On Fri, 15 Aug 1997 12:03:31 -0600  Michael L. Hitch wrote:

> On Aug 15, 10:33am, Simon Burge wrote:
> > On Thu, 14 Aug 1997 14:34:34 -0700  Jonathan Stone wrote:
> > 
> > > 
> > > Here it is. Totally uuntested:
> > > 
> > > /*	$NetBSD: modf.S,v 1.7 1996/09/17 01:32:26 jonathan Exp $	*/
> > >
> > > [ code deleted ]
> > 
> > This doesn't work either - I put the new modf.S in, did a "make all install"
> > in libc, and re-run ldconfig (necessary?).  I copied Sean's test program
> > to zz.c, and got the following:
> 
>   I rebuilt libc.so, installed it, and it also still got a segment violation.
> However, the code that caused the segment violation was still the old code.
> After I rebooted, then it worked OK.

Finally got there!  The behavior below really confuses me:

vlad:/usr/lib 5# ls -l libc.*
-r--r--r--  1 bin   bin    2041460 Aug 15 10:13 libc.a
lrwxrwxrwx  1 root  wheel       13 Aug 15 10:13 libc.so@ -> libc.so.12.17
lrwxrwxrwx  1 root  wheel       13 Aug 16 20:44 libc.so.12@ -> libc.so.12.17
-r--r--r--  1 bin   bin    1017399 Aug 15 10:13 libc.so.12.17
-r--r--r--  1 bin   bin    1017383 Aug 14 12:41 libc.so.12.17.XXX
vlad:/usr/lib 6# ldconfig
vlad:/usr/lib 7# ls -l libc.*
-r--r--r--  1 bin   bin    2041460 Aug 15 10:13 libc.a
lrwxrwxrwx  1 root  wheel       13 Aug 15 10:13 libc.so@ -> libc.so.12.17
lrwxrwxrwx  1 root  wheel       17 Aug 16 20:44 libc.so.12@ -> libc.so.12.17.XXX
-r--r--r--  1 bin   bin    1017399 Aug 15 10:13 libc.so.12.17
-r--r--r--  1 bin   bin    1017383 Aug 14 12:41 libc.so.12.17.XXX

libc.so.12.17.XXX is the libc with the old modf.S.  Notice how ldconfig
moved the libc.so.12 symlink to point to it, instead of libc.so.12.17?
I was just about to reply back to Michael saying whatever he did didn't
work for me.  Speaking of which, we still don't have the source to
ldconfig, do we?  My ldconfig binary is dated June 4, 1996, and the
ldconfig in the current source tree doesn't build on the pmax (it's part
of GNU ld, which doesn't have a mips directory in it's arch directory).

I'll now start up a compile of perl5, and have a look at the rename
problem Sean also mentioned.

Simon.