[WebAssembly] Provide WasmFunction content offset informatin.

[yurydelendik]

Summary

WasmObjectWriter mostly operates with function segments offsets that do not
include their size fields. WasmObjectFile needs to have and provide this
information to the lld to maintain proper R_WEBASSEMBLY_FUNCTION_OFFSET_I32
relocations entries.

[sbc100]

lib/Object/WasmObjectFile.cpp

878 ↗

(On Diff #146370)

As an alternative, perhaps we could just move the const uint8_t *FunctionStart = Ptr; down a couple of lines so that the function body no longer includes its size?

The downside would be that the linker would need to re-encode the size for each function.

@aardappel was recently looking making objdump -d work and he ran into issues relating to pointer the actual start of the instructions.

Having said all that, this seems like a perfectly reasonable interim fix.

[aardappel]

lib/Object/WasmObjectFile.cpp

878 ↗

(On Diff #146370)

I could already fix objdump to give correct output for wasm disassembly, but that would require a some wasm-specific parsing in objdump. Question is, can we do this more generically such that objdump does things correctly automatically?
Some wasm specific code may be unavoidable, if we'd also like to interleave a disassembly with local/arg annotations etc.

[yurydelendik]

lib/Object/WasmObjectFile.cpp

878 ↗

(On Diff #146370)

It looks like it is possible to exclude function size from data and use CodeSectionOffset as a start of the function. (LLD's CodeSection needs to be changed to generate function size/header) Would you like to do that instead?

[sbc100]

Should I land this now?

[yurydelendik]

Yes, please.

llvm/trunk/include/llvm/BinaryFormat/Wasm.h

 };
 
 struct WasmFunction {
   uint32_t Index;
   std::vector<WasmLocalDecl> Locals;
   ArrayRef<uint8_t> Body;
   uint32_t CodeSectionOffset;
   uint32_t Size;
+  uint32_t CodeOffset;  // start of Locals and Body
   StringRef SymbolName; // from the "linking" section
   StringRef DebugName; // from the "name" section
   uint32_t Comdat; // from the "comdat info" section
 };
 
 struct WasmDataSegment {
   uint32_t MemoryIndex;
   WasmInitExpr Offset;

llvm/trunk/lib/Object/WasmObjectFile.cpp

Error WasmObjectFile::parseCodeSection(const uint8_t *Ptr, const uint8_t *End) {
   }
 
   while (FunctionCount--) {
     wasm::WasmFunction Function;
     const uint8_t *FunctionStart = Ptr;
     uint32_t Size = readVaruint32(Ptr);
     const uint8_t *FunctionEnd = Ptr + Size;
 
+    Function.CodeOffset = Ptr - FunctionStart;
     Function.Index = NumImportedFunctions + Functions.size();
     Function.CodeSectionOffset = FunctionStart - CodeSectionStart;
     Function.Size = FunctionEnd - FunctionStart;
 
     uint32_t NumLocalDecls = readVaruint32(Ptr);
     Function.Locals.reserve(NumLocalDecls);
     while (NumLocalDecls--) {
       wasm::WasmLocalDecl Decl;