No error bubbling?

[jupyter-widgets-contrib]

No error bubbling?

[kenseehart]

Here's my workaround:

import traceback

class MyUI:
    def __init__(self, canvas, model):
        self.canvas = canvas
        self.model = model
        canvas.on_mouse_down(self.mouse_down)

    def mouse_down(self, x, y):
        try:
            do_some_stuff()
        except:
            self.print_traceback()
            
    def print_traceback(self):
        self.canvas.fill_style = '#ff8888'
        self.canvas.fill_rect(10, 10, 300, 300)
        self.canvas.fill_style = '#000000'
        s = traceback.format_exc()
        for i, si in enumerate(s.split('\n')):
            canvas.fill_text(si, 20, 30+15*i)    

Works great. Now my code is debugable.

[martinRenou]

Thanks a lot for opening an issue. I suppose this is unexpected behavior. I am not sure what is causing this. I will have a look in two weeks.

[janfreyberg]

that's a really neat workaround, thanks @kenseehart

[martinRenou]

That is really weird. I could also see something similar in ipyleaflet today, see jupyter-widgets/ipyleaflet#455 (comment).

I really don't know where this comes from.

Maybe you already know about it, but you can use the Output widget in order to capture the stream explicitly, and put it where you want.

[janfreyberg]

Yes - the issue is more for debugging when it's hard to know where to capture output. I don't know why this might be happening, but the jupyter lab log console is discussed here: jupyterlab/jupyterlab#7386

They specifically mention un-handled messages being logged; I don't know how a message being "handled" is defined though.

[kenseehart]

Looks like the CallbackDispatcher is consuming stdin and stdout for some reason. It's not about exceptions per se. Probably the exception is be handled in the normal way, but when the traceback is written to stdout, it gets consumed.

[ianhi]

The CallbackDispatcher is being created a little bit differently in ipycanvas and the ipywidgets button. In the button it is created in the constructor:
https://github.com/jupyter-widgets/ipywidgets/blob/6be18d9b75353f7b4a1c328c6ea06d8959f978f6/ipywidgets/widgets/widget_button.py#L57-L60

whereas here it is wrapped in at traitlets Instance function and explicitly added:
https://github.com/martinRenou/ipycanvas/blob/a97d7810745afb61e4daed71bd92b998304e36fd/ipycanvas/canvas.py#L180

[martinRenou]

@ianhi do you think that could be the cause of the issue?

Indeed having the object instantiated as a kind-of static property does not sound right at all.

[ianhi]

@martinRenou oops I forgot to include any context for that comparison. Yes, my intention was to imply that this may be the cause as this seems to be the only relevant difference between ipycanvas and the ipywidgets.Button

[ahrm]

Hi! Why is this issue closed? Is it fixed in the new versions?

[ianhi]

Hi @ahrm this issue is still open. You see this at the top of the page: