From: Tom Lane <tgl@sss.pgh.pa.us>
To: 1003931418@qq.com
cc: pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #19423: canonicalize_path does not function correctly when processing paths that contain . and ..
In-reply-to: <19423-a45c4524977e1704@postgresql.org>
References: <19423-a45c4524977e1704@postgresql.org>
Comments: In-reply-to PG Bug reporting form <noreply@postgresql.org>
	message dated "Tue, 03 Mar 2026 07:27:44 +0000"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <1003729.1772548104.1@sss.pgh.pa.us>
Date: Tue, 03 Mar 2026 09:28:24 -0500
Message-ID: <1003730.1772548104@sss.pgh.pa.us>
Archived-At: <https://www.postgresql.org/message-id/1003730.1772548104%40sss.pgh.pa.us>
Precedence: bulk

PG Bug reporting form <noreply@postgresql.org> writes:
> When the input parameter path of the function is
> /tmp/root_hack4_test_tablespace/./../../test, the expected result after
> processing by canonicalize_path is /test, but the actual result remains
> /tmp/root_hack4_test_tablespace/./../../test.

AFAICT it does what you're expecting in v15 and later.  The previous
implementation did not promise to clean up '.' and '..' everywhere;
it only said

  *            o  remove trailing '.'
  *            o  process trailing '..' ourselves

> The implementation of this function differs significantly between version
> 14, version 15, and later versions.

See commit c10f830c511f0ba3e6f4c9d99f444d39e30440c8, particularly
the commit message's explanation of why not back-patch it.

			regards, tom lane