Don't put new packages through updates-testing

Michael Schwendt mschwendt.tmp0701.nospam at arcor.de
Sat Jun 2 09:50:21 UTC 2007


On Sat, 02 Jun 2007 14:55:37 +0530, Rahul Sundaram wrote:

> Michael Schwendt wrote:
> > On Sat, 02 Jun 2007 14:17:40 +0530, Rahul Sundaram wrote:
> > 
> >> The base functionality must still be tested.
> > 
> > That is much too vague.
> 
> We need to sit down and define the details more precisely if we agree 
> with the fundamental idea.

Don't you fear that it would drive away reviewers?

Back in time we've had to be very careful not to raise the bar too high.
What has changed so that the Fedora Project is willing to increase the
work for volunteer reviewers and packagers?

There are many packages in the collection, which have been packaged only
because they are build- or run-time requirements. For some packages, even
minor version updates result in a unified diff of at least 2MiB in size.
For some programs apparently even a large bunch of upstream developers
cannot perform enough testing.

> > It's "should" here again.
> > 
> > The app may work fine in the reviewers environment, but malfunction in a
> > different environment.
> 
> Sure but the reviewer shouldn't be skipping this step. Agree?

The packager should not skip it either, but some do. So what? It's too
easy for them to excuse themselves with a brief "cannot reproduce that,
works for me" and a quick fix as an update. Quick updates are not possible
anymore, however, because the self-proclaimed Fedora Bureaucracy
task-force has set up new hurdles and is working on producing even
newer ones.




More information about the fedora-devel-list mailing list