On Fri, 23 Oct 2020 14:26:07 GMT, Maurizio Cimadamore <mcimadam...@openjdk.org> wrote:
>> It currently fails on x86_32 with `java.lang.IllegalStateException: >> Misaligned access at address: 12`. I checked that once JDK-8254162 >> integrates, that issue is gone. Until then, having clean x86_32 testing is >> beneficial for other work. >> >> Testing: >> - [x] Affected test on Linux x86_64 (still passes) >> - [x] Affected test on Linux x86_32 (now ignored) > > I'm not sure I get as to why the test fails on 32 bits - in the sense that it > does not seem to rely on platform specific assumptions (at least on a quick > look). On top of my head I'm not aware of what might have caused this to > start working again in JDK-8254162, so maybe some more investigation is > required? > > At the very least we should only disable the problematic spliterator test, > which is, I believe "testSegmentSpliterator" - and leave the other running, > which is useful, as this is an important stress test for spliterators. Ok - I understand at least why JDK-8254162 fixed it - JDK-8254162 uses a new API to access memory (MemoryAccess) which disables alignment checks, so that's why the issue doesn't happen. But that means that there could be a latent issue in the test. Looking deeper. ------------- PR: https://git.openjdk.java.net/jdk/pull/836
