[FFmpeg-devel] [PATCH 03/11] fate/cover-art: Add test for muxing cover arts to FLAC

Andreas Rheinhardt andreas.rheinhardt at outlook.com
Tue Aug 31 17:10:03 EEST 2021


Andreas Rheinhardt:
> Also covers muxing and demuxing of nonstandard FLAC channel layouts
> and the multi-dim-quant option of the FLAC encoder
> (all of which was hitherto uncovered).
> 
> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt at outlook.com>
> ---
>  tests/fate/cover-art.mak            |  10 +++
>  tests/ref/fate/cover-art-flac-remux | 111 ++++++++++++++++++++++++++++
>  2 files changed, 121 insertions(+)
>  create mode 100644 tests/ref/fate/cover-art-flac-remux
> 
> diff --git a/tests/fate/cover-art.mak b/tests/fate/cover-art.mak
> index c42a4bc499..cbcc03fabf 100644
> --- a/tests/fate/cover-art.mak
> +++ b/tests/fate/cover-art.mak
> @@ -46,6 +46,16 @@ FATE_COVER_ART_REMUX-$(call ALLYES, FILE_PROTOCOL MP3_DEMUXER MJPEG_DECODER \
>                         += fate-cover-art-mp3-id3v2-remux
>  fate-cover-art-mp3-id3v2-remux: CMD = transcode mp3 $(TARGET_SAMPLES)/exif/embedded_small.mp3 mp3 "-map 0 -map 0:v -map 0:v -c:a copy -filter:v:0 scale -filter:v:2 scale -c:v:0 bmp -c:v:1 copy -c:v:2 png -metadata:s:v:0 comment=Band/Orchestra" "-map 0 -c copy -t 0.1" "" "-show_entries stream_tags:stream_disposition=attached_pic:stream=index,codec_name"
>  
> +# Also covers muxing and demuxing of nonstandard channel layouts into FLAC.
> +FATE_COVER_ART_REMUX-$(call ALLYES, FILE_PROTOCOL MOV_DEMUXER OGG_DEMUXER   \
> +                                    ALAC_DECODER MJPEG_DECODER SCALE_FILTER \
> +                                    CHANNELMAP_FILTER ARESAMPLE_FILTER      \
> +                                    FLAC_ENCODER BMP_ENCODER PNG_ENCODER    \
> +                                    FLAC_MUXER FLAC_DEMUXER FLAC_DECODER    \
> +                                    FRAMECRC_MUXER PIPE_PROTOCOL)           \
> +                       += fate-cover-art-flac-remux
> +fate-cover-art-flac-remux: CMD = transcode mov $(TARGET_SAMPLES)/lossless-audio/inside.m4a flac "-map 0 -map 1:v -map 1:v -af channelmap=channel_layout=FL+FC,aresample -c:a flac -multi_dim_quant 1 -c:v:0 copy -metadata:s:v:0 comment=Illustration -metadata:s:v:0 title=OpenMusic  -filter:v:1 scale -c:v:1 png -metadata:s:v:1 title=landscape -c:v:2 copy -filter:v:3 scale -metadata:s:v:2 title=portrait -c:v:3 bmp  -metadata:s:v:3 comment=Conductor -c:v:4 copy -t 0.5" "-map 0 -map 0:a -c:a:0 copy -c:v copy" "" "-show_entries format_tags:stream_tags:stream_disposition=attached_pic:stream=index,codec_name" "-f ogg -i $(TARGET_SAMPLES)/cover_art/ogg_vorbiscomment_cover.opus"
> +
>  FCA_TEMP-$(call ALLYES, RAWVIDEO_MUXER FILE_PROTOCOL) = $(FATE_COVER_ART-yes)
>  FATE_COVER_ART = $(FCA_TEMP-yes)
>  $(FATE_COVER_ART): CMP = oneline
> diff --git a/tests/ref/fate/cover-art-flac-remux b/tests/ref/fate/cover-art-flac-remux
> new file mode 100644
> index 0000000000..07a8400128
> --- /dev/null
> +++ b/tests/ref/fate/cover-art-flac-remux
> @@ -0,0 +1,111 @@
> +26a0d04dadbdd2703a78c603f7901bc9 *tests/data/fate/cover-art-flac-remux.flac
> +1103460 tests/data/fate/cover-art-flac-remux.flac
> +#extradata 0:       34, 0x92060af7
> +#tb 0: 1/44100
> +#media_type 0: audio
> +#codec_id 0: flac
> +#sample_rate 0: 44100
> +#channel_layout 0: 5
> +#channel_layout_name 0: 2 channels (FL+FC)
> +#tb 1: 1/90000
> +#media_type 1: video
> +#codec_id 1: png
> +#dimensions 1: 200x200
> +#sar 1: 2834/2834
> +#tb 2: 1/90000
> +#media_type 2: video
> +#codec_id 2: png
> +#dimensions 2: 485x359
> +#sar 2: 1/1
> +#tb 3: 1/90000
> +#media_type 3: video
> +#codec_id 3: mjpeg
> +#dimensions 3: 199x300
> +#sar 3: 72/72
> +#tb 4: 1/90000
> +#media_type 4: video
> +#codec_id 4: bmp
> +#dimensions 4: 485x359
> +#sar 4: 0/1
> +#tb 5: 1/90000
> +#media_type 5: video
> +#codec_id 5: mjpeg
> +#dimensions 5: 199x300
> +#sar 5: 72/72
> +#tb 6: 1/44100
> +#media_type 6: audio
> +#codec_id 6: pcm_s16le
> +#sample_rate 6: 44100
> +#channel_layout 6: 5
> +#channel_layout_name 6: 2 channels (FL+FC)
> +0,          0,          0,     4608,       14, 0x242b036d
> +1,          0,          0,        0,    66888, 0x1fd72743
> +2,          0,          0,        0,   438706, 0x8bcb753b
> +3,          0,          0,        0,    21671, 0x7d1c6030
> +4,          0,          0,        0,   522758, 0x7a63572f
> +5,          0,          0,        0,    21671, 0x7d1c6030
> +6,          0,          0,     4608,    18432, 0x00000000
> +0,       4608,       4608,     4608,     7636, 0xc3775d85
> +6,       4608,       4608,     4608,    18432, 0xaaab4d62
> +0,       9216,       9216,     4608,     5700, 0x8b9a3cc7
> +6,       9216,       9216,     4608,    18432, 0x49410f52
> +0,      13824,      13824,     4608,     5375, 0x1fa6c33b
> +6,      13824,      13824,     4608,    18432, 0x2d52e439
> +0,      18432,      18432,     3618,     4060, 0x59fd7219
> +6,      18432,      18432,     3618,    14472, 0xd2d47505
> 

This test gives different results depending upon whether
lpc_apply_welch_window_sse2 is used or not. Here is the diff:

--- ./tests/ref/fate/cover-art-flac-remux	2021-08-31 03:50:13.923155025
+0000

+++ tests/data/fate/cover-art-flac-remux	2021-08-31 04:17:10.016165436 +0000

@@ -1,4 +1,4 @@

-26a0d04dadbdd2703a78c603f7901bc9 *tests/data/fate/cover-art-flac-remux.flac

+6f275651fdb5af696a95f9f0a1afbadb *tests/data/fate/cover-art-flac-remux.flac

 1103460 tests/data/fate/cover-art-flac-remux.flac

 #extradata 0:       34, 0x92060af7

 #tb 0: 1/44100

@@ -51,7 +51,7 @@

 6,       9216,       9216,     4608,    18432, 0x49410f52

 0,      13824,      13824,     4608,     5375, 0x1fa6c33b

 6,      13824,      13824,     4608,    18432, 0x2d52e439

-0,      18432,      18432,     3618,     4060, 0x59fd7219

+0,      18432,      18432,     3618,     4060, 0xcb396324

 6,      18432,      18432,     3618,    14472, 0xd2d47505

 [STREAM]

 index=0


I presume that it has something to do with this being the small last
frame and that maybe said function overreads a bit?
It also happens when one disables the uncommon option "-multi_dim_quant
1"; in fact the difference is more pronounced if one does so: Using
lpc_apply_welch_window_sse2 results in a one byte bigger frame. Notice
that the decoded audio is always identical.
It would be nice if someone with assembly skills could look into this.

- Andreas


More information about the ffmpeg-devel mailing list