Date: Wed, 24 Sep 2025 17:13:44 +0200
From: Christoph Berg <myon@debian.org>
To: Daniel Gustafsson <daniel@yesql.se>
Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: "openssl" should not be optional
Message-ID: <aNQKqD8bHmy05Z3p@msg.df7cb.de>
References: <aNPSp1-RIAs3skZm@msg.df7cb.de>
 <52A12672-72CC-47FC-A0FA-DDFF7C23F2D2@yesql.se>
 <aNPX867RbttSeNkl@msg.df7cb.de>
 <EB87D949-F9D4-4A8E-B003-BB0C18848F83@yesql.se>
 <aNPbVscSfN6SlRrJ@msg.df7cb.de>
 <DB9931FA-8A28-4958-BC00-3667F527FCFC@yesql.se>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <DB9931FA-8A28-4958-BC00-3667F527FCFC@yesql.se>
Archived-At: 
 <https://www.postgresql.org/message-id/aNQKqD8bHmy05Z3p%40msg.df7cb.de>
Precedence: bulk

Re: Daniel Gustafsson
> Looking at this I was reminded that we already handle this by using a fallback
> and the test worked all along.  The message for this was quite poorly worded
> though, and used a warning instead of a note.  The attached will try to detect
> openssl being missing before trying to run it, and will skip the warning
> message if the fallback is used (which really isn't a warning in the first
> place).

Thanks, I just built the postgresql-18 again with this patch (and
openssl not installed [*]). It passes fine now.

In the meantime, I also got the report that postgresql-17 is not
failing in that environment, so the problem is new in 18.

> The ERROR in 003_sslinfo is intentional, we are testing that processing fails
> by passing an invalid value.

Ah, I was mentioning that in the original report because it only
showed up in the failing log, but that's just because the non-failing
build does not go scraping the test log files. That made the problem
look bigger than it actually was.

Thanks,
Christoph

[*] future builds will have openssl as build-dependency.