Re: Python independant binary extensions
On Fri, Jun 30, 2006 at 06:43:22PM +0200, Josselin Mouette <joss@debian.org> wrote:
> Le vendredi 30 juin 2006 à 17:38 +0200, Mike Hommey a écrit :
> > On Fri, Jun 30, 2006 at 04:16:40PM +0200, Josselin Mouette <joss@debian.org> wrote:
> > > Python-support already checks the md5sums of the files to install, but
> > > it excludes the .so because files are moved to /usr/share and it would
> > > violate the FHS. I also thought this would never happen.
> > >
> > > I can add the handling of this case by putting them e.g.
> > > in /usr/lib/python-support/python-foo/all/.
> >
> > What if, let's say, python 2.3 and 2.4 versions are identical, but 2.5
> > is different ?
> >
> > Also, don't forget the files are different until they are stripped. So
> > the basis for differenciation shouldn't be an md5sum...
>
> It's kind of chicken-and-egg problem. If dh_pysupport is run before
> dh_strip, it won't detect this similarity. If it is run after dh_strip,
> it won't be possible to use dh_strip --dbg-package as gdb wouldn't be
> able to find the symbols at their final location.
>
> If the case is sporadic enough, I guess we should just let go or handle
> the case by hand. If it is not, I can try to modify python-support so
> that it moves files in /usr/lib/debug as well, but it's kind of ugly.
Or you could compare all ELF sections except .debug.* or something
along these lines. Or strip in a temporary directory, but that'd mean
the stripping work would be done twice...
Mike
Reply to: