[SYCL] Enable undefined-symbol test for the opencl and gpu #944
Conversation
| @@ -1,4 +1,4 @@ | |||
| // XFAIL: cuda || hip || (opencl && gpu) | |||
| // XFAIL: cuda || hip | |||
There was a problem hiding this comment.
Could you clarify why this test still fails in Jenkins environment?
There was a problem hiding this comment.
It seems to me, that test become pass only on OCL GEN9 Linux test suite. If it is related to the intel/llvm#5867, then test will fail on windows, drivers were updated only for linux.
May be, we can specialize test suites on which this test should pass? I don't know how to do it.
There was a problem hiding this comment.
The test fails on Linux as well. I suggest we investigate the difference between test behavior in Jenkins and GitHub Actions environments.
There was a problem hiding this comment.
Yes, but I don't understand which test suite is used on jenkins for linux.
Yes, we should investigate. I'll try to figure out.
There was a problem hiding this comment.
If I understand correctly, drivers in Github Actions docker container for OCL GEN9 more recent, than in the jenkins.
OCL GEN9 Github Actions have version 22.11.22682, that matches version from the uplift PR for 'linux_staging'. Can be seen in raw logs from here
Jenkins have version 22.10.22597.
Anyway, we can't enable test only for specific driver version. Should we disable it for now, while it fails on some driver versions?
There was a problem hiding this comment.
You seem to be right. I think we must align driver version for both systems before merging any patches.
@alexbatashev, could you clarify why GHA is using more recent version than buildbot for pre-commit? I assumed it should be used only for nightly jobs.
There is a PR uplifting Jenkins version to 22.11.*, but I see new failures in pre-commit for this change. intel/llvm#5897
Probably the safer way is to rollback the driver version for GHA and commit intel/llvm#5897 together with this patch once all pre-commit issues are addressed.
There was a problem hiding this comment.
I reverted driver version in GHA precommit to 22.10.*.
This patch should go together with intel/llvm#5897.
bader
dismissed
their stale review
My concern is addressed by intel/llvm@fa054cc.
Test become unexpectedly pass, possibly, after the driver update in the intel/llvm#5867