Incorrect database update of genre

[caasland]

I've been using rockbox for a few days on  Sansa e280 and noticed that a couple genre's are misread by the database.

For example, one song shows up in Genre 'Me' when it should he 'Metal'. I have the same problem podcasts showing up in "Pod" or "Podc" when they should be in "Podcast".

The tags are correct, and both the original firmware and MediaMonkey displays the correct Genre's.

Should I create a bug report out of this, or has it been resolved (I wasn't able to find another similar problem).

[soap]

When you say the tags are correct...what kind of tags are we talking about?
ID3v1?  ID3v2.3?  ID3v2.4?  Other?
I strongly suspect if you view the raw tags of the files in question you will find truncated versions of them.

[caasland]

Using mp3tag I can view the v1 and v2.3 tags which looks fine.

Is there another way to view the "raw" tags?

[bascule]

Have you got lots of info in any of the other metadata fields (like really long artist/album names, or long comment strings)?

I seem to recall that there is a fixed (total) size of metadata read per file and that the rest just gets discarded (cut off) when it's full...

[caasland]

The comments section is pretty long.

so ... Does that imply that if one field has lots of data, the Rockbox software will discard data from other fields?

[LinusN]

so ... Does that imply that if one field has lots of data, the Rockbox software will discard data from other fields?

Yes, that is unfortunately correct...

[caasland]

Was this a design choice, a hardware limitation, or "we haven't gotten around to fixing that yet" sort of thing?

i.e. if I roll my sleeves up and get into the code, would this be something I could fix? I ask, because there are additional things I would like to change (like bookmarking in the database, re-arranging menu items, implementing "delete on play" etc).

[bascule]

I may be completely wrong here, but I think it's more of a trade-off at the current time. Remembering that all metadata goes on the buffer and thus uses up space that could be used for audio data and maybe coupled with the fact that memory is statically allocated, so there has to be a cut-off somewhere, creates the current situation.

I'm sure it could be altered (if not what you would call fixed) for your own purposes by a bit of code tweaking (e.g., not reading the comments field at all; using a larger metadata buffer for all files etc.)

It would need to have demonstrable benefits for all to be included into the SVN trunk, however.

But if you are going to add bookmarking for the Database at the same time, then please roll those sleeves up right now and get on with the job

[caasland]

What I would recommend is ensuring the important stuff (album, artist, genre, etc) has enough space, then add the comments or other fields as space allows.

At the moment, the easiest option was to write a MM script to strip comments.

Also created a few more MM scripts to make the Sansa work for me .. so no need to roll up my sleeves (yet).