Hi Raffaello, > On Aug 7, 2020, at 1:03 AM, Raffaello Giulietti > <raffaello.giulie...@gmail.com> wrote: > > my previous explanation is (luckily) wrong. > > The reason the new tests pass on the old source is because the underlying > input stream correctly throws IOOBE which is catched by the tests.
OK. > There's one corner case that would not pass, however, so I added > testNegativeOffsetZeroLength() to exercise that. It passes on the new source > but fails on the old. > > I also added two additional corner cases for too large combinations of offset > and length. This patch looks good. I am running it through our CI tests but I don’t expect to see any problems. If the test run succeeds aside from any failures unrelated to this code then I can push this. Thanks, Brian
