Fix issue 104: Only call setNativeRange() in response to a mutation when the editor has focus

[kloots]

This is a fix for issue #1404.

[jhchen]

Need to investigate a bit more but two current thoughts:

• I generally prefer to avoid setTimeout as it makes code non-deterministic and hides the real stack trace when issues occur. But some tasks cannot be done without it but this warrants exploring alternatives.

• There is a TODO above this code that iirc was because of a Firefox issue but it could be the case that it would be best not to have this SCROLL_UPDATE handler at all

[kloots]

@jhchen thanks for the reply.

I saw that TODO in the comment above, but lacked the context as to why the SCROLL_UPDATE handler existed in the first place. Can you recall? If it no longer needs to exist, then that would be the best solution.

If it needs to remain, I can't think of a better solution than leveraging setTimeout(). I'm with you on the downsides. But for this specific problem (ensuring document.activeElement is accurate) I'm not aware of another solution.

If there's anything I can do to move this forward, please let me know.

[jhchen]

I opted for another way to fix the issue but thank you for the detailed report in #1404 -- it made the investigation a lot easier.

The code was originally added to deal with if you select-all (cmd-a) and typed a character, Firefox would lose the selection. There was a logical inconsistency with how the fix as it should only care about the case where it has focus but getNativeRange() returns the range regardless.