[Lldb-commits] [PATCH] D121967: [LLDB][NativePDB] Create inline function decls

[Zequan Wu via Phabricator via lldb-commits]

zequanwu added inline comments.


================
Comment at: lldb/source/Plugins/SymbolFile/NativePDB/PdbAstBuilder.cpp:1155
+          TypeDeserializer::deserializeAs<StringIdRecord>(parent_cvt, sir));
+      parent = GetOrCreateNamespaceDecl(sir.String.data(), *parent);
+    }
----------------
rnk wrote:
> Can these be nested? You may need to split these on `::` and create  or look up nested namespace decls.
Yes, there can be nested and then the `StringIddRecord` would contains all of that. 
For example, function `foo`'s `ParentScope` is the following.
` 0x1070 | LF_STRING_ID [size = 16] ID: <no type>, String: A::B`


================
Comment at: lldb/source/Plugins/SymbolFile/NativePDB/PdbAstBuilder.cpp:1158
+
+    CVType func_type_cvt = m_index.tpi().getType(func_ti);
+    if (func_type_cvt.kind() == LF_PROCEDURE) {
----------------
rnk wrote:
> This only requires `func_ti` as an input. Does it need to be part of the case? Should we compute param_count for methods too?
Removed. param_count is computed outside the switch but it's not useful for method decls. Method decls are created via `TypeSystemClang::AddMethodToCXXRecordType` which will add parameter decls.


================
Comment at: lldb/source/Plugins/SymbolFile/NativePDB/PdbAstBuilder.cpp:1163
+          TypeDeserializer::deserializeAs<ProcedureRecord>(func_type_cvt, pr));
+      param_count = pr.getParameterCount();
+    }
----------------
rnk wrote:
> This is overwritten later, so I think you can remove this if block.
Yeah, I forgot to remove this if block. It's not necessary anymore.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D121967/new/

https://reviews.llvm.org/D121967