On 2013-06-14 09:21:12 -0700, Jeff Davis wrote:
> On Fri, 2013-06-14 at 16:10 +0200, Andres Freund wrote:
> > Jeff Davis has a patch pending
> > (1365493015.7580.3240.camel@sussancws0025) that passes the buffer_std
> > flag down to MarkBufferDirtyHint() for exactly that reason. I thought we
> > were on track committing that, but rereading the thread it doesn't look
> > that way.
> >
> > Jeff, care to update that patch?
>
> Rebased and attached. Changed so all callers use buffer_std=true except
> those in freespace.c and fsmpage.c.
>
> Simon, did you (or anyone else) have an objection to this patch? If not,
> I'll go ahead and commit it tomorrow morning.
I'd like to see a comment around the memcpys in XLogSaveBufferForHint()
that mentions that they are safe in a non std buffer due to
XLogCheckBuffer setting an appropriate hole/offset. Or make an explicit
change of the copy algorithm there.
Btw, if you touch that code, I'd vote for renaming XLOG_HINT to XLOG_FPI
or something like that. I find the former name confusing...
Greetings,
Andres Freund
-- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services