On 11 Aug, 22:50, William Stein <wst...@gmail.com> wrote:
>
> It's actually interesting to summarize, the specific constraints I am aware 
> of:
>
>    1. A major government agency -- we can't use Sage unless you
> provide a version that contains no binary components and that builds
> from source using the latest released version of GCC.

>  -- William

Currently the default compiler people will get on 't2' is gcc 4.2.4,
not the latest which is 4.4.1, although I have installed 4.4.1 on
there, and it works ok. The reason I changed from what was 4.4.0 to
4.2.4 is that 4.2.4 did cause 20 test failures in mpfr, which Sun
admits is a bug in memset().

I've since updated the mpfr.spkg file, so that on sun4v machines, a
patch developed by one of the mpfr developers gets applied. (For
performance reasons, I only did this on sun4v). As such, there is no
reason to use 4.2.4 any more. I could make the default the latest
4.4.1, so we can ensure Sage is building with the latest gcc.

Given we have settled on a minimum compiler version of 4.0.1, there
might be some benefit in making the default compiler 4.0.1, so we
become aware of any packages which make use of code that will not work
on 4.0.1.

I don't like changing the default compiler too much, as it is likly to
break any builds people have in progress. Given 't2' is down (even the
LOM), it might be a good time to change it, as 4.2.4 is

* Not necessary any more to get the MPFR tests to pass.
* Is not the latest gcc
* Is not the oldest gcc we expect Sage to build on.

One thing to note is that I'm using 4.4.0 at home, so I tend to test
Sage on almost the latest gcc, and could easily change that to the
very latest version of gcc.

So the default compiler on t2 could be:

* 4.4.1 (the  latest)
* 4.0.1 (the oldest Sage is expected to build with)
* Leave it at 4.2.4, which is roughly halfway between the oldest and
latest versions.

Any thoughts?

Dave




--~--~---------~--~----~------------~-------~--~----~
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
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---

Reply via email to