Unwinding failure

[Mark Mitchell]

Richard --

  According to my binary search, it looks like this change:

Sun Apr 25 12:30:50 1999  Richard Henderson  <rth@cygnus.com>

	* calls.c (emit_call_1): New arg rounded_stack_size; update callers.
	Update pending_stack_adjust based on this value.
	(compute_argument_block_size): Include pending_stack_adjust in
	PREFERRED_STACK_BOUNDARY alignment.
	* function.c (assign_parms): Don't round to PREFERRED_STACK_BOUNDARY.

is responsible for the current failures of g++.eh/unwind1.C on
x86-linux-gnu.  I know that you checked in:

Sun Apr 25 13:06:13 1999  Richard Henderson  <rth@cygnus.com>

	* function.c (assign_parms/STACK_BYTES): Revert last change,
	and that of 19 Nov.

but that didn't fix this problem.  Perhaps something in the
emit_call_1 changes is not propogating to the DWARF 2 debugging stuff?

-- 
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com