RE: Daylight Savings Time and the CA Server

[johill@lanl.gov (Jeff Hill)]

Stephanie,

Both the osiTime and the osiTimer codes store time internally in a format
which is independent of the time zone and day light savings time. Internally
they store the elapsed GMT delay since the EPICS epoch, which should
be invariant of local differences in time.

Nevertheless, if your system manager sets the time back instead of setting
a flag to convert to day light savings I completely agree that you
should expect discontinuities in the logical flow of time stamps, and
also problems with one hour additional delays in timers that are currently
running (waiting to expire).

Also, it is possible that when a time stamp is converted to
an ANSI "struct tm" or a string with the current date, hours, min, sec,
etc that mistakes could be made. This conversion has been handled in
the past by tsDefs.h, and therefore the EPICS_TS_MIN_WEST needed to be
set, and there were concerns about proper handling of leap seconds.
Recent versions of osiTime use the ANSI C library to perform these
conversions. Therefore, they will rely on a day light savings time flag
in VMS to compute the correct local date and time.

Jeff

>
> Has anybody run a CA server over a daylight-savings-time change
> (in particular, fall back) with no problems?  Looking at the
> timer and time code, it appears to me that all our timers will wait
> an extra hour once the VMS system manager sets time back by an hour
> (which is how daylight savings time is handled here).  Perhaps I need
> to study the code a bit more.
>
> Thanks in advance,
>
> Stephanie Allison
>