[codeview] Fix handling of S_HEAPALLOCSITE

The type index is from the TPI stream, not the IPI stream. Fix the
dumper, fix type index discovery, and add a test in LLD.

Also improve the log message we emit when we fail to rewrite type
indices in LLD. That's how I found this bug.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@316461 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Reid Kleckner 2017-10-24 17:02:40 +00:00
parent 70d05d486e
commit 91db8008a9
2 changed files with 2 additions and 3 deletions

View File

@ -413,8 +413,7 @@ static bool discoverTypeIndices(ArrayRef<uint8_t> Content, SymbolKind Kind,
Refs.push_back({TiRefKind::IndexRef, 8, 1}); // ID of inlinee
break;
case SymbolKind::S_HEAPALLOCSITE:
// FIXME: It's not clear if this is a type or item reference.
Refs.push_back({TiRefKind::IndexRef, 8, 1}); // signature
Refs.push_back({TiRefKind::TypeRef, 8, 1}); // UDT allocated
break;
// Defranges don't have types, just registers and code offsets.

View File

@ -615,7 +615,7 @@ Error MinimalSymbolDumper::visitKnownRecord(CVSymbol &CVR, FrameProcSym &FP) {
Error MinimalSymbolDumper::visitKnownRecord(CVSymbol &CVR,
HeapAllocationSiteSym &HAS) {
AutoIndent Indent(P, 7);
P.formatLine("type = {0}, addr = {1} call size = {2}", idIndex(HAS.Type),
P.formatLine("type = {0}, addr = {1} call size = {2}", typeIndex(HAS.Type),
formatSegmentOffset(HAS.Segment, HAS.CodeOffset),
HAS.CallInstructionSize);
return Error::success();