On Tue, Oct 6, 2015 at 5:32 AM, Andres Freund <andres@anarazel.de> wrote:
> The problem is basically that cache invalidations can arrive while
> you're building new cache entries. Everytime you e.g. open a relation
> cache invalidations can arrive. Assume you'd written the code like:
> You're avoiding that by only entering into the hashtable *after* the
> lookup. And I think that deserves a comment.
Thanks, revised patch attached.
P.