Mailinglist Archive: opensuse-buildservice (248 mails)

< Previous Next >
Re: [opensuse-buildservice] Re: [opensuse-factory] Devel packages not a links to Factory
  • From: Michal Vyskocil <mvyskocil@xxxxxxx>
  • Date: Wed, 14 Oct 2009 10:26:37 +0200
  • Message-id: <200910141026.43692.mvyskocil@xxxxxxx>
On Tuesday 13 of October 2009 16:50:05 Richard Guenther wrote:
On Tue, 13 Oct 2009, Richard Guenther wrote:
On Tue, 13 Oct 2009, Michal Vyskocil wrote:
Hi,

during work on fix for apache2-mod_jk I found that even if the
Apache:Modules/apache2-mod_jk is marked as devel project, it contains
completely different package with newest version and a different
history, which makes it as devel project unusable. So I converted it to
source link and submit a fix again.

Anyway I used an attached Python script to find all packages not have
the _link in a filelist - their maintainers would consider to fix it. I
also attached a list of affected packages.

BTW: is there any simple way how to convert the package to source link?
I used the sequence

osc linkpac openSUSE:Factory $PACKAGE $PROJECT
osc rremove $PROJECT $PACKAGE `osc ls $PROJECT $PACKAGE | grep -v
_link`

but it messed the commit log, because every one deletion is one commit.

Well, luckily you didn't catch devel:gcc/gdb. And of course a setup
this way may be completely on purpose! It definitely is for gdb.

Thats not a luck - that is a list of packages not contains a _link, what gdb
is not. And I don't want to force anyone to fix it, this is just JFI.


Btw. the very much most important case in that this is the preferred
way is if you track an upstream rpm package and do not want to have
gazillions of merge errors just because the Factory package gets its
spec file munged completely on each checkin.

Yes, I'm aware about it. But I'm not sure how the semantics is? Is the _link
just for the default submission target of osc submitreq? Will the --diff and
other thinks work? I'm thinking about same workflow for some of my packages
(unsurprisingly from Fedora too), so it's the reason why I'm asking.


Ok, that had to be said.

.spec file changing on checkin is EVIL. EVIL. EVIL. (can't repeat
it often enough).

fully agreed - do we know what's the reason for that? If some parts of spec
don't match some requirements, it should be told using rpmlint, or maybe in
rejected checkin.

Regards
Michal Vyskocil
< Previous Next >