[llvm] [Dexter] Adapt to upcoming lldb stepping behavior (PR #108127)

[Jason Molenda via llvm-commits]

jasonmolenda wrote:

> Looks like a sensible way to solve the problem, thanks for this! Just so I understand correctly, in cases where we set breakpoints but exclusively advance using `go()`, as with the ConditionalController, the change in LLDB's behaviour will have no effect - is that correct?

The lldb `go` method does `SBProcess.Continue()`.  If we are moving between the breakpoints with Continues, the behavior is the same as it was before.  We execute until we hit the breakpoint instruction, and then when we resume the thread (step/continue/etc), the breakpoint is silently stepped past before execution is resumed.  The issue I'm fixing in this PR only comes up because the `step` stops on the breakpoint instruction that hasn't been executed yet.  I did run the dexter tests on my desktop with the stepping change and they passed, we'll see what happens when it hits the CI bots but I think this is sufficient.  Thanks!



https://github.com/llvm/llvm-project/pull/108127