"Heikki Linnakangas" <heikki@enterprisedb.com> writes:
> On second thought, we do call PageGetHeapFreeSpace without holding a
> lock in heap_page_prune_opt as well, so it better be safe. Looking
> closer at PageGetHeapFreeSpace, I think it is. The
> return value can be bogus, of course.
> That's worth noting in the comments:
On closer look, the comments in heap_page_prune_opt seem to address
the issue already, and I'm not sure why we should comment
PageGetHeapFreeSpace this way but not all the other variants in
bufpage.c.
regards, tom lane