The BIT_AND_EXPR handling in extract_range_from_binary_expr_1 was using value_range_constant_singleton without first checking whether the range was a constant. The earlier handling was correctly guarded:
/* If either input range contains only non-negative values we can truncate the result range maximum to the respective maximum of the input range. */ if (int_cst_range0 && tree_int_cst_sgn (vr0.min) >= 0) wmax = wi::min (wmax, vr0.max, TYPE_SIGN (expr_type)); if (int_cst_range1 && tree_int_cst_sgn (vr1.min) >= 0) wmax = wi::min (wmax, vr1.max, TYPE_SIGN (expr_type)); so this patch uses the same guards again. Existing tests showed the need for this once polynomial constants are added. Tested on aarch64-linux-gnu, x86_64-linux-gnu and powerpc64le-linus-gnu. OK to install? Richard 2017-09-20 Richard Sandiford <richard.sandif...@linaro.org> Alan Hayward <alan.hayw...@arm.com> David Sherwood <david.sherw...@arm.com> gcc/ * tree-vrp.c (extract_range_from_binary_expr_1): Check int_cst_rangeN before calling value_range_constant_singleton (&vrN). Index: gcc/tree-vrp.c =================================================================== --- gcc/tree-vrp.c 2017-09-20 13:04:29.932873038 +0100 +++ gcc/tree-vrp.c 2017-09-20 13:07:32.589667751 +0100 @@ -2930,9 +2930,11 @@ extract_range_from_binary_expr_1 (value_ = wi::set_bit_in_zero (TYPE_PRECISION (expr_type) - 1, TYPE_PRECISION (expr_type)); if (!TYPE_UNSIGNED (expr_type) - && ((value_range_constant_singleton (&vr0) + && ((int_cst_range0 + && value_range_constant_singleton (&vr0) && !wi::cmps (vr0.min, sign_bit)) - || (value_range_constant_singleton (&vr1) + || (int_cst_range1 + && value_range_constant_singleton (&vr1) && !wi::cmps (vr1.min, sign_bit)))) { min = TYPE_MIN_VALUE (expr_type);