Re: [Gnash-dev] [PATCH 2/2] Work around rendering problem in GemCraft Chapter One

[Sandro Santilli]

On Wed, Jan 30, 2013 at 12:56:22AM +0100, address@hidden wrote:
> Sandro Santilli wrote:
> > 
> > testsuite/actionscript.all/BitmapData.as has some tests using getPixel
> > to verify the color of pixels in a rendered bitmap.
> > 
> > You can build the tests at all supported SWF versions with
> > 
> >   make -C testsuite/actionscript.all BitmapData-v{4,5,6,7}.swf
> > 
> > Make sure you've run "make check" before that as the generation of
> > those SWF also need the presence of a Dejagnu.swf file which is
> > created as part of that. You'll need libming installed and found
> > at ./configure time.
> 
> Thanks for the explanation. I have now run the testsuite and got
> 3 unexpected successes (lines 538, 559, 573) with the patch.
> 
> AFAICS these are already testing for exactly the problematic case
> I tried to fix.

Thanks a lot, your patch was pushed in master with commit
880ee12f40397b45d4f96935fc03d8a81ecb8e26

> A fourth xcheck at line 574 still fails, though. I'm not sure
> I can help with that.

You can see what gnash gets by running like this:

utilities/gprocessor -v testsuite/actionscript.all/BitmapData-v8.swf |
 grep XFAIL

You'll notice that those tests are not giving much info about what's
expected and what's obtained. There's room to improve the test.

I've added some traces in the pixel color checker and I see:

25169:1] 328 TRACE: R E:0 O:11
25169:1] 328 TRACE: G E:0 O:11
25169:1] 329 TRACE: B E:255 O:254

It could be really that we need an higher tolerance.
Don't really know. Hopefully Benjamin Wolsey will help
you, as he made that testcase.

--strk;