Re: Question about fixing #512075
On Sun, Jan 25, 2009 at 10:18:19PM +0100, Aurelien Jarno wrote:
> On Sun, Jan 25, 2009 at 01:22:19PM +0100, Adeodato Simó wrote:
> * Thomas Weber [Sat, 24 Jan 2009 22:08:31 +0100]:
>
> > I've patched 1.0.6 instead, it turned out to be only a one-line change.
> > Sadly, the package now FTBFS on ARM only,
> > /usr/bin/ld: cannot find -lgfortranbegin
>
> > Sigh, back to square one.
>
> Okay, CC'ing the arm buildd maintainers to see if the recognice
> something from the log:
>
> http://buildd.debian.org/~luk/status/package.php?p=octave-symbolic#fail-octave-symbolic-arm
>
> I am able to reproduce the problem locally. The problem is that the
> build system is using -L/usr/lib/gcc/arm-linux-gnu/4.3.1 while this
> directory doesn't exists.
>
> The directory is also wrong on other systems, but as g++ and thus
> g++-4.3 is used (g++-4.1 is used on arm), the directory containing
> -lgfortranbegin is already in the include path.
>
> I don't see any reason to compile this package with g++-4.1 on arm
> anymore (the ICE have been fixed a long time ago), so the best
> solution is probably to build with the default compiler on arm, like it
> is done on armel.
This might require a rebuild of Octave and every arch package depending
on it (the gcc path is hardwired in the mkoctfile script used for
building these addons), but see below.
The reason for forcing gcc to 4.1 was
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=467503
(This crashed Octave itself on ARM).
If this is fixed in gcc-4.3, maybe that bug report should be closed, or
the information added?
Anyway, what's debian-release's opinion on the package? I'll investigate
the possibility of overriding the /usr/bin/gcc- stuff for this one
package.
Thomas
Reply to: