search

google-code-export / los-cocos

Automatically exported from code.google.com/p/los-cocos
Other
1 stars 0 forks source link

ZoomTransition scene can misbehave with some hardware #164

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago 
What steps will reproduce the problem?
1. run the stock test_transition_zoom.py

What is the expected output? What do you see instead?
Expected:
The src scene (grossini over blue background) going more transparent,
the dst scene going more visible,
ending with the dst scene in full view

Seen with gpu nvidia gforce 6200:
  seems that the dst scene is also going more transparent, transition ends with a full black window.

Seen with ati 3200 integrated gpu:
  works as expected

cocos 0.3 , 0.4, r1064

win xp 32 bits, python 2.6.5, pyglet 1.1.4

Additional notes:
I need to recheck this in both machines, I noticed the problem while working on 
issue 159 out of town.

Original issue reported on code.google.com by ccanepacc@gmail.com on 26 Nov 2010 at 2:47

GoogleCodeExporter commented 9 years ago 
integrated nvidia 6150SE  shows the defect
integrated intel 945G don't shows the defect

The problem is not in the library code, it is in the test code:

The BackgroundLayer defined there don't sets the current gl color, hence it 
could pick up the one set for another nodes (here an auxiliar node used in the 
transition).

Other related situation with bad behavior happens adding a particle system or a 
tinted sprite over the background layer: the layer would be tinted with the 
current open gl color (including alpha).

The fix is easy: add a 
            glColor4ub(255, 255, 255, 255)

as the first line in the 'draw' method of BackgroundLayer.

But, there is another thing to consider:
    BackgroundLayer appears aprox 60 times in the test_*
    With four implementations
        . the simplest one does not call transform in the draw method, hence it would behave like a peculiar CocosNode (not following .rotation, .position, .scale)
        . the next applies transform but have this color leak problem
        . two special cases

Being that the test_*.py can and should be used to experiment by library users, 
it should expose 'good' objects, not quirky ones.

So I think the fix for this issue should be backported to the other appearances.

I will fix in a few commits by

. change the special cases to use another name / become a common case
  ( test_custom_on_resize.py , test\test_transform.py )

. add a script to bulk upgrade the common cases to good behaved one
  ( tools\uniform.py )

. bulk upgrade, retesting

Original comment by ccanepacc@gmail.com on 23 Mar 2012 at 2:17

GoogleCodeExporter commented 9 years ago 
This issue was closed by revision r1208.

Original comment by ccanepacc@gmail.com on 2 Jul 2012 at 7:09

GoogleCodeExporter commented 9 years ago 
Fixed for all scripts with the defect at r1208
snapcompare tell rendered images don't changed when running in a good behaved 
tesbed.

We should have care to not introduce same weakness in new tests

Original comment by ccanepacc@gmail.com on 2 Jul 2012 at 7:13

GoogleCodeExporter commented 9 years ago 
fixed in cocos 0.5.5, closing

Original comment by ccanepacc@gmail.com on 14 Aug 2012 at 12:53