#musicbrainz

I've run into this a few times now, where on a release there's composer and arranger credits given for some tracks but no specific track artist. Which should I be using in cases like this?

or better yet, do we have a style guide for this sort of thing?

Hi yayako, it depends on the type of release; See https://musicbrainz.org/doc/Style/Specific_type...

For example, if it is classical music, just use the composer name as recommended in https://musicbrainz.org/doc/Style/Classical/Tra...

Hi ! I have an issue with our replication (that started yesterday). I saw a constraint violation on tag_idx_name. I thought I did something wrong so I recreated the db using recreatedb.sh but I have the same problem. Is it a known issue ? Did I do anything wrong ?

The impacted package seems to be: "This packet was produced (or begins) at 2023-10-23 16:03:16.672738+00"

Hi Matthieu52, the database has been moved to a new server yesterday, so it is more likely an issue on our side, thanks for reporting.

I saw that this replication packet is indeed the one that was generated right after the planned maintenance

According to the mirror.log, https://metabrainz.org/api/musicbrainz/replicat... might be the culprit

yeah, a tag index was corrupted in the main DB yesterday, which allowed some tags with duplicate names to be added...I will work on fixing the replication packet in about an hour when I'm awake (?)

Thanks !

I opened a ticket for follow-up: https://tickets.metabrainz.org/browse/MBS-13333

MBS-13333: Replication is blocked by tag duplicates

yvanzo: thanks, was there a replicated mirror somewhere on our infrastructure that I could test with?

bitmap: no, the one on wolf is out-of-sync, I can update it though.

alright, I'll set one up locally too

fixing the packets by hand is possible, but it won't help people who have the packet half-applied (up until the broken seqid). so I think the best option is to simply have the replication script drop tag_idx_name before applying this sequence and then re-create it after the last broken packet is applied. users will have to update their code

people using mbslave (I'm not sure how many that is) will require manual intervention

maybe I'll think of a better solution while I'm setting up my mirror

currently recreating the mirror DB on wolf

thanks...I actually don't have enough space here to create another DB

maybe I can get away with only importing the derived dump

in the meantime I'm writing a script to reindex all indexes in production that are using either the default or musicbrainz collation...I already did this for UNIQUE indexes yesterday (the ones that could cause replication issues) but I'll do the rest now

ok, everything reindexed

atj: I'm getting 403s from atisket when i'm connected to my VPN - I assume this is intentional, for spam reasons?

Sciencentistguy: can you have a look at the forum thread? I posted about it. If you use the same VPN server I can remove that specific IP from the blocklist.

ah, yeah sorry i forgot there was a forum thread - its fine, i can just switch server temporarily while the ban-evader persists

Hi yvanzo, thanks! It's a soundtrack release, so style says I should use the composer, but it feels a little weird to leave out the arranger especially when a feat. vocalist is involved with the arrangement. Here's the credits for the release: https://media.vgm.io/albums/83/119138/119138-8c...

Do you think it'd still be best to relegate vocalist and arranger to the recording relationship in this case?

I'm actually not sure if in a case like this the arranger is intended to be the track artist, because that's how some other Japanese soundtrack releases on MB seem to be...

yayako: you might have misunderstood yvanzo - it only should have the composer in the assist field if it's a classical release:)

*artist field

aerozol: I checked the Soundtrack style page which says "If there are no track artists, use the release artist instead.", what gives me pause with a release like this is that it does specify "arranger" on the album art, and Japanese releases on MB seem to be split on whether arranger should be used in the artist field of a recording/track... that I've seen

I think arranger is probably the most suitable here, unless you mean I make the track credits something like "compose:name arrange:name"?

precisely as written, I mean

I don't think I can do it like that though, since I can't prefix the first name with a join phrase

Hmm, sorry, I can’t read Japanese so I didn’t see that. You can prefix the first name, but it’s clunky imo. I’ll see if there’s any JP editors on the Discord who’ve come afcross this

no problem, thanks!!

yayako: I know neither about soundtracks nor about Japanese releases. You might have better chances asking on the forums. In any case, adding relationships is worth it.

Definitely will be adding the proper composer/arranger relationships once I get the release in, for sure! I always try to put as much data as I can when I work on a release