Re: [Pkg-octave-devel] broken m68k build

["Christian T. Steigies" <cts@debian.org>]

On Sat, Feb 26, 2005 at 12:51:42AM +0100, Richard Zidlicky wrote:
> On Fri, Feb 25, 2005 at 09:15:18PM +0100, Rafael Laboissiere wrote:
> > [Cc: to the Debian m68k crew and debian-admin]
> > 
> > * John W. Eaton <jwe@bevo.che.wisc.edu> [2005-02-25 14:34]:
> > 
> > > For some time now, the m68k build of octave2.1 has been failing with
> > > 
> > >   /usr/bin/g++ -c  -fPIC -I. -I.. -I../liboctave -I../src -I../libcruft/misc  -I../glob -I../glob -DHAVE_CONFIG_H  -Wall -W -Wshadow -O1 -g0 ls-mat5.cc -o pic/ls-mat5.o
> > >   /tmp/cchqGuhc.s: Assembler messages:
> > >   /tmp/cchqGuhc.s:20898: Error: operands mismatch -- statement `move.b %a4,%d1' ignored
> > > 
> > > I know that the proper fix for this problem is to fix the bug in the
> > > compiler that is generating the bad code, but I'd also be willing to
> > > take a look at the problem and see if I can come up with a
> > > workaround.  To do that, I'd need ssh access to an m68k system with
> > > all the current build tools.
> 
> I guess I already have the fixes for the compiler, can you see if
> the appended patches make any difference? Alternatively, send me
> .i file and compile flags to test.

You mean recompile gcc, then recompile octave? That would take a long while...

It is a regression from g++3.3.3, since that file compiles with that version
of g++, see:
http://people.debian.org/~cts/octave2.1/

Your ii files should also be there, if that helps.

I started to build octave2.1-2.1.66 outside the chroot, with the older g++,
maybe it works. It has been picked up by another buildd already, maybe that
build succeeds as well, but I doubt it.

Christian