Opened 16 months ago
Closed 16 months ago
#11448 closed defect (fixed)
[Regression] "swscale" ignores "param0"/"param1" after 7.1
| Reported by: | b7j6abhi3q | Owned by: | |
|---|---|---|---|
| Priority: | important | Component: | swscale |
| Version: | git-master | Keywords: | |
| Cc: | MasterQuestionable | Blocked By: | |
| Blocking: | Reproduced by developer: | yes | |
| Analyzed by developer: | yes |
Description (last modified by )
Summary of the bug: Passing different values of param0 and param1 to the scale filter produces identical output when it shouldn't.
How to reproduce:
% ffmpeg -i input -vf scale=-2:540:flags=bicubic:param0=1:param1=0 -f hash - SHA256=7083e8d529bc9639a0db1d9aff4253442fa227ac632e19151e863b91611c200d % ffmpeg -i input -vf scale=-2:540:flags=bicubic:param0=0:param1=0.5 -f hash - SHA256=7083e8d529bc9639a0db1d9aff4253442fa227ac632e19151e863b91611c200d ffmpeg version N-118406-g957eb2323a-20250203
Change History (7)
comment:1 by , 16 months ago
| Description: | modified (diff) |
|---|
comment:2 by , 16 months ago
| Analyzed by developer: | set |
|---|---|
| Cc: | added |
| Summary: | swscale ignores param0/param1 → "swscale" ignores "param0"/"param1"? |
| Version: | unspecified → git-master |
comment:4 by , 16 months ago
I should have mentioned the same command works in 7.1, more detailed example in which case the difference is also visually obvious if output to file:
ffmpeg version n7.1-184-gdc07f98934-20250206
% ffmpeg -i carphone_qcif.y4m -vf scale=1440:1080:flags=bicubic,setsar=1 -f hash - SHA256=f8d1a22960bd14de8933cdd55995dbfe553ab0d3e7fde78c307e7bd92b08f60c % ffmpeg -i carphone_qcif.y4m -vf scale=1440:1080:flags=bicubic:param0=1:param1=0,setsar=1 -f hash - SHA256=6556d0e7f59dde9d7c73ff2f9632bfbc388806f0022146a994ea24852b01506 % ffmpeg -i carphone_qcif.y4m -vf scale=1440:1080:flags=bicubic:param0=0:param1=0,setsar=1 -f hash - SHA256=90261f5618ceba055a46a5e6d18c422d36802576df554795816819e31ab3affb
ffmpeg version N-118431-g6da82b4485-20250206
% ffmpeg -i carphone_qcif.y4m -vf scale=1440:1080:flags=bicubic,setsar=1 -f hash - SHA256=f8d1a22960bd14de8933cdd55995dbfe553ab0d3e7fde78c307e7bd92b08f60c % ffmpeg -i carphone_qcif.y4m -vf scale=1440:1080:flags=bicubic:param0=1:param1=0,setsar=1 -f hash - SHA256=f8d1a22960bd14de8933cdd55995dbfe553ab0d3e7fde78c307e7bd92b08f60c % ffmpeg -i carphone_qcif.y4m -vf scale=1440:1080:flags=bicubic:param0=0:param1=0,setsar=1 -f hash - SHA256=f8d1a22960bd14de8933cdd55995dbfe553ab0d3e7fde78c307e7bd92b08f60c
sample from https://media.xiph.org/video/derf/y4m/carphone_qcif.y4m
The problem occurred with every input video that I tested and also with the lanczos filter.
comment:5 by , 16 months ago
| Status: | new → open |
|---|---|
| Summary: | "swscale" ignores "param0"/"param1"? → [Regression] "swscale" ignores "param0"/"param1" after 7.1 |
͏ Regression window:
͏ https://github.com/FFmpeg/FFmpeg/commits?since=2025-01-23&until=2025-02-03
͏ ; prominently 2025-02-02 ~ 2025-02-03
͏ (keyword "scale")
͏ Please investigate the exact commit?
͏ Thanks.
͏ ----
͏ Seemingly not within previous window..?
͏ https://github.com/FFmpeg/FFmpeg/commits/master/libswscale/graph.c?until=2025-02-07
͏ https://patchwork.ffmpeg.org/project/ffmpeg/patch/20250206215857.17894-1-jamrial@gmail.com/
comment:6 by , 16 months ago
| Priority: | normal → important |
|---|
A fix was sent: https://patchwork.ffmpeg.org/project/ffmpeg/list/?series=13878
comment:7 by , 16 months ago
| Reproduced by developer: | set |
|---|---|
| Resolution: | → fixed |
| Status: | open → closed |
Fixed in 268d0b6527cba1ebac1f44347578617341f85c35.



͏ "The specified values are specific of some scaling algorithms and ignored by others."
͏ Per: https://ffmpeg.org/ffmpeg-scaler.html
͏ Please further verify?
͏ See also: https://trac.ffmpeg.org/ticket/10993#comment:2