[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Manpower in Debian Common Lisp Team



Hi Kambiz,

Le jeudi 31 mars 2022 à 16:17 +0200, Kambiz Darabi a écrit :

> > - enable the --with-sb-linkable-runtime feature on armel, armhf and
> > arm64, this is now possible since SBCL 2.2.0 (and also possibly on
> > kfreebsd-amd64 and kfreebsd-i386, though SBCL currently does not build
> > there). You just need to expand the list of architectures on line 31 of
> > debian/rules. Note that this is just a suggestion, I don’t know if this
> > will work or if this is useful to users (but someone requested it for
> > x86 in the past).
> 
> done

Note that I don’t see that change in your repository.

> > Alternatively, it’s possible to have autopkgtests (and lintian,
> > piuparts, reprotest…) run automatically on salsa.debian.org at every
> > git push. See https://salsa.debian.org/salsa-ci-team/pipeline for
> > instructions on how to enable it.
> 
> I don't have the permissions to modify the common-lisp-team/sbcl CI/CD
> settings, would you please set
> 
> Settings > CI/CD > Generall Pipelines > CI/CD configuration file
> 
> to "debian/salsa-ci.yml"?

I have granted you “Maintainer” permissions in the common-lisp-team
group on salsa.debian.org. My understanding is that you should now be
able to change the CI/CD settings of the common-lisp-team/sbcl
repository. Please let me know if this is not the case.

> In my fork, the i386 build fails with an error:
> 
> backtrace.c:610: Error: bad register name `%rbp'
> 
> https://salsa.debian.org/darabi-guest/sbcl/-/jobs/2624274
> 
> which I probably have to report upstream, right?

Looking at the source code, it seems that the LISP_FEATURE_X86_64 macro is defined when building the i386 package:
https://salsa.debian.org/darabi-guest/sbcl/-/blob/master/src/runtime/backtrace.c#L607

Actually make-config.sh detects an x86_64 architecture, see the job
log:
https://salsa.debian.org/darabi-guest/sbcl/-/jobs/2624274#L1326

Obviously this is incorrect. The detection is apparently based on the
output of 'uname -m'. On my machine, inside an i386 chroot, the output
is i686, which should lead to the correct outcome. And I don’t see any
recent change in make-config.sh that could have caused the problem. So
my guess is that the environment in which i386 packages are built
inside the Salsa CI runners is different, and causes the incorrect
detection.

I would therefore ignore this problem for the time being, try an
upload, and see whether the package builds on i386 build daemons. If it
does, then the problem is specific to the CI runners (in which case it
could make sense to investigate why).

Best wishes,

-- 
⢀⣴⠾⠻⢶⣦⠀  Sébastien Villemot
⣾⠁⢠⠒⠀⣿⡁  Debian Developer
⢿⡄⠘⠷⠚⠋⠀  https://sebastien.villemot.name
⠈⠳⣄⠀⠀⠀⠀  https://www.debian.org

Attachment: signature.asc
Description: This is a digitally signed message part


Reply to: