Re: ARGB cursors (was Re: Adjusting the 2.4 schedule)

[Keith Packard <keithp keithp com>]

Around 15 o'clock on Jun 26, Owen Taylor wrote:

>  But the exact interpretation is unclear. For XFree86,
>  XQueryBestCursor() will clamp it's returned result to the maximum
>  allowed hardware size, but use software fallbacks for larger cursors.
>  I think other servers may deal with over-large cursors in less-graceful
>  ways.

It's worse than you imagine -- XFree86 will occasionally do "bad things" if
you send too large a cursor.  It has a software fallback, but the ARGB
cursors uncovered many bugs in that fallback demonstrating how little it
was ever actually used.

In addition, enabling DRI or Xv anywhere on the screen should cause the X
server switch back to hardware cursors to avoid "issues" where software
cursors overlay DRI or YUV screen areas.  This, of course, will again
truncate cursors to the hardware size.  But, these code paths haven't been 
well tested and problems are likely.

Executive overview: respect the cursor size limits or face uncertain peril.

> Could gtkdnd.c use cursors instead of icons to display the DND
> drag icon? (Ignoring the problem of overlaying two cursors into
> a single cursor)

That would be nice -- I think it's pretty practical if we assume X cursors 
originate on the client side of the wire (which they will after we figure 
out how cursor themes work).  A standard cursor theme which used two-color 
cursors would make this practical even on legacy X servers.  I'd like to 
avoid depending on the X server for this -- cursor XIDs aren't stable 
enough for naming.

-keith