[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
Re: [ptp-dev] debug design
|
> We need to think about how the org.eclipse.ptp.debug.external component is
> going to interface with the external debug manager. Currently you're
> starting the external debugger at the end of a pipe, sending commands and
> reading responses from a command line interface (right?). I would like to
> get away from this model, since it is complex, inefficient, and very
> susceptable to minor changes in the CLI. I see two alternatives:
>
Agree with you on this one.... I did it that way because the only
reusable code that is readily available is the GDB/MI code (after all,
it's only for testing)
> 1. Start the external debugger as a separate process, then use TCP/IP to
> communicate with it. The downside with this approach is the need to
> develop a protocol, and a Java implementation of the protocol. However,
> you could probably use the classic guard socket protocol with some
> modifications. The upside is that TCP/IP support in Java is good and easy
> to use.
>
> 2. Use JNI to call a C library interface to the external debug manager.
> The downside with this approach is the complexity of JNI, and the need to
> deal with threading issues, etc. The upside is that you can call an API
> directly and avoid the overhead of packing/unpacking messages.
>
> Comments? Preferences?
>
I prefer the first one.... I think it's better....
I do believe what you mean with TCP/IP is a full blown,
connection-oriented socket ?
(rather than, for example, connectionless XML-RPC ?
or a high-level framework like beep http://www.beepcore.org/ ?)
Donny
> Greg
> _______________________________________________
> ptp-dev mailing list
> ptp-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/ptp-dev
>
