ajeffrey: trying Servo3D, but it doesn't start. I see that in the logs:

well done, it totally locks the device. It's stuck with the app starting animation, now way to get out of it

had to hard reboot

xclaesse: oh that's not good

xclaesse: I'll have a look to see what's going on with that panic.

ajeffrey: "panicked" is like raising an exception in Rust ?

xclaesse: Rust has two classes of exception,

error results, which are expected to happen sometimes,

and panics, which are intended never to happen.

Servo panic means something has gone very wrong.

ok, so like an assert()

Yes, failed asserts generate a panic.

Panics tear down the thread that panicked, so they're pretty serious.

ajeffrey: with Servo2D am I supposed to get a video?

I only see a big black screen, and typing an URL doesn't load anything

xclaesse: looks like the same panic

I'll see if I can replicate it at my end

If I can I'll also look to see if older builds are working.

This is one of the problems about not being able to test the ML1 in CI, sigh.

If we had a way to test on the device, then we could make sure the nightly build works before uploading it.

It only cost 3k$ to put a device in your CI runners :D

not sure it will survive for long if you let it running 24/7

xclaesse: and work out where to host it.

I don't imagine AWS is going to give us an ML1 image any time soon.

you can run on your own host, can't you ?

xclaesse: if we're going to gate on it in CI then we'd need something a bit more robust than putting a machine in my office on the public internet.

xclaesse: rats, I downloaded and installed the latest nightly mpk, and it all just worked.

so we're getting to play the game of "it works on my end", sigh.

:/

indeed.

xclaesse: looking at the panic, it's the ScriptThread init that's panicking, could this be something to do with security settings?

we've had problems with the JS JIT not playing well with ML security settings, e.g. forum.magicleap.com/hc/en-us/commun…egmentation-fault-in-release-builds

if you're feeling brave, you could try launching the app in the debugger and doing "break rust_panic" to install a breakpoint at the point of panic, that way we could get a backtrace.

I've got no time for that atm, I just wanted to give it a try out of curiosity. Sorry :/

FYI, the mpk I'm using is the one from download.servo.org that I re-signed

xclaesse: okay, the PR with the new signing cert hasn't merged yet, once it does you could try again to see if it's a resigning issue?

ajeffrey: sure, ping me when the mpk is ready ;-)

xclaesse: will do

ajeffrey: oooh, found a reason for a big perf issue ! The audio sink is busy looping

instead of waiting on a cond until an audio buffer is available, it spin in a loop a retry until it gets an output buffer

found that while working on getting spatial sound working

xclaesse: ah, that is interesting!