#musicbrainz-devel

bitmap: you could make it default to the user's current user interface locale

kepstin-laptop: I figured it should default to english, since it would replace the "translate foreign artist names to English" checkbox and use the same config variable

only the new option would use a generic description like "translate artist names to this locale:" or something, and have a dropdown menu

the list of locales we allow for aliases is rather large though, and it's not clear to me what should be done if someone selected "English United States" (if that should even be listed)

well, the match should generally first try the exact locale the user picked (e.g. en_US), then generic language (e.g. en), then maybe fallback to sort name?

of course.

but that's nitpicking :P

yup

that reminds me of the other issue with the alias locale thing; there's no way to specify what script an alias is in if the language that's set can be written in multiple scripts

not sure how many languages are customarily written in multiple scripts

japanese vs. kana, I suppose

hmm. how does it deal with the same name being used in multiple languages?

say, some artist with a japanese name who uses the same latin transliteration on german, french, and english releases?

can you add the same alias multiple times, each time with a different locale set?

Japanese is generally considered to be a single script, albeit composed of others.

not according to our database

hiragana, katakana, hiragana+katakana, and japanese

are all four separate scripts :P

I'm guessing it's based on ISO-15924.

it's based on some ISO standard

so probably

"Japanese" is meant to refer to the whole of the script. Hiragana, katakana and kanji.

even so

Katakana as a separate entry is useful for, say, Ainu.

yeah

even so, the difference between an alias in "Japanese" and one in "Hiragana+Katakana" is meaningful

See http://musicbrainz.org/release/94cc33be-6d7c-49... :P

presumably that didn't require dumping into the release editor, but :P

ianmcorvidae, it is, but maybe not in the artist locale alias bit for translating in picard.

Err, no, it didn't. I forget these things. :P

I'm tired. My brain is dead from maths. Other plausible excuses.

less math, more haskell

or something

Sadly, Haskell doesn't get me a better grade in my courses.

choose better courses :P

Other courses don't get me a maths degree. :P

And hell, even the CS department at my uni would be scared by Haskell.

choose a better school/degree, then :P

haha

clearly 'choose a better school' is the one here

I didn't have much choice when I chose. And now I'm too close to graduating to go elsewhere.

yeah

I understand these things, it's just more fun to pester you

I may as well finish and then go work on another degree that won't benefit from Haskell.

I'm hoping you meant "will" there :P

well, parsing

haskell

good pairing there XD

I didn't mean "will" there. I'm not a computer scientist. :P

but but ALL degrees should benefit from Haskell ;)

ianmcorvidae, I would like you to evaluate the truth of a statement of logic for me: "should" = "do"

Huh. I don't think we ever learned statements could take that truth value.

Should make these proofs I have to write a lot easier to manage.

haha

"The sum for n running from 0 of B_n/n! t^n increases without bound. Sure, why not? Therefore, the radius of convergence..."

Good night, fair ianmcorvidae.

good night

probably myself soon, as well

hello!

goodevening ruaoks

hi warp!

ocharles: I get a crash when the release editor tries to access ->entity on an Edit::Artist::Create instance.

ocharles: did anything change with those edits recently? (this occurs when trying to reproduce MBS-3414).

warp: ping

warp: read your email please!

ok!

ruaok: about the search servers needing regular kicking?

yep.

thanks.

ok, I'll try to keep an eye on it.

thx

warp: yes

entity no longer exists

you have to use entity_id and $c->model('Artist')->get_by_id

but I thought I fixed all references to that?

ocharles: well, apparently not :)

where's the reference to ->entity?

oh, I see some, bah

would you like me to fix them?

Wizard/ReleaseEditor.pm, around line 930, missing entities

ocharles: yes please. you can reproduce the crash with the instructions in MBS-3414

warp: pushed, but I can't reproduce that ticket :)

ocharles: pushed in master?

yea

ok, I'll give that a try.

no, still get it.

[error] Caught exception in MusicBrainz::Server::Controller::ReleaseEditor::Edit->edit "Can't locate object method "entity" via package "MusicBrainz::Server::Edit::Artist::Create" at /home/warp/code/mb/script/../lib/MusicBrainz/Server/Wizard/ReleaseEditor.pm line 929."

oh, probably forgot to pull.

ocharles: yes, you fixed it. artist was properly created, etc..

magic!

wow, test.musicbrainz.org is a bit messed up.

warp: how so?

http://test.musicbrainz.org/search?query=the+lo... doesn't fill the width

http://test.musicbrainz.org/release/aff4a693-59... has broken javascript.

well, re the width stuff, that's cause I was playing around trying to get overflow stuff working

time for a clean out though I think :)

but we should be releasing today anyway

so we can just reset

I have to run off to the muncipality whatever before it closes at 15:00

bbl

ijabz_: hi

djce: hi dave

Just chatting to our java guru here at work about our search server troubles

you know, where it runs out of file descriptors (I assume you're aware of this)

he suggests using "findbugs" http://findbugs.sourceforge.net/ on the codebase

"probably caused by bad error handling code in Java. Probably a try/catch block without a finally statement to close the file handle / socket."

Right, I was aware that one of the servers was running out of file handles, but do both do it ?

I think so, yes.

I'll check the logs.

My code doesn't really directly deal with files themselves, file handling is done by the lucene library

and I would be surprised if it has this kind of error

Yeah, check logs and let me know

I'm pretty sure it affects both servers

and if you open a resource via lucene, you'd still have to close it

I'll have a look but AFAIK I only open one resource per index, i.e 7 resources and thats it

once opened they remain open for the duration of search server running

Also, can configure search servers to use mapped mode ?init=mmap which doesn't use files (at least not in the same way)

Haven't received anything from ruaok about this, so Im in the dark about what his investigations uncovered

all in the chatlogs

though this is why I'm trying to push people to discuss things via tickets

much easier to track stuff

+ 1 on that

from logs , seems ruaok thinks that tomcat could actually be the problem

i know he wants to change from tomcat, but I'm not sure why that is

Well assuming something on the search server is leaking file handles sockets, and its not the search server code, or the lucene libs, or JVM it only leaves tomcat

Could try updating to latest version though first

Also we could update Lucene, although will introduce deprecations to be cleared yup I think it will still work , I'll check in a minute

[06:15] kepstin-laptopthat reminds me of the other issue with the alias locale thing; there's no way to specify what script an alias is in if the language that's set can be written in multiple scripts

in india they have idk 3-4 languages, and 9-13 scripts afaik