[2051] in Kerberos_V5_Development
comments on HP and BSDI problems
daemon@ATHENA.MIT.EDU (Sam Hartman)
Mon Dec 2 01:27:47 1996
Date: Mon, 2 Dec 1996 01:27:40 -0500
From: Sam Hartman <hartmans@MIT.EDU>
To: krbdev@MIT.EDU
I do not know if we want to consider the HP problem a show
stopper or not. Also, if we don't consider it a show stopper, is it
reasonable to release binaries if the patch that is applied is in the
notes section?
Also, if we do decide to fix this problem, Irecommend that we
feature-testify Doug's patch before applying. It currently checks for
hpux; I think it might be reasonable to check for the appropriate
entries in the structure using an AC_TRY_COMPILE in configure.in.
I'm not sure how I missed Doug's patch; I looked through themail
groups it could have possibly gone to and failed to find it, although
it may exceed my expire timeout; I'll try to be more careful in the
future.
I noticed a bug about BSDI and loginpaths.h; I don't know if
Paul will run into this, but if he does, we may need to deal with
that. A BSDI specific patch is probably reasonable for 1.0, although
we should seriously consider looking at making loginpaths.h portable.
Unfortunately, both these issues (pty on HP and loginpaths.h
on BSDI) are bugs in inherently OS-dependent code. While I have tried
to make libpty portable, every new platform has introduced new
bogosities that require new configure test. I am not surprised, nor
particularly disappointed that our code does not work on platforms we
don't have access to. Hopefully, nothing will be too difficult to fix
in the future. The only major problem I see is back-porting telnet
and kshd to the Cray. (I half-removed a lot of poorly abstracted
Cray-specific code from both because I didn't know how to
re-abstract.)
If we do have the joy of another thaw, there isn't really a
need to build both NetBSD 1.1 and 1.2. For i386, I chose to build 1.1
because it will work on 1.2 platforms, I think; if people would
rather, I can build 1.2 for both.
