[Lldb-commits] [PATCH] D91238: Recognize __builtin_debugtrap on arm64, advance pc past it so users can continue easily

Wed Nov 11 00:55:43 PST 2020

labath added a reviewer: jingham.
labath added a comment.

Including Jim, as he was involved in the discussion last time this came around http://lists.llvm.org/pipermail/lldb-dev/2020-March/016017.html.

This functionality would not work on linux/lldb-server currently, though I would love if it did -- I've found the x86 behavior of continuing after an int3 to be pretty handy, and was annoyed by the fact it does not work the same way on arm.

That said, I'm not sure that implementing this in the server is the right approach. The previous discussion seemed to converge on doing that in the client, and I think the reasoning behind that is sound (like, ensuring that the PC we report will be the same as the one extracted by the native crash/backtrace tools when the process when the process hits that instruction without the debugger present). It would also mean that this behavior would automatically work with all stubs, even third party ones we have no control over.

Maybe we don't even have to do the really fancy thing of adjusting the PC just before we continue. The initial implementation could just adjust the PC immediately after a pc. That way, if this will be a client feature, we could easily go back and implement something more fancy...

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D91238/new/

https://reviews.llvm.org/D91238