diff --git a/lld/MachO/Driver.h b/lld/MachO/Driver.h --- a/lld/MachO/Driver.h +++ b/lld/MachO/Driver.h @@ -81,9 +81,8 @@ notFounds.insert(path.str()); } - // Writes the dependencies to specified path. - // The content is sorted by its Op Code, then within each section, - // alphabetical order. + // Writes the dependencies to specified path. The content is first sorted by + // OpCode and then by the filename (in alphabetical order). void write(llvm::StringRef version, const llvm::SetVector &inputs, llvm::StringRef output); diff --git a/lld/MachO/InputFiles.h b/lld/MachO/InputFiles.h --- a/lld/MachO/InputFiles.h +++ b/lld/MachO/InputFiles.h @@ -177,6 +177,7 @@ void parseLoadCommands(MemoryBufferRef mb); void parseReexports(const llvm::MachO::InterfaceFile &interface); + bool isReferenced() const { return numReferencedSymbols > 0; } static bool classof(const InputFile *f) { return f->kind() == DylibKind; } @@ -187,21 +188,17 @@ uint32_t compatibilityVersion = 0; uint32_t currentVersion = 0; int64_t ordinal = 0; // Ordinal numbering starts from 1, so 0 is a sentinel + unsigned numReferencedSymbols = 0; RefState refState; bool reexport = false; bool forceNeeded = false; bool forceWeakImport = false; bool deadStrippable = false; bool explicitlyLinked = false; - - unsigned numReferencedSymbols = 0; - - bool isReferenced() const { return numReferencedSymbols > 0; } - // An executable can be used as a bundle loader that will load the output // file being linked, and that contains symbols referenced, but not // implemented in the bundle. When used like this, it is very similar - // to a Dylib, so we re-used the same class to represent it. + // to a dylib, so we've used the same class to represent it. bool isBundleLoader; private: diff --git a/lld/MachO/OutputSection.h b/lld/MachO/OutputSection.h --- a/lld/MachO/OutputSection.h +++ b/lld/MachO/OutputSection.h @@ -58,13 +58,23 @@ // Unneeded sections are omitted entirely (header and body). virtual bool isNeeded() const { return true; } - virtual void finalize() { - // TODO investigate refactoring synthetic section finalization logic into - // overrides of this function. - } + // The implementations of this method can assume that it is only called right + // before addresses get assigned to this particular OutputSection. In + // particular, this means that it gets called only after addresses have been + // assigned to output sections that occur earlier in the output binary. + // Naturally, this means different sections' finalize() methods cannot execute + // concurrently with each other. As such, avoid using this method for + // operations that do not require this strict sequential guarantee. + // + // Operations that need to occur late in the linking process, but which do not + // need the sequential guarantee, should be named `finalizeContents()`. See + // e.g. LinkEditSection::finalizeContents() and + // CStringSection::finalizeContents(). + virtual void finalize() {} virtual void writeTo(uint8_t *buf) const = 0; + // Handle section$start$ and section$end$ symbols. void assignAddressesToStartEndSymbols(); StringRef name; diff --git a/lld/MachO/SyntheticSections.h b/lld/MachO/SyntheticSections.h --- a/lld/MachO/SyntheticSections.h +++ b/lld/MachO/SyntheticSections.h @@ -62,6 +62,8 @@ align = target->wordSize; } + // Implementations of this method can assume that the regular (non-__LINKEDIT) + // sections already have their addresses assigned. virtual void finalizeContents() {} // Sections in __LINKEDIT are special: their offsets are recorded in the diff --git a/lld/MachO/UnwindInfoSection.cpp b/lld/MachO/UnwindInfoSection.cpp --- a/lld/MachO/UnwindInfoSection.cpp +++ b/lld/MachO/UnwindInfoSection.cpp @@ -392,7 +392,7 @@ } // Scan the __LD,__compact_unwind entries and compute the space needs of -// __TEXT,__unwind_info and __TEXT,__eh_frame +// __TEXT,__unwind_info and __TEXT,__eh_frame. template void UnwindInfoSectionImpl::finalize() { if (symbols.empty()) return; diff --git a/lld/MachO/Writer.cpp b/lld/MachO/Writer.cpp --- a/lld/MachO/Writer.cpp +++ b/lld/MachO/Writer.cpp @@ -1108,8 +1108,10 @@ treatSpecialUndefineds(); if (config->entry && !isa(config->entry)) prepareBranchTarget(config->entry); + // Canonicalization of all pointers to InputSections should be handled by - // these two methods. + // these two scan* methods. I.e. from this point onward, for all live + // InputSections, we should have `isec->canonical() == isec`. scanSymbols(); scanRelocations(); @@ -1119,6 +1121,8 @@ if (in.stubHelper->isNeeded()) in.stubHelper->setup(); + // At this point, we should know exactly which output sections are needed, + // courtesy of scanSymbols() and scanRelocations(). createOutputSections(); // After this point, we create no new segments; HOWEVER, we might @@ -1140,11 +1144,10 @@ void macho::createSyntheticSections() { in.header = make(); - if (config->dedupLiterals) { + if (config->dedupLiterals) in.cStringSection = make(); - } else { + else in.cStringSection = make(); - } in.wordLiteralSection = config->dedupLiterals ? make() : nullptr; in.rebase = make();