Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Fri, 11 May 2007 17:30:54 -0500

Hi Ralph,

Ralph Lange wrote:

I think I found something that smells like a bug in the vxWorks "local"

clock (or in libCom/osi/os/vxWorks/iocClock).

This smells like a similar issue to the SNS daylight savings problem.

Please try setting the EPICS_TIMEZONE environment variable instead of

the TIMEZONE variable from the st.cmd script and see if the same

behaviour occurs - Dave Thompson found that this made a difference, such

that before iocInit() you must use EPICS_TIMEZONE, not TIMEZONE.

If I've understood that problem properly it stems from the fact that

ANSI C doesn't provide a version of mktime() that takes in UTC, and the

vxWorks version of mktime() is broken. I started working on a new

epicsTimeLoadTimeInit::epicsTimeLoadTimeInit() routine in epicsTime.cpp

(Dave's problem seemed to be associated with the code that calculates

the values of epicsEpochOffset and epicsEpochOffsetAsAnUnsignedLong),

but haven't finished it yet.

- Andrew
--
The right to be heard does not automatically include
the right to be taken seriously. -- Hubert H. Humphrey

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System