"Dan Harkless" <[EMAIL PROTECTED]> wrote:
> D'oh! Sorry about that, Scott (and anyone else affected). Stupid of me to
> think my prototype would be free of conflicts on all UNIXen. That's what I
> get for coding in the middle of the night. I've fixed it so now configure
> looks in <grp.h> and <unistd.h> for the initgroups() declaration and only
> provides its own prototype if it's not in either of those places.
>
> I'm not familiar with the FreeBSD port -- does nmh require changes to the
> main source tree to work on FreeBSD, or are you just putting together some
> kind of binary package? Just trying to gauge how urgent it is to get out a
> 1.0.4 with my fix (wonder if any other OSes choke on that unconditional
> initgroups() prototype...).
The FreeBSD "port" includes a Makefile and a few little patches to get
things to build and install into the right places automagically. The
standard binary package would get built from this "port". If you
want, I could email you the port files to show you what I mean....it's
just a makefile, a few small patch files and a little packaging system
glue.
FYI, what I do in the port is:
1) A few knobs in the Makefile to pass some settings through to the
configure --with-pager, --with-editor & --with-mts settings.
2) A hack in the Makefile to allow overriding of the spool locking
style. It defaults to FLOCK_LOCKING (since that's what the stock
'mail.local' program says it uses).
3) A patch to patch up acconfig.h for FreeBSD - ifdef out the *_LOCKING
defines since I pass them in via CFLAGS from the makefile, define
the BSD42/BSD44 & MAILGROUP defines, change the BACKUP_PREFIX to
"#" (for MH backwards compatibility - I think I got talked into
that one a while back).
4) Patches to add & use an INSTALL_SCRIPT var to some Makefile.in files for
use instead of INSTALL_PROGRAM. It adds this:
test -z "$INSTALL_SCRIPT" && INSTALL_SCRIPT='${INSTALL} -m 755'
AC_SUBST(INSTALL_SCRIPT)
to the configure.in file. IIRC, the INSTALL_PROGRAM setting did
something (probably stripping) that didn't work for scripts.
5) A patch to ifdef out the extra initgroups() prototype in slocal.c.
So, until things are completely autoconf'd, #2 & #3 will remain
necessary. We could fix #4 in the distribution. And it sounds like
you've got #5 covered.
Thanx,
[EMAIL PROTECTED]
