Closed agocke closed 2 years ago
Do we need to revert the interface change? I understand that it's weird and may cause problems. On the other hand, the same limitation exists in Roslyn, and so it's unlikely that the trimmer will encounter such pattern. That said - it would be interesting to be able to optionally disable validation as a perf improvement.
Nope, we don't have to take this, I thought I would propose it regardless.
FYI, Roslyn disables validation due to this behavior.
I'll probably just sit down, fix this in S.R.Metadata, and we can update to the new package once it's published.
We can keep this pull request open as a reminder.
This did catch the bug around sorting GenericParameters table that was real. Do you think we got all the value out of this that we could?
Cc @vitek-karas