Improvements for onerror line 0 and line 1 errors

[JAB Creations]

Saw this...
http://html5.org/tools/web-apps-tracker ... 13&to=6914

...and came to this...
http://www.w3.org/TR/html5/webappapis.h ... ipt-errors

I log JavaScript errors to a database and look over them. Unfortunately about half are useless line 0 or line 1 errors with no realistic way to determine how the error was triggered and thus making the error report useless.

Secondly it would be exceptionally useful if the onerror event emulated PHP's error handling. A function name and perhaps a back-tracking ability (what called what, did function a1 call b2 and b2 spawned the error? etc).

Having these two issues covered by the spec would greatly improve error handling.

[zcorpan]

Browsers could include the stack trace in the message argument. Is that good enough?

[JAB Creations]

Had to look up stack trace but was what I thought it would be; that would include the list of say functions in a separate object/array from from the say again function name where the error occurred?

[zcorpan]

Yep. My thinking was that the whole trace could be dumped in the message (as plain text).

[JAB Creations]

Cool. What about JavaScript errors spawned from browser plugins/extensions/etc? I think a lot of the JavaScript errors in my site's log seem to be spawned from those.

[JamesMGreene]

I feel that @zcorpan's proposal may be a sufficient workaround but it's not a valid solution. Here is my counter-proposal: https://gist.github.com/3ded0f6e7f0a658b9394

[arabella]

Yeah, I guess the spec could be a bit clearer about the canvas element. Could you file a spec bug, please?



emma