[Lldb-commits] [lldb] [lldb] Provide lr value in faulting frame on arm64 (PR #138805)

[Jason Molenda via lldb-commits]

jasonmolenda wrote:

> > I also needed to remove the brk #0xf000 instruction because lldb is not able to step over it -- it just ends up spinning forever. FWICS, it's not actually necessary now that you're stepping through the test. Nonetheless, this is a bug -- one that @DavidSpickett might be interested in :)
> 
> Yes I've been asked a few times about it. GDB also has problems with it, but probably does something slightly more useful. I'll speak to folks on that side and see how we could handle this.

@DavidSpickett sorry I missed this discussion a few weeks ago.  Yeah, in debugserver aarch64 backend we check for the current instruction being `brk #0xf000` specifically and move $pc to the next instruction.  Customers were using `__builtin_debugtrap()` in their code to force a break in the debugger, and wanted to continue stepping after hitting that, like they could on Intel systems (where the pc was past the interrupt instruction).  I wasn't sure if it was best to special case this in debugserver, or up in lldb in some architecture dependent stepping plan when we're stopped at this instruction, I may have chosen poorly.

https://github.com/llvm/llvm-project/pull/138805