RFC: MIT kerberos and the gssd in main
- Reply: Cy Schubert : "Re: RFC: MIT kerberos and the gssd in main"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Sat, 02 Aug 2025 23:17:11 UTC
Hi,
I've lost track of the discussions (or even where they are
taking place, so I am going to post here and hope the
discussion stays here.
My personal preference (feel free to discuss this) is that,
when MK_MITKRB5 == "yes" for the buildworld/installworld..
- The .h files under /usr/include are exactly the same ones that
"pkg install krb5" generates and under the exact same names.
(No Heimdal .h files under /usr/include and no renaming or
putting them in a different subdir.)
- The libraries under /usr/lib are exactly the same ones that
"pkg install krb5" generates and under the exact same names.
I think this will minimize confusion. Yes, anything that links to
libgssapi will need to be fixed (Makefile plus ???) since there
is no such library for MIT, but at least people will see what needs
to be fixed. (There are a lot of places where code knows where
MIT puts .h files and which MIT kerberos library names are used.)
However (and this is the more important part for me), I'd like
a resolution w.r.t.what file names and where they go soon, so
I can get a patch for gssd.c needed to make it work for MIT
straightened out.
I do now have code that works when linked to the libraries
in /usr/local/lib, using the MIT kerberos .h files.
Thanks for any comments, rick
ps: Unless someone complains about doing so, I intend to
tweak /usr/src/usr.sbin/Makefile so that it only builds the
gssd when both MK_GSSAPI and MK_KERBEROS_SUPPORT
are not "no". This allows me to get rid of the cruft in gssd.c
that makes it build for the MK_KERBEROS_SUPPORT == "no"
case, since it won't do anything useful without kerberos anyhow.