Help! GNU Make 3.58 won't compile under 3005

[Chip Salzenberg]

I am so confused.

A long time ago, in a galaxy far far away (named "3003"), I compiled
GNU Make 3.58.  It worked fine.  Everyone was happy.

Recently, we installed the 3005 update.  Now when I compile GNU Make,
it doesn't work.  Everyone is sad.

Tracking this error down has been an adventure.

When I compile without optimization, a simple "for" loop in main()
is executing all wrong.  The loop looks like this:

  register unsigned int i;

  /* ... */

  for (i = 0; environ[i] != 0; ++i)
    {
      register char *ep = environ[i];
      while (*ep++ != '=')
	;
      (void) define_variable (environ[i], ep - environ[i] - 1, ep, o_env, 1);
    }

A debugging printf() reveals that the value of "i" is getting trashed
during the subroutine call!  That problem should not be possible.

It gets worse.  When I compile GNU Make with the optimizer on, it dies
with the message "Killed" -- *unless* I run it with the "-d" flag, in
which case it is fine (if verbose).  I'm so confused!

I'd try to track down the bug myself, but without a "-S" flag for cc,
I can't verify the compilers output.  :-(

What's happening to my cc?!
-- 
Chip Salzenberg at Teltronics/TCT     <chip at tct.com>, <uunet!pdn!tct!chip>
          perl -e 'sub do { print "extinct!\n"; }   do do()'