Le 05/11/2011 14:31, Jeroen Demeyer a écrit :
On 2011-11-04 08:19, Julien Puydt wrote:
but rather a more deliberately fragile one :
- check very-specific-arch1, and set magic options if so ;
...
- check very-specific-archN, and set magic options if so ;
- if we aren't on a known arch, or a fly farted too near a known one :
FAIL.
This is *so* true. The problem here is that most Sage developers simply
don't know about these issues and simply don't know or don't care about
how to solve this.
Eh... but as mathematicians, they know the value of having a result
whose hypotheses are somehow minimal, and that is the very same
principle, so I think there is hope here.
I'm more hopeful about upstream than you are : the problem I mentioned
is more there because they're not conscious of the issue, rather than
because they don't want to fix it -- or too stupid to get it.
Another problem on the way to cross-compilation is that some packages
want to run tests at compile-time to adjust settings : those tests will
have to be disabled when cross-compiling.
I think it's totally unrealistic to think that Sage will ever work under
cross-compilers. There are so many spkgs which are written without
giving thought to cross-compiling. So you would need to fix every such
spkg (or convince upstream to do it, which is even harder).
Why do you think I don't give a date for the end of the work? ;-)
Still, I think things aren't that dark :
- sage will probably evolve and stop trying to recompile what is almost
a full distribution : it will use what's already on the system ;
- what will be left will be much smaller ;
As an example, ekiga uses already-built win32 libs for most of its deps
(glib, gtk+, opal, ptlib, etc), so we only had to make sure our own
codebase was right -- not the full tree of deps.
Snark on #sagemath
(yes, last time I wrote #sage-devel, mistaking the mailing-list and the
channel)
--
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org
