- figure out what to do on Windows and GTK+ if we don't have menus but the user wants a menubar (zero-height widget? don't bother? complain?)
- bin.c
	- find a way to consolidate the duplicate code across OSs
	- find a way to move the has parent check at the beginning of binDestroy()
	- determine whether or not margins count in preferredSize() when there is no main control
- menu item state change while the menu is visible (not in response to user action)
	- OS X: requires an [NSMenu update] call after changing state
	- Windows: requires getting current menu state when amending it in sync()
- figure out where we should return HRESULTs
- Windows: don't abort if a cleanup function fails?
- should uiWindowOnClosing() really return an int?
	- if so, should the default return 0 instead of 1?
- note that uiInitOptions should be initialized to zero
	- have a cbSize member
- drop "Page" from uiTab method names? (uiTabAppendPage() -> uiTabAppend())
- override dock menu quit item to act like our app menu quit item
- consider calling setAppleMenu: for the application menu; it doesn't seem to make much of a difference but
- release OS X properties in dealloc
	- switch away from properties or figure out what incantation is necessary for 32-bit
	- figure out strong/weak vs assign/copy/retain
- figure out what to do in OS X terminate:
- clean up bin handling
	- is binSetOSParent() even needed?
- readd the allocation tracking as just a list of pointers
- add a final cleanup function (uiUninit() or uiCleanup())
- whenever a list of things is destroyed, each successive item must be removed as it is destroyed, otherwise we might wind up in a situation where we access items after they're freed
- make the name of the variable to refer to a single tab page consistent (already decided to make them all `page`)
- make sure uiEntryOnChanged() is not triggered when calling uiEntrySetText()
- clean up windows resizing logic
	- make it so that only top-level window resizes trigger an update; container resizes do not update
		- windows resizing logic is simply not comprehensive enough (no null resizes allowed) to do things
		- we control resizes of all children so we can reliably update after a resize
		- we already need to do this in uiContainer :/
- make it so Windows API calls that do logLastError(), etc. abort whatever they're doing and not try to continue, just like wintable
- 32-bit Mac OS X support (requires lots of code changes)
	- change the build system to be more receptive to arch changes

ultimately:
- add some sort of runtime type checking
- MAYBE readd lifetime handling/destruction blocking
- figure out how to integrate leak checking into the source tree
- assign control IDs on windows
	- GWL(P)_ID
	- related? [12:25] <ZeroOne> And the blue outline on those buttons [ALL clicked buttons on Windows 7] won't go away
		- I get this too
- SWP_NOCOPYBITS to avoid button redraw issues on Windows when not in tab, but only when making resize faster
- opposite side alignment control in uiBox
- collapse multiple consecutive menu separators?
	- including for quit/preferences/about?
		- or not even have separators there?
			- in which case, hide leading or trailing separators?
- generate libui.lib and related files
- rename newcontrol.[cm] to control.[cm] or something else?
- make OS-specific headers generated by an IDL

notes to self
- explicitly document label position at top-left corner
- mark that uiControlShow() on a uiWindow() will bring to front and give keyboard focus because of OS X
	- make sure ShowWindow() is sufficient for zorder on Windows
- document that you can't use InsertBefore functions to insert before a nonexistent index (that includes if an array is empty)
	- reconsider this, as the pointer array code does work with the first invalid index...
		- we would need to test everything else with it
- note that uiTabInsertPageBefore() does NOT change the current tab page (it may change its index if inserting before the current page)