Stacktraces are missing (at least on lua 5.2) for LuaErrors about mlua HOT 4 CLOSED

I found a better way without attaching Option<String> to a WrapperError and seems it works better
fe39ae0

from mlua.

Thanks for the bug report!

I'm working to fix, and will release 0.6.0-beta.2 soon (sorry, I probably will not backport to 0.5, but hope it's ok).

In 0.6 all functions that can trigger longjmp are moved to the C shim and it's safe to call luaL_traceback (from the C side, checking available stack first for Lua <= 5.3, in 5.4 it's implemented using buffers).
I also like the idea to create a custom (and stack independent) implementation of luaL_traceback.

from mlua.

@khvzak Hey, thanks for your reply. I experimented somewhat in #45. This seems to work fine in my use case, but I’m just coming back to Rust after a ... seven year break (time flies!).

I suspect your implementation will be more sound, I’d still appreciate a review if you like.

from mlua.

I'm preparing v0.6.0-beta.2 release and pushed a commit that should improve stacktrace reporting in the module mode.
I have not yet published it. Could you please check from the master branch does it looks as you expected?
Thanks.

It will report callback error + stacktrace for a most top callback that failed and the main cause, skipping intermediate calls.

--
I don't much like the current approach, but seems not many options available. Ideally tracebacks should be attached to Errors but it would be super breaking change (like wrapping error to Error+Traceback or adding field to each Enum variants).

from mlua.