On Tue, 11 May 2021 03:25:35 -0700, Ed Jaffe wrote:
>
>We experienced significant clock drift on our z13s.  (I can't understand
>why IBM makes STP optional and pricey. All Z machines should be able to
>synchronize to NIST.)
>
>This discrepancy wreaked havoc for our folks trying to use the mainframe
>as a GIT repository, so we decided to correct the issue in CLOCKxx
>exactly as described in this thread.
> 
Might it be better to tweak CVTLSO rather than CLOCKxx?
o Wouldn't that keep GMT better aligned with LT?
o Decrementing CVTLSO might cause intervals to appear inflated.
o Incrementing CVTLSO might cause duplicate/ambiguous timestamps.
o How does CLOCKxx affect OMVS timestamps?  How do your UNIX
  timestamps compare to TIME LT/GMT?

How does STIMER behave when the interval spans a DST or leap second
boundary?  Is this documented?

How does the TIME macro make the STCK and fetch of CVTLSO/CVTLDTO
appear atomic lest one of the latter change between accesses? I know it's
a tiny window, but Murphy.

-- gil

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to