On Fri, Oct 2, 2015 at 4:11 PM, Peter Geoghegan <pg(at)heroku(dot)com> wrote:
> Actually, isn't that another bug? The fact that we don't do the same
> from within gc_qtexts() in normal cases, even with an exclusive lock
> held? We do this:
Ah, no. We check pgss->gc_count in any case, so it should be fine.
That will also make it safe to do the unlink() as outlined already,
because a new qtext_load_file() call from
pg_stat_statements_internal() (due to gc_count bump) will allocate the
file again by name.
--
Peter Geoghegan