This patch adds some assertions against sext (.., 0) and zext (..., 0). The former is undefined at the sext_hwi level and the latter is disallowed for consistency with the former.
Also, set_bit (rightly IMO) can't handle bit >= precision. For precision <= HOST_BITS_PER_WIDE_INT it would invoke undefined behaviour while for other precisions I think it would crash. A case with precision <= HOST_BITS_PER_WIDE_INT showed up in java (fix posted separately). Tested on x86_64-linux-gnu and powerpc64-linux-gnu. OK to install? Thanks, Richard Index: gcc/wide-int.h =================================================================== --- gcc/wide-int.h 2014-05-02 16:28:09.561842842 +0100 +++ gcc/wide-int.h 2014-05-02 16:44:04.015463718 +0100 @@ -2046,6 +2046,8 @@ wi::sext (const T &x, unsigned int offse unsigned int precision = get_precision (result); WIDE_INT_REF_FOR (T) xi (x, precision); + gcc_checking_assert (offset != 0); + if (offset <= HOST_BITS_PER_WIDE_INT) { val[0] = sext_hwi (xi.ulow (), offset); @@ -2065,6 +2067,8 @@ wi::zext (const T &x, unsigned int offse unsigned int precision = get_precision (result); WIDE_INT_REF_FOR (T) xi (x, precision); + gcc_checking_assert (offset != 0); + /* This is not just an optimization, it is actually required to maintain canonization. */ if (offset >= precision) @@ -2102,6 +2106,9 @@ wi::set_bit (const T &x, unsigned int bi WI_UNARY_RESULT_VAR (result, val, T, x); unsigned int precision = get_precision (result); WIDE_INT_REF_FOR (T) xi (x, precision); + + gcc_checking_assert (bit < precision); + if (precision <= HOST_BITS_PER_WIDE_INT) { val[0] = xi.ulow () | ((unsigned HOST_WIDE_INT) 1 << bit);