This fixes PR54031. While technically the source is invalid
(a pointer is not aligned as it should) it wasn't my plan to
exploit this with the get_object_alignment change. In fact
it is CCPs wrongdoing that it uses get_object_alignment_1
instead of get_pointer_alignemnt_1 when looking at a pointer.
Bootstrapped on x86_64-unknown-linux-gnu, testing in progress.
Richard.
2012-07-20 Richard Guenther <rguent...@suse.de>
PR tree-optimization/54031
* tree-ssa-ccp.c (get_value_from_alignment): Use
get_pointer_alignment_1.
Index: gcc/tree-ssa-ccp.c
===================================================================
*** gcc/tree-ssa-ccp.c (revision 189704)
--- gcc/tree-ssa-ccp.c (working copy)
*************** get_value_from_alignment (tree expr)
*** 510,516 ****
gcc_assert (TREE_CODE (expr) == ADDR_EXPR);
! get_object_alignment_1 (TREE_OPERAND (expr, 0), &align, &bitpos);
val.mask
= double_int_and_not (POINTER_TYPE_P (type) || TYPE_UNSIGNED (type)
? double_int_mask (TYPE_PRECISION (type))
--- 510,516 ----
gcc_assert (TREE_CODE (expr) == ADDR_EXPR);
! get_pointer_alignment_1 (expr, &align, &bitpos);
val.mask
= double_int_and_not (POINTER_TYPE_P (type) || TYPE_UNSIGNED (type)
? double_int_mask (TYPE_PRECISION (type))
