[aarch64-port-dev ] jdk8: random null pointer exceptions from javac

[Andrew Dinn]

On 04/11/15 17:18, Edward Nevill wrote:

Ok, so let me try to summarize what you are saying to be sure I have
understood correctly:

1. These changes apply to jdk8 (because the bug has been seen in jdk8).

2. You are changing the definition of the output allocation operand
iRegNNoSP to use a register class which excludes rfp.

3. This stops it being treated like iRegINoSp which allows integer
outputs to be generated in rfp.

4. It also puts iRegNNoSP on a par with iRegPNoSP which also uses a
register class that excludes rfp.

5. And the reason for doing this is that you think the GC fails to find
oops when they are saved in rfp.

Is that correct? Is so then I am still unsure why this same problem does
not arise on jdk9.

The definition for iRegNNoSp in jdk9 uses a register class which
conditionally includes rfp iff PreserveFramePointer is true. So, in jdk9
narrow oops can be stored in rfp when PreserveFramePointer is true.

Why do we not see the same problem on jdk9?

Is there a disparity in the GC behaviour which means this works on jdk9?

Or is it just that the default for PreserveFramePointer is false so we
never see the problem?

regards,


Andrew Dinn
-----------