Mark Wong <mark(at)2ndQuadrant(dot)com> writes:
> The following I've enabled force_parallel_mode for HEAD, 11, 10, and
> 9.6:
Thanks Mark!
In theory, the stack trace we now have from shoveler proves that parallel
mode has nothing to do with this, because the crash happens during
planning (specifically, inlining SQL functions). I wonder though if
it's possible that previous force_parallel_mode queries have had some
undesirable effect on the process's stack allocation. Just grasping
at straws, because it's sure not clear what's going wrong.
regards, tom lane