Oracle broken on 1.6

106 views
Skip to first unread message

Shai Berger

unread,
Oct 4, 2013, 11:43:55 AM10/4/13
to django-d...@googlegroups.com
Hi,

It looks like the fix to #21165[1] broke the tests on Oracle (seems like the
behavior with respect to timezone information changed) in 1.6.X. I just saw
the Jenkins report[2], I hope to look at it later tonight.

Thanks,
Shai.

[1]https://github.com/django/django/commit/5252885494079cf28a337644a87e61b19340f09c
[2]http://ci.djangoproject.com/job/Django%20Oracle/246/

Ramiro Morales

unread,
Oct 4, 2013, 12:39:51 PM10/4/13
to django-d...@googlegroups.com
Shai,
Oops, I'm the committer of that one.

That particular commit only touches a test unrelated to the ones failing (and makes it use some test decorator infrastructure added [1]recently to solve failures on Windows).

Plus, the additions made don't look like something that would leak some kind of state that could explain the observed failures.

From what I can gather from the Jenkins UI most of the last Oracle 1.6.x builds have been extremely unstable, some of them even didn't start. I suspect the failures started with some older commit but only appeared now.

But I'm also aware anything is possible in the Land of the Oracle when the magic of Time Zone Spells are involved, so don't quote me on the above affirmation.

Please feel free to contact me via IRC (nick: cramm) if you think I can be of any help debugging and solving these issues.

--
Ramiro Morales
@ramiromorales

1. https://github.com/django/django/commit/5b97b99a014443b255cf8ab0467864c5874027da

Shai Berger

unread,
Oct 5, 2013, 4:43:09 AM10/5/13
to django-d...@googlegroups.com
Hi,

On Friday 04 October 2013 19:39:51 Ramiro Morales wrote:
>
> On Fri, Oct 4, 2013 at 12:43 PM, Shai Berger <sh...@platonix.com> wrote:
> > Hi,
> >
> > It looks like the fix to #21165[1] broke the tests on Oracle (seems like
> > the behavior with respect to timezone information changed) in 1.6.X. I
> > just saw the Jenkins report[2], I hope to look at it later tonight.
> >

And now it seems the operative phrase there was "looks like". I had some time
to look at the relevant commits, and indeed none of them seem to be able to
cause that breakage, and on my own test environment, everything passes.

My current guess is that something broke the Jenkins test environment, I'll
look more into it later.

Sorry for the noise,

Shai.
Reply all
Reply to author
Forward
0 new messages