avm2: Implement Matrix3D with 2D support only #18888
Conversation
cookie-s
force-pushed
the
matrix3d-transformation-with-2d-only
branch
from
7473740 to
ac547ca
Compare
cookie-s
changed the title
kjarosh
added
A-avm2
There was a problem hiding this comment.
This log is taken from Adobe Flash Player 32.
cookie-s
force-pushed
the
matrix3d-transformation-with-2d-only
branch
from
ac547ca to
69f6554
Compare
|
#18908 produced Git conflict with this PR and it actually implemented a kind of "2D/3D switch" inside DisplayObject. Rebased and followed its design now. |
cookie-s
force-pushed
the
matrix3d-transformation-with-2d-only
branch
3 times, most recently
from
a3a3018 to
6417e5f
Compare
|
Rebased. Thanks to #18911 , this PR gets much more slim. |
cookie-s
force-pushed
the
matrix3d-transformation-with-2d-only
branch
from
796ced1 to
35413ad
Compare
danielhjacobs
added
the
waiting-on-review
Replace stubs for flash.geom.Transform.matrix3D getter/setter with an actual implementation of Matrix3D with limited support. This implementation is just a proxy to the existing 2D matrix implementation. Therefore transformations beyond 2D transformation works differently from the expected result.
cookie-s
force-pushed
the
matrix3d-transformation-with-2d-only
branch
from
35413ad to
236c1cb
Compare
|
Rebased. |
| @@ -1,7 +1,8 @@ | |||
| num_frames = 1 | |||
|
|
|||
| [image_comparisons.output] | |||
| tolerance = 0 | |||
| tolerance = 10 | |||
| max_outliers = 4000 | |||
|
|
|||
| [player_options] | |||
| with_renderer = { optional = true, sample_count = 1 } | |||
There was a problem hiding this comment.
I believe the FP screenshot was done using high quality, in which case the sample count should be 4. This will hopefully make images closer to each other (so that we can possibly drop max outliers)
Also this commit can be safely squashed with the previous test commit
There was a problem hiding this comment.
Yes, I took the screenshot with high quality. I see, I didn't know the meaning of sample_count. Now fixed and squashed to 770d63e !
However, max_outliers cannot be dropped so much
Image 'output' succeeded: 3690 outliers found, max difference 255
Here is the diff created by imgtests:
The edges of the objects may be slightly different?
There was a problem hiding this comment.
Yeah, you're right. I've been trying to debug it but it seems that Flash just renders those objects differently with Matrix3D compared to Matrix, even if they represent the same transformation. I think we have to leave high max outliers and just add FIXME/TODO above it which describes the issue
There was a problem hiding this comment.
Oh, I didn't notice that 2D and 3D had the difference. True. Thank you, I added a comment: 81de5b6#diff-1cc83ff3bc64c1ef783edf3b104e090e3ea173a7298b6a3322ccdd7538fe80ecR5
| trace("sprite3D.transform.matrix3D.rawData", sprite3D.transform.matrix3D.rawData); | ||
| trace("mat3D.rawData", mat3D.rawData); | ||
|
|
||
| //// FIXME: matrix3D.rawData should be updated by x/y update. |
There was a problem hiding this comment.
I believe that's already the case, it's just that we don't reset tx,ty on matrix3D = null and the assertions below fail.
We could also do some more assertions related to non-indentity 3D matrices, I think we can safely do that without modifying the code
There was a problem hiding this comment.
cookie-s
force-pushed
the
matrix3d-transformation-with-2d-only
branch
2 times, most recently
from
f17cb03 to
019476d
Compare
Co-authored-by: Kamil Jarosz <[email protected]>
.matrix = A, .matrix = null, then .matrix3D has the info of the matrix A. .matrix3D = A, .matrix3D = null, then .matrix is reset to the identity.
cookie-s
force-pushed
the
matrix3d-transformation-with-2d-only
branch
from
019476d to
af60217
Compare
Retry #18810 .
Partially resolves #8033 .
#18810 (comment)
Commits
perspectiveProjection,matrix3Dstubs #18908 : 2D/3D switch ( https://docs.ruffle.rs/en_US/FlashPlatform/reference/actionscript/3/flash/geom/Transform.html#matrix3D ).