Re: octave2.9 ftbfs during linking stage
On 2007-06-28 11:16 +0200, Thomas Weber wrote:
> Am Donnerstag, 28. Juni 2007 10:55:07 schrieb Wookey:
> > On 2007-06-26 17:54 +0200, Thomas Weber wrote:
> > There have been issues with the gcc4.2 toolchain not building, which I
> > understand are now fixed, but octave seems to be using gcc3.4 and I'm
> > not aware of any recent linking problems there.
> No. It uses g77 for Fortan (thus the 3.4 stuff), but normal gcc otherwise
> (which means gcc 4.2 currently).
OK. (BTW for the new armel port all g77 stuff will be moved to
gfortran, because gcc3.4/g77 doesn't support arm EABI, and isn't going
to. I don't know if this will 'just work' for octave or not? I mention
this here just because it seemed a good moment).
> > It looks like the octave link is quite big, which may be tickling some
> > kind of bug, or possibly just running out of memory on the build
> > machine. That machine is a 256MB netwinder. Is octave lots of C++?
> Yes. The part that's not C++ is Fortran.
> > That can really make such a machine struggle (although it's less of a
> > problem with gcc3 than gcc4).
> Is there a simple workaround, ideally applicable to all architectures?
The simplest is 'buildds with more than 256MB ram'. There is a 512MB
arm buildd commissioned and waiting to be added to the buildd
rotation. There is a DSA ticket for this and after discussions at
debconf I expect it to happen in the next couple of weeks.
You can send in a request to email@example.com to the effect that
octave will only be built on machines with more memory.
However we don't really know if this is actually the problem. It built
OK for etch with gcc 4.1. Someone doing a manual build to confirm that
it is OK would be best before asking for admin changes.
As I say I should be able to do this very soon, but not right now as
my good machines are in bits.
Principal hats: Balloonz - Toby Churchill - Aleph One - Debian