wrap callback call in try catch

[SimonDanisch]

I remember reading, that errors in c callbacks yield undefined behavior! Since I ran into some undefined behavior with GLFW, I figured that it might be this.
I'm not 100% if that's actually true, so we shouldn't use this try catch if it's false. We should probably also try to give some more information & backtrace.

Sorry to ping you directly, but @yuyichao, you're probably the only one who would know the answer to this question with certainty!

[jayschwa]

I remember reading, that errors in c callbacks yield undefined behavior!

Do you know where? The default error callback throws an error and has always seemed to work fine.

Since I ran into some undefined behavior with GLFW

Can you provide some more details?

[SimonDanisch]

Hm this is a bit complicated...
So I read that thread in discourse a while ago - and it's not unlikely that I miss-remember the details or got the context wrong! Sadly, I cant find it anymore. Anyways, remember when upgrading to 1.0, I had a bunch of very weird silent corruptions? I never pinned them down, but it looked a lot like undefined behavior in the event callbacks.
I also pinned down some of those to Observables, which I use to propagate the GLFW events - so callbacks I register with Observables, end up being called from within a GLFW callback.
And I got some really weird bugs, were julia just silently quit or froze, which was fixed by introducing a try catch in the Observable callbacks, and printing the error instead of throwing it!
Since I read that discourse thread, I came up with the theory, that the actual source of corruption might have been in throwing inside the c-callback.
Since Julia currently has problems with catching errors this might also have a different cause! That's why I hope someone else could give us some clarity on what could have caused this!

[SimonDanisch]

This hasn't turned up anymore