On Thu, 12 Sep 2024 10:27:49 GMT, Raffaello Giulietti <rgiulie...@openjdk.org> wrote:
>> I see (with obvious adaptations and one more `case`). > > The implementation of the shift methods in `MutableBigInteger` has several > control flow paths, depending on whether a new `int[]` is needed, whether > `arraycopy()` can be used, etc. > > Relying on random tests is good if the "search space" (the number of control > flow paths) is too big for a systematic approach. But here I think it should > be possible to exercise all paths with a dozen or so well targeted tests, in > addition to the existing ones. > > These are white-box tests, so the internal structure of the algorithms is > supposed to be known and this knowledge should be exploited in trying to hit > the weak points, if there are. So, a test should be done for every possible path of computation of `MBI.leftShift()` method... ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/20008#discussion_r1756662312
