Fix the build of cramjam a subdependency of fastavro

[pyrco]

No description provided.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 83.70%. Comparing base (1701dcf) to head (14538e5).
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #145   +/-   ##
=======================================
  Coverage   83.70%   83.70%           
=======================================
  Files          34       34           
  Lines        3485     3485           
=======================================
  Hits         2917     2917           
  Misses        568      568           

Flag	Coverage Δ
unittests	83.70% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[pyrco]

This fixes it for pypy3.9 on Linux. The windows fail seems rather more complicated. We could make that build not required anymore?

[yunzheng]

this sadly doesn't fix the windows pypy3.9. Maybe we should pin cramjam to an older version for now?

[yunzheng]

looks like 3.9 support was removed from cramjam. See milesgranger/cramjam#177

[JSCU-CNI]

It seems like pypy3.9 is no longer maintained (or at least distributed) in favor of pypy3.10 since version 7.3.17

We will keep the buildbots building PyPY3.9 until the end of August

https://doc.pypy.org/en/latest/release-v7.3.17.html#pypy-versions-and-speed-pypy-org

[yunzheng]

@pyrco I've pinned cramjam for now for pypy3.9, which seems to be the easiest to do for now.

I think you can also remove the extra deb package now. But i leave that up to you.

[pyrco]

Let's go for the pinning option and see when that fails ;-) If it does there are some other options, which would disable avro support in flow.record on pypy3.9 & windows.

In the meantime we should look into dropping support for pypy3.9. We want to keep CPython 3.9 around for a bit though. The current idea is to stop supporting that once it is out of support upstream.