Re: Building against testing is good (was: Propossed Project: Odyssey)
On Wed, Oct 24, 2001 at 06:45:38PM +0200, Bernd Eckenfels wrote:
> On Wed, Oct 24, 2001 at 12:00:37PM -0400, H. S. Teoh wrote:
> > I believe the default behaviour of the debhelper scripts is responsible
> > for this. I've never myself directly specified that my packages depend on
> > the latest possible libc, but I do run the latest libc, and so when I
> > build my packages, debhelper just picks up that version number.
>
> That is why it is a good idea to build packages on testing instead of
> unstable. That way it is much more likely, that the package has all
> dependencies resolved and can move to testing, quickly after upload.
I believe this advice is incorrect.
If you don't build against the lib versions in unstable, then the libs
will never move to testing. They won't move to testing, because doing
so would break packages (the packages which depend on the old
versions).
This proviso does not apply if the old lib is also available as an
oldlibs package. However, policy is (I think) only to provide oldlibs
packages when the old version actually released: e.g. we don't expect
the libgal maintainer to keep oldlibs packages of libgal9,10,11,12....
For this reason (and possibly others) I thought the advice was to
compile against packages in unstable. However, I Cc: this to the
release manager for clarification on that point...
Incidentally, this is not really germane to the point made earlier in
this thread, which is about unnecessary versioned Build-Dependencies.
I think probably versioned Build-Deps should only be used in a
situation of last resort (I don't use any in balsa, I don't think).
Jules
Reply to: