Re: [Rhythmbox-devel] possible audio file corruption with Rhythmbox

[Jonathan Matthew <jonathan d14n org>]

On Wed, May 2, 2012 at 3:18 AM, Wolfgang Jeltsch
<2mw1nfaj acme softbase org> wrote:
> Hello,
>
> Rhythmbox seems to cause data corruption in Ogg Vorbis files under
> certain circumstances. I downloaded a video from YouTube using
> youtube-dl, which gave me a *.webm file. I then extracted the audio
> track using FFmpeg as follows:
>
>    ffmpeg -i $NAME.webm -vn -acodec copy $NAME.ogg
>
> That is, I told FFmpeg to just copy the Vorbis data, not re-encode the
> audio. The resulting *.ogg file could be played without problems.
>
> I moved the *.ogg file to my music folder. I started Rhythmbox and
> changed the metadata of the file. When exiting Rhythmbox,
> rhythmbox-metadata crashed with a SIGSEGV. When listening to the *.ogg
> audio file afterwards, I recognized short gaps at certain positions,
> each maybe about 0.1 seconds long. These gaps hadn’t shown up before.
> They also couldn’t be heard when playing the file with ogg123.
>
> So I suppose that Rhythmbox had corrupted the actual audio data when
> trying to change the metadata. I cannot tell whether the *.ogg file
> produced by FFmpeg wasn’t already corrupted in some way, but even if it
> was, I would have expected Rhythmbox to not further corrupt it.
>
> Of course, my assumptions can be wrong. Is there another explanation for
> the described problems? Have you heard of problems like this before?

Rhythmbox doesn't intentionally modify files without explicit user
action, but there have been bugs in the past that caused it to do so.
Once it's doing that, there's always the possibility that it'll damage
the files, though takes some steps to protect against that.

It sounds like you're experiencing at least one bug in rhythmbox.
Please open a bug report at
https://bugzilla.gnome.org/enter_bug.cgi?product=rhythmbox detailing
how you're reproducing this problem. If you could run 'rhythmbox -D
metadata' in a terminal and trigger the problem, the resulting output
would be helpful.

> Assuming that the file had been corrupted, I wonder whether it was only
> this file. Could it be that Rhythmbox has also corrupted files whose
> metadata I didn’t change at this point? After all, the message at
>
>    <http://mail.gnome.org/archives/rhythmbox-devel/2011-May/msg00046.html>
>
> claims that Rhythmbox sometimes changes audio files without the user
> knowing about this. Is this true? I had expected Rhythmbox to only
> change an audio file when I explicitly change its metadata.
>
> Now I wonder to what extend my music collection could be damaged.
> Normally, I would look at file modification times to see whether
> Rhythmbox has changed other audio files. However, Rhythmbox doesn’t seem
> to update a file’s modification time when setting metadata. At least,
> the modification time of the supposedly corrupted file was still at its
> old value after I had changed the metadata.
>
> Is it true that Rhythmbox changes audio files without updating their
> modification times? If yes, how can I tell if Rhythmbox has modified a
> file?

This was an unintentional side effect - it tries to preserve the
permissions and ownership of the original file, but was keeping the
mtime too. I've fixed this so mtime is appropriately updated now.