Re: [sqlsmith] Failed assertion in joinrels.c

From: Dilip Kumar <dilipbalaut(at)gmail(dot)com>
To: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
Cc: Robert Haas <robertmhaas(at)gmail(dot)com>, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com>, Michael Paquier <michael(dot)paquier(at)gmail(dot)com>, Andreas Seltenreich <seltenreich(at)gmx(dot)de>, Piotr Stefaniak <postgres(at)piotr-stefaniak(dot)me>, PostgreSQL mailing lists <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: [sqlsmith] Failed assertion in joinrels.c
Date: 2016-09-09 04:49:05
Message-ID: CAFiTN-sL+TLhf3e8pnT27i2qAHAGUbRhQf3_kXeS2UED6wp9mg@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox
Thread:
Lists: pgsql-hackers

On Thu, Sep 8, 2016 at 7:21 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> We could make it work like that without breaking the ABI if we were
> to add a NOERROR bit to the allowed "flags". However, after looking
> around a bit I'm no longer convinced what I said above is a good idea.
> In particular, if we put the responsibility of reporting the error on
> callers then we'll lose the ability to distinguish "no such pg_type OID"
> from "type exists but it's only a shell". So I'm now thinking it's okay
> to promote lookup_type_cache's elog to a full ereport, especially since
> it's already using ereport for some of the related error cases such as
> the shell-type failure.

Okay..
>
> That still leaves us with what to do for domain_in. A really simple
> fix would be to move the InitDomainConstraintRef call before the
> getBaseTypeAndTypmod call, because the former fetches the typcache
> entry and would then benefit from lookup_type_cache's ereport.
> But that seems a bit magic.
>
> I'm tempted to propose that domain_state_setup start with
>
> typentry = lookup_type_cache(domainType, TYPECACHE_DOMAIN_INFO);
> if (typentry->typtype != TYPTYPE_DOMAIN)
> ereport(ERROR,
> (errcode(ERRCODE_DATATYPE_MISMATCH),
> errmsg("type %s is not a domain",
> format_type_be(domainType))));
>
> removing the need to verify that after getBaseTypeAndTypmod.

Seems like a better option, done it this way..
attaching the modified patch..
>
> The cache loading is basically free because InitDomainConstraintRef
> would do it anyway; so the extra cost here is only one dynahash search.
> You could imagine buying back those cycles by teaching the typcache
> to be able to cache the result of getBaseTypeAndTypmod, but I'm doubtful
> that we really care. This whole setup sequence only happens once per
> query anyway.

Agreed.

--
Regards,
Dilip Kumar
EnterpriseDB: http://www.enterprisedb.com

Attachment Content-Type Size
cache_lookup_failure_v5.patch application/octet-stream 4.2 KB

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Mithun Cy 2016-09-09 05:50:59 Re: Patch: Implement failover on libpq connect level.
Previous Message Tom Lane 2016-09-09 04:35:42 Re: feature request: explain "with details" option