On Fri, 30 Oct 2026 12:53:44 GMT, Jorn Vernee <[email protected]> wrote:
>> Ah I see -- the native lib was already there. Still, we should make sure to >> test against a library compiled with some (C) optimizations > > The test fails without the fix. The native function is declared to accept > `int`, but we pass a smaller type. This essentially allows us to observe the > widening done by the bindings. I've reworked the test a bit to drop the dependency on the native library altogether. We always widen sub-int values to int in the binding recipe, so we just need to make sure this widening takes the signedness into account on platforms where it's required. Rather than relying on compiler optimizations, the test now inspects the value passed to the native side more directly. I've split the platforms that widen to long into a separate test as well. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/33417#discussion_r4425020814
