On Tue, Sep 29, 2015 at 1:17 AM, Christophe JAILLET <[email protected]> wrote: > > However, I was wondering if we could call tzset() and use tzname[0] or > tzname[1] instead of the hard coded "UTC+0". According to [1], tzset() first > looks at TZ, but, if not set, then at the localtime file. > This way, we could define TZ if it can be a performance boost in some cases > and avoid the change of behavior on existing systems. > > > However, > - I don't know if this is portable > - I'm unsure about how daylight saving time will be handled
A better guess would be nice. When the runtime doesn't match the test suite, some tests (ssl/varlookup.t) can fail when the timezone is behind/ahead on day of the week from UTC.
