From owner-cvs-user  Mon Sep 15 01:11:55 1997
Return-Path: <owner-cvs-user>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id BAA11372
          for cvs-user-outgoing; Mon, 15 Sep 1997 01:11:55 -0700 (PDT)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id BAA11361;
          Mon, 15 Sep 1997 01:11:41 -0700 (PDT)
Received: (from bde@localhost) by godzilla.zeta.org.au (8.8.5/8.6.9) id SAA05669; Mon, 15 Sep 1997 18:05:58 +1000
Date: Mon, 15 Sep 1997 18:05:58 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199709150805.SAA05669@godzilla.zeta.org.au>
To: bde@zeta.org.au, rgrimes@gndrsh.aac.dev.com
Subject: Re: cvs commit: src Makefile
Cc: cvs-all@FreeBSD.ORG, cvs-committers@FreeBSD.ORG, cvs-user@FreeBSD.ORG,
        rgrimes@FreeBSD.ORG
Sender: owner-cvs-user@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

>> >  Log:
>> >  When calling the src/include Makefile use the shorthand ``all beforeinstall''
>> >  inplace of ``all installhdrs symlinks'' so those of us who run with
>> >  SHARED=copies get the right results.
>> 
>> This just undoes rev.1.136 and defeats the point of the installhdrs target,
>> which is to support `make install' forcing installation of symlinks here
>> (for the temporary include directory for `make world').
>> 
>> SHARED=copies is the default.
>
>No, it does not do that.  Take a very careful look, the forcing of

Yes it does.  Of course I looked carefully before commenting.  You broke
the main `includes' target.

>the symlinks for the temporary includes occurs someplace else, namely
>in the ``bootstrap:'' which still has the call to ``symlinks'', which
>DTRT.

The `bootstrap' target only installs very temporary (and not quite right)
includes, just enough to build the bootstrap tools.  The `includes'
target is built a little later.  This must use symlinks so that the
header times don't change for the NOCLEAN case.  (`make bootstrap'
has blown away any files installed by a previous run with SHARED=copies.
The !NOCLEAN case isn't important since everything gets blown away
anyway.)

>revision1.136 broke work that was done over 4 years ago, and until

It only broke the undocumented internal `includes' target.

>then had worked correctly, I just made it work again.  As things
>where _all_ systems ended up with links in /usr/include if they
>ran ``make includes''.

Don't use undocumented internal targets.

Bruce