[980] in Kerberos-V5-bugs
random pl3 stuff..
daemon@ATHENA.MIT.EDU (*Hobbit*)
Wed Dec 7 05:46:45 1994
Date: Wed, 7 Dec 1994 05:46:39 -0500
From: *Hobbit* <hobbit@asylum.sf.ca.us>
To: krb5-bugs@MIT.EDU
Tom mentioned there being a new patchlevel out, but I couldn't find it.
I'm playing with pl3 on a Linux box, and ran into some things you've probably
already heard about, but in case I'm wrong...
The "clean" targets in util/et and util/ss blow away awk files that are
needed. Since "make distclean" seems an appropriate way to back out of
a failed configure or make run, this gets in the way. I'm not sure why it
didn't hose me yet on the sunos box; the clean targets look the same.
In any case, these .awk files are *not* in the top level "src" directory
as Makefile.in would have you believe. I'm not sure what was intended here,
so I don't know what to change.
Again, in util/ss and util/et, Makefile.in assumes SED = /bin/sed. This
doesn't exist under linux, it's /usr/bin/sed. How 'bout an AC_PROG_SED
autoconf macro, maybe, as with awk?
Autoconf seems to mainly use HAVE_xxx as its indicator defines, but there
are still instances of USE_xxx or HAS_xxx sprinkled throughout the kerberos
code. Is this intentional, or are they going to be made more consistent
someday? I see all three formats [and a few others] in hunks of source code
from everywhere, but the gnu people seem to have settled on HAVE_xxx for
features, calls, and .h file names alike. Notable occurrences I found
in K5 were USE_DIRENT_H, HAS_UTIMES, HAS_STRDUP ...
In appl/bsd, configure produces a makefile that tries to define "srandom"
as "srand48". This conflicts with the Linux .h files, and the compiler
complains of redefinition. [Hi MarkE! your name was in there...]
In appl/bsd/krlogind.c and krshd.c, the symbol PF_UNSPEC is used, which
isn't defined in my socket.h. This is more likely to be a linux problem,
since socket.h there has AF_UNSPEC and defines for PF_everything-else
point back to AF_everything-else, but not for UNSPEC. The other socket
code in that directory just uses 0, which is probably more portable!
Krlogind.c also has some termios trouble, with redefine warnings of
TIOCPKT_xxx, but it built anyway...
What is the "official" linux development platform? I've got kernel 1.1.54
here and everything else from slackware 2.0.0 [i think -- it bloody changes
every week], and don't discount the possibility that I introduced some
skew by running a newer kernel with *very slightly* older binaries.
_H*
