https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113226
--- Comment #2 from Patrick Palka <ppalka at gcc dot gnu.org> --- (In reply to Patrick Palka from comment #1) > Huh, how bizarre. > > > i == 1, j == -100, i*j == 4294967196, max_type(i) == 1, max_type(i)*j == > > -100 > > Here i and j are just ordinary 'long long', so I don't get why i*j is > 4294967196 instead of -100? Everything else, in particular that int64_t(max_type(i)*j) is -100, seems correct/expected to me. FWIW that expression computes the product of the corresponding promoted/sign-extended 65-bit precision values, and the overall check is analogous to int32_t i = 1, j = -100; assert (int64_t(i*j) == int64_t(i)*j); except the two precisions are 64/65 bits instead of 32/64 bits. (When shorten_p is true, the overall check is analogous to assert (i*j == int32_t(int64_t(i)*j)) instead.)
