Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.debian.devel > #113508 > unrolled thread

How painful is a lib(fuse) .so version bump for a distribution

Started byBernd Schubert <bernd.schubert@fastmail.fm>
First post2024-09-25 22:00 +0200
Last post2024-09-27 11:30 +0200
Articles 6 — 4 participants

Back to article view | Back to linux.debian.devel

Contents

  How painful is a lib(fuse) .so version bump for a distribution Bernd Schubert <bernd.schubert@fastmail.fm> - 2024-09-25 22:00 +0200
    Re: How painful is a lib(fuse) .so version bump for a distribution Andrey Rakhmatullin <wrar@debian.org> - 2024-09-25 22:40 +0200
      Re: How painful is a lib(fuse) .so version bump for a distribution Ben Hutchings <ben@decadent.org.uk> - 2024-09-26 00:50 +0200
        Re: How painful is a lib(fuse) .so version bump for a distribution Andrey Rakhmatullin <wrar@debian.org> - 2024-09-26 08:40 +0200
    Re: How painful is a lib(fuse) .so version bump for a distribution László Böszörményi (GCS) <gcs@debian.org> - 2024-09-26 21:20 +0200
      Re: How painful is a lib(fuse) .so version bump for a distribution Bernd Schubert <bernd.schubert@fastmail.fm> - 2024-09-27 11:30 +0200

#113508 — How painful is a lib(fuse) .so version bump for a distribution

FromBernd Schubert <bernd.schubert@fastmail.fm>
Date2024-09-25 22:00 +0200
SubjectHow painful is a lib(fuse) .so version bump for a distribution
Message-ID<JqGl3-eNB3-7@gated-at.bofh.it>
Hello,

I would like to ask how painful is a library (libfuse) .so for a
distribution?


As you can see here https://github.com/libfuse/libfuse/pull/1038
there are some arguments not to do so.
The pull request above explains why and adds even more breakage,
when we already have chance for that. But it would be good to
know how much trouble it is for a distribution.


Thanks,
Bernd

[toc] | [next] | [standalone]

#113509

FromAndrey Rakhmatullin <wrar@debian.org>
Date2024-09-25 22:40 +0200
Message-ID<JqGXM-eO2L-11@gated-at.bofh.it>
In reply to#113508

[Multipart message — attachments visible in raw view] — view raw

On Wed, Sep 25, 2024 at 09:38:26PM +0200, Bernd Schubert wrote:
> I would like to ask how painful is a library (libfuse) .so for a
> distribution?

Not at all, if all revdeps build fine with the bumped version.

"I'm assuming that distributions do not recompile packages against a new
library version, unless absolutely needed - if I'm right library ABI
changes cause silent issues." sounds very wrong to me. We always
recompile packages against a new soname when that is uploaded, and as we
try to keep packages uptodate a new soname is normally uploaded every time
it's bumped by the upstream.

-- 
WBR, wRAR

[toc] | [prev] | [next] | [standalone]

#113513

FromBen Hutchings <ben@decadent.org.uk>
Date2024-09-26 00:50 +0200
Message-ID<JqIZz-ePhM-5@gated-at.bofh.it>
In reply to#113509

[Multipart message — attachments visible in raw view] — view raw

On Thu, 2024-09-26 at 01:31 +0500, Andrey Rakhmatullin wrote:
> On Wed, Sep 25, 2024 at 09:38:26PM +0200, Bernd Schubert wrote:
> > I would like to ask how painful is a library (libfuse) .so for a
> > distribution?
> 
> Not at all, if all revdeps build fine with the bumped version.
> 
> "I'm assuming that distributions do not recompile packages against a new
> library version, unless absolutely needed - if I'm right library ABI
> changes cause silent issues." sounds very wrong to me. We always
> recompile packages against a new soname when that is uploaded, and as we
> try to keep packages uptodate a new soname is normally uploaded every time
> it's bumped by the upstream.

But we *don't* rebuild when there is not an soversion bump, so the
previous unintended ABI changes in libfuse3-3 may well have caused
silent issues for its reverse-dependencies.

Ben.

-- 
Ben Hutchings
If the facts do not conform to your theory, they must be disposed of.

[toc] | [prev] | [next] | [standalone]

#113515

FromAndrey Rakhmatullin <wrar@debian.org>
Date2024-09-26 08:40 +0200
Message-ID<JqQkp-eUqi-7@gated-at.bofh.it>
In reply to#113513

[Multipart message — attachments visible in raw view] — view raw

On Thu, Sep 26, 2024 at 12:48:03AM +0200, Ben Hutchings wrote:
> > > I would like to ask how painful is a library (libfuse) .so for a
> > > distribution?
> > 
> > Not at all, if all revdeps build fine with the bumped version.
> > 
> > "I'm assuming that distributions do not recompile packages against a new
> > library version, unless absolutely needed - if I'm right library ABI
> > changes cause silent issues." sounds very wrong to me. We always
> > recompile packages against a new soname when that is uploaded, and as we
> > try to keep packages uptodate a new soname is normally uploaded every time
> > it's bumped by the upstream.
> 
> But we *don't* rebuild when there is not an soversion bump, so the
> previous unintended ABI changes in libfuse3-3 may well have caused
> silent issues for its reverse-dependencies.

Right, when read like that it's correct.

-- 
WBR, wRAR

[toc] | [prev] | [next] | [standalone]

#113519

FromLászló Böszörményi (GCS) <gcs@debian.org>
Date2024-09-26 21:20 +0200
Message-ID<Jr2bT-f1CT-1@gated-at.bofh.it>
In reply to#113508
On Wed, Sep 25, 2024 at 9:46 PM Bernd Schubert
<bernd.schubert@fastmail.fm> wrote:
> I would like to ask how painful is a library (libfuse) .so for a
> distribution?
 In what sense? Upstream ABI breakages don't help, I wait for 3.17 at
least if that helps - not upgrading it to middle versions.

> As you can see here https://github.com/libfuse/libfuse/pull/1038
> there are some arguments not to do so.
> The pull request above explains why and adds even more breakage,
> when we already have chance for that. But it would be good to
> know how much trouble it is for a distribution.
 Debian doesn't recompile dependent packages for new releases. Doing
library transitions (due to new soname) is an everyday thing for us.
The workflow is documented [1], I will not repeat it here. There are
forty-two dependent packages that need to be rebuilt for such a case,
not a big number. I'm open to schedule the transition if upstream
changes the soname.

Regards,
Laszlo/GCS
[1] https://wiki.debian.org/Teams/ReleaseTeam/Transitions

[toc] | [prev] | [next] | [standalone]

#113527

FromBernd Schubert <bernd.schubert@fastmail.fm>
Date2024-09-27 11:30 +0200
Message-ID<Jrfsu-faur-5@gated-at.bofh.it>
In reply to#113519
Hi László,

On 9/26/24 21:12, László Böszörményi (GCS) wrote:
> On Wed, Sep 25, 2024 at 9:46 PM Bernd Schubert
> <bernd.schubert@fastmail.fm> wrote:
>> I would like to ask how painful is a library (libfuse) .so for a
>> distribution?
>  In what sense? Upstream ABI breakages don't help, I wait for 3.17 at
> least if that helps - not upgrading it to middle versions.
> 
>> As you can see here https://github.com/libfuse/libfuse/pull/1038
>> there are some arguments not to do so.
>> The pull request above explains why and adds even more breakage,
>> when we already have chance for that. But it would be good to
>> know how much trouble it is for a distribution.
>  Debian doesn't recompile dependent packages for new releases. Doing
> library transitions (due to new soname) is an everyday thing for us.
> The workflow is documented [1], I will not repeat it here. There are
> forty-two dependent packages that need to be rebuilt for such a case,
> not a big number. I'm open to schedule the transition if upstream
> changes the soname.

thank you very much! So no issue for Debian :)


Best,
Bernd

[toc] | [prev] | [standalone]

Back to top | Article view | linux.debian.devel

csiph-web