Hi Paul,
On 26/11/2015 7:55 PM, Paul Sandoz wrote:
Hi,
This is a request for an optimistic review to fork the sun.misc.Unsafe and
jdk.internal.misc.Unsafe native method tables so that we can evolve the latter
e.g. for VarHandles unsafe work
http://cr.openjdk.java.net/~psandoz/jdk9/JDK-8143628-unsafe-native-jdk/
http://cr.openjdk.java.net/~psandoz/jdk9/JDK-8143628-unsafe-native-hotspot/
Code changes seem fine.
Tests .... don't think we use testng with any hotspot tests so not sure
about that part. Also wondering how these tests will access jdk.internal
once we're using the modular image?
Thanks,
David
I would like to go through hs-comp since this patch blocks VarHandles work that
will also go through hs-comp.
Access tests for both Unsafes are included. The tests are generated using the
simple regex-based stream preprocessor (Spp.java) that is also used by nio and
it follows the same approach as the nio tests where the generated test code,
the test template, and the script to execute the template are in the repo.
Core-libs is included because I have added an opt-in command line option (-nel)
to Spp.java to not preserve line numbers between the template and generated
source code, as sometimes the additional empty lines can be a distraction when
browsing the source. When this option is not used it should not result in any
changes in output. I also fixed up the TEXT regex pattern so embedded variables
work within generic type variable declarations (required for VarHandles which
also uses templates to generate source code).
The review request is optimistic since this patch is blocked on:
https://bugs.openjdk.java.net/browse/JDK-8143628
C1 LinearScan asserts when compiling two back-to-back CompareAndSwapLongs
which causes test failures on 32 bit fastdebug builds (linux and windows). I am
reluctant to modify the tests to workaround (it’s good they found a bug that
goes way back to at least 7u1!), and i am hopeful it can be fixed soon (since
Roland has found the cause).
Paul.
