Re: [PATCH] dynahash: add memory allocation failure check

m(dot)korotkov(at)postgrespro(dot)ru writes:
> I found a case of potential NULL pointer dereference.
> In src/backend/utils/hash/dynahash.c in function HTAB *hash_create() the
> result of the DynaHashAlloc() is used unsafely.
> The function DynaHashAlloc() calls MemoryContextAllocExtended() with
> MCXT_ALLOC_NO_OOM and can return a NULL pointer.

Ugh, that's a stupid bug. Evidently my fault, too (9c911ec06).

> Added the pointer check for avoiding a potential problem.

This doesn't seem like a nice way to fix it. The code right there is
assuming palloc semantics, which was okay before 9c911ec06, but is so
no longer. I think the right thing is to put it back to palloc
semantics, which means not using DynaHashAlloc there, as attached.

regards, tom lane