class file reconstitutor / LVT

[Coleen Phillimore]

Hi Thomas, I was going to answer over the weekend but had to check on 
something.

I like the idea of cutting down the patch size with the LVT change and 
(hopefully) the RC_TRACE change a lot.

To check things into openjdk there are new rules (which I can't find 
exactly) which says you have to have someone check in something for you 
8 times, then you're a committer and then someone has to check in 
something 32 times before you're a reviewer.

Anyway, I think you need to send Mark.Reinhold at oracle.com your openjdk 
user name and some secret ssh code so that you can check things in.    
And then you get on this list:  http://db.openjdk.java.net/people   Mark 
probably wont' respond right away because he's at JavaOne.

In the meantime, you can send out a code review as a patch file to: 
hotspot-dev at openjdk.java.net

There are ways to get webrevs out to the openjdk list but you have to 
have the secret password and name first.

Thanks,
Coleen

On 10/1/2011 5:14 PM, Thomas Wuerthinger wrote:
> Coleen,
>
> could you please brief me on how to proceed on this? What is the 
> required process for pushing the patch for the LVT bug fix to the 
> OpenJDK repo. Should this be done by you or can I do it somehow 
> myself? This small and simple patch might be the right way to get me 
> introduced to the process.
>
> Cutting down on the final patch size is probably a good idea (if it is 
> not a too painful process to get a smaller changeset in). So we might 
> consider gradually bringing the changes to other parts of the VM in 
> place before doing the final swap of the jvmtiRedefineClasses.cpp/hpp 
> files.
>
> - thomas
>
>
> On 10/1/11 7:46 PM, Daniel D. Daugherty wrote:
>> Makes sense to me...
>>
>> For the LVT bug, here is a bug that can be used:
>>
>> 7064927 4/3 retransformClasses() does not pass in LocalVariableTable 
>> of a method
>>
>> to track the issue...
>>
>> Dan
>>
>>
>> On 10/1/11 7:17 AM, Thomas Wuerthinger wrote:
>>> Coleen, Dan,
>>>
>>> I'm currently doing some further patch clean up. During that I found 
>>> the changes to jvmtiClassFileReconstitutor that are basically just 
>>> the addition of the LVT (local variable table) to the class file 
>>> reconstitutor and resolve the comment "FIXME: for now no LVT" (patch 
>>> in the attachment). Does it make sense to commit things like that 
>>> separately in advance in order to reduce the size of the final 
>>> enhanced class redefinition patch?
>>>
>>> @Coleen: It could also make sense to create a separate patch that 
>>> replaces the TRACE_RANGE with the new tracing, because this would 
>>> significantly reduce the number of touched files of the full class 
>>> redefinition patch.
>>>
>>> - thomas
>