Bytecode generation, Source code mappings, JCov, Future

[Alex Rau]

On 22.04.2008, at 00:16, Robert Field wrote:

> I ported JCov to java.lang.instrument (a fully Java API) not to JVM  
> TI (a native API).  And it is now in active use internal to Sun.   
> There were plans to make it open source, but I haven't heard  
> anything about this being done -- though I don't track the project  
> anymore.
>
> I terms of character position, this is something that has been  
> actively requested for years -- it would need to be implemented by  
> javac.  This is needed for tools like debuggers, profilers,  
> coverage analysis, etc.  But it such an interface has not been  
> formally defined.  javac does have or did have an unofficial mode  
> for generating character position info (for JCov, as you mention),  
> but I am not familiar with it or its status.  Compiler folk...
>
> -Robert
>

Thanks for clarification ! That underlines my impression that the  
character position related stuff embedded in the existing "JCov"  
implementation is *not* superseded by any JVMTI (or as you said  
java.lang.instrument) port and that getting character positions for  
post-compilation tools is a useful feature.

My patch sent to compiler-dev at openjdk.java.net yesterday provides a  
basic interface for making character positions in a simple format  
accessible to "user-space".

Best regards,


  Alex


>
> Uday Dhanikonda wrote:
>> Alex
>>
>> There was an effort by Robert recently to rewrite JCov using JVM TI.
>>
>> Robert can you comment on this.
>>
>> Thanks
>> Uday
>>