Re: [AD] 5.0.8-pre branch

[Peter W. <nov...@gm...>]

On Sat, 10 Nov 2012 20:16:25 +0100, Elias Pschernig <eli...@gm...> wrote:
> The 555 modes are likely an Intel bug in the HD3000. We probably should
> remove the 15-bit modes from Allegro completely (16-bit modes like 5551
> and 1555 work fine so those we would keep).

We can deprecate them in the documentation if it's decided.

> And we use non-premultiplied alpha in the test program for some reason,
> which is not supported with the OSX loader so the alpha values are off:
> 
> http://www.allegro5.org/tests/run%20on%202012-08-31%2000:33osx/test%20png%20%5Bsw%5D.png
> http://www.allegro5.org/tests/run%20on%202012-08-31%2000:33osx/test%20tga%20%5Bsw%5D.png
> 
> Difference + Normalize in Gimp:
> 
> http://allegro5.org/diff.png
> 
> I think SiegeLord had an idea how to fix it by loading the image twice,
> once in alpha mode and that way extracting the alpha channel - but it
> would be some effort to test it and libpng works fine.

Ok.

> Both demos work fine.
> 
> 
> Some of the examples exit with a warning about an uncommitted
> CATransaction. No idea what that is about.
> 
> ex_joystick_events crashes. ex_joystick_hotplugging freezes. There is
> no joystick but it's still bad style :P

Seems OS X specific.

> ex_noframe crashes on exit.

Seems OS X specific.

> ex_fs_resize changes the mode of the desktop then remains like that
> without any Allegro window or the ability to do anything. But my
> opinion is that mode-changing is useless anyway. ex_fs_window works
> fine.

This prompted me to check ex_fs_resize on Windows, and instantly hit a
number of bugs.  I'm pushing back the release another week to give the
fixes a chance to settle.

Thanks for testing, and to Jon and Trent.

Peter