Request for review: 7174978: NPG: Fix bactrace builder for class redefinition

serguei.spitsyn at serguei.spitsyn at
Tue Dec 11 14:31:19 PST 2012


It looks good in general.
Just some questions below.


1339 void java_lang_Throwable::mark_on_stack(oop throwable) {
            . . .
1352       if (method == NULL) return;

  Would it be more safe to continue instead of return?
    1352       if (method == NULL) continue;


   63 void Backtrace::do_unloading() {

  I guess, this can be called at a safepoint only.
  Would it make sense to place a comment or an assert?

I see you already created a new unit test for this:


On 12/10/12 1:15 PM, Coleen Phillimore wrote:
> I have updated this webrev to include cleanups suggested by John Rose 
> for the anonymous class fix.   Please review before I add more to this!!
> open webrev at
> bug link at
> Thanks,
> Coleen
> On 12/05/2012 02:23 PM, Coleen Phillimore wrote:
>> Summary: Save the set of backtraces to use for on stack method 
>> walking for redefine classes.
>> I also moved metadataOnStackMark class to it's own file because it's 
>> not only used for redefine classes.   Some metadata can be 
>> individually deallocated (eg. the Method* created in the relocator).
>> open webrev at
>> bug link at
>> Ran test that will be added to the jdk/tests in 
>> java/lang/instrument/ (to be checked in 
>> separately).
>> thanks,
>> Coleen

-------------- next part --------------
An HTML attachment was scrubbed...

More information about the hotspot-runtime-dev mailing list