search

scikit-hep / uproot3

ROOT I/O in pure Python and NumPy.
BSD 3-Clause "New" or "Revised" License
314 stars 67 forks source link

Add support for Python 3.9 #526

Closed eduardo-rodrigues closed 3 years ago

eduardo-rodrigues commented 3 years ago

Final PR after similar PRs in uproot-methods and awkward-array for the uproot 3 suite to support Python 3.9, which is relevant as a LTS version.

eduardo-rodrigues commented 3 years ago

Random time-outs in Travis are responsible for the 1-2 failures. In other terms, this is actually ready to be merged IMO.

Looking forward to the subsequent release of the uproot 3 suite for Python 3.9 :-).

eduardo-rodrigues commented 3 years ago

@chrisburr, it may be better to revert to conda. There are issues with what you replaced with and, as said, things are in fact all OK in reality.

chrisburr commented 3 years ago

@eduardo-rodrigues They're not random failures. All of the failures were for the Python 2.7 job and retrying didn't help.

So far as I know all of the old mamba issues have been fixed so I'm interested in reporting anything that remains upstream. I'll give up and revert if it starts to become time consuming to fix.

eduardo-rodrigues commented 3 years ago

Understood, thanks.

chrisburr commented 3 years ago

@jpivarski It looks like my changes worked but I can revert them if you'd prefer?

FYI: Travis CI is shutting down .org at the end of the year and there is no longer a free tier on .com (I guess it doesn't really matter given uproot4 is mostly ready to take over).

chrisburr commented 3 years ago

Apparently you merged as I was writing that, I'll assume you're happy 😄

eduardo-rodrigues commented 3 years ago

Thanks for the info @chrisburr. Most org packages now use Azure pipelines or GitHub actions, hence not a problem.

jpivarski commented 3 years ago

@chrisburr Sorry, our actions crossed!

The awkward0/uproot3 branches will be replaced with awkward1/uproot4 by December 1 (https://indico.cern.ch/event/960587/contributions/4070318/), so it's okay if it hobbles along.

I had no idea that Travis will be shutting down. It has been a lot of trouble—Azure Pipelines and GitHub Actions have been a lot easier to deal with. But I guess the timing is perfect: I wouldn't want to port these tests.

Thanks again for your help!

jpivarski commented 3 years ago

Deployment with Travis took 11 hours! (No errors, just slow. Each task was 1‒5 minutes; it just took a long time to get any time on them. I guess that's why they're giving up the free service, especially with so many competitors.)