2009-06-21 11:36:54 +08:00
|
|
|
//===- SourceMgr.cpp - Manager for Simple Source Buffers & Diagnostics ----===//
|
2009-03-13 15:05:43 +08:00
|
|
|
//
|
|
|
|
// The LLVM Compiler Infrastructure
|
|
|
|
//
|
|
|
|
// This file is distributed under the University of Illinois Open Source
|
|
|
|
// License. See LICENSE.TXT for details.
|
|
|
|
//
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
//
|
2009-06-21 11:36:54 +08:00
|
|
|
// This file implements the SourceMgr class. This class is used as a simple
|
|
|
|
// substrate for diagnostics, #include handling, and other low level things for
|
|
|
|
// simple parsers.
|
2009-03-13 15:05:43 +08:00
|
|
|
//
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
|
2017-06-06 19:49:48 +08:00
|
|
|
#include "llvm/Support/SourceMgr.h"
|
2017-02-16 06:17:02 +08:00
|
|
|
#include "llvm/ADT/ArrayRef.h"
|
2016-08-12 08:18:03 +08:00
|
|
|
#include "llvm/ADT/STLExtras.h"
|
2017-06-06 19:49:48 +08:00
|
|
|
#include "llvm/ADT/SmallVector.h"
|
2017-02-16 06:17:02 +08:00
|
|
|
#include "llvm/ADT/StringRef.h"
|
2012-12-04 00:50:05 +08:00
|
|
|
#include "llvm/ADT/Twine.h"
|
2017-02-16 06:17:02 +08:00
|
|
|
#include "llvm/Support/ErrorOr.h"
|
2013-01-11 02:50:15 +08:00
|
|
|
#include "llvm/Support/Locale.h"
|
2012-12-04 00:50:05 +08:00
|
|
|
#include "llvm/Support/MemoryBuffer.h"
|
2014-05-16 21:16:30 +08:00
|
|
|
#include "llvm/Support/Path.h"
|
2017-02-16 06:17:02 +08:00
|
|
|
#include "llvm/Support/SMLoc.h"
|
2017-06-06 19:49:48 +08:00
|
|
|
#include "llvm/Support/raw_ostream.h"
|
2017-02-16 06:17:02 +08:00
|
|
|
#include <algorithm>
|
|
|
|
#include <cassert>
|
|
|
|
#include <cstddef>
|
[Support] Make line-number cache robust against access patterns.
Summary:
The LLVM SourceMgr class (which is used indirectly by Swift, though not Clang)
has a routine for looking up line numbers of SMLocs. This routine uses a
shared, special-purpose cache that handles exactly one access pattern
efficiently: looking up the line number of an SMLoc that points into the same
buffer as the last query made to the SourceMgr, at a location in the buffer at
or ahead of the last query.
When this works it's fine, but when it fails it's catastrophic for performancer:
one recent out-of-order access from a Swift utility routine ran for tens of
seconds, spending 99% of its time repeatedly scanning buffers for '\n'.
This change removes the shared cache from the SourceMgr and installs a new
cache in each SrcBuffer. The per-SrcBuffer caches are also "full", in the sense
that rather than caching a single last-query pointer, they cache _all_ the
line-ending offsets, in a binary-searchable array, such that once it's
populated (on first access), all subsequent access patterns run at the same
speed.
Performance measurements I've done show this is actually a little bit faster on
real codebases (though only a couple fractions of a percent). Memory usage is
up by a few tens to hundreds of bytes per SrcBuffer that has a line lookup done
on it; I've attempted to minimize this by using dynamic selection of integer
sized when storing offset arrays. But the main motive here is to
make-impossible the cases we don't always see, that show up by surprise when
there is an out-of-order access pattern.
Reviewers: jordan_rose
Reviewed By: jordan_rose
Subscribers: probinson, llvm-commits
Differential Revision: https://reviews.llvm.org/D45003
llvm-svn: 329470
2018-04-07 08:44:02 +08:00
|
|
|
#include <limits>
|
2017-02-16 06:17:02 +08:00
|
|
|
#include <memory>
|
|
|
|
#include <string>
|
|
|
|
#include <utility>
|
|
|
|
|
2009-03-13 15:05:43 +08:00
|
|
|
using namespace llvm;
|
|
|
|
|
2013-01-11 02:50:15 +08:00
|
|
|
static const size_t TabStop = 8;
|
|
|
|
|
2014-07-09 16:30:15 +08:00
|
|
|
unsigned SourceMgr::AddIncludeFile(const std::string &Filename,
|
|
|
|
SMLoc IncludeLoc,
|
|
|
|
std::string &IncludedFile) {
|
2011-06-01 21:10:15 +08:00
|
|
|
IncludedFile = Filename;
|
2014-07-07 01:43:13 +08:00
|
|
|
ErrorOr<std::unique_ptr<MemoryBuffer>> NewBufOrErr =
|
2014-11-06 09:13:27 +08:00
|
|
|
MemoryBuffer::getFile(IncludedFile);
|
2009-06-21 13:06:04 +08:00
|
|
|
|
|
|
|
// If the file didn't exist directly, see if it's in an include path.
|
2014-07-07 01:43:13 +08:00
|
|
|
for (unsigned i = 0, e = IncludeDirectories.size(); i != e && !NewBufOrErr;
|
|
|
|
++i) {
|
|
|
|
IncludedFile =
|
|
|
|
IncludeDirectories[i] + sys::path::get_separator().data() + Filename;
|
2014-11-06 09:13:27 +08:00
|
|
|
NewBufOrErr = MemoryBuffer::getFile(IncludedFile);
|
2009-06-21 13:06:04 +08:00
|
|
|
}
|
2010-01-27 18:13:11 +08:00
|
|
|
|
2014-07-07 01:43:13 +08:00
|
|
|
if (!NewBufOrErr)
|
2014-07-06 18:33:31 +08:00
|
|
|
return 0;
|
2009-06-21 13:06:04 +08:00
|
|
|
|
2014-08-22 04:44:56 +08:00
|
|
|
return AddNewSourceBuffer(std::move(*NewBufOrErr), IncludeLoc);
|
2009-06-21 13:06:04 +08:00
|
|
|
}
|
|
|
|
|
2014-07-06 18:33:31 +08:00
|
|
|
unsigned SourceMgr::FindBufferContainingLoc(SMLoc Loc) const {
|
2009-03-13 15:05:43 +08:00
|
|
|
for (unsigned i = 0, e = Buffers.size(); i != e; ++i)
|
2009-03-14 00:01:53 +08:00
|
|
|
if (Loc.getPointer() >= Buffers[i].Buffer->getBufferStart() &&
|
2009-03-19 04:36:45 +08:00
|
|
|
// Use <= here so that a pointer to the null at the end of the buffer
|
|
|
|
// is included as part of the buffer.
|
|
|
|
Loc.getPointer() <= Buffers[i].Buffer->getBufferEnd())
|
2014-07-06 18:33:31 +08:00
|
|
|
return i + 1;
|
|
|
|
return 0;
|
2009-03-13 15:05:43 +08:00
|
|
|
}
|
|
|
|
|
[Support] Make line-number cache robust against access patterns.
Summary:
The LLVM SourceMgr class (which is used indirectly by Swift, though not Clang)
has a routine for looking up line numbers of SMLocs. This routine uses a
shared, special-purpose cache that handles exactly one access pattern
efficiently: looking up the line number of an SMLoc that points into the same
buffer as the last query made to the SourceMgr, at a location in the buffer at
or ahead of the last query.
When this works it's fine, but when it fails it's catastrophic for performancer:
one recent out-of-order access from a Swift utility routine ran for tens of
seconds, spending 99% of its time repeatedly scanning buffers for '\n'.
This change removes the shared cache from the SourceMgr and installs a new
cache in each SrcBuffer. The per-SrcBuffer caches are also "full", in the sense
that rather than caching a single last-query pointer, they cache _all_ the
line-ending offsets, in a binary-searchable array, such that once it's
populated (on first access), all subsequent access patterns run at the same
speed.
Performance measurements I've done show this is actually a little bit faster on
real codebases (though only a couple fractions of a percent). Memory usage is
up by a few tens to hundreds of bytes per SrcBuffer that has a line lookup done
on it; I've attempted to minimize this by using dynamic selection of integer
sized when storing offset arrays. But the main motive here is to
make-impossible the cases we don't always see, that show up by surprise when
there is an out-of-order access pattern.
Reviewers: jordan_rose
Reviewed By: jordan_rose
Subscribers: probinson, llvm-commits
Differential Revision: https://reviews.llvm.org/D45003
llvm-svn: 329470
2018-04-07 08:44:02 +08:00
|
|
|
template <typename T>
|
|
|
|
unsigned SourceMgr::SrcBuffer::getLineNumber(const char *Ptr) const {
|
|
|
|
|
|
|
|
// Ensure OffsetCache is allocated and populated with offsets of all the
|
|
|
|
// '\n' bytes.
|
|
|
|
std::vector<T> *Offsets = nullptr;
|
|
|
|
if (OffsetCache.isNull()) {
|
|
|
|
Offsets = new std::vector<T>();
|
|
|
|
OffsetCache = Offsets;
|
|
|
|
size_t Sz = Buffer->getBufferSize();
|
|
|
|
assert(Sz <= std::numeric_limits<T>::max());
|
|
|
|
StringRef S = Buffer->getBuffer();
|
|
|
|
for (size_t N = 0; N < Sz; ++N) {
|
|
|
|
if (S[N] == '\n') {
|
|
|
|
Offsets->push_back(static_cast<T>(N));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
Offsets = OffsetCache.get<std::vector<T> *>();
|
|
|
|
}
|
|
|
|
|
|
|
|
const char *BufStart = Buffer->getBufferStart();
|
|
|
|
assert(Ptr >= BufStart && Ptr <= Buffer->getBufferEnd());
|
|
|
|
ptrdiff_t PtrDiff = Ptr - BufStart;
|
|
|
|
assert(PtrDiff >= 0 && static_cast<size_t>(PtrDiff) <= std::numeric_limits<T>::max());
|
|
|
|
T PtrOffset = static_cast<T>(PtrDiff);
|
|
|
|
|
|
|
|
// std::lower_bound returns the first EOL offset that's not-less-than
|
|
|
|
// PtrOffset, meaning the EOL that _ends the line_ that PtrOffset is on
|
|
|
|
// (including if PtrOffset refers to the EOL itself). If there's no such
|
|
|
|
// EOL, returns end().
|
|
|
|
auto EOL = std::lower_bound(Offsets->begin(), Offsets->end(), PtrOffset);
|
|
|
|
|
|
|
|
// Lines count from 1, so add 1 to the distance from the 0th line.
|
|
|
|
return (1 + (EOL - Offsets->begin()));
|
|
|
|
}
|
|
|
|
|
|
|
|
SourceMgr::SrcBuffer::SrcBuffer(SourceMgr::SrcBuffer &&Other)
|
|
|
|
: Buffer(std::move(Other.Buffer)),
|
|
|
|
OffsetCache(Other.OffsetCache),
|
|
|
|
IncludeLoc(Other.IncludeLoc) {
|
|
|
|
Other.OffsetCache = nullptr;
|
|
|
|
}
|
|
|
|
|
|
|
|
SourceMgr::SrcBuffer::~SrcBuffer() {
|
|
|
|
if (!OffsetCache.isNull()) {
|
|
|
|
if (OffsetCache.is<std::vector<uint8_t>*>())
|
|
|
|
delete OffsetCache.get<std::vector<uint8_t>*>();
|
|
|
|
else if (OffsetCache.is<std::vector<uint16_t>*>())
|
|
|
|
delete OffsetCache.get<std::vector<uint16_t>*>();
|
|
|
|
else if (OffsetCache.is<std::vector<uint32_t>*>())
|
|
|
|
delete OffsetCache.get<std::vector<uint32_t>*>();
|
|
|
|
else
|
|
|
|
delete OffsetCache.get<std::vector<uint64_t>*>();
|
|
|
|
OffsetCache = nullptr;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-05-06 06:17:32 +08:00
|
|
|
std::pair<unsigned, unsigned>
|
2014-07-06 18:33:31 +08:00
|
|
|
SourceMgr::getLineAndColumn(SMLoc Loc, unsigned BufferID) const {
|
|
|
|
if (!BufferID)
|
|
|
|
BufferID = FindBufferContainingLoc(Loc);
|
|
|
|
assert(BufferID && "Invalid Location!");
|
2010-01-27 18:13:11 +08:00
|
|
|
|
[Support] Make line-number cache robust against access patterns.
Summary:
The LLVM SourceMgr class (which is used indirectly by Swift, though not Clang)
has a routine for looking up line numbers of SMLocs. This routine uses a
shared, special-purpose cache that handles exactly one access pattern
efficiently: looking up the line number of an SMLoc that points into the same
buffer as the last query made to the SourceMgr, at a location in the buffer at
or ahead of the last query.
When this works it's fine, but when it fails it's catastrophic for performancer:
one recent out-of-order access from a Swift utility routine ran for tens of
seconds, spending 99% of its time repeatedly scanning buffers for '\n'.
This change removes the shared cache from the SourceMgr and installs a new
cache in each SrcBuffer. The per-SrcBuffer caches are also "full", in the sense
that rather than caching a single last-query pointer, they cache _all_ the
line-ending offsets, in a binary-searchable array, such that once it's
populated (on first access), all subsequent access patterns run at the same
speed.
Performance measurements I've done show this is actually a little bit faster on
real codebases (though only a couple fractions of a percent). Memory usage is
up by a few tens to hundreds of bytes per SrcBuffer that has a line lookup done
on it; I've attempted to minimize this by using dynamic selection of integer
sized when storing offset arrays. But the main motive here is to
make-impossible the cases we don't always see, that show up by surprise when
there is an out-of-order access pattern.
Reviewers: jordan_rose
Reviewed By: jordan_rose
Subscribers: probinson, llvm-commits
Differential Revision: https://reviews.llvm.org/D45003
llvm-svn: 329470
2018-04-07 08:44:02 +08:00
|
|
|
auto &SB = getBufferInfo(BufferID);
|
|
|
|
const char *Ptr = Loc.getPointer();
|
|
|
|
|
|
|
|
size_t Sz = SB.Buffer->getBufferSize();
|
|
|
|
assert(Sz <= std::numeric_limits<uint64_t>::max());
|
|
|
|
unsigned LineNo;
|
|
|
|
if (Sz <= std::numeric_limits<uint8_t>::max())
|
|
|
|
LineNo = SB.getLineNumber<uint8_t>(Ptr);
|
|
|
|
else if (Sz <= std::numeric_limits<uint16_t>::max())
|
|
|
|
LineNo = SB.getLineNumber<uint16_t>(Ptr);
|
|
|
|
else if (Sz <= std::numeric_limits<uint32_t>::max())
|
|
|
|
LineNo = SB.getLineNumber<uint32_t>(Ptr);
|
|
|
|
else
|
|
|
|
LineNo = SB.getLineNumber<uint64_t>(Ptr);
|
|
|
|
|
|
|
|
const char *BufStart = SB.Buffer->getBufferStart();
|
2012-05-06 06:17:32 +08:00
|
|
|
size_t NewlineOffs = StringRef(BufStart, Ptr-BufStart).find_last_of("\n\r");
|
2012-05-08 02:12:42 +08:00
|
|
|
if (NewlineOffs == StringRef::npos) NewlineOffs = ~(size_t)0;
|
2012-05-06 06:17:32 +08:00
|
|
|
return std::make_pair(LineNo, Ptr-BufStart-NewlineOffs);
|
2009-03-13 15:05:43 +08:00
|
|
|
}
|
|
|
|
|
2009-07-03 06:24:20 +08:00
|
|
|
void SourceMgr::PrintIncludeStack(SMLoc IncludeLoc, raw_ostream &OS) const {
|
2009-06-21 11:39:35 +08:00
|
|
|
if (IncludeLoc == SMLoc()) return; // Top of stack.
|
2010-01-27 18:13:11 +08:00
|
|
|
|
2014-07-06 18:33:31 +08:00
|
|
|
unsigned CurBuf = FindBufferContainingLoc(IncludeLoc);
|
|
|
|
assert(CurBuf && "Invalid or unspecified location!");
|
2009-03-13 15:05:43 +08:00
|
|
|
|
2009-07-03 06:24:20 +08:00
|
|
|
PrintIncludeStack(getBufferInfo(CurBuf).IncludeLoc, OS);
|
2010-01-27 18:13:11 +08:00
|
|
|
|
2009-07-03 06:24:20 +08:00
|
|
|
OS << "Included from "
|
|
|
|
<< getBufferInfo(CurBuf).Buffer->getBufferIdentifier()
|
|
|
|
<< ":" << FindLineNumber(IncludeLoc, CurBuf) << ":\n";
|
2009-03-13 15:05:43 +08:00
|
|
|
}
|
|
|
|
|
2011-10-16 13:43:57 +08:00
|
|
|
SMDiagnostic SourceMgr::GetMessage(SMLoc Loc, SourceMgr::DiagKind Kind,
|
2011-10-16 13:47:55 +08:00
|
|
|
const Twine &Msg,
|
2013-01-11 02:50:15 +08:00
|
|
|
ArrayRef<SMRange> Ranges,
|
|
|
|
ArrayRef<SMFixIt> FixIts) const {
|
2009-03-13 15:05:43 +08:00
|
|
|
// First thing to do: find the current buffer containing the specified
|
2012-05-07 00:20:49 +08:00
|
|
|
// location to pull out the source line.
|
2011-10-16 12:47:35 +08:00
|
|
|
SmallVector<std::pair<unsigned, unsigned>, 4> ColRanges;
|
2012-05-07 00:20:49 +08:00
|
|
|
std::pair<unsigned, unsigned> LineAndCol;
|
2016-10-02 00:38:28 +08:00
|
|
|
StringRef BufferID = "<unknown>";
|
2012-05-07 00:20:49 +08:00
|
|
|
std::string LineStr;
|
|
|
|
|
|
|
|
if (Loc.isValid()) {
|
2014-07-06 18:33:31 +08:00
|
|
|
unsigned CurBuf = FindBufferContainingLoc(Loc);
|
|
|
|
assert(CurBuf && "Invalid or unspecified location!");
|
2012-05-07 00:20:49 +08:00
|
|
|
|
2014-06-25 08:41:15 +08:00
|
|
|
const MemoryBuffer *CurMB = getMemoryBuffer(CurBuf);
|
2012-05-07 00:20:49 +08:00
|
|
|
BufferID = CurMB->getBufferIdentifier();
|
2011-10-16 12:47:35 +08:00
|
|
|
|
2012-05-07 00:20:49 +08:00
|
|
|
// Scan backward to find the start of the line.
|
|
|
|
const char *LineStart = Loc.getPointer();
|
|
|
|
const char *BufStart = CurMB->getBufferStart();
|
|
|
|
while (LineStart != BufStart && LineStart[-1] != '\n' &&
|
|
|
|
LineStart[-1] != '\r')
|
|
|
|
--LineStart;
|
|
|
|
|
|
|
|
// Get the end of the line.
|
|
|
|
const char *LineEnd = Loc.getPointer();
|
|
|
|
const char *BufEnd = CurMB->getBufferEnd();
|
|
|
|
while (LineEnd != BufEnd && LineEnd[0] != '\n' && LineEnd[0] != '\r')
|
|
|
|
++LineEnd;
|
|
|
|
LineStr = std::string(LineStart, LineEnd);
|
|
|
|
|
|
|
|
// Convert any ranges to column ranges that only intersect the line of the
|
|
|
|
// location.
|
|
|
|
for (unsigned i = 0, e = Ranges.size(); i != e; ++i) {
|
|
|
|
SMRange R = Ranges[i];
|
|
|
|
if (!R.isValid()) continue;
|
|
|
|
|
|
|
|
// If the line doesn't contain any part of the range, then ignore it.
|
|
|
|
if (R.Start.getPointer() > LineEnd || R.End.getPointer() < LineStart)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
// Ignore pieces of the range that go onto other lines.
|
|
|
|
if (R.Start.getPointer() < LineStart)
|
|
|
|
R.Start = SMLoc::getFromPointer(LineStart);
|
|
|
|
if (R.End.getPointer() > LineEnd)
|
|
|
|
R.End = SMLoc::getFromPointer(LineEnd);
|
|
|
|
|
|
|
|
// Translate from SMLoc ranges to column ranges.
|
2013-01-11 02:50:15 +08:00
|
|
|
// FIXME: Handle multibyte characters.
|
2012-05-07 00:20:49 +08:00
|
|
|
ColRanges.push_back(std::make_pair(R.Start.getPointer()-LineStart,
|
|
|
|
R.End.getPointer()-LineStart));
|
|
|
|
}
|
|
|
|
|
|
|
|
LineAndCol = getLineAndColumn(Loc, CurBuf);
|
2011-10-16 12:47:35 +08:00
|
|
|
}
|
2012-05-07 00:20:49 +08:00
|
|
|
|
|
|
|
return SMDiagnostic(*this, Loc, BufferID, LineAndCol.first,
|
2012-05-06 06:17:32 +08:00
|
|
|
LineAndCol.second-1, Kind, Msg.str(),
|
2013-01-11 02:50:15 +08:00
|
|
|
LineStr, ColRanges, FixIts);
|
2009-07-03 07:08:13 +08:00
|
|
|
}
|
|
|
|
|
2014-06-17 10:15:40 +08:00
|
|
|
void SourceMgr::PrintMessage(raw_ostream &OS, const SMDiagnostic &Diagnostic,
|
|
|
|
bool ShowColors) const {
|
2010-04-06 08:26:48 +08:00
|
|
|
// Report the message with the diagnostic handler if present.
|
|
|
|
if (DiagHandler) {
|
2011-10-16 13:43:57 +08:00
|
|
|
DiagHandler(Diagnostic, DiagContext);
|
2010-04-06 08:26:48 +08:00
|
|
|
return;
|
|
|
|
}
|
2010-12-10 01:37:32 +08:00
|
|
|
|
2014-06-17 10:15:40 +08:00
|
|
|
if (Diagnostic.getLoc().isValid()) {
|
2014-07-06 18:33:31 +08:00
|
|
|
unsigned CurBuf = FindBufferContainingLoc(Diagnostic.getLoc());
|
|
|
|
assert(CurBuf && "Invalid or unspecified location!");
|
2012-05-07 00:20:49 +08:00
|
|
|
PrintIncludeStack(getBufferInfo(CurBuf).IncludeLoc, OS);
|
|
|
|
}
|
2009-07-03 07:08:13 +08:00
|
|
|
|
2014-04-07 12:17:22 +08:00
|
|
|
Diagnostic.print(nullptr, OS, ShowColors);
|
2009-03-13 15:05:43 +08:00
|
|
|
}
|
2009-07-03 06:24:20 +08:00
|
|
|
|
2014-06-17 10:15:40 +08:00
|
|
|
void SourceMgr::PrintMessage(raw_ostream &OS, SMLoc Loc,
|
|
|
|
SourceMgr::DiagKind Kind,
|
|
|
|
const Twine &Msg, ArrayRef<SMRange> Ranges,
|
|
|
|
ArrayRef<SMFixIt> FixIts, bool ShowColors) const {
|
|
|
|
PrintMessage(OS, GetMessage(Loc, Kind, Msg, Ranges, FixIts), ShowColors);
|
|
|
|
}
|
|
|
|
|
2013-09-28 05:09:25 +08:00
|
|
|
void SourceMgr::PrintMessage(SMLoc Loc, SourceMgr::DiagKind Kind,
|
|
|
|
const Twine &Msg, ArrayRef<SMRange> Ranges,
|
|
|
|
ArrayRef<SMFixIt> FixIts, bool ShowColors) const {
|
2017-02-16 06:17:02 +08:00
|
|
|
PrintMessage(errs(), Loc, Kind, Msg, Ranges, FixIts, ShowColors);
|
2013-09-28 05:09:25 +08:00
|
|
|
}
|
|
|
|
|
2009-07-03 06:24:20 +08:00
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
// SMDiagnostic Implementation
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
|
2013-01-11 02:50:15 +08:00
|
|
|
SMDiagnostic::SMDiagnostic(const SourceMgr &sm, SMLoc L, StringRef FN,
|
2011-10-16 13:43:57 +08:00
|
|
|
int Line, int Col, SourceMgr::DiagKind Kind,
|
2013-01-11 02:50:15 +08:00
|
|
|
StringRef Msg, StringRef LineStr,
|
2017-02-16 06:17:02 +08:00
|
|
|
ArrayRef<std::pair<unsigned,unsigned>> Ranges,
|
2013-01-11 02:50:15 +08:00
|
|
|
ArrayRef<SMFixIt> Hints)
|
2011-10-16 13:43:57 +08:00
|
|
|
: SM(&sm), Loc(L), Filename(FN), LineNo(Line), ColumnNo(Col), Kind(Kind),
|
2013-01-11 02:50:15 +08:00
|
|
|
Message(Msg), LineContents(LineStr), Ranges(Ranges.vec()),
|
|
|
|
FixIts(Hints.begin(), Hints.end()) {
|
2018-04-09 00:46:22 +08:00
|
|
|
llvm::sort(FixIts.begin(), FixIts.end());
|
2013-01-11 02:50:15 +08:00
|
|
|
}
|
|
|
|
|
2013-02-15 20:30:38 +08:00
|
|
|
static void buildFixItLine(std::string &CaretLine, std::string &FixItLine,
|
|
|
|
ArrayRef<SMFixIt> FixIts, ArrayRef<char> SourceLine){
|
2013-01-11 02:50:15 +08:00
|
|
|
if (FixIts.empty())
|
|
|
|
return;
|
|
|
|
|
|
|
|
const char *LineStart = SourceLine.begin();
|
|
|
|
const char *LineEnd = SourceLine.end();
|
|
|
|
|
|
|
|
size_t PrevHintEndCol = 0;
|
|
|
|
|
|
|
|
for (ArrayRef<SMFixIt>::iterator I = FixIts.begin(), E = FixIts.end();
|
|
|
|
I != E; ++I) {
|
|
|
|
// If the fixit contains a newline or tab, ignore it.
|
|
|
|
if (I->getText().find_first_of("\n\r\t") != StringRef::npos)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
SMRange R = I->getRange();
|
|
|
|
|
|
|
|
// If the line doesn't contain any part of the range, then ignore it.
|
|
|
|
if (R.Start.getPointer() > LineEnd || R.End.getPointer() < LineStart)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
// Translate from SMLoc to column.
|
|
|
|
// Ignore pieces of the range that go onto other lines.
|
|
|
|
// FIXME: Handle multibyte characters in the source line.
|
|
|
|
unsigned FirstCol;
|
|
|
|
if (R.Start.getPointer() < LineStart)
|
|
|
|
FirstCol = 0;
|
|
|
|
else
|
|
|
|
FirstCol = R.Start.getPointer() - LineStart;
|
|
|
|
|
|
|
|
// If we inserted a long previous hint, push this one forwards, and add
|
|
|
|
// an extra space to show that this is not part of the previous
|
|
|
|
// completion. This is sort of the best we can do when two hints appear
|
|
|
|
// to overlap.
|
|
|
|
//
|
|
|
|
// Note that if this hint is located immediately after the previous
|
|
|
|
// hint, no space will be added, since the location is more important.
|
|
|
|
unsigned HintCol = FirstCol;
|
|
|
|
if (HintCol < PrevHintEndCol)
|
|
|
|
HintCol = PrevHintEndCol + 1;
|
|
|
|
|
|
|
|
// FIXME: This assertion is intended to catch unintended use of multibyte
|
|
|
|
// characters in fixits. If we decide to do this, we'll have to track
|
|
|
|
// separate byte widths for the source and fixit lines.
|
2017-02-16 06:17:02 +08:00
|
|
|
assert((size_t)sys::locale::columnWidth(I->getText()) ==
|
2013-01-11 02:50:15 +08:00
|
|
|
I->getText().size());
|
|
|
|
|
|
|
|
// This relies on one byte per column in our fixit hints.
|
|
|
|
unsigned LastColumnModified = HintCol + I->getText().size();
|
|
|
|
if (LastColumnModified > FixItLine.size())
|
|
|
|
FixItLine.resize(LastColumnModified, ' ');
|
|
|
|
|
|
|
|
std::copy(I->getText().begin(), I->getText().end(),
|
|
|
|
FixItLine.begin() + HintCol);
|
|
|
|
|
|
|
|
PrevHintEndCol = LastColumnModified;
|
|
|
|
|
|
|
|
// For replacements, mark the removal range with '~'.
|
|
|
|
// FIXME: Handle multibyte characters in the source line.
|
|
|
|
unsigned LastCol;
|
|
|
|
if (R.End.getPointer() >= LineEnd)
|
|
|
|
LastCol = LineEnd - LineStart;
|
|
|
|
else
|
|
|
|
LastCol = R.End.getPointer() - LineStart;
|
|
|
|
|
|
|
|
std::fill(&CaretLine[FirstCol], &CaretLine[LastCol], '~');
|
|
|
|
}
|
2011-10-16 13:43:57 +08:00
|
|
|
}
|
2011-10-16 12:47:35 +08:00
|
|
|
|
2013-01-11 02:50:15 +08:00
|
|
|
static void printSourceLine(raw_ostream &S, StringRef LineContents) {
|
|
|
|
// Print out the source line one character at a time, so we can expand tabs.
|
|
|
|
for (unsigned i = 0, e = LineContents.size(), OutCol = 0; i != e; ++i) {
|
|
|
|
if (LineContents[i] != '\t') {
|
|
|
|
S << LineContents[i];
|
|
|
|
++OutCol;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
// If we have a tab, emit at least one space, then round up to 8 columns.
|
|
|
|
do {
|
|
|
|
S << ' ';
|
|
|
|
++OutCol;
|
|
|
|
} while ((OutCol % TabStop) != 0);
|
|
|
|
}
|
|
|
|
S << '\n';
|
|
|
|
}
|
2011-10-16 12:47:35 +08:00
|
|
|
|
2013-01-11 10:37:55 +08:00
|
|
|
static bool isNonASCII(char c) {
|
|
|
|
return c & 0x80;
|
|
|
|
}
|
|
|
|
|
2015-06-16 04:30:22 +08:00
|
|
|
void SMDiagnostic::print(const char *ProgName, raw_ostream &S, bool ShowColors,
|
|
|
|
bool ShowKindLabel) const {
|
2012-07-21 02:29:44 +08:00
|
|
|
// Display colors only if OS supports colors.
|
|
|
|
ShowColors &= S.has_colors();
|
2012-04-19 03:04:15 +08:00
|
|
|
|
|
|
|
if (ShowColors)
|
|
|
|
S.changeColor(raw_ostream::SAVEDCOLOR, true);
|
|
|
|
|
2009-07-03 06:24:20 +08:00
|
|
|
if (ProgName && ProgName[0])
|
|
|
|
S << ProgName << ": ";
|
|
|
|
|
2010-01-21 18:13:27 +08:00
|
|
|
if (!Filename.empty()) {
|
|
|
|
if (Filename == "-")
|
|
|
|
S << "<stdin>";
|
|
|
|
else
|
|
|
|
S << Filename;
|
2010-01-27 18:13:11 +08:00
|
|
|
|
2010-01-21 18:13:27 +08:00
|
|
|
if (LineNo != -1) {
|
|
|
|
S << ':' << LineNo;
|
|
|
|
if (ColumnNo != -1)
|
|
|
|
S << ':' << (ColumnNo+1);
|
|
|
|
}
|
|
|
|
S << ": ";
|
2009-07-03 06:24:20 +08:00
|
|
|
}
|
2010-01-27 18:13:11 +08:00
|
|
|
|
2015-06-16 04:30:22 +08:00
|
|
|
if (ShowKindLabel) {
|
|
|
|
switch (Kind) {
|
|
|
|
case SourceMgr::DK_Error:
|
|
|
|
if (ShowColors)
|
|
|
|
S.changeColor(raw_ostream::RED, true);
|
|
|
|
S << "error: ";
|
|
|
|
break;
|
|
|
|
case SourceMgr::DK_Warning:
|
|
|
|
if (ShowColors)
|
|
|
|
S.changeColor(raw_ostream::MAGENTA, true);
|
|
|
|
S << "warning: ";
|
|
|
|
break;
|
|
|
|
case SourceMgr::DK_Note:
|
|
|
|
if (ShowColors)
|
|
|
|
S.changeColor(raw_ostream::BLACK, true);
|
|
|
|
S << "note: ";
|
|
|
|
break;
|
2017-10-13 07:56:02 +08:00
|
|
|
case SourceMgr::DK_Remark:
|
|
|
|
if (ShowColors)
|
|
|
|
S.changeColor(raw_ostream::BLUE, true);
|
|
|
|
S << "remark: ";
|
|
|
|
break;
|
2015-06-16 04:30:22 +08:00
|
|
|
}
|
2012-04-19 03:04:15 +08:00
|
|
|
|
2015-06-16 04:30:22 +08:00
|
|
|
if (ShowColors) {
|
|
|
|
S.resetColor();
|
|
|
|
S.changeColor(raw_ostream::SAVEDCOLOR, true);
|
|
|
|
}
|
2012-04-19 03:04:15 +08:00
|
|
|
}
|
|
|
|
|
2010-01-21 18:13:27 +08:00
|
|
|
S << Message << '\n';
|
2009-11-23 06:08:00 +08:00
|
|
|
|
2012-04-19 03:04:15 +08:00
|
|
|
if (ShowColors)
|
|
|
|
S.resetColor();
|
|
|
|
|
2011-10-16 13:47:55 +08:00
|
|
|
if (LineNo == -1 || ColumnNo == -1)
|
2011-10-16 12:47:35 +08:00
|
|
|
return;
|
2010-01-27 18:13:11 +08:00
|
|
|
|
2013-01-11 10:37:55 +08:00
|
|
|
// FIXME: If there are multibyte or multi-column characters in the source, all
|
|
|
|
// our ranges will be wrong. To do this properly, we'll need a byte-to-column
|
|
|
|
// map like Clang's TextDiagnostic. For now, we'll just handle tabs by
|
|
|
|
// expanding them later, and bail out rather than show incorrect ranges and
|
|
|
|
// misaligned fixits for any other odd characters.
|
2016-08-12 08:18:03 +08:00
|
|
|
if (find_if(LineContents, isNonASCII) != LineContents.end()) {
|
2013-01-11 02:50:15 +08:00
|
|
|
printSourceLine(S, LineContents);
|
|
|
|
return;
|
|
|
|
}
|
2013-01-11 10:37:55 +08:00
|
|
|
size_t NumColumns = LineContents.size();
|
2013-01-11 02:50:15 +08:00
|
|
|
|
2011-10-16 12:47:35 +08:00
|
|
|
// Build the line with the caret and ranges.
|
2013-01-11 02:50:15 +08:00
|
|
|
std::string CaretLine(NumColumns+1, ' ');
|
2011-10-16 12:47:35 +08:00
|
|
|
|
|
|
|
// Expand any ranges.
|
|
|
|
for (unsigned r = 0, e = Ranges.size(); r != e; ++r) {
|
|
|
|
std::pair<unsigned, unsigned> R = Ranges[r];
|
2013-01-11 02:50:15 +08:00
|
|
|
std::fill(&CaretLine[R.first],
|
|
|
|
&CaretLine[std::min((size_t)R.second, CaretLine.size())],
|
|
|
|
'~');
|
2011-10-16 12:47:35 +08:00
|
|
|
}
|
2013-01-11 02:50:15 +08:00
|
|
|
|
|
|
|
// Add any fix-its.
|
|
|
|
// FIXME: Find the beginning of the line properly for multibyte characters.
|
|
|
|
std::string FixItInsertionLine;
|
|
|
|
buildFixItLine(CaretLine, FixItInsertionLine, FixIts,
|
|
|
|
makeArrayRef(Loc.getPointer() - ColumnNo,
|
|
|
|
LineContents.size()));
|
|
|
|
|
2011-10-16 12:47:35 +08:00
|
|
|
// Finally, plop on the caret.
|
2013-01-11 02:50:15 +08:00
|
|
|
if (unsigned(ColumnNo) <= NumColumns)
|
2011-10-16 12:47:35 +08:00
|
|
|
CaretLine[ColumnNo] = '^';
|
|
|
|
else
|
2013-01-11 02:50:15 +08:00
|
|
|
CaretLine[NumColumns] = '^';
|
2011-10-16 12:47:35 +08:00
|
|
|
|
|
|
|
// ... and remove trailing whitespace so the output doesn't wrap for it. We
|
|
|
|
// know that the line isn't completely empty because it has the caret in it at
|
|
|
|
// least.
|
|
|
|
CaretLine.erase(CaretLine.find_last_not_of(' ')+1);
|
|
|
|
|
2013-01-11 02:50:15 +08:00
|
|
|
printSourceLine(S, LineContents);
|
2011-10-16 12:47:35 +08:00
|
|
|
|
2012-04-19 03:04:15 +08:00
|
|
|
if (ShowColors)
|
|
|
|
S.changeColor(raw_ostream::GREEN, true);
|
|
|
|
|
2011-10-16 12:47:35 +08:00
|
|
|
// Print out the caret line, matching tabs in the source line.
|
|
|
|
for (unsigned i = 0, e = CaretLine.size(), OutCol = 0; i != e; ++i) {
|
|
|
|
if (i >= LineContents.size() || LineContents[i] != '\t') {
|
|
|
|
S << CaretLine[i];
|
|
|
|
++OutCol;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Okay, we have a tab. Insert the appropriate number of characters.
|
|
|
|
do {
|
|
|
|
S << CaretLine[i];
|
|
|
|
++OutCol;
|
2013-01-11 02:50:15 +08:00
|
|
|
} while ((OutCol % TabStop) != 0);
|
2009-07-03 06:24:20 +08:00
|
|
|
}
|
2013-01-11 02:50:15 +08:00
|
|
|
S << '\n';
|
2012-04-19 03:04:15 +08:00
|
|
|
|
|
|
|
if (ShowColors)
|
|
|
|
S.resetColor();
|
2013-01-11 02:50:15 +08:00
|
|
|
|
|
|
|
// Print out the replacement line, matching tabs in the source line.
|
|
|
|
if (FixItInsertionLine.empty())
|
|
|
|
return;
|
2011-10-16 12:47:35 +08:00
|
|
|
|
2013-09-28 05:24:36 +08:00
|
|
|
for (size_t i = 0, e = FixItInsertionLine.size(), OutCol = 0; i < e; ++i) {
|
2013-01-11 02:50:15 +08:00
|
|
|
if (i >= LineContents.size() || LineContents[i] != '\t') {
|
|
|
|
S << FixItInsertionLine[i];
|
|
|
|
++OutCol;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Okay, we have a tab. Insert the appropriate number of characters.
|
|
|
|
do {
|
|
|
|
S << FixItInsertionLine[i];
|
|
|
|
// FIXME: This is trying not to break up replacements, but then to re-sync
|
|
|
|
// with the tabs between replacements. This will fail, though, if two
|
|
|
|
// fix-it replacements are exactly adjacent, or if a fix-it contains a
|
|
|
|
// space. Really we should be precomputing column widths, which we'll
|
|
|
|
// need anyway for multibyte chars.
|
|
|
|
if (FixItInsertionLine[i] != ' ')
|
|
|
|
++i;
|
|
|
|
++OutCol;
|
|
|
|
} while (((OutCol % TabStop) != 0) && i != e);
|
|
|
|
}
|
2011-10-16 12:47:35 +08:00
|
|
|
S << '\n';
|
2009-07-03 06:24:20 +08:00
|
|
|
}
|