I've tried to test this but v4 doesn't apply cleanly after pure2 is applied, and v4 doesn't include enough to test it (applying v4 only causes test failures).

I reviewed v4 and it looks fine in general.  I do see that there are changes in it unrelated to this issue, but they are PEP8 changes so I'm not objecting, but ideally that would be split out into a separate patch.

I think that this code will incorrectly detect something like '%s %%wall clock' as a date spec because it contains '%w', but strptime would consider that '%' followed by the string 'wall'.  A subtle edge case, but worth considering.  Maybe it needs to strip out %% first then look for the % sequences?  Or perhaps just do the conversion and if the Y/M/D fields are set in then decide that it included a date spec, or if the HMS are set then say that it has the time spec included?