Thursday, 2025-01-02

*** jhorstmann is now known as Guest478610:52
*** jhorstmann is now known as Guest478911:21
clarkbthis is fun it appears that the latest pip that runs with python3.6 vendors a pep517 that understands pep660 but the latest setuptools that runs under python3.6 does not17:09
clarkbso now I'm down a big rabbit hole of vendoring the pep660 implementation in setuptools into pbr so that we can continue to have python3.6 support17:09
fungithat's a rather odd, and surprising, gap17:11
fungii really wish "editable" installs just weren't a thing, but lots of people seem to find them useful17:14
clarkbhttps://zuul.opendev.org/t/openstack/build/92784bf21bf24cd39de0bee7528e1a66/log/job-output.txt#694 that is my takeaway from this error anyway17:29
clarkbsetuptools 64.0.0 or newer is needed for pep660 but python3.6 can only do setuptools 59.x.y17:30
clarkbbut the pip there is happily using this code path instead of the legacy one17:30
clarkbthen it explodes17:30
clarkbthis is quickly becoming whackamole where you fix the python3 problem by vnedoring newer python3 code but that breaks python2...17:49
clarkbI'm hoping that we can kinda flip the script though and make modern python use of pbr less hacky even if it means making old python use of pbr more hacky18:05
fungiyeah, ideally pbr has to do a lot less since modern python packaging is starting to catch up to pbr's features18:06
clarkbside note it is interesting to see how python3 isms "pollute" code bases when they don't really add any value. Like ['foo', *opts, *otherlist] vs ['foo'] + opts + otherslist18:06
tonybat what point do we "give up" and create pbr2 that supports python 2.7 and keep pbr for python3?23:02
tonyband could that actually help?23:02
fungiit doesn't seem like python 2.7 is causing any real problem, it's that for pip and setuptools there is a nexus at python 3.6 where projects with a pyproject.toml can't be installed in editable mode23:16
fungiand pbr wants to have a pyproject.toml and be installed in editable mode for certain tests while supporting python 3.623:17
clarkbtonyb: fungi: right in the thread that brought this all up (its a pbr change) I basically said we can/should consider that when pbr on python2 becomes a problem23:56
clarkbso far it hasn't been a problem. The reason to avoid it is then everything using it has to flip flop the dep one way or the other dependning on how we do it23:56
clarkbthe problem I've run up against today is pypa not making their releases align between tools23:57

Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!