Ignore ERFA dubious year warning

[taldcroft]

Description

This patch sets warnings to globally ignore the ERFA dubious year warning that gets emitted for UTC dates either before around 1950 or well after the last known leap second (e.g. CxoTime('2099:001')). This warning is conservatively indicating that UTC is not well-defined in those regimes, but we don't care.

One annoying side effect is that this is really global, meaning that once cxotime is imported then it applies also to pure astropy.time.Time or any calls to the erfa package directly. Another strategy would be to catch these warnings individually in application code where they get emitted. But overall I think this warning is useless in the context of Chandra operations work.

Testing

• Passes unit tests on MacOS
• Functional testing

Functional testing

Prior to the patch running acisfp_check produced two ERFA warnings about dubious year. With this patch in place the following is warning free:

(ska3-2021.2rc4) ➜ env PYTHONPATH=/Users/aldcroft/git/cxotime acisfp_check \
   --outdir=out \
   --oflsdir=${SKA}/data/ska_testr/test_loads/2019/MAY2019/oflsa \
   --nlet_file=${SKA}/data/ska_testr/test_loads/2019/MAY2019/oflsa/NonLoadTrackedEvents.txt \
   --state-builder=sql \
   --run-start=2019:135

[javierggt]

why does such a date appear in acis packages anyway?

[javierggt]

The code is simple enough and I believe it does what we expect. Anyway I wanted to check, just to see myself.

When you run that command with the older version, where do you see the warnings? I ran the command as you ran it, which is also the command in ska_testr/packages/acisfp_check/test_regress.sh, but I see no log file anywhere.

[jeanconn]

With regard to why the date comes up, I think in the ACIS package context there's stuff that uses year 2098 or 2099 for "end of time" type things like:

jeanconn-fido> tail -n3 /proj/sot/ska/data/ska_testr/test_loads/2019/MAY2019/oflsa/dahtbon_history.rdb 
2016:202:04:12:23.359	0
2098:001:00:00:00.000	0
2098:001:00:00:01.000	0

It isn't uncommon for us to use a 2099 time more directly as the last state from cmd_states.

I don't know if there might be something convenient with more limited scope to work with those "end of time" times, but ignoring these ERFA warns seems reasonable to me. If we want a separate warning from cxotime when working with a fictitious future date that could be some kind of enhancement request.

[jeanconn]

I'm fine with this. I note I also ran one of the full acis module tests and saw the warnings stay away.