labath wrote: Yeah, I plan to look into Function::GetAddressRange and see how we can make it do something reasonable for these kinds of functions (probably by having it return a list of ranges). That will likely involve looking at the users of that API as well, although I'm less than enthusiastic about making this work with the instruction scanner -- the reason for that (at least for our use case) I don't see much use for it. We already have tools (like profilers) which rely on the correctness of unwind info (even in the async case), so I don't think we gain much by using it. Clang already produces correct eh_frame for these discontinuous functions (for x86 at least, we're in the process of fixing aarch64), and hand-written functions are not likely to be split, and even if they are, unless the user writes correct debug info for them, we won't be able to detect that they are split. At that point the user might as well write the eh_frame section as well...
https://github.com/llvm/llvm-project/pull/111409 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
