Re: [cdt-debug-dev] determining which breakpoint was hit

["Alain Magloire" <alain@xxxxxxx>]

> 
> > What are you trying to achieve ?
> 
> I'm trying to write a plugin that can automate a task to be performed
> after a specified breakpoint (from the user's Breakpoint View) is hit,
> during a debug session. I want to do it generically for all debug
> models. So I'm up at the Eclipse Debug interfaces like IBreakpoint. I
> did this, and it works with JDT using the IThread.getBreakpoints() call,
> which is the only documented way I see to find out which breakpoint a
> thread is suspended at, but CDT (1.2 at least) just returns null in that
> function.
> 
> In the messing with internals I shouldn't be messing with department:
> The BreakpointHit object can point me to the Breakpoint object, but I
> need the CBreakpoint (that implements IBreakpoint). From what I can
> tell, the CBreakpointManager is the only thing that knows about the
> mappings between CDI breakpoints and CBreakpoints, and its method
> getBreakpoint() will give me what I want. And the only public way to get
> that manager is the CDebugTarget.getAdapter call, but for that I have to
> know ICBreakpointManager.class(), which I guess I can do with a
> CDT-dependent plugin. So that should work as a nasty temporary solution.
> :)
> 
> The real solution requires a change to CDT...
> 

Yes, Now I see what you mean.  Make a PR for it in bugzilla.
When implementing the Eclipse debug core interface we probably
did not implement all the methods and some are still stubs.

> >From the documentation of IThread.getBreakpoints:
> 
> "Returns the breakpoints that caused this thread to suspend, or an empty
> collection if this thread is not suspended or was not suspended by a
> breakpoint. Usually a single breakpoint will be returned, but this
> collection can contain more than one breakpoint if two breakpoints are
> at the same location in a program."
> 
> In GDB I guess all threads get suspended when a breakpoint is hit? So it
> might be valid by this description to return the same breakpoint from
> all IThreads. This would be quite easy to do given the data I see being
> maintained already. If you wanted to be more elegant and only mention
> the breakpoint in the thread where the breakpoint actually occurred,
> that might be harder (I don't know off-hand if GDB even reports that?).
> 

The information is there, GDB/MI/CDI etc .. returns the breakpoint that
triggers the suspend ... It looks like we just did not implement the
IThread.getBreakpoints() ... currently it is always returning null 8-(.

Make a PR and we will remedy in the head.
BTW, if you find other places that we did not respect the Eclipse
Platform debug model please fire a PR.

Thanks for the heads up.