#metabrainz

ruaok: any news on the CAA cache?

it'll be fine

monkey huesound doesn't load for me at all now

From https://test.listenbrainz.org/huesound/ ?

lucifer: not yet. this non-technical discussions can move quite slowly.

Works on test.lb monkey

oh ok! 👍

I only deployed it on test.LB, not on bono

I was deploying on bono from my end

ruaok any config changes to deploy on bono needed?

likely.

one sec

code changes may be needed too.

Sure no hurry

If it's too much effort, I would like to leave it out

Can I deploy on test.lb btw?

akshat: comment out these lines: https://github.com/metabrainz/listenbrainz-serv...

then you config.py needs to specify lb_db for the database names, rather than just db.

akshat: might be easier for you to work with monkey to deploy to test.lb.

Cool ruaok, the bono changes required seem minimal and I could do that

monkey let me know the process for test.lb when you're free. No hurry

Sure thing! I've been letting Github actions build the image for me:

Find the "Run workflow" button, search for the branch (caa-color-browse) and then set the docker image tag to "test": https://github.com/metabrainz/listenbrainz-serv...

!m lucifer

You're doing good work, lucifer!

we haven't started throwing you down that rabbit hole, akshat. it will take more time than a short convo to get you bootstrapped int our production setup

Hahah okayy

lucifer we are about to reach 10k downloads on mb android app soon. I would like to have a blog and some additional features out by then. We have around a month before that happens. Would you suggest any major updates for it? Currently we have the theme changes and some basic stuff

Well, it's not that tricky now that githug builds the images for us. Once the action is finished, SSH into the kiss server and you're two commands away: sudo su and once logged in as root: `~/docker-server-configs/scripts/update_services.sh test listenbrainz-web`

Or alternatively, for today, you can ask me to deploy :p

Great monkey. But I do feel that currently I don't have the use case to deploy on test because that would mean I commit everything and have things done to test.

But for things like testing frontend stuff maybe after making a tweak, I just want to test the functionality and for that bono has been great

I hear you

Yuss

lucifer: I can't get the tests on this to pass, ever.

same BS connection is closed error. is this an oppotunity for us to test something?

[listenbrainz-server] 14amCap1712 merged pull request #1679 (03master…timestamp-check): Reject listens before 2005 https://github.com/metabrainz/listenbrainz-serv...

lucifer: did you find the original code that rejected these listens?

akshat: sounds good. i don't have any particular ones in mind currently. feel free to suggest any or ask in forums.

ruaok: which PR?

alastairp: i tried but couldn't find it.

lucifer: https://github.com/metabrainz/listenbrainz-serv...

ruaok: it looks like only one test is failing due to that db issue. other 2 are failing an assert.

maybe related to this change

right. but my question was this: is this an opportunity for us to investigate?

or should I just go and fix the asserts and keep pressing re-test?

i think fixing asserts is fine. i'd really like for us to fix this error but have no clue where to start.

k, I'll fix the asserts then.

i occassionally get this error locally too so my hunch is this might be related to a comparitvely slow system but not sure.

plausible, yes.

ah

MIN_ID = 1033430400 # approx when audioscrobbler was created

🎉

this was it (oct 2002)

where the heck did you find that?

I went to the first commit and went forward from there :-P

i was searching for last and fm so makes sense i couldn't find it

but we didn't actually reject listens with an earlier date - we just used it as the lowest value that you could _filter_ by

my problem as well.

now that we found this, should we amend 2005 to 2002 then?

any idea why lfm has bad data?

re reject: LFM clearly has scrobbles with timestamps of 1970. probably best to reject

is this something that we can ask them about?

them no. RJ/russ, yes.

oh, speaking of lastfm - https://twitter.com/jonty/status/14519462916096... this is interesting

we should make one

well, at the very least perhaps someone (be it one of us or this user) could post on the lfm forums saying "wtf my data seems broken"

unless the issue is that lfm doesn't validate this data either, and it was a broken client who submitted it

that line chart looks nice, we have a bar chart but for this line chart looks better. definitely we should add one.

they do have 3 : Timestamp too far in the past. so i expect them to verify

interesting

maybe old scrobbling api didn't?

duh. i spent half hour figuring out why spotify reader tests were failing only to realise we don't setup timescale writer in unit test.

I feel ya

I'd absolutely love to have lucifer in Barcelona for 2 weeks to have a cleanup hack-week

when are we going to make that happen?

that would be fun. and really hard at the same time. we'd never leave the office, because hacking!

hehe :D

lucifer: when do you have two weeks free>

?

and you've been vaxxed, yes?

when do you have 2 weeks free, alastairp ?

yeah, i have been vaxxed. two weeks free not sure currently because uni is in session.

well, I guess the whole thing can be planned when we have a window for this.

yup sounds good! 😄

I will make 2 weeks free whenever we can do it

ideally not over Christmas/Reyes, but otherwise pretty open

ball is in your court, lucifer.

ideally with at least 2 months notice.

lucifer: alastairp : how do you two feel about this? https://github.com/metabrainz/listenbrainz-serv...

sounds good. i'll get back to you on the timeline once I get my schedule for rest of session.

k

ruaok: lgtm but acc to me this will probably eat up the original stack trace. if this is just for that resource is closed error, then its probably fine. because the stack trace is not useful for that anyways.

if it gets in the way, we can always remove it.

it re-raises the error though, and str(err) is just going to be the final item in the stacktrace anyway. what's the motivation for adding the extra log line?

but this has made debugging tests very frustrating.

👍

or did you intend to not include the `raise`?

alastairp: something about the tests eats the error message and I get no meaningful error.

this way I get a message in the test debugging when code in this block fails.

ah, interesting. sure, if this helps you get a notification about the error then that's fine

ok, then that PR is ready.

as lucifer says, you won't get a specific traceback, perhaps you could use the traceback module to get it if you need it

lucifer: once we merge that we'll need to dump a full dump to catch up. or we could be lazy about it and just wait a few days.

🦥

I wonder if we need a sentinel that we can add to comments indicating our various hacks so that we can go back and re-address them in the future to see if they've fixed

some time ago, i faced this issue. full traces not showing up. i setup a personal account on sentry.io and configured sentry in my test runs. turned out to be useful some times.

ruaok: fine with waiting. less than a week left.

k. a lot less hassle.

sorry rak.

Nov 1/2 everything should be better.

> or did you intend to not include the `raise`?

No takebacks, we were promised a `raise`

🤣

it's in the context of an exception, and you're clearly an exceptional case, reosarevok

That's me. An exceptional error.

Exquisite

exceptions: things that shouldn't happen in normal day-to-day situations

finally! the office fibre connection has finally been changed into my name. took 4 months.

now to see how long it take for them to start billing the right account.

ruaok: btw the stripe PR is ready for review.

ah yes. thanks for the nudge.

how about the two TODO items without a check?

ruaok: Yeah, I use https://beets.io/ to tag my music, so the tags should be equivalent to what picard would give. And sure, I can wait until early November for stats to catch up.

yes one is once you take a look, let me know if somewhere docs could be improved.

rak: really? those listens come from tagged tracks? remind me you LB user name again, something is amiss.

the other one is that stripe advises to return a HTTP 200 response as soon as possible. currently what we do is , return HTTP 200 after inserting in db. is that fine or should we respond before even inserting?

lucifer: is this the response of MeB to stripe in a webhook?

that sounds fine to me. I am guessing they don't want you to wait minutes or hours.

alastairp: yes

I think that "as soon as possible" in this context also includes "as soon as you've actually done something useful with the data"

that

ah yes, makes sense.

ruaok: ryanakca . I'm scrobbling with mpdscribble via the proxy interface

rak: looking at https://listenbrainz.org/user/ryanakca/ -- all those listens come from beets tagged files?

iirc the proxy interface ignores much stuff. https://api.listenbrainz.org/1/user/ryanakca/li...

I wonder if we have bugs in the proxy interface that isn't copying everything

any chance to use the real API and not the proxy interface?

any mpd plugins/client that use the native LB API?

ruaok: Everything on that first page, at least. There are some scrobbles a few pages back that are from a browser extension / youtube

because if your stuff is tagged, there is little point in complaining about a less than ideal mapping.

alastairp: There's https://github.com/kori/wylt , but I struggled a bit to get it working. I can take another look at it.

I was thinking of just writing one myself in a few weeks after my paper deadline.

yeah, because if you really care about your LB experience, don't use that compat api