Tree VRP bug: internal compiler error: in build_int_cst_wide, at tree.c:886

Mon Mar 19 16:29:00 GMT 2007

On 3/19/07, Andrew Haley <aph@redhat.com> wrote:
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31264
>> This happens because we have a VIEW_CONVERT EXPR(double->long) that is
> used in a conditional. VRP tries to register an edge assertion for
> this expression:
>> <view_convert_expr 0x2aaaae08c6c0
> type <integer_type 0x2aaaadfee3c0 long public DI
> size <integer_cst 0x2aaaadfe19f0 constant invariant 64>
> unit size <integer_cst 0x2aaaadfe1a20 constant invariant 8>
> align 64 symtab 0 alias set -1 canonical type 0x2aaaadfee3c0 precision 64 min <integer_cst 0x2aaaadfe1930 -9223372036854775808> max <integer_cst 0x2aaaadfe1990 9223372036854775807>
> pointer_to_this <pointer_type 0x2aaaae00e300> chain <type_decl 0x2aaaadff2270 long>>
>> arg 0 <ssa_name 0x2aaaae0a8f00
> type <real_type 0x2aaaadffa6c0 double DF size <integer_cst 0x2aaaadfe19f0 64> unit size <integer_cst 0x2aaaadfe1a20 8>
> align 64 symtab 0 alias set -1 canonical type 0x2aaaadffa6c0 precision 64
> chain <type_decl 0x2aaaadff2d00 double>>
> visited var <parm_decl 0x2aaaadfe7960 D.774> def_stmt <phi_node 0x2aaaae0b2d00>
>> Which eventually works its way down to:
>> if (!has_single_use (op))
> {
> val = build_int_cst (TREE_TYPE (op), 0);
> register_new_assert_for (op, code, val, NULL, e, bsi);
> retval = true;
> }
>> and this fails with an ICE because TREE_TYPE (op) is real_type, and
> you can't build_int_cst with a real_type.
>> The simplest way of fixing this is simply to disallow
> VIEW_CONVERT_EXPR in this context. Comments?
>> Index: tree-vrp.c
> ===================================================================
> --- tree-vrp.c (revision 122839)
> +++ tree-vrp.c (working copy)
> @@ -3485,7 +3485,6 @@
> }
> else if (TREE_CODE (rhs) == NOP_EXPR
> || TREE_CODE (rhs) == CONVERT_EXPR
> - || TREE_CODE (rhs) == VIEW_CONVERT_EXPR
> || TREE_CODE (rhs) == NON_LVALUE_EXPR)
> {
> /* Recurse through the type conversion. */

It's indeed broken to look through VIEW_CONVERT_EXPRs here. The patch looks
obviously correct.
Thanks,
Richard.