Camelia, the Perl 6 bug

IRC log for #darcs, 2013-07-09

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:34 mizu_no_oto joined #darcs
00:41 favonia joined #darcs
02:56 jgross_ joined #darcs
03:30 mizu_no_oto joined #darcs
03:57 preflex joined #darcs
05:53 dolio joined #darcs
06:23 nomeata joined #darcs
06:23 javier_rooster joined #darcs
06:29 javier_rooster joined #darcs
06:32 lelit joined #darcs
06:34 javier_rooster joined #darcs
06:39 javier_rooster joined #darcs
06:44 javier_rooster joined #darcs
06:45 gh_ joined #darcs
06:51 raichoo joined #darcs
07:06 adnap joined #darcs
07:51 alexei joined #darcs
08:55 owst joined #darcs
09:13 kaol joined #darcs
09:13 haasn joined #darcs
09:15 owst joined #darcs
09:15 adnap joined #darcs
09:15 gh_ joined #darcs
09:15 lelit joined #darcs
09:15 preflex joined #darcs
09:15 alexsuraci_ joined #darcs
09:15 Igloo joined #darcs
09:15 teratorn joined #darcs
09:15 mornfall joined #darcs
09:15 MasseR_ joined #darcs
10:59 gh_ joined #darcs
11:08 gh_ hi
11:54 Heffalump hi
12:25 konundra joined #darcs
12:55 donri joined #darcs
13:03 burp joined #darcs
13:15 uniquenick joined #darcs
14:08 mizu_no_oto joined #darcs
14:22 Gowilla joined #darcs
14:47 dolio joined #darcs
14:51 bsrk joined #darcs
15:14 raichoo joined #darcs
16:19 bsrk joined #darcs
16:48 konundra joined #darcs
16:50 bsrk joined #darcs
16:51 mizu_no_oto joined #darcs
16:53 * Heffalump appears
16:54 Heffalump bsrk: how are things?
16:54 bsrk Hi
16:54 Heffalump I have a couple of things to discuss as well as general progress - (a) frequency of blogposts (b) are there going to be any other periods of unavailabilty after this week?
16:56 bsrk I don't think there will be more periods of unavailability, atleast one's that I can forsee.
16:59 Heffalump ok. Re blogposts, we agreed on every two weeks which I viewed as the minimum necessary. You haven't really kept up with that, so I think it would be best to aim for at least weekly, which means I can start hassling you a week after the last blogpost. I hope it would also make them easier to write as there would be less for each one.
17:01 bsrk Sorry :-( Lets shift to once per week then.
17:01 Heffalump cool, thanks
17:01 Heffalump ok, onto code :-) I see there are some new paches in your repo, want to update me on what you've done?
17:01 Heffalump [about to go underground]
17:03 sm hi guys. Maybe I can check my oauth config with bsrk while Heffalump is underground
17:03 bsrk okay
17:04 bsrk You need to register your application at Google and Github respectivily
17:05 bsrk You can register Github's in your user settings
17:06 sm bsrk: I'm just testing google so far. I wasn't sure if I have the callback urls right. I seemed to see a register screen after logging in with google
17:06 bsrk Similarly, for google go to api console in developer's site
17:07 bsrk code.google.com/apis/console
17:07 sm yes. how does https://gist.github.com/simonmichael/5959175 look to you ?
17:08 bsrk yes, they look good to me
17:09 sm ok.. could we do a quick test of login on the dev instance,  http://hub.joyful.com
17:10 sm it may mess up your account on hub.darcs.net until I fix it. I'll test mine, hang on
17:10 bsrk It seems to be  working for me.
17:11 sm so right now "simon" is not associated with google; I click log in, google; I see the register form. That's working as intended ?
17:12 bsrk Yes, that is how it was intended. If you try to log in with an account that is not associated it will redirect to register
17:12 bsrk Login as usual, and then link the account with google
17:12 sm I find it a little confusing
17:13 bsrk Yeah, I thought there was a message saying this. :-)
17:13 sm I wonder do other oauth sites have the same flow ?
17:15 bsrk Yes, I think so.
17:16 bsrk Hmm. I did put in a message, but am using the old variable.
17:16 sm oh boom, I broke production again - see http://hub.darcs.net
17:17 sm time to apply your password schema change there
17:17 bsrk okay
17:20 konundra joined #darcs
17:21 sm urgh.. I'm having to pull everything
17:22 sm your patches are nicely grouped for reading, but they don't leave the code in a buildable state - you probably knew that
17:23 bsrk they don't?
17:23 bsrk I thought they did.
17:24 sm nope. not to worry
17:24 bsrk Ah, if you mean on a patch by patch basis, then yes.
17:24 sm right
17:25 bsrk but if you take then as a feature group, it is ought to work.
17:25 sm yes indeed
17:25 sm I thought I could just pull the password field change, but no
17:26 sm last chance for me to release 1.1 before merging your stuff. I didn't want 1.1 to include a schema change. Is anybody else running darcsden, or is it a non-issue ?
17:27 sm non-issue, I think
17:29 * Heffalump reappears
17:29 Heffalump bsrk: what happened to the documentation you were going to write for oauth?
17:30 Heffalump sm: I'm running it on my machine, but I could live with blownig away my db
17:30 bsrk I wrote it somewhere, but it is not in the repo. :-)
17:31 Heffalump bsrk: please fix that :-)
17:31 sm Heffalump: good, I figured
17:32 Heffalump sm: on the other hand the question of schema upgrades needs to be addressed at some point, and it's easier to shake out the bugs in that process with low-impact users.
17:32 bsrk sm, I fixed missing message issue, but the patch is stuck in my machine for now. :-(
17:32 Heffalump but I don't really care either way
17:33 bsrk Also, there are some other patches in darcsden-gsoc that fix bugs like these
17:33 sm I feel like not solving the whole migration/db issue right now.. someone might replace it all with sqlite soon
17:33 * sm wonders who could be bribed...
17:34 sm one sec, I'm still bringing up production servers
17:34 Heffalump sm: I've got some changes nearly ready to submit which abstract the Couch interface in a dumb way.
17:34 Heffalump i.e. just replace each call to COuch with a typeclass method.
17:34 Heffalump my idea was that we can then evolve that typeclass to be a nicer interface
17:35 Heffalump bsrk: ok, so what have you been doing recently?
17:36 Heffalump also, it would be good to check your progress against the plan
17:37 bsrk Mostly just bugfixes and ui tweaks; edit file is sort of usable now.
17:38 Heffalump the code review with sm agreed on you writing some tests, didn't it?
17:38 bsrk Yes, thats right.
17:39 Heffalump ok, so in terms of features, you're up to the week 4 (file editing) but there's still some polishing left for some?
17:42 bsrk That is basically it. Features upto file editing can withstand basic usage; but none of them have fully been realized either.
17:43 Heffalump what week of GSoC proper are we in now?
17:45 bsrk GSoC coding period started 22 days ago, so this is the start of the 4th week.
17:47 Heffalump ok - do you think you'll be able to get all the features up to week 4 properly polished by the end of the week?
17:48 bsrk polished is relative - They are doing what they are supposed to do, but it could be more.
17:48 bsrk For example, OAuth could import more data
17:49 Heffalump I was thinking about things like documentation and tests, at least.
17:49 sm I think we'll get lots of polish ideas once this is live on darcs hub
17:49 bsrk Oh
17:50 Heffalump yeah - I guess I'm thinking of "finished as a feature from the pov of GSoC" even if there is nice-to-have follow-on work
17:54 sm ok! after the obligatory deps tussle, the oauth and password reminder changes are merged and deployed on hub.darcs.net
17:54 sm thanks a lot bsrk
17:54 Heffalump yay!
17:55 bsrk Documentation seems straight forward. tests are a little murkier, so I'll do those basic ones we agreed on.
17:55 Heffalump yes, that sounds fin
17:55 bsrk sm: OAuth includes github?
17:55 Heffalump not everything has to be tested, but where agreed/appropriate they do need to be done
17:56 sm bsrk: yes.. oh I haven't set up the env vars for either yet. I'll do that
17:57 Heffalump generally I have the sense that there's a bit of a risk of you falling behind soon, given that you're only just on schedule having started early. So I'd like to just make sure we keep a close eye on progress.
17:57 bsrk sm: there seems to be some css problems too..
17:57 Heffalump avoid problems before they happen rather than having to play catch-up later
17:58 bsrk heffalump: I agree
17:58 Heffalump do you have a list of todo items at a detailed level (e.g. test X, document oauth so that someone else can set it up from scratch, etc)?
18:00 bsrk Yes, 1) document external configurations for these features 2) login tests 3) relations between repositories and it's forks
18:01 Heffalump could you put that on the wiki or somewhere else I can track it easily?
18:01 Heffalump does that mean that editing is done?
18:02 sm dangit why is darcs hub not seeing my env vars
18:03 bsrk Yeah, editing is pretty much done in the sense of a Gsoc feature.
18:03 Heffalump tests?
18:04 bsrk It's mostly UI?
18:04 Heffalump darcsden itself is mostly UI, that's not an excuse for it to have no tests
18:05 bsrk sm: I think restarting the instance will make the problem go away.
18:05 Heffalump as we've discussed before, the tests can either drive the server via http, or they can go via a lower layer and test the underlying code directly
18:06 Heffalump and I understand that things like github and oauth are even harder to test because they require external services too. But editing is quite self-contained and there are potentially issues with patches been recorded correctly and the like.
18:07 sm bsrk: not in this setup (running from supervisord). I need to spend more time on this, I have hidden the github and google features for now
18:07 Heffalump I was never very comfortable with the env vars adding yet another config mechanism.
18:07 sm hub.darcs.net should be fully functional and stable again at this point (let me know if not)
18:08 sm I don't like them, but some do (heroku). Is there any alternative ?
18:08 sm I guess a config file
18:08 Heffalump my biggest problem is that at the moment some things are configured by env vars and some things by compiled in stuff in Settings.hs
18:09 Heffalump setting env vars can be a pain (especially if there are several to manage) and compiled-in stuff is also a pain, so we have two different and painful mechanisms :-)
18:09 sm yes there's a problem with a lot of settings being compile time (Settings.hs) which would be better as runtime (darcsden.cfg and/or command-line flags)
18:09 Heffalump one config file sounds nice to me, but there may be other good options too
18:11 Heffalump bsrk: we haven't really finished, but my train will be arriving soon so let's arrange the next meeting while the rest of the conversation continues. Friday would suit me, or possibly Thursday though the time would be more unpreditable.
18:11 bsrk Okay, friday it is.
18:12 Heffalump can you put the todo items on the wiki and keep that up to date?
18:12 sm Heffalump: your db abstraction work sounds good, probably a useful step even if we drop couchdb (which seems likely to me) ?
18:13 sm we shouldn't overengineer this either, if it's not going to affect a lot of folks
18:13 bsrk heffalump: okay
18:13 Heffalump sm: yeah, that's the idea. I'd like to have something really simple even if really slow for small installations.
18:13 Heffalump btw, is all the session state in redis? i.e. if redis stays up and darcsden restarts, do clients see anything break?
18:14 Heffalump assuming that the restart itself is clean
18:14 sm I assume things keep working smoothly, eg they'd still be logged in
18:15 Heffalump which sort of suggests that thethe dumb alternative to redis should be an on-disk store rather than an in-memory store, though of course that makes it more persistent than redis if you don't take special steps there.
18:15 sm or maybe that's in cookies, I dunno
18:15 sm I'm pretty sure we could use the main db for sessions as well if we want
18:15 Heffalump if it's in cookies it would imply that actual credentials are in cookies which sounds very dangerous, though I don't really have an instinctive feel for web stuff
18:17 sm my feeling is we should try consolidating all db stuff into persistent + sqlite or postgres if needed, then see how much of the repo metadata we can move out to _darcs
18:18 Heffalump makes sense
18:18 Heffalump biaw
18:18 sm persistent will give us static guarantees on db and automatic migrations
18:18 sm ok, see you Heffalump
18:19 sm me too.. see you bsrk
18:19 alexei joined #darcs
18:26 favonia joined #darcs
19:03 favonia joined #darcs
19:13 mizu_no_oto joined #darcs
19:29 dolio joined #darcs
20:06 lispy joined #darcs
20:16 lispy_ joined #darcs
21:21 mizu_no_oto joined #darcs
22:00 mizu_no_oto joined #darcs
22:09 konundra joined #darcs
22:17 felipe joined #darcs
22:40 dolio joined #darcs
22:41 favonia joined #darcs
23:43 pr joined #darcs
23:43 pr joined #darcs
23:44 kaol joined #darcs
23:56 dolio joined #darcs

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary