On Wed, Sep 18, 2019 at 8:49 AM Richard Sandiford
<richard.sandif...@arm.com> wrote:
>
> loc_list_for_tree_1 and add_const_value_attribute currently ICE
> on POLY_INTs. loc_list_for_tree_1 can do something sensible but
> add_const_value_attribute has to punt, since the constant there
> needs to be a link-time rather than load-time or run-time constant.
>
> This is tested by later SVE patches.
>
> Tested on aarch64-linux-gnu with SVE (with and without follow-on
> patches) and x86_64-linux-gnu. OK to install?
OK.
Richard.
> Richard
>
>
> 2019-09-18 Richard Sandiford <richard.sandif...@arm.com>
>
> gcc/
> * dwarf2out.c (loc_list_from_tree_1): Handle POLY_INT_CST.
> (add_const_value_attribute): Handle CONST_POLY_INT.
>
> Index: gcc/dwarf2out.c
> ===================================================================
> --- gcc/dwarf2out.c 2019-09-17 15:27:11.374066662 +0100
> +++ gcc/dwarf2out.c 2019-09-18 07:47:42.297132785 +0100
> @@ -18568,6 +18568,24 @@ loc_list_from_tree_1 (tree loc, int want
> }
> break;
>
> + case POLY_INT_CST:
> + {
> + if (want_address)
> + {
> + expansion_failed (loc, NULL_RTX,
> + "constant address with a runtime component");
> + return 0;
> + }
> + poly_int64 value;
> + if (!poly_int_tree_p (loc, &value))
> + {
> + expansion_failed (loc, NULL_RTX, "constant too big");
> + return 0;
> + }
> + ret = int_loc_descriptor (value);
> + }
> + break;
> +
> case CONSTRUCTOR:
> case REAL_CST:
> case STRING_CST:
> @@ -19684,6 +19702,7 @@ add_const_value_attribute (dw_die_ref di
> case MINUS:
> case SIGN_EXTEND:
> case ZERO_EXTEND:
> + case CONST_POLY_INT:
> return false;
>
> case MEM:
