Custom Event error handler | Voters

Custom Event error handler

planned

Tapple Gao

Luau scripts have a lot more ways to crash than LSL scripts. I wrote this as a proof of concept on making LLTimers and LLEvents crash proof:
https://gist.github.com/tapple/70b874b3801d6cbf103559287e85b964
It might be nice if there was a function like LLEvents:setDefaultErrorHandler(errorHandler)
 that did the same thing the above code does, without needing to wrap/monkey patch anything.

February 17, 2026

Tapple Gao

I wish we had scriptable access to the default slua error handler, so I could wrap my coroutines with it. It's similar to this function:

local function errorHandler(msg)
    ll.Say(DEBUG_CHANNEL, debug.traceback(msg, 2))
end

except

the traceback is not broadcast publicly on DEBUG_CHANNEL, only to the owner (this is not replicatable with
ll.RegionSayTo(ll.GetOwner(), DEBUG_CHANNEL, msg)
; see the wiki's caveats
the traceback is only visible to the owner if the script is modifiable. I could probably this using
bit32.btest(ll.GetInventoryPermMask(ll.GetScriptName(), MASK_BASE), PERM_MODIFY
. But I don't know if this logic matches what the sim does due to insufficient testing

I'd want to use the error handler like:

local function coroLoudWrap(f)
    -- works the same as coroutine.wrap, but if the
    -- coroutine has an unhandled error, print a stack
    -- trace, just like how the main thread does
    return function(...)
        return coroutine.resume(coroutine.create(xpcall), f, errorHandler, ...)
    end
end

Honestly, I also wish

coroLoudWrap

was a builtin function, or

coroutine.wrap

took a second argument that enabled this behavior. I don't like how using coroutines breaks error trackbacks unless you're very careful. Error trackbacks should be the easy path, not the hard one

Tapple Gao

very related: it would be nice if all events started in a coroutine so that I don't need to add that wrapper to every event handler.

putting all events in a coroutine would have the side effect of making exceptions nonfatal, since coroutines are implicitly pcalled

Tapple Gao

DetectedEvent might complicate coroutines by default, as it is currently backed by global state rather than instanced state, and thus won't survive a yield

Harold Linden

updated the status to

planned

Thanks! I already have plans for something like this, design TBD :)

Tapple Gao

I haven't structured my slua scripts around linkset data and lljson yet, but I plan to. I'm not sure how much code I want to spend verifying that my linkset data has a sane format vs just leaving it to some high-level code similar to the above that deletes the linkset data key if there ends up being a problem using it.

I'll probably write some sort of linkset data wrapper that works similar to python's with block. That would be a lot more specific than "what to do for general unhandled errors". This wrapper would probably solve most random crashes I can currently imagine, while I'm this early in my slua journey.