Darren Tucker wrote:

> Douglas E. Engert wrote:
>> If configure was to just test for all the files

> That's what it does now.

Your, right. I was looking at the proposed patch to configure you sent
on 2/24 that added the check_gssapi_gssapi_h_first

>> then let the
>> #ifdefs in the source test and include the gssapi/xxx versions first
>> then it should work. Configure would be much simpler.

> That has exactly the same problem as currently only with the paths
> reversed.
> If there is a platform that has the system headers in
> /usr/include/gssapi/gssapi.h and the locally-built headers in, eg,
> /usr/local/krb/include/gssapi.h then we're right back where we started.
> Now, I don't know if such a platform exists, but if it does then chances
> are it's working now and your proposed change will break it.

Your configure patch is looking better, what about a
--with-gssapi-include-path= option?

Personally, I am willing to live with the symlink on HP.

The new IETF kitten group should address some of the issues
with gssapi.h and its friends.


Douglas E. Engert
Argonne National Laboratory
9700 South Cass Avenue
Argonne, Illinois 60439
(630) 252-5444

openssh-unix-dev mailing list