| Re: [cdt-core-dev] Proposed Launch Changes |
Hi!
I would like to hold off doing something like this as we plan to support a general remote launching mechanism. Thoughts?
There is going to be a tradeoff between complexity of plug-in implementation and flexibility. The challenge is going to be finding the sweet spot that gives lots of power to the user without requiring them to understand CDT internals.
I think you are on the right track by proposing methods. I've found that these discussions can't proceed in the abstract or you get swept up in generalities. Seeing a specific proposal for launch plugins to CDT would be quite helpful. I've also found it more effective for someone knowledgeable to come up with a proposal rather than doing it in committee. Any volunteers?
The only other comment I would have is that you may want to split run vs. debug launching. Until your roadmap says you are doing your own debugger, you will be built on an external debugger. That means that debug launch should always be a local process launch with the right parameters or commands passed to the debugger that may target the debugger at a remote target. I don't seem much value in a CDT connecting to a remote GDB. Perhaps for remote Linux debug, but then Linux already has a pretty nice remote debug GDB agent, so it would seem a little excessive to me.
In contrast, run launching may well be doing a remote launch because there is no complex intermediary to do the remote launch. You can bet that our user class of the plugin we are running here is not operating on the debug command line.
Thanks! -Chris