Camelia, the Perl 6 bug

IRC log for #mojo, 2010-08-14

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

All times shown according to UTC.

Time Nick Message
00:24 stephen joined #mojo
01:23 ashleydev joined #mojo
02:41 janus joined #mojo
02:52 MojoGuest864 joined #mojo
02:52 MojoGuest864 From: http://news.ycombinator.com/item?id=1277067 (12 hits)
02:53 MojoGuest864 joined #mojo
02:53 MojoGuest864 From: http://news.ycombinator.com/item?id=1277067 (13 hits)
02:53 MojoGuest437 joined #mojo
02:53 MojoGuest437 From: http://news.ycombinator.com/item?id=1277067 (15 hits)
02:54 MojoGuest563 joined #mojo
02:54 MojoGuest563 From: http://twitter.com/justinvi​ncent/statuses/12550383961 (2 hits)
02:54 MojoGuest563 left #mojo
02:55 MojoGuest563 joined #mojo
02:55 MojoGuest563 From: http://twitter.com/justinvi​ncent/statuses/12550383961 (3 hits)
02:55 MojoGuest563 left #mojo
03:43 ask joined #mojo
04:03 tl joined #mojo
07:33 baton8 joined #mojo
07:35 vti sri:         (in cleanup) set kevent failed: No such file or directory at /Users/vti/dev/mojo/lib/Mojo/IOLoop.pm line 776, <DATA> line 285 during global destruction.
07:36 sri vti: harmless
07:37 vti i have a nice plugin
07:37 vti give me couple minutes
07:43 yko joined #mojo
08:19 perlrocks Twitter: "#TweetButton plugin for #Mojolicious http://github.com/vti/mojol​icious-plugin-tweet_button #Perl" --vtivti http://twitter.com/vtivti/status/21135314751
08:21 sri vti++ # fun example
08:21 sri vti: but you need a better module skeleton
08:21 sri makefile and readme make a huge difference
08:22 vti ok...
08:23 vti cp ~/dev/mojo/Makefile.PL .
08:23 vti ;p
08:24 janus moin
08:24 janus another day, another http parser.  this time, with lemon :)
08:26 janus waaaaaay nicer than yacc and/or bison
08:27 sri parsing http is rather boring, the problem is interpretation
08:29 janus that depends imho
08:30 janus (oh wait, that's interpreting already ;))
08:31 perlrocks Twitter: "RT @vtivti #TweetButton plugin for #Mojolicious http://is.gd/ehaVy #Perl" --zakame http://twitter.com/zakame/status/21135797671
08:31 sri http 1.1 has so many edge cases it's not even funny
08:32 janus exactly the reason why i want to be as near to the protocol as possible
08:32 sri Expect: 100-continue is one of my favorites
08:33 sri a client needs to be prepared to handle a 100 Continue response even without requesting it for POST and PUT...
08:33 janus also the reason why i think your IOLoop/daemon is exactly the right direction
08:34 sri or the whole idea of HEAD requests...
08:34 sri you need to keep the request around to be able to intrepret the response right
08:34 janus sure
08:34 sri so screwed up
08:35 janus :)
08:36 sri btw. feel free to improve our http parser while you are at it :)
08:37 janus at least you identified it and named'em transactions... more than other are able to do ;)
08:38 sri think i should have said tokenizing instead of parsing above
08:39 sri Range: bytes=0-0 or so would have been cool instead of HEAD imo
08:40 sri http would be so easy to clean up
08:41 janus isn't that what mojo stands for? :)
08:42 sri i mean the spec
08:42 janus i know
08:42 sri ;p
08:44 janus good implementations show how protocols should/could be improved in my opinion
08:46 Htbaa joined #mojo
08:47 perlrocks Twitter: "RT @vtivti: #TweetButton plugin for #Mojolicious http://github.com/vti/mojol​icious-plugin-tweet_button #Perl" --kraih http://twitter.com/kraih/status/21136480775
08:47 janus all http implementations suck, mojo just sucks less? *g*
08:48 janus ... thinking of mutt/mime/... oh wait... http ;D
08:49 sri i hope, but atm i think we might have too much abstraction
08:49 sri instantiation is sadly still very expensive in perl
08:53 janus isn't the highest price in that regard user code? ideally it could be devided into startup cost and as less as possible per request?
08:54 sri ye of course, in a real app it doesn't matter much
08:54 sri but playing with micro benchmarks is fun :)
08:55 sri (not to mention maintainability)
08:55 sri optimized code is always a maintainability nightmare
08:56 janus but there're two kinds of optimizations: conceptional ones and tricks'n'hacks
08:57 sri true, i mean only performance hacks
08:59 sri a good optimization was getting rid of client side pipelining support
09:06 janus i love the simpletons (simple singletons without any dependent crap) *g*
09:06 janus reading mojo code is every single time a joy, i really have to say that!
09:06 sri \o/
09:09 janus hrhr, reminds me of marketing pages a la ``what our customers say'' ;D
09:11 sri actually, you should tweet your quote above
09:11 sri just write it a again and prefix it with !tweet
09:11 sri :)
09:12 sri !tweet ☁
09:12 janus ah right (no account there)
09:12 perlrocks sri: http://twitter.com/perlrocks/status/21137513348
09:12 sri we have a channel account
09:12 janus !tweet reading mojo code is every single time a joy, i really have to say that!
09:12 perlrocks janus: http://twitter.com/perlrocks/status/21137519335
09:13 sri :)
09:13 vti clean code equals Mojolicious ;p
09:15 sri code is never clean enough
09:18 vti same with bugs... :)
09:18 janus they can never be dead enough? :)
09:18 sri :)
09:20 baton8 joined #mojo
09:22 janus when did you shorten the code paths that much?! very very nice! :)
09:24 janus sri: i think i see a pattern that might lead to being ``scary''...
09:24 sri heh
09:25 janus remember what god said at the end of futurama where bender met him? ;)
09:25 janus unfortunately so true
09:27 janus ``If you do things right, people won't be sure you've done anything at all.''
09:27 janus and that's scary if there still is something that must have be dont by someone ;)
09:27 janus s/nt/ne/
09:29 janus in that regard, god is like being the best administrator ever... noone will notices because things just work *g*
09:29 janus ok, now drifting off
09:29 janus not the first time though :)
09:29 sri wow...way too philosophical for this early hour...
09:30 janus what can i say... one thing leads to another
09:35 janus speaking of code paths... are you able to test kqueue on osx too?
09:35 und3f sri, after 5e3c19f8b764a7fd0f02c4bc58c30bc8285a7878 commit fastcgi method become slowly and response in 30-40 seconds
09:35 und3f But works nice with e8cf2afeadde
09:37 sri janus: tests will always use poll atm
09:39 sri und3f: there was no fastcgi change in that commit
09:39 und3f yes, but fastcgi become slowly after that commit
09:40 sri doesn't make sense
09:40 und3f actually i got in log message that page is rendered
09:40 und3f but no result in browser
09:40 sri no idea
09:41 und3f i'
09:41 und3f i'll check code
09:41 und3f possibly find something
09:41 sri likely something in your code, tests and stuff only got faster
09:42 perlrocks Twitter: "mojolicious" (ja) --ykw http://twitter.com/ykw/status/21138738820
09:43 und3f sri, but site works perfect with daemon
09:43 sri sounds very strange, considering only stuff in mojolicious changed in that commit
09:45 janus openbsd w/ kqueue passes all tests
09:46 sri janus: no surprise, i'm developing with kqueue enabled :)
09:46 sri but good to know
09:48 und3f sri, just generated lite app doesn't work too
09:49 sri und3f: strange
09:50 und3f sri, but everything works with e8cf2afeadde70630aec1a49df337fb7308e37c6 commit
09:50 sri und3f: are you absolutely sure it was this commit? http://github.com/kraih/mojo/commit/5e​3c19f8b764a7fd0f02c4bc58c30bc8285a7878
09:50 und3f yes. this commit
09:51 sri there must be something else
09:51 janus sri: no suprise then it works, good to know too ;)
10:07 und3f sri, i found change that cause that slowness
10:07 und3f -    my $m = MojoX::Routes::Match->new($c->tx, $path);
10:07 und3f 63
10:07 und3f +    my $m = MojoX::Routes::Match->new($c, $path);
10:08 und3f If i change it back - fastcgi works fast again
10:08 und3f sri, how route dispatching connected to fastcgi?
10:09 sri there is no connection
10:09 sri it doesn't make sense
10:12 und3f sri, can you help me with mojo architecture? Who call's MojoX::Dispatcher::Routes:: dispatch?
10:12 sri Mojolicious.pm
10:41 und3f dunno :(
10:51 und3f mojo::server debug' tell that everything wrote
10:51 und3f but browser still waiting :|
10:52 und3f i also see response in wireshark
11:18 perlrocks Twitter: "RT @vtivti #TweetButton plugin for #Mojolicious http://github.com/vti/mojol​icious-plugin-tweet_button #Perl" --sharifulin http://twitter.com/sharifulin/status/21142701320
11:38 und3f sri, i've straced both variants of mojo_lite_app. Only difference is that workable variant calls 'close' on fcgi socket
11:38 und3f but i can't find where
11:40 sri why would it close the socket? something is very wrong there
11:40 sri do you depend on cgi like behavior?
11:41 sri as in the fastcgi script has to die as soon as it handled a request?
11:42 sri if thats true the whole thing might point to some kind of leak keeping the process alive
11:42 und3f sri, http://pastebin.ca/1916881
11:43 sri but thats assuming you really screwed up your fastcgi setup
11:43 und3f that diff from non-workable to workable
11:43 sri what is that supposed to tell me?
11:43 und3f there is close(4)
11:43 sri so?
11:45 sri why would there be a close? thats the only thing i've seen so far that seems wrong ;p
11:46 sri and read what i wrote above
11:46 sri maybe you should step back and just look for a memory leak
11:49 und3f sri, i am testing on clean just generated lite app
11:50 sri i know
11:52 perlrocks Twitter: "Mojolicious-Plugin-TweetButton-0.0001, Вячеслав Тихановский. CPAN: http://tinyurl.com/25sd83v" --cpan_linked http://twitter.com/cpan_linked/status/21144221593
11:52 vti i guess that's my name
11:53 sri now thats a broken twitter bot
11:53 sri they should have just used mojolicious...
11:53 vti ye
11:54 vti there is already 0.0002 btw...
11:55 perlrocks Twitter: "Mojolicious-Plugin-TweetButton-0.0002, Вячеслав Тихановский. CPAN: http://tinyurl.com/2chtz84" --cpan_linked http://twitter.com/cpan_linked/status/21144396889
11:56 vti right...
11:58 sri !tweet @cpan_linked your unicode support seems broken, that wouldn't have happened with #mojolicious. ☺
11:58 perlrocks Twitter: "@cpan_linked your unicode support seems broken, that wouldn't have happened with #mojolicious. ☺ --sri" --perlrocks http://twitter.com/perlrocks/status/21144526443
11:58 perlrocks sri: http://twitter.com/perlrocks/status/21144526443
11:58 vti :D
11:59 spleenjack joined #mojo
12:01 sri people in here don't tweet enough!
12:01 * sri cracks the whip
12:54 korshak joined #mojo
12:58 korshak joined #mojo
13:04 baton8 joined #mojo
13:37 GitHub105 joined #mojo
13:37 GitHub105 mojo: master Sebastian Riedel * 86a5bf8 (172 files in 41 dirs): fixed a design flaw in Mojo::IOLoop that prevented optimization by replacing the write_cb attribute with a write method - http://bit.ly/bJBXk2
13:37 GitHub105 left #mojo
13:38 sri vti: you were interested in that i think
13:39 sri the new ioloop synopsis and the examples show how easy it is
13:40 sri like examples/microhttpd.pl
13:41 sri i prolly broke lots of stuff, please test :)
13:41 vti oO
13:42 sri oh, i've also removed all unnecessary copyright notes
13:44 sri write calls are not ideal yet i think, profiling and possibly optimization patches would be very welcome
13:47 vti i was going to try vnc :)
13:48 GitHub160 joined #mojo
13:48 GitHub160 mojo: master Sebastian Riedel * c70039e (1 files in 1 dirs): make example more realistic - http://bit.ly/dbR337
13:48 GitHub160 left #mojo
13:53 sri feels way more natural to use ->write
14:12 GitHub179 joined #mojo
14:12 GitHub179 mojo: master Sebastian Riedel * c3cf599 (1 files in 1 dirs): fixed memory leak - http://bit.ly/brJ9yG
14:12 GitHub179 left #mojo
14:13 sri und3f[A]: that will fix your problem
14:13 sri but your fastcgi setup is broken, since your processes are not persistent
14:13 sri your proc manager restarts the fastcgi process after each request
14:32 GitHub31 joined #mojo
14:32 GitHub31 mojo: master Sebastian Riedel * da9819c (1 files in 1 dirs): made client quite a bit faster - http://bit.ly/bCGasH
14:32 GitHub31 left #mojo
14:33 sri almost let the client be slower after the big change...
14:34 sri now it can properly bypass kqueue/poll/epoll for writing
14:37 sri vti: what are your thoughts about ->write?
14:38 Htbaa joined #mojo
14:39 vti sri: i like it, but i will tell you more when i start rewriting stuff ;p
14:39 sri vti: ok, make sure to mention any other design flaws you might see, now is the time to fix them all at once! :)
14:42 sri http://github.com/kraih/mojo/blob/mast​er/examples/flash-policy-server.pl#L32
14:42 sri isn't it pretty? could amlost be a oneliner :)
14:43 vti ye, it is much more intuitive
14:46 sri and makes all examples twice as fast :)
14:49 vti too bad we don't have karma bot here ;p
14:57 sri ye, so much wasted karma
15:10 sri i just don't get why twitter would enforce oauth
15:11 sri limiting basic auth to https seems like the perfect solution until oauth 2.0
16:39 MojoGuest756 joined #mojo
16:43 MojoGuest799 joined #mojo
17:03 perlrocks Twitter: "Improved #mojolicious performance by a few percent today and fixed a small design flaw enabling many more optimizations. :) #perl" --kraih http://twitter.com/kraih/status/21164374045
17:06 yko joined #mojo
17:39 ashleydev joined #mojo
17:42 GitHub7 joined #mojo
17:42 GitHub7 mojo: master Sebastian Riedel * 61e39b0 (6 files in 2 dirs): cleanup - http://bit.ly/agM0hR
17:42 GitHub7 left #mojo
18:25 spleenjack joined #mojo
18:36 esskar joined #mojo
18:52 GitHub22 joined #mojo
18:52 GitHub22 mojo: master Sebastian Riedel * 1b4dcea (4 files in 4 dirs): added support for route name generation - http://bit.ly/dy6ZxR
18:52 GitHub22 left #mojo
18:53 sri get '/foo' => '*';   # will have the name "foo"
18:54 sri i'm a little afraid of too much magic here
18:57 GitHub80 joined #mojo
18:57 GitHub80 mojo: master Sebastian Riedel * 48fb4ec (1 files in 1 dirs): updated changes - http://bit.ly/9L0aBP
18:57 GitHub80 left #mojo
19:06 GitHub116 joined #mojo
19:06 GitHub116 mojo: master Sebastian Riedel * cafa73b (2 files in 2 dirs): updated pod - http://bit.ly/c4dZq7
19:06 GitHub116 left #mojo
19:22 spleenjack1 joined #mojo
20:07 spleenjack joined #mojo
20:23 perlite joined #mojo
21:06 jawnsy joined #mojo
21:58 ask joined #mojo
22:29 [1]awnstudio joined #mojo
23:00 [1]awnstudio joined #mojo
23:04 su-bzero joined #mojo
23:04 marcus joined #mojo
23:41 ask joined #mojo
23:57 ashleydev joined #mojo
23:58 ashleydev joined #mojo

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