https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120048
--- Comment #5 from Andrew Macleod <amacleod at redhat dot com> --- It is trapping when asking for the type() of an UNDEFINED range. Looking up the call chain: Breakpoint 2, ipcp_store_vr_results () at /gcc/master/gcc/gcc/ipa-cp.cc:6366 6366 tmp.update_bitmask (bm); (gdb) p tmp.dump(stderr) [irange] short int [0, 0] so tmp starts with a range of [0,0] and then an attempt is made to apply a bitmask (gdb) p bm.dump(stderr) MASK 0xfffe VALUE 0x1$7 = void The bitmask that is being set indicates that we KNOW bit 0 is the value 1. This is incongruent a range of [0,0] and the new code which removes the lower bit ranges of a specified bitmask says [0,0] is not a value range for this bitmask, and we end up making tmp UNDEFINED. It seems like the bitmask is being incorrectly created for this range? 6358 value_range tmp = plats->m_value_range.m_vr; 6359 tree type = ipa_get_type (info, i); 6360 irange_bitmask bm (wide_int::from (bits->get_value (), 6361 TYPE_PRECISION (type), 6362 TYPE_SIGN (type)), 6363 wide_int::from (bits->get_mask (), 6364 TYPE_PRECISION (type), 6365 TYPE_SIGN (type))); 6366 tmp.update_bitmask (bm); 6367 ipa_vr vr (tmp); 6368 ts->m_vr->quick_push (vr); Why is it even creating a bitmask?
