-
Notifications
You must be signed in to change notification settings - Fork 4.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Baseline two failing tests to make outerloop green again #61280
Conversation
Tagging subscribers to this area: @hoyosjs Issue Details/cc @dotnet/runtime-infrastructure
|
Are you still seeing |
@AndyAyersMS - well, I saw the failure in my two outerloop runs from last night. If it was actually fixed last night we don't really need the GenericsTest exclusion anymore; I didn't notice the bug was closed. In light of your clarification I'll remove the exclusion from issues.targets and I'll rerun the job to double-check. |
…per Andy Ayers' PR feedback
c1fa023
to
6c28714
Compare
@AndyAyersMS - interesting, so the test now only fails on arm64 in Crossgen2 mode complaining about mismatch between the compiler-calculated and runtime type layout for Vector: Testing Vector<bool> Type System.Numerics.Vector`1: expected HFA type 00000004, actual 00000000 I can confirm I no longer see the bug on Linux x64. The new failure is common to Windows and Linux. |
The new failure has to be something different (though perhaps similar). Maybe we're letting some detail of the build machine bleed through? |
Well, I guess it would be easiest to start from the semantic side - what is the 'expected' HFA value for Vector<bool> on arm64? Is it 4 (as Crossgen2 expects) or 0 (as runtime sets on the type)? |
@trylek Is this exclusion still needed? I'm confused on the state of things given the conversation. |
Well, for the GenericTest, that still fails but on other platforms and with a different bug. The test45929 continues failing on arm, albeit non-deterministically (sometimes it passes). I'll update the PR with the new exclusion for GenericTest based on outcomes of the above outerloop test. |
OK, my bad, I messed up the 2nd iteration - when I sent out the first iteration and Andy Ayers pointed out that the GenericTest should be fixed now, I inadvertently deleted a different exclusion for the GenericTest from the issues.targets file because I didn't notice there were two different ones. That is fixed now and that should make the remaining outerloop failures disappear. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So you're just excluding the one test on windows arm32...?
Yes, exactly, as you apparently fixed the GenericTest its exclusion is no longer necessary. The pre-existing exclusion is still in place though, the issue #60036 remains in place. It's currently assigned to me but I must admit I'm not an arm64 expert so I need help in figuring out what exactly needs fixing there - i.o.w., which HFA value is the "right one", that should imply whether we need to fix the runtime or Crossgen2 to put them in sync. |
Formatting was fixed by Bruce in #61334. Not even sure why it is running here. |
OK, so we're back to two test exclusions now ;-). |
/cc @dotnet/runtime-infrastructure