Zefram wrote:
For example, suppose you have a rule (not expressible in POSIX form) applying from 2012 onwards that has transitions in March and September, plus some one-off transitions in November 2013.
The idea behind the recent changes is that all the rules one can write can be expressed in extended-POSIX form; if that's good enough, then the scenario you describe is no longer possible. Can you supply an example illustrating the problem? Does this have anything to do with yearistype.sh? Come to think of it, there is one situation the zic code doesn't cover -- perpetual DST. The code claims that this is a POSIX no-no, but on further thought that's incorrect; for example, for San Luis's perpetual-DST (before the database changed yesterday) we can use TZ='WART4WARST,J1/0,J365/24'. I'll come up with a patch to fix this.