On 06/21/10 10:09 PM, Robert Bradshaw wrote:
On Jun 21, 2010, at 2:03 PM, Dr. David Kirkby wrote:
It seem strange we are only checking about 20% of the packages, even
if SAGE_CHECK is yes.
Yes, I guess you could say we trust the upstream testing too much.
But many of the upstream packages have no test suite.
But failures can be caused my many things.
* Compiler bugs
* Operating systems bugs
* Operating systems which are mis-configured.
* An almost endless list of other things.
So actually testing the packages is a good idea. Clearly the doc tests do not
pick up many possible issues in Sage.
If the python results are anything to go by, 3 systems give 3 different set of
failures.
1) John's Mac (OS X I assume)
test_asynchat test_ctypes test_distutils test_locale test_smtplib
2) My Blade 1000 (Solaris 10 SPARC)
test_distutils test_float test_hotshot test_multiprocessing test_sunaudiodev
3) My Ultra 27 (OpenSolaris on x64)
test_distutils test_float test_multiprocessing test_posix test_zlib
The only common one is "test_distutils" which has failed on all 3 tested
systems.
I found the python testing quite useful, as there was a bug
http://trac.sagemath.org/sage_trac/ticket/8440
which was preventing '_multiprocessing' from building on Solaris. I worked
around that, so now '_multiprocessing' builds, but I note
'test_multiprocessing' fails the test suite on both Solaris machines.
So perhaps my "fix" was not as good as I had hoped! That looks like it will be
something that needs reporting upstream.
I'll
add an spkg-check to Cython the next time I update that package.
Good.
- Robert
MPIR, MPFR and now Python (once #9295 is committed) are all covered. I'd like to
know what else is pretty critical. I guess Cython must one be one such package,
but what else?
I'm 99% sure if we run the Maxima test suite there will be several failures due
to numerical errors, but I don't know how to run that from the command line
(i.e. spkg-check). I think you need to fire up maxima first.
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
