Re: Time issues on Linux

[Andrew Johnson <anj@aps.anl.gov>]

Hi Ralph,

I thought that's what was probably causing your issue.

On 09/27/2016 08:28 AM, Ralph Lange wrote:
> In libCom/osi's epicsTime module, I see epicsTimeToTM() and
> epicsTimeToGMTM() methods that allow to select if the timezone mechanism
> should be used or not.
> 
> However, the other direction only offers epicsTimeFromTM() that honors
> timezones, and eventually calls mktime().
> 
> Would you agree that epicsTimeFromGMTM() and the matching
> epicsTime::epicsTime(const gm_tm_nano_sec &tm) should be added?

Yes, with our own implementation in epicsTime.cpp. I would suggest
stealing the algorithm for timegm() from

ftp://ftp.netbsd.org/pub/pkgsrc/current/pkgsrc/pkgtools/libnbcompat/files/timegm.c
without explicitly copying their code (ours is going to be different
anyway given that it's producing an epicsTime).

As this change will only be adding new methods/routines I'd be happy to
take it on the 3.14 branch (hopefully in time for the 3.14.12.6 release
at the end of October).

This routine could be useful in epicsTimeTest, IIRC there are some
things we don't test because of the effect of the time-zone on results.

- Andrew

-- 
Arguing for surveillance because you have nothing to hide is no
different than making the claim, "I don't care about freedom of
speech because I have nothing to say." -- Edward Snowdon