#ReflectiveAccessByInstrumentationAgents

[Andrew Dinn]

On 06/05/16 08:59, Alan Bateman wrote:
> On 06/05/2016 08:32, Peter Levart wrote:
>> I think that not giving the agent such API just makes life for agent
>> writers harder but does not prevent them to break encapsulation in
>> arbitrary ways anyway. 
> 
> We've been over this a few times, it's on the issues list, but we want
> to see how far we can get without adding this method.

Ok, I'll be that guinea pig.

> On injecting helpers then if the agent can inject an initializer +
> helper method into any public class in the module then it has full
> control. It doesn't matter if it's a concealed package because the
> initializer (which the agent can trigger to execute without access) just
> needs to export the package to the agent.

Yes, that's probably the best way to restrict access if you assume the
agent is itself in a Jigsaw module. But I doubt all (or even most?)
agents will want to use Jigsaw. They certainly won't if they want to
implement functionality that is useful to pre- and post-JDK9 runtimes
since that will force the devs down the path of dual-source and
dual-produuct. That is a very unattractive proposition since it
complicates both development and deployment.

regards,


Andrew Dinn
-----------
Senior Principal Software Engineer
Red Hat UK Ltd
Registered in England and Wales under Company Registration No. 03798903
Directors: Michael Cunningham, Michael ("Mike") O'Neill, Eric Shander