Re: C++ transition stumbling blocks?
On Tue, Jan 21, 2003 at 07:35:43PM +0100, Andreas Rottmann wrote:
> Junichi Uekawa <dancer@netfort.gr.jp> writes:
>
> [aspell stuff snipped]
>
> Another thing that holds a few packages back (among them zinf, which I
> really would like to upload a new version) is id3lib, which also FTBFS
> with g++ 3.2. This seems fixed with 3.8.2, upstream, however. I'll
> probably NMU, if I get no response from the maintainer (CC'ed).
I was planning to do that last night however I've been feeling very under
the weather recently. I apologize for not doing it sooner.
> One thing that strikes me wrt this package is that there are two
> source packages id3lib (upstream version 3.7.x) and id3lib3.8.0. This
> source packge is named after the *full* (major.minor.micro) version
> number, which is also done with the resulting binary packages. This is
> not reasonable, IMO.
I'd suggest that anyone who NMUs id3lib learn when upstream breaks the ABI -
because it is most certainly not when they release a new major version.
Don't go changing the soname around on me either. I know it's messed up - it
needs to stay messed up unless you want to be gratuitously incompatible with
everyone else.
Either learn about this fast or don't NMU.
> I suggest the following resolution: have id3lib3.8.0 removed from the
> archive, id3lib upgraded to 3.8.2 and all packages that don't build
> against it fixed. They'd have to be fixed anyway, since I assume
> id3lib 3.7.x doesn't build with g++ 3.2, too. I also suggest the
> id3lib maintainer read the libpkg-guide[0].
Yeah! Let's immediately make packages uninstallable! Woohoo!
(I've already made that mistake once - hint: you'll need an entirely
independant source package for this upload.)
--
Robert Woodcock - rcw@debian.org
"It's not a crime to be incompetent, and that's Ebbers' defense, I guess"
-- Richard Tilton, Bankruptcy Attorney, on Worldcom's former CEO
Reply to: