[SYCL] Fix segfault on program exit when user thread is not finished yet #7908
Conversation
…used by intel#6837) Signed-off-by: Tikhomirova, Kseniya <[email protected]>
Signed-off-by: Tikhomirova, Kseniya <[email protected]>
Signed-off-by: Tikhomirova, Kseniya <[email protected]>
Signed-off-by: Tikhomirova, Kseniya <[email protected]>
Signed-off-by: Tikhomirova, Kseniya <[email protected]>
|
/testwin |
Signed-off-by: Tikhomirova, Kseniya <[email protected]>
Signed-off-by: Tikhomirova, Kseniya <[email protected]>
Signed-off-by: Tikhomirova, Kseniya <[email protected]>
Signed-off-by: Tikhomirova, Kseniya <[email protected]>
|
/summary:run |
Signed-off-by: Tikhomirova, Kseniya <[email protected]>
bader
changed the title
|
@KseniyaTikhomirova @romanovvlad This PR is leading to a bunch of TimeOuts on llvm-test-suite when we test Windows. Pretty much every PR since it's been merged ( for example: intel/llvm-test-suite#1500 ). There are a couple of tests that consistently lock up ( HostInteropTask/host-task-failure.cpp ) and others that are inconsistent. Meaning, I can get between 7 to 12 of them to lock up if I run the test suite locally, but many of those same tests pass if run individually. The "/testwin" directive only makes sure something is buildable on windows and the in-tree tests are run, but not the llvm-test-suite tests on Windows. Likewise, my last understanding is that "summary:run" doesn't do anything presently either. |
|
@cperkinsintel, I unintentionally moved thread pool drain out of win32 check. FIxed here #7973 |
PR #7908 accidentally introduced a behavior that causes lots of timeouts on the CI system on Windows. Signed-off-by: Tikhomirova, Kseniya <[email protected]>
(caused by #6837)
Signed-off-by: Tikhomirova, Kseniya [email protected]