Content-Type: text/plain;
	charset=us-ascii
Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3776.700.51.11.1\))
Subject: Re: disabled SSL log_like tests
From: Daniel Gustafsson <daniel@yesql.se>
In-Reply-To: <3320404.1746633879@sss.pgh.pa.us>
Date: Wed, 7 May 2025 19:53:38 +0200
Cc: Thomas Munro <thomas.munro@gmail.com>,
 Andrew Dunstan <andrew@dunslane.net>,
 PostgreSQL Hackers <pgsql-hackers@postgresql.org>,
 Jacob Champion <jacob.champion@enterprisedb.com>
Content-Transfer-Encoding: quoted-printable
Message-Id: <8E3E58BF-CC99-4C45-9FB8-3BA83AA6FDC8@yesql.se>
References: <984fca80-85a8-4c6f-a5cc-bb860950b435@dunslane.net>
 <2199758.1744901785@sss.pgh.pa.us>
 <1ce11d3f-624c-4003-a032-1b10cc138305@dunslane.net>
 <2814408.1745005558@sss.pgh.pa.us> <2859105.1745015195@sss.pgh.pa.us>
 <1120735.1745350422@sss.pgh.pa.us>
 <CA+hUKG+fLqyweHqFSBcErueUVT0vDuSNWui-ySz3+d_APmq7dw@mail.gmail.com>
 <3058990.1746485124@sss.pgh.pa.us> <3227092.1746580691@sss.pgh.pa.us>
 <CA+hUKGJkUJAwckjBr-=B9oUdDgjSEH+3njd__OnPoPuAtze3Qg@mail.gmail.com>
 <3248136.1746592452@sss.pgh.pa.us>
 <3004EFB2-8FFA-429A-AC5E-E3F40E002A94@yesql.se>
 <3320404.1746633879@sss.pgh.pa.us>
To: Tom Lane <tgl@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/8E3E58BF-CC99-4C45-9FB8-3BA83AA6FDC8%40yesql.se>
Precedence: bulk

> On 7 May 2025, at 18:04, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>=20
> Daniel Gustafsson <daniel@yesql.se> writes:
>>> On 7 May 2025, at 06:34, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> I couldn't help noticing that the backtraces went through
>>> lib/libssl/tls13_legacy.c, which doesn't give a warm feeling
>>> about how supported they think our usage is (and perhaps also
>>> explains why they didn't detect this bug themselves).
>=20
>> Since we no longer support 1.0.2 we can apply something like the =
(lightly
>> tested) attached which should be a no-op as we already use =
TLS_method() but via
>> an alias.
>=20
> Yeah, I saw that SSLv23_method() was merely an alias for TLS_method()
> in LibreSSL as well.  That means unfortunately that your proposal is
> just cosmetic and doesn't get us out of using code that they're
> calling "legacy".  I wonder what it would take to get to the "modern"
> code paths.

AFAICT (it's not documented what I can see) the Libressl folks consider =
code
inherited from OpenSSL legacy.  Using current OpenSSL API's and moving =
away
from deprecated API's is probably our best bet.  On that note, =
TLS_method is
the current API to use in both OpenSSL and Libressl according to their
respective documentations.

A separate be-secure-libressl.c could move to use their libtls instead =
of the
libssl OpenSSL compatibility library, which may be an interesting =
excercise but
a very different project from what is discussed here.

--
Daniel Gustafsson