Baseline more pri1 tests with native AOT#111068
Conversation
I originally thought this found a real bug but this is just a really weird test exercising not just slow delegate creation path, but also quite invalid IL: https://github.com/dotnet/runtime/blob/3d88f276b13ae4d2ffa6a66b02f32171fdc825b5/src/tests/JIT/Methodical/delegate/_simpleoddpower.il#L510-L517 (Notice that despite the newobj, the actual object instance is of a completely different type and the method is not on that type.)
|
Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas |
Should we delete this test instead? |
@dotnet/jit-contrib for opinion on the deletion since it's a JIT test. |
@dotnet/jit-contrib any opinion about deleting the test? If no opinion, I would prefer committing as-is (only disable on native AOT). |
|
@dotnet/ilc-contrib could someone have a look at this test disablement? |
Sorry for the lack of response on our end -- I don't think we have an opinion on this. |
Thanks, assumed that. It is invalid IL, but there's probably some reason for it. I don't mind if native AOT doesn't have matching behavior for invalid IL so I'm fine just disabling it, just need someone to agree and sign off. |
|
Thank you! |
* main: JIT: fix case where implied subrange assertions can get lost in morph (dotnet#112020) Propose new async API (dotnet#110420) Move the diagnostic print for stack overflow test failure (dotnet#112001) JIT: Fold more nullchecks (dotnet#111985) Baseline more pri1 tests (dotnet#111068)
3 participants
I originally thought this found a real bug but this is just a really weird test exercising not just slow delegate creation path, but also quite invalid IL:
runtime/src/tests/JIT/Methodical/delegate/_simpleoddpower.il
Lines 510 to 517 in 3d88f27
(Notice that despite the newobj, the actual object instance is of a completely different type and the method is not on that type, took a while to realize the newobj is useless and unrelated.)
Cc @dotnet/ilc-contrib