Hi Chris,
Thanks for fixing this. It looks good to me.
- Ioi
On 4/9/20 9:44 PM, Chris Plummer wrote:
Hello,
Please review the following:
https://bugs.openjdk.java.net/browse/JDK-8235220
http://cr.openjdk.java.net/~cjplummer/8235220/webrev.00
First off, thanks to Ioi for tracking this one down and proposing the
fix.
The test is executing the clhsdb "scanoops" command, which scans the
specified heap address range looking for objects of the specified type
(or all objects if no type is specified). The test determines the
address range by first using clhsdb "universe" to get the start and
end of the eden space. scanoops then iterates over this entire range,
calling RobustOopDeterminator.oopLooksValid() on each oop as it
iterates over the range. The problem is eventually you'll end up past
the initialized part of the heap. That causes
VirtualBaseConstructor.findDynamicTypeForAddress() to return null,
resulting in throwing a WrongTypeException.
RobustOopDeterminator.oopLooksValid() should catch this exception and
return false when it happens.
The CR mentions a few different failure modes. This only fixes the
WrongTypeException failure. The NullPointerException is only on OSX
and I believe is the same as JDK-8241158 [1], which happens when
dumping the heap, so it seems probable that it is also turning up when
scanning part of the heap. I think this bug was preventing us from
ever seeing the WrongTypeException on OSX. Then there also JDK-8230731
[2] on Windows, which seems to prevent ever seeing the
WrongTypeException on Windows.
[1] https://bugs.openjdk.java.net/browse/JDK-8241158
[2] https://bugs.openjdk.java.net/browse/JDK-8230731
thanks,
Chris
