[1004] in Kerberos-V5-bugs
Re: Porting of V5 to AIX 4.1
daemon@ATHENA.MIT.EDU (epeisach@MIT.EDU)
Thu Dec 22 21:48:34 1994
From: epeisach@MIT.EDU
Date: Thu, 22 Dec 1994 21:48:16 -0500
To: alf@minnie.kgn.ibm.com
Cc: krb5-bugs@MIT.EDU
Ted will probably say the same with most of these....
a) make clean nuking some sources - has already been fixed
1) I'm not sure what you mean exactly, but the generated makefiles will
now run autoconf correctly - --topdir is passed properly now... (as well
as the proper configdir...
2) autoconf will be version 2.1 in the next release (although, I think
the acgernal.m4 needed some fixes to deal with Ultrix's broken shell.
----> Hint - I think the top line of aclocal.m4 should have an
AC_VERSION construct - to ensure the proper autoconf is used...
3) The sed problem - Is the problem that SVR4 sed can't handle more than
one comment line? I've just noticed it on a differnent platform but
haven't reported it yet... If it's the same, is a presed sanitizer
required?
4) make clean nuking sources - fixed already
5) re: mk_cmds installable - I'll leave that for someone else....
6) const stuff - ditto
7) gcc seems to be the award winner... Optimizer seems to work on all
platforms where native cc breaks...
8/9 - someone else?
10) The test for -lsocket.... I reported this before wrt SGI machines
where the sys V compatibility libraries are present. (hence libsocket).
gethostbyname is broken in it.... As you mention, the order may be
specific. Perhaps what is needed is a KRB_SOCKET_LIBS definition in
aclocal.m4 which will do the right thing? That way if it is determined
to be broken on a particular platform it can be fixed globally... I
would also wish for a --with-opt-netlib=???? which would be added into
the link line - say for a working resolver library..... Again, could be
implemented here...
Ezra
