"Hiroshi Inoue" <Inoue@tpf.co.jp> writes:
> Oh,I was just looking at heapoverride stuff quite accidentally.
> Yes, this call is ugly and should be replaced by CommandCounterIncrement().
OK, I'm running a build now with setheapoverride calls removed.
Will see what happens.
About half of the setheapoverride calls surrounded heap_update()
(formerly called heap_replace()) calls. AFAICS there is no need
for these calls unless heap_update itself needs them --- but there
are many calls to heap_update that do not have setheapoverride.
Perhaps heap_replace once needed setheapoverride but no longer does?
I am going to try just removing these calls without adding a
CommandCounterIncrement to replace them. If anyone knows that
this is a bad idea, let me know!
regards, tom lane