On Fri, Aug 03, 2007 at 09:01:22AM -0400, Andrew Dunstan wrote:
> >>>>[ patch to use pg_strftime in xlog.c ]
> >>>>
> >>>This code deliberately does not use pg_strftime, for the same reasons
> >>>that elog.c doesn't use it.
> >>>
> >>>I'm inclined to think that an appropriate fix is the same as we use in
> >>>elog.c, ie, don't use %Z at all under Windows.
> >>>
> >>Eh? Do you mean this change?
> >>"%Y-%m-%d %H:%M:%S %Z" to "%Y-%m-%d %H:%M:%S"
> >>That tzname is expressed here does not regard me as a problem.
> >>Probably, elog.c has still more nearly another problem.
> >>
> >
> >Having talked a bit off-list with Hiroshi-san, he came up with the
> >suggestion taht we should be logging this information in UTC/GMT instead of
> >the servers timezone (for all cases, not just win32). That would make
> >things equally "safe" wrt changes in the pg timezone, and always
> >predictable.
> >
> >Thoughts on this?
> >
> >
> >
>
> Are you just talking about xlog.c? That wouldn't be an acceptable change
> in elog.c, although we could provide additional escapes to output UTC if
> needed.
Yes, just talking about xlog.c.
//Magnus