Wrong file location when loading metadata from egg-info about importlib_metadata

[jaraco]

In GitLab by @warsaw on Jan 1, 2019, 19:42

It seems to me like importlib_metadata has to trust the metadata that it finds. I don't think we should be in the game of guessing the intent. If SOURCES has bogus metadata, that seems to be the real bug.

[jaraco]

In GitLab by @jaraco on Feb 23, 2019, 17:53

I think I may have made a mistake in assuming that SOURCES points to files as installed, but now that I think about it more, just based on the name, it's understandable that 'SOURCES' would refer to the location of the files in the SOURCE and not in the distribution or installed package. As a result, it's probably inappropriate for importlib_metadata to return those values to satisfy files().

maybe importlib_metadata should not trust the location of SOURCES from egg-info

That would be my instinct. Importlib_metadata should probably just raise a NotImplementedError or similar.

[jaraco]

In GitLab by @warsaw on Feb 23, 2019, 23:45

@jaraco If SOURCES can't be trusted, what's the approach for getting accurate installed file location metadata from eggs? Is there some other file that captures this? Should eggs be explicitly not supported? Do we have to query the file system at run time? (I don't think I'd like that.)

[jaraco]

In GitLab by @jaraco on Feb 25, 2019, 01:50

what's the approach for getting accurate installed file location metadata from eggs?

Given that the most common usage for eggs these days is the "editable installed" package, I'm not even sure it's a meaningful thing to indicate what was "installed". I'm leaning toward not supporting files at all for eggs.

[jaraco]

In GitLab by @warsaw on Mar 18, 2019, 17:34

@jaraco Given your last comment, should we close this issue as won't fix and release 0.9? This is the last issue open for that milestone.

[jaraco]

In GitLab by @jaraco on Apr 4, 2019, 20:07

closed