[7919] in Athena Bugs
Re: rsh -x / ld problems
daemon@ATHENA.MIT.EDU (Richard Basch)
Wed Aug 14 07:39:36 1991
Date: Wed, 14 Aug 91 07:39:26 -0400
To: tytso@MIT.EDU
Cc: "Mark W. Eichin" <eichin@MIT.EDU>, testers@MIT.EDU, network@MIT.EDU,
In-Reply-To: Theodore Ts'o's message of Mon, 12 Aug 91 16:53:40 -0400,
From: "Richard Basch" <basch@MIT.EDU>
This is both a response and another bug report:
After several attempts and a couple hours of work, the old-style .klogin
files are once again supported.
I don't understand how Kerberos EVER compiled. The loader let you get
away with multiply defined routines (there is no way that it could not
have needed to link in both .o files that defined strsave()).
Does anyone know of any dependencies on kuserok() besides login and
rkinitd (outside of the Kerberos sources)? We are trying to minimize
recompilation.
------------
The VAX, RT, and MIPS loaders seem to be broken and allow some multiply
defined routines to be overlooked. The contorted example was:
login.o
main
strsave
(lots of others)
libkrb.a
kuserok.o
kuserok - calls ParmCompare
kparse.o
ParmCompare
strsave
Of course, kshd relied on strsave() being defined in the Kerberos
library, just to complicate our lives... we added a copy in kshd and
made the one in kparse.o a static, and all the problems went away on the
RIOS. The RIOS was absolutely correct... the VAX, RT, and MIPS were all
wrong.
-R
