Test failure: Loader/classloader/explicitlayout/Regressions/13362/Github13362/Github13362.cmd (Verify_FieldOffset) #102868
Comments
v-wenyuxu
added
os-mac-os-x
dotnet-policy-service
bot
added
the
untriaged
|
@MichalStrehovsky This looks like a regression from #102810 |
If it's from that, it would be an existing bug exposed by precompiling more stuff. I can "fix" it by adding back the blocking for ReadyToRun only, or I can leave this for crossgen team to fix properly. Do you have an opinion/preference? |
|
Failed in: runtime-coreclr r2r 20240529.1 Failed tests: Error message: Stack trace:
|
|
Failed in: runtime-coreclr r2r-extra 20240601.1 Failed tests: Error message: Stack trace:
More failures
|
MichalStrehovsky
added a commit
to MichalStrehovsky/runtime
that referenced
this issue
Jun 3, 2024
The relaxation of type layout check in dotnet#102810 exposed a bug in the type layout algorithm that doesn't correctly handle doing type layout in this situation (it doesn't match what CoreCLR VM computes). This restores the old behavior for ReadyToRun that just throws on this. The compiler will then stop precompiling the method and fall back to runtime JIT. Works around dotnet#102868 (the bug still exists, this is a valid situation that we should be able to precompile, we just won't).
|
Failed in: runtime-coreclr r2r-extra 20240602.1 Failed tests: Error message: Stack trace:
More failures
|
MichalStrehovsky
added a commit
that referenced
this issue
Jun 4, 2024
The relaxation of type layout check in #102810 exposed a bug in the type layout algorithm that doesn't correctly handle doing type layout in this situation (it doesn't match what CoreCLR VM computes). This restores the old behavior for ReadyToRun that just throws on this. The compiler will then stop precompiling the method and fall back to runtime JIT. Works around #102868 (the bug still exists, this is a valid situation that we should be able to precompile, we just won't).
MichalStrehovsky
removed
the
blocking-outerloop
|
Workaround was added in #102982. Dropping the blocking-outerloop label. We might still want to remove the workaround and fix the type layout mismatch (or at least understand the root cause to make sure this is the only thing that can surface the discrepancy). |
|
Failed in: runtime-coreclr r2r 20240603.1 Failed tests: Error message: Stack trace:
More failures
|
This is different test with different failure mode. This issue is specifically about Could you please open a new issue on the GitHub_65690 failure? |
jkotas
changed the title
|
This hasn't repro'd recently, and I wasn't able to reproduce it either locally. The most recent pipelines have occasional failures in the leg described by this issue, but they are failures different from this one. |
See #102868 (comment) for reason. |
I saw that but it was reported as failed again after you posted that workaround. There is a comment with more failures after that. |
It takes a while for a change to propagate through the system. The failure reported was likely from a test run that started before the workaround was committed. You should see the failure if you revert #102982 and trigger crossgen2 run. |
Oh so a race condition between test run and fix application, got it. Will take a look without the workaround in the future then. |
jeffschwMSFT
removed
the
untriaged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Failed in: runtime-coreclr outerloop 20240529.7
Failed tests:
Error message:
Stack trace:
The text was updated successfully, but these errors were encountered: