On Feb 13, 3:09 pm, "Dr. David Kirkby" <david.kir...@onetel.net>
wrote:
> mabshoff wrote:
Hi David,
> > Once more for the record: Sun is *not* sponsoring the Solaris port of
> > Sage, the DoD does that.
>
> Sorry, I mis-read William's presentation
> "Funding from Microsoft, UW, NSF, DoD, Google, Sun, private donors etc".
Yes, but Sun gave Sage the new hardware at a rather excellent price.
> It's good to see somebody (DoD)is funding the port. I think there could
> be a lot of interest in this from Solaris people, who tend to often come
> from scientific backgrounds.
I can see that, too. There is certainly still enough large Sparc SMP
hardware around since if you need a large SMP box you can either pick
Itanium, Sparc or Power6 and when I mean large SMP I am taking 0.5+ TB
RAM or more, so 8 sockets won't cut it :)
> > Sun has recently offered some resources like
> > access to large machines to help out which is very welcome.
>
> Yes, I'm sure it is useful. There is some serious looking hardware in
> that Sage presentation on the Sun web site.
Ironically it is running Linux :)
> > <SNIP>
>
> > By the way: Starting from 3.3 on we will release beta binaries for 32
> > bit Solaris 10 on Sparc as well as x86 with a build in toolchain. 64
> > bit support isn't far out and once I am in Seattle after SD13 at the
> > end of the months my goal will be to get all the remaining doctests
> > that are broken fixed. By Sage 4.0 (probably out around SD 15 in May)
> > we will add full Tier 1 support for 32 and 64 bit Solaris on Sparc as
> > well as Intel CPUs.
>
> Does the source compile with reasonable ease on Solaris as 32-bit now?
Yes, but I should post some details.
> Last time I looked at the source, which was several months back, there
> were several issues, which you knew how to fix, but had not had time to
> incorporate into the source. Rather than keep applying the fixes with
> each new version, I thought I'd wait until they were incorporated. I
> don't mind testing and reporting failures, but if there are many known
> issues, and known solutions which have not been incorporated, I'd rather
> wait until they are.
All the fixes are in and I ended up fixing a gcc 4.3 specific bug you
ran into in Matplotlib. That one was quite strange to say the least.
One problem on Solaris at the moment is that Solaris' default "which"
does not return a non-zero return code when $FOO isn't found, so all
the detection at the start of the build phase does not work and things
do easily blow up later. I have a patch for that, but it isn't clean
yet.
I do build the following from sources, i.e. no funny g prefix like the
Solaris versions:
* gcc 4.3.2 build with GNU as and GNU ld (switching to 4.3.3 soon).
gcc 4.2.x is *broken*, i.e. numpy will segfault when compiled with it
on Solaris
* binutils 2.18 or higher
* gnu make
There a more needed bits at the moment if you want Sage to doctest
well and also do stuff like -bdist since there are still GNUisms in
some shell scripts. Overall look at
http://wiki.sagemath.org/solaris/toolchain
I have precompiled toolchains I use for x86 and Sparc Solaris 10 in
http://sage.math.washington.edu/home/mabshoff/solaris-binaries/
The current 3.3.alpha6 status can be seen at
http://wiki.sagemath.org/solaris/sage-3.3
Note that two important patches to that are mandatory to make doctests
pass are not on that page yet, but they should be after 3.3.rc1. The
older 3.2.3 Solaris porting page has the fixes fixed in with other now
merged fixes, but they are hard to extract unless you know what you
are looking for. As I mentioned there will be 3.3 binaries with all
Solaris specific fixes not yet in standard Sage with integrated
toolchain in one tar.gz once the 3.3 release is done.
Cheers,
Michael
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@googlegroups.com
To unsubscribe from this group, send email to
sage-devel-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
