Re: built packages based on Apr 15 snapshot
Ben Collins writes:
> On Mon, Apr 16, 2001 at 05:38:17PM -0600, Matt Taggart wrote:
> > Hi debian-gcc,
> >
> > FYI...
> >
> > I built newer gcc-3.0 packages for hppa based on a cvs snapshot most recently
> > synced up with upstream cvs on 2001-04-15. So basically I just replaced the
> > gcc-20010403.tar.bz2 tarball with a newer one. When building I ran into a
> > patch failure with debian/patches/sonames.dpatch because
> > libjava/libtool-version changed upstream slightly,
> >
> > - 1:0:0
> > + 2:0:0
> >
> > After fixing sonames.dpatch all patches apply(with some offsets) and
> > everything build/works fine. Some patches will need to be regen'd to get rid
> > of the offsets but that should be easy.
>
> BTW, I'm a bit concerned about the soname stuff being used for hppa.
> What happens when we get a final ABI, and want to get rid of the hack?
> Wont it break things on hppa? Should we resolve this now?
Until now it's not clear if ABI's change in the branch. In the past
ABI numbers were not changed during the development leading to the
releases. To play safe, I did choose other ABI numbers for the
snapshots.
If the ABI of a package doesn't change until the 3.0 release:
- ln -s libfoo.so.<upstream num> libfoo.so.<our num>
- provide libfoo<our num>
If the ABI changes, make the package libfoo<upstream num> which
doesn't conflict with the already uploaded package.
What can we do if we use now the upstream num and the ABI still
changes? (At least that could be the case for ia64 exception handling,
which could go into the branch).
Unfortunately 3.0 was moved to testing before we did resolve this...
Reply to: