Re: Emit a warning if the extension's GUC is set incorrectly

From: Kyotaro Horiguchi <horikyota(dot)ntt(at)gmail(dot)com>
To: Shinya11(dot)Kato(at)oss(dot)nttdata(dot)com
Cc: masao(dot)fujii(at)oss(dot)nttdata(dot)com, bharath(dot)rupireddyforpostgres(at)gmail(dot)com, daniel(at)yesql(dot)se, pgsql-hackers(at)postgresql(dot)org
Subject: Re: Emit a warning if the extension's GUC is set incorrectly
Date: 2021-12-21 01:50:07
Message-ID: 20211221.105007.916160321220747245.horikyota.ntt@gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

At Fri, 17 Dec 2021 11:25:22 +0900, Shinya Kato <Shinya11(dot)Kato(at)oss(dot)nttdata(dot)com> wrote in
> On 2021-12-17 01:55, Fujii Masao wrote:
> > I'm still not sure why you were thinking that ERROR is more proper
> > here.
>
> Since I get an ERROR when I set the wrong normal GUCs, I thought I
> should also get an ERROR when I set the wrong extension's GUCs.
> However, there are likely to be harmful effects, and I may need to
> think again.

The GUC placeholder mechanism is evidently designed so that server
allows to define variables unknown to the server before loading
extension modules. ERRORing out that variables at the timing is
contradicting the objective for the placeholder mechanism.

Addition to that EmitWarningsOnPlaceholders()'s objective is to warn
for variables that shouldn't be of a *known* namespace. However, the
patch is intending to check out variable definitions of all namespaces
that are seen in configuration file, but it is undeterminable at that
time whether each of the namespaces is either just wrong or unknown
yet. And the latter is, as mentioned above, what we are intending to
allow.

As the concusion, I think we cannot rule-out "wrong" namespaces unless
we find any means to distinguish whether a namespace is wrong or
not-yet-defined before loading extension modules.

regards.

--
Kyotaro Horiguchi
NTT Open Source Software Center

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Greg Nancarrow 2021-12-21 01:55:52 Re: PublicationActions - use bit flags.
Previous Message wangw.fnst@fujitsu.com 2021-12-21 01:31:42 RE: Confused comment about drop replica identity index