On 06/20/2012 11:05 AM, kcrisman wrote:
On Tuesday, June 19, 2012 10:21:31 PM UTC-4, Stephen Montgomery-Smith wrote:
On 06/19/2012 07:29 PM, kcrisman wrote:
>
>
> On Tuesday, June 19, 2012 5:28:29 PM UTC-4, Stephen
Montgomery-Smith wrote:
>
> When I run "sage testall" for sage-5.1.beta5 on FreeBSD/amd64, I get
> the
>
>
> Does this imply you get no other failures? Even close to that
would be
> great news.
I have four other errors, which are all floating point errors, and
it is
debatable whether they are failures.
> I'm curious how many of the patches in tickets in the FreeBSD
component
> [1] you actually need with 5.1.beta5 in order for Sage to be a-ok.
I did not use #5873, #7831, #9606, #12059. All the others I used.
Can I just confirm that this is with the *Sage* versions of these
packages (R, numpy, matplotlib), and not system versions? We'd like Sage
to build from scratch if possible.
The only thing I used not from sage was atlas. I don't even have the
FreeBSD version of R or matplotlib installed, so those couldn't even be
used by mistake.
Same for the patches/spkgs for the other tickets. Some (e.g. the cephes
one) will need some work just to use patch instead of cp, check for exit
codes, etc., but maybe this is closer to being merged now.
Also, which version of FreeBSD are we talking? Not that we have to
support older versions per se, but if it does end up being "supported"
then we'd like to warn users which one - some of the relevant tickets
mention that this would make a difference.
This was built on FreeBSD-8.3-stable/amd64. I have tried building it on
http://www.redports.org/ and I can confirm it built and tested properly
on FreeBSD-9/amd64.
Some tests fail for FreeBSD/i386. I plan to look into these at a later
date. My guess is that they are issues with cython.
I used a work around for #9606, which I will describe below.
For this reason, I created my own versions of cc, c++, gfortran, etc,
and sh, in $SAGE_ROOT/bin. sh was merely a link to /usr/local/bin/bash
(hence no need for #9606). For cc, etc, I placed a script something
like:
Did you try doing sage -sh with your built version without making this
link? #9606 suggests that it's not an issue any more, based on the new
code in spkg/bin/sage.
I am now trying to build sage without the link of sh to bash.
But I don't see how to get around the link for gcc, etc. Also we will
need a link of make to gmake, since bsdmake which FreeBSD uses by
default is highly incompatible with gnu-make.
Finally, anyone who has FreeBSD with an up-to-date set of ports can try
out sage. Currently it is at sage-5.0.1. Just do
cd /usr/ports/math/sage
make && make test
--
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
