[2068] in Kerberos_V5_Development
platform-specific patches for the 1.0 release
daemon@ATHENA.MIT.EDU (Theodore Y. Ts'o)
Tue Dec 3 19:52:01 1996
Date: Tue, 3 Dec 1996 19:51:42 -0500
From: "Theodore Y. Ts'o" <tytso@MIT.EDU>
To: krbdev@MIT.EDU
Cc: Ken Hornstein <kenh@cmf.nrl.navy.mil>, Paul A Vixie <paul@vix.com>
Cc: Doug Engert <deengert@anl.gov>, Andrew Hobson <ahobson@mindspring.com>
Cc: Dave McGuire <mcguire@neurotica.com>
After discussion with the MIT Kerberos team (both in e-mail and some
private discussions), it seems the most reasonable to just simply
include platform-specific patches for those two platforms (HP and BSDI)
where they've been necessary.
Specifically they are:
* Doug Engert's patch to ./util/pty/update_utmp.c so that it
will work under HPUX
* Paul Vixie's patch to remove #include <malloc.h> from
kadmin/testing/util/tcl_kadm5.c for BSDI. (Or
workaround of simply creating a dummy malloc.h file)
* Paul Vixie's patch for tcl 7.4 vs 7.5 for BSDI.
Any patches or special workarounds should be included in the Notes
section of the README.buildinfo file.
Also, gcc warnings about signed vs. unsigned char problems inside the
DES layer are known; we'll worry about them for post 1.0, but they don't
cause any problems in practice.
Yes, we're making compromises, but I really want to push this release
out the door before the IETF meeting next week. In particular, trying
to support which the MIT core developmenet team doesn't have access to,
platforms such as HP and BSDI, is going to take enough time that if we
can support them using some simple platform-specific patches, that may
be our best bet for now. A last minute attempt to support NetBSD (just
before code freeze) caused show-stopper bugs for the Alpha and linux
platforms which pushed back the 1.0 pre-release external testing, and
that's caused us to get a bit gun-shy.
So if you're developing patches for platforms where you need a few
patches, go ahead and make them; please try to keep them as short and as
low-risk as possible. Platform specific #ifdef's (such as #ifdef hpux)
are preferred, because the chance that the patch will screw up other
platforms is limited. Just include the patches in the README.buildinfo
file.
Also, please let us know if you're planning on doing this, so we can
update the install.texinfo file, in the OS compatibility section, with
appropriate warnings that patches might be necessary for the 1.0
release.
Thanks!!
- Ted
P.S. Doug, given that you've submitted patches for the HPUX platforms,
is there any chance you could generate a binary distribution for HPUX
10?
