Uh oh!
There was an error while loading. Please reload this page.
MNT add CI jobs to run integration tests for downstream projects: loky, joblib, distributed - #236
Conversation
Codecov Report
@@ Coverage Diff @@## master #236 +/- ##
=======================================
Coverage 85.12% 85.12% =======================================
Files 1 1 Lines 585 585 Branches 117 117 =======================================
Hits 498 498 Misses 63 63 Partials 24 24Continue to review full report at Codecov.
|
ogrisel
left a comment
There was a problem hiding this comment.
Thanks for tackling this. Here are some comments.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
pierreglaser
commented
Jan 25, 2019
Mmh. So the good news is that there does not seem to be any syntax error. Also, |
Uh oh!
There was an error while loading. Please reload this page.
this seems to fix test_duplicate_clients
pierreglaser
commented
Jan 28, 2019
lesteve
commented
Jan 28, 2019
No suggestion off the top of my head, sorry ... |
There was a problem hiding this comment.
Instead of running those integration tests as cron jobs, I am thinking we could run them conditionally whenever the commit message has a special flag. For instance, for the joblib entry:
commit_message =~ /(ci-downstream|ci-joblib)/https://docs.travis-ci.com/user/conditions-v1
WDYT?
pierreglaser
commented
Jan 29, 2019
This sounds great to tackle issues early. But it will be harder for us to spot a breaking change if we miss it initially. We would at least have to push empty commits with |
ogrisel
commented
Jan 29, 2019
Yes but we can do that in a PR that we do not merge to avoid polluting master with empty commits. |
ogrisel
commented
Jan 29, 2019
So be it. It's a tradeoff between long CI jobs, unwanted noisy notifications (in case the downstream master branches are failing randomly for causes unrelated to cloudpickle) and thinking about doing manual triggers prior to making a cloudpickle release. |
pierreglaser
commented
Jan 29, 2019
OK. I'll change the trigger and make a few test commits inside the PR with the appropriate flags. |
64a122b to
a3c86bbCompare
ogrisel
left a comment
There was a problem hiding this comment.
Actually, changing my mind on the commit message tagging syntax:
Uh oh!
There was an error while loading. Please reload this page.
9afceb6 to
f1c2196Comparef1c2196 to
cacd7a0Compareogrisel
commented
Jan 29, 2019
It seems to work. Can you just try with |
ogrisel
commented
Jan 29, 2019
Did you cancel the |
pierreglaser
commented
Jan 29, 2019
Nope, actually I thought you did. Might be because of auto cancelling builds when more recent commits are pushed? |
pierreglaser
commented
Jan 29, 2019
Besides that the filter seems to work indeed. |
ogrisel
commented
Jan 29, 2019
Good, let's wait for the final build to complete. |
Merged. Thanks @pierreglaser! |
This PR configures cron-jobs that build and launch the test suite of downtream projects that rely on
cloudpickle. The main purpose is to find out quickly if we introduce breaking changes.Ideally we would rather test the stable (PyPI) rather than the latest (master branch) version, but each time, a project-specific issue was making it at least clumsy, at worst impossible. (I can detail upon request).