python-irodsclient
python-irodsclient copied to clipboard
irods
Investigate interpreter segfault TODO comment for Python 3.6 vs 3.11
See comment in the following code. https://github.com/irods/python-irodsclient/blob/86a8f11a9399db29774c4096d83bba733a024ab6/irods/test/data_obj_test.py#L741-L744
This may apply to other tests (i.e. collection_test.py).
FWIW, I would recommend dropping support for 3.6 as part of #480 as it has been EOL since the end of 2021 (https://devguide.python.org/versions/#unsupported-versions), and so we won't need to worry about this.
FWIW, I would recommend dropping support for 3.6 as part of #480 as it has been EOL since the end of 2021 (https://devguide.python.org/versions/#unsupported-versions), and so we won't need to worry about this.
Yes please. We currently advertise 3.4, 3.5, and 3.6. Let's drop all three for 2.1.0.
https://github.com/irods/python-irodsclient/blob/0c9600e898b33e184c97f1dfc219375074e201a0/setup.py#L32-L34
We'll keep 2.7 (and 3.7, also EOL) until we get to our 3.0.0 release.
So perhaps we move this issue to 3.0.0 and take care of all EOL issues at once?
I was more thinking of dropping support in 3.0.0. If we drop support for anything that came after python 2.7, that would be kind of weird.
Yes, I think moving this to the 3.0.0 milestone makes sense.
We want to remove support for < Python 3.8 for PRC 3.0.0.
Aka, we'll support 3.8 and later.
We want to remove support for
< Python 3.8for PRC 3.0.0.Aka, we'll support 3.8 and later.
Does this still apply? We should currently support 3.8+ instead of 3.9+? In the current version of the issue description @alanking has given Python 3.9 as the cutoff for the Python PRC should now minimally support, but I kind of think we should still support 3.8+, as 3.8 is the system-Python on ubuntu 20.04 and we still actively publish packages for that OS.
With that in mind, I'm currently still investigating whether 3.8 has the segfault issue.
Update, Python 3.8.0 still segfaults. Going on to 3.9, since ubuntu20.04 has ready-installable packages for that version.
Ok, I've established 3.9.5 (the version of system Python 3.9 that is ready installable on ubuntu 20.04) does not segfault in context of this issue. Will investigate 3.9.0 and if it's stil true of that version, it should cinch 3.9+ as the new definitively supportable Python for latest PRC.
Ok, in the clear for 3.9.0 - no segmentation fault.
We'll work on checking off the boxes and make Python 3.9 the cutoff for upcoming PRC v3.1.0
make Python 3.9 the cutoff for upcoming PRC v3.1.0
Yes please.
The packages/OSs we publish for the server are independent from this client library.
make Python 3.9 the cutoff for upcoming PRC v3.1.0
Yes please.
The packages/OSs we publish for the server are independent from this client library.
True, although there will be a number of new things for e.g. Ubuntu 20 system admins to watch now like making sure an SSL development support lib is installed (think there are a couple options for that one?) as well as zlib1g-dev and maybe others before building a custom Python. Otherwise pip won't install.
But i'm pretty sure those are way outside our scope for this library, right?
If you think not, then let's open a separate issue to discuss any particular pitfalls.
On re-examination, yes - our README is presently just a Python iRODS Client usage guide , not including directions to management of the OS's facilities (like builds of Python) and packages that we depend on. It's big enough already, and if there's a need later for a README-install.md we'll know it.
@d-w-moore confirmed that Python versions >3.8 do not exhibit the issue here. The next release will be listed as compatible only with Python versions >3.8, so the tearDownClass method has been stubbed out and the TODO comment has been removed.
If that sounds accurate, please close this one, @d-w-moore. Thanks