[FFmpeg-trac] #5586(undetermined:new): bug wit

[FFmpeg]

#5586: bug wit
-------------------------------------+-------------------------------------
             Reporter:  guenzoman    |                     Type:  defect
               Status:  new          |                 Priority:  important
            Component:               |                  Version:
  undetermined                       |  unspecified
             Keywords:  blend        |               Blocked By:
             Blocking:               |  Reproduced by developer:  0
Analyzed by developer:  0            |
-------------------------------------+-------------------------------------
 Summary of the bug:
 When i use the blend filter, the Y variable in the formula works
 differently in 2.x and 3.x builds.
 How to reproduce:
 video 1: https://dl.dropboxusercontent.com/u/4096865/1.mp4
 video 2: https://dl.dropboxusercontent.com/u/4096865/2.mp4
 {{{
 $ ffmpeg -i 1.mp4 -i 2.mp4 -filter_complex
 "blend=all_expr='if(gte((T+1)*H-Y,H),A,B)'" out.mp4
 }}}

 Basically, i tried all static 3.x builds and some static 2.x builds.

 In 3.x builds, it produces a video like this:
 https://dl.dropboxusercontent.com/u/4096865/3x%20blend.mp4

 In 2.x builds, it produces a video like this:
 https://dl.dropboxusercontent.com/u/4096865/2x%20blend.mp4

 It's like if the Y variable were like a mod(Y,44), because this command
 bellow produces a similar results, in 2.x builds:
 {{{
 $ ffmpeg -i 1.mp4 -i 2.mp4 -filter_complex
 "blend=all_expr='if(gte((T+1)*H-mod(Y,44),H),A,B)'" out.mp4
 }}}
 and this video:
 https://dl.dropboxusercontent.com/u/4096865/2x%20blend%20with%20mod.mp4

--
Ticket URL: <https://trac.ffmpeg.org/ticket/5586>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker