Has anyone kept track of all the timezone bugs and made a regression test suite that can be applied to a C library to make sure it handles time zones correctly for the various definitions of correct?