Camelia, the Perl 6 bug

IRC log for #mojo, 2012-09-19

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

All times shown according to UTC.

Time Nick Message
00:18 Jason_ joined #mojo
00:19 gurugeek joined #mojo
00:55 laouji joined #mojo
01:08 jnap joined #mojo
01:10 xaka joined #mojo
01:19 vel joined #mojo
01:28 * tempire optimizes
01:32 biz joined #mojo
01:45 l3l1p joined #mojo
01:57 d4rkie joined #mojo
02:28 jnap joined #mojo
02:32 l3l1p joined #mojo
02:43 noganex_ joined #mojo
02:52 xaka joined #mojo
02:52 crab hi tempire. what are you optimising?
02:53 tempire my desk.
02:53 tempire and posture
02:55 arpadszasz joined #mojo
03:14 l3l1p joined #mojo
03:47 d4rkie joined #mojo
03:51 crab are they optimal yet?
03:58 D4RK-PH0ENiX joined #mojo
04:09 zpmorgan_ joined #mojo
04:26 Foxcool joined #mojo
04:29 l3l1p joined #mojo
04:59 d4rkie joined #mojo
05:26 pingup3rl joined #mojo
05:33 AmeliePoulain joined #mojo
05:40 * marcus optimizes his attitude
05:44 xaka joined #mojo
05:45 inokenty joined #mojo
05:54 xaka joined #mojo
05:55 spleenjack joined #mojo
06:09 xaka joined #mojo
06:10 laouji joined #mojo
06:18 Vandal joined #mojo
06:24 l3l1p joined #mojo
06:24 xaka joined #mojo
06:52 fhelmber_ joined #mojo
07:17 judofyr joined #mojo
07:25 pingup3rl joined #mojo
07:38 batman joined #mojo
07:45 Adurah joined #mojo
07:45 andrefs joined #mojo
08:13 dminus_ joined #mojo
08:26 arpadszasz joined #mojo
08:37 kmx sri: ad our last discussion - https://github.com/kmx/mojo-server-metyl
08:48 Britzel_ joined #mojo
08:53 Britzel joined #mojo
08:56 andrefs joined #mojo
09:01 cosmincx joined #mojo
09:03 baton8 joined #mojo
09:11 * tempire puts the jhthorsen dotcloud blog post on mojocasts.com
09:17 nour any way to change the name of the cookie from mojolicious to something else ?
09:22 crab edit the source
09:23 crab ;-)
09:24 crab there's probably an accessor to set the name. look at the documentation.
09:26 Kwa http://search.cpan.org/~sri/Mojolici​ous-3.42/lib/Mojolicious/Sessions.pm
09:38 sawtooth joined #mojo
09:40 dvinciguerra joined #mojo
09:42 dvinciguerra joined #mojo
10:35 andrefs joined #mojo
10:50 kmx joined #mojo
11:03 pau4o joined #mojo
11:20 asarch joined #mojo
11:22 sri kmx: it doesn't work on max os x
11:22 sri "Thread 3 terminated abnormally: Can't create listen socket: Address already in use at repo/mojo/lib/Mojo/IOLoop.pm line 142 thread 3"
11:24 sri umm mac os x
11:24 purl mac os x is a BSD so it's whatever any other POSIX system uses
11:26 kmx sri: well, Mojo::IOLoop should not start listening at all, the listening is done in dispatcher thread
11:26 kmx sri: please try to revert to Mojo 3.42
11:27 sri nothing changed in 3.43
11:27 kmx sri: all testing I have done was on MS Windows box
11:28 kmx sri: I'll try to do a bit more testing late, gonna leave for some $work affairs right now
11:28 sri perl 5.12.4 btw
11:29 kmx sri: I am on 5.16.1
11:31 kitt_vl joined #mojo
11:32 kmx sri: I can reproduce on my linux + 5.12.3 - will investigate later
11:35 marcus https://gist.github.com/3749197 woot?
12:33 mire_ joined #mojo
12:34 nic Anyone trying out the threaded server is going to be someone quite adventurous, so probably wouldn't mind a requirement on 5.16.1
12:36 nic zpmorgan_: Mojolicious::Plugin::Run::Open3 works well, and should be multiplatform
12:36 nic but somehow I've lost a bundle of code out of Mojolicious::Plugin::Run
12:37 nic so I'll fix that tomorrow, then we should have something to continue testing with
12:38 nic (just a warning not to spend time on the code you pulled the other day; it's known-broken ;) )
12:38 nic (sorry)
12:39 zpmorgan_ nice!
12:39 zpmorgan_ I personally wouldn't mind that 5.16.1 requirement
12:39 nic oh, re 5.16.1, I was meaning kmx's code
12:40 nic so far I've only been testing ::Run with 5.16.1, but should start testing lower versions on Friday
12:53 judofyr joined #mojo
12:56 judofyr joined #mojo
12:59 nour how can i test for memory leaks
13:11 d4rkie joined #mojo
13:11 sri \o\
13:11 sri /o/
13:20 vel joined #mojo
13:22 bluescreen joined #mojo
13:35 jnap joined #mojo
13:39 Foxcool joined #mojo
13:55 judofyr joined #mojo
13:58 judofyr_ joined #mojo
13:59 gryphon joined #mojo
14:07 kmx sri: I have hopefully fixed "can't create listening..." problem
14:07 * sri is busy exploring the latest xkcd :)
14:13 bjoernfan How do I sign up for the mailinglist..?
14:15 bjoernfan Or is that Google Group-thingie not a proper mailing list?
14:23 memowe sri: thanks. #timekiller
14:24 wjh joined #mojo
14:24 sri that xkcd is goning to cost millions in lost productivity
14:26 Kwa I've just spent 10 minutes on it and only turned it off because the boss came in.  I couldn't help but click on prev...
14:26 Britzel_ joined #mojo
14:27 Kwa Now random...
14:28 Leandr joined #mojo
14:32 vervain I've started with the traditional 'follow left wall' algorithm ;-)
14:33 sri damn... two of these four comments are absolutelt horrible http://blogs.perl.org/users/peter_martini/201​2/09/subroutine-signatures---the-plan-v1.html
14:33 sri that's why we can't have nice things!
14:33 sri where do people like the come from?
14:34 memowe sri, Kwa: spoiler http://i.imgur.com/GDJPL.png
14:34 sri vervain: that way you'll go straight through the caves and miss the whole middle part ;p
14:34 vervain *nod*... I did say staring with... I can't help myself :-)
14:35 vervain Of course this is only after seeing if 'ripping' was trivial... doesn't look like it is. :-)
14:36 vervain "The whole system is rigged against the middle class and families trying to climb out of poverty!" <giggle>
14:37 sri i have come here to chew bubblegum, and i'm all out of gum!
14:38 vervain That's a shame.
14:39 sri \o/
14:44 knshaum joined #mojo
14:58 pau4o joined #mojo
15:06 labrown joined #mojo
15:09 espent joined #mojo
15:15 l3l1p joined #mojo
15:15 schoenfe1d joined #mojo
15:15 schoenfe1d hi
15:17 aptituz I have trouble with a controller which is logged as beeing unexisting. If I load exactly that controller with a use statement in the applications pm its working fine. any idea what could be wrong?
15:17 sri o/
15:20 Kwa aptituz: You sure it's a problem with the controller, rather than a problem with the route?
15:22 cosimo joined #mojo
15:23 aptituz no, I am not (did I imply that?). well, however, the one other yet existing controller is working fine
15:26 Kwa aptituz: What's the exact error you're getting. I'm only guessing based on what you've said that it might be a routing problem, rather than a controller problem.
15:27 aptituz sec, putting together a paste with the messages (and what I think might be interesting apart from that)
15:30 aptituz here we go: http://paste.debian.net/190912/
15:31 aptituz Kwa: do you see anything wrong with the route?
15:32 vervain how about ->to('tag#index')?
15:32 vervain Not htat it should matter... just pointing out a shortcut for you to consider.
15:33 aptituz yeah, I'm aware of the short-cut, but I do not really like it
15:33 aptituz if I come and revisit my code in a few monthes I probably would not remember what it means.
15:34 mire_ joined #mojo
15:36 aptituz hm. the :id placeholder seems to be the culprit
15:39 Kwa aptituz: Oh?
15:39 Kwa aptituz: I was thinking along the lines of the namespace you've set might not be in your path.
15:39 aptituz I have not specified a default value, since it should not be optional. however, if I specify a default value it "works"
15:40 Kwa Excellent if that's what the problem is. :)
15:41 aptituz Kwa: isn't it visible from the paste, that the namespace is in the path? ;)
15:41 Britzel joined #mojo
15:42 aptituz hm, well, probably it would even be better to specify a regular expression for the id, instead of nothing
15:42 sri @INC is not visible in the paste
15:44 aptituz indeed, but default @INC usually includes '.', doesn't it?
15:44 sri i hope not
15:45 sri also... what is '.'? most web servers change the cwd
15:46 Britzel left #mojo
15:47 aptituz well, with your "i hope not" statement you seriously confused me :D its been this way on Debian since I'm doing perl, which is somewhat about 10 years.
15:48 aptituz (IIRC)
15:49 sri oh, actually i'm wrong, '.' is in @INC
15:50 sri how awkward
15:50 sri if you depend on that your apps will never ever be portable
15:54 aptituz well, one shouldn't depend on that for distribution, indeed. but for development it seems fine to me ;)
15:54 Britzel joined #mojo
15:59 sri until something does a chdir, and you didn't notice
16:10 xaka joined #mojo
16:14 abra joined #mojo
16:22 biz joined #mojo
16:25 jzawodn joined #mojo
16:28 judofyr joined #mojo
16:28 sri kmx: looks like hypnotoad is about twice as fast as the thread based server
16:30 sri memory usage is 45mb for hypnotoad and 80mb for the thread based server
16:30 sri (with the micro benchmark test app)
16:32 sri the thread based server is only slightly faster than a single threaded daemon
16:33 nour when a plugin gets loaded into mojo does mojo have access to the instance after register is called? wondering because I'm referencing the application from it
16:33 sri oh, and that was with keep alive
16:33 sri without keep alive the thread based server gets completely demolished
16:34 sri it's like 20 times slower then
16:34 sri but that should be fixable with multi accept
16:37 sri oh, maybe i should mention a few numbers :)
16:38 sri this is on a dual core macbook, with keep alive we have hypnotoad at 1800 rps, metyl at 1100 rps, and daemon at 1000 rps
16:39 sri without keep alive it's hypnotoad at 1100 rps, metyl at 50 rps, and daemon at 800 rps
16:41 sri memory usage is 45mb for hypnotoad with 5 processes, 80mb for metyl with one process, and 20mb for daemon with one process
16:42 sri all running on a threaded perl 5.12.4
16:42 pau4o left #mojo
16:43 sri both hypnotoad and daemon gain between 50 and 100 rps on a non-threaded perl 5.16.1 here
16:45 sri hmm
16:46 sri yea, so daemon on a non-threaded perl 5.16.1 beats metyl on a threaded perl 5.12.4 by a few rps
16:48 Britzel joined #mojo
16:51 sri switching the recurring timer from 0.5 to 0.025 increases rps to 1300 for metyl
16:55 jnap joined #mojo
17:04 fhelmber_ joined #mojo
17:07 rem_lex|pivo joined #mojo
17:09 vishwanathms joined #mojo
17:20 andrefs joined #mojo
17:28 asarch joined #mojo
17:50 dpetrov_ joined #mojo
18:04 jwang joined #mojo
18:19 dminus joined #mojo
18:19 gryphon joined #mojo
18:24 tolid_air joined #mojo
18:29 purl joined #mojo
18:33 batman joined #mojo
18:35 kmx sri: thanks for nice benchmark, I have put those numbers into metyl's readme
18:37 sri kmx: i still think you should theoretically be able to beat hypnotoad
18:38 kmx sri: I do not even dare to think about such a thing like beating hypnotoad :)
18:38 sri you just need multi accept and better load balancing
18:38 kmx not sure what factor has EV
18:38 sri up to 100 rps with hypnotoad
18:39 kmx sri: I was thinking how different it would be if Mojo::Reactor will support idle event
18:40 sri wouldn't work so well i think
18:40 sri you want many sockets active concurrently
18:40 kmx as each thread is checking newly accepted connections only once in 0.5s
18:41 sri when i set it to 0.025 it only made a small difference
18:41 kmx ok
18:41 Adurah Something about coro.
18:42 sri i think with idle event sockets would just be handled more sequentially
18:44 sri Adurah: what about coro?
18:45 Adurah It's better than threads, or so its creator says, heh.
18:45 sri coroutines don't work like you think they do
18:46 Adurah You don't know what I think...!
18:47 sri coroutines are not an alternative to threads
18:47 Adurah Now you must argue with the Coro module creator!
18:47 sri you brought it up
18:48 kmx Adurah: my goal was to check ithreads way mostly as it might be potentially handy in the future for production deployment on MS Windows
18:48 Adurah Windows process emulation...! Oh no!
18:48 batman what is metyl?
18:49 sri batman: a hack https://github.com/kmx/mojo-server-metyl
18:49 batman *looking*
18:52 batman hm... not sure if i get -why-
18:52 sri i find the central acceptor design interesting, but i believe it would work a lot better with sendfd
18:52 batman anyway... a completely different thing: would it be an idea for Mojo::Base -base; to export weaken?
18:53 sri batman: i think that has been discussed before
18:53 batman ok
18:53 sri don't remember any details though
18:53 batman that's fine
18:54 kmx sri: in fact metyl is doing a kind of sendfd - dispatcher passes accepted sockets to any of free workers (not via sendfd but via Threads::Queue::Duplex)
18:55 sri kmx: well, with sendfd you would have a handle to wake up the worker
18:55 sri which eliminates all the delays
18:55 kmx but worker is already running mojo loop
18:56 kmx so I need somehow to step in
18:56 sri you would have a readable watcher on the sendfd pipe, whenever a new fd comes in the loop wakes up right away
18:56 andrefs joined #mojo
18:57 kmx right with pipe it would work unfortunately I pass socket via Threads::Queue::Duplex not a pipe
18:58 kmx but of course I can change that part
18:58 kmx anyway pipes do work differently on Windows
18:58 sri there are no pipes on windows i heard :)
18:59 sri at least IO::Pipely uses sockets to emulate them afair
18:59 kmx there are named pipe which a completely another kind of animal :)
18:59 sri ah
19:00 kmx anyway I would prefer something portable
19:03 Leandr joined #mojo
19:10 vifo joined #mojo
19:10 vifo Hey folks!
19:10 * osprey_ still want socks support in useragent
19:11 vifo Is there a way to combine POE with Mojo? I'd finally like to implement a job server (each job being processed separately), with a RESTful interface handled by Mojo. Are there any similar approaches?
19:12 sri vifo: easy, you just have to fix POE::Loop::EV :)
19:12 vifo sri: you're joking, right? :)
19:12 sri not at all
19:12 batman left #mojo
19:13 vifo what do you mean with fix?
19:13 sri http://mojolicio.us/perldoc/Mojoli​cious/Guides/Cookbook#Event_loops
19:13 sri this would work for POE too
19:13 sri POE::Loop::EV needs to be updated for the latest EV api i believe
19:13 batman joined #mojo
19:16 vifo but how exactly would this work then? i am comfortable with mojo (i think at least), but new to POE (it does not have to be POE anyway, i am simply searching for a way to 1. control the job server via HTTP/Mojo, 2. have kind of "master" job controller, 3. have jobs run separated)
19:17 vifo would Mojolicious itself provide this?
19:18 sri see Resque, i don't think you need a event loop at all :S
19:19 vifo ruby is right now not an option :)
19:20 vifo yep, but this is exactly what I am searching for (well maybe, without Redis)
19:20 vifo or what I'd like to implement with Perl/Mojo and some other components
19:21 sri https://metacpan.org/module/Resque
19:25 biz joined #mojo
19:36 vifo joined #mojo
19:42 baton8 joined #mojo
19:53 mire_ joined #mojo
20:08 vifo joined #mojo
20:11 bluescreen_ joined #mojo
20:31 xaka joined #mojo
20:37 vifo joined #mojo
20:51 jnap joined #mojo
20:53 jnap left #mojo
20:56 kmx joined #mojo
21:38 lukep joined #mojo
21:41 bluescreen joined #mojo
21:44 perlite_ joined #mojo
21:58 vifo joined #mojo
22:26 vifo joined #mojo
22:31 davido joined #mojo
22:33 davido_ joined #mojo
22:41 davido_ A discussion on Mojolicious going on at PerlMonks: http://www.perlmonks.org/?node_id=994418
22:45 Adura joined #mojo
22:46 vifo joined #mojo
23:00 sri davido++ # excellent answer
23:02 * sri is quite happy with how security flaws have been handled in the past, other frameworks prefer to hide theirs completely
23:14 vifo joined #mojo
23:32 tempire holy thorough answers batman
23:32 tempire er, davido
23:32 tempire davido++
23:32 tempire you deserve a reward
23:33 * tempire assigns this car: http://www.craigslist.org/ab​out/best/sea/2979885438.html
23:36 vifo joined #mojo
23:47 vervain The CL posting _must_ be in the best of CL.  That's flippin' brilliant. :)  I wish I was still young and creative like that. But oh no, now my kids steal it all.  Quantum Physics suck. ;)
23:59 vifo joined #mojo

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