Bug#671866: gmp: causes gcc ICEs on amd64 and kfreebsd-amd64
jcristau at debian.org
Tue May 8 19:26:40 UTC 2012
On Tue, May 8, 2012 at 19:43:41 +0200, Laurent Fousse wrote:
> * Julien Cristau [Tue, May 08, 2012 at 07:19:15PM +0200]:
> > On Mon, May 7, 2012 at 22:09:55 -0500, Steve M. Robbins wrote:
> > > Interesting. It could well be the configure option. For the past 6
> > > years, Debian built using --disable-fat on amd64. I changed it
> > > because there are both bug workarounds and code speedups that are
> > > processor-based. So it makes more sense for the code to detect the
> > > processor at runtime.
> > >
> > As I'd like this fixed ASAP to get the buildds back in working
> > condition, I'll be preparing a NMU with --disable-fat. Please yell if
> > you think I shouldn't upload it.
> Could you confirm that it fixes the issue? Anyway, I see no problem
> with that.
Just re-tested, on a kvm guest, trying to build the hello package:
- fresh sid build chroot, gcc aborts
- with a local no-change rebuild of libgmp10, gcc aborts
- with a --disable-fat libgmp10, build works fine
NMU diff below.
diff -u gmp-5.0.5+dfsg/debian/changelog gmp-5.0.5+dfsg/debian/changelog
@@ -1,3 +1,11 @@
+gmp (2:5.0.5+dfsg-1.1) unstable; urgency=low
+ * Non-maintainer upload.
+ * Revert to --disable-fat, as it seems to cause gcc to abort on some
+ machines (closes: #671866)
+ -- Julien Cristau <jcristau at debian.org> Tue, 08 May 2012 19:24:03 +0200
gmp (2:5.0.5+dfsg-1) unstable; urgency=low
* New upstream.
diff -u gmp-5.0.5+dfsg/debian/rules gmp-5.0.5+dfsg/debian/rules
@@ -44,7 +44,7 @@
lib64dir = usr/lib64
-confflags = --prefix=/usr --enable-cxx --enable-fat
+confflags = --prefix=/usr --enable-cxx --disable-fat
ifeq ($(DEB_BUILD_GNU_TYPE), $(DEB_HOST_GNU_TYPE))
AR = ar
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 836 bytes
Desc: Digital signature
More information about the debian-science-maintainers