Summary: | sci-libs/openblas-0.3.23[-test] runs ./openblas_utest during src_compile() | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | cyrillic <cyrilmaley> |
Component: | Current packages | Assignee: | Gentoo Science Related Packages <sci> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | lumin |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
URL: | https://github.com/xianyi/OpenBLAS/issues/4017 | ||
See Also: | https://bugs.gentoo.org/show_bug.cgi?id=910174 | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: | build.log.xz |
Description
cyrillic
2023-07-16 01:56:01 UTC
Created attachment 865564 [details]
build.log.xz
The problem goes away if I revert that last patch from bug 910174 If I compile with gcc-13 instead of clang-16 it fails the fork safety test instead of hanging : ... TEST 35/37 potrf:bug_695 [OK] TEST 36/37 kernel_regress:skx_avx [OK] TEST 37/37 fork:safety_after_fork_in_parent [FAIL] ERR: Failed to fork process. RESULTS: 37 tests (36 ok, 1 failed, 0 skipped) ran in 59 ms make[1]: *** [Makefile:52: run_test] Error 1 make[1]: Leaving directory '/tmp/portage/sci-libs/openblas-0.3.23/work/OpenBLAS-0.3.23/utest' make: *** [Makefile:157: tests] Error 2 * ERROR: sci-libs/openblas-0.3.23::gentoo failed (compile phase): * emake failed The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=c730b8fb4941b27b62b776f61ced930f2fbbd64b commit c730b8fb4941b27b62b776f61ced930f2fbbd64b Author: Michael Orlitzky <mjo@gentoo.org> AuthorDate: 2023-07-17 02:56:35 +0000 Commit: Michael Orlitzky <mjo@gentoo.org> CommitDate: 2023-07-17 02:56:35 +0000 sci-libs/openblas: don't build tests in src_compile. Bug: https://bugs.gentoo.org/910415 Signed-off-by: Michael Orlitzky <mjo@gentoo.org> sci-libs/openblas/openblas-0.3.23.ebuild | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) I'm not sure how the patch caused that, because "tests" was *always* a prerequisite of "all" (the default Makefile target). Anyway, now it's running "emake shared" explicitly to avoid the "all" target, so hopefully that part is fixed. That still leaves the fork safety test to debug... There's an OpenBLAS issue for the hang ($url), and now an LLVM issue for the failed assertion too: https://github.com/llvm/llvm-project/issues/63908 I would think that building the tests in src_compile() would be okay. It is just running the tests that should hold off until src_test(). In any case, openblas::gentoo seems to be working for me once again. |