zas: not sure why, but these errors have become fairly common inside the json dumps container on isaac (happened 22 times this month): "Can't connect to metabrainz.org:443 (Connection reset by peer)"
2025-10-21 29442, 2025
bitmap[m]
I'm adding some retry logic to admin/replication/LoadReplicationChanges so that it doesn't crash the script
When did the format of the CAA JSON change? The 'id' for an image used to be a string, it now appears to be a number. libcaa is broken for retrieving that as a result.
So I expect it was somewhere in between those years?
2025-10-21 29436, 2025
reosarevok[m]
But if that's the case, libcaa should have been broken for a long time 😅
2025-10-21 29422, 2025
adhawkins
reosarevok[m]: I suspect it has been, and just nobody uses it!
2025-10-21 29435, 2025
adhawkins
I use it in flactag, but currently only to request '/front' which doesn't require the ID.
2025-10-21 29450, 2025
pite_ has quit
2025-10-21 29407, 2025
reosarevok[m]
Oh, I see
2025-10-21 29448, 2025
reosarevok[m]
In that case, maybe! bitmap might still know more but given how old that is... I think accepting both makes sense, but we might want to try and reindex old images to have a number everywhere, bitmap?
2025-10-21 29401, 2025
pite joined the channel
2025-10-21 29427, 2025
adhawkins
To be honest, handling both is less of a change to libcaa, otherwise the interface and ABI changes due to the different return type.
2025-10-21 29401, 2025
Maxr1998_ has quit
2025-10-21 29447, 2025
Maxr1998 joined the channel
2025-10-21 29406, 2025
bitmap[m]
adhawkins (IRC): I would guess it changed to consistently output numbers when we deployed the Python artwork-indexer in 2024. though I don't think the previous Perl CAA-indexer consistently outputted them as strings, either (that's a common problem with JSON encoding in Perl if you don't force the context, which we weren't doing)
I agree number makes more sense in theory, though those are bigints (64-bit) which some languages will have trouble parsing by default
2025-10-21 29448, 2025
bitmap[m]
though currently the biggest cover art ID is only 1348490566944, which is between 2^40 and 2^41, so less than MAX_SAFE_INTEGER in JS (2^53 - 1) at least
2025-10-21 29401, 2025
bitmap[m]
adhawkins (IRC): anyway, I would certainly handle both cases as we aren't going to be reindexing everything anytime soon
I know it's late so don't expect an answer anytime soon, but since I see your emoji reaction... when you have a moment, could you please make a thumbnail for the art creator page that we can put on the explore page? I think it's time to promote the art creator so that it's not hidden at the bottom of the page.
