* Gerry Lawrence <gwlp...@gmail.com> [2010-12-21 20:05]: > Focus follows mouse - which I can get on any unix, easily, and > any microsoft, with a little more effort, is impossible on the > MAC.
It's actually *technically impossible* to implement Focus Follows Mouse on MacOS. Do you want to know why? It's because of the menu bar at the top of the screen. * <http://steve-yegge.blogspot.com/2008/04/settling-osx-focus-follows-mouse-debate.html>: > The problem is that Macs, always and forever, have put the menu > bar of the focused application at the top of the screen. The > menu bars of unfocused applications are hidden and are not in > any way user-interactible. > > As you might expect, this UI assumption has been baked into the > Mac APIs from the very beginning. Programmers will take > advantage of any axiom they can in order to get things working, > so over time this has turned from a UI design assumption into > an architectural "feature". > > In particular, when an app is in the background, its menu > structure may not be intact, and the app may be in a state that > assumes it will not be receiving any keyboard input. One > concrete example mentioned in the thread was that when an app > is in the background, the child menu items do not have parent > links (although the parents still have pointers to the > children). > > This has serious ramifications for focus-follows-mouse. There > are certain built-in hotkeys that can activate menu entries, > and apps are also free to define their own. If you try to > activate a menu in a background application, it could in theory > wind up crashing the app, if the app is assuming an intact menu > structure and is traversing bottom-up rather than top-down. > > You could attempt a knee-jerk solution by allowing Control and > Alt through, but deny the Command modifier, since that's the > most common menu-activation modifier (I think). But there's > another class of applications (Emacs included) that dynamically > generates at least part of its menu structure based on the data > content. For instance, the Emacs Imenu package generates a list > of jump targets from a source-code buffer. Even typing a new > function definition could still trigger a rebuild of the IMenu, > which (for all anyone knows) could crash Emacs. Regards, -- Aristotle Pagaltzis // <http://plasmasturm.org/>
