Conversation
This contains the fix for the one failing test. There are likely many more things that broke.
|
@RosalynLP @voisey Pandas loves to break the user interface in all sorts of weird ways (and sometimes even change semantics so that you get wrong result). I have fixed the instance that is failing in the tests, but there are other things like in particular the theory covariance stuff that may be broken now. Unfortunately we also don't have tests for those (and the corresponding PR was closed). Could you please try to run some representative reports and see that they work? |
|
Updating pandas scares me |
Will do |
This should avoid travis timeouts and also help identifying failures.
|
It seems this still times out even if you download the resources first (and add a lot of noise to the logs). cc @wilsonmr . |
|
ok so I added some more prints and ran conda build changing the test script to have
and it at least appears that it's getting stuck in |
|
Is this ready? as per #697 I think there is a bit of urgency because at the moment it's a bit of a mess for people trying to transition to new reportengine |
|
I am going to check some theorycov reports now to see if they still work |
|
Should we merge this then? |
|
Yes, please. |
|
I wonder if afterwards we should send email suggesting people update environments/create new ones? Seeing as nothing appears to be failing seems ok to merge |
Shouldn't environments work fine automatically? When you update nnpdf, both pandas and reportengine should be updated as well. And if you don't update everything that used to work should keep working... |


This contains the fix for the one failing test. There are likely many
more things that broke.
Closes #690.
Closes #697.