mktime in a C program

[labadie_1]

Alpha Vms 7.3-1 unpatched (no ACRTL or TDF patch or any, this is a test machine)

A program using mktime gives a one hour drift from the system time (minus one hour).
When I ask for a time after the next time change (last Sunday of October), it is plus one hour.
Anyway, always incorrect.

I seem to recall that tzset had a relationship with that. At
http://h71000.www7.hp.com/commercial/c/docs/5763p053.html

but if somebody has an example of using tzset, it would help.

[John Gillings]

Gerard,

There have been several fixes to timezones, so it would be best to get your system up to rev first.

Check what your timezones set to, then check the definitions. If the definitions are wrong, have a look at the timezone update procedures at:

http://ftp.hp.com.au

Although these are for Australian & New Zealand timezones, I'm sure you can use them as templates to fix yours.

(and if you're running DECnet+, don't forget to fix those zone definitions too!)

[Hein van den Heuvel]

>>> A program using mktime gives a one hour drift from the system time (minus one hour).


Or.. it is returning UTC.
In Greenwich it one hour earlier then local for you no?

fwiw,
Hein.

[labadie_1]

I think the setup is correct (for France) on the node

$ sh log *zone*

(LNM$PROCESS_TABLE)

(LNM$JOB_818E2080)

(LNM$GROUP_000001)

(LNM$SYSTEM_TABLE)

"SYS$TIMEZONE_DAYLIGHT_SAVING" = "1"
"SYS$TIMEZONE_DIFFERENTIAL" = "7200"
"SYS$TIMEZONE_NAME" = "MET DST"
"SYS$TIMEZONE_RULE" = "MET-1MET DST-2,M3.4.0/02,M10.5.0/03"

(LNM$SYSCLUSTER_TABLE)

(DECW$LOGICAL_NAMES)


I have found in comp.os.vms a post by David Jones which says

You've got it backwards, decc$fix_time is NOT adding the differential while
time() is subtracting it (in order to give the illusion that the system clock
is UTC). The ctime() expects its argument to be UTC and then adjusts the
displayed time returned by whatever the current timezone setting (TZSET) is.
The documentation should be clearer that decc$fix_time() does not do any
timezone adjustments, but I think it's use is discouraged anyway.

The crux of the problem is that VMS's system clock uses local time rather than
UTC. The timestamps in the file system are local time also, which causes
similar grief with the times returned in stat structures.


David L. Jones

[Vouters · Accepted solution]

Hi Gerard,

Can you test this program on your computer?
Also use the ctime function to get the local time.
Regards,
Philippe

[labadie_1]

Thanks a lot Philippe, this is exactly what I was looking for.