[17213] in Kerberos_V5_Development
Re: [PATCH] Make krb5-config suppress CFLAGS output and omit extra
daemon@ATHENA.MIT.EDU (Russ Allbery)
Mon Sep 26 12:03:54 2011
From: Russ Allbery <rra@stanford.edu>
To: "krbdev\@mit.edu" <krbdev@mit.edu>
In-Reply-To: <1317050237.2101.53.camel@t410> (Greg Hudson's message of "Mon,
26 Sep 2011 11:17:17 -0400")
Date: Mon, 26 Sep 2011 09:03:49 -0700
Message-ID: <87d3enuw7u.fsf@windlord.stanford.edu>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Errors-To: krbdev-bounces@mit.edu
Greg Hudson <ghudson@mit.edu> writes:
> This is more complicated. First, is do_deps an established convention?
> pkg-config uses the --static flag to decide whether to output dependency
> libraries (on the theory that you need them for static linking but not
> shared linking).
curl-config uses --static-libs as the flag to get this information.
I think at this point including all the internal dependencies is only
needed for static linking. I'm not certain that's true across all of the
stranger platforms (like HP-UX or AIX), but I suspect it is. I would
therefore omit all the internal dependencies except to the degree that one
is supporting static linking.
> Third, the provided patch only omits external dependencies, not internal
> ones (or com_err). An app using only GSSAPI functions should depend
> only on libgssapi_krb5, not on libkrb5 and libk5crypto. However, it's
> likely that apps have come to expect the current behavior--especially as
> we don't appear to properly support things like "krb5-config --libs
> gssapi krb5". So, this problem may be intractable.
I think you should go ahead and break clients that use krb5-config --libs
gssapi and then expect to have Kerberos symbols available. The interface
is pretty clearly defined ("GSSAPI application with Kerberos 5 bindings"),
and clients should be running both krb5-config --libs gssapi and
krb5-config --libs krb5 and combining them. (That's what my applications
that need both already do.) There may be some build system breakage for
people who did the wrong thing, but it's a lot cleaner as an interface.
--
Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>
_______________________________________________
krbdev mailing list krbdev@mit.edu
https://mailman.mit.edu/mailman/listinfo/krbdev
