Re: Please hint krb5 1.7 into testing and retain libkrb53 in testing
Sam Hartman wrote:
>>>>>> "Luk" == Luk Claes <email@example.com> writes:
> Luk> Sam Hartman wrote:
> Luk> Hi
> >> This is an update on my message about the krb5 transition.
> >> Currently, if krb5 1.7 from unstable were to migrate into
> >> testing and if the libkrb53 binary package were maintained in
> >> testing (it disappears from the krb5 source package between
> >> testing and unstable), I believe we have high confidence that:
> >> 1) nothing would break
> Luk> I'm not so sure this is the case as I tried it last night and
> Luk> britney refused to transition it because of making packages
> Luk> uninstallable in testing. I guess that's because some
> Luk> packages that try to migrate together depend on krb5 already
> Luk> and are not installable...
> *sigh* You need to maintain libdes425-3 in testing. I thought
> libkrb53 included both libkrb4.so.2 and libdes425.so.3, but it depends
> on libdes425-3 for the second.
> )(Sorry, there were multiple different breakdowns of this around)
> If that's not it then then please read on; otherwise sorry for wasting
> your time.
> even if you manage to fix this with uploads to unstable, I'd like to
> understand a bit more about what went wrong here.
> I don't like it when I give people advice that might break things and thought I had very high confidence in my statement:-)
> You tried to hint krb5 into testing while retaining the libkrb53
> binarypackage alrready in testing and something became uninstallable?
Ah, that's the reason. I thought there were different source packages
involved (for libkrb53), but apparently you ask to do a gross hack to
keep the libkrb53 binary package in while migrating krb5 and its
remaining binary packages. This is possible, though I'd rather not do that.
If we would do something like that we would probably retain all the
binary packages from the previous source package btw.
Sorry for misunderstanding, though krb5 is almost ready to migrate
without having to leave libkrb53 in testing, so I'd prefer to do that.