-------- Original-Nachricht -------- > Datum: Thu, 05 Nov 2009 16:25:26 +0100 > Von: "Axel Dörfler" <axeld@xxxxxxxxxxxxxxxx> > "Ingo Weinhold" <ingo_weinhold@xxxxxx> wrote: > > > "Ingo Weinhold" <ingo_weinhold@xxxxxx> wrote: > > > > I just tried it myself under Haiku and there seems to be an issue > > > > with the time zone not being considered correctly (in GMT+1 I > > > > have > > > > to specify "%+50 mins%" to get the files modified at most ten > > > > minutes ago). > [...] > > > Maybe parsedate() is to blame for this. The BFS time stuff > > > currently > > > doesn't care about timezones, which you could also call a bug, > > > though. > > parsedate() gets the current time as a parameter, so for relative > > time > > strings, it can hardly do anything wrong. BQuery passes time(NULL) to > > it, > > which should be the local time. > > AFAIK time(NULL) returns the system time, which is probably UTC on your > system? Indeed it is. > > It's really the question what time should be passed to the kernel > > (local or > > UTC) and how BFS handles it. > > BFS just don't care right now, but I think it would make sense to always > use UTC as a time base there. Or rather the hardware time -- otherwise the kernel would need to be able to do timezone conversions. I guess this makes sense, given that the stat structure also seems to contain hardware time values. > Currently, BFS uses real_time_clock_usecs() which should be equivalent > to time(NULL) besides the higher resolution. That raises the question where the one hour offset comes from on my machine, though. CU, Ingo