From: Tom Lane <tgl@sss.pgh.pa.us>
To: =?utf-8?Q?=C3=81lvaro?= Herrera <alvherre@kurilemu.de>
cc: Andrew Dunstan <andrew@dunslane.net>,
        PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: fix pg_mkdir_p to tolerate concurrent directory creation
In-reply-to: <ajVLauHFakqDsXsY@alvherre.pgsql>
References: <ajVLauHFakqDsXsY@alvherre.pgsql>
Comments: In-reply-to =?utf-8?Q?=C3=81lvaro?= Herrera <alvherre@kurilemu.de>
	message dated "Fri, 19 Jun 2026 16:01:47 +0200"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <1472097.1781880662.1@sss.pgh.pa.us>
Date: Fri, 19 Jun 2026 10:51:02 -0400
Message-ID: <1472098.1781880662@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/1472098.1781880662%40sss.pgh.pa.us>
Precedence: bulk

=?utf-8?Q?=C3=81lvaro?= Herrera <alvherre@kurilemu.de> writes:
> Hmm, should the 'break's in the loop be 'continue's?  If you're creating
> path a/b/c, and a concurrent process beats you but is only attempting to
> do a/b, then with the break you'll fail to create the final c.

Maybe I'm confused, but I think it's right as-is.  The "break"s are
places where we're giving up and reporting a hard error.  If we want
to continue to the next directory level, we just fall through ---
and doing that with "continue" would be wrong because it'd miss the
step at the bottom of the loop:

		if (!last)
			*p = '/';


			regards, tom lane