TAI => UTC => TAI does not round-trip for certain dates about erfa

[taldcroft]

The upstream SOFA library now has a candidate patch for this issue. This is currently in test and review by the SOFA board. They expect a release no earlier than late October 2013.

[eteq]

So should we just wait until the SOFA release @taldcroft, given that it will probably be after astropy 0.3, or would you rather apply the patch here?

[timj]

Can we get a clue from @scottransom ?

[taldcroft]

@eteq - Sorry for being a bit vague, but the candidate patch isn't formally available yet. I was offered a preview, but I had the understanding that they really didn't want the code to get out prior to approval by the SOFA board. So I declined and said we would just wait for the SOFA release. (Otherwise their patch would probably be visible to the world as an ERFA PR).

This is a long way of saying, yes, this issue will probably stay open beyond the 0.3 release. The current milestone of "No milestone" is appropriate.

[eteq]

Hah, ok, fair enough, @taldcroft. Guess this will need to be added to "known issues" in astropy for 0.3, then. Hopefully by 0.3.1 it'll be out.

[scottransom]

Yes, the board is supposed to be reviewing the candidate patch in SOFA for this issue. It is a slow process with the board, though...

[eteq]

As per the discussion in astropy/astropy#1031, it looks like this was fixed in the latest version, which we have currently in master. From http://www.iausofa.org/current_changes.html :

Corrected due to bug with TAI->UTC 1960-1972: d2dtf.c, dtf2d.c, taiutc.c and utctai.c.

So I'll close this now. @taldcroft - if you think the issue still exists in some form, feel free to re-open.