On Thu, 15 May 2025 09:49:10 GMT, Per Minborg <[email protected]> wrote:
> This PR improves the robustness of tests on slow machines such as "zero"
> variants. The timeout was increased to 10 minutes for both tests. There are
> many combinations in `TestAccessModes`, so the test takes some time even on a
> fast machine. In `StdLibTest`, there are numerous native calls, and these can
> be much slower on "zero" machines. In the latter test class, I made some
> attempts to improve performance. On my M1 machine, the test was about 10%
> faster with the changes (but I only did one run, so this might be in the
> error margin).
test/jdk/java/foreign/StdLibTest.java line 56:
> 54: static final Linker abi = Linker.nativeLinker();
> 55:
> 56: private static final StdLibHelper STD_LIB_HELPER = new StdLibHelper();
The `STD_LIB_HELPER` is now a constant.
test/jdk/java/foreign/StdLibTest.java line 145:
> 143: static final class StdLibHelper {
> 144:
> 145: static final MethodHandle strcat =
> abi.downcallHandle(abi.defaultLookup().findOrThrow("strcat"),
Drive-by cleanup to blessed order.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/25248#discussion_r2090771644
PR Review Comment: https://git.openjdk.org/jdk/pull/25248#discussion_r2090769634
