On Wed, Dec 17, 2025 at 08:29:17PM +0000, Jonathan Wakely wrote: > One reason for hoisting this class out of the _Select_uint_least_t > template was that it should be easier to add proper tests for the > operators now. I didn't do that as part of this fix, because the > priority was just to fix the failing tests, and because the tests for > std::uniform_real_distribution and std::linear_congruential_engine > contain checks that the results are as expected, which confirms that > the arithmetic is giving the expected answers for the current uses of > the class.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119595#c2 contains my quick and hackish testcase for that int128.h implementation, I was basically testing the class on x86_64 against unsigned __int128 using pseudo-random inputs for each of the operations and comparing whether they computed the same result. Jakub
