[clang] [llvm] Alisonz/tmp2 (PR #204668)
Alison Zhang via cfe-commits
cfe-commits at lists.llvm.org
Thu Jun 18 13:46:15 PDT 2026
https://github.com/azhan92 updated https://github.com/llvm/llvm-project/pull/204668
>From 785f4a51a8e7696a273ac12a6ae1fbbf30f04c1c Mon Sep 17 00:00:00 2001
From: alisonzhang <alisonzhang at ibm.com>
Date: Thu, 18 Jun 2026 12:00:14 -0400
Subject: [PATCH 1/8] Changes from fexec-charset PR
---
clang/include/clang/Lex/Preprocessor.h | 4 ++++
clang/include/clang/Lex/TextEncoding.h | 23 +++++++++++++++++++++++
clang/lib/Lex/CMakeLists.txt | 1 +
clang/lib/Lex/TextEncoding.cpp | 18 ++++++++++++++++++
4 files changed, 46 insertions(+)
create mode 100644 clang/include/clang/Lex/TextEncoding.h
create mode 100644 clang/lib/Lex/TextEncoding.cpp
diff --git a/clang/include/clang/Lex/Preprocessor.h b/clang/include/clang/Lex/Preprocessor.h
index 8b684e85eb1c1..28a14eb76f8ef 100644
--- a/clang/include/clang/Lex/Preprocessor.h
+++ b/clang/include/clang/Lex/Preprocessor.h
@@ -30,6 +30,7 @@
#include "clang/Lex/ModuleMap.h"
#include "clang/Lex/PPCallbacks.h"
#include "clang/Lex/PPEmbedParameters.h"
+#include "clang/Lex/TextEncoding.h"
#include "clang/Lex/Token.h"
#include "clang/Lex/TokenLexer.h"
#include "clang/Support/Compiler.h"
@@ -198,6 +199,7 @@ class Preprocessor {
std::unique_ptr<ScratchBuffer> ScratchBuf;
HeaderSearch &HeaderInfo;
ModuleLoader &TheModuleLoader;
+ TextEncoding TE;
/// External source of macros.
ExternalPreprocessorSource *ExternalSource;
@@ -1265,6 +1267,8 @@ class Preprocessor {
Builtin::Context &getBuiltinInfo() { return *BuiltinInfo; }
llvm::BumpPtrAllocator &getPreprocessorAllocator() { return BP; }
+ TextEncoding &getTextEncoding() { return TE; }
+
void setExternalSource(ExternalPreprocessorSource *Source) {
ExternalSource = Source;
}
diff --git a/clang/include/clang/Lex/TextEncoding.h b/clang/include/clang/Lex/TextEncoding.h
new file mode 100644
index 0000000000000..ef718aec4d6d6
--- /dev/null
+++ b/clang/include/clang/Lex/TextEncoding.h
@@ -0,0 +1,23 @@
+//===-- clang/Lex/TextEncoding.h - Text Conversion Config -*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_CLANG_LEX_TEXTENCODING_H
+#define LLVM_CLANG_LEX_TEXTENCODING_H
+
+#include "clang/Basic/LangOptions.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Support/TextEncoding.h"
+
+enum ConversionAction { CA_NoConversion };
+
+class TextEncoding {
+public:
+ llvm::TextEncodingConverter *getConverter(ConversionAction Action) const;
+};
+
+#endif
diff --git a/clang/lib/Lex/CMakeLists.txt b/clang/lib/Lex/CMakeLists.txt
index f61737cd68021..7b0be7249cd99 100644
--- a/clang/lib/Lex/CMakeLists.txt
+++ b/clang/lib/Lex/CMakeLists.txt
@@ -29,6 +29,7 @@ add_clang_library(clangLex
Preprocessor.cpp
PreprocessorLexer.cpp
ScratchBuffer.cpp
+ TextEncoding.cpp
TokenConcatenation.cpp
TokenLexer.cpp
diff --git a/clang/lib/Lex/TextEncoding.cpp b/clang/lib/Lex/TextEncoding.cpp
new file mode 100644
index 0000000000000..33e5436367014
--- /dev/null
+++ b/clang/lib/Lex/TextEncoding.cpp
@@ -0,0 +1,18 @@
+//===--- TextEncoding.cpp -------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "clang/Lex/TextEncoding.h"
+#include "clang/Basic/DiagnosticDriver.h"
+
+llvm::TextEncodingConverter *
+TextEncoding::getConverter(ConversionAction Action) const {
+ switch (Action) {
+ default:
+ return nullptr;
+ }
+}
>From e8270826b0c143247bf62459e20e9643d7468c29 Mon Sep 17 00:00:00 2001
From: alisonzhang <alisonzhang at ibm.com>
Date: Thu, 18 Jun 2026 12:05:21 -0400
Subject: [PATCH 2/8] Changes from finput-charset PR
---
.../clang/Basic/DiagnosticCommonKinds.td | 3 +
clang/include/clang/Basic/SourceManager.h | 10 +-
.../include/clang/Frontend/CompilerInstance.h | 1 +
clang/include/clang/Lex/TextEncoding.h | 4 +-
clang/lib/Basic/SourceManager.cpp | 104 ++++++++++++++----
clang/lib/Frontend/CompilerInstance.cpp | 20 ++--
.../lib/Frontend/VerifyDiagnosticConsumer.cpp | 4 +-
clang/lib/Lex/ModuleMap.cpp | 5 +-
clang/lib/Lex/PPDirectives.cpp | 7 +-
clang/lib/Lex/Preprocessor.cpp | 5 +-
clang/lib/Lex/TextEncoding.cpp | 2 +
clang/lib/Serialization/ASTReader.cpp | 7 +-
12 files changed, 132 insertions(+), 40 deletions(-)
diff --git a/clang/include/clang/Basic/DiagnosticCommonKinds.td b/clang/include/clang/Basic/DiagnosticCommonKinds.td
index f2ed2f4698b8d..8ebac3908b465 100644
--- a/clang/include/clang/Basic/DiagnosticCommonKinds.td
+++ b/clang/include/clang/Basic/DiagnosticCommonKinds.td
@@ -417,6 +417,9 @@ def note_file_sloc_usage : Note<
"%plural{0:|: plus %2B (%human2B) for macro expansions}2">;
def note_file_misc_sloc_usage : Note<
"%0 additional files entered using a total of %1B (%human1B) of space">;
+def warn_charset_conversion_failed : Warning<
+ "conversion from source encoding failed for '%0': %1; interpreting as IBM-1047">,
+ InGroup<DiagGroup<"charset-conversion-failed">>;
// Modules
def err_module_format_unhandled : Error<
diff --git a/clang/include/clang/Basic/SourceManager.h b/clang/include/clang/Basic/SourceManager.h
index 4217b8683da1e..f7d91d612e4ab 100644
--- a/clang/include/clang/Basic/SourceManager.h
+++ b/clang/include/clang/Basic/SourceManager.h
@@ -50,6 +50,7 @@
#include "llvm/Support/Allocator.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/MemoryBuffer.h"
+#include "llvm/Support/TextEncoding.h"
#include <cassert>
#include <cstddef>
#include <map>
@@ -156,6 +157,11 @@ class alignas(8) ContentCache {
/// FIXME: Remove this once OrigEntry is a FileEntryRef with a stable name.
StringRef Filename;
+ /// Information on whether this is associated with a FileID for a file (as
+ /// opposed to a buffer) and, if so, what conversion (if any) was requested.
+ llvm::PointerIntPair<llvm::TextEncodingConverter *, 1u, bool>
+ FileIDConverterInfo;
+
/// A bump pointer allocated array of offsets for each source line.
///
/// This is lazily computed. The lines are owned by the SourceManager
@@ -918,6 +924,7 @@ class SourceManager : public RefCountedBase<SourceManager> {
/// being \#included from the specified IncludePosition.
FileID createFileID(FileEntryRef SourceFile, SourceLocation IncludePos,
SrcMgr::CharacteristicKind FileCharacter,
+ llvm::TextEncodingConverter *Converter = nullptr,
int LoadedID = 0,
SourceLocation::UIntTy LoadedOffset = 0);
@@ -942,7 +949,8 @@ class SourceManager : public RefCountedBase<SourceManager> {
/// Get the FileID for \p SourceFile if it exists. Otherwise, create a
/// new FileID for the \p SourceFile.
FileID getOrCreateFileID(FileEntryRef SourceFile,
- SrcMgr::CharacteristicKind FileCharacter);
+ SrcMgr::CharacteristicKind FileCharacter,
+ llvm::TextEncodingConverter *Converter = nullptr);
/// Creates an expansion SLocEntry for the substitution of an argument into a
/// function-like macro's body. Returns the start of the expansion.
diff --git a/clang/include/clang/Frontend/CompilerInstance.h b/clang/include/clang/Frontend/CompilerInstance.h
index bb0eddb918623..89a0d066afd6f 100644
--- a/clang/include/clang/Frontend/CompilerInstance.h
+++ b/clang/include/clang/Frontend/CompilerInstance.h
@@ -864,6 +864,7 @@ class CompilerInstance : public ModuleLoader {
///
/// \return True on success.
static bool InitializeSourceManager(const FrontendInputFile &Input,
+ llvm::TextEncodingConverter *,
DiagnosticsEngine &Diags,
FileManager &FileMgr,
SourceManager &SourceMgr);
diff --git a/clang/include/clang/Lex/TextEncoding.h b/clang/include/clang/Lex/TextEncoding.h
index ef718aec4d6d6..3e7653580e994 100644
--- a/clang/include/clang/Lex/TextEncoding.h
+++ b/clang/include/clang/Lex/TextEncoding.h
@@ -13,9 +13,11 @@
#include "llvm/ADT/StringRef.h"
#include "llvm/Support/TextEncoding.h"
-enum ConversionAction { CA_NoConversion };
+enum ConversionAction { CA_NoConversion, CA_FromInputEncoding };
class TextEncoding {
+std::unique_ptr<llvm::TextEncodingConverter> FromInputEncodingConverter;
+
public:
llvm::TextEncodingConverter *getConverter(ConversionAction Action) const;
};
diff --git a/clang/lib/Basic/SourceManager.cpp b/clang/lib/Basic/SourceManager.cpp
index b6cc6ec9365f5..950186866c86d 100644
--- a/clang/lib/Basic/SourceManager.cpp
+++ b/clang/lib/Basic/SourceManager.cpp
@@ -16,6 +16,7 @@
#include "clang/Basic/LLVM.h"
#include "clang/Basic/SourceLocation.h"
#include "clang/Basic/SourceManagerInternals.h"
+#include "clang/Lex/TextEncoding.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/MapVector.h"
#include "llvm/ADT/STLExtras.h"
@@ -31,6 +32,7 @@
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/MemoryBuffer.h"
#include "llvm/Support/raw_ostream.h"
+#include "llvm/Support/SmallVectorMemoryBuffer.h"
#include <algorithm>
#include <cassert>
#include <cstddef>
@@ -136,7 +138,51 @@ ContentCache::getBufferOrNone(DiagnosticsEngine &Diag, FileManager &FM,
Buffer = std::move(*BufferOrError);
- // Check that the file's size fits in an 'unsigned' (with room for a
+ // Unless this is a named pipe (in which case we can handle a mismatch),
+ // check that the file's size is the same as in the file entry (which may
+ // have come from a stat cache).
+ assert(Buffer->getBufferSize() >= (size_t)ContentsEntry->getSize());
+ if (!ContentsEntry->isNamedPipe() &&
+ Buffer->getBufferSize() < (size_t)ContentsEntry->getSize()) {
+ Diag.Report(Loc, diag::err_file_modified) << ContentsEntry->getName();
+
+ return std::nullopt;
+ }
+
+ // Convert source from the input charset to UTF-8 if necessary.
+ llvm::TextEncodingConverter *Converter = FileIDConverterInfo.getPointer();
+ if (Converter) {
+ StringRef OriginalBuf = Buffer->getBuffer();
+ llvm::SmallString<0> UTF8Buf;
+ UTF8Buf.reserve(OriginalBuf.size() + 1);
+
+ std::error_code EC = Converter->convert(OriginalBuf, UTF8Buf);
+ if (EC) {
+ // If conversion fails, emit a warning and fall back to interpreting the
+ // file as UTF-8 without conversion.
+ //
+ // This allows the compiler to accept system or third-party headers that
+ // are encoded in UTF-8 even if conversion to the option-specified input
+ // charset failed.
+ //
+ // Diagnostics already exist when files are not well-formed UTF-8.
+ //
+ // TODO: Add input byte offset information.
+ //
+ // TODO: Consider adjusting the message to omit the "interpreting as
+ // UTF-8" recovery description if the warning has been upgraded to an
+ // error.
+ Diag.Report(Loc, diag::warn_charset_conversion_failed)
+ << ContentsEntry->getName() << EC.message();
+ } else {
+ // TODO: Reclaim memory if the buffer size exceeds the content.
+ auto NewBuf = std::make_unique<llvm::SmallVectorMemoryBuffer>(
+ std::move(UTF8Buf), Buffer->getBufferIdentifier());
+ Buffer = std::move(NewBuf);
+ }
+ }
+
+ // Check that the buffer's size fits in an 'unsigned' (with room for a
// past-the-end value). This is deeply regrettable, but various parts of
// Clang (including elsewhere in this file!) use 'unsigned' to represent file
// offsets, line numbers, string literal lengths, and so on, and fail
@@ -151,22 +197,15 @@ ContentCache::getBufferOrNone(DiagnosticsEngine &Diag, FileManager &FM,
return std::nullopt;
}
- // Unless this is a named pipe (in which case we can handle a mismatch),
- // check that the file's size is the same as in the file entry (which may
- // have come from a stat cache).
- // The buffer will always be larger than the file size on z/OS in the presence
- // of characters outside the base character set.
- assert(Buffer->getBufferSize() >= (size_t)ContentsEntry->getSize());
- if (!ContentsEntry->isNamedPipe() &&
- Buffer->getBufferSize() < (size_t)ContentsEntry->getSize()) {
- Diag.Report(Loc, diag::err_file_modified) << ContentsEntry->getName();
-
- return std::nullopt;
- }
-
- // If the buffer is valid, check to see if it has a UTF Byte Order Mark
- // (BOM). We only support UTF-8 with and without a BOM right now. See
- // http://en.wikipedia.org/wiki/Byte_order_mark for more information.
+ // If the buffer is valid, check to see if it has a UTF Byte Order Mark (BOM)
+ // Note that any conversion requested using `-finput-charset` (if successful)
+ // has already occurred, so we are expecting UTF-8 with or without a BOM.
+ //
+ // In theory, if we see a non-UTF-8 BOM, we can assume that an appropriate
+ // conversion was not supplied via `-finput-charset` and we could try to
+ // convert based on the BOM.
+ //
+ // See http://en.wikipedia.org/wiki/Byte_order_mark for more information.
StringRef BufStr = Buffer->getBuffer();
const char *InvalidBOM = getInvalidBOM(BufStr);
@@ -537,15 +576,30 @@ FileID SourceManager::getNextFileID(FileID FID) const {
/// being \#included from the specified IncludePosition.
FileID SourceManager::createFileID(FileEntryRef SourceFile,
SourceLocation IncludePos,
- SrcMgr::CharacteristicKind FileCharacter,
+ SrcMgr::CharacteristicKind FileCharacter,
+ llvm::TextEncodingConverter *Converter,
int LoadedID,
SourceLocation::UIntTy LoadedOffset) {
SrcMgr::ContentCache &IR = getOrCreateContentCache(SourceFile,
isSystem(FileCharacter));
+ #ifndef NDEBUG
+ // Either the content cache has never been used for a FileID (and, if we are
+ // being asked to use a converter, there should be no valid buffer set up for
+ // it) or the conversion (or lack thereof) should be the same as that used
+ // previously.
+ auto [CacheConverter, CacheUsedByFileID] = IR.FileIDConverterInfo;
+ if (CacheUsedByFileID)
+ assert(CacheConverter == Converter);
+ else
+ assert(!Converter || IR.IsBufferInvalid || !IR.getBufferIfLoaded());
+#endif
+ IR.FileIDConverterInfo.setPointerAndInt(Converter, true);
+
// If this is a named pipe, immediately load the buffer to ensure subsequent
// calls to ContentCache::getSize() are accurate.
- if (IR.ContentsEntry->isNamedPipe())
+ // Do the same if character-encoding conversion was requested.
+ if (IR.ContentsEntry->isNamedPipe() || Converter)
(void)IR.getBufferOrNone(Diag, getFileManager(), SourceLocation());
return createFileIDImpl(IR, SourceFile.getName(), IncludePos, FileCharacter,
@@ -583,10 +637,12 @@ FileID SourceManager::createFileID(const llvm::MemoryBufferRef &Buffer,
/// new FileID for the \p SourceFile.
FileID
SourceManager::getOrCreateFileID(FileEntryRef SourceFile,
- SrcMgr::CharacteristicKind FileCharacter) {
+ SrcMgr::CharacteristicKind FileCharacter,
+ llvm::TextEncodingConverter *Converter) {
FileID ID = translateFile(SourceFile);
- return ID.isValid() ? ID : createFileID(SourceFile, SourceLocation(),
- FileCharacter);
+ return ID.isValid() ? ID
+ : createFileID(SourceFile, SourceLocation(),
+ FileCharacter, Converter);
}
/// createFileID - Create a new FileID for the specified ContentCache and
@@ -2340,8 +2396,8 @@ SourceManagerForFile::SourceManagerForFile(StringRef FileName,
std::make_unique<DiagnosticsEngine>(DiagnosticIDs::create(), *DiagOpts);
SourceMgr = std::make_unique<SourceManager>(*Diagnostics, *FileMgr);
FileEntryRef FE = llvm::cantFail(FileMgr->getFileRef(FileName));
- FileID ID =
- SourceMgr->createFileID(FE, SourceLocation(), clang::SrcMgr::C_User);
+ FileID ID = SourceMgr->createFileID(
+ FE, SourceLocation(), clang::SrcMgr::C_User, /*Converter=*/nullptr);
assert(ID.isValid());
SourceMgr->setMainFileID(ID);
}
diff --git a/clang/lib/Frontend/CompilerInstance.cpp b/clang/lib/Frontend/CompilerInstance.cpp
index 8aee45b5dc644..008bdb5bdeb0d 100644
--- a/clang/lib/Frontend/CompilerInstance.cpp
+++ b/clang/lib/Frontend/CompilerInstance.cpp
@@ -34,6 +34,7 @@
#include "clang/Lex/HeaderSearch.h"
#include "clang/Lex/Preprocessor.h"
#include "clang/Lex/PreprocessorOptions.h"
+#include "clang/Lex/TextEncoding.h"
#include "clang/Sema/CodeCompleteConsumer.h"
#include "clang/Sema/ParsedAttr.h"
#include "clang/Sema/Sema.h"
@@ -912,15 +913,20 @@ CompilerInstance::createOutputFileImpl(StringRef OutputPath, bool Binary,
// Initialization Utilities
bool CompilerInstance::InitializeSourceManager(const FrontendInputFile &Input){
- return InitializeSourceManager(Input, getDiagnostics(), getFileManager(),
- getSourceManager());
+ // Retrieve the converter to the internal charset if it exists.
+ llvm::TextEncodingConverter *Converter =
+ hasPreprocessor() ? getPreprocessor().getTextEncoding().getConverter(
+ CA_FromInputEncoding)
+ : nullptr;
+
+ return InitializeSourceManager(Input, Converter, getDiagnostics(),
+ getFileManager(), getSourceManager());
}
// static
-bool CompilerInstance::InitializeSourceManager(const FrontendInputFile &Input,
- DiagnosticsEngine &Diags,
- FileManager &FileMgr,
- SourceManager &SourceMgr) {
+bool CompilerInstance::InitializeSourceManager(
+ const FrontendInputFile &Input, llvm::TextEncodingConverter *Converter,
+ DiagnosticsEngine &Diags, FileManager &FileMgr, SourceManager &SourceMgr) {
SrcMgr::CharacteristicKind Kind =
Input.getKind().getFormat() == InputKind::ModuleMap
? Input.isSystem() ? SrcMgr::C_System_ModuleMap
@@ -950,7 +956,7 @@ bool CompilerInstance::InitializeSourceManager(const FrontendInputFile &Input,
}
SourceMgr.setMainFileID(
- SourceMgr.createFileID(*FileOrErr, SourceLocation(), Kind));
+ SourceMgr.createFileID(*FileOrErr, SourceLocation(), Kind, Converter));
assert(SourceMgr.getMainFileID().isValid() &&
"Couldn't establish MainFileID!");
diff --git a/clang/lib/Frontend/VerifyDiagnosticConsumer.cpp b/clang/lib/Frontend/VerifyDiagnosticConsumer.cpp
index 1bfe644b2525a..01e3b20e0c7cb 100644
--- a/clang/lib/Frontend/VerifyDiagnosticConsumer.cpp
+++ b/clang/lib/Frontend/VerifyDiagnosticConsumer.cpp
@@ -610,8 +610,10 @@ static bool ParseDirective(StringRef S, ExpectedData *ED, SourceManager &SM,
}
FileID FID = SM.translateFile(*File);
+ // FIXME: Figure out character-encoding converter treatment.
if (FID.isInvalid())
- FID = SM.createFileID(*File, Pos, SrcMgr::C_User);
+ FID = SM.createFileID(*File, Pos, SrcMgr::C_User,
+ /*Converter=*/nullptr);
if (PH.Next(Line) && Line > 0)
ExpectedLoc = SM.translateLineCol(FID, Line, 1);
diff --git a/clang/lib/Lex/ModuleMap.cpp b/clang/lib/Lex/ModuleMap.cpp
index 6c07386f89010..c7c1e04b76ea1 100644
--- a/clang/lib/Lex/ModuleMap.cpp
+++ b/clang/lib/Lex/ModuleMap.cpp
@@ -1473,7 +1473,10 @@ bool ModuleMap::parseModuleMapFile(FileEntryRef File, bool IsSystem,
if (LocalFID.isInvalid()) {
auto FileCharacter =
IsSystem ? SrcMgr::C_System_ModuleMap : SrcMgr::C_User_ModuleMap;
- LocalFID = SourceMgr.createFileID(File, ExternModuleLoc, FileCharacter);
+ // FIXME: Module map files are also textual "source files". For consistency,
+ // conversion should occur.
+ LocalFID = SourceMgr.createFileID(File, ExternModuleLoc, FileCharacter,
+ /*Converter=*/nullptr);
}
ID = LocalFID;
}
diff --git a/clang/lib/Lex/PPDirectives.cpp b/clang/lib/Lex/PPDirectives.cpp
index eb21a510dcf83..9f42ad12655e1 100644
--- a/clang/lib/Lex/PPDirectives.cpp
+++ b/clang/lib/Lex/PPDirectives.cpp
@@ -2796,7 +2796,12 @@ Preprocessor::ImportAction Preprocessor::HandleHeaderIncludeOrImport(
// position on the file where it will be included and after the expansions.
if (IncludePos.isMacroID())
IncludePos = SourceMgr.getExpansionRange(IncludePos).getEnd();
- FileID FID = SourceMgr.createFileID(*File, IncludePos, FileCharacter);
+ // Retrieve the converter to the internal charset if it exists.
+ llvm::TextEncodingConverter *Converter =
+ getTextEncoding().getConverter(CA_FromInputEncoding);
+
+ FileID FID =
+ SourceMgr.createFileID(*File, IncludePos, FileCharacter, Converter);
if (!FID.isValid()) {
TheModuleLoader.HadFatalFailure = true;
return ImportAction::Failure;
diff --git a/clang/lib/Lex/Preprocessor.cpp b/clang/lib/Lex/Preprocessor.cpp
index 1e21b4a94cea3..c5e32468bcc7e 100644
--- a/clang/lib/Lex/Preprocessor.cpp
+++ b/clang/lib/Lex/Preprocessor.cpp
@@ -649,8 +649,9 @@ void Preprocessor::EnterMainSourceFile() {
<< PPOpts.PCHThroughHeader;
return;
}
- setPCHThroughHeaderFileID(
- SourceMgr.createFileID(*File, SourceLocation(), SrcMgr::C_User));
+ // FIXME: Figure out character-encoding converter treatment.
+ setPCHThroughHeaderFileID(SourceMgr.createFileID(
+ *File, SourceLocation(), SrcMgr::C_User, /*Converter=*/nullptr));
}
// Skip tokens from the Predefines and if needed the main file.
diff --git a/clang/lib/Lex/TextEncoding.cpp b/clang/lib/Lex/TextEncoding.cpp
index 33e5436367014..7a0fea3798277 100644
--- a/clang/lib/Lex/TextEncoding.cpp
+++ b/clang/lib/Lex/TextEncoding.cpp
@@ -12,6 +12,8 @@
llvm::TextEncodingConverter *
TextEncoding::getConverter(ConversionAction Action) const {
switch (Action) {
+ case CA_FromInputEncoding:
+ return FromInputEncodingConverter.get();
default:
return nullptr;
}
diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index f8a6a38bb9b5c..379622e92984b 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -2002,8 +2002,11 @@ bool ASTReader::ReadSLocEntry(int ID) {
}
SrcMgr::CharacteristicKind
FileCharacter = (SrcMgr::CharacteristicKind)Record[2];
- FileID FID = SourceMgr.createFileID(*File, IncludeLoc, FileCharacter, ID,
- BaseOffset + Record[0]);
+ // Note: If conversion was originally necessary, OverriddenBuffer should be
+ // true and the associated handling will trigger.
+ FileID FID = SourceMgr.createFileID(*File, IncludeLoc, FileCharacter,
+ /*Converter=*/nullptr, ID,
+ BaseOffset + Record[0]);
SrcMgr::FileInfo &FileInfo = SourceMgr.getSLocEntry(FID).getFile();
FileInfo.NumCreatedFIDs = Record[5];
if (Record[3])
>From 569cd4065cdf76720c6636f9ff4af975f25f5d7d Mon Sep 17 00:00:00 2001
From: alisonzhang <alisonzhang at ibm.com>
Date: Thu, 18 Jun 2026 11:18:31 -0400
Subject: [PATCH 3/8] Add getEncodingNameFromFileTag function
---
llvm/include/llvm/Support/AutoConvert.h | 19 +++++++++++++++++++
1 file changed, 19 insertions(+)
diff --git a/llvm/include/llvm/Support/AutoConvert.h b/llvm/include/llvm/Support/AutoConvert.h
index d68b0e8b515e0..337befec1b352 100644
--- a/llvm/include/llvm/Support/AutoConvert.h
+++ b/llvm/include/llvm/Support/AutoConvert.h
@@ -105,6 +105,25 @@ inline ErrorOr<bool> needConversion(const Twine &FileName, const int FD = -1) {
return false;
}
+inline ErrorOr<std::string>
+getEncodingNameFromFileTag(const Twine &FileName, const int FD = -1) {
+#ifdef __MVS__
+ ErrorOr<__ccsid_t> TagOrErr = getzOSFileTag(FileName, FD);
+ if (!TagOrErr)
+ return TagOrErr.getError();
+
+ __ccsid_t Tag = *TagOrErr;
+ if (Tag == 0)
+ return std::string(); // Return empty string for no tag
+
+ char Buffer[16];
+ snprintf(Buffer, sizeof(Buffer), "%03d", Tag);
+ return std::string(Buffer);
+#else
+ return std::string(); // Return empty string for non-MVS platforms
+#endif
+}
+
} /* namespace llvm */
#endif /* __cplusplus */
>From 97ec4133c8977d91cad6ff26bda05f5672fdc1a7 Mon Sep 17 00:00:00 2001
From: alisonzhang <alisonzhang at ibm.com>
Date: Thu, 18 Jun 2026 12:17:48 -0400
Subject: [PATCH 4/8] Add TextEncodingConverter cache
---
llvm/include/llvm/Support/TextEncoding.h | 15 ++++++++
llvm/lib/Support/TextEncoding.cpp | 45 ++++++++++++++++++++++++
2 files changed, 60 insertions(+)
diff --git a/llvm/include/llvm/Support/TextEncoding.h b/llvm/include/llvm/Support/TextEncoding.h
index 8a304910aa5dd..6ca37ce6e7a4e 100644
--- a/llvm/include/llvm/Support/TextEncoding.h
+++ b/llvm/include/llvm/Support/TextEncoding.h
@@ -21,8 +21,10 @@
#include "llvm/Support/Compiler.h"
#include "llvm/Support/ErrorOr.h"
+#include <memory>
#include <string>
#include <system_error>
+#include <utility>
namespace llvm {
@@ -137,6 +139,19 @@ class TextEncodingConverter {
}
};
+/// Cache for TextEncodingConverter instances.
+class TextEncodingConverterCache {
+ public:
+ /// Get or create a cached TextEncodingConverter.
+ /// If the converter exists in the cache, returns it. Otherwise, creates a new
+ /// converter, caches it, and returns it.
+ /// \param[in] SourceEncoding the source character encoding name
+ /// \param[in] TargetEncoding the target character encoding name
+ /// \return pointer to the converter or an error code
+ LLVM_ABI static ErrorOr<TextEncodingConverter *>
+ getOrCreateConverter(StringRef SourceEncoding, StringRef TargetEncoding);
+};
+
} // namespace llvm
#endif
diff --git a/llvm/lib/Support/TextEncoding.cpp b/llvm/lib/Support/TextEncoding.cpp
index d36f02c1300b9..7797624ab278b 100644
--- a/llvm/lib/Support/TextEncoding.cpp
+++ b/llvm/lib/Support/TextEncoding.cpp
@@ -16,8 +16,11 @@
#include "llvm/ADT/SmallString.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/StringExtras.h"
+#include "llvm/ADT/StringMap.h"
#include "llvm/Support/ConvertEBCDIC.h"
+#include "llvm/Support/ManagedStatic.h"
#include <system_error>
+#include <utility>
#if HAVE_ICU
#if HAVE_WINDOWS_ICU
@@ -356,3 +359,45 @@ ErrorOr<TextEncodingConverter> TextEncodingConverter::create(StringRef From,
return std::make_error_code(std::errc::invalid_argument);
#endif
}
+
+namespace {
+// Global cache for TextEncodingConverter instances
+// Use StringMap which is designed for string keys
+using ConverterCache = StringMap<std::unique_ptr<TextEncodingConverter>>;
+
+static ManagedStatic<ConverterCache> GlobalConverterCache;
+} // namespace
+
+ErrorOr<TextEncodingConverter *>
+TextEncodingConverterCache::getOrCreateConverter(StringRef SourceEncoding,
+ StringRef TargetEncoding) {
+ // Don't create a converter if source and target are the same
+ if (SourceEncoding == TargetEncoding)
+ return nullptr;
+
+ // Create cache key by concatenating source and target with a separator
+ SmallString<64> Key;
+ Key = SourceEncoding;
+ Key += " -> ";
+ Key += TargetEncoding;
+
+ // Check if converter already exists
+ auto Iter = GlobalConverterCache->find(Key);
+ if (Iter != GlobalConverterCache->end()) {
+ return Iter->second.get();
+ }
+
+ // Create a new converter
+ ErrorOr<TextEncodingConverter> ErrorOrConverter =
+ TextEncodingConverter::create(SourceEncoding, TargetEncoding);
+ if (!ErrorOrConverter) {
+ return ErrorOrConverter.getError();
+ }
+
+ // Insert into cache and return pointer
+ auto NewConverter =
+ std::make_unique<TextEncodingConverter>(std::move(*ErrorOrConverter));
+ TextEncodingConverter *Result = NewConverter.get();
+ GlobalConverterCache->try_emplace(Key, std::move(NewConverter));
+ return Result;
+}
>From 2de55673f253d89bba98bd99fbff8fcc5a166105 Mon Sep 17 00:00:00 2001
From: alisonzhang <alisonzhang at ibm.com>
Date: Thu, 18 Jun 2026 11:31:00 -0400
Subject: [PATCH 5/8] Get filetag and create converter
---
clang/lib/Basic/SourceManager.cpp | 14 ++++++++++++++
1 file changed, 14 insertions(+)
diff --git a/clang/lib/Basic/SourceManager.cpp b/clang/lib/Basic/SourceManager.cpp
index 950186866c86d..945bc3e4177e2 100644
--- a/clang/lib/Basic/SourceManager.cpp
+++ b/clang/lib/Basic/SourceManager.cpp
@@ -583,6 +583,20 @@ FileID SourceManager::createFileID(FileEntryRef SourceFile,
SrcMgr::ContentCache &IR = getOrCreateContentCache(SourceFile,
isSystem(FileCharacter));
+ llvm::ErrorOr<std::string> Ccsid =
+ llvm::getEncodingNameFromFileTag(SourceFile.getName());
+ if (!Ccsid) {
+ Diag.Report(SourceLocation(), diag::err_cannot_open_file)
+ << SourceFile.getName() << Ccsid.getError().message();
+ return FileID();
+ }
+ if (!Ccsid->empty()) {
+ llvm::ErrorOr<llvm::TextEncodingConverter *> FileTagConverter =
+ llvm::TextEncodingConverterCache::getOrCreateConverter(*Ccsid, "UTF-8");
+ if (FileTagConverter)
+ Converter = *FileTagConverter;
+ }
+
#ifndef NDEBUG
// Either the content cache has never been used for a FileID (and, if we are
// being asked to use a converter, there should be no valid buffer set up for
>From f46342c689c8a87fd0b12d38320a1ff149af3051 Mon Sep 17 00:00:00 2001
From: alisonzhang <alisonzhang at ibm.com>
Date: Thu, 18 Jun 2026 15:44:13 -0400
Subject: [PATCH 6/8] Disable autoconversion for tagged files
---
clang/lib/Basic/SourceManager.cpp | 11 ++++++++++-
1 file changed, 10 insertions(+), 1 deletion(-)
diff --git a/clang/lib/Basic/SourceManager.cpp b/clang/lib/Basic/SourceManager.cpp
index 945bc3e4177e2..7af5b48481a1a 100644
--- a/clang/lib/Basic/SourceManager.cpp
+++ b/clang/lib/Basic/SourceManager.cpp
@@ -122,7 +122,16 @@ ContentCache::getBufferOrNone(DiagnosticsEngine &Diag, FileManager &FM,
// return paths.
IsBufferInvalid = true;
- auto BufferOrError = FM.getBufferForFile(*ContentsEntry, IsFileVolatile);
+ // If a converter is set, open the file in binary mode to get raw bytes
+ // and avoid platform-specific auto-conversion (e.g., EBCDIC->ASCII on z/OS,
+ // CRLF->LF on Windows). The explicit converter will handle all transformations.
+ bool NeedsExplicitConversion = FileIDConverterInfo.getPointer() != nullptr;
+ bool IsText = !NeedsExplicitConversion;
+
+ auto BufferOrError = FM.getBufferForFile(*ContentsEntry, IsFileVolatile,
+ /*RequiresNullTerminator=*/true,
+ /*MaybeLimit=*/std::nullopt,
+ IsText);
// If we were unable to open the file, then we are in an inconsistent
// situation where the content cache referenced a file which no longer
>From 067516a4f6fa9f27aa7ce9b3b3fdb79605b385ca Mon Sep 17 00:00:00 2001
From: alisonzhang <alisonzhang at ibm.com>
Date: Thu, 18 Jun 2026 16:23:33 -0400
Subject: [PATCH 7/8] Debug output from bob
---
clang/lib/Basic/FileManager.cpp | 2 ++
clang/lib/Basic/SourceManager.cpp | 27 +++++++++++++++++++++++++-
llvm/lib/Support/VirtualFileSystem.cpp | 6 ++++++
3 files changed, 34 insertions(+), 1 deletion(-)
diff --git a/clang/lib/Basic/FileManager.cpp b/clang/lib/Basic/FileManager.cpp
index 8fb3ba0a27aad..50471b43ef96a 100644
--- a/clang/lib/Basic/FileManager.cpp
+++ b/clang/lib/Basic/FileManager.cpp
@@ -601,6 +601,8 @@ std::error_code FileManager::getStatValue(StringRef Path,
//
// Because of this, check to see if the file exists with 'open'. If the
// open succeeds, use fstat to get the stat info.
+ llvm::errs() << "[DEBUG] FileManager: Opening file '" << Path << "' in "
+ << (IsText ? "TEXT" : "BINARY") << " mode\n";
auto OwnedFile =
IsText ? FS->openFileForRead(Path) : FS->openFileForReadBinary(Path);
diff --git a/clang/lib/Basic/SourceManager.cpp b/clang/lib/Basic/SourceManager.cpp
index 7af5b48481a1a..b91abc4c928eb 100644
--- a/clang/lib/Basic/SourceManager.cpp
+++ b/clang/lib/Basic/SourceManager.cpp
@@ -128,6 +128,11 @@ ContentCache::getBufferOrNone(DiagnosticsEngine &Diag, FileManager &FM,
bool NeedsExplicitConversion = FileIDConverterInfo.getPointer() != nullptr;
bool IsText = !NeedsExplicitConversion;
+ llvm::errs() << "[DEBUG] ContentCache::getBufferOrNone: Opening file '"
+ << ContentsEntry->getName() << "' - "
+ << (IsText ? "TEXT mode (no converter)" : "BINARY mode (converter present)")
+ << "\n";
+
auto BufferOrError = FM.getBufferForFile(*ContentsEntry, IsFileVolatile,
/*RequiresNullTerminator=*/true,
/*MaybeLimit=*/std::nullopt,
@@ -161,11 +166,18 @@ ContentCache::getBufferOrNone(DiagnosticsEngine &Diag, FileManager &FM,
// Convert source from the input charset to UTF-8 if necessary.
llvm::TextEncodingConverter *Converter = FileIDConverterInfo.getPointer();
if (Converter) {
+ llvm::errs() << "[DEBUG] SourceManager: Using converter for file '"
+ << ContentsEntry->getName() << "'\n";
StringRef OriginalBuf = Buffer->getBuffer();
+ llvm::errs() << "[DEBUG] SourceManager: Original buffer size: "
+ << OriginalBuf.size() << " bytes\n";
llvm::SmallString<0> UTF8Buf;
UTF8Buf.reserve(OriginalBuf.size() + 1);
std::error_code EC = Converter->convert(OriginalBuf, UTF8Buf);
+ llvm::errs() << "[DEBUG] SourceManager: Conversion "
+ << (EC ? "FAILED" : "succeeded")
+ << ", UTF8 buffer size: " << UTF8Buf.size() << " bytes\n";
if (EC) {
// If conversion fails, emit a warning and fall back to interpreting the
// file as UTF-8 without conversion.
@@ -600,10 +612,18 @@ FileID SourceManager::createFileID(FileEntryRef SourceFile,
return FileID();
}
if (!Ccsid->empty()) {
+ llvm::errs() << "[DEBUG] SourceManager::createFileID: File '" << SourceFile.getName()
+ << "' has encoding tag: '" << *Ccsid << "'\n";
llvm::ErrorOr<llvm::TextEncodingConverter *> FileTagConverter =
llvm::TextEncodingConverterCache::getOrCreateConverter(*Ccsid, "UTF-8");
- if (FileTagConverter)
+ if (FileTagConverter) {
Converter = *FileTagConverter;
+ llvm::errs() << "[DEBUG] SourceManager::createFileID: Converter obtained for '"
+ << *Ccsid << "' -> UTF-8\n";
+ } else {
+ llvm::errs() << "[DEBUG] SourceManager::createFileID: Failed to get converter: "
+ << FileTagConverter.getError().message() << "\n";
+ }
}
#ifndef NDEBUG
@@ -617,6 +637,11 @@ FileID SourceManager::createFileID(FileEntryRef SourceFile,
else
assert(!Converter || IR.IsBufferInvalid || !IR.getBufferIfLoaded());
#endif
+
+ if (Converter) {
+ llvm::errs() << "[DEBUG] SourceManager::createFileID: Setting converter for file '"
+ << SourceFile.getName() << "'\n";
+ }
IR.FileIDConverterInfo.setPointerAndInt(Converter, true);
// If this is a named pipe, immediately load the buffer to ensure subsequent
diff --git a/llvm/lib/Support/VirtualFileSystem.cpp b/llvm/lib/Support/VirtualFileSystem.cpp
index 42e8bb4f9958e..219fbd2097543 100644
--- a/llvm/lib/Support/VirtualFileSystem.cpp
+++ b/llvm/lib/Support/VirtualFileSystem.cpp
@@ -119,6 +119,8 @@ ErrorOr<std::unique_ptr<MemoryBuffer>>
FileSystem::getBufferForFile(const llvm::Twine &Name, int64_t FileSize,
bool RequiresNullTerminator, bool IsVolatile,
bool IsText) {
+ llvm::errs() << "[DEBUG] FileSystem::getBufferForFile: Opening '" << Name
+ << "' in " << (IsText ? "TEXT" : "BINARY") << " mode\n";
auto F = IsText ? openFileForRead(Name) : openFileForReadBinary(Name);
if (!F)
return F.getError();
@@ -348,6 +350,8 @@ ErrorOr<Status> RealFileSystem::status(const Twine &Path) {
ErrorOr<std::unique_ptr<File>>
RealFileSystem::openFileForRead(const Twine &Name) {
+ llvm::errs() << "[DEBUG] RealFileSystem::openFileForRead: Opening '"
+ << Name << "' in TEXT mode\n";
auto BypassSandbox = sys::sandbox::scopedDisable();
return openFileForReadWithFlags(Name, sys::fs::OF_Text);
@@ -355,6 +359,8 @@ RealFileSystem::openFileForRead(const Twine &Name) {
ErrorOr<std::unique_ptr<File>>
RealFileSystem::openFileForReadBinary(const Twine &Name) {
+ llvm::errs() << "[DEBUG] RealFileSystem::openFileForReadBinary: Opening '"
+ << Name << "' in BINARY mode\n";
auto BypassSandbox = sys::sandbox::scopedDisable();
return openFileForReadWithFlags(Name, sys::fs::OF_None);
>From ed912a41a25ceab606c8ebbf7be8f93a0eb8213d Mon Sep 17 00:00:00 2001
From: alisonzhang <alisonzhang at ibm.com>
Date: Thu, 18 Jun 2026 16:45:59 -0400
Subject: [PATCH 8/8] Changes from Bob
---
llvm/lib/Support/TextEncoding.cpp | 30 ++++++++++++++++++++++--------
1 file changed, 22 insertions(+), 8 deletions(-)
diff --git a/llvm/lib/Support/TextEncoding.cpp b/llvm/lib/Support/TextEncoding.cpp
index 7797624ab278b..f1801a9c9b69e 100644
--- a/llvm/lib/Support/TextEncoding.cpp
+++ b/llvm/lib/Support/TextEncoding.cpp
@@ -365,7 +365,12 @@ namespace {
// Use StringMap which is designed for string keys
using ConverterCache = StringMap<std::unique_ptr<TextEncodingConverter>>;
-static ManagedStatic<ConverterCache> GlobalConverterCache;
+struct ConverterCacheData {
+ ConverterCache Cache;
+ std::shared_mutex Mutex;
+};
+
+static ManagedStatic<ConverterCacheData> GlobalConverterCache;
} // namespace
ErrorOr<TextEncodingConverter *>
@@ -381,23 +386,32 @@ TextEncodingConverterCache::getOrCreateConverter(StringRef SourceEncoding,
Key += " -> ";
Key += TargetEncoding;
- // Check if converter already exists
- auto Iter = GlobalConverterCache->find(Key);
- if (Iter != GlobalConverterCache->end()) {
- return Iter->second.get();
+ // First, try to find existing converter with shared lock (allows concurrent reads)
+ {
+ std::shared_lock<std::shared_mutex> ReadLock(GlobalConverterCache->Mutex);
+ auto Iter = GlobalConverterCache->Cache.find(Key);
+ if (Iter != GlobalConverterCache->Cache.end())
+ return Iter->second.get();
}
+ // Not found, need to create - acquire unique lock for writing
+ std::unique_lock<std::shared_mutex> WriteLock(GlobalConverterCache->Mutex);
+
+ // Double-check: another thread might have created it while we were waiting
+ auto Iter = GlobalConverterCache->Cache.find(Key);
+ if (Iter != GlobalConverterCache->Cache.end())
+ return Iter->second.get();
+
// Create a new converter
ErrorOr<TextEncodingConverter> ErrorOrConverter =
TextEncodingConverter::create(SourceEncoding, TargetEncoding);
- if (!ErrorOrConverter) {
+ if (!ErrorOrConverter)
return ErrorOrConverter.getError();
- }
// Insert into cache and return pointer
auto NewConverter =
std::make_unique<TextEncodingConverter>(std::move(*ErrorOrConverter));
TextEncodingConverter *Result = NewConverter.get();
- GlobalConverterCache->try_emplace(Key, std::move(NewConverter));
+ GlobalConverterCache->Cache.try_emplace(Key, std::move(NewConverter));
return Result;
}
More information about the cfe-commits
mailing list