[Lldb-commits] [lldb] Make breakpoint stop reason more accurate for function breakpoints (PR #130841)

[via lldb-commits]

satyajanga wrote:

> Pretty good start. There can be more than one breakpoint at a location and we need to be sure we test these cases. Like we can have an exception breakpoint (which might be at `cxa_throw`) and we can have a function breakpoint for this as well. We need to make sure we document how we are doing to display this. If we must pick one, then lets be clear about:
> 
> * exception breakpoints take precedence
> * then function breakpoints
> * then source file and line breakpoints
> 
> If we have a free form text we can attach to the stop notification it might be nice to see "function breakpoint & exception breakpoint", but if we have to pick just one, then we need to be clear about what we are going to do.


Agree that there can be more than one breakpoint at a location. 

but 'reason' field in StoppedEvent only takes one string from fixed set of strings. 
>    reason: 'step' | 'breakpoint' | 'exception' | 'pause' | 'entry' | 'goto' | 'function breakpoint' | 'data breakpoint' | 'instruction breakpoint' | string;
https://microsoft.github.io/debug-adapter-protocol/specification#Events_Stopped

So the existing logic for instruction_breakpoint only sets the reason as 'instruction breakpoint' if all the breakpoints are instruction_breakpoints. 
Current behavior
1. if all the breakpoints are exception breakpoints set reason = 'exception'
2. if all the breakpoints are function breakpoints set reason = 'function breakpoint'
3.  if all the breakpoints are instruction breakpoints set reason = 'instruction breakpoint'
4. if there are combination of one more breakpoints , then the reason will be 'breakpoint' 

I think this is a reasonable behavior, I will document this in the code. @clayborg  let me know what you think?



https://github.com/llvm/llvm-project/pull/130841