[Scummvm-devel] problems with 16 bit cursors

[Max Horn]

Hi Andre,

not sure what the current state of your proposal is, nor whether anybody beside Marcus responded to you (maybe in private / on IRC). Anyway, I recommend submitting this as a pull request, that way it won't be forgotten (as it will hang around on the pull request page), and people can discuss there.

bye,
Max

Am 06.03.2011 um 19:54 schrieb Andre Heider:

> hi list,
> 
> i ran into a problem i'd like to solve. this only affects backends
> which utilize hardware textures combined with frontends using 16bit
> cursors.
> 
> when a frontend sets a 16bit cursor, it can use any pixelformat it
> wants. all of our frontends are nice though, and use the same format
> as the screen.
> but that can result in eg. rgb565, which is good for the game screen,
> but not so good for the cursor because of missing alpha bits. backends
> have a problem then. they have to convert the pixel buffer in
> setMouseCursor() to a format with alpha. even when forcing such a
> format, frontends don't set alpha bits. so a backend has to set those
> bits based on the keycolor - which is double madness. the fact that
> most 16bit cursors are animated only makes it worse.
> 
> so this problem only exists on non-CLUT8 modes, with a focus on alpha
> bits. a solution is only needed when USE_RGB_COLORS is enabled and the
> game uses 16bit cursors.
> 
> what would be nice is a similar mechanism to what we do with the screen formats:
> frontend and backend provide a list of supported modes, and common
> code picks one based on those two lists.
> 
> i whipped up some patches with a proposal to solve this, see:
> https://github.com/dhewg/scummvm/commits/16bit (the last 6 patches)
> 
> i hope that throwing a few patches in here is enough to illustrate
> what i'm trying to solve, the diffs are not that big.
> 
> docs are missing or insufficient, some additional notes:
> - there is a default implementation of the new
> OSystem::getSupportedCursorFormats(). When not overwriting, it should
> behave like its without the patches
> - the new getCursorPixelFormat() is only supposed to be called after
> setting up a non-CLUT8 mode
> - the patches in engines/scumm are by no means final and not supposed
> to merged as-is. some parts are using the wrong colors, the changes in
> there are to provide an example of how it could look like for a
> frontend (and a way for me to test the changes)
> - support for more than one alpha bit is possible (groovie providing 4
> bits came up on this list recently)
> - the android patches show how nice and easy it would be for affected backends
> - the wii backend will have less overhead and intermediate buffers
> too. it also allows a ARGB3444 cursor without degrading the screen
> format from RGB565 to ARGB3444
> - at least the opengl will gain too, probably others too
> 
> engines using 16bit cursors:
> - gob
> - groovie
> - mohawk
> - scumm
> 
> this is not a major problem, but it doesn't allow me to implement the
> 16bit support on the android backend as efficiently as i'd like :)
> 
> any comments on this proposal are appreciated ;)
> andre
> 
> ------------------------------------------------------------------------------
> What You Don't Know About Data Connectivity CAN Hurt You
> This paper provides an overview of data connectivity, details
> its effect on application quality, and explores various alternative
> solutions. http://p.sf.net/sfu/progress-d2d
> _______________________________________________
> Scummvm-devel mailing list
> Scummvm-devel at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/scummvm-devel
>