Re: BUG #9371: pg_dump acquiring ROW EXCLUSIVE locks on tables

Dean Rasheed <dean(dot)a(dot)rasheed(at)gmail(dot)com> writes:
> Here's a patch for HEAD along those lines.

> I've tested it on our production data and confirmed that with this
> patch pg_dump no longer acquires exclusive locks. I think this should
> be back-patched, since we do promise that pg_dump does not block other
> readers or writers.

I think this patch looks generally sane, and I agree that the excess
locking is a bug worthy of being back-patched. But is anyone concerned
about changing the signature of AcquireRewriteLocks() in back branches?
I can't immediately think of a reason why extensions might be calling it,
but ...

We could avoid a signature change in back branches by making
AcquireRewriteLocks() into a wrapper around some new function.
But I don't want to do it like that in HEAD, so this would create
a divergence between HEAD and back branches.

I'm inclined to think a signature change is OK. Objections?

regards, tom lane