[2207] in Kerberos_V5_Development
Re: V1_0_WIN32_BRANCH with Win16
daemon@ATHENA.MIT.EDU (Doug Engert)
Thu Feb 6 18:02:30 1997
Date: Thu, 6 Feb 1997 17:01:29 -0600
From: Doug Engert <DEEngert@anl.gov>
To: "Richard Basch" <basch@lehman.com>
Cc: deengert@anl.gov, tytso@MIT.EDU, krbdev@MIT.EDU
In-Reply-To: <199702030434.XAA09737@badger.lehman.com>
Richard Basch writes:
>
> I have looked a bit more at the Unknown code... basically index 148 of
> the comerr table (lib/krb5/error_tables/krb5_err.c or
> include/krb5_err.h) indicates that code 148 is a clock skew. I suspect
> the Win16 time computation has problems under NT, or may even be broken.
>
No it is not broken, it is just looking for the TZ= in the environment.
More on the blocking_hook_problem in cns.c.
The compiler options in the src/config/windows.in for none DLL routines
are wrong! This may effect other routines as well.
It has ##WIN16##CCOPTS2=/AL /Zp /G2s /Os /Zi /Od /W3 ...
This will not work, since the /GA flag is missing to generate Windows
prologs.
The windows/cns/Makefile.in from the krb5-1.0 release provided its own
flags which adddress this:
CFLAGS_RELEASE = /nologo /W3 /AL /GAs /G2 /Zp /O2 /DNDEBUG=1
The windows\cns\Makefile.in in the krb5-nt-alpha1 relies on the flags
set in the src/config/windows.in
So whomever decied to rewrite the makefiles for the utilites, should
go back to the krb5-1.0 and look close at what else might have been
changed, and settle up on a good set of flags to be used.
My temporary solution was to add the /GA flag and add the _export back
on the blocking_hook_proc as well, as was done with the krb5-1.0.
It seems to have fixed the problem.
--
Douglas E. Engert <DEEngert@anl.gov>
Argonne National Laboratory
9700 South Cass Avenue
Argonne, Illinois 60439
(630) 252-5444
