On Tuesday, September 06, 2011 15:31:26 Dmitry Olshansky wrote: > On 06.09.2011 7:36, Daniel Murphy wrote: > > The pull-request auto-tester is now available for druntime and phobos, > > as well as dmd. > > > > http://yebblies.com/results/ > > Looks like there is a problem with Phobos datetime on test machine, off > by 1 hour - DST problem?
I'm working on it. It's a combination of an actual bug surrounding DST changes and the fact that you can't actually test that SysTime(dateTime) is going to be equal to dateTime during a DST change, because the OS is inconsistent between time zones on what it does when the requested time is during a DST switch (or even within an hour or two of it in some cases). Any time that you try and set a property of a SysTime, if that property would put it during a DST switch, there are going to be issues, and as far as I can tell, there is no way to get rid of them all. It's quite disgusting really. Adding or subtracting from a time is fine, and because SysTime holds its time internally in UTC, it avoids quite a few problems, but anytime that you have to convert it to its time zone and back again (as must be done when setting its properties), there are issues around DST. I could make it work if the OS were actually consistent, but on Linux at least, it isn't. Bleh. In any case. You don't want to hear/read me ranting about it. I'll have a solution fairly soon, but I think that there's always going to be a risk of the std.datetime unit tests failing in some time zone or other just because one of the times tested happens to fall on one of that time zone's DST switches. - Jonathan M Davis _______________________________________________ phobos mailing list [email protected] http://lists.puremagic.com/mailman/listinfo/phobos
