Repository navigation
RFC: Rearrange OSError logic so it makes sense #19636
Description
This is a follow-up of #19631 . We need to audit the logic path in astropy/config/configuration.py so things like OSError check still make sense, or if not, move it to a place that does. xref #19631 (comment)
Activity
neutrinoceros commentedon May 3, 2026
Starting to inquire.
First note: the
except OSErrorbranch is never reached even on 7.2.x and withHOMEunset, in the sense that the test suite doesn't cover it. This means we cannot just remove thistry/exceptblock and let CI validate the approach.neutrinoceros commentedon May 3, 2026
Our next clue: this
except OSErrorwas last updated in #10877, and introduced in #8402.neutrinoceros commentedon May 3, 2026
There are however 4 tests currently failing on main if
HOMEis unset (all inastropy/config). It's probably worth looking this up first.neutrinoceros commentedon May 3, 2026
it's actually 20 if I update
ignore_config_paths_global_stateso it only mocks HOME if already setneutrinoceros commentedon May 3, 2026
I think I found a bigger, more concrete problem from here. It gets its own ticket at #19648