-
-
Notifications
You must be signed in to change notification settings - Fork 18.5k
TST: dateutil testing failures on windows #7337
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
@dbew lmk what debug info I can provide |
@jreback I'll look into this but we don't have python on windows here. I might have to try at home. These failures
look like a difference in naming dateutil timezones between windows and linux. e.g. in linux
so As a start there, please can you run the following and paste the output?
The others, I'm not sure about right now. |
|
Hmm. On Linux, I get
It's a shame that the name changes on windows - you get The code that's failing is in
I think we'd be fine changing that last line to
and the tests pass on linux for me. However, we rely on being able to round trip from
which boils down to:
I've branched the head of pydata/pandas here and made this change. It works locally but I can't test on windows until I get home. I'll dig into the other errors when I get python installed under windows. |
i'll test a bit later, you might find this a good ref for installing on windows: https://github.com/pydata/pandas/tree/master/scripts/windows_builder |
@dbew ok that seemed to fix all the tests except one, and this one is trivial: https://github.com/pydata/pandas/blob/master/pandas/tools/tests/test_pivot.py#L500 when testing, use so pls make that changed and do a PR and should be good to go |
@dbew I am going to merge your fix...no need to do a PR |
ok, great. That test isn't in a file I changed in the original PR - strange that it's failing now. Looking at the code, I can't see how that's related to the dateutil stuff either. Could it be related to some of the other changes going in recently? I've done a pull request for the first part, here |
oops, we commented at the same time. I'll cancel the PR. Thanks for merging the fix. |
merged via 6629446 |
Thank you. |
seems 26-32 is failing (i'll post if other builds are as well), with a lot less failures though. |
I'll fix this last one
|
to make this even more complicated. looks like and only on windows (as I think the 2.6 build DOES run with 1.5 on travis). |
happens on 33-32/33-64 with dateutil 2.2, so this is a windows issue then (and not 2.6/dateutil) |
These ones
are due to a bug in dateutli on windows. You can get the
It looks like the correct UTC timezone in dateutil on windows is The other failures are slightly different, I don't yet understand why they're a problem on windows but not linux. I'm going to look into these a bit further. |
can o u fix |
I think this is just a test issue. The way In This means that it's very unlikely that I'm looking at fixing the tests up now. Was it just the python 26/ win32 build that failed on these tests? |
no....fails on 3.2/3.3 as well (but NOT 2.7), odd |
That's odd - because I can see the failures in py27/win64 under anaconda. Anyway, I've dug in a bit more and I think using It's also consistent with the test in I'll update the tests and if it looks good, I'll update the docs to show how to construct the |
@dbew it could me just reading the setup... I have a window on my pc at home which shows all the runs (kind of like travis) and just prints out failures; its possible those tests are not being run for some reason. go ahead and submit a PR and I can give a test tongiht |
Ok. It looks like the change to tzutc is working. I'll update the docs and submit a PR as soon as I can. |
gr8! glad you got it working on windows! |
cc @dbew
(I am just posting 27-64, but failing on most windows versions). prob the same error.
I think has to do with 32-bit ints (which is the default on windows even for 64-bit versions), being treated as 64-bit.
The text was updated successfully, but these errors were encountered: