[SYCL] Fix postcommit

[sarnex]

Totally broken on all platforms right now.

/__w/llvm/llvm/src/sycl/source/detail/physical_mem_impl.hpp:33:3: error: default label in switch which covers all enumeration values [-Werror,-Wcovered-switch-default]
   33 |   default:

and

#14372

[sarnex]

Merging now to unbreak CI.

@steffenlarsen FYI

[bader]

@sarnex, pre-commit has failed. I would wait pre-commit results before merging.

[sarnex]

 /__w/llvm/llvm/src/sycl/source/detail/physical_mem_impl.hpp:34:1: error: control reaches end of non-void function [-Werror=return-type]

sigh

[aelovikov-intel]

@sarnex, pre-commit has failed. I would wait pre-commit results before merging.

That's not feasible because pre-commit runs take many hours to finish...

[sarnex]

Yeah my bad on this one. I'll fix it and try to run precommit when I can but for high priority issues like this there is always some risk/reward tradeoff.

[bader]

Yeah my bad on this one. I'll fix it and try to run precommit when I can but for high priority issues like this there is always some risk/reward traceoff.

I don't think "high priority issue" justifies the risk. We have fixes for high priority issues every day.

Do you mean: we need to merge some change urgently? Could you clarify what the reason for the urgency with this PR? Failing tests in post-commit does not sound to me as a good reason to take the risk to break pre-commit testing.

The much less risky change would be reverting the commit breaking post-commit validation. But even revert commit must pass pre-commit testing.

[sarnex]

Could you clarify what the reason for the urgency with this PR?

Postcommit build on multiple platforms was failing, and a developer reached out to me internally telling me about the problem and that they would like this fixed asap. In general if I wasn't confident on the fix and it being trivial I would just revert, but my confidence was incorrect in this case, and that was my bad.