Information about possible JDB enhancements

[Chris Plummer]

Hi Jakob,

You are correct in assuming that in general jdb is low priority. We 
mostly rely on it for testing jdi (there are quite a few jdb tests we 
maintain mainly for this purpose), and also for doing some targeted 
experimenting that might not be quite as easy to do with a full IDE. We 
do welcome bug fixes and enhancements, but we need to weigh their cost 
w.r.t. the time it takes to understand and review the changes, and also 
consider the risk of breaking something. I don't see an issue with a 
simple change like your empty command handling suggestion as being a 
problem. However, conditional breakpoints seems like a stretch. It's not 
a JDI feature, so will likely take quite a bit of work to support in 
jdb. I don't see this as something we'd want to spend time on, but 
others might disagree. Certainly it's open to discussion.

BTW, it looks like conditional expressions were at one point a 
consideration for JDWP (and therefore could have been leveraged by JDI). 
If you look in the EventRequest.set in the JDWP spec:

https://docs.oracle.com/javase/10/docs/specs/jdwp/jdwp-protocol.html#JDWP_EventRequest_Set

You'll find that modkind 2 is the filter for "Conditional on 
expression". However, the argument type is exprID, which is not defined 
anywhere, and the spec just says "For the future". The debug agent 
actually accepts the filter and saves it along with the exprID, but when 
it comes time to apply filters, it just ignores it:

             case JDWP_REQUEST_MODIFIER(Conditional):
/***
                 if (...  filter->u.Conditional.exprID ...) {
                     return JNI_FALSE;
                 }
***/
                 break;


thanks,

Chris

On 6/18/21 12:57 PM, daniel.daugherty at oracle.com wrote:
> Forwarding to serviceability-dev at ...
>
> Dan
>
>
> On 6/18/21 3:31 PM, Jakob Cornell wrote:
>> Hi all,
>>
>> I'm hoping to become an OpenJDK contributor in order to make some 
>> enhancements to JDB, which is my go-to tool for Java debugging. As an 
>> example, JDB currently ignores empty commands, and I hope to make a 
>> change so that entering an empty command results in the previous 
>> command being rerun (as users coming from GDB will expect).  I would 
>> also be interested in investigating implementation of conditional 
>> breakpoints, assuming JDI supports this (I'm not familiar with it).
>>
>> I get the sense that these changes may not be priorities for the 
>> OpenJDK development community (or perhaps may not even be desired), 
>> and so I'm writing to try to get a sense of whether I'd be able to 
>> find someone willing to sponsor changes like these.  If the answer is 
>> no then I shouldn't bother to look into creating patches.  Can 
>> anybody speak to the feasability of this?  Also, is there a more 
>> targeted mailing list I should use?
>>
>> Thanks,
>> Jakob
>