[10] RFR(M): 8176506: C2: loop unswitching and unsafe accesses cause crash

Volker Simonis volker.simonis at gmail.com
Wed May 31 09:10:08 UTC 2017

On Mon, May 29, 2017 at 10:10 AM, Roland Westrelin <rwestrel at redhat.com> wrote:
> Hi Volker,
> Thanks for reviewing this again.
>> I think it's good that you now use SIGILL on all platforms. The change
>> looks good except the following minor nits I've already mentioned in
>> my first review. I'll leave it up to you if you want to fix them and
>> in the case you do there's no need for a new webrev.
> Sorry, I forgot about those. I will fix the comments.
>> test/compiler/unsafe/TestMaybeNullUnsafeAccess.java
>> - wouldn't it be safer to run the test with -Xbatch and
>> -XX:-UseOnStackReplacement for any case and to make it evident that
>> the test relays on the fact that test1() and test2() are both
>> compiled but not inlined ?
> Ok. Do you suggest adding a comment that states that test1/test2
> shouldn't be inlined?

No I suggest using:

 * @run main/othervm -Xbatch -XX:-UseOnStackReplacement

instead of:

 * @run main/othervm TestMaybeNullUnsafeAccess

in test/compiler/unsafe/TestMaybeNullUnsafeAccess.java

>> - you should also update the copyright on most files you've touched.
> I don't think I've updated a copyright in years. I thought it was ok to
> let bulk periodic copyright updates take care of that.
> Roland.

More information about the hotspot-compiler-dev mailing list