[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-serviceability-dev
Subject:    RFR (XS): 8005592: ClassLoaderDataGraph::_unloading incorrectly defined as nonstatic in vmStructs
From:       mikael.vidstedt () oracle ! com (Mikael Vidstedt)
Date:       2012-12-29 1:26:26
Message-ID: 50DE46C2.8070406 () oracle ! com
[Download RAW message or body]


Thanks for the review!

And, yes, I asked myself that same question about what should actually 
be listed in vmstructs when I found this. The only criterion I know is 
that if the SA uses it needs to be there. However, I think there's a 
fairly large bunch of data and types that are in the vmstructs arrays 
which are not used by the SA at all.

We know we don't like the SA + vmStructs "mess" and want something else 
to replace it. Meanwhile I guess fixing what's there will have to do.

Cheers,
Mikael

On 12/28/2012 3:40 PM, David Holmes wrote:
> Mikael,
>
> This looks good to me.
>
> It does make me wonder (not for the first time) exactly what criteria 
> determines which fields from which classes appear in VMStructs?
>
> David
>
> On 29/12/2012 8:24 AM, Mikael Vidstedt wrote:
>>
>> Please review the following change.
>>
>> Background:
>>
>> The _unloading field is a static field in ClassLoaderDataGraph (in
>> classLoaderData.hpp) and should therefore be defined using static_field,
>> as opposed to nonstatic_field, in vmStructs.
>>
>> Apart from changing from nonstatic_field to static_field I also added an
>> assert in the CHECK_NONSTATIC_VM_STRUCT_ENTRY macro to make sure any
>> field offsets are within the bounds of the corresponding structs. The
>> assert triggers for _unloading before the change to static_field.
>>
>> The change passes JPRT.
>>
>> http://cr.openjdk.java.net/~mikael/8005592/webrev.00/
>>
>> Thanks,
>> Mikael
>>


[prev in list] [next in list] [prev in thread] [next in thread]