3419f0a
@@ -68,7 +68,7 @@
%ifarch s390x
%global testargs --exclude-regex KokkosCore_UnitTest_StackTraceTest
%endif
- %ctest %{?testargs}
+ %ctest %{?testargs} --timeout 6000
%files
%doc README.md
Signed-off-by: David Abdurachmanov davidlt@rivosinc.com
(Adding @davidlt @tekkamanninja )
Note an alternative proposal is to modify the %ctest macro to add this parameter implicitly. See: https://src.fedoraproject.org/rpms/snapshot/pull-request/3#comment-183686
%ctest
Looks good to me. I will hold off with merging to see the follow-up on %ctest modification proposal.
The more I think about this. I would really like to see meson_test and ctest macros have a global timeout multiplier for riscv64. It will be way easier to remove it later on.
Note this should be ifarch riscv64 too. Default is 1500 IIRC. A timeout on any other arch most likely would be a regression. We only need more time on riscv64.
Yes I'll look at that next, once I've finished on mold.
mold
Interestingly when I ran the test suite on x86-64, tests also timed out ...
The following tests FAILED: 4 - Kokkos_CoreUnitTest_OpenMP (Timeout) 25 - Kokkos_ContainersUnitTest_Serial (Timeout) 26 - Kokkos_ContainersUnitTest_OpenMP (Timeout) 32 - Kokkos_AlgorithmsUnitTest_StdSet_C (Timeout) 33 - Kokkos_AlgorithmsUnitTest_StdSet_D (Timeout) 34 - Kokkos_AlgorithmsUnitTest_StdSet_E (Timeout) 35 - Kokkos_AlgorithmsUnitTest_StdSet_Team_A (Timeout) 36 - Kokkos_AlgorithmsUnitTest_StdSet_Team_B (Timeout) 37 - Kokkos_AlgorithmsUnitTest_StdSet_Team_C (Timeout) 40 - Kokkos_AlgorithmsUnitTest_StdSet_Team_F (Timeout) 41 - Kokkos_AlgorithmsUnitTest_StdSet_Team_G (Timeout) 42 - Kokkos_AlgorithmsUnitTest_StdSet_Team_H (Timeout) 43 - Kokkos_AlgorithmsUnitTest_StdSet_Team_I (Timeout) 44 - Kokkos_AlgorithmsUnitTest_StdSet_Team_L (Timeout) Errors while running CTest
This is on a 16 core AMD server so I would hope it can keep up.
Anyway I'm still planning a patch to cmake macros to modify %ctest on riscv64.
Alternate proposal: https://src.fedoraproject.org/rpms/cmake/pull-request/37
I had something similar on EPEL9, but there the toolchain (specifically OpenMP) was just slow.
Pull-Request has been merged by junghans
Signed-off-by: David Abdurachmanov davidlt@rivosinc.com