Bug#784009: Segmentation fault

Magnus Holmgren holmgren at debian.org
Sun May 10 17:06:48 UTC 2015


måndagen den 4 maj 2015 19.32.52 skrev  Andreas Metzler:
> On 2015-05-03 Magnus Holmgren <holmgren at debian.org> wrote:
> > lördagen den 2 maj 2015 17.47.21 skrev  Andreas Metzler:
> > > On 2015-05-02 Niels Möller <nisse at lysator.liu.se> wrote:
> [...]
> 
> >> /I/ think that would help, afaict we would need to either package a
> >> nettle-2.7 with versioned symbols or patch Debian's version.
> > 
> > Not sure how this situation is normally handled, but when Nettle 3.1
> > is uploaded to sid, new versions of GnuTLS and other packages
> > linking against it should follow soon after [1], so the problem is
> > temporary.
> 
> Hello,
> "temporary" can take surprisingly long if one of the linking packages
> suddenly develops a build error. ;-)
> And one can get a surising amount of bug reports from people doing
> partial ugrades, too.

Yeah, some additional testing and some coordination and/or swift (bin)NMUing 
may be in order, but I say let's get this over with. :-)

> > In testing there should never be more than one version as
> > there's only one source package.
> 
> I somehow missed that fact in my considerations. (I think that at some
> point the RM considered changing the testing migration infrastructure
> to allow keeping old, "orphaned" binaries around to simplify
> transitions. - I am not sure whether this was ever implemented.)

You may be right. We should keep in mind that all binary packages need to be 
accompanied in the archive by the corresponding sources, but that's not 
impossible to do.

> > I'm not aware of any special
> > provisions for transitions of libraries *without* symbol versions,
> > but since Nettle does now. A 2.7.x with symbol versions might
> > still be helpful to some if you meant for it to be uploaded to jessie (and
> > stable point releases of various Debian derivatives).
> 
> I do not think we have ever done that, making a stable-update just to
> introduce versioned symbols. - Needs doublechecking with -release.

No, I don't think that's an accepted cause for a stable update. All depending 
packages would need to be rebuilt and re-uploaded as well for it to be 
meaningful.

> > [1] That is, Nettle 3.1 will be uploaded to sid when the current
> > upload to experimental has been cleared by the FTP masters and
> > GnuTLS is ready to follow, which I guess is after the guile-1.8
> > transition, but perhaps an upload of 3.4.0 to experimental is in
> > order before then? Last I checked, all other packages linking
> > against nettle needed no code changes.
> 
> I think GnuTLS being listed as part of the guile-1.8 removal is an
> error in the reporting script, GnuTLS has moved to guile-2.0 in 2013.
> OTOH I hope we can have gnutls linked against nettle 3.x without
> needing to update to 3.4.x. (Which would couple together two
> transitions. And 3.4 ist still a development release.)
> <http://lists.gnutls.org/pipermail/gnutls-devel/2015-May/007583.html>

A rather significant patch, but since someone has already done it, will you 
use it? 3.1-1 has entered experimental, as you may have seen. Should I upload 
3.1.1-1 to unstable or to experimental for now?

-- 
Magnus Holmgren        holmgren at debian.org
Debian Developer 



More information about the Pkg-gnutls-maint mailing list