Home | History | Annotate | Line # | Download | only in Frontend
TextDiagnostic.cpp revision 1.1.1.2
      1 //===--- TextDiagnostic.cpp - Text Diagnostic Pretty-Printing -------------===//
      2 //
      3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
      4 // See https://llvm.org/LICENSE.txt for license information.
      5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
      6 //
      7 //===----------------------------------------------------------------------===//
      8 
      9 #include "clang/Frontend/TextDiagnostic.h"
     10 #include "clang/Basic/CharInfo.h"
     11 #include "clang/Basic/DiagnosticOptions.h"
     12 #include "clang/Basic/FileManager.h"
     13 #include "clang/Basic/SourceManager.h"
     14 #include "clang/Lex/Lexer.h"
     15 #include "llvm/ADT/SmallString.h"
     16 #include "llvm/ADT/StringExtras.h"
     17 #include "llvm/Support/ConvertUTF.h"
     18 #include "llvm/Support/ErrorHandling.h"
     19 #include "llvm/Support/Locale.h"
     20 #include "llvm/Support/Path.h"
     21 #include "llvm/Support/raw_ostream.h"
     22 #include <algorithm>
     23 
     24 using namespace clang;
     25 
     26 static const enum raw_ostream::Colors noteColor =
     27   raw_ostream::BLACK;
     28 static const enum raw_ostream::Colors remarkColor =
     29   raw_ostream::BLUE;
     30 static const enum raw_ostream::Colors fixitColor =
     31   raw_ostream::GREEN;
     32 static const enum raw_ostream::Colors caretColor =
     33   raw_ostream::GREEN;
     34 static const enum raw_ostream::Colors warningColor =
     35   raw_ostream::MAGENTA;
     36 static const enum raw_ostream::Colors templateColor =
     37   raw_ostream::CYAN;
     38 static const enum raw_ostream::Colors errorColor = raw_ostream::RED;
     39 static const enum raw_ostream::Colors fatalColor = raw_ostream::RED;
     40 // Used for changing only the bold attribute.
     41 static const enum raw_ostream::Colors savedColor =
     42   raw_ostream::SAVEDCOLOR;
     43 
     44 /// Add highlights to differences in template strings.
     45 static void applyTemplateHighlighting(raw_ostream &OS, StringRef Str,
     46                                       bool &Normal, bool Bold) {
     47   while (1) {
     48     size_t Pos = Str.find(ToggleHighlight);
     49     OS << Str.slice(0, Pos);
     50     if (Pos == StringRef::npos)
     51       break;
     52 
     53     Str = Str.substr(Pos + 1);
     54     if (Normal)
     55       OS.changeColor(templateColor, true);
     56     else {
     57       OS.resetColor();
     58       if (Bold)
     59         OS.changeColor(savedColor, true);
     60     }
     61     Normal = !Normal;
     62   }
     63 }
     64 
     65 /// Number of spaces to indent when word-wrapping.
     66 const unsigned WordWrapIndentation = 6;
     67 
     68 static int bytesSincePreviousTabOrLineBegin(StringRef SourceLine, size_t i) {
     69   int bytes = 0;
     70   while (0<i) {
     71     if (SourceLine[--i]=='\t')
     72       break;
     73     ++bytes;
     74   }
     75   return bytes;
     76 }
     77 
     78 /// returns a printable representation of first item from input range
     79 ///
     80 /// This function returns a printable representation of the next item in a line
     81 ///  of source. If the next byte begins a valid and printable character, that
     82 ///  character is returned along with 'true'.
     83 ///
     84 /// Otherwise, if the next byte begins a valid, but unprintable character, a
     85 ///  printable, escaped representation of the character is returned, along with
     86 ///  'false'. Otherwise a printable, escaped representation of the next byte
     87 ///  is returned along with 'false'.
     88 ///
     89 /// \note The index is updated to be used with a subsequent call to
     90 ///        printableTextForNextCharacter.
     91 ///
     92 /// \param SourceLine The line of source
     93 /// \param i Pointer to byte index,
     94 /// \param TabStop used to expand tabs
     95 /// \return pair(printable text, 'true' iff original text was printable)
     96 ///
     97 static std::pair<SmallString<16>, bool>
     98 printableTextForNextCharacter(StringRef SourceLine, size_t *i,
     99                               unsigned TabStop) {
    100   assert(i && "i must not be null");
    101   assert(*i<SourceLine.size() && "must point to a valid index");
    102 
    103   if (SourceLine[*i]=='\t') {
    104     assert(0 < TabStop && TabStop <= DiagnosticOptions::MaxTabStop &&
    105            "Invalid -ftabstop value");
    106     unsigned col = bytesSincePreviousTabOrLineBegin(SourceLine, *i);
    107     unsigned NumSpaces = TabStop - col%TabStop;
    108     assert(0 < NumSpaces && NumSpaces <= TabStop
    109            && "Invalid computation of space amt");
    110     ++(*i);
    111 
    112     SmallString<16> expandedTab;
    113     expandedTab.assign(NumSpaces, ' ');
    114     return std::make_pair(expandedTab, true);
    115   }
    116 
    117   unsigned char const *begin, *end;
    118   begin = reinterpret_cast<unsigned char const *>(&*(SourceLine.begin() + *i));
    119   end = begin + (SourceLine.size() - *i);
    120 
    121   if (llvm::isLegalUTF8Sequence(begin, end)) {
    122     llvm::UTF32 c;
    123     llvm::UTF32 *cptr = &c;
    124     unsigned char const *original_begin = begin;
    125     unsigned char const *cp_end =
    126         begin + llvm::getNumBytesForUTF8(SourceLine[*i]);
    127 
    128     llvm::ConversionResult res = llvm::ConvertUTF8toUTF32(
    129         &begin, cp_end, &cptr, cptr + 1, llvm::strictConversion);
    130     (void)res;
    131     assert(llvm::conversionOK == res);
    132     assert(0 < begin-original_begin
    133            && "we must be further along in the string now");
    134     *i += begin-original_begin;
    135 
    136     if (!llvm::sys::locale::isPrint(c)) {
    137       // If next character is valid UTF-8, but not printable
    138       SmallString<16> expandedCP("<U+>");
    139       while (c) {
    140         expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(c%16));
    141         c/=16;
    142       }
    143       while (expandedCP.size() < 8)
    144         expandedCP.insert(expandedCP.begin()+3, llvm::hexdigit(0));
    145       return std::make_pair(expandedCP, false);
    146     }
    147 
    148     // If next character is valid UTF-8, and printable
    149     return std::make_pair(SmallString<16>(original_begin, cp_end), true);
    150 
    151   }
    152 
    153   // If next byte is not valid UTF-8 (and therefore not printable)
    154   SmallString<16> expandedByte("<XX>");
    155   unsigned char byte = SourceLine[*i];
    156   expandedByte[1] = llvm::hexdigit(byte / 16);
    157   expandedByte[2] = llvm::hexdigit(byte % 16);
    158   ++(*i);
    159   return std::make_pair(expandedByte, false);
    160 }
    161 
    162 static void expandTabs(std::string &SourceLine, unsigned TabStop) {
    163   size_t i = SourceLine.size();
    164   while (i>0) {
    165     i--;
    166     if (SourceLine[i]!='\t')
    167       continue;
    168     size_t tmp_i = i;
    169     std::pair<SmallString<16>,bool> res
    170       = printableTextForNextCharacter(SourceLine, &tmp_i, TabStop);
    171     SourceLine.replace(i, 1, res.first.c_str());
    172   }
    173 }
    174 
    175 /// This function takes a raw source line and produces a mapping from the bytes
    176 ///  of the printable representation of the line to the columns those printable
    177 ///  characters will appear at (numbering the first column as 0).
    178 ///
    179 /// If a byte 'i' corresponds to multiple columns (e.g. the byte contains a tab
    180 ///  character) then the array will map that byte to the first column the
    181 ///  tab appears at and the next value in the map will have been incremented
    182 ///  more than once.
    183 ///
    184 /// If a byte is the first in a sequence of bytes that together map to a single
    185 ///  entity in the output, then the array will map that byte to the appropriate
    186 ///  column while the subsequent bytes will be -1.
    187 ///
    188 /// The last element in the array does not correspond to any byte in the input
    189 ///  and instead is the number of columns needed to display the source
    190 ///
    191 /// example: (given a tabstop of 8)
    192 ///
    193 ///    "a \t \u3042" -> {0,1,2,8,9,-1,-1,11}
    194 ///
    195 ///  (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
    196 ///   display)
    197 static void byteToColumn(StringRef SourceLine, unsigned TabStop,
    198                          SmallVectorImpl<int> &out) {
    199   out.clear();
    200 
    201   if (SourceLine.empty()) {
    202     out.resize(1u,0);
    203     return;
    204   }
    205 
    206   out.resize(SourceLine.size()+1, -1);
    207 
    208   int columns = 0;
    209   size_t i = 0;
    210   while (i<SourceLine.size()) {
    211     out[i] = columns;
    212     std::pair<SmallString<16>,bool> res
    213       = printableTextForNextCharacter(SourceLine, &i, TabStop);
    214     columns += llvm::sys::locale::columnWidth(res.first);
    215   }
    216   out.back() = columns;
    217 }
    218 
    219 /// This function takes a raw source line and produces a mapping from columns
    220 ///  to the byte of the source line that produced the character displaying at
    221 ///  that column. This is the inverse of the mapping produced by byteToColumn()
    222 ///
    223 /// The last element in the array is the number of bytes in the source string
    224 ///
    225 /// example: (given a tabstop of 8)
    226 ///
    227 ///    "a \t \u3042" -> {0,1,2,-1,-1,-1,-1,-1,3,4,-1,7}
    228 ///
    229 ///  (\\u3042 is represented in UTF-8 by three bytes and takes two columns to
    230 ///   display)
    231 static void columnToByte(StringRef SourceLine, unsigned TabStop,
    232                          SmallVectorImpl<int> &out) {
    233   out.clear();
    234 
    235   if (SourceLine.empty()) {
    236     out.resize(1u, 0);
    237     return;
    238   }
    239 
    240   int columns = 0;
    241   size_t i = 0;
    242   while (i<SourceLine.size()) {
    243     out.resize(columns+1, -1);
    244     out.back() = i;
    245     std::pair<SmallString<16>,bool> res
    246       = printableTextForNextCharacter(SourceLine, &i, TabStop);
    247     columns += llvm::sys::locale::columnWidth(res.first);
    248   }
    249   out.resize(columns+1, -1);
    250   out.back() = i;
    251 }
    252 
    253 namespace {
    254 struct SourceColumnMap {
    255   SourceColumnMap(StringRef SourceLine, unsigned TabStop)
    256   : m_SourceLine(SourceLine) {
    257 
    258     ::byteToColumn(SourceLine, TabStop, m_byteToColumn);
    259     ::columnToByte(SourceLine, TabStop, m_columnToByte);
    260 
    261     assert(m_byteToColumn.size()==SourceLine.size()+1);
    262     assert(0 < m_byteToColumn.size() && 0 < m_columnToByte.size());
    263     assert(m_byteToColumn.size()
    264            == static_cast<unsigned>(m_columnToByte.back()+1));
    265     assert(static_cast<unsigned>(m_byteToColumn.back()+1)
    266            == m_columnToByte.size());
    267   }
    268   int columns() const { return m_byteToColumn.back(); }
    269   int bytes() const { return m_columnToByte.back(); }
    270 
    271   /// Map a byte to the column which it is at the start of, or return -1
    272   /// if it is not at the start of a column (for a UTF-8 trailing byte).
    273   int byteToColumn(int n) const {
    274     assert(0<=n && n<static_cast<int>(m_byteToColumn.size()));
    275     return m_byteToColumn[n];
    276   }
    277 
    278   /// Map a byte to the first column which contains it.
    279   int byteToContainingColumn(int N) const {
    280     assert(0 <= N && N < static_cast<int>(m_byteToColumn.size()));
    281     while (m_byteToColumn[N] == -1)
    282       --N;
    283     return m_byteToColumn[N];
    284   }
    285 
    286   /// Map a column to the byte which starts the column, or return -1 if
    287   /// the column the second or subsequent column of an expanded tab or similar
    288   /// multi-column entity.
    289   int columnToByte(int n) const {
    290     assert(0<=n && n<static_cast<int>(m_columnToByte.size()));
    291     return m_columnToByte[n];
    292   }
    293 
    294   /// Map from a byte index to the next byte which starts a column.
    295   int startOfNextColumn(int N) const {
    296     assert(0 <= N && N < static_cast<int>(m_byteToColumn.size() - 1));
    297     while (byteToColumn(++N) == -1) {}
    298     return N;
    299   }
    300 
    301   /// Map from a byte index to the previous byte which starts a column.
    302   int startOfPreviousColumn(int N) const {
    303     assert(0 < N && N < static_cast<int>(m_byteToColumn.size()));
    304     while (byteToColumn(--N) == -1) {}
    305     return N;
    306   }
    307 
    308   StringRef getSourceLine() const {
    309     return m_SourceLine;
    310   }
    311 
    312 private:
    313   const std::string m_SourceLine;
    314   SmallVector<int,200> m_byteToColumn;
    315   SmallVector<int,200> m_columnToByte;
    316 };
    317 } // end anonymous namespace
    318 
    319 /// When the source code line we want to print is too long for
    320 /// the terminal, select the "interesting" region.
    321 static void selectInterestingSourceRegion(std::string &SourceLine,
    322                                           std::string &CaretLine,
    323                                           std::string &FixItInsertionLine,
    324                                           unsigned Columns,
    325                                           const SourceColumnMap &map) {
    326   unsigned CaretColumns = CaretLine.size();
    327   unsigned FixItColumns = llvm::sys::locale::columnWidth(FixItInsertionLine);
    328   unsigned MaxColumns = std::max(static_cast<unsigned>(map.columns()),
    329                                  std::max(CaretColumns, FixItColumns));
    330   // if the number of columns is less than the desired number we're done
    331   if (MaxColumns <= Columns)
    332     return;
    333 
    334   // No special characters are allowed in CaretLine.
    335   assert(CaretLine.end() ==
    336          llvm::find_if(CaretLine, [](char c) { return c < ' ' || '~' < c; }));
    337 
    338   // Find the slice that we need to display the full caret line
    339   // correctly.
    340   unsigned CaretStart = 0, CaretEnd = CaretLine.size();
    341   for (; CaretStart != CaretEnd; ++CaretStart)
    342     if (!isWhitespace(CaretLine[CaretStart]))
    343       break;
    344 
    345   for (; CaretEnd != CaretStart; --CaretEnd)
    346     if (!isWhitespace(CaretLine[CaretEnd - 1]))
    347       break;
    348 
    349   // caret has already been inserted into CaretLine so the above whitespace
    350   // check is guaranteed to include the caret
    351 
    352   // If we have a fix-it line, make sure the slice includes all of the
    353   // fix-it information.
    354   if (!FixItInsertionLine.empty()) {
    355     unsigned FixItStart = 0, FixItEnd = FixItInsertionLine.size();
    356     for (; FixItStart != FixItEnd; ++FixItStart)
    357       if (!isWhitespace(FixItInsertionLine[FixItStart]))
    358         break;
    359 
    360     for (; FixItEnd != FixItStart; --FixItEnd)
    361       if (!isWhitespace(FixItInsertionLine[FixItEnd - 1]))
    362         break;
    363 
    364     // We can safely use the byte offset FixItStart as the column offset
    365     // because the characters up until FixItStart are all ASCII whitespace
    366     // characters.
    367     unsigned FixItStartCol = FixItStart;
    368     unsigned FixItEndCol
    369       = llvm::sys::locale::columnWidth(FixItInsertionLine.substr(0, FixItEnd));
    370 
    371     CaretStart = std::min(FixItStartCol, CaretStart);
    372     CaretEnd = std::max(FixItEndCol, CaretEnd);
    373   }
    374 
    375   // CaretEnd may have been set at the middle of a character
    376   // If it's not at a character's first column then advance it past the current
    377   //   character.
    378   while (static_cast<int>(CaretEnd) < map.columns() &&
    379          -1 == map.columnToByte(CaretEnd))
    380     ++CaretEnd;
    381 
    382   assert((static_cast<int>(CaretStart) > map.columns() ||
    383           -1!=map.columnToByte(CaretStart)) &&
    384          "CaretStart must not point to a column in the middle of a source"
    385          " line character");
    386   assert((static_cast<int>(CaretEnd) > map.columns() ||
    387           -1!=map.columnToByte(CaretEnd)) &&
    388          "CaretEnd must not point to a column in the middle of a source line"
    389          " character");
    390 
    391   // CaretLine[CaretStart, CaretEnd) contains all of the interesting
    392   // parts of the caret line. While this slice is smaller than the
    393   // number of columns we have, try to grow the slice to encompass
    394   // more context.
    395 
    396   unsigned SourceStart = map.columnToByte(std::min<unsigned>(CaretStart,
    397                                                              map.columns()));
    398   unsigned SourceEnd = map.columnToByte(std::min<unsigned>(CaretEnd,
    399                                                            map.columns()));
    400 
    401   unsigned CaretColumnsOutsideSource = CaretEnd-CaretStart
    402     - (map.byteToColumn(SourceEnd)-map.byteToColumn(SourceStart));
    403 
    404   char const *front_ellipse = "  ...";
    405   char const *front_space   = "     ";
    406   char const *back_ellipse = "...";
    407   unsigned ellipses_space = strlen(front_ellipse) + strlen(back_ellipse);
    408 
    409   unsigned TargetColumns = Columns;
    410   // Give us extra room for the ellipses
    411   //  and any of the caret line that extends past the source
    412   if (TargetColumns > ellipses_space+CaretColumnsOutsideSource)
    413     TargetColumns -= ellipses_space+CaretColumnsOutsideSource;
    414 
    415   while (SourceStart>0 || SourceEnd<SourceLine.size()) {
    416     bool ExpandedRegion = false;
    417 
    418     if (SourceStart>0) {
    419       unsigned NewStart = map.startOfPreviousColumn(SourceStart);
    420 
    421       // Skip over any whitespace we see here; we're looking for
    422       // another bit of interesting text.
    423       // FIXME: Detect non-ASCII whitespace characters too.
    424       while (NewStart && isWhitespace(SourceLine[NewStart]))
    425         NewStart = map.startOfPreviousColumn(NewStart);
    426 
    427       // Skip over this bit of "interesting" text.
    428       while (NewStart) {
    429         unsigned Prev = map.startOfPreviousColumn(NewStart);
    430         if (isWhitespace(SourceLine[Prev]))
    431           break;
    432         NewStart = Prev;
    433       }
    434 
    435       assert(map.byteToColumn(NewStart) != -1);
    436       unsigned NewColumns = map.byteToColumn(SourceEnd) -
    437                               map.byteToColumn(NewStart);
    438       if (NewColumns <= TargetColumns) {
    439         SourceStart = NewStart;
    440         ExpandedRegion = true;
    441       }
    442     }
    443 
    444     if (SourceEnd<SourceLine.size()) {
    445       unsigned NewEnd = map.startOfNextColumn(SourceEnd);
    446 
    447       // Skip over any whitespace we see here; we're looking for
    448       // another bit of interesting text.
    449       // FIXME: Detect non-ASCII whitespace characters too.
    450       while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
    451         NewEnd = map.startOfNextColumn(NewEnd);
    452 
    453       // Skip over this bit of "interesting" text.
    454       while (NewEnd < SourceLine.size() && isWhitespace(SourceLine[NewEnd]))
    455         NewEnd = map.startOfNextColumn(NewEnd);
    456 
    457       assert(map.byteToColumn(NewEnd) != -1);
    458       unsigned NewColumns = map.byteToColumn(NewEnd) -
    459                               map.byteToColumn(SourceStart);
    460       if (NewColumns <= TargetColumns) {
    461         SourceEnd = NewEnd;
    462         ExpandedRegion = true;
    463       }
    464     }
    465 
    466     if (!ExpandedRegion)
    467       break;
    468   }
    469 
    470   CaretStart = map.byteToColumn(SourceStart);
    471   CaretEnd = map.byteToColumn(SourceEnd) + CaretColumnsOutsideSource;
    472 
    473   // [CaretStart, CaretEnd) is the slice we want. Update the various
    474   // output lines to show only this slice, with two-space padding
    475   // before the lines so that it looks nicer.
    476 
    477   assert(CaretStart!=(unsigned)-1 && CaretEnd!=(unsigned)-1 &&
    478          SourceStart!=(unsigned)-1 && SourceEnd!=(unsigned)-1);
    479   assert(SourceStart <= SourceEnd);
    480   assert(CaretStart <= CaretEnd);
    481 
    482   unsigned BackColumnsRemoved
    483     = map.byteToColumn(SourceLine.size())-map.byteToColumn(SourceEnd);
    484   unsigned FrontColumnsRemoved = CaretStart;
    485   unsigned ColumnsKept = CaretEnd-CaretStart;
    486 
    487   // We checked up front that the line needed truncation
    488   assert(FrontColumnsRemoved+ColumnsKept+BackColumnsRemoved > Columns);
    489 
    490   // The line needs some truncation, and we'd prefer to keep the front
    491   //  if possible, so remove the back
    492   if (BackColumnsRemoved > strlen(back_ellipse))
    493     SourceLine.replace(SourceEnd, std::string::npos, back_ellipse);
    494 
    495   // If that's enough then we're done
    496   if (FrontColumnsRemoved+ColumnsKept <= Columns)
    497     return;
    498 
    499   // Otherwise remove the front as well
    500   if (FrontColumnsRemoved > strlen(front_ellipse)) {
    501     SourceLine.replace(0, SourceStart, front_ellipse);
    502     CaretLine.replace(0, CaretStart, front_space);
    503     if (!FixItInsertionLine.empty())
    504       FixItInsertionLine.replace(0, CaretStart, front_space);
    505   }
    506 }
    507 
    508 /// Skip over whitespace in the string, starting at the given
    509 /// index.
    510 ///
    511 /// \returns The index of the first non-whitespace character that is
    512 /// greater than or equal to Idx or, if no such character exists,
    513 /// returns the end of the string.
    514 static unsigned skipWhitespace(unsigned Idx, StringRef Str, unsigned Length) {
    515   while (Idx < Length && isWhitespace(Str[Idx]))
    516     ++Idx;
    517   return Idx;
    518 }
    519 
    520 /// If the given character is the start of some kind of
    521 /// balanced punctuation (e.g., quotes or parentheses), return the
    522 /// character that will terminate the punctuation.
    523 ///
    524 /// \returns The ending punctuation character, if any, or the NULL
    525 /// character if the input character does not start any punctuation.
    526 static inline char findMatchingPunctuation(char c) {
    527   switch (c) {
    528   case '\'': return '\'';
    529   case '`': return '\'';
    530   case '"':  return '"';
    531   case '(':  return ')';
    532   case '[': return ']';
    533   case '{': return '}';
    534   default: break;
    535   }
    536 
    537   return 0;
    538 }
    539 
    540 /// Find the end of the word starting at the given offset
    541 /// within a string.
    542 ///
    543 /// \returns the index pointing one character past the end of the
    544 /// word.
    545 static unsigned findEndOfWord(unsigned Start, StringRef Str,
    546                               unsigned Length, unsigned Column,
    547                               unsigned Columns) {
    548   assert(Start < Str.size() && "Invalid start position!");
    549   unsigned End = Start + 1;
    550 
    551   // If we are already at the end of the string, take that as the word.
    552   if (End == Str.size())
    553     return End;
    554 
    555   // Determine if the start of the string is actually opening
    556   // punctuation, e.g., a quote or parentheses.
    557   char EndPunct = findMatchingPunctuation(Str[Start]);
    558   if (!EndPunct) {
    559     // This is a normal word. Just find the first space character.
    560     while (End < Length && !isWhitespace(Str[End]))
    561       ++End;
    562     return End;
    563   }
    564 
    565   // We have the start of a balanced punctuation sequence (quotes,
    566   // parentheses, etc.). Determine the full sequence is.
    567   SmallString<16> PunctuationEndStack;
    568   PunctuationEndStack.push_back(EndPunct);
    569   while (End < Length && !PunctuationEndStack.empty()) {
    570     if (Str[End] == PunctuationEndStack.back())
    571       PunctuationEndStack.pop_back();
    572     else if (char SubEndPunct = findMatchingPunctuation(Str[End]))
    573       PunctuationEndStack.push_back(SubEndPunct);
    574 
    575     ++End;
    576   }
    577 
    578   // Find the first space character after the punctuation ended.
    579   while (End < Length && !isWhitespace(Str[End]))
    580     ++End;
    581 
    582   unsigned PunctWordLength = End - Start;
    583   if (// If the word fits on this line
    584       Column + PunctWordLength <= Columns ||
    585       // ... or the word is "short enough" to take up the next line
    586       // without too much ugly white space
    587       PunctWordLength < Columns/3)
    588     return End; // Take the whole thing as a single "word".
    589 
    590   // The whole quoted/parenthesized string is too long to print as a
    591   // single "word". Instead, find the "word" that starts just after
    592   // the punctuation and use that end-point instead. This will recurse
    593   // until it finds something small enough to consider a word.
    594   return findEndOfWord(Start + 1, Str, Length, Column + 1, Columns);
    595 }
    596 
    597 /// Print the given string to a stream, word-wrapping it to
    598 /// some number of columns in the process.
    599 ///
    600 /// \param OS the stream to which the word-wrapping string will be
    601 /// emitted.
    602 /// \param Str the string to word-wrap and output.
    603 /// \param Columns the number of columns to word-wrap to.
    604 /// \param Column the column number at which the first character of \p
    605 /// Str will be printed. This will be non-zero when part of the first
    606 /// line has already been printed.
    607 /// \param Bold if the current text should be bold
    608 /// \param Indentation the number of spaces to indent any lines beyond
    609 /// the first line.
    610 /// \returns true if word-wrapping was required, or false if the
    611 /// string fit on the first line.
    612 static bool printWordWrapped(raw_ostream &OS, StringRef Str,
    613                              unsigned Columns,
    614                              unsigned Column = 0,
    615                              bool Bold = false,
    616                              unsigned Indentation = WordWrapIndentation) {
    617   const unsigned Length = std::min(Str.find('\n'), Str.size());
    618   bool TextNormal = true;
    619 
    620   // The string used to indent each line.
    621   SmallString<16> IndentStr;
    622   IndentStr.assign(Indentation, ' ');
    623   bool Wrapped = false;
    624   for (unsigned WordStart = 0, WordEnd; WordStart < Length;
    625        WordStart = WordEnd) {
    626     // Find the beginning of the next word.
    627     WordStart = skipWhitespace(WordStart, Str, Length);
    628     if (WordStart == Length)
    629       break;
    630 
    631     // Find the end of this word.
    632     WordEnd = findEndOfWord(WordStart, Str, Length, Column, Columns);
    633 
    634     // Does this word fit on the current line?
    635     unsigned WordLength = WordEnd - WordStart;
    636     if (Column + WordLength < Columns) {
    637       // This word fits on the current line; print it there.
    638       if (WordStart) {
    639         OS << ' ';
    640         Column += 1;
    641       }
    642       applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
    643                                 TextNormal, Bold);
    644       Column += WordLength;
    645       continue;
    646     }
    647 
    648     // This word does not fit on the current line, so wrap to the next
    649     // line.
    650     OS << '\n';
    651     OS.write(&IndentStr[0], Indentation);
    652     applyTemplateHighlighting(OS, Str.substr(WordStart, WordLength),
    653                               TextNormal, Bold);
    654     Column = Indentation + WordLength;
    655     Wrapped = true;
    656   }
    657 
    658   // Append any remaning text from the message with its existing formatting.
    659   applyTemplateHighlighting(OS, Str.substr(Length), TextNormal, Bold);
    660 
    661   assert(TextNormal && "Text highlighted at end of diagnostic message.");
    662 
    663   return Wrapped;
    664 }
    665 
    666 TextDiagnostic::TextDiagnostic(raw_ostream &OS,
    667                                const LangOptions &LangOpts,
    668                                DiagnosticOptions *DiagOpts)
    669   : DiagnosticRenderer(LangOpts, DiagOpts), OS(OS) {}
    670 
    671 TextDiagnostic::~TextDiagnostic() {}
    672 
    673 void TextDiagnostic::emitDiagnosticMessage(
    674     FullSourceLoc Loc, PresumedLoc PLoc, DiagnosticsEngine::Level Level,
    675     StringRef Message, ArrayRef<clang::CharSourceRange> Ranges,
    676     DiagOrStoredDiag D) {
    677   uint64_t StartOfLocationInfo = OS.tell();
    678 
    679   // Emit the location of this particular diagnostic.
    680   if (Loc.isValid())
    681     emitDiagnosticLoc(Loc, PLoc, Level, Ranges);
    682 
    683   if (DiagOpts->ShowColors)
    684     OS.resetColor();
    685 
    686   if (DiagOpts->ShowLevel)
    687     printDiagnosticLevel(OS, Level, DiagOpts->ShowColors);
    688   printDiagnosticMessage(OS,
    689                          /*IsSupplemental*/ Level == DiagnosticsEngine::Note,
    690                          Message, OS.tell() - StartOfLocationInfo,
    691                          DiagOpts->MessageLength, DiagOpts->ShowColors);
    692 }
    693 
    694 /*static*/ void
    695 TextDiagnostic::printDiagnosticLevel(raw_ostream &OS,
    696                                      DiagnosticsEngine::Level Level,
    697                                      bool ShowColors) {
    698   if (ShowColors) {
    699     // Print diagnostic category in bold and color
    700     switch (Level) {
    701     case DiagnosticsEngine::Ignored:
    702       llvm_unreachable("Invalid diagnostic type");
    703     case DiagnosticsEngine::Note:    OS.changeColor(noteColor, true); break;
    704     case DiagnosticsEngine::Remark:  OS.changeColor(remarkColor, true); break;
    705     case DiagnosticsEngine::Warning: OS.changeColor(warningColor, true); break;
    706     case DiagnosticsEngine::Error:   OS.changeColor(errorColor, true); break;
    707     case DiagnosticsEngine::Fatal:   OS.changeColor(fatalColor, true); break;
    708     }
    709   }
    710 
    711   switch (Level) {
    712   case DiagnosticsEngine::Ignored:
    713     llvm_unreachable("Invalid diagnostic type");
    714   case DiagnosticsEngine::Note:    OS << "note: "; break;
    715   case DiagnosticsEngine::Remark:  OS << "remark: "; break;
    716   case DiagnosticsEngine::Warning: OS << "warning: "; break;
    717   case DiagnosticsEngine::Error:   OS << "error: "; break;
    718   case DiagnosticsEngine::Fatal:   OS << "fatal error: "; break;
    719   }
    720 
    721   if (ShowColors)
    722     OS.resetColor();
    723 }
    724 
    725 /*static*/
    726 void TextDiagnostic::printDiagnosticMessage(raw_ostream &OS,
    727                                             bool IsSupplemental,
    728                                             StringRef Message,
    729                                             unsigned CurrentColumn,
    730                                             unsigned Columns, bool ShowColors) {
    731   bool Bold = false;
    732   if (ShowColors && !IsSupplemental) {
    733     // Print primary diagnostic messages in bold and without color, to visually
    734     // indicate the transition from continuation notes and other output.
    735     OS.changeColor(savedColor, true);
    736     Bold = true;
    737   }
    738 
    739   if (Columns)
    740     printWordWrapped(OS, Message, Columns, CurrentColumn, Bold);
    741   else {
    742     bool Normal = true;
    743     applyTemplateHighlighting(OS, Message, Normal, Bold);
    744     assert(Normal && "Formatting should have returned to normal");
    745   }
    746 
    747   if (ShowColors)
    748     OS.resetColor();
    749   OS << '\n';
    750 }
    751 
    752 void TextDiagnostic::emitFilename(StringRef Filename, const SourceManager &SM) {
    753 #ifdef _WIN32
    754   SmallString<4096> TmpFilename;
    755 #endif
    756   if (DiagOpts->AbsolutePath) {
    757     auto File = SM.getFileManager().getFile(Filename);
    758     if (File) {
    759       // We want to print a simplified absolute path, i. e. without "dots".
    760       //
    761       // The hardest part here are the paths like "<part1>/<link>/../<part2>".
    762       // On Unix-like systems, we cannot just collapse "<link>/..", because
    763       // paths are resolved sequentially, and, thereby, the path
    764       // "<part1>/<part2>" may point to a different location. That is why
    765       // we use FileManager::getCanonicalName(), which expands all indirections
    766       // with llvm::sys::fs::real_path() and caches the result.
    767       //
    768       // On the other hand, it would be better to preserve as much of the
    769       // original path as possible, because that helps a user to recognize it.
    770       // real_path() expands all links, which sometimes too much. Luckily,
    771       // on Windows we can just use llvm::sys::path::remove_dots(), because,
    772       // on that system, both aforementioned paths point to the same place.
    773 #ifdef _WIN32
    774       TmpFilename = (*File)->getName();
    775       llvm::sys::fs::make_absolute(TmpFilename);
    776       llvm::sys::path::native(TmpFilename);
    777       llvm::sys::path::remove_dots(TmpFilename, /* remove_dot_dot */ true);
    778       Filename = StringRef(TmpFilename.data(), TmpFilename.size());
    779 #else
    780       Filename = SM.getFileManager().getCanonicalName(*File);
    781 #endif
    782     }
    783   }
    784 
    785   OS << Filename;
    786 }
    787 
    788 /// Print out the file/line/column information and include trace.
    789 ///
    790 /// This method handlen the emission of the diagnostic location information.
    791 /// This includes extracting as much location information as is present for
    792 /// the diagnostic and printing it, as well as any include stack or source
    793 /// ranges necessary.
    794 void TextDiagnostic::emitDiagnosticLoc(FullSourceLoc Loc, PresumedLoc PLoc,
    795                                        DiagnosticsEngine::Level Level,
    796                                        ArrayRef<CharSourceRange> Ranges) {
    797   if (PLoc.isInvalid()) {
    798     // At least print the file name if available:
    799     FileID FID = Loc.getFileID();
    800     if (FID.isValid()) {
    801       const FileEntry *FE = Loc.getFileEntry();
    802       if (FE && FE->isValid()) {
    803         emitFilename(FE->getName(), Loc.getManager());
    804         OS << ": ";
    805       }
    806     }
    807     return;
    808   }
    809   unsigned LineNo = PLoc.getLine();
    810 
    811   if (!DiagOpts->ShowLocation)
    812     return;
    813 
    814   if (DiagOpts->ShowColors)
    815     OS.changeColor(savedColor, true);
    816 
    817   emitFilename(PLoc.getFilename(), Loc.getManager());
    818   switch (DiagOpts->getFormat()) {
    819   case DiagnosticOptions::Clang:
    820     if (DiagOpts->ShowLine)
    821       OS << ':' << LineNo;
    822     break;
    823   case DiagnosticOptions::MSVC:  OS << '('  << LineNo; break;
    824   case DiagnosticOptions::Vi:    OS << " +" << LineNo; break;
    825   }
    826 
    827   if (DiagOpts->ShowColumn)
    828     // Compute the column number.
    829     if (unsigned ColNo = PLoc.getColumn()) {
    830       if (DiagOpts->getFormat() == DiagnosticOptions::MSVC) {
    831         OS << ',';
    832         // Visual Studio 2010 or earlier expects column number to be off by one
    833         if (LangOpts.MSCompatibilityVersion &&
    834             !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2012))
    835           ColNo--;
    836       } else
    837         OS << ':';
    838       OS << ColNo;
    839     }
    840   switch (DiagOpts->getFormat()) {
    841   case DiagnosticOptions::Clang:
    842   case DiagnosticOptions::Vi:    OS << ':';    break;
    843   case DiagnosticOptions::MSVC:
    844     // MSVC2013 and before print 'file(4) : error'. MSVC2015 gets rid of the
    845     // space and prints 'file(4): error'.
    846     OS << ')';
    847     if (LangOpts.MSCompatibilityVersion &&
    848         !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2015))
    849       OS << ' ';
    850     OS << ':';
    851     break;
    852   }
    853 
    854   if (DiagOpts->ShowSourceRanges && !Ranges.empty()) {
    855     FileID CaretFileID = Loc.getExpansionLoc().getFileID();
    856     bool PrintedRange = false;
    857 
    858     for (ArrayRef<CharSourceRange>::const_iterator RI = Ranges.begin(),
    859          RE = Ranges.end();
    860          RI != RE; ++RI) {
    861       // Ignore invalid ranges.
    862       if (!RI->isValid()) continue;
    863 
    864       auto &SM = Loc.getManager();
    865       SourceLocation B = SM.getExpansionLoc(RI->getBegin());
    866       CharSourceRange ERange = SM.getExpansionRange(RI->getEnd());
    867       SourceLocation E = ERange.getEnd();
    868       bool IsTokenRange = ERange.isTokenRange();
    869 
    870       std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(B);
    871       std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(E);
    872 
    873       // If the start or end of the range is in another file, just discard
    874       // it.
    875       if (BInfo.first != CaretFileID || EInfo.first != CaretFileID)
    876         continue;
    877 
    878       // Add in the length of the token, so that we cover multi-char
    879       // tokens.
    880       unsigned TokSize = 0;
    881       if (IsTokenRange)
    882         TokSize = Lexer::MeasureTokenLength(E, SM, LangOpts);
    883 
    884       FullSourceLoc BF(B, SM), EF(E, SM);
    885       OS << '{'
    886          << BF.getLineNumber() << ':' << BF.getColumnNumber() << '-'
    887          << EF.getLineNumber() << ':' << (EF.getColumnNumber() + TokSize)
    888          << '}';
    889       PrintedRange = true;
    890     }
    891 
    892     if (PrintedRange)
    893       OS << ':';
    894   }
    895   OS << ' ';
    896 }
    897 
    898 void TextDiagnostic::emitIncludeLocation(FullSourceLoc Loc, PresumedLoc PLoc) {
    899   if (DiagOpts->ShowLocation && PLoc.isValid())
    900     OS << "In file included from " << PLoc.getFilename() << ':'
    901        << PLoc.getLine() << ":\n";
    902   else
    903     OS << "In included file:\n";
    904 }
    905 
    906 void TextDiagnostic::emitImportLocation(FullSourceLoc Loc, PresumedLoc PLoc,
    907                                         StringRef ModuleName) {
    908   if (DiagOpts->ShowLocation && PLoc.isValid())
    909     OS << "In module '" << ModuleName << "' imported from "
    910        << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
    911   else
    912     OS << "In module '" << ModuleName << "':\n";
    913 }
    914 
    915 void TextDiagnostic::emitBuildingModuleLocation(FullSourceLoc Loc,
    916                                                 PresumedLoc PLoc,
    917                                                 StringRef ModuleName) {
    918   if (DiagOpts->ShowLocation && PLoc.isValid())
    919     OS << "While building module '" << ModuleName << "' imported from "
    920       << PLoc.getFilename() << ':' << PLoc.getLine() << ":\n";
    921   else
    922     OS << "While building module '" << ModuleName << "':\n";
    923 }
    924 
    925 /// Find the suitable set of lines to show to include a set of ranges.
    926 static llvm::Optional<std::pair<unsigned, unsigned>>
    927 findLinesForRange(const CharSourceRange &R, FileID FID,
    928                   const SourceManager &SM) {
    929   if (!R.isValid()) return None;
    930 
    931   SourceLocation Begin = R.getBegin();
    932   SourceLocation End = R.getEnd();
    933   if (SM.getFileID(Begin) != FID || SM.getFileID(End) != FID)
    934     return None;
    935 
    936   return std::make_pair(SM.getExpansionLineNumber(Begin),
    937                         SM.getExpansionLineNumber(End));
    938 }
    939 
    940 /// Add as much of range B into range A as possible without exceeding a maximum
    941 /// size of MaxRange. Ranges are inclusive.
    942 static std::pair<unsigned, unsigned>
    943 maybeAddRange(std::pair<unsigned, unsigned> A, std::pair<unsigned, unsigned> B,
    944               unsigned MaxRange) {
    945   // If A is already the maximum size, we're done.
    946   unsigned Slack = MaxRange - (A.second - A.first + 1);
    947   if (Slack == 0)
    948     return A;
    949 
    950   // Easy case: merge succeeds within MaxRange.
    951   unsigned Min = std::min(A.first, B.first);
    952   unsigned Max = std::max(A.second, B.second);
    953   if (Max - Min + 1 <= MaxRange)
    954     return {Min, Max};
    955 
    956   // If we can't reach B from A within MaxRange, there's nothing to do.
    957   // Don't add lines to the range that contain nothing interesting.
    958   if ((B.first > A.first && B.first - A.first + 1 > MaxRange) ||
    959       (B.second < A.second && A.second - B.second + 1 > MaxRange))
    960     return A;
    961 
    962   // Otherwise, expand A towards B to produce a range of size MaxRange. We
    963   // attempt to expand by the same amount in both directions if B strictly
    964   // contains A.
    965 
    966   // Expand downwards by up to half the available amount, then upwards as
    967   // much as possible, then downwards as much as possible.
    968   A.second = std::min(A.second + (Slack + 1) / 2, Max);
    969   Slack = MaxRange - (A.second - A.first + 1);
    970   A.first = std::max(Min + Slack, A.first) - Slack;
    971   A.second = std::min(A.first + MaxRange - 1, Max);
    972   return A;
    973 }
    974 
    975 /// Highlight a SourceRange (with ~'s) for any characters on LineNo.
    976 static void highlightRange(const CharSourceRange &R,
    977                            unsigned LineNo, FileID FID,
    978                            const SourceColumnMap &map,
    979                            std::string &CaretLine,
    980                            const SourceManager &SM,
    981                            const LangOptions &LangOpts) {
    982   if (!R.isValid()) return;
    983 
    984   SourceLocation Begin = R.getBegin();
    985   SourceLocation End = R.getEnd();
    986 
    987   unsigned StartLineNo = SM.getExpansionLineNumber(Begin);
    988   if (StartLineNo > LineNo || SM.getFileID(Begin) != FID)
    989     return;  // No intersection.
    990 
    991   unsigned EndLineNo = SM.getExpansionLineNumber(End);
    992   if (EndLineNo < LineNo || SM.getFileID(End) != FID)
    993     return;  // No intersection.
    994 
    995   // Compute the column number of the start.
    996   unsigned StartColNo = 0;
    997   if (StartLineNo == LineNo) {
    998     StartColNo = SM.getExpansionColumnNumber(Begin);
    999     if (StartColNo) --StartColNo;  // Zero base the col #.
   1000   }
   1001 
   1002   // Compute the column number of the end.
   1003   unsigned EndColNo = map.getSourceLine().size();
   1004   if (EndLineNo == LineNo) {
   1005     EndColNo = SM.getExpansionColumnNumber(End);
   1006     if (EndColNo) {
   1007       --EndColNo;  // Zero base the col #.
   1008 
   1009       // Add in the length of the token, so that we cover multi-char tokens if
   1010       // this is a token range.
   1011       if (R.isTokenRange())
   1012         EndColNo += Lexer::MeasureTokenLength(End, SM, LangOpts);
   1013     } else {
   1014       EndColNo = CaretLine.size();
   1015     }
   1016   }
   1017 
   1018   assert(StartColNo <= EndColNo && "Invalid range!");
   1019 
   1020   // Check that a token range does not highlight only whitespace.
   1021   if (R.isTokenRange()) {
   1022     // Pick the first non-whitespace column.
   1023     while (StartColNo < map.getSourceLine().size() &&
   1024            (map.getSourceLine()[StartColNo] == ' ' ||
   1025             map.getSourceLine()[StartColNo] == '\t'))
   1026       StartColNo = map.startOfNextColumn(StartColNo);
   1027 
   1028     // Pick the last non-whitespace column.
   1029     if (EndColNo > map.getSourceLine().size())
   1030       EndColNo = map.getSourceLine().size();
   1031     while (EndColNo &&
   1032            (map.getSourceLine()[EndColNo-1] == ' ' ||
   1033             map.getSourceLine()[EndColNo-1] == '\t'))
   1034       EndColNo = map.startOfPreviousColumn(EndColNo);
   1035 
   1036     // If the start/end passed each other, then we are trying to highlight a
   1037     // range that just exists in whitespace. That most likely means we have
   1038     // a multi-line highlighting range that covers a blank line.
   1039     if (StartColNo > EndColNo) {
   1040       assert(StartLineNo != EndLineNo && "trying to highlight whitespace");
   1041       StartColNo = EndColNo;
   1042     }
   1043   }
   1044 
   1045   assert(StartColNo <= map.getSourceLine().size() && "Invalid range!");
   1046   assert(EndColNo <= map.getSourceLine().size() && "Invalid range!");
   1047 
   1048   // Fill the range with ~'s.
   1049   StartColNo = map.byteToContainingColumn(StartColNo);
   1050   EndColNo = map.byteToContainingColumn(EndColNo);
   1051 
   1052   assert(StartColNo <= EndColNo && "Invalid range!");
   1053   if (CaretLine.size() < EndColNo)
   1054     CaretLine.resize(EndColNo,' ');
   1055   std::fill(CaretLine.begin()+StartColNo,CaretLine.begin()+EndColNo,'~');
   1056 }
   1057 
   1058 static std::string buildFixItInsertionLine(FileID FID,
   1059                                            unsigned LineNo,
   1060                                            const SourceColumnMap &map,
   1061                                            ArrayRef<FixItHint> Hints,
   1062                                            const SourceManager &SM,
   1063                                            const DiagnosticOptions *DiagOpts) {
   1064   std::string FixItInsertionLine;
   1065   if (Hints.empty() || !DiagOpts->ShowFixits)
   1066     return FixItInsertionLine;
   1067   unsigned PrevHintEndCol = 0;
   1068 
   1069   for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
   1070        I != E; ++I) {
   1071     if (!I->CodeToInsert.empty()) {
   1072       // We have an insertion hint. Determine whether the inserted
   1073       // code contains no newlines and is on the same line as the caret.
   1074       std::pair<FileID, unsigned> HintLocInfo
   1075         = SM.getDecomposedExpansionLoc(I->RemoveRange.getBegin());
   1076       if (FID == HintLocInfo.first &&
   1077           LineNo == SM.getLineNumber(HintLocInfo.first, HintLocInfo.second) &&
   1078           StringRef(I->CodeToInsert).find_first_of("\n\r") == StringRef::npos) {
   1079         // Insert the new code into the line just below the code
   1080         // that the user wrote.
   1081         // Note: When modifying this function, be very careful about what is a
   1082         // "column" (printed width, platform-dependent) and what is a
   1083         // "byte offset" (SourceManager "column").
   1084         unsigned HintByteOffset
   1085           = SM.getColumnNumber(HintLocInfo.first, HintLocInfo.second) - 1;
   1086 
   1087         // The hint must start inside the source or right at the end
   1088         assert(HintByteOffset < static_cast<unsigned>(map.bytes())+1);
   1089         unsigned HintCol = map.byteToContainingColumn(HintByteOffset);
   1090 
   1091         // If we inserted a long previous hint, push this one forwards, and add
   1092         // an extra space to show that this is not part of the previous
   1093         // completion. This is sort of the best we can do when two hints appear
   1094         // to overlap.
   1095         //
   1096         // Note that if this hint is located immediately after the previous
   1097         // hint, no space will be added, since the location is more important.
   1098         if (HintCol < PrevHintEndCol)
   1099           HintCol = PrevHintEndCol + 1;
   1100 
   1101         // This should NOT use HintByteOffset, because the source might have
   1102         // Unicode characters in earlier columns.
   1103         unsigned NewFixItLineSize = FixItInsertionLine.size() +
   1104           (HintCol - PrevHintEndCol) + I->CodeToInsert.size();
   1105         if (NewFixItLineSize > FixItInsertionLine.size())
   1106           FixItInsertionLine.resize(NewFixItLineSize, ' ');
   1107 
   1108         std::copy(I->CodeToInsert.begin(), I->CodeToInsert.end(),
   1109                   FixItInsertionLine.end() - I->CodeToInsert.size());
   1110 
   1111         PrevHintEndCol =
   1112           HintCol + llvm::sys::locale::columnWidth(I->CodeToInsert);
   1113       }
   1114     }
   1115   }
   1116 
   1117   expandTabs(FixItInsertionLine, DiagOpts->TabStop);
   1118 
   1119   return FixItInsertionLine;
   1120 }
   1121 
   1122 /// Emit a code snippet and caret line.
   1123 ///
   1124 /// This routine emits a single line's code snippet and caret line..
   1125 ///
   1126 /// \param Loc The location for the caret.
   1127 /// \param Ranges The underlined ranges for this code snippet.
   1128 /// \param Hints The FixIt hints active for this diagnostic.
   1129 void TextDiagnostic::emitSnippetAndCaret(
   1130     FullSourceLoc Loc, DiagnosticsEngine::Level Level,
   1131     SmallVectorImpl<CharSourceRange> &Ranges, ArrayRef<FixItHint> Hints) {
   1132   assert(Loc.isValid() && "must have a valid source location here");
   1133   assert(Loc.isFileID() && "must have a file location here");
   1134 
   1135   // If caret diagnostics are enabled and we have location, we want to
   1136   // emit the caret.  However, we only do this if the location moved
   1137   // from the last diagnostic, if the last diagnostic was a note that
   1138   // was part of a different warning or error diagnostic, or if the
   1139   // diagnostic has ranges.  We don't want to emit the same caret
   1140   // multiple times if one loc has multiple diagnostics.
   1141   if (!DiagOpts->ShowCarets)
   1142     return;
   1143   if (Loc == LastLoc && Ranges.empty() && Hints.empty() &&
   1144       (LastLevel != DiagnosticsEngine::Note || Level == LastLevel))
   1145     return;
   1146 
   1147   // Decompose the location into a FID/Offset pair.
   1148   std::pair<FileID, unsigned> LocInfo = Loc.getDecomposedLoc();
   1149   FileID FID = LocInfo.first;
   1150   const SourceManager &SM = Loc.getManager();
   1151 
   1152   // Get information about the buffer it points into.
   1153   bool Invalid = false;
   1154   StringRef BufData = Loc.getBufferData(&Invalid);
   1155   if (Invalid)
   1156     return;
   1157 
   1158   unsigned CaretLineNo = Loc.getLineNumber();
   1159   unsigned CaretColNo = Loc.getColumnNumber();
   1160 
   1161   // Arbitrarily stop showing snippets when the line is too long.
   1162   static const size_t MaxLineLengthToPrint = 4096;
   1163   if (CaretColNo > MaxLineLengthToPrint)
   1164     return;
   1165 
   1166   // Find the set of lines to include.
   1167   const unsigned MaxLines = DiagOpts->SnippetLineLimit;
   1168   std::pair<unsigned, unsigned> Lines = {CaretLineNo, CaretLineNo};
   1169   for (SmallVectorImpl<CharSourceRange>::iterator I = Ranges.begin(),
   1170                                                   E = Ranges.end();
   1171        I != E; ++I)
   1172     if (auto OptionalRange = findLinesForRange(*I, FID, SM))
   1173       Lines = maybeAddRange(Lines, *OptionalRange, MaxLines);
   1174 
   1175   for (unsigned LineNo = Lines.first; LineNo != Lines.second + 1; ++LineNo) {
   1176     const char *BufStart = BufData.data();
   1177     const char *BufEnd = BufStart + BufData.size();
   1178 
   1179     // Rewind from the current position to the start of the line.
   1180     const char *LineStart =
   1181         BufStart +
   1182         SM.getDecomposedLoc(SM.translateLineCol(FID, LineNo, 1)).second;
   1183     if (LineStart == BufEnd)
   1184       break;
   1185 
   1186     // Compute the line end.
   1187     const char *LineEnd = LineStart;
   1188     while (*LineEnd != '\n' && *LineEnd != '\r' && LineEnd != BufEnd)
   1189       ++LineEnd;
   1190 
   1191     // Arbitrarily stop showing snippets when the line is too long.
   1192     // FIXME: Don't print any lines in this case.
   1193     if (size_t(LineEnd - LineStart) > MaxLineLengthToPrint)
   1194       return;
   1195 
   1196     // Trim trailing null-bytes.
   1197     StringRef Line(LineStart, LineEnd - LineStart);
   1198     while (!Line.empty() && Line.back() == '\0' &&
   1199            (LineNo != CaretLineNo || Line.size() > CaretColNo))
   1200       Line = Line.drop_back();
   1201 
   1202     // Copy the line of code into an std::string for ease of manipulation.
   1203     std::string SourceLine(Line.begin(), Line.end());
   1204 
   1205     // Build the byte to column map.
   1206     const SourceColumnMap sourceColMap(SourceLine, DiagOpts->TabStop);
   1207 
   1208     // Create a line for the caret that is filled with spaces that is the same
   1209     // number of columns as the line of source code.
   1210     std::string CaretLine(sourceColMap.columns(), ' ');
   1211 
   1212     // Highlight all of the characters covered by Ranges with ~ characters.
   1213     for (SmallVectorImpl<CharSourceRange>::iterator I = Ranges.begin(),
   1214                                                     E = Ranges.end();
   1215          I != E; ++I)
   1216       highlightRange(*I, LineNo, FID, sourceColMap, CaretLine, SM, LangOpts);
   1217 
   1218     // Next, insert the caret itself.
   1219     if (CaretLineNo == LineNo) {
   1220       CaretColNo = sourceColMap.byteToContainingColumn(CaretColNo - 1);
   1221       if (CaretLine.size() < CaretColNo + 1)
   1222         CaretLine.resize(CaretColNo + 1, ' ');
   1223       CaretLine[CaretColNo] = '^';
   1224     }
   1225 
   1226     std::string FixItInsertionLine = buildFixItInsertionLine(
   1227         FID, LineNo, sourceColMap, Hints, SM, DiagOpts.get());
   1228 
   1229     // If the source line is too long for our terminal, select only the
   1230     // "interesting" source region within that line.
   1231     unsigned Columns = DiagOpts->MessageLength;
   1232     if (Columns)
   1233       selectInterestingSourceRegion(SourceLine, CaretLine, FixItInsertionLine,
   1234                                     Columns, sourceColMap);
   1235 
   1236     // If we are in -fdiagnostics-print-source-range-info mode, we are trying
   1237     // to produce easily machine parsable output.  Add a space before the
   1238     // source line and the caret to make it trivial to tell the main diagnostic
   1239     // line from what the user is intended to see.
   1240     if (DiagOpts->ShowSourceRanges) {
   1241       SourceLine = ' ' + SourceLine;
   1242       CaretLine = ' ' + CaretLine;
   1243     }
   1244 
   1245     // Finally, remove any blank spaces from the end of CaretLine.
   1246     while (!CaretLine.empty() && CaretLine[CaretLine.size() - 1] == ' ')
   1247       CaretLine.erase(CaretLine.end() - 1);
   1248 
   1249     // Emit what we have computed.
   1250     emitSnippet(SourceLine);
   1251 
   1252     if (!CaretLine.empty()) {
   1253       if (DiagOpts->ShowColors)
   1254         OS.changeColor(caretColor, true);
   1255       OS << CaretLine << '\n';
   1256       if (DiagOpts->ShowColors)
   1257         OS.resetColor();
   1258     }
   1259 
   1260     if (!FixItInsertionLine.empty()) {
   1261       if (DiagOpts->ShowColors)
   1262         // Print fixit line in color
   1263         OS.changeColor(fixitColor, false);
   1264       if (DiagOpts->ShowSourceRanges)
   1265         OS << ' ';
   1266       OS << FixItInsertionLine << '\n';
   1267       if (DiagOpts->ShowColors)
   1268         OS.resetColor();
   1269     }
   1270   }
   1271 
   1272   // Print out any parseable fixit information requested by the options.
   1273   emitParseableFixits(Hints, SM);
   1274 }
   1275 
   1276 void TextDiagnostic::emitSnippet(StringRef line) {
   1277   if (line.empty())
   1278     return;
   1279 
   1280   size_t i = 0;
   1281 
   1282   std::string to_print;
   1283   bool print_reversed = false;
   1284 
   1285   while (i<line.size()) {
   1286     std::pair<SmallString<16>,bool> res
   1287         = printableTextForNextCharacter(line, &i, DiagOpts->TabStop);
   1288     bool was_printable = res.second;
   1289 
   1290     if (DiagOpts->ShowColors && was_printable == print_reversed) {
   1291       if (print_reversed)
   1292         OS.reverseColor();
   1293       OS << to_print;
   1294       to_print.clear();
   1295       if (DiagOpts->ShowColors)
   1296         OS.resetColor();
   1297     }
   1298 
   1299     print_reversed = !was_printable;
   1300     to_print += res.first.str();
   1301   }
   1302 
   1303   if (print_reversed && DiagOpts->ShowColors)
   1304     OS.reverseColor();
   1305   OS << to_print;
   1306   if (print_reversed && DiagOpts->ShowColors)
   1307     OS.resetColor();
   1308 
   1309   OS << '\n';
   1310 }
   1311 
   1312 void TextDiagnostic::emitParseableFixits(ArrayRef<FixItHint> Hints,
   1313                                          const SourceManager &SM) {
   1314   if (!DiagOpts->ShowParseableFixits)
   1315     return;
   1316 
   1317   // We follow FixItRewriter's example in not (yet) handling
   1318   // fix-its in macros.
   1319   for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
   1320        I != E; ++I) {
   1321     if (I->RemoveRange.isInvalid() ||
   1322         I->RemoveRange.getBegin().isMacroID() ||
   1323         I->RemoveRange.getEnd().isMacroID())
   1324       return;
   1325   }
   1326 
   1327   for (ArrayRef<FixItHint>::iterator I = Hints.begin(), E = Hints.end();
   1328        I != E; ++I) {
   1329     SourceLocation BLoc = I->RemoveRange.getBegin();
   1330     SourceLocation ELoc = I->RemoveRange.getEnd();
   1331 
   1332     std::pair<FileID, unsigned> BInfo = SM.getDecomposedLoc(BLoc);
   1333     std::pair<FileID, unsigned> EInfo = SM.getDecomposedLoc(ELoc);
   1334 
   1335     // Adjust for token ranges.
   1336     if (I->RemoveRange.isTokenRange())
   1337       EInfo.second += Lexer::MeasureTokenLength(ELoc, SM, LangOpts);
   1338 
   1339     // We specifically do not do word-wrapping or tab-expansion here,
   1340     // because this is supposed to be easy to parse.
   1341     PresumedLoc PLoc = SM.getPresumedLoc(BLoc);
   1342     if (PLoc.isInvalid())
   1343       break;
   1344 
   1345     OS << "fix-it:\"";
   1346     OS.write_escaped(PLoc.getFilename());
   1347     OS << "\":{" << SM.getLineNumber(BInfo.first, BInfo.second)
   1348       << ':' << SM.getColumnNumber(BInfo.first, BInfo.second)
   1349       << '-' << SM.getLineNumber(EInfo.first, EInfo.second)
   1350       << ':' << SM.getColumnNumber(EInfo.first, EInfo.second)
   1351       << "}:\"";
   1352     OS.write_escaped(I->CodeToInsert);
   1353     OS << "\"\n";
   1354   }
   1355 }
   1356