[SYCL][E2E] Disable fill_any_size.cpp on FPGA #16590
Conversation
Signed-off-by: Sarnie, Nick <[email protected]>
|
Any idea why/when it started to fail? With these generic issue info I'm afraid we might be losing crucial information when regressions happen. |
|
The first commit where I see it failing (it's sporadaic btw) is ac9e5d9, but since it's sporadic I have no idea if that actually caused it. My stance tends to be as CI maintainers we should unbreak the CI ASAP and let code/test owners do the actual technical investigation. Of course if we have an idea of what caused it or some relevant info we should add it, but I would rather we not do any advanced analysis, here I only notice the above pattern with that commit. I'll add any patterns I see in future changes like this, thanks. |
Unfortunately, that only works in theory... |
sycl/test-e2e/USM/fill_any_size.cpp
Outdated
| // UNSUPPORTED: accelerator | ||
| // UNSUPPORTED-TRACKER: https://github.com/intel/llvm/issues/16589 |
There was a problem hiding this comment.
I think this is just the existing oneapi-src/unified-runtime#2440 above. fpga/cpu share most of the implementation, and the flakyness matches the behavior described there.
There was a problem hiding this comment.
ok, ill close my issue and use this tracker, thanks
Signed-off-by: Sarnie, Nick <[email protected]>
I believe it started failing two days ago when this test was re-enabled in #16544. |
|
Nice find! |
It's failing in postcommit.