On Mon, 2009-05-18 at 12:13 +0200, Markus Duft wrote:
> On Mon, 2009-05-18 at 11:56 +0200, Fabian Groffen wrote:
> > On 18-05-2009 10:39:44 +0200, Markus Duft wrote:
> > > > Maybe this is something totally different, but doesn't the
> > > > eclass/portage somewhere *remove* the executable bit of shared
> > > > libraries?
> > > 
> > > hmm. have to look that up, but i guess what you mean is the handling
> > > of .la files?. however it is strange because it is the _only_ .so that
> > > is not executable. and of course non-executable shared libraries don't
> > > work.
> > > 
> > > i'd like to avoid a chmod a+x :)
> > 
> > Please check misc-functions.sh and isolated-functions.sh
> > 
> 
> mhm, right; /opt/gentoo/usr/lib/portage/bin/misc-functions.sh
> (install_qa_check()) does all i want, and the code seems to be
> well-written enough to get libintl.so*, and make it executable. however
> it doesn't (and as i said, libintl is the only one) :( .
> 
> also it seems portage calls install_qa_check rather unconditionally, so
> it shouldn't be the case that any pre_src_install (or any other hook)
> makes the call to the function disappear.
> 
> I'll do further investigation.

something strange is happening here. at the time the install_qa_check
verifies the executable bit on libintl.so.8.0.2, it seems to _be_
executable... but when it gets copied/moved to the final destination,
the executable bit gets lost somehow - i can't find an explanation for
it right now... any ideas?

Cheers, Markus

> 
> Cheers, Markus
> 
> > 
> 
>