1 //===--- TextDiagnostic.cpp - Text Diagnostic Pretty-Printing -------------===//
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
7 //===----------------------------------------------------------------------===//
9 #include "clang/Frontend/TextDiagnostic.h"
10 #include "clang/Basic/CharInfo.h"
11 #include "clang/Basic/DiagnosticOptions.h"
12 #include "clang/Basic/FileManager.h"
13 #include "clang/Basic/SourceManager.h"
14 #include "clang/Lex/Lexer.h"
15 #include "llvm/ADT/SmallString.h"
16 #include "llvm/ADT/StringExtras.h"
17 #include "llvm/Support/ConvertUTF.h"
18 #include "llvm/Support/ErrorHandling.h"
19 #include "llvm/Support/Locale.h"
20 #include "llvm/Support/Path.h"
21 #include "llvm/Support/raw_ostream.h"
24 using namespace clang;
26 static const enum raw_ostream::Colors noteColor =
28 static const enum raw_ostream::Colors remarkColor =
30 static const enum raw_ostream::Colors fixitColor =
32 static const enum raw_ostream::Colors caretColor =
34 static const enum raw_ostream::Colors warningColor =
36 static const enum raw_ostream::Colors templateColor =
38 static const enum raw_ostream::Colors errorColor = raw_ostream::RED;
39 static const enum raw_ostream::Colors fatalColor = raw_ostream::RED;
40 // Used for changing only the bold attribute.
41 static const enum raw_ostream::Colors savedColor =
42 raw_ostream::SAVEDCOLOR;
44 /// Add highlights to differences in template strings.
45 static void applyTemplateHighlighting(raw_ostream &OS, StringRef Str,
46 bool &Normal, bool Bold) {
48 size_t Pos = Str.find(ToggleHighlight);
49 OS << Str.slice(0, Pos);
50 if (Pos == StringRef::npos)
53 Str = Str.substr(Pos + 1);
55 OS.changeColor(templateColor, true);
59 OS.changeColor(savedColor, true);
65 /// Number of spaces to indent when word-wrapping.
66 const unsigned WordWrapIndentation = 6;
68 static int bytesSincePreviousTabOrLineBegin(StringRef SourceLine, size_t i) {
71 if (SourceLine[--i]=='\t')
78 /// returns a printable representation of first item from input range
80 /// This function returns a printable representation of the next item in a line
81 /// of source. If the next byte begins a valid and printable character, that
82 /// character is returned along with 'true'.
84 /// Otherwise, if the next byte begins a valid, but unprintable character, a
85 /// printable, escaped representation of the character is returned, along with
86 /// 'false'. Otherwise a printable, escaped representation of the next byte
87 /// is returned along with 'false'.
89 /// \note The index is updated to be used with a subsequent call to
90 /// printableTextForNextCharacter.
92 /// \param SourceLine The line of source
93 /// \param i Pointer to byte index,
94 /// \param TabStop used to expand tabs
95 /// \return pair(printable text, 'true' iff original text was printable)
97 static std::pair<SmallString<16>, bool>
98 printableTextForNextCharacter(StringRef SourceLine, size_t *i,
100 assert(i && "i must not be null");
101 assert(*i<SourceLine.size() && "must point to a valid index");
103 if (SourceLine[*i]=='\t') {
104 assert(0 < TabStop && TabStop <= DiagnosticOptions::MaxTabStop &&
105 "Invalid -ftabstop value");
106 unsigned col = bytesSincePreviousTabOrLineBegin(SourceLine, *i);
107 unsigned NumSpaces = TabStop - col%TabStop;
108 assert(0 < NumSpaces && NumSpaces <= TabStop
109 && "Invalid computation of space amt");
112 SmallString<16> expandedTab;
113 expandedTab.assign(NumSpaces, ' ');
114 return std::make_pair(expandedTab, true);
117 unsigned char const *begin, *end;
118 begin = reinterpret_cast<unsigned char const *>(&*(SourceLine.begin() + *i));
119 end = begin + (SourceLine.size() - *i);
121 if (llvm::isLegalUTF8Sequence(begin, end)) {
123 llvm::UTF32 *cptr = &c;
124 unsigned char const *original_begin = begin;
125 unsigned char const *cp_end =
126 begin + llvm::getNumBytesForUTF8(SourceLine[*i]);
128 llvm::ConversionResult res = llvm::ConvertUTF8toUTF32(
129 &begin, cp_end, &cptr, cptr + 1, llvm::strictConversion);
131 assert(llvm::conversionOK == res);
132 assert(0 < begin-original_begin
133 && "we must be further along in the string now");
134 *i += begin-original_begin;
136 if (!llvm::sys::locale::isPrint(c)) {
137 // If next character is valid UTF-8, but not printable
138 SmallString<16> expandedCP("<U+>");
140 expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(c%16));
143 while (expandedCP.size() < 8)
144 expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(0));
145 return std::make_pair(expandedCP, false);
148 // If next character is valid UTF-8, and printable
149 return std::make_pair(SmallString<16>(original_begin, cp_end), true);
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);
159 return std::make_pair(expandedByte, false);
162 static void expandTabs(std::string &SourceLine, unsigned TabStop) {
163 size_t i = SourceLine.size();
166 if (SourceLine[i]!='\t')
169 std::pair<SmallString<16>,bool> res
170 = printableTextForNextCharacter(SourceLine, &tmp_i, TabStop);
171 SourceLine.replace(i, 1, res.first.c_str());
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).
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
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.
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
191 /// example: (given a tabstop of 8)
193 /// "a \t \u3042" -> {0,1,2,8,9,-1,-1,11}
195 /// (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
197 static void byteToColumn(StringRef SourceLine, unsigned TabStop,
198 SmallVectorImpl<int> &out) {
201 if (SourceLine.empty()) {
206 out.resize(SourceLine.size()+1, -1);
210 while (i<SourceLine.size()) {
212 std::pair<SmallString<16>,bool> res
213 = printableTextForNextCharacter(SourceLine, &i, TabStop);
214 columns += llvm::sys::locale::columnWidth(res.first);
216 out.back() = columns;
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()
223 /// The last element in the array is the number of bytes in the source string
225 /// example: (given a tabstop of 8)
227 /// "a \t \u3042" -> {0,1,2,-1,-1,-1,-1,-1,3,4,-1,7}
229 /// (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
231 static void columnToByte(StringRef SourceLine, unsigned TabStop,
232 SmallVectorImpl<int> &out) {
235 if (SourceLine.empty()) {
242 while (i<SourceLine.size()) {
243 out.resize(columns+1, -1);
245 std::pair<SmallString<16>,bool> res
246 = printableTextForNextCharacter(SourceLine, &i, TabStop);
247 columns += llvm::sys::locale::columnWidth(res.first);
249 out.resize(columns+1, -1);
254 struct SourceColumnMap {
255 SourceColumnMap(StringRef SourceLine, unsigned TabStop)
256 : m_SourceLine(SourceLine) {
258 ::byteToColumn(SourceLine, TabStop, m_byteToColumn);
259 ::columnToByte(SourceLine, TabStop, m_columnToByte);
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());
268 int columns() const { return m_byteToColumn.back(); }
269 int bytes() const { return m_columnToByte.back(); }
271 /// Map a byte to the column which it is at the start of, or return -1
272 /// if it is not at the start of a column (for a UTF-8 trailing byte).
273 int byteToColumn(int n) const {
274 assert(0<=n && n<static_cast<int>(m_byteToColumn.size()));
275 return m_byteToColumn[n];
278 /// Map a byte to the first column which contains it.
279 int byteToContainingColumn(int N) const {
280 assert(0 <= N && N < static_cast<int>(m_byteToColumn.size()));
281 while (m_byteToColumn[N] == -1)
283 return m_byteToColumn[N];
286 /// Map a column to the byte which starts the column, or return -1 if
287 /// the column the second or subsequent column of an expanded tab or similar
288 /// multi-column entity.
289 int columnToByte(int n) const {
290 assert(0<=n && n<static_cast<int>(m_columnToByte.size()));
291 return m_columnToByte[n];
294 /// Map from a byte index to the next byte which starts a column.
295 int startOfNextColumn(int N) const {
296 assert(0 <= N && N < static_cast<int>(m_byteToColumn.size() - 1));
297 while (byteToColumn(++N) == -1) {}
301 /// Map from a byte index to the previous byte which starts a column.
302 int startOfPreviousColumn(int N) const {
303 assert(0 < N && N < static_cast<int>(m_byteToColumn.size()));
304 while (byteToColumn(--N) == -1) {}
308 StringRef getSourceLine() const {
313 const std::string m_SourceLine;
314 SmallVector<int,200> m_byteToColumn;
315 SmallVector<int,200> m_columnToByte;
317 } // end anonymous namespace
319 /// When the source code line we want to print is too long for
320 /// the terminal, select the "interesting" region.
321 static void selectInterestingSourceRegion(std::string &SourceLine,
322 std::string &CaretLine,
323 std::string &FixItInsertionLine,
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)
334 // No special characters are allowed in CaretLine.
335 assert(CaretLine.end() ==
336 llvm::find_if(CaretLine, [](char c) { return c < ' ' || '~' < c; }));
338 // Find the slice that we need to display the full caret line
340 unsigned CaretStart = 0, CaretEnd = CaretLine.size();
341 for (; CaretStart != CaretEnd; ++CaretStart)
342 if (!isWhitespace(CaretLine[CaretStart]))
345 for (; CaretEnd != CaretStart; --CaretEnd)
346 if (!isWhitespace(CaretLine[CaretEnd - 1]))
349 // caret has already been inserted into CaretLine so the above whitespace
350 // check is guaranteed to include the caret
352 // If we have a fix-it line, make sure the slice includes all of the
353 // fix-it information.
354 if (!FixItInsertionLine.empty()) {
355 unsigned FixItStart = 0, FixItEnd = FixItInsertionLine.size();
356 for (; FixItStart != FixItEnd; ++FixItStart)
357 if (!isWhitespace(FixItInsertionLine[FixItStart]))
360 for (; FixItEnd != FixItStart; --FixItEnd)
361 if (!isWhitespace(FixItInsertionLine[FixItEnd - 1]))
364 // We can safely use the byte offset FixItStart as the column offset
365 // because the characters up until FixItStart are all ASCII whitespace
367 unsigned FixItStartCol = FixItStart;
369 = llvm::sys::locale::columnWidth(FixItInsertionLine.substr(0, FixItEnd));
371 CaretStart = std::min(FixItStartCol, CaretStart);
372 CaretEnd = std::max(FixItEndCol, CaretEnd);
375 // CaretEnd may have been set at the middle of a character
376 // If it's not at a character's first column then advance it past the current
378 while (static_cast<int>(CaretEnd) < map.columns() &&
379 -1 == map.columnToByte(CaretEnd))
382 assert((static_cast<int>(CaretStart) > map.columns() ||
383 -1!=map.columnToByte(CaretStart)) &&
384 "CaretStart must not point to a column in the middle of a source"
386 assert((static_cast<int>(CaretEnd) > map.columns() ||
387 -1!=map.columnToByte(CaretEnd)) &&
388 "CaretEnd must not point to a column in the middle of a source line"
391 // CaretLine[CaretStart, CaretEnd) contains all of the interesting
392 // parts of the caret line. While this slice is smaller than the
393 // number of columns we have, try to grow the slice to encompass
396 unsigned SourceStart = map.columnToByte(std::min<unsigned>(CaretStart,
398 unsigned SourceEnd = map.columnToByte(std::min<unsigned>(CaretEnd,
401 unsigned CaretColumnsOutsideSource = CaretEnd-CaretStart
402 - (map.byteToColumn(SourceEnd)-map.byteToColumn(SourceStart));
404 char const *front_ellipse = " ...";
405 char const *front_space = " ";
406 char const *back_ellipse = "...";
407 unsigned ellipses_space = strlen(front_ellipse) + strlen(back_ellipse);
409 unsigned TargetColumns = Columns;
410 // Give us extra room for the ellipses
411 // and any of the caret line that extends past the source
412 if (TargetColumns > ellipses_space+CaretColumnsOutsideSource)
413 TargetColumns -= ellipses_space+CaretColumnsOutsideSource;
415 while (SourceStart>0 || SourceEnd<SourceLine.size()) {
416 bool ExpandedRegion = false;
419 unsigned NewStart = map.startOfPreviousColumn(SourceStart);
421 // Skip over any whitespace we see here; we're looking for
422 // another bit of interesting text.
423 // FIXME: Detect non-ASCII whitespace characters too.
424 while (NewStart && isWhitespace(SourceLine[NewStart]))
425 NewStart = map.startOfPreviousColumn(NewStart);
427 // Skip over this bit of "interesting" text.
429 unsigned Prev = map.startOfPreviousColumn(NewStart);
430 if (isWhitespace(SourceLine[Prev]))
435 assert(map.byteToColumn(NewStart) != -1);
436 unsigned NewColumns = map.byteToColumn(SourceEnd) -
437 map.byteToColumn(NewStart);
438 if (NewColumns <= TargetColumns) {
439 SourceStart = NewStart;
440 ExpandedRegion = true;
444 if (SourceEnd<SourceLine.size()) {
445 unsigned NewEnd = map.startOfNextColumn(SourceEnd);
447 // Skip over any whitespace we see here; we're looking for
448 // another bit of interesting text.
449 // FIXME: Detect non-ASCII whitespace characters too.
450 while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
451 NewEnd = map.startOfNextColumn(NewEnd);
453 // Skip over this bit of "interesting" text.
454 while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
455 NewEnd = map.startOfNextColumn(NewEnd);
457 assert(map.byteToColumn(NewEnd) != -1);
458 unsigned NewColumns = map.byteToColumn(NewEnd) -
459 map.byteToColumn(SourceStart);
460 if (NewColumns <= TargetColumns) {
462 ExpandedRegion = true;
470 CaretStart = map.byteToColumn(SourceStart);
471 CaretEnd = map.byteToColumn(SourceEnd) + CaretColumnsOutsideSource;
473 // [CaretStart, CaretEnd) is the slice we want. Update the various
474 // output lines to show only this slice, with two-space padding
475 // before the lines so that it looks nicer.
477 assert(CaretStart!=(unsigned)-1 && CaretEnd!=(unsigned)-1 &&
478 SourceStart!=(unsigned)-1 && SourceEnd!=(unsigned)-1);
479 assert(SourceStart <= SourceEnd);
480 assert(CaretStart <= CaretEnd);
482 unsigned BackColumnsRemoved
483 = map.byteToColumn(SourceLine.size())-map.byteToColumn(SourceEnd);
484 unsigned FrontColumnsRemoved = CaretStart;
485 unsigned ColumnsKept = CaretEnd-CaretStart;
487 // We checked up front that the line needed truncation
488 assert(FrontColumnsRemoved+ColumnsKept+BackColumnsRemoved > Columns);
490 // The line needs some truncation, and we'd prefer to keep the front
491 // if possible, so remove the back
492 if (BackColumnsRemoved > strlen(back_ellipse))
493 SourceLine.replace(SourceEnd, std::string::npos, back_ellipse);
495 // If that's enough then we're done
496 if (FrontColumnsRemoved+ColumnsKept <= Columns)
499 // Otherwise remove the front as well
500 if (FrontColumnsRemoved > strlen(front_ellipse)) {
501 SourceLine.replace(0, SourceStart, front_ellipse);
502 CaretLine.replace(0, CaretStart, front_space);
503 if (!FixItInsertionLine.empty())
504 FixItInsertionLine.replace(0, CaretStart, front_space);
508 /// Skip over whitespace in the string, starting at the given
511 /// \returns The index of the first non-whitespace character that is
512 /// greater than or equal to Idx or, if no such character exists,
513 /// returns the end of the string.
514 static unsigned skipWhitespace(unsigned Idx, StringRef Str, unsigned Length) {
515 while (Idx < Length && isWhitespace(Str[Idx]))
520 /// If the given character is the start of some kind of
521 /// balanced punctuation (e.g., quotes or parentheses), return the
522 /// character that will terminate the punctuation.
524 /// \returns The ending punctuation character, if any, or the NULL
525 /// character if the input character does not start any punctuation.
526 static inline char findMatchingPunctuation(char c) {
528 case '\'': return '\'';
529 case '`': return '\'';
530 case '"': return '"';
531 case '(': return ')';
532 case '[': return ']';
533 case '{': return '}';
540 /// Find the end of the word starting at the given offset
543 /// \returns the index pointing one character past the end of the
545 static unsigned findEndOfWord(unsigned Start, StringRef Str,
546 unsigned Length, unsigned Column,
548 assert(Start < Str.size() && "Invalid start position!");
549 unsigned End = Start + 1;
551 // If we are already at the end of the string, take that as the word.
552 if (End == Str.size())
555 // Determine if the start of the string is actually opening
556 // punctuation, e.g., a quote or parentheses.
557 char EndPunct = findMatchingPunctuation(Str[Start]);
559 // This is a normal word. Just find the first space character.
560 while (End < Length && !isWhitespace(Str[End]))
565 // We have the start of a balanced punctuation sequence (quotes,
566 // parentheses, etc.). Determine the full sequence is.
567 SmallString<16> PunctuationEndStack;
568 PunctuationEndStack.push_back(EndPunct);
569 while (End < Length && !PunctuationEndStack.empty()) {
570 if (Str[End] == PunctuationEndStack.back())
571 PunctuationEndStack.pop_back();
572 else if (char SubEndPunct = findMatchingPunctuation(Str[End]))
573 PunctuationEndStack.push_back(SubEndPunct);
578 // Find the first space character after the punctuation ended.
579 while (End < Length && !isWhitespace(Str[End]))
582 unsigned PunctWordLength = End - Start;
583 if (// If the word fits on this line
584 Column + PunctWordLength <= Columns ||
585 // ... or the word is "short enough" to take up the next line
586 // without too much ugly white space
587 PunctWordLength < Columns/3)
588 return End; // Take the whole thing as a single "word".
590 // The whole quoted/parenthesized string is too long to print as a
591 // single "word". Instead, find the "word" that starts just after
592 // the punctuation and use that end-point instead. This will recurse
593 // until it finds something small enough to consider a word.
594 return findEndOfWord(Start + 1, Str, Length, Column + 1, Columns);
597 /// Print the given string to a stream, word-wrapping it to
598 /// some number of columns in the process.
600 /// \param OS the stream to which the word-wrapping string will be
602 /// \param Str the string to word-wrap and output.
603 /// \param Columns the number of columns to word-wrap to.
604 /// \param Column the column number at which the first character of \p
605 /// Str will be printed. This will be non-zero when part of the first
606 /// line has already been printed.
607 /// \param Bold if the current text should be bold
608 /// \param Indentation the number of spaces to indent any lines beyond
610 /// \returns true if word-wrapping was required, or false if the
611 /// string fit on the first line.
612 static bool printWordWrapped(raw_ostream &OS, StringRef Str,
616 unsigned Indentation = WordWrapIndentation) {
617 const unsigned Length = std::min(Str.find('\n'), Str.size());
618 bool TextNormal = true;
620 // The string used to indent each line.
621 SmallString<16> IndentStr;
622 IndentStr.assign(Indentation, ' ');
623 bool Wrapped = false;
624 for (unsigned WordStart = 0, WordEnd; WordStart < Length;
625 WordStart = WordEnd) {
626 // Find the beginning of the next word.
627 WordStart = skipWhitespace(WordStart, Str, Length);
628 if (WordStart == Length)
631 // Find the end of this word.
632 WordEnd = findEndOfWord(WordStart, Str, Length, Column, Columns);
634 // Does this word fit on the current line?
635 unsigned WordLength = WordEnd - WordStart;
636 if (Column + WordLength < Columns) {
637 // This word fits on the current line; print it there.
642 applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
644 Column += WordLength;
648 // This word does not fit on the current line, so wrap to the next
651 OS.write(&IndentStr[0], Indentation);
652 applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
654 Column = Indentation + WordLength;
658 // Append any remaning text from the message with its existing formatting.
659 applyTemplateHighlighting(OS, Str.substr(Length), TextNormal, Bold);
661 assert(TextNormal && "Text highlighted at end of diagnostic message.");
666 TextDiagnostic::TextDiagnostic(raw_ostream &OS,
667 const LangOptions &LangOpts,
668 DiagnosticOptions *DiagOpts)
669 : DiagnosticRenderer(LangOpts, DiagOpts), OS(OS) {}
671 TextDiagnostic::~TextDiagnostic() {}
673 void TextDiagnostic::emitDiagnosticMessage(
674 FullSourceLoc Loc, PresumedLoc PLoc, DiagnosticsEngine::Level Level,
675 StringRef Message, ArrayRef<clang::CharSourceRange> Ranges,
676 DiagOrStoredDiag D) {
677 uint64_t StartOfLocationInfo = OS.tell();
679 // Emit the location of this particular diagnostic.
681 emitDiagnosticLoc(Loc, PLoc, Level, Ranges);
683 if (DiagOpts->ShowColors)
686 printDiagnosticLevel(OS, Level, DiagOpts->ShowColors,
687 DiagOpts->CLFallbackMode);
688 printDiagnosticMessage(OS,
689 /*IsSupplemental*/ Level == DiagnosticsEngine::Note,
690 Message, OS.tell() - StartOfLocationInfo,
691 DiagOpts->MessageLength, DiagOpts->ShowColors);
695 TextDiagnostic::printDiagnosticLevel(raw_ostream &OS,
696 DiagnosticsEngine::Level Level,
698 bool CLFallbackMode) {
700 // Print diagnostic category in bold and color
702 case DiagnosticsEngine::Ignored:
703 llvm_unreachable("Invalid diagnostic type");
704 case DiagnosticsEngine::Note: OS.changeColor(noteColor, true); break;
705 case DiagnosticsEngine::Remark: OS.changeColor(remarkColor, true); break;
706 case DiagnosticsEngine::Warning: OS.changeColor(warningColor, true); break;
707 case DiagnosticsEngine::Error: OS.changeColor(errorColor, true); break;
708 case DiagnosticsEngine::Fatal: OS.changeColor(fatalColor, true); break;
713 case DiagnosticsEngine::Ignored:
714 llvm_unreachable("Invalid diagnostic type");
715 case DiagnosticsEngine::Note: OS << "note"; break;
716 case DiagnosticsEngine::Remark: OS << "remark"; break;
717 case DiagnosticsEngine::Warning: OS << "warning"; break;
718 case DiagnosticsEngine::Error: OS << "error"; break;
719 case DiagnosticsEngine::Fatal: OS << "fatal error"; break;
722 // In clang-cl /fallback mode, print diagnostics as "error(clang):". This
723 // makes it more clear whether a message is coming from clang or cl.exe,
724 // and it prevents MSBuild from concluding that the build failed just because
725 // there is an "error:" in the output.
736 void TextDiagnostic::printDiagnosticMessage(raw_ostream &OS,
739 unsigned CurrentColumn,
740 unsigned Columns, bool ShowColors) {
742 if (ShowColors && !IsSupplemental) {
743 // Print primary diagnostic messages in bold and without color, to visually
744 // indicate the transition from continuation notes and other output.
745 OS.changeColor(savedColor, true);
750 printWordWrapped(OS, Message, Columns, CurrentColumn, Bold);
753 applyTemplateHighlighting(OS, Message, Normal, Bold);
754 assert(Normal && "Formatting should have returned to normal");
762 void TextDiagnostic::emitFilename(StringRef Filename, const SourceManager &SM) {
763 SmallVector<char, 128> AbsoluteFilename;
764 if (DiagOpts->AbsolutePath) {
765 const DirectoryEntry *Dir = SM.getFileManager().getDirectory(
766 llvm::sys::path::parent_path(Filename));
768 // We want to print a simplified absolute path, i. e. without "dots".
770 // The hardest part here are the paths like "<part1>/<link>/../<part2>".
771 // On Unix-like systems, we cannot just collapse "<link>/..", because
772 // paths are resolved sequentially, and, thereby, the path
773 // "<part1>/<part2>" may point to a different location. That is why
774 // we use FileManager::getCanonicalName(), which expands all indirections
775 // with llvm::sys::fs::real_path() and caches the result.
777 // On the other hand, it would be better to preserve as much of the
778 // original path as possible, because that helps a user to recognize it.
779 // real_path() expands all links, which sometimes too much. Luckily,
780 // on Windows we can just use llvm::sys::path::remove_dots(), because,
781 // on that system, both aforementioned paths point to the same place.
783 SmallString<4096> DirName = Dir->getName();
784 llvm::sys::fs::make_absolute(DirName);
785 llvm::sys::path::native(DirName);
786 llvm::sys::path::remove_dots(DirName, /* remove_dot_dot */ true);
788 StringRef DirName = SM.getFileManager().getCanonicalName(Dir);
790 llvm::sys::path::append(AbsoluteFilename, DirName,
791 llvm::sys::path::filename(Filename));
792 Filename = StringRef(AbsoluteFilename.data(), AbsoluteFilename.size());
799 /// Print out the file/line/column information and include trace.
801 /// This method handlen the emission of the diagnostic location information.
802 /// This includes extracting as much location information as is present for
803 /// the diagnostic and printing it, as well as any include stack or source
804 /// ranges necessary.
805 void TextDiagnostic::emitDiagnosticLoc(FullSourceLoc Loc, PresumedLoc PLoc,
806 DiagnosticsEngine::Level Level,
807 ArrayRef<CharSourceRange> Ranges) {
808 if (PLoc.isInvalid()) {
809 // At least print the file name if available:
810 FileID FID = Loc.getFileID();
812 const FileEntry *FE = Loc.getFileEntry();
813 if (FE && FE->isValid()) {
814 emitFilename(FE->getName(), Loc.getManager());
820 unsigned LineNo = PLoc.getLine();
822 if (!DiagOpts->ShowLocation)
825 if (DiagOpts->ShowColors)
826 OS.changeColor(savedColor, true);
828 emitFilename(PLoc.getFilename(), Loc.getManager());
829 switch (DiagOpts->getFormat()) {
830 case DiagnosticOptions::Clang: OS << ':' << LineNo; break;
831 case DiagnosticOptions::MSVC: OS << '(' << LineNo; break;
832 case DiagnosticOptions::Vi: OS << " +" << LineNo; break;
835 if (DiagOpts->ShowColumn)
836 // Compute the column number.
837 if (unsigned ColNo = PLoc.getColumn()) {
838 if (DiagOpts->getFormat() == DiagnosticOptions::MSVC) {
840 // Visual Studio 2010 or earlier expects column number to be off by one
841 if (LangOpts.MSCompatibilityVersion &&
842 !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2012))
848 switch (DiagOpts->getFormat()) {
849 case DiagnosticOptions::Clang:
850 case DiagnosticOptions::Vi: OS << ':'; break;
851 case DiagnosticOptions::MSVC:
852 // MSVC2013 and before print 'file(4) : error'. MSVC2015 gets rid of the
853 // space and prints 'file(4): error'.
855 if (LangOpts.MSCompatibilityVersion &&
856 !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2015))
862 if (DiagOpts->ShowSourceRanges && !Ranges.empty()) {
863 FileID CaretFileID = Loc.getExpansionLoc().getFileID();
864 bool PrintedRange = false;
866 for (ArrayRef<CharSourceRange>::const_iterator RI = Ranges.begin(),
869 // Ignore invalid ranges.
870 if (!RI->isValid()) continue;
872 auto &SM = Loc.getManager();
873 SourceLocation B = SM.getExpansionLoc(RI->getBegin());
874 CharSourceRange ERange = SM.getExpansionRange(RI->getEnd());
875 SourceLocation E = ERange.getEnd();
876 bool IsTokenRange = ERange.isTokenRange();
878 std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(B);
879 std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(E);
881 // If the start or end of the range is in another file, just discard
883 if (BInfo.first != CaretFileID || EInfo.first != CaretFileID)
886 // Add in the length of the token, so that we cover multi-char
888 unsigned TokSize = 0;
890 TokSize = Lexer::MeasureTokenLength(E, SM, LangOpts);
892 FullSourceLoc BF(B, SM), EF(E, SM);
894 << BF.getLineNumber() << ':' << BF.getColumnNumber() << '-'
895 << EF.getLineNumber() << ':' << (EF.getColumnNumber() + TokSize)
906 void TextDiagnostic::emitIncludeLocation(FullSourceLoc Loc, PresumedLoc PLoc) {
907 if (DiagOpts->ShowLocation && PLoc.isValid())
908 OS << "In file included from " << PLoc.getFilename() << ':'
909 << PLoc.getLine() << ":\n";
911 OS << "In included file:\n";
914 void TextDiagnostic::emitImportLocation(FullSourceLoc Loc, PresumedLoc PLoc,
915 StringRef ModuleName) {
916 if (DiagOpts->ShowLocation && PLoc.isValid())
917 OS << "In module '" << ModuleName << "' imported from "
918 << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
920 OS << "In module '" << ModuleName << "':\n";
923 void TextDiagnostic::emitBuildingModuleLocation(FullSourceLoc Loc,
925 StringRef ModuleName) {
926 if (DiagOpts->ShowLocation && PLoc.isValid())
927 OS << "While building module '" << ModuleName << "' imported from "
928 << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
930 OS << "While building module '" << ModuleName << "':\n";
933 /// Find the suitable set of lines to show to include a set of ranges.
934 static llvm::Optional<std::pair<unsigned, unsigned>>
935 findLinesForRange(const CharSourceRange &R, FileID FID,
936 const SourceManager &SM) {
937 if (!R.isValid()) return None;
939 SourceLocation Begin = R.getBegin();
940 SourceLocation End = R.getEnd();
941 if (SM.getFileID(Begin) != FID || SM.getFileID(End) != FID)
944 return std::make_pair(SM.getExpansionLineNumber(Begin),
945 SM.getExpansionLineNumber(End));
948 /// Add as much of range B into range A as possible without exceeding a maximum
949 /// size of MaxRange. Ranges are inclusive.
950 static std::pair<unsigned, unsigned>
951 maybeAddRange(std::pair<unsigned, unsigned> A, std::pair<unsigned, unsigned> B,
953 // If A is already the maximum size, we're done.
954 unsigned Slack = MaxRange - (A.second - A.first + 1);
958 // Easy case: merge succeeds within MaxRange.
959 unsigned Min = std::min(A.first, B.first);
960 unsigned Max = std::max(A.second, B.second);
961 if (Max - Min + 1 <= MaxRange)
964 // If we can't reach B from A within MaxRange, there's nothing to do.
965 // Don't add lines to the range that contain nothing interesting.
966 if ((B.first > A.first && B.first - A.first + 1 > MaxRange) ||
967 (B.second < A.second && A.second - B.second + 1 > MaxRange))
970 // Otherwise, expand A towards B to produce a range of size MaxRange. We
971 // attempt to expand by the same amount in both directions if B strictly
974 // Expand downwards by up to half the available amount, then upwards as
975 // much as possible, then downwards as much as possible.
976 A.second = std::min(A.second + (Slack + 1) / 2, Max);
977 Slack = MaxRange - (A.second - A.first + 1);
978 A.first = std::max(Min + Slack, A.first) - Slack;
979 A.second = std::min(A.first + MaxRange - 1, Max);
983 /// Highlight a SourceRange (with ~'s) for any characters on LineNo.
984 static void highlightRange(const CharSourceRange &R,
985 unsigned LineNo, FileID FID,
986 const SourceColumnMap &map,
987 std::string &CaretLine,
988 const SourceManager &SM,
989 const LangOptions &LangOpts) {
990 if (!R.isValid()) return;
992 SourceLocation Begin = R.getBegin();
993 SourceLocation End = R.getEnd();
995 unsigned StartLineNo = SM.getExpansionLineNumber(Begin);
996 if (StartLineNo > LineNo || SM.getFileID(Begin) != FID)
997 return; // No intersection.
999 unsigned EndLineNo = SM.getExpansionLineNumber(End);
1000 if (EndLineNo < LineNo || SM.getFileID(End) != FID)
1001 return; // No intersection.
1003 // Compute the column number of the start.
1004 unsigned StartColNo = 0;
1005 if (StartLineNo == LineNo) {
1006 StartColNo = SM.getExpansionColumnNumber(Begin);
1007 if (StartColNo) --StartColNo; // Zero base the col #.
1010 // Compute the column number of the end.
1011 unsigned EndColNo = map.getSourceLine().size();
1012 if (EndLineNo == LineNo) {
1013 EndColNo = SM.getExpansionColumnNumber(End);
1015 --EndColNo; // Zero base the col #.
1017 // Add in the length of the token, so that we cover multi-char tokens if
1018 // this is a token range.
1019 if (R.isTokenRange())
1020 EndColNo += Lexer::MeasureTokenLength(End, SM, LangOpts);
1022 EndColNo = CaretLine.size();
1026 assert(StartColNo <= EndColNo && "Invalid range!");
1028 // Check that a token range does not highlight only whitespace.
1029 if (R.isTokenRange()) {
1030 // Pick the first non-whitespace column.
1031 while (StartColNo < map.getSourceLine().size() &&
1032 (map.getSourceLine()[StartColNo] == ' ' ||
1033 map.getSourceLine()[StartColNo] == '\t'))
1034 StartColNo = map.startOfNextColumn(StartColNo);
1036 // Pick the last non-whitespace column.
1037 if (EndColNo > map.getSourceLine().size())
1038 EndColNo = map.getSourceLine().size();
1040 (map.getSourceLine()[EndColNo-1] == ' ' ||
1041 map.getSourceLine()[EndColNo-1] == '\t'))
1042 EndColNo = map.startOfPreviousColumn(EndColNo);
1044 // If the start/end passed each other, then we are trying to highlight a
1045 // range that just exists in whitespace. That most likely means we have
1046 // a multi-line highlighting range that covers a blank line.
1047 if (StartColNo > EndColNo) {
1048 assert(StartLineNo != EndLineNo && "trying to highlight whitespace");
1049 StartColNo = EndColNo;
1053 assert(StartColNo <= map.getSourceLine().size() && "Invalid range!");
1054 assert(EndColNo <= map.getSourceLine().size() && "Invalid range!");
1056 // Fill the range with ~'s.
1057 StartColNo = map.byteToContainingColumn(StartColNo);
1058 EndColNo = map.byteToContainingColumn(EndColNo);
1060 assert(StartColNo <= EndColNo && "Invalid range!");
1061 if (CaretLine.size() < EndColNo)
1062 CaretLine.resize(EndColNo,' ');
1063 std::fill(CaretLine.begin()+StartColNo,CaretLine.begin()+EndColNo,'~');
1066 static std::string buildFixItInsertionLine(FileID FID,
1068 const SourceColumnMap &map,
1069 ArrayRef<FixItHint> Hints,
1070 const SourceManager &SM,
1071 const DiagnosticOptions *DiagOpts) {
1072 std::string FixItInsertionLine;
1073 if (Hints.empty() || !DiagOpts->ShowFixits)
1074 return FixItInsertionLine;
1075 unsigned PrevHintEndCol = 0;
1077 for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1079 if (!I->CodeToInsert.empty()) {
1080 // We have an insertion hint. Determine whether the inserted
1081 // code contains no newlines and is on the same line as the caret.
1082 std::pair<FileID, unsigned> HintLocInfo
1083 = SM.getDecomposedExpansionLoc(I->RemoveRange.getBegin());
1084 if (FID == HintLocInfo.first &&
1085 LineNo == SM.getLineNumber(HintLocInfo.first, HintLocInfo.second) &&
1086 StringRef(I->CodeToInsert).find_first_of("\n\r") == StringRef::npos) {
1087 // Insert the new code into the line just below the code
1088 // that the user wrote.
1089 // Note: When modifying this function, be very careful about what is a
1090 // "column" (printed width, platform-dependent) and what is a
1091 // "byte offset" (SourceManager "column").
1092 unsigned HintByteOffset
1093 = SM.getColumnNumber(HintLocInfo.first, HintLocInfo.second) - 1;
1095 // The hint must start inside the source or right at the end
1096 assert(HintByteOffset < static_cast<unsigned>(map.bytes())+1);
1097 unsigned HintCol = map.byteToContainingColumn(HintByteOffset);
1099 // If we inserted a long previous hint, push this one forwards, and add
1100 // an extra space to show that this is not part of the previous
1101 // completion. This is sort of the best we can do when two hints appear
1104 // Note that if this hint is located immediately after the previous
1105 // hint, no space will be added, since the location is more important.
1106 if (HintCol < PrevHintEndCol)
1107 HintCol = PrevHintEndCol + 1;
1109 // This should NOT use HintByteOffset, because the source might have
1110 // Unicode characters in earlier columns.
1111 unsigned NewFixItLineSize = FixItInsertionLine.size() +
1112 (HintCol - PrevHintEndCol) + I->CodeToInsert.size();
1113 if (NewFixItLineSize > FixItInsertionLine.size())
1114 FixItInsertionLine.resize(NewFixItLineSize, ' ');
1116 std::copy(I->CodeToInsert.begin(), I->CodeToInsert.end(),
1117 FixItInsertionLine.end() - I->CodeToInsert.size());
1120 HintCol + llvm::sys::locale::columnWidth(I->CodeToInsert);
1125 expandTabs(FixItInsertionLine, DiagOpts->TabStop);
1127 return FixItInsertionLine;
1130 /// Emit a code snippet and caret line.
1132 /// This routine emits a single line's code snippet and caret line..
1134 /// \param Loc The location for the caret.
1135 /// \param Ranges The underlined ranges for this code snippet.
1136 /// \param Hints The FixIt hints active for this diagnostic.
1137 void TextDiagnostic::emitSnippetAndCaret(
1138 FullSourceLoc Loc, DiagnosticsEngine::Level Level,
1139 SmallVectorImpl<CharSourceRange> &Ranges, ArrayRef<FixItHint> Hints) {
1140 assert(Loc.isValid() && "must have a valid source location here");
1141 assert(Loc.isFileID() && "must have a file location here");
1143 // If caret diagnostics are enabled and we have location, we want to
1144 // emit the caret. However, we only do this if the location moved
1145 // from the last diagnostic, if the last diagnostic was a note that
1146 // was part of a different warning or error diagnostic, or if the
1147 // diagnostic has ranges. We don't want to emit the same caret
1148 // multiple times if one loc has multiple diagnostics.
1149 if (!DiagOpts->ShowCarets)
1151 if (Loc == LastLoc && Ranges.empty() && Hints.empty() &&
1152 (LastLevel != DiagnosticsEngine::Note || Level == LastLevel))
1155 // Decompose the location into a FID/Offset pair.
1156 std::pair<FileID, unsigned> LocInfo = Loc.getDecomposedLoc();
1157 FileID FID = LocInfo.first;
1158 const SourceManager &SM = Loc.getManager();
1160 // Get information about the buffer it points into.
1161 bool Invalid = false;
1162 StringRef BufData = Loc.getBufferData(&Invalid);
1166 unsigned CaretLineNo = Loc.getLineNumber();
1167 unsigned CaretColNo = Loc.getColumnNumber();
1169 // Arbitrarily stop showing snippets when the line is too long.
1170 static const size_t MaxLineLengthToPrint = 4096;
1171 if (CaretColNo > MaxLineLengthToPrint)
1174 // Find the set of lines to include.
1175 const unsigned MaxLines = DiagOpts->SnippetLineLimit;
1176 std::pair<unsigned, unsigned> Lines = {CaretLineNo, CaretLineNo};
1177 for (SmallVectorImpl<CharSourceRange>::iterator I = Ranges.begin(),
1180 if (auto OptionalRange = findLinesForRange(*I, FID, SM))
1181 Lines = maybeAddRange(Lines, *OptionalRange, MaxLines);
1183 for (unsigned LineNo = Lines.first; LineNo != Lines.second + 1; ++LineNo) {
1184 const char *BufStart = BufData.data();
1185 const char *BufEnd = BufStart + BufData.size();
1187 // Rewind from the current position to the start of the line.
1188 const char *LineStart =
1190 SM.getDecomposedLoc(SM.translateLineCol(FID, LineNo, 1)).second;
1191 if (LineStart == BufEnd)
1194 // Compute the line end.
1195 const char *LineEnd = LineStart;
1196 while (*LineEnd != '\n' && *LineEnd != '\r' && LineEnd != BufEnd)
1199 // Arbitrarily stop showing snippets when the line is too long.
1200 // FIXME: Don't print any lines in this case.
1201 if (size_t(LineEnd - LineStart) > MaxLineLengthToPrint)
1204 // Trim trailing null-bytes.
1205 StringRef Line(LineStart, LineEnd - LineStart);
1206 while (!Line.empty() && Line.back() == '\0' &&
1207 (LineNo != CaretLineNo || Line.size() > CaretColNo))
1208 Line = Line.drop_back();
1210 // Copy the line of code into an std::string for ease of manipulation.
1211 std::string SourceLine(Line.begin(), Line.end());
1213 // Build the byte to column map.
1214 const SourceColumnMap sourceColMap(SourceLine, DiagOpts->TabStop);
1216 // Create a line for the caret that is filled with spaces that is the same
1217 // number of columns as the line of source code.
1218 std::string CaretLine(sourceColMap.columns(), ' ');
1220 // Highlight all of the characters covered by Ranges with ~ characters.
1221 for (SmallVectorImpl<CharSourceRange>::iterator I = Ranges.begin(),
1224 highlightRange(*I, LineNo, FID, sourceColMap, CaretLine, SM, LangOpts);
1226 // Next, insert the caret itself.
1227 if (CaretLineNo == LineNo) {
1228 CaretColNo = sourceColMap.byteToContainingColumn(CaretColNo - 1);
1229 if (CaretLine.size() < CaretColNo + 1)
1230 CaretLine.resize(CaretColNo + 1, ' ');
1231 CaretLine[CaretColNo] = '^';
1234 std::string FixItInsertionLine = buildFixItInsertionLine(
1235 FID, LineNo, sourceColMap, Hints, SM, DiagOpts.get());
1237 // If the source line is too long for our terminal, select only the
1238 // "interesting" source region within that line.
1239 unsigned Columns = DiagOpts->MessageLength;
1241 selectInterestingSourceRegion(SourceLine, CaretLine, FixItInsertionLine,
1242 Columns, sourceColMap);
1244 // If we are in -fdiagnostics-print-source-range-info mode, we are trying
1245 // to produce easily machine parsable output. Add a space before the
1246 // source line and the caret to make it trivial to tell the main diagnostic
1247 // line from what the user is intended to see.
1248 if (DiagOpts->ShowSourceRanges) {
1249 SourceLine = ' ' + SourceLine;
1250 CaretLine = ' ' + CaretLine;
1253 // Finally, remove any blank spaces from the end of CaretLine.
1254 while (!CaretLine.empty() && CaretLine[CaretLine.size() - 1] == ' ')
1255 CaretLine.erase(CaretLine.end() - 1);
1257 // Emit what we have computed.
1258 emitSnippet(SourceLine);
1260 if (!CaretLine.empty()) {
1261 if (DiagOpts->ShowColors)
1262 OS.changeColor(caretColor, true);
1263 OS << CaretLine << '\n';
1264 if (DiagOpts->ShowColors)
1268 if (!FixItInsertionLine.empty()) {
1269 if (DiagOpts->ShowColors)
1270 // Print fixit line in color
1271 OS.changeColor(fixitColor, false);
1272 if (DiagOpts->ShowSourceRanges)
1274 OS << FixItInsertionLine << '\n';
1275 if (DiagOpts->ShowColors)
1280 // Print out any parseable fixit information requested by the options.
1281 emitParseableFixits(Hints, SM);
1284 void TextDiagnostic::emitSnippet(StringRef line) {
1290 std::string to_print;
1291 bool print_reversed = false;
1293 while (i<line.size()) {
1294 std::pair<SmallString<16>,bool> res
1295 = printableTextForNextCharacter(line, &i, DiagOpts->TabStop);
1296 bool was_printable = res.second;
1298 if (DiagOpts->ShowColors && was_printable == print_reversed) {
1303 if (DiagOpts->ShowColors)
1307 print_reversed = !was_printable;
1308 to_print += res.first.str();
1311 if (print_reversed && DiagOpts->ShowColors)
1314 if (print_reversed && DiagOpts->ShowColors)
1320 void TextDiagnostic::emitParseableFixits(ArrayRef<FixItHint> Hints,
1321 const SourceManager &SM) {
1322 if (!DiagOpts->ShowParseableFixits)
1325 // We follow FixItRewriter's example in not (yet) handling
1326 // fix-its in macros.
1327 for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1329 if (I->RemoveRange.isInvalid() ||
1330 I->RemoveRange.getBegin().isMacroID() ||
1331 I->RemoveRange.getEnd().isMacroID())
1335 for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
1337 SourceLocation BLoc = I->RemoveRange.getBegin();
1338 SourceLocation ELoc = I->RemoveRange.getEnd();
1340 std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(BLoc);
1341 std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(ELoc);
1343 // Adjust for token ranges.
1344 if (I->RemoveRange.isTokenRange())
1345 EInfo.second += Lexer::MeasureTokenLength(ELoc, SM, LangOpts);
1347 // We specifically do not do word-wrapping or tab-expansion here,
1348 // because this is supposed to be easy to parse.
1349 PresumedLoc PLoc = SM.getPresumedLoc(BLoc);
1350 if (PLoc.isInvalid())
1354 OS.write_escaped(PLoc.getFilename());
1355 OS << "\":{" << SM.getLineNumber(BInfo.first, BInfo.second)
1356 << ':' << SM.getColumnNumber(BInfo.first, BInfo.second)
1357 << '-' << SM.getLineNumber(EInfo.first, EInfo.second)
1358 << ':' << SM.getColumnNumber(EInfo.first, EInfo.second)
1360 OS.write_escaped(I->CodeToInsert);