Better testing of tagged tars

[Andreas K. Huettel]

Am Montag, 11. Februar 2013, 00:24:51 schrieb Albert Astals Cid:
> El Diumenge, 10 de febrer de 2013, a les 08:15:40, Martin Gräßlin va 
escriure:
> > On Saturday 09 February 2013 23:08:50 Albert Astals Cid wrote:
> > > Of course another option is lifting the requirement for the
> > > pre-packages not being publicly available, after all the packages will
> > > most likely be the real thing, so if everyone agrees it is better
> > > lifting this requirement, we can do it, the fact that *I* personally
> > > like it the way it is doesn't mean it's the better way.
> > 
> > With my bugzilla user hat on I'm afraid of that. It would mean we get bug
> > reports for an unreleased version. That's bound to create confusion  - we
> > would not be able to trust the version field any more. In case of a
> > re-spin it will get just worse - different tar balls with the same
> > version information.
> 
> Another option is just release the tarballs once and don't do any respin at
> all. After all we have build.kde.org that builds the stuff so we are kind
> of "confident" it builds, if anything fails to build or something big is
> found we can add it as a note (+ kde-packager mail) to the info page like
> we did with the nepomuk thing for 4.10.0 http://kde.org/info/4.10.0.php
> 
> That seems like a "sensible" compromise to me.
> 
>  * We release only one tarball
>  * Distros still can pick up build or bugfixes (as they will do anyway
> either we include them in a respin tarball or not)
>  * We can "silently" release the *only one* tarball a few days in advance
> to get distros to package for the release day
> 
> Comments?
> 

Actually, I think the current procedure is OK as it is. Sure, things can go 
wrong sometime, but as long as someone of the knowledgeable upstream 
developers cares and quickly codes a fix or workaround, that is not really a 
big problem. No distro will stick with "vanilla tarballs" just for the fun of 
it if upstream has accepted a patch that solves a real problem into the 
stable/bugfix branch.

[As for the longer testing... well, 1) in the case of .0 versions, that's what 
betas and rc's are for, and 2) in the case of .X versions, X>0, well, it's 
called the "stable branch" for a reason- don't mess with it!

[Actually I've been working on automatically extracting some statistics how 
large a percentage of the code is changed between git tags, since I was 
curious how well "the code converges" in different packages. Will finish this 
once I get around to it... (No Perl.)]]

-- 

Andreas K. Huettel
Gentoo Linux developer 
dilfridge at gentoo.org
http://www.akhuettel.de/

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.kde.org/pipermail/release-team/attachments/20130213/6b4c5e25/attachment.sig>