Date: Fri, 28 May 2021 12:08:29 -0500
From: Justin Pryzby <pryzby@telsasoft.com>
To: Andres Freund <andres@anarazel.de>
Cc: pgsql-hackers@postgresql.org, Tom Lane <tgl@sss.pgh.pa.us>
Subject: Re: terminate called after throwing an instance of 'std::bad_alloc'
Message-ID: <20210528170829.GK2082@telsasoft.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210517002936.GD373@telsasoft.com>
User-Agent: Mutt/1.9.4 (2018-02-28)
Archived-At: 
 <https://www.postgresql.org/message-id/20210528170829.GK2082%40telsasoft.com>
Precedence: bulk

There's a memory leak affecting JIT expressions, even when inlining,
optimization, and tuple deforming are disabled.

The server that got OOM after installing PG13.3 (because your patch wasn't
applied) still gets OOM even with inline_above_cost=-1, optimize_above_cost=-1,
and deforming=off.

Actually I think this is an even more severe leak, even though it seems to be
infrequent.  The inlining leak is slow enough that I don't think that it,
alone, would cause a crash.  It's possible there's a race condition or
something that makes this hard to notice or reproduce, or the inlining leak is
just easier to find.

I guess your patch avoids this issue, too, since it didn't OOM with your patch.

-- 
Justin