Michael Paquier <mich...@paquier.xyz> writes: > On Mon, Oct 21, 2019 at 12:47:40AM -0400, Tom Lane wrote: >> This seems like a real and possibly serious bug to me. Backend sigsetjmp >> callers *must* clear error_context_stack (or restore it to a previous >> value), because if it isn't NULL it's surely pointing at garbage, ie a >> local variable that's no longer part of the valid stack.
> Sure. From my recollection of memories we never set it in autovacuum > code paths (including index entry deletions), so I don't think that we > have an actual live bug here. Uh ... what about, say, auto-analyze on an expression index? That could call user-defined PL functions and thus reach just about all of the backend. regards, tom lane