[39] in Kerberos-V5-bugs
Re: Krb5 changes for Sun
daemon@ATHENA.MIT.EDU (John T Kohl)
Wed Oct 24 15:07:28 1990
Date: Wed, 24 Oct 90 15:07:00 -0400
From: John T Kohl <jtkohl@ATHENA.MIT.EDU>
To: prb@ATHENA.MIT.EDU
Cc: marks@sun.com, krb5-bugs@ATHENA.MIT.EDU
In-Reply-To: Paul Borman's message of Wed, 24 Oct 90 14:55:11 EDT,
> From: prb@ATHENA.MIT.EDU
> Date: Wed, 24 Oct 90 14:55:11 EDT
> It is unfortunate that by doing:
> cd lib ; make clean ; make
> you end up with a libkrb5.a that is not usable (e.g. it does not have the
> .o's from lib/../error_tables...
read the discuss meeting for some details on a discussion of exactly
this topic.
> ss/cmd_tbl.c
> ss/listen.c
> ss/list_rqs.c
> libos/read_pwd.c
> These all have problems with the return value from signal. I
> have
> #define HasVoidSignalReturn YES
> in my sun.cf file, but that does not appear to help.
The voidSignalReturn isn't (currently) done by HasVoidSignalReturn; it's
done with <krb5/posix-conf.h>. When we upgrade the include files such
that they are generated per-build tree, this problem should disappear.
> inlcude/krb5/ext-proto.h
> Added before and #includes:
> #ifdef SYSV
> #define index strchr
> #define rindex strrchr
> #endif
> Changed the include for string?.h
> #ifdef SYSV
> #include <string.h>
> #else
> #include <strings.h>
> #endif
<string.h> should ALWAYS be the right thing; on BSD4.3 systems, it has
the str? routines and an #include <strings.h>
> # these next three are all about looking between -1 and unsigned.
Those changes create problems that will get fixed a different way (by
choosing a suitable compile-time value for the default encryption types).
Thanks,
John
