When compiling NativeAOT, is the deferred information stored in a PDB file with field pointers like in C++?

[dadavadd]

NativeAOT PDB file loaded into IDA:

C++ pdb loaded into IDA:

Or does information about the fields disappear during compilation, even from a PDB file?

[ghost]

Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas See info in area-owners.md if you want to be subscribed.

Issue Details

---

NativeAOT PDB file loaded into IDA:  C++ pdb loaded into IDA:  Or does the information about the fields disappear during compilation?

Author:	dadavadd
Assignees:	-
Labels:	`untriaged`, `area-NativeAOT-coreclr`
Milestone:	-

[SingleAccretion]

Yes, the AOT compiler does produce debugging information for fields, among other things. It can also produce information needed by reflection for fields that may be reflected on (this is usually a rather small subset of all fields) in a custom format that is stored directly in the binary.

[dadavadd]

Yes, the AOT compiler does produce debugging information for fields, among other things. It can also produce information needed by reflection for fields that may be reflected on (this is usually a rather small subset of all fields) in a custom format that is stored directly in the binary.

Is there any way to view them in IDA PRO while loading debugging information (.pdb)?

[SingleAccretion]

I do not know (this would be an IDA question). I would presume that the "shape" of debug information produced is sufficiently different from C++'s that it is not picked up by the decompiler by default.

[dadavadd]

Indeed, this is noticeable with strings. They are marked in ida pro as "unk_1234567".

I do not know (this would be an IDA question). I would presume that the "shape" of debug information produced is sufficiently different from C++'s that it is not picked up by the decompiler by default.

Indeed, this is noticeable with strings. They are marked in ida pro as "unk_1234567".

[MichalStrehovsky]

Indeed, this is noticeable with strings. They are marked in ida pro as "unk_1234567".

The string in Native AOT look nothing like typical strings in native languages - first is a pointer to the vtable, next is the size of the string, followed by the characters. IDA would need to learn to understand these.

[Thealexbarney]

I remember back in the CoreRT days IDA was a lot better at loading debug info. It got worse somewhere along the way, although I seem to recall IDA loading a lot more stuff from DWARF info compared to PDBs back then.

[MichalStrehovsky]

I think the question has been answered and I don't see anything else actionable.