Re: [AD] stretch_blit bug

[Elias Pschernig <elias@xxxxxxxxxx>]

On Sat, 2007-05-05 at 08:55 -0600, Trent Gamblin wrote:
> On Thu, 2007-04-26 at 15:40 -0600, I wrote:
> > ... Everything appears to work just as the
> > original, except I'm not able to test Mode-X right now ...
> 
> I had a chance to test things out in Mode-X. Here is what I found:
> 
> I found it impossible to enter Mode-X unless Allegro is compiled with
> assembly. I tried under real DOS and Linux console. If that's the case,
> then the C version of the Mode-X stretchers will never be called.
> However:
> 
> Neither mine nor the original C versions of the Mode-X stretchers worked
> properly at all, when forced into use. Under DOS they fill with a solid
> color and under Linux they segfault.
> 
> The only version that I was able to get to work properly was the
> assembly version under DOS. Under Linux the assembly version gave
> incorrect colors and sizes.
> 
> I'm attaching a patch that fixes the inaccuracy of the current
> stretch_blit implementation. It shouldn't break anything that works
> currently. The fact that the C Mode-X stretchers don't work shouldn't
> matter, as they don't appear to be used (and they were broken to begin
> with).

I confirmed that it fixes the bug in the original testcase, and putting
it into my speed test it also is about 20% faster :) Therefore applied
to SVN.

-- 
Elias Pschernig