Bugzilla Bug 226621

Nothing unusual besides a gcc-4.3 regression.
Other archs already have it stable

Thanks!

kpogre-1.5* won't build with that version and that is no regression...should I
go for kpogre-1.6.6?

well, the problem is that I committed kpogre-1.6.6 10 days ago. If you want to
mark it stable before the usual 30 days, please go ahead.

x86 stable

amd64 stable, we don't have a stable kpogre so I'll leave that alone.

alpha/ia64/sparc stable, closing

Well, I missed to fix the monlitic koffice ebuild, which specifies postgres? (
<dev-libs/libpqxx-2.6.9 ), yet.

You are aware that you have to check the repository, not to break the stable
tree, are you!?

(In reply to comment #0)
> Nothing unusual besides a gcc-4.3 regression.
> Other archs already have it stable

Honestly this isn't a valid reason to rush it at all. Not carefully checking
the whole stable tree and waiting for all issues to be fixed before going ahead
is quite unacceptable. I wonder why no one of those stabilizing for their
secific architecture stumbled about it. Is the QA approach of Gentoo _that_ low
nowadays?!

Just adding a reference to bug 229093 - "media-gfx/showimg-0.9.5 won't build
against dev-libs/libpqxx-2.6.9".

libpqxx-2.6.9 just went back to unstable. Really, a few years back, you'd got a
real kick in the ass for such an action.

come on, you had more than 3 months time to fix that stuff. And marking the
thing stable doesn't break the stable tree, only removing 2.6.8 would. So,
please revert your changes and close this bug.

Um, media-gfx/showimg-0.9.5 is stable and won't build against
dev-libs/libpqxx-2.6.9. Unless I'm misunderstanding, that means
dev-libs/libpqxx-2.6.9 breaks the stable tree, no?

No, that means showimg is broken.

(In reply to comment #10)
> come on, you had more than 3 months time to fix that stuff. And marking the
> thing stable doesn't break the stable tree, only removing 2.6.8 would.

You seem to have a completely broken understanding of what are the requirements
of the stable tree and what are your responsibilities with regards to
stabilization of an ebuild other ebuilds depend on.

1) The latest stable ebuild of any package:slot combination requiring the
ebuild you want to mark stable has to work flawlessly with exactly the ebuild
you want to mark stable (under the condition it is the highest version within
the slot).  No, some stable ebuilds working with libpqxx 2.6.8 only and others
with 2.6.9 is not o.k. and exactly the definition of a broken stable tree. 

2) It is the maintainers responsibility to check the tree for potential issues,
with the usual procedure to create a tracker bug (if necessary as in this case)
and cc affected maintainers ask for others to help you in case of unmaintained
packages, e.g. via the gentoo-dev mailing list and test as much stuff as you
can yourself.

3) Every arch maintainer is held to do exactly these checks for the
architectures he feels responsible for, himself. Otherwise we could mark
directly stable and wouldn't need the seconfd pair of eyes.

And btw.: No one had any time to do anything about this stabilization request,
because no maintainer of affected packages has been informed.

cc'ing other maintainers, who broke the tree on their architecture.

(In reply to comment #15)
> cc'ing other maintainers, who broke the tree on their architecture.

I did what? :)

@carlo: Em, no. koffice has this in the DEPEND: "<dev-libs/libpqxx-2.6.9". As
long as dev-libs/libpqxx-2.6.8 stays in the tree everything's fine since the
package manager should not show the update to dev-libs/libpqxx-2.6.9 if koffice
is installed or will be installed.
And for the showimg case: change the dependency to the above and all will be
fine until you fix it.

(In reply to comment #15)
> cc'ing other maintainers, who broke the tree on their architecture.
> 

Would you mind explaining what I did wrong wrt this bug?

From the libpqcxx ChangeLog:

  13 Aug 2007; Tobias Scherbaum <dertobi123@gentoo.org>
  libpqxx-2.6.9.ebuild:
  ppc. stable

Especially note "August" and "2007". 

(In reply to comment #17)
> @carlo: Em, no. koffice has this in the DEPEND: "<dev-libs/libpqxx-2.6.9". As
> long as dev-libs/libpqxx-2.6.8 stays in the tree everything's fine since the
> package manager should not show the update to dev-libs/libpqxx-2.6.9 if
> koffice is installed or will be installed.

The keyword here is 'should'. Your ideal does not match reality. Hence it
breaks the tree.

@zlin: well, this is the package managers problem (or it's devs), not mine.

Carlo is in a way right, that such up- and downgrade cycles (emerge pulls in
2.6.9 as latest stable, koffice in the next run 2.6.8) are not acceptable in
stable profile.  I did check the stable profile that's why I asked for kpogre,
so koffice slipped me in a way.  This is of course the maintainers problem, but
normally I as a architecture dev do refuse such a package until the maintainers
has it fixed.
So go on and propose an actual solution instead of fighting about if the
unstabling was the right action to do.

Ok, please get this moving somehow...as 2.6.9 is the only version fixed for GCC
4.3.  Thanks.

It's a shame that showimg includes a beta copy of kexi. It seems to be no
longer maintained, and I think it would be difficult to get showimg to use an
external version of kexi. I have attached a fix for showimg to compile against
libpqxx-2.6.9 on bug #229093

Created an attachment (id=178889) [details]
unofficial ebuild for version 3.0

For what I understand, gcc-4.3 cannot go stable until dev-libs/libpqxx-2.6.9
goes stable, as it blocks app-office/kexi.

kexi can also work on libpqxx-2.6.8 as soon as postgresql team patches it.
Dropping severity to normal

I can't really figure out what you guys are arguing about, but I would like to
compile koffice. It fails because it wants to downgrade to libpqxx-2.6.8 which
will not compile.

Any idea when this will be resolved?

*** This bug has been marked as a duplicate of bug 249248 ***