On 09/20/10 05:04 PM, John H Palmieri wrote:
Should we change the spkg-check file for the python spkg? Are there
*any* machines on which it passes?
When I asked people to test, not a single person got them all pass.
I've tried vanilla python-2.6.6
and python-2.7, and while I can get python-2.7 to pass self-tests on
one machine (sage.math), it fails on every other machine I've tried: a
Mac OS X 10.6 machine, t2.math, and a handful of skynet machines. I
think python-2.6.6 fails on all of these, including sage.math.
It would be nice if you could do
$ SAGE_CHECK=yes
$ export SAGE_CHECK
$ make
and have some chance of Sage building successfully, but the python
spkg prevents this from happening, and on many linux and OS X
machines, it may be the only obstacle. (R caused trouble for me on
Solaris, also.)
How about we have an environment variable
SAGE_SKIP_CHECK_python
SAGE_SKIP_CHECK_r
SAGE_SKIP_CHECK_atlas
or similar for any packages which are known to regularly fail? That way, one
could disable any test one wanted. There are probably a few we would want to -
Python being one. I think R might be another.
I would not suggest going around adding them to every package, but in the case
of Python at least, there does seem to be a good case for this. There may be
instances where the tests take a very long time, and you are not going to do the
same package every time.
IMHO, it would be good if we run all the maxima self-tests, though of course
some will fail on ECL.
So what should we do with the python spkg-check file? It would be too
drastic to ignore all failures. Of the many tests which python runs,
are there some whose failures we can safely ignore, and so we should
just skip them? For example, test_distutils fails on sage.math (in
2.6.6), and test_mailbox fails on the skynet machines; are these
failures acceptable?
It's hard to know. The best you can do is check the Python bug list, and if its
not reported, report the bug. That said, I've not had a lot of success at
getting Python bugs fixed.
(Along these lines, we might get some improvement by switching to
Python 2.7. Is anyone working on this?)
We might get some more headaches though. The 2.6.x series, where x>0, have all
been bug-fix-only releases. The 2.7 might add more features,
--
John
BTW, one other thing to note is that running the self-tests, and getting an exit
code of 0, does not in general mean the tests have all passed. Cliquer is one
package where this is so, but there are others too.
Dave
--
To post to this group, send an email to [email protected]
To unsubscribe from this group, send an email to
[email protected]
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org
