On Tue, Feb 20, 2007 at 12:13:07AM +0100, David Härdeman wrote:
> >Ok. Then there's left the error produced by evms_activate after doing
> >what it needed to do ("libgcc_s.so.1 must be installed for
> >pthread_cancel to work\nAborted") which I mentioned in my original
> >report, perhaps that's something that needs to be added to initramfs?
>
> How did you get evms_activate into the initramfs image? The hook scripts
> usually call a shell function which automagically adds all necessary
> libraries as well...which it seems that it hasn't done.
mkinitramfs installed it there. Note that the evms_activate binary
actually printed the error message, it wasn't a dynamic linker error.
I think glibc tries to dlopen() the library, prints the error after
the failure and aborts. That can't obviously be caught by the
mkinitramfs magic.
I think pthreads needs it for at least pthread_cancel to work at least
on amd64. Might need it for some other pthread stuff too, the library
seems to define other (local) weak symbols in addition to
pthread_cancel, perhaps most notably pthreads_mutex_(un)lock and
pthread_once.
Perhaps it should be investigated if the library is needed on other
arches too and possibly included in the initramfs by default in those
arches as it's probably needed by many threaded programs.
Sami
signature.asc
Description: Digital signature
