Just media conv

[EtomicBomb]

No description provided.

[EtomicBomb]

I am still working on this one so we shouldn't merge it until its in better shape

[vagos]

@EtomicBomb Good job! I tried running on my end in the container and it both checks fail. Does it pass on your end?

[EtomicBomb]

Yeah they both pass. That is curious

[EtomicBomb]

Can you send the md5sum of the inputs and outputs? @vagos

[EtomicBomb]

I pulled ghcr.io/binpash/benchmarks:main:
docker build -t bench2 -f- . <<EOF
FROM ghcr.io/binpash/benchmarks:main
COPY . .
RUN ["bash"]
EOF

and ran ./main.sh media-conv --small, ./main.sh media-conv, and everything passes. Can you provide more info about the error you're experiencing and about your workflow @vagos?

[vagos]

I see this one is also failing in CI

[EtomicBomb]

True. Testing locally on an ubuntu 22.04 image I get all of the mp3s failing to match.

[EtomicBomb]

I looked at one of the mp3s. Running md5sum on the whole file, the hashes don't match between debian and ubuntu. Hashing just the audio streams with ffmpeg -i ./results/to_mp3.small/file_example_WAV_10MG.wav.mp3 -map 0:a -f md5 - 2>/dev/null makes them match. The images have different versions of ffmpeg. I wonder if hashing the audio streams is robust enough.

ffmpeg version 4.4.2-0ubuntu0.22.04.1 Copyright (c) 2000-2021 the FFmpeg developers
built with gcc 11 (Ubuntu 11.2.0-19ubuntu1)

(from ubuntu-latest)

vs.

ffmpeg version 5.1.6-0+deb12u1 Copyright (c) 2000-2024 the FFmpeg developers
built with gcc 12 (Debian 12.2.0-14)

(from debian 12.7)

[vagos]

Great catch! Do you think you can make the CI use our container?

Hashing the audio streams is 100% okay (I think this will probably save us headaches in the long run).
I imagine this just requires changing the verify.sh script.

[vagos]

If we also want to use our container for the CI, make a separate PR :)

[EtomicBomb]

@vagos The benchmark passes CI now.