https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92131
--- Comment #13 from rguenther at suse dot de <rguenther at suse dot de> --- On Thu, 17 Oct 2019, jakub at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92131 > > --- Comment #12 from Jakub Jelinek <jakub at gcc dot gnu.org> --- > (In reply to Eric Botcazou from comment #11) > > > From what I can see, the weird + 0x7ffffffffffffffe is created when > > > extract_range_from_plus_minus_expr is called with PLUS_EXPR and VARYING > > > and > > > long int [-INF, e.7_8 + -1] ranges. > > > /* Build the symbolic bounds if needed. */ > > > adjust_symbolic_bound (min, code, expr_type, > > > sym_min_op0, sym_min_op1, > > > neg_min_op0, neg_min_op1); > > > adjust_symbolic_bound (max, code, expr_type, > > > sym_max_op0, sym_max_op1, > > > neg_max_op0, neg_max_op1); > > > in there. max_op0 is 9223372036854775807 (maximum of VARYING), and > > > max_op1 > > > is -1 (which is not the upper bound, just an offset against the symbolic). > > > combine_bound adds this into wmax, so 9223372036854775807 + -1 gives > > > 0x7ffffffffffffffe and that is what is used. > > > > This looks like the bug, i.e. combine_bound shouldn't be called on VARYING. > > Perhaps, but it would still be called for > long int [-9223372036854775805, +INF] + long int [-INF, e.7_8 + -1], for which > it would still end up with long int [-INF, e.7_8 + 9223372036854775806]. We cannot really add range bounds to symbolics because those bounds are not added in the IL and thus those adds are prone to overflowing. We can only do _1 = _2 + 1; with _2 = [ -1, _3 + 4] derive _1 = [ 0, _3 + 5] but not for _1 = _2 + _4; with _4 = [ 0, 1 ] derive _1 = [ -1, _3 + 5] Richard.
