Re: Bug#27334: libc resolver problem solved (critical bug)
On Wed, 18 Nov 1998, Marcelo E. Magallon wrote:
> Auch.
>
> That means a new gcc-compiled libc6 package won't have this
> __register_frame_info function. That would break a large ammount of
> packages.
Thats right. For example any binary using libstdc++2.9 will fail, because
libstdc++2.9 needs that symbol. Of course the simple fix for this is to
recompile egcs.
> One idea:
>
> * Compile libc6 2.0.7u-6 with gcc and put a (>= 2.0.7u-6) on the -dev
> package.
> * Make a tiny i386-only (other archs don't have this problem) library
> lib__register_frame_info.so.0 that provides the missing function (where do
> we obtain this function from? The egcs source?); preload this.
there is another one iirc: __unregister_frame_info
> * Before releasing slink, make sure nothing has a Dependency like libc6 (>=
> 2.0.7u-4) or libc6 (>= 2.0.7u-5). File a massive critical bug report
> against this packages -- if the maintainer can confirm the package doesn't
> suffer from this (objdump is your friend here), close the bug.
It is libc6 (>= 2.0.7u)
> * When all this bugs have been closed, release a new libc6 with this extra
> library removed (we don't want to carry this forever, do we?)
>
> The last point is the most problematic one, because there's no way to assure
> noone will upgrade the libc6 package leaving an old package installed. The
> third point could be automated unpacking the packages on master and avoid
> the nasty unnecessary bug reports.
>
> Am I nuts?
I dont think so. Btw there is another way: compile libc with a recent
egcs, and then we don't have problems with compatibility (checking this
now, waiting for the compile to finish), only with other dists... but
that's bad too.
--
Madarasz Gergely gorgo@caesar.elte.hu gorgo@linux.rulez.org
It's practically impossible to look at a penguin and feel angry.
Egy pingvinre gyakorlatilag lehetetlen haragosan nezni.
HuLUG: http://mlf.linux.rulez.org/
Reply to: