On Mon, Oct 14, 2024 at 06:26:23PM +0100, Richard W.M. Jones wrote:
On Mon, Oct 14, 2024 at 06:21:32PM +0100, Richard W.M. Jones wrote:
> On Mon, Oct 14, 2024 at 09:48:51AM +0100, Daniel P. Berrangé wrote:
> > On Sat, Oct 12, 2024 at 11:57:57AM +0100, Richard W.M. Jones wrote:
> > > [I couldn't find out where to send questions / bugs related
> > > to libvirt CI, so here I am ...]
> > >
> > >
https://gitlab.com/nbdkit/libnbd/-/pipelines/1493147187
> > >
> > > It would be much easier to fix the above pipeline (which doesn't
> > > reproduce locally) if I could download artifacts for the failing
> > > builds. However we don't seem to have them for macOS and FreeBSD.
> > > They work fine for Linux. Why, and can they be generated somehow?
> >
> > This is related to the way those platforms are built. The job on GitLab
> > CI is just a facade, which launches the real job on Cirrus CI. All that
> > the GitLab CI job gets back is the build log output.
> >
> > On the plus side, Cirrus CI has an *awesome* feature where you can login
> > to the Cirrus CI interface directly, and select "re-run with shell
access"
> > option against the failed job.
> >
> > This will spawn the job again and give you an interactive shell session
> > in the browser, which makes it fairly easy to debug wierd problems.
>
> I wasn't able to get this to work (no "Re-run" button as discussed in
> chat). Can anyone grab the sh/*.log files from the latest build here?
>
>
https://gitlab.com/nbdkit/libnbd/-/jobs/8081421582
... or I was thinking we should do what we do in Fedora builds, and
have a bit of script which dumps all the log files when the tests
fail. This is a better and more general solution:
https://src.fedoraproject.org/rpms/libnbd/blob/rawhide/f/libnbd.spec#_286
I will have a look at how easy this is to add to ci/build.sh. I
wonder if that file is generated by libvirt-ci? It doesn't have the
usual "THIS FILE WAS AUTO-GENERATED" comment so I guess not.