dlav-sc wrote: > One kind of testing we need to add here - given the performance degradation > this feature entails - is that if you have no software watchpoints, we don't > do this single step all threads, then if you set a software watchpoint we > start single-stepping all threads and then when you remove the software > watchpoint we stop the single-stepping again. It would also be nice to try a small program and run it from start to finish in this mode with a software watchpoint set. It would be good to find out how generally stable that is...
Yeah, I have consider adding such tests as well, but I currently have no idea how to implement them. I used lldb step logs to observe the thread plan stack and to ensure that the WatchpointThreadPlan is used only when necessary. However, I haven't found a way to obtain the state of the thread plan stack via the python api or the lldb command-line interface to use in a test. We could potentially parse the logs in tests to get the thread plan stack, but that seems too unreliable. Please share your thoughts, if you have any ideas on the matter. https://github.com/llvm/llvm-project/pull/151195 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
