Have -fnew-exceptions actually use the runtime field, and not generate runtime checking code inside each handler.

Have -fnew-exceptions actually use the runtime field, and not generate
runtime checking code inside each handler.
THis works at low opt, but not with optimization at the moment.
Tue Jun 23 10:06:07 EDT 1998  Andrew MacLeod  (amacleod@cygnus.com)
	* eh-common.h (struct __eh_info): Remove coerced value field.
	* libgcc2.c (find_exception_handler): Don't set coerced_value field.
	* except.c (get_dynamic_handler_chain, get_dynamic_cleanup_chain): Use
	POINTER_SIZE instead of Pmode.
	(expand_start_all_catch): Call start_catch_handler() if we are not
	using new style exceptions.
1998-06-22  Andrew MacLeod  (amacleod@cygnus.com)
	* parse.y (function_try_block): Don't call start_catch_handler.
	* except.c (call_eh_info): Remove coerced field from declaration.
	(build_eh_type_type_ref): New function to create an address of a
	rtti function for the new style exception tables.
	(expand_start_catch_block): Split function, this contains the
	common part.
	(process_start_catch_block_old): New function to perform the rest
	of expand_start_catch_block under old style exceptions.
	(process_start_catch_block_old): New function to perform the rest
	of expand_start_catch_block under new style exceptions.
	(expand_end_catch_block): Only pop the false label off the stack under
	the old style of exceptions.
	* semantics.c (finish_try_block): Don't call start_catch_handler.
	* exception.cc (struct cp_eh_info): Add original_value field.
	(__cplus_type_matcher): Perform type matching on the original exception
	value, and if we have a match, set the current value.
	(__cp_push_exception): Set the original expcetion value.

From-SVN: r20671