ut2004 makes a 2D window for the splash screen, which sets the screen
surface's pixels field. Then we tear down that video mode and create a GL
context, and the Quartz target isn't resetting the pixels field to NULL.
When you just create a GL window, the structure is memset'd to zero the
first time through, so unless you hit ut2004's codepath, you won't see the
bug. :)
Without this patch, quitting a windowed ut2003/ut2004 game makes the OS dump a
warning about a bogus free() to stderr, but it doesn't actually crash. All we
need to do is explicitly initialize the current->pixels field.
From: Max Horn
Subject: Re: Fix for OSX bug
The problem is that our actual client area is realized as a
subview of the "content view" of the window. That's normal so far. The
odd part is that, while our subview starts out filling it parent 100%,
upon resizes it gets changed to something smaller....
Anyway, here is a revised patch which simply enforces the size of
window_view by calling setFrameSize.
From: "Mike Gorchak"
Subject: Batch of the QNX6 fixes for the SDL
1. Updated readme.QNX
2. Fixed crashes during intensive window updating under fast machines (got over 200 rectangles for update).
3. Fixed double-buffered fullscreen modes, now it works as needed.
4. Fixed Photon detection algorithm.
5. Fixed HWSURFACE update function.
6. Added SDL_PHOTON_FULLSCREEN_REFRESH environment variable support for control refresh rates under Photon.
7. Added 640x400 fullscreen mode emulation via 640x480 (if videodriver not supports original 640x400 mode of course) shifted by 40 vertical pixels from begin, to center it. It's needed for some old DOS games which ran in doubled 320x200 mode.
8. Added available video ram amount support.
8. Added hardware surface allocation/deallocation support if current videomode and videodriver supports it.
9. Added hardware filling support.
10. Added hardware blits support (simple and colorkeyed).
And I've added to testvidinfo test color-keyed blits benchmark (maybe need to add alpha blits benchmark too ?). Currently Photon not supporting any alpha hardware blittings (all drivers lack of alpha blitting code support, only software alpha blitting exist in photon, which is hundreds times slowest than the SDL's one). So I've not added the alpha support. I suppose new QNX 6.3 will have the hardware alpha support, so when it will be done, I'll add alpha support.
From: Max Horn
Subject: Modifier key fix
The internal modifier state can get out of sync with reality. To
trigger this, do for example this:
1) Launch an SDL app
2) Alt-click on the desktop (this will hide the SDL app).
3) Bring the SDL app back to the front
4) SDL will still think alt is pressed (and as such will treat left
clicks like middle clicks). If you press and release alt, it'll be fine
again.
The attached patch cures this by rechecking the modifier state whenever
we process an event.
From: Dmitry Yakimov
Subject: [SDL] [PATCH] SDL bug patch
Let me introduce 2 fixes to SDL.
1. Preventing great slowdown on fast machines while hardware
flipping (it's obviously bug).
2. Setting up 85 Hz of monitor if supported. The reason is that
Win98 by default sets lowest frequency 60 Hz.
And we can't set up maximum frequency because some users can have
wrong monitor drivers.
This is important for shareware and commercial programs.
From: Jens Leuschner
Subject: Small patch for libSDL
In newer XFree86 releases in the German keyboard the "AltGr"-Key no
longer produces the keysym "Mode_switch", instead it produces
"ISO_Level3_Shift".
I have created a small patch against SDL-1.2.6 (which cleanly applies to
the current CVS-SDL too) that makes SDL recognize ISO_Level3_Shift as
AltGR key.
