Jason Grout wrote:
On 04/01/2010 01:56 AM, William Stein wrote:
Sage didn't get to where it is now and won't get to where it needs to
go by such an attitude of not supporting platforms. If anything, we
need to solidly support far more platforms than we currently support.
William, what is your definition of the statement, for example, that
Sage officially supports Suse?
That's obviously for William to answer.
It sounds like David's definition is that every release of Sage is
tested on Suse before release by at least compiling it and running all
doctests without errors (David, correct me if I'm misinterpreting your
views!). How does your definition differ, if it does?
Jason, you are not mis-representing my views. That would be my close to my
definition, though I would suggest it is modified a little, to be more precise.
We state the version of openSUSE we support, not just 'openSUSE'. Ideally that
would be the latest stable release, but not necessarily so. I would not suggest
holding up a release of Sage just because SUSE have released a new version, and
nobody has had the time to set up that installation. Neither would I remove SUSE
support if Sage does not build on the latest release, but builds on an earlier
version.
Of course, if we chose to support older and never versions of openSUSE, that is
better, as long as Sage is tested on them.
Since in the case of openSUSE, there is official Sage support for both 32-bit
and 64-bit, then it should be checked on both 32-bit and 64-bit installations.
We should also state the GCC version we support for each platform. That
statement is clarified by saying "Any gcc >=4.0.1 should work, but Sage is
tested by building with the gcc version stated".
Of course, the fact a platform is not officially supported does not mean that
Sage will not work on it, but simply that we don't have the resources to test
every Sage release on it.
I believe the 'supported' definition used by MATLAB is reasonable
http://www.mathworks.com/support/sysreq/current_release/linux.html
(not necessarily the same operating systems or versions, but just the general
principle).
===============================================================
===============================================================
* Ubuntu 8.04, 8.10, 9.04, and 9.10
* Red Hat Enterprise Linux 5.x
* SUSE Linux Enterprise Desktop 11.x
* Debian 5.x
* The listed distributions are those Linux distributions that MathWorks products
have been validated against. It is likely that other distributions with kernel
version 2.6 or later and glibc version 2.5 or later can successfully run
MathWorks products, but The MathWorks will be in a limited position to provide
technical support for those distributions.
===============================================================
===============================================================
By adopting something similar to this, I see four advantages:
1) We ensure releases are not made which are broken on so-called supported
platforms.
2) Sage releases are not held up forever, because nobody is able to verify Sage
works on them.
3) If people want to set up a Sage server for access by students, then can be
reasonably certain that new releases of Sage will work on that operating system
version.
4) If people get problems building Sage on one of the supported platforms, we
can indicate to them that they try a specific version of GCC, as that is known
to work. If they still report problems, we can point out that Sage has been
verified to work on that platform, and it is therefore likely to be a problem
with their installation.
As and when the resources become available to test on more platforms, we should
do so. If it is no longer practical to test on a certain version of the
operating system, then official support should be dropped for that.
The main limit on testing Sage would usually be peoples time. In the case of
Itanium, it might be limited by the hardware resources available, as few people
have access to machines with Itanium processors. If that is the case, then we
should not consider Itanium as a supported platform.
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
URL: http://www.sagemath.org
To unsubscribe, reply using "remove me" as the subject.
