Re: [PATCH] fix a performance issue with multiple logical-decoding walsenders

On Mon, Jan 6, 2020 at 7:57 PM Pierre Ducroquet <[email protected]> wrote:
>
> On Monday, January 6, 2020 6:57:33 PM CET Tom Lane wrote:
> > Pierre Ducroquet <[email protected]> writes:
> > > Attached to this email is a patch with better comments regarding the
> > > XLogSendLogical change.
> >
> > Hi,
> >   This patch entirely fails to apply for me (and for the cfbot too).
> > It looks like (a) it's missing a final newline and (b) all the tabs
> > have been mangled into spaces, and not correctly mangled either.
> > I could probably reconstruct a workable patch if I had to, but
> > it seems likely that it'd be easier for you to resend it with a
> > little more care about attaching an unmodified attachment.
> >
> > As for the question of back-patching, it seems to me that it'd
> > likely be reasonable to put this into v12, but probably not
> > further back.  There will be no interest in back-patching
> > commit cfdf4dc4f, and it seems like the argument for this
> > patch is relatively weak without that.
> >
> >                       regards, tom lane
>
> Hi
>
> My deepest apologies for the patch being broken, I messed up when transferring
> it between my computers after altering the comments. The verbatim one attached
> to this email applies with no issue on current HEAD.
> The patch regarding PostmasterIsAlive is completely pointless since v12 where
> the function was rewritten, and was included only to help reproduce the issue
> on older versions. Back-patching the walsender patch further than v12 would
> imply back-patching all the machinery introduced for PostmasterIsAlive
> (9f09529952) or another intrusive change there, a too big risk indeed.

+1, backpatch to v12 looks sensible.