[SDL] Win32 Palette odd case bug & fix

[Manny]

Using our app and the nifty sailboat testpalette program it works either 
way ... in windib mode - seems there is a difficulty with directx 
palettes(?) (windowed & fullscreen) but I haven't had a chance to investigate.

InitVideo does create a screen_pal for the window if the display is 8 bit, 
so this is just matching that behavior if the desktop is changed from 
outside after the Init, so I think it's all ok.

What I'm doing (and what didn't work without the change) is if the desktop 
gets switched to 8 bit while the program is running, then I make an 8 bit 
surface and set the 332 palette.  Setting the palette would always fail if 
the desktop wasn't 8 bit when SDL was initialized.  Then  I make a 
backbuffer of the original depth so everything keeps working and dither the 
backbuffer to the 332.  Not great (though surprisingly ok), but before we'd 
just bail with a "hey you changed the display to 8 bit" sort of message, so 
it's an improvement of sorts.  Pretty much what SDL does automatically, but 
with a dither.

Thanks,
--Manny

At 06:50 AM 6/7/2001 -0700, you wrote:
> > SDL currently does not handle the case that the program is running in a
> > window at depth > 8, then the user changes their desktop to 8 bit and the
> > SDL window is recreated accordingly.  The result is the window is created
> > in 8 bit and has no "screen_pal"
>
> > It seems supporting this case is as simple as moving this code in
> > SDL_dibVideo.c (line 521) down below "#endif /* 
> !NO_CHANGEDISPLAYSETTINGS */"
> > i.e. have it apply to all windows, not just changed display settings:
>
>I have committed this code in CVS.  It does change the semantics of the
>video code by now using a palette in windowed mode, which I'm not sure
>is the intended behavior.  Can you check to make sure the new code works
>properly both in fullscreen mode and windowed at 8-bpp?
>
>Thanks!
>         -Sam Lantinga, Lead Programmer, Loki Software, Inc.