Errors are not forwarded to window.onerror

[ACXgit]

Errors inside components do not reach window.onerror event handler while using the Mithril router, mounting components with m.route. Instead they do when the component is mounted with m.mount.

This didn't happen in 1.x.

Mithril version:
2.0.4

Browser and OS:
Firefox 81 64bit, Windows 10

Code

[flems] m.route with Mithril 1.1.7, working as expected

[flems] m.route with Mithril 2.0.4, not forwarding

[flems] m.mount with Mithril 2.0.4, working as expected

Context

Without errors being forwarded to window.onerror it's impossible to catch them globally and log them, i.e. to an application monitoring and error tracking platform like Sentry.

[dead-claudia]

Found the issue:

• The initial route isn't #!/, so it has to update the route.
• Whenever the route updates, it then has to redraw.
• Redraws are all run in sequence, with errors captured and logged.

I feel the issue may be mitigated in two parts:

1. Resolve the initial route before mounting, to avoid an initial blank render. Edit: Not necessary. Just the second.
2. Change async redraws to call requestAnimationFrame once for each root rather than once for all roots, and just use the array of scheduled roots to avoid double-scheduling roots. (It's worth noting no user code can run between animation frame callbacks, and microtasks are delayed until after paint, so this change would likely not be user-breaking.)

[kfule]

(memo)
On the first drawing of “m.mount”, the user-defined window.onerror is being called because it does not pass the following try-catch.
On the other hand, in “m.route”, m.redraw seems to be called at the same time as the initial drawing, and the following try-catch seems to catch the error. This appears to be the difference between m.mount and m.route.
https://github.com/MithrilJS/mithril.js/blob/v2.2.13/api/mount-redraw.js#L13

At least, by removing the above try-catch, user-defined window.onerror is also called in m.route case.
(Considering the error handling during repaint, this may be a bug(?) during first paint in m.mount case.)