[1992] in Kerberos-V5-bugs
Re: minor adjustments to krb5-beta6
daemon@ATHENA.MIT.EDU (Dave Blob)
Wed Jun 12 14:48:42 1996
Date: Wed, 12 Jun 1996 13:47:53 -0500 (CDT)
From: Dave Blob <blob@syl.dl.nec.com>
To: Sam Hartman <hartmans@MIT.EDU>
Cc: krb5-bugs@MIT.EDU
In-Reply-To: <tslpw74gb4i.fsf@tertius.mit.edu>
Reply-To: blob@syl.dl.nec.com (David Blob)
> Sam (Sam Hartman <hartmans@mit.edu>) writes:
>>>>>> "Dave" == Dave Blob <blob@syl.dl.nec.com> writes:
Dave> - On BSDi 1.1, ar needs the T flag in lib/gssapi, since some
Dave> filenames are greater than 15 characters...
Sam> I see two solutions here: shorten the file names
Sam> (possibly prefered) or include the t flag.
Sam> Questions: does are fail or warn without the t flag?
On BSDi 2.0, the truncation gets done automagically & you get a warning:
> ar: warning: g_acquire_cred.o truncated to g_acquire_cred.
On BSDi 1.1, there's not truncation or warning. Ranlib fails with:
> ranlib: libgssapi.a: Inappropriate file type or format
Sam> How can we determine if a particular OS supports ar -t in a
Sam> non-os-dependent manner? (I realize you could just check to
Sam> see if ar returns a non-zero xit status when given the -t flag,
Sam> but are there any OSes where ar -t works but does something we
Sam> don't expect or desire?)
Just so we're clear here, it's gotta be the -T flag. -t is for a table
listing on almost every OS I checked... Anyhow, I checked all our
machines and none of them use -T otherwise. Since that includes a
linux box, I assume that include gnu's ar. If shortening the
problems, maybe just a blurb in the doc/OS-notes that says:
On BSDI 1.1, make with ARADD="ar crTuv"
Although this may mess up libupdate which sets ARADD by hand (you could
change that to only set it if it wasn't already set)... Still though, I
think shortening the filenames is the bigger win.
Dave> - On most BSDs, you need -DMISSING_ERROR_PROTO
Dave> -DMISSING_CLEARERR_PROTO, although it isn't documented...
Sam> There is a configure test we wrote for NetBSD that
Sam> should detect this and define the appropriate symbols. I would
Sam> appreciate if you could look at configure or aclocal.m4 and
Sam> determine why this test fails on BSDI.
*** aclocal.m4.orig Wed Jun 12 13:22:17 1996
--- aclocal.m4 Wed Jun 12 13:21:55 1996
***************
*** 1091,1097 ****
[$2();],
eval krb5_cv_missing_$2_prototype=yes,
eval krb5_cv_missing_$2_prototype=no)))
! if eval "test \"`echo '$krb5_cv_missing_'$2_prototype`\" = yes"; then
AC_MSG_RESULT(yes)
$4
else
--- 1091,1097 ----
[$2();],
eval krb5_cv_missing_$2_prototype=yes,
eval krb5_cv_missing_$2_prototype=no)))
! if eval "test \"`echo '$krb5_cv_missing_'$1_prototype`\" = yes"; then
AC_MSG_RESULT(yes)
$4
else
Also:
- src/config/config.guess fails for both bsdi's, but the one in
util/autoconf does not...
- mk_cmds made in util/ss has the full path to the build directory in
it, which messes up things like moving the hierarchy you compiled in
to another disk...
- Dave
