#musicbrainz-devel

I'm not doing a merge, I'm deploying beta

and seeing as you only 'fab beta' after merging to beta, then that assumption makes sense to me

If you want to just restart beta, then maybe we should add a restart target too

I have 10 musicbrainz checkouts spread out over four machines, the one I'm deploying from is rarely the same as the one I did the merge in.

Project musicbrainz-data build #200: STILL FAILING in 2 min 46 sec: http://ci.musicbrainz.org/job/musicbrainz-data/...

* Oliver Charles: jenkins.sh: Use CREATE EXTENSION

well I'm not sure I feel comfortable automating merges

if anything, put in a check that local matches origin

sure, I don't want to automate merges. I don't want to do any merging, I just want to deploy :)

I also want to know why this javascript is still not minifying correctly. grmbl.

ocharles: do you know what to do about the "Can't locate IO/All/Temp.pm" when trying to run compile_resources?

yea, you have to stick CARTON_HOME into the environment and fuss around with it that way

oh right

that should be done in admin/socket-deploy.sh though

Project musicbrainz-data build #201: FIXED in 3 min 22 sec: http://ci.musicbrainz.org/job/musicbrainz-data/...

export PERL_CARTON_PATH=$mb_server/local

that?

wahey, musicbrainz-data learnt about ISNIs, and it only took 2 hours to go from "my machine has no dependencies" to "done"

that's a pretty good turnaround :)

(I hate that compile_resources just says OK even if the compilation failed)

I think the exit code is 0 then, too

warp: any progress?

nikki: I confirmed myself for the summit

cool :D

nikki: happy enough with lolo?

from what I've seen, yes

fantastico

:)

one in one out server policy

nikki: yes.

cya tomorrow!

I dunno what's wrong with rika but there's two `postgres: musicbrainz musicbrainz_db_static [local] SELECT` hovering at ~100% CPU

is there a way to see what the queries are? I'm not sure if my sandbox is causing them or what

nikki: unbroken!

Project musicbrainz-server_beta build #564: STILL FAILING in 2 min 56 sec: http://ci.musicbrainz.org/job/musicbrainz-serve...

warp: Don't use reserved words as variable names in lib/filesize.js.

nm, I figured out how to see the queries. both are stuck on selecting release cover art.

now I can make lots of tickets for it >:D

:)

you're gonna be sick of cover art soon :P (assuming you aren't already XD)

http://tickets.musicbrainz.org/browse/MBS-6611 is the one for the filesizes at least

nikki: thanks!

also 12, 13, 14, 15 and 16 :P

I've put the ones I think are more important in the next fix version

12 sounds like a lot of work.

the others shouldn't be.

and 17 XD

I think that was everything...

now to upload stuff! :D

I'm leaving 12 and 17 unassigned

the other's I will try to fix tomorrow.

great :)

s/'//

and thanks :D

d

:)

warp: assigned 18 to you too

ianmcorvidae: (in case you don't see the chatlog) there's two postgres queries on rika that need killing (9212, 9359)

ianmcorvidae: although it'd be nice to figure out why they're stalled, because this is a common occurrence :/

aaaand now there's three. they seem to be spawned whenever I try to load a release group page (the query is from load_for_release_groups in lib/MusicBrainz/Server/Data/Artwork.pm)

hopefully the CPU doesn't combust in the meantime

warp: still around?

(or rather, back?)

warp: anyway, you might want to look at http://tickets.musicbrainz.org/browse/MBS-6619 while we're getting errors, since it's not very easy to reproduce otherwise

nikki: I'm back for a few minutes now.

no slowdown error when I try it now, so I guess I missed it :(

is there a way I can find out what it's returning when it gets stuck?

(I mean, so you know what it should be looking for)

I'd like to know what headers and status code it is returning

you'd have to open the developer tools of your browser (before the request) and go to the network tab.

the status code should be 503

ah, indeed.

but the question is whether that 503 makes it to the javascript. if the slowdown doesn't have CORS headers then it will be difficult to deal with it properly.

but I should just fake a 503 tomorrow with and without CORS headers and check if it deals with either of them properly.

there should be some event fired due to the request failure...

I seem to have found the network stuff

I'll keep an eye out

kepstin-work: if the request is denied because of CORS the browser gives you no information at all. it wouldn't surprise me if no event fired.

but anyway, I'll simulate it tomorrow.

right now I'm going back to the big blue room.

hmm. the cors check is supposed to be done via a HEAD request prior to doing the real request tho

isn't it?

so the upload won't run unless the cors check already passed

kepstin-work: if the response to the HEAD request had CORS headers that doesn't mean the slowdown response also has CORS headers.

any response without CORS headers is something the browser will stay silent about, it will not inform javascript of those response because javascript does not have permission to know.

hmm, you're right, it is supposed to revalidate the access control headers on the response to the real request

according to the cors spec, failing a cors check is supposed to act in the same way as if a network error occurred and the server couldn't be contacted. i think.

(it would probably be worth testing the ajax uploader with an unreachable/offline server too, since that theoretically should use the same error path, I think...)

warp: https://pastee.org/daux2

warp: also https://pastee.org/xpgjk

warp: tsk. you didn't set the tickets to in beta testing

http://slexaxton.github.io/Jed/ looks very cool

so instead of having a huge text_strings.tt that we have to manually edit, we could extract the strings directly from the javascript and convert the .po files to json

which I imagine could be cached the same way text_strings.tt is, if we wanted

ocharles: lolo can connect to rika, is what I'd meant, reading the conversation -- reverse tunnel lolo -> rika and then nikki can do a normal forward tunnel to rika

we just used /etc/hosts

it looks like it's set, yeah

but for everyone's future reference, the way it works is that non-rika stuff can connect to rika but not the other way around

e.g. the way I get data dumps onto it is I log into scooby and scp them from scooby -- can't do it from rika, since it can't connect that direction

heh that site of colours based on covers.. led me to like http://blog.jenniferdewalt.com/post/56319597560...

insane. it's I've wasted like an hour here :P

my fave so far is http://jenniferdewalt.com/endless_throbbers.html

\o/ I'm back :D