Re: kerberos is the new sqlite: disable, force mit, or ?

[Taylor R Campbell <campbell+netbsd-tech-pkg%mumble.net@localhost>]

> Date: Wed, 4 Sep 2024 12:39:14 +0000
> From: Taylor R Campbell <campbell+netbsd-tech-pkg%mumble.net@localhost>
> 
> Oops.  I'm wrong.  It will cause postgresql16-client to fail to build,
> which is a regression.  OK, let me see if there's another way to
> address this.

So, not having had time in the last week to work on this, I think for
now we should apply the patch I suggested, but also remove `gssapi'
from the suggested options if KRB5_TYPE is anything other than
mit-krb5.

This isn't a regression from the status quo -- that regression
happened last year when the gssapi option was quietly broken at
runtime.  It just means that, instead of a mysterious crash at
runtime, the functionality is not present.

Then, perhaps we can convince pgsql upstream to simply have degraded
functionality, but not loss of gssapi altogether, if the new API is
missing, and also convince heimdal upstream to release a new version
that implements the missing API.