"Simon Riggs" <simon@2ndquadrant.com> writes:
> On Wed, 2006-11-01 at 12:05 -0500, Tom Lane wrote:
>> I don't think this works.
> Surely you are pointing out a bug, no?
> If a backend did crash, the init file would be wrong and we'd get
> exactly the same wrong relfilenode errors we got after that PITR.
Yeah, which is the same bug we've got now. They're both WAL-replay-
doesn't-fix-the-init-file cases.
> The issue must surely be that the patch isn't wrong per se, just that
> RelationCacheInitFileInvalidate is called too late and that requires an
> additional fix.
No. In the non-crash situation there is sufficient interlocking to
avoid a problem, and I feel no desire to redesign that mechanism.
Trying to do it before commit would create its own issues, anyway:
someone could install a new init file before you finish committing.
regards, tom lane