[Lldb-commits] [lldb] [lldb] Merge/unify ABI-provided AArch64 unwind plans (PR #139545)

[Pavel Labath via lldb-commits]

labath wrote:

> Nice unification, thanks. I'm sure the pc=lr rule was me not trusting the algorithms over in RegisterContextUnwind to do the right thing if the rule wasn't listed. Having just pushed all that code around for a day, I know this kind of thing is unneeded, but harmless if it's present.

Well.. as it stands now, it's kind of needed because without it, your test would break in CompareUnwindPlansForIdenticalInitialPCLocation, as it doesn't recognise that `SetReturnAddressRegister(LR)` is equivalent to `SetRegisterLocationToRegister(PC, LR)`. 

This is kind of why I have suggested in https://discourse.llvm.org/t/unhappiness-with-the-lldb-unwinder-register-passing-up-the-stack-interrup-trap-sigtramp-frames/86058/3 to standardise on the `pc=lr` rules. I wrote that before I encountered this problem, but seeing we had plans which employed both of these strategies (and they both mostly worked) has made me more confident this could work.


https://github.com/llvm/llvm-project/pull/139545