + I->CodeToInsert.size();
1056 if (NewFixItLineSize > FixItInsertionLine.size())
1057 FixItInsertionLine.resize(NewFixItLineSize, ' ');
1059 std::copy(I->CodeToInsert.begin(), I->CodeToInsert.end(),
1060 FixItInsertionLine.end() - I->CodeToInsert.size());
1063 HintCol + llvm::sys::locale::columnWidth(I->CodeToInsert);
1065 FixItInsertionLine.clear();
1071 expandTabs(FixItInsertionLine, DiagOpts->TabStop);
1073 return FixItInsertionLine;
1076 /// \brief Emit a code snippet and caret line.
1078 /// This routine emits a single line's code snippet and caret line..
1080 /// \param Loc The location for the caret.
1081 /// \param Ranges The underlined ranges for this code snippet.
1082 /// \param Hints The FixIt hints active for this diagnostic.
1083 void TextDiagnostic::emitSnippetAndCaret(
1084 SourceLocation Loc, DiagnosticsEngine::Level Level,
1085 SmallVectorImpl<CharSourceRange>& Ranges,
1086 ArrayRef<FixItHint> Hints,
1087 const SourceManager &SM) {
1088 assert(Loc.isValid() && "must have a valid source location here");
1089 assert(Loc.isFileID() && "must have a file location here");
1091 // If caret diagnostics are enabled and we have location, we want to
1092 // emit the caret. However, we only do this if the location moved
1093 // from the last diagnostic, if the last diagnostic was a note that
1094 // was part of a different warning or error diagnostic, or if the
1095 // diagnostic has ranges. We don't want to emit the same caret
1096 // multiple times if one loc has multiple diagnostics.
1097 if (!DiagOpts->ShowCarets)
1099 if (Loc == LastLoc && Ranges.empty() && Hints.empty() &&
1100 (LastLevel != DiagnosticsEngine::Note || Level == LastLevel))
1103 // Decompose the location into a FID/Offset pair.
1104 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1105 FileID FID = LocInfo.first;
1106 unsigned FileOffset = LocInfo.second;
1108 // Get information about the buffer it points into.
1109 bool Invalid = false;
1110 StringRef BufData = SM.getBufferData(FID, &Invalid);
1114 const char *BufStart = BufData.data();
1115 const char *BufEnd = BufStart + BufData.size();
1117 unsigned LineNo = SM.getLineNumber(FID, FileOffset);
1118 unsigned ColNo = SM.getColumnNumber(FID, FileOffset);
1120 // Arbitrarily stop showing snippets when the line is too long.
1121 static const size_t MaxLineLengthToPrint = 4096;
1122 if (ColNo > MaxLineLengthToPrint)
1125 // Rewind from the current position to the start of the line.
1126 const char *TokPtr = BufStart+FileOffset;
1127 const char *LineStart = TokPtr-ColNo+1; // Column # is 1-based.
1129 // Compute the line end. Scan forward from the error position to the end of
1131 const char *LineEnd = TokPtr;
1132 while (*LineEnd != '\n' && *LineEnd != '\r' && LineEnd != BufEnd)
1135 // Arbitrarily stop showing snippets when the line is too long.
1136 if (size_t(LineEnd - LineStart) > MaxLineLengthToPrint)
1139 // Trim trailing null-bytes.
1140 StringRef Line(LineStart, LineEnd - LineStart);
1141 while (Line.size() > ColNo && Line.back() == '\0')
1142 Line = Line.drop_back();
1144 // Copy the line of code into an std::string for ease of manipulation.
1145 std::string SourceLine(Line.begin(), Line.end());
1147 // Build the byte to column map.
1148 const SourceColumnMap sourceColMap(SourceLine, DiagOpts->TabStop);
1150 // Create a line for the caret that is filled with spaces that is the same
1151 // number of columns as the line of source code.
1152 std::string CaretLine(sourceColMap.columns(), ' ');
1154 // Highlight all of the characters covered by Ranges with ~ characters.
1155 for (SmallVectorImpl<CharSourceRange>::iterator I = Ranges.begin(),
1158 highlightRange(*I, LineNo, FID, sourceColMap, CaretLine, SM, LangOpts);
1160 // Next, insert the caret itself.
1161 ColNo = sourceColMap.byteToContainingColumn(ColNo-1);
1162 if (CaretLine.size()<ColNo+1)
1163 CaretLine.resize(ColNo+1, ' ');
1164 CaretLine[ColNo] = '^';
1166 std::string FixItInsertionLine = buildFixItInsertionLine(LineNo,
1171 // If the source line is too long for our terminal, select only the
1172 // "interesting" source region within that line.
1173 unsigned Columns = DiagOpts->MessageLength;
1175 selectInterestingSourceRegion(SourceLine, CaretLine, FixItInsertionLine,
1176 Columns, sourceColMap);
1178 // If we are in -fdiagnostics-print-source-range-info mode, we are trying
1179 // to produce easily machine parsable output. Add a space before the
1180 // source line and the caret to make it trivial to tell the main diagnostic
1181 // line from what the user is intended to see.
1182 if (DiagOpts->ShowSourceRanges) {
1183 SourceLine = ' ' + SourceLine;
1184 CaretLine = ' ' + CaretLine;
1187 // Finally, remove any blank spaces from the end of CaretLine.
1188 while (CaretLine[CaretLine.size()-1] == ' ')
1189 CaretLine.erase(CaretLine.end()-1);
1191 // Emit what we have computed.
1192 emitSnippet(SourceLine);
1194 if (DiagOpts->ShowColors)
1195 OS.changeColor(caretColor, true);
1196 OS << CaretLine << '\n';
1197 if (DiagOpts->ShowColors)
1200 if (!FixItInsertionLine.empty()) {
1201 if (DiagOpts->ShowColors)
1202 // Print fixit line in color
1203 OS.changeColor(fixitColor, false);
1204 if (DiagOpts->ShowSourceRanges)
1206 OS << FixItInsertionLine << '\n';
1207 if (DiagOpts->ShowColors)
1211 // Print out any parseable fixit information requested by the options.
1212 emitParseableFixits(Hints, SM);
1215 void TextDiagnostic::emitSnippet(StringRef line) {
1221 std::string to_print;
1222 bool print_reversed = false;
1224 while (i<line.size()) {
1225 std::pair<SmallString<16>,bool> res
1226 = printableTextForNextCharacter(line, &i, DiagOpts->TabStop);
1227 bool was_printable = res.second;
1229 if (DiagOpts->ShowColors && was_printable == print_reversed) {
1234 if (DiagOpts->ShowColors)
1238 print_reversed = !was_printable;
1239 to_print += res.first.str();
1242 if (print_reversed && DiagOpts->ShowColors)
1245 if (print_reversed && DiagOpts->ShowColors)
1251 void TextDiagnostic::emitParseableFixits(ArrayRef<FixItHint> Hints,
1252 const SourceManager &SM) {
1253 if (!DiagOpts->ShowParseableFixits)
1256 // We follow FixItRewriter's example in not (yet) handling
1257 // fix-its in macros.
1258 for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1260 if (I->RemoveRange.isInvalid() ||
1261 I->RemoveRange.getBegin().isMacroID() ||
1262 I->RemoveRange.getEnd().isMacroID())
1266 for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1268 SourceLocation BLoc = I->RemoveRange.getBegin();
1269 SourceLocation ELoc = I->RemoveRange.getEnd();
1271 std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(BLoc);
1272 std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(ELoc);
1274 // Adjust for token ranges.
1275 if (I->RemoveRange.isTokenRange())
1276 EInfo.second += Lexer::MeasureTokenLength(ELoc, SM, LangOpts);
1278 // We specifically do not do word-wrapping or tab-expansion here,
1279 // because this is supposed to be easy to parse.
1280 PresumedLoc PLoc = SM.getPresumedLoc(BLoc);
1281 if (PLoc.isInvalid())
1285 OS.write_escaped(PLoc.getFilename());
1286 OS << "\":{" << SM.getLineNumber(BInfo.first, BInfo.second)
1287 << ':' << SM.getColumnNumber(BInfo.first, BInfo.second)
1288 << '-' << SM.getLineNumber(EInfo.first, EInfo.second)
1289 << ':' << SM.getColumnNumber(EInfo.first, EInfo.second)
1291 OS.write_escaped(I->CodeToInsert);