Steve Langasek | 1 May 2005 02:42
Picon
Favicon

Re: libsasl2 and libmysql* updates for sarge

On Sat, Apr 30, 2005 at 10:03:01AM -0300, Henrique de Moraes Holschuh wrote:
> On Mon, 25 Apr 2005, Steve Langasek wrote:
> > Since dovecot, libnss-mysql, and libnss-mysql-bg have already been updated to
> > use libmysqlclient12, the main reason for *not* updating cyrus-sasl2, the
> > MTAs, and the ftp daemons (i.e.: segfaults from mixing and matching) no
> > longer applies.  I would accept an update of cyrus-sasl2 built against
> > libmysqlclient12, but I would prefer to see at least exim4 and postfix
> > updated first.

> Do you want an cyrus-sasl2 upload that conflicts with all packages depending
> on both cyrus-sasl2 and libmysqlclient{!12} ?  I will bump the shlibdep info
> as well.

I'd rather not have either conflicts or a shlibdep bump here, because I
think there's a diminishing-returns argument against it: conflicts in
particular make upgrades painful, and shlibdeps have no bearing on plugins.

> > Even if they weren't all done together, there's still a possibility of
> > segfaults because of libnss-mysql right now, so it's worth transitioning as

> The truth is libnss-mysql (and any other libnss-* of the sort) should
> conflict with any libmysqlclient (or any other lib of the sort) it is not
> linked to.  This is utter braindamage that can only be really fixed by
> enforcing a versioned-symbols-or-die rule.

It's braindamage that we've lived with for a while, without major incident;
while I want to eliminate as much of it as possible, it doesn't seem to be
enough of a problem in practice to justify making it harder to upgrade in
the general case.

(Continue reading)

Picon
Favicon

Re: libsasl2 and libmysql* updates for sarge

On Sat, 30 Apr 2005, Steve Langasek wrote:
> particular make upgrades painful, and shlibdeps have no bearing on plugins.

Ok, no conflicts, no shlibdep bump.

> > The truth is libnss-mysql (and any other libnss-* of the sort) should
> > conflict with any libmysqlclient (or any other lib of the sort) it is not
> > linked to.  This is utter braindamage that can only be really fixed by
> > enforcing a versioned-symbols-or-die rule.
> 
> It's braindamage that we've lived with for a while, without major incident;

And that we will kill in etch, one way or the other (and I'd rather we kill
it using versioned symbols).  This is something post-sarge, obviously.

> > > Incidentally, I think cyrus-sasl3-mit's build-dep on libmysqlclient10-dev is
> > > spurious and should be dropped.
> 
> > Please file a bug, I am not the maintainer for cyrus-sasl2-mit...
> 
> Oh, but you offered to take care of it, I thought :)

Only for cyrus-sasl2.  Sam Hartman is NOT semi-MIA AFAIK, so there is no
reason for me to go messing with his packages...

--

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh
(Continue reading)


Gmane