I wrote:
> Ugh. Looks like I broke this in 84f5c2908, by not thinking about the
> possibility that a CALL's argument expressions would need an up-to-date
> snapshot.
Concretely, the attached seems to take care of it. I'll refrain
from pushing this till later, though, since v14 is frozen for rc1
for a few hours more.
regards, tom lane