[Bug 1710] Crash SIGSEGV in C2 Compilerthread when compiling one specific method

[bugzilla-daemon at icedtea.classpath.org]

http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=1710

--- Comment #9 from Stefan Huehner <stefan at huehner.org> ---
A fix for this issue is now available it seems.

If you check this jdk7u hotspot commit:
http://hg.openjdk.java.net/jdk7u/jdk7u/hotspot/rev/7801eaea32d2

8038785: hot workaround fix for a crash in C2 compiler at Node::rematerialize
Summary: This fix will introduce a performance regression. It should be
reverted after proper fix will be found.

As i am urgently looking for that fix to get fixed in openjdk-6 i did a
test-build of a patched icedtea 1.13.1 with that fix.

Good news is that i could not reproduce the bug (using the reproducer attached
here) with that patched build.

Note: The jdk7u patch does not apply perfect to openjdk-6.

The patch does delete 2 hunks of code and 1 line is different in that deleted
part.

I changed that patch to just delete the same block of code and the resulting
code does compile fine and runs my reproducer without crashing.

Note that i did my testing using the ubuntu 12.04 openjdk-6 source package
which runs some big (slow) testsuite during the build also.

See modified diff attached here.

Also the 1 line difference i mentioned above is the following:
--          RegionNode *r = new (C) RegionNode(edges+1);
+-          RegionNode *r = new (C, edges+1) RegionNode(edges+1);

First line is the one in jdk7, second the one in jdk6.

Could someone with some hotspot knowledge please check if just deleting both
hunks is still the right thing to do in openjdk-6?

Lastly i would really like to see this fixed (and published in ubuntu-12.04)
before end of this month, as in openjdk-6 it is a regression introduced in the
6b27 .. 7b30 timeframe.

If i can help with any work related to that just let me know.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/distro-pkg-dev/attachments/20140414/4471630f/attachment.html>