Log file Viewer
#pypa-dev logs for Thursday the 21st of May, 2020
[13:08:50] <ronny> jaraco: i wonder if it would be ok for a setuptools file finder to ralk up the call tree to find the distribution/command objects in order to read the setuptools_scm configuration
[13:11:41] <pradyunsg> dstufft: could I get the pypi release access for https://pypi.org/project/scripttest/?
[13:16:40] <jaraco> ronny: I’m guessing you want to provide the user more control over which files end up in their project. The philosophy of setuptools file finders is to be dead simple (just enumerate all files relative to ‘.') and leave it to setuptools options to interpret that information.
[13:17:13] <jaraco> So on one hand, I would encourage users to follow that philosophy and avoid more complicated scenarios.
[13:17:22] <ronny> the problem is that they pick up wrong git repos and setups when i cant control them
[13:17:46] <ronny> so in the simple case they work, but pip kills things, and downloads within other git repos kill things
[13:18:10] <jaraco> But assuming that there’s a real need that you want to support, I would recommend against walking the call tree to try to extract properties from the Distribution object.
[13:18:19] <ronny> and right now i have no way to figure in setuptools_scm if the finder is doing will or breaking
[13:18:26] <jaraco> Even if you could make that work reliably, it’s not an interface that Setuptools is going to want to support.
[13:20:24] <jaraco> It sounds like you’re concerned about a particular use-case, where pip is building from an sdist and happens to be in a git repo.
[13:20:55] <jaraco> In that case, it sounds like pip is not effectively isolating the build… and perhaps the problem is already solved by `pip install —use-pep517`
[13:20:58] <ronny> so right now i cant even tell if the file finder is using the right/belonging git repo, and downstream users get errors/problems due to that
[13:25:46] <jaraco> No problem. I need to get to work too. I’d say try to distill the problem to its essence, maybe put together a minimum reproducer, and describe where the users’ and setuptools_scm’s expectations aren’t met. It may be that setuptools_scm should have more sophistication, or it may be that the file finder protocol is insufficient to address this use-case.
[13:26:47] <travis-ci> Change view : https://github.com/pypa/pip/compare/3ec6b824ddbd...80c640b5a892
[15:01:39] <travis-ci> Change view : https://github.com/pypa/pip/compare/80c640b5a892...90168f0b2442
[15:38:07] <travis-ci> Change view : https://github.com/pypa/pip/compare/ba9b38355ecc...278ac2d67ffa
[15:42:05] <ronny> pradyunsg: im still looking for a way to get the actual correct egg-infos from the unisolated source tree in case of a broken isolated build
[15:42:48] <ronny> pradyunsg: and there is the general topic of setuptools file finders not having enough context to know if they happen in the repo of origin, or somewhere painfull (dbian packaging repos&co)
[15:45:02] <ronny> pradyunsg: i hope to get setuptools_scm 4.0 out of the window today so i can focus on COT - cogs of testing
[16:38:10] <travis-ci> Change view : https://github.com/pypa/pip/compare/a14f9aa8ca9d...76b865155ee6
[23:07:19] <travis-ci> Change view : https://github.com/pypa/pip/compare/a14f9aa8ca9d322083e28434a951aff2ffb2e9ea...76b865155ee646927e838c377b46643b51ccd2fd