Python 3.6 support #2512
Python 3.6 support #2512
Conversation
|
@richafrank @ssanderson Friendly ping. 😇 |
|
This only fails because the retarded CI build process takes more than an hour. |
|
bump @richafrank @ssanderson |
|
Hey @Peque sorry for the (very) long response delay here. I've opened a general issue on some of the goals and challenges we've had around adding support for newer Python versions here: #2616. It looks like the status on this PR is that the Travis build is working (!), but that Appveyor was timing out. Any idea on what was causing the timeouts? |
|
@ssanderson Thanks for your reply. In my experience, AppVeyor is not-very-fast (compared to Travis), at least the free workers allocated to FOSS projects. That, plus the not-very-fast test suite in Zipline is what may be causing the timeouts. |
|
@ssanderson Probably just clicking on "rerun" will make the tests pass in AppVeyor too. |
|
Hmm. I clicked rerun and it seems like the appveyor build still died with a timeout after an hour. If I'm reading the build log in https://ci.appveyor.com/project/Quantopian/zipline/builds/30111660/job/ub9twt0bed0688wm?fullLog=true correctly, it looks like we're not even making it through the install, let alone running the tests. |
|
For comparison, a recent appveyor build on another branch just took about 19 minutes, so it seems like something's meaningfully different about the changes here. |
|
@ssanderson I'll have a quick look at it tomorrow. |
While testing examples, positions are not in the expected order.
Peque
force-pushed
the
python36
branch
2 times, most recently
from
191e839 to
13d3e80
Compare
|
@ssanderson My conclusion is: AppVeyor is selectively slow. 😂 Would you mind merging these commits into a branch in your own fork and open a new PR to see if AppVeyor finishes when you open the PR instead of me? See #2619. |
|
@ssanderson Friendly ping. 😊 |
|
hey @Peque. @richafrank and I looked at this for a bit this morning and couldn't see anything obvious that would cause your Appveyor builds to run slower than other ones. There have also been other builds from non-Q contributors that have succeeded in the interim (e.g., #2430), so I don't have a good working theory here yet. |
|
@ssanderson Have you tried fetching this branch from your fork and opening a PR from your account? Just out of curiosity. Maybe those other non-Q contributors had paying AppVeyor accounts too. 🤷♀️ |
|
Closing this after: #2643 |
Will take a look at Python 3.7 once this gets merged.