[SciPy-User] test issues with 0.10

[Pauli Virtanen]

14.11.2011 21:17, Paul Anton Letnes kirjoitti:
[clip]
>> What was the last version of scipy that did work for you?
> 
> 0.10.b2 was working fine, although not all tests passed.

Would be great if you can check which commit between v0.10.0b2 and
v0.10.0 introduces the issue.

However, I'm suspecting that it's a memory corruption issue, which can
mean that the commit found to unveil the issue has nothing to do with
the problem itself.

(Git can help you with finding the faulty commit:
http://book.git-scm.com/5_finding_issues_-_git_bisect.html

git bisect start
git bisect good v0.10.0b2
git bisect bad v0.10.0
...)

> I figured it would be ironed out before release. (I did not
> use any of the problematic sub-modules anyway.)

Well, those ARPACK failures which we could reproduce were addressed.

I'm a bit surprised to see that you still get 'f' mode failures in the
ARPACK tests --- it switches to double precision on OSX, so the
corresponding 'd' mode tests should fail, too. I'd double check this
kind of results by a completely clean build, i.e., remove the scipy
installation and the build directory, and rebuild.

-- 
Pauli Virtanen