Hello.

I'm encountering an issue with the PDBs generated by NativeAOT: while they correctly list all fields of user-defined classes, they often report an incorrect total type size. This inconsistency makes it difficult to analyze memory layout using reverse engineering or debugging tools such as IDA Pro, since such tools rely on accurate type sizes to determine field boundaries and class structure.

Repro scenario
In my project, I have a class like this:

public class WindowsClipboardMonitor
{
    private readonly PCWSTR _className;
    private readonly HWND _hwnd;
    private readonly Thread _messageLoopThread;
    private readonly uint _messageLoopThreadId;
    private readonly bool _isDisposed;
    private readonly CancellationToken _token;
    private readonly Func<string, Task> ClipboardUpdate;
}

After compiling the app with NativeAOT and inspecting the resulting PDB using IDA Pro, I observe the following:

Type: ClipboardTranslator_Core_ClipboardTranslator_Core_ClipboardHandler_WindowsClipboardMonitor | Index: 19481 | Size: 8
Field: Object
Field: +0020: ClipboardTranslator_Core_Windows_Win32_Foundation_PCWSTR _className
Field: +0028: ClipboardTranslator_Core_Windows_Win32_Foundation_HWND _hwnd
Field: +0008: (S_P_CoreLib_System_Threading_Thread)* _messageLoopThread
Field: +0018: UInt32 _messageLoopThreadId
Field: +001C: Boolean8 _isDisposed
Field: +0030: S_P_CoreLib_System_Threading_CancellationToken _token
Field: +0010: (S_P_CoreLib_System_Func_2<String__S_P_CoreLib_System_Threading_Tasks_Task>)* ClipboardUpdate

From this layout, it's clear the actual type size should be at least 0x38 (56 bytes), not 8. But NativeAOT reports the size as 8.

To confirm that the issue is only with the reported size, I manually patched the corresponding byte in the .pdb file (or in-memory via IDA) to change the reported type size from 0x08 to 0x38.

After doing this, IDA Pro immediately recognizes the full layout, and all fields are displayed correctly, including those beyond the 8-byte boundary.

This confirms that:

• The field list and their offsets are correct in the PDB.
• The only incorrect part is the type's reported size.
• Tools depending on type size use that value to stop parsing fields.

This issue does not affect value types (structs). All struct sizes in the same PDB are reported correctly, and their fields are fully visible in IDA without patching.