Marten van Kerkwijk

:+1: to removing `codecov` altogether. Couldn't agree more with https://hynek.me/articles/ditch-codecov-python/ > Unfortunately, Codecov has grown very flaky. I have lost any confidence in the fact when it fails a build...

The docstring is indeed not very clear, but the `__iadd__` operator does actually return an array; it is just that `value` is added in-place and then `self` is returned: ```...

@tagliagiulio - astropy follows the IAU recommendation for leap seconds, which is to spread them out equally over the day. See https://docs.astropy.org/en/latest/api/astropy.time.TimeUnix.html#astropy.time.TimeUnix This has the disadvantage of breaking exact correspondence...

I'm basing my comment on the Standards of Fundamental Astronomy (which done by the IAU); see http://www.iausofa.org/sofa_ts_c.pdf, in particular section 4.2 (with more general explanation in 3.5.1). A PR with...

Sorry to have been misleading here! The SOFA guide is really all I saw but clearly I didn't read it well enough to remember more than that I had seen...

More concretely, I think it makes sense to perhaps have a small subsection on leap seconds on the main documentation page, to which classes like `TimeUnix` can refer.

Weird. It is not hypothesis itself, since the version used for predeps is the same as for other runs (6.103.1). But might it be a numpy 2.0 issue? Possibly data...

OK, so then all the `str(path)` in `setup_package` can go too?

Bit far from what I usually work on, but if there's a move to UTF-8 (which is logical), wouldn't it make most sense to just mirror that move and leave...

@charris, @HaoZeke - not sure you are the right ones, but most fixes for `f2py` recently seem to have been by you, so I was hoping you might be willing...