14 GCC HEAD regressions, 5 new, with your patch on 2004-04-05T12:48:20Z.

[Geoff Keating]

It looks like these failures:

> The new failures are:
> native g++.sum g++.dg/ext/altivec-1.C
> native gcc.sum gcc.dg/altivec-2.c
> native gcc.sum gcc.dg/altivec-4.c
> native gcc.sum gcc.dg/altivec-7.c
> native gcc.sum gcc.dg/altivec-8.c

are due to this patch:

> +2004-04-05  Richard Kenner  <kenner@vlsi1.ultra.nyu.edu>
> +
> +	* c-decl.c (build_compound_literal): Use TYPE_READONLY.
> +	* emit-rtl.c (set_mem_attributes_minus_bitpos): Likewise.
> +	* objc/objc-act.c (adorn_decl, gen_declspecs): Likewise.
> +	* c-typeck.c (decl_constant_value): Don't access DECL_INITIAL of a
> +	PARM_DECL.
> +	* calls.c (flags_from_decl_or_type): Use TYPE_READONLY and do so only
> +	for a type.
> +	* print-tree.c (print_node): Properly handle side-effects, readonly,
> +	and constant flags.
> +	* tree.c (build1_stat, build_expr_wfl): Only look at TREE_SIDE_EFFECTS
> +	and TREE_CONSTANT if not a type.
> +	* tree.h (IS_NON_TYPE_CODE_CLASS): New macro.
> +	(IS_EXPR_CODE_CLASS): Write 'E', not 'e'.
> +	(NON_TYPE_CHECK): New macro.
> +	(TREE_SIDE_EFFECT, TREE_READONLY, TREE_CONSTANT: Add check.


-- 
- Geoffrey Keating <geoffk@geoffk.org>