Opened 4 years ago
Closed 4 years ago
#22100 closed defect (fixed)
Openblas testsuite fails on OSX
Reported by: | vbraun | Owned by: | |
---|---|---|---|
Priority: | major | Milestone: | sage-7.5 |
Component: | linear algebra | Keywords: | |
Cc: | fbissey | Merged in: | |
Authors: | François Bissey | Reviewers: | Volker Braun |
Report Upstream: | N/A | Work issues: | |
Branch: | ea0c0f7 (Commits, GitHub, GitLab) | Commit: | ea0c0f7d1687b1fa8ede2df6407d385b6bdd6628 |
Dependencies: | #22021 | Stopgaps: |
Description
Fails with
gfortran -m128bit-long-double -Wall -m64 -L/Users/buildslave-sage/slave/sage_git/build/local/lib -Wl,-rpath,/Users/buildslave-sage/slave/sage_git/build/local/lib -o dblat2 dblat2.o ../libopenblas_sandybridgep-r0.2.19.a -lpthread -lgfortran -lpthread -lgfortran ld: file too small (length=0) file '../libopenblas_sandybridgep-r0.2.19.a' for architecture x86_64 collect2: error: ld returned 1 exit status gfortran -m128bit-long-double -Wall -m64 -L/Users/buildslave-sage/slave/sage_git/build/local/lib -Wl,-rpath,/Users/buildslave-sage/slave/sage_git/build/local/lib -o cblat2 cblat2.o ../libopenblas_sandybridgep-r0.2.19.a -lpthread -lgfortran -lpthread -lgfortran make[4]: *** [sblat1] Error 1
Change History (9)
comment:1 Changed 4 years ago by
comment:2 Changed 4 years ago by
OK, so spkg-install
ran ok but no spkg-check
is failing. For exactly the same reason and I should say at exactly the same point. This is because we already run the test suite of openblas in spkg-install
as it seems to be part of the default target. And we have a failure in spkg-check
because make
was not run with the same arguments as in spkg-install
- notably USE_THREAD=0
. There are three things we could do at this stage:
- fix
spkg-install
so the test suite is not run and make surespkg-check
is run with same options. Because there is no configuration as such the settings are not preserved from one run to another. - remove
spkg-check
and make a note that the test suite is run duringspkg-install
. By far the easiest solution. - status quo, fix
spkg-check
to use the same options asspkg-install
and keep running the test suite in bothspkg-install
andspkg-check
.
Other notes: this output means that the last patch I didn't forward upstream is what is needed to fix the build in #22021. A problem with ifdef
, I cannot really forward that patch upstream without knowing precisely what shell was making it fail. I also would have to make the fix comprehensive over other makefile and possibly other use of ifdef
.
comment:3 follow-up: ↓ 4 Changed 4 years ago by
FWIW I'm using the default shell on OSX, i.e. an ancient bash (version 3.2.57)
comment:4 in reply to: ↑ 3 Changed 4 years ago by
Replying to vbraun:
FWIW I'm using the default shell on OSX, i.e. an ancient bash (version 3.2.57)
OK when I make a PR upstream, I can say something about the environment. In the mean time, out of the three options, what do you favor?
comment:5 follow-up: ↓ 6 Changed 4 years ago by
Is there an easy way to not run the testsuite in spkg-install? then that would be nice to minimise install time. If not then I'd just remove spkg-check since its redundant...
comment:6 in reply to: ↑ 5 Changed 4 years ago by
Replying to vbraun:
Is there an easy way to not run the testsuite in spkg-install? then that would be nice to minimise install time. If not then I'd just remove spkg-check since its redundant...
The main makefile has
all :: libs netlib tests shared
so I am guessing we could call make on libs netlib shared
. I'll put something together to try.
comment:7 Changed 4 years ago by
- Branch set to u/fbissey/openblas_tests
- Commit set to ea0c0f7d1687b1fa8ede2df6407d385b6bdd6628
- Dependencies set to #22021
- Status changed from new to needs_review
comment:8 Changed 4 years ago by
- Reviewers set to Volker Braun
- Status changed from needs_review to positive_review
Thanks, lgtm!
comment:9 Changed 4 years ago by
- Branch changed from u/fbissey/openblas_tests to ea0c0f7d1687b1fa8ede2df6407d385b6bdd6628
- Resolution set to fixed
- Status changed from positive_review to closed
More log