Closed balay closed 1 year ago
I don't understand what is going on here and cannot reproduce.
I have added ExaGO GitHub source code references in exago@develop
that correspond to the build failures. These lines call .c_str()
where I linked, and these are changes added 5 months ago, so not quite sure what is going wrong here.
Is the right ExaGO source code being used?
You can try out a build with: https://gitlab.com/xsdk-project/spack-xsdk/-/tree/xsdk-1.0.0
Where its using:
+ version("1.6.0", branch="main", submodules=True)
[with dependencies on newer versions of hiop, petsc added]
So main
is used here. Is develop
the branch to use?
Is there a way to add the concretization/spec being installed of ExaGO to the build logs
Attaching spck spec
this build.
Is the right ExaGO source code being used?
You can try out a build with: https://gitlab.com/xsdk-project/spack-xsdk/-/tree/xsdk-1.0.0
Where its using:
+ version("1.6.0", branch="main", submodules=True)
[with dependencies on newer versions of hiop, petsc added]
So
main
is used here. Isdevelop
the branch to use?Is there a way to add the concretization/spec being installed of ExaGO to the build logs
Attaching
spck spec
this build.
Main has this bug, but develop does not. Please use develop, and we will try and get a release once we have fully moved over to GitHub.
Switching to 'develop' branch - I still get a failure..
https://gitlab.com/xsdk-project/spack-xsdk/-/jobs/4987095719
Okay apologies for taking so long to get to this. We now have better compiler flags in our internal gcc pipelines that catch this type of issue and so hopefully this isn't as much of a problem moving forward.
Try again with latest exago@develop in GitHub and you should be good to go.
To update - exago builds continue to fail.
BTW: Forgot to mention the above 2 failures are with gcc. Perhaps they worked before?
The intel builds also fail - for ex: https://gitlab.com/xsdk-project/spack-xsdk/-/jobs/5089552335
@balay we now have GitHub action pipelines that catch these build failures before xSDK pipelines catch them (hopefully).
exago@develop
now has a fix that should let the pipelines go through, and with spack PR https://github.com/spack/spack/pull/40188, you can also build exago~tests
in situations like this to avoid the problem altogether.
Closing as latest pipelines are going through just fine. Need to create a new issue for gcc@11+ compilation issues
ref: https://gitlab.com/xsdk-project/spack-xsdk/-/jobs/4969753954
spack-build-out.txt