Re: [email@example.com: Re: Woody retrospective and Sarge introspective]
Le Wed, Jul 31, 2002 at 02:47:50PM +0100, Mark Brown écrivait:
> So after testing with the versions in unstable for a while package gets
> built and reuploaded against a potentially completely different set of
Yes, but they still have to go through the "testing scripts" ensuring
that no new RC bugs have been found on those recompiled package and that
they compile on all arches.
> That seems more than a little suboptimal since it reduces the
> benefits of the testing that has been done. It also does nothing to
> address problems with undeclared dependencies breaking testing.
Partially it does, because people testing t-p-u will file RC bugs against the
package because they are testing in an environment that is 80% "testing"
(and 20% "t-p-u"). The bug is likely to be detected before it enters
I expect t-p-u to be quite small compared to a complete distribution
- packages would enter testing more easily (we don't have dependencies
- t-p-u could be purged of packages that are not able to enter
testing after a certain delay
(t-p-u is not meant to be a complete distribution and is not meant to
be used to compile things against it)
> Plus, either source only uploads will have to be supported or people
> will have to have give themselves a testing environment to do things in
Yes. But it's getting easy to setup a testing chroot for your builds.
Raphaël Hertzog -+- http://strasbourg.linuxfr.org/~raphael/
Formation Linux et logiciel libre : http://www.logidee.com