Simon Conseil

No magic :D, we must add `:noindex:` to the duplicated directives, and this PR just propagates `:noindex:` to autodoc.

> is there an astropy PR that uses this feature? Would be nice to see if it really works before we merge. @pllim - no, I could do that indeed...

Things are going a little bit too fast I think ... there are reasons for the way flake8 was configured. flake8 was run with a small subset of errors in...

@nstarman - Sure, one of the benefit of using black is that the formatting should then be pep8 compliant, except maybe a few specific cases. So I guess once black...

The checksum issue is probably the same as #4587 indeed. About the funpack/BITPIX issue, when updating the hdu and writing the file, astropy adds a ZTENSION keyword which moves ZBITPIX...

> I wonder if there isn't a way to somehow store the initial list of keywords for each HDU when loading compressed file, and later restore only those particular keywords...

I don't see the warning when testing locally with the change from #13753. Can you try again now that it is merged @Pharisaeus ?

Yes sounds like a funpack issue, which merges the compressed extension header back into the primary HDU: ``` ❯ dfits -x 0 invalid.fits.fz | grep 'DATE\|xtension' EXTEND = T /...

@svank - Sorry this PR felt off the radar and there are now conflicts ... Do you think you rebase to get it merged for v5.2 (before the end of...

It feels weird to use `tox.ini` for things that are not tox ... if really pytest deprecates using `setup.cfg`, which is not the case yet, and if support of `pyproject.toml`...