Re: relcache reference leak on refresh materialized view concurrently

yamamoto@valinux.co.jp (YAMAMOTO Takashi) writes:
> the following patch fixes missing index_close in case
> the mv has non-unique indexes.

Hm ... I see the leak, I think, but isn't this an extraordinarily
complex patch?  Looks like adding

    else
    {
        index_close(indexRel, NoLock);
    }

at the bottom of the "if" would also fix the problem, and would be
far easier to verify.

            regards, tom lane