Re: 'nother testing question
On Fri, 23 Nov 2001, Jules Bean wrote:
> What's keeping libtool out of testing? It's up-to-date on all
> release candidate architectures, update_excuses lists not problems
> with it. But it's not going in...
> Presumably it's not going in because if it did, it would break too
> much? So that would be packages using libltdl0?
Yes, that's right.
> So does that mean *every* single library package in debian which uses
> libtool (which is most libraries that are trying to be cross-platform,
> I suspect) must be recompiled against the new libtool, before the new
> libtool can get into testing?
> That seems to be balsa, elastic, gql, pspell, according to my woody
> Packages file.
> elastic in sid doesn't depend on libtool at all. gql has been
> recompiled. balsa has been recompiled (I know, I did it!). pspell has
> been recompiled.
> So it isn't those, is it?
> Or is it the case that there has never been any point at which all of
> those packages were valid candidates at the same time for the big
> install to happen?
This is the case - all the packages that depend on this library mustn't
have more RC bugs than their versions in testing, they must be recompiled
onall architectures and they must have already waited their 2/5/10 days at
the same time (and that's why big library updates to testing are nearly
impossible without manual interaction). If you look more detailed at
testing you see that ajt often does some tricks to get things working.
There's a similar problem e.g. with Guile (libguile6 -> libguile9).
> wondering vaguely if this testing idea works as well as it should...
Get my GPG key: finger email@example.com | gpg --import
Fingerprint: B29C E71E FE19 6755 5C8A 84D4 99FC EA98 4F12 B400