LLDB cannot show backtrace at `rust_begin_unwind`

[lilyball]

When using LLDB to debug rust, breaking on rust_begin_unwind only provides the following backtrace:

* thread #2: tid = 0xef8f7, 0x00000001027a2efe librustrt.dylib`rust_begin_unwind(token=839147) + 14 at rust_builtin.cpp:563, stop reason = breakpoint 1.1
    frame #0: 0x00000001027a2efe librustrt.dylib`rust_begin_unwind(token=839147) + 14 at rust_builtin.cpp:563
    frame #1: 0x0000000100093678 libstd-6c65cf4b443341b1-0.8-pre.dylib`rt::task::__extensions__::meth_25045::begin_unwind::_7c12263797ed078::_0$x2e8$x2dpre + 72

GDB is able to show the full stack.

I believe LLDB uses a different method of generating a backtrace than GDB, one that is considered much safer, but is also more sensitive. However I don't know the details.

Unfortunately, this is going to become a big issue for anyone using Rust on OS X, as the upcoming version of the developer tools (Xcode 5) is dropping GDB (and GCC) entirely.

[lilyball]

GDB says frame #2 looks like

#2  0xb6a825ef in upcall_call_shim_on_c_stack (args=0xb42fddf8, fn_ptr=0xb7c5ab50) at src/rt/rust_upcall.cpp:42

It's possible that the issue is purely with this function, rather than being a more general issue with debug info.

[alexcrichton]

Closing, I believe that the experience with rust_fail is greatly improved. I also haven't recently run into this at all (and I've been using lldb heavily)

Feel free to reopen if it's still here!