[Pkg-iscsi-maintainers] Bug#784092: open-iscsi: udebs for some archs completely empty

[Ritesh Raj Sarraf]

Hello Christian,

Apologies for the delayed reply.



On Sunday 03 May 2015 04:51 AM, Christian Seiler wrote:
> Package: open-iscsi
> Version: 2.0.873+git0.3b4b4500-8
>
> (I'm reporting this to keep track of the issue. I've found this by
> chance while improving the packaging.)
>
> udebs are currently only built for select architectures. Unfortunately,
> there are two separate lists that have gotten out of sync.
>
> The first list is in debian/control and tells Debian's build tools for
> which architectures the udeb should be built at all. The list of
> architectures there is:
>
> amd64 arm64 i386 ia64 mips mipsel powerpc s390x ppc64el ppc64 armhf
>
> The second list is in debian/rules. It is used to make the
> determination whether to populate the udeb or not. That list only
> contains the following architectures:
>
> amd64 arm64 i386 ia64 mips mipsel powerpc
>
> This means that for the following 4 architectures the udeb is built but
> never populated during build: s390x ppc64el ppc64 armhf
>
> You can see that if you look at the package sizes for different
> architectures:
> https://packages.debian.org/unstable/open-iscsi-udeb
> Most udebs are a few 100 K large, on those 4 archictectures they are
> less than 1 K large (which is probably just useless metadata).
>
> @Ritesh:
>
> I've already fixed this in my local packaging (will push to git on
> alioth soon), where I have gotten rid of the separate list in
> debian/rules (making it impossible for the bug to resurface), but two
> questions remain for me:
>
>  1. I think we should ask the release team to make a stable update for
>     Jessie's first point release, because this is really bad. The
>     installer won't be able to provide iSCSI on those 4 architectures
>     at all.
>
>     If you agree, we should do an upload to unstable first with my
>     packaging changes that also fix the bug (we need to fix it in
>     unstable first before a PU will be accepted) and then prepare a
>     targeted fix for Jessie (by just adding the missing archs to
>     debian/rules). Normally I would not have suggested an unstable
>     upload so soon already (I would have liked to have more changes in
>     git beforehand), but this bug seems rather nasty to me.
>
>     So if you are onboard, I'll run gbp dch on just my packaging
>     changes so far in unstable so that a finished -9 package is in git.
>     You can then do an unstable upload, I'll prepare a targeted upload
>     for Jessie, push that into git, ask the release team for approval
>     (which my guess is they'll grant for this type of bug) and then you
>     could upload the specifically fixed version to jessie-p-u.

Yes. That is the right approach. Let's get it into Unstable. The stable
team anyways would ask the same.

>
>  2. Is there a reason why the udeb is not Architecture: any instead of
>     the specific ones? I mean, the binaries are the same ones as in the
>     normal deb package (no separate build), so I don't see why one
>     needs this in the first place? Is this perhaps a relic from the
>     time where the package still built kernel modules?
>
>     Because I'd really like to get rid of the specific architecture
>     list in general, that will reduce the maintenance burden in the
>     long run - so ideally I'd change that to any or linux-any or so.
>
> Thoughts?
>
> Christian
>
If memory serves me correct, the problem was that open-iscsi's
dependency, scsi-modules, wasn't available on all arches in the installer.

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=759817

You should go through the bug report and then check the relevant
subsystem's current status.

-- 
Ritesh Raj Sarraf | http://people.debian.org/~rrs
Debian - The Universal Operating System


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: OpenPGP digital signature
URL: <http://lists.alioth.debian.org/pipermail/pkg-iscsi-maintainers/attachments/20150512/77ab5cfa/attachment.sig>