182 pending F11 stable updates. WTF?

Michael Schwendt mschwendt at gmail.com
Mon May 11 19:25:25 UTC 2009


On Mon, 11 May 2009 09:37:30 -0700, Jesse wrote:

> I'd really like to know what people's thoughts are on making the
> overrides more known.  One problem we have with them as well is that
> maintainers don't let us know when the build is done, so that we can
> remove the override to prevent poisoning further builds.

How does it help? For example, you tag a library pkg that bumps a SONAME,
then you can rebuild deps against the new SONAME, and when done you want
to untag the lib, so any other packager would build against the old SONAME
again? What am I missing? You need the buildroot override tags to rebuild
deps. Releasing builds into updates-testing would give time to detect
broken deps. Currently. Skipping updates-testing would remove that chance.

As soon as somebody requests to override the buildroots, there must be a
complete plan on whether it will be possible to prepare updates for *all*
deps (isn't it true that most problems originate in upgrades which require
a chain of upgrades of deps?). Or else you need something like "atomic
buildroot override build-jobs". Which either build in an isolated modified
buildroot. Or which lock down the normal buildroot, so only a specific set
of packages as requested by a packager may be rebuilt against temporary
buildroot modifications.

 




More information about the fedora-devel-list mailing list