On Thu, Nov 27, 2008 at 08:37:31PM +0100, Farkas Levente wrote: I suspect they start out the same, but prelink stupidity kicks in. $ rpm -qlvp mingw32-binutils-2.18.50_20080109_2-8.fc9.x86_64.rpm | grep ar\$ -rwxr-xr-x 2 root root 622560 Sep 24 16:52 /usr/bin/i686-pc-mingw32-ar -rwxr-xr-x 2 root root 622560 Sep 24 16:52 /usr/i686-pc-mingw32/bin/ar Unfortunately although I keep going on about how prelink is the most completely broken concept ever, it persists in all versions of Fedora. Rich. -- Richard Jones, Emerging Technologies, Red Hat http://et.redhat.com/~rjones virt-p2v converts physical machines to virtual machines. Boot with a live CD or over the network (PXE) and turn machines into Xen guests. http://et.redhat.com/~rjones/virt-p2v

On Thu, Nov 27, 2008 at 09:45:59PM +0100, Farkas Levente wrote: Perhaps a hard link, but if it's not causing any actual bug, I'd rather leave it how it is. Rich. -- Richard Jones, Emerging Technologies, Red Hat http://et.redhat.com/~rjones Read my OCaml programming blog: http://camltastic.blogspot.com/ Fedora now supports 68 OCaml packages (the OPEN alternative to F#) http://cocan.org/getting_started_with_ocaml_on_red_hat_and_fedora

On Fri, Nov 28, 2008 at 11:44:11AM +0100, Farkas Levente wrote: Hard links might be a bit more robust, but really is this causing a bug? I'm very wary of changing these details of what gcc is doing. I don't understand why gcc needs to keep two copies of the binaries, and I don't understand why it puts a copy in /usr/i686-pc-mingw32/bin/ (a non-standard location), but my lack of understanding means I don't want to mess with this. Rich. -- Richard Jones, Emerging Technologies, Red Hat http://et.redhat.com/~rjones virt-p2v converts physical machines to virtual machines. Boot with a live CD or over the network (PXE) and turn machines into Xen guests. http://et.redhat.com/~rjones/virt-p2v

Daniel P. Berrange wrote: but why not just create symlinks? is there anything what we can loose? in the %install section for all files in /usr/i686-pc-mingw32/bin can be a symlink. actually it's not cause a bug, but after hardlink it's getting confusing (i just discover this as i set the new macros and find two different bins). from the binutils buildlog it's clear they also create link (although hard links), but then rpm forget about it:-( here is one such part of the build log (i check all file in /usr/i686-pc-mingw32/bin are a hard link during the build process): ------------------------------------------------- mkdir -p -- /home/lfarkas/rpm/BUILDROOT/mingw32-binutils-2.18.50_20080109_2-8.fc11.i386/usr/i686-pc-mingw32/bin for i in nm-new strip-new ar ranlib dlltool objdump objcopy; do \ if [ -f $i ]; then \ j=`echo $i | sed -e 's/-new//'`; \ k=`echo $j | sed 's,^,i686-pc-mingw32-,'`; \ if [ "/usr/bin/$k" != "/usr/i686-pc-mingw32/bin/$j" ]; then \ rm -f /home/lfarkas/rpm/BUILDROOT/mingw32-binutils-2.18.50_20080109_2-8.fc11.i386/usr/i686-pc-mingw32/bin/$j; \ ln /home/lfarkas/rpm/BUILDROOT/mingw32-binutils-2.18.50_20080109_2-8.fc11.i386/usr/bin/$k /home/lfarkas/rpm/BUILDROOT/mingw32-binutils-2.18.50_20080109_2-8.fc11.i386/usr/i686-pc-mingw32/bin/$j || /bin/sh ./libtool --mode=install /usr/bin/install -c $i /home/lfarkas/rpm/BUILDROOT/mingw32-binutils-2.18.50_20080109_2-8.fc11.i386/usr/i686-pc-mingw32/bin/$j; \ fi; \ else true; \ fi; \ done ------------------------------------------------- -- Levente "Si vis pacem para bellum!"