From: Tom Lane <tgl@sss.pgh.pa.us>
To: Michael Paquier <michael@paquier.xyz>
cc: Robert Haas <robertmhaas@gmail.com>, Andres Freund <andres@anarazel.de>,
        Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>,
        PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: Issue with cancel_before_shmem_exit while searching to remove a
 particular registered exit callbacks
In-reply-to: <20200907074057.GG4867@paquier.xyz>
References: 
 <CA+TgmoY8hrNE-sO+B95M8AD6CL9VohSMik2VZkrnkbANRDS5gw@mail.gmail.com>
 <20200807212022.wocdoo2mmzhjsjm3@alap3.anarazel.de>
 <CA+TgmoYgNMc7x814iih==wH-jN-4bO_FH_xWBdV5HEK=6e3kDQ@mail.gmail.com>
 <3195298.1597070661@sss.pgh.pa.us>
 <CA+TgmobBH1SGmQY6G9x4nth03bJXyg2FXEbmNqs-_aRp1peAgQ@mail.gmail.com>
 <3250875.1597088493@sss.pgh.pa.us>
 <CA+Tgmoaz=e0CEOVpFTMRtanXX+0P-1uRrYtKdWktQedLtGs6Qg@mail.gmail.com>
 <20200811001102.bwcwmrwgirxyec6i@alap3.anarazel.de>
 <3354159.1597106762@sss.pgh.pa.us>
 <CA+TgmobSwOUTonpKUKaVLiK2XGLw4OcVn+hyffJKBNbZcJpACw@mail.gmail.com>
 <20200907074057.GG4867@paquier.xyz>
Comments: In-reply-to Michael Paquier <michael@paquier.xyz>
	message dated "Mon, 07 Sep 2020 16:40:57 +0900"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <1671026.1599492021.1@sss.pgh.pa.us>
Date: Mon, 07 Sep 2020 11:20:21 -0400
Message-ID: <1671027.1599492021@sss.pgh.pa.us>
Precedence: bulk

Michael Paquier <michael@paquier.xyz> writes:
> Is there still something that needs to absolutely be done here knowing
> that we have bab1500 that got rid of the root issue?  Can the CF entry
> be marked as committed?

I think there is agreement that we're not going to change
cancel_before_shmem_exit's restriction to only allow LIFO popping.
So we should improve its comment to explain why.  The other thing
that seems legitimately on-the-table for this CF entry is whether
we should change cancel_before_shmem_exit to complain, rather than
silently do nothing, if it fails to pop the stack.  Bharath's
last patchset proposed to add an elog(DEBUG3) complaint, which
seems to me to be just about entirely useless.  I'd make it an
ERROR, or maybe an Assert.

			regards, tom lane