Question on status of a few bugs in OpenJDK 6
Vladimir Kozlov
vladimir.kozlov at oracle.com
Thu Mar 8 10:55:40 PST 2012
> Could you please do me a favor and review "7032314: Allow to generate CallLeafNoFPNode in IdealKit",
> see if it's safe to be applied it to HS20? Just this one, independent of other loopPredicate-related fixes.
Yes, it is safe if you mean next changes:
http://cr.openjdk.java.net/~kvn/7032314/webrev
I don't see and remember any bugs related to this so it should be safe. It was just code refactoring.
Vladimir
On 3/8/12 10:31 AM, Krystal Mok wrote:
> Hi Vladimir,
>
> Inline:
>
> On Fri, Mar 9, 2012 at 2:04 AM, Vladimir Kozlov <vladimir.kozlov at oracle.com <mailto:vladimir.kozlov at oracle.com>> wrote:
>
> Hi, Kris
>
> As I know, only 2 bugs from these affect JDK6.
>
> Got it. That's reassuring, thanks a lot :-)
>
> First is 5091921 but the fix is too big to backport it to jdk6 and it also rely on new code in HS21 VM which is only
> present in JDK7
>
> So it was too big...I did see that it relies on a new loopPredicate.cpp file, with later patches to that file, too.
>
> (we stop delivering new VM into jdk6 updates after HS20 VM).
>
> Yes, I know. We're still using JDK6 in production, and somehow we need to patch the VM a little bit for a couple of
> features, so we've been using HSX20 as the base.
>
> Second bug 7044738 was backported into product jdk6u28. I don't know if OpenJDK6 community took it.
>
> As of now, nope. HS20-b12 in OpenJDK 6 update 24 doesn't have that yet.
>
> The rest is bug tail of 5091921 and related changes so they do not affect jdk6.
>
> Could you please do me a favor and review "7032314: Allow to generate CallLeafNoFPNode in IdealKit", see if it's safe to
> be applied it to HS20? Just this one, independent of other loopPredicate-related fixes.
>
> Regards,
> Vladimir
>
> Thanks,
> Kris Mok
>
>
> On 3/8/12 3:37 AM, Krystal Mok wrote:
>
> Hi all,
>
> Guess I'm asking on the wrong list, but I don't know which list I should be sending this to for OpenJDK 6-only
> issues.
> The jdk6-dev list is mostly silent now, doesn't seem to be the right place, neither.
> Maybe I should be asking the sustaining team?
>
> I was looking at what changes went into recent OpenJDK 6 releases. Turns out OpenJDK 6 build 23 is the closest
> version
> to JDK 6 update 25, and OpenJDK 6b24 is the closest version to 6u30. Reading through the release changes [1][2],
> though,
> tells that:
> 1. The loop optimization bug that crashes the VM is fixed:
> 7070134: Hotspot crashes with sigsegv from PorterStemmer
> 2. A few other related bugs aren't included in OpenJDK 6 yet, as of now:
> 5091921: Sign flip issues in loop optimizer
> 7052494: Eclipse test fails on JDK 7 b142
> 7004535: Clone loop predicate during loop unswitch
> 7068051: SIGSEGV in PhaseIdealLoop::build_loop___late_post on T5440
> 7044738: Loop unroll optimization causes incorrect result
> 7077439: Possible reference through NULL in loopPrdicate.cpp:726
>
> I suppose some of these fixes are actually in the product JDK 6u29+ already, like 7044738.
>
> So, questions:
> * Could anyone help identify whether these bugs affect OpenJDK 6 or not?
> * If they do, what are the plans for porting the fixes back to OpenJDK 6?
> * And what is the general plan for future maintence of OpenJDK 6?
> * Is there any place in the public that we can track what's going to be backported into OpenJDK 6?
> * Will it sort of EOL along with the product JDK 6?
>
> Thanks,
> Kris Mok
>
> [1]: https://blogs.oracle.com/kto/__resource/OpenJDK6-B23-Changes.__html
> <https://blogs.oracle.com/kto/resource/OpenJDK6-B23-Changes.html>
> [2]: https://blogs.oracle.com/__openjdk6releases/resource/__OpenJDK6-B24-Changes.html
> <https://blogs.oracle.com/openjdk6releases/resource/OpenJDK6-B24-Changes.html>
>
>
More information about the hotspot-dev
mailing list