I've fixed some of the tests and consolidated some common code. I haven't looked at METH_O yet because I'm not familiar with this stuff, and I can't find a usable function to get the length of structure sequences which includes the non-visible elements. I also don't recall why the ifdef was changed, but I can look into that. As for the regexes, there's probably no limit on how many hours an offset can be, so it may not make much sense to be too clever about it.

I'd like to look at the glibc code for mktime to verify my assumptions. Right now I don't have that much confidence in the new functions in the patch because I may have misunderstood how certain C library functions work. However, I have tested the revised code in DST and non-DST for CEST and CET respectively (using User Mode Linux).
File Added: time-3.diff