On Sun, Aug 9, 2009 at 12:12 AM, Ondrej Certik<[email protected]> wrote:
> On Sun, Aug 9, 2009 at 12:01 AM, Ondrej Certik<[email protected]> wrote:
>> On Fri, Aug 7, 2009 at 12:19 PM, Ondrej Certik<[email protected]> wrote:
>>> On Thu, Aug 6, 2009 at 12:20 PM, Vinzent
>>> Steinberg<[email protected]> wrote:
>>>>
>>>> http://buildbot.sympy.org/builders/amd64-py2.5-sympy-tests/builds/287/steps/shell/logs/stdio
>>>>
>>>> Could someone with access to an amd64 machine please debug the cause
>>>> of this?
>>>
>>> I don't have time to fix this currently. In the worst case we'll just
>>> revert the patches, if you don't know how to fix this. My amd64 boxes
>>> are now inaccessible, but once we fix it, I'll give you an account, so
>>> that you can debug this.
>>
>> There are several failures, but I just took one and after a bit of
>> debugging, this is what fails for me in the current sympy:
>>
>> $ bin/test sympy/mpmath/tests/test_hp.py
>> sympy/mpmath/tests/test_matrices_mpmath.py
>> ============================= test process starts
>> ==============================
>> executable: /usr/bin/python (2.6.2-final-0)
>>
>> sympy/mpmath/tests/test_hp.py[3] ...
>> [OK]
>> sympy/mpmath/tests/test_matrices_mpmath.py[8] ...F....
>> [FAIL]
>>
>> ________________________________________________________________________________
>> _______ sympy/mpmath/tests/test_matrices_mpmath.py:test_matrix_creation
>> ________
>> File "/home/ondrej/repos/sympy/sympy/mpmath/tests/test_matrices_mpmath.py",
>> line 98, in test_matrix_creation
>> [one/3, one/4, one/5]])
>> AssertionError
>>
>>
>> However, this works:
>>
>> $ bin/test sympy/mpmath/tests/test_matrices_mpmath.py
>> ============================= test process starts
>> ==============================
>> executable: /usr/bin/python (2.6.2-final-0)
>>
>> sympy/mpmath/tests/test_matrices_mpmath.py[8] ........
>> [OK]
>>
>> =================== tests finished: 8 passed in 0.07 seconds
>> ===================
>>
>>
>> So mpmath or sympy is doing something that it shouldn't have been doing.
>
>
> A quick bisect shows that this commit broke it:
>
> $ git bisect gooddc6736a56da7c054a7fc00e486b258bd73bade24 is first bad commit
> commit dc6736a56da7c054a7fc00e486b258bd73bade24
> Author: Fredrik Johansson <[email protected]>
> Date: Fri Feb 6 14:50:44 2009 +0100
>
> Update sympy.mpmath to mpmath-svn
>
> * A few fixes in evalf.py were needed. In particular, I simplified the
> extrapolating
> numerical summation code to just call mpmath.nsum instead. It is
> a bit slower,
> but works better.
> * statistics.Normal now uses mpmath.erfinv instead of the ad-hoc
> implementation
> * A few other small interface incompatibilities in mpmath were adjusted for
>
> All tests pass, but I could have broken some imports etc that cause it to
> pass only by accident on my machine.
>
> The new mpmath docs are not included in this patch.
>
> Also, no particular support has been added for any of the several
> new functions
> in mpmath.
>
> Signed-off-by: Ondrej Certik <[email protected]>
>
> :040000 040000 e028279435be3c2c1206b5f95211d0b5145ffe2f
> 891d34833ad510bafb7c0da77d109eb378bb8149 M sympy
>
>
>
> Fredrik, any ideas how to get this fixed? Can you reproduce it?
>
> It was broken from this commit, it's just that the tests were never
> run in this order. The order was reversed by this patch:
>
> 334b0a5 (test: run all tests containing any of the given strings in
> their path, 2009-07-27)
>
> I don't currently understand why, so I am looking into this and try to
> fix it. Nevertheless, the above thing is still a bug.
I have figured that out -- if you use get_paths(), you need to sort
the list after that, just like at other places in runtests.py. So this
simple patch fixes it:
diff --git a/sympy/utilities/runtests.py b/sympy/utilities/runtests.py
index 66ba79f..92e1eeb 100644
--- a/sympy/utilities/runtests.py
+++ b/sympy/utilities/runtests.py
@@ -89,6 +89,8 @@ def test(*args, **kwargs):
mypaths = []
for p in t.get_paths(dir='sympy'):
mypaths.extend(glob(p))
+ mypaths = list(set(mypaths))
+ mypaths.sort()
t.add_paths([p for p in mypaths if any(a in p for a in args)])
return t.test()
This is a really critical thing, since the tests were failing on our
buildbots, so I pushed that in. Now all the tests pass again on
buildbots, so we don't have to revert any patches.
Nevertheless, the above bug in mpmath (and apparently also in nseries,
see the buildbot logs) stays.
What we could do, if anyone has time to implement that, is to test the
testsuite in a random order, it would probably reveal couple more
similar bugs.
Ondrej
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"sympy" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to [email protected]
For more options, visit this group at http://groups.google.com/group/sympy?hl=en
-~----------~----~----~----~------~----~------~--~---
