Camelia, the Perl 6 bug

IRC log for #mojo, 2011-04-09

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

All times shown according to UTC.

Time Nick Message
00:05 perlrocks Twitter: "OK, so today I've started to learn #mojolicious - web framework written in #perl. So far feels much better than #django..." --sirmacik http://twitter.com/sirmaci​k/status/56507952420491264
00:57 dku left #mojo
00:58 hidek left #mojo
01:17 hidek joined #mojo
02:20 shurex joined #mojo
06:35 yakudzo left #mojo
06:35 yakudzo joined #mojo
08:00 * sri yawns
08:14 arpadszasz joined #mojo
08:19 shurex left #mojo
09:35 arpadszasz left #mojo
09:45 fish_ i use net::telnet in a mojolicious::lite app and somehow it seems to die without the enclosing evals catching it. at least mojolicious' error handler should handle this, right?
09:45 fish_ i found nothing in the logfiles beside the fact that the fcgi proccess dies
09:46 fish_ and: i couldn't reproduce that with the mojolicious daemon..
09:47 fish_ the only thing i see in the mojolicious logfile is the requestimer / handling the route, than 'dispatching callback'. than it, according to the lighttpd logfile, it dies
09:50 sri if you can't reproduce it with the mojolicious daemon you've got a problem
09:50 sri could mean one of your modules messes with globals
09:53 fish_ does mojolicious handle only errors in functions like get '/' {}? i 'mess with globals', if you mean things like using a global variable for the Net::Telnet instance..
09:53 sri it's an eval cage
09:54 fish_ yes, i know. but for the whole mojolicious app or just for the coderef i hand over to, for example, get?
09:54 sri for the handler
09:55 sri more than the coderef, less than the app
09:55 fish_ ok
09:56 sri when i say globals i mean file descriptors and stuff
09:56 fish_ yeah, my strace give me some futher hints. its a EPIPE caused by writing to a closed fd
09:56 fish_ but the writing happens on a handler
09:58 sri anyway, if you're doing something the eval cage can't catch you're doing something evil
10:00 fish_ sri: well, im doing it like this: get '/' => sub { .. $self->render(_status()) }; ... sub _status { .. _nps('/s') .. } ... sub _nps { .. $telnet->cmd(shift) .. }
10:00 fish_ where $telnet is a global instance of Net::Telnet
10:00 fish_ is there a fundamental problem with this?
10:00 sri don't know how net::telnet works, so i can't say
10:01 sri there are millions of things it could screw up
10:01 fish_ ok, but your guess would be some serious 'evil' within Net::Telnet..?
10:02 sri guess, yes
10:02 fish_ (i even set Errmode to 'return' which should take care that it never dies on error..)
10:02 fish_ sri: ok, thanks for the hints :)
10:03 sri die would not cause problems
10:04 sri stuff like changing global PerlIO settings would
10:05 sri reassigning fd0...
10:06 fish_ it uses syswrite for writing the data, probably that dies (and Net::Telnet is not handling this at all. that not nice, but shouldn't cause this problem)
10:10 fish_ sri: ok, i'll dig into Net::Telnet looking for somthing suspicious
10:11 sri can be almost impossible to track down, considering it only affects fastcgi
10:55 * moritz can haz small doc fix, pull request sent
10:58 GitHub155 joined #mojo
10:58 GitHub155 mojo: master Moritz Lenz * b203724 (1 files in 1 dirs): [cookbook] CGI ScriptAlias needs as trailing slash, otherwise only the root URL works
10:58 GitHub155 mojo: master Sebastian Riedel * fefa5f9 (1 files in 1 dirs): cleanup
10:58 GitHub155 mojo: master commits 1cecda2...fefa5f9 - http://bit.ly/g9Yr85
10:58 GitHub155 left #mojo
11:08 moritz when I deploy as FastCGI on apache, most requests are quite fast, and some few take ages
11:09 moritz which aren't slow when I run it with the built-in daemon
11:09 sri how long are "ages"?
11:09 moritz like, 20s
11:10 sri weird
11:10 moritz nothing in the error.log
11:10 sri i was thinking maybe it's restarting the processes, but that's just weird
11:10 sri oh
11:10 sri which version?
11:10 purl which version is that?
11:11 moritz of what? mojo?
11:11 sri ye
11:11 moritz quite recent version from git (this week)
11:12 sri ok, then the bugfix is in
11:12 sri no idea
11:12 fish_ moritz: strace?
11:12 purl strace is http://sourceforge.net/projec​t/showfiles.php?group_id=2861
11:53 sri looks like we have the first negative side effect of lazy loading client/useragent
11:54 sri few months ago sigpipe would have been ignored and fishs app would have just worked
11:55 moritz oh and i have some ajax which updates part of my pages
11:56 moritz when deployed with fastcgi, it takes a long time between the arrival of the response and actual updating
11:56 moritz WTF?
11:57 * sri hates fastcgi
11:57 moritz sri: what else would you use together with apache? mod_perl?
11:57 sri why apache at all?
11:58 moritz because that's what the (shared) server is running
11:58 sri mod_proxy?
11:58 purl it has been said that mod_proxy is nice
11:58 moritz and then? hypnotoad?
11:58 * sri nods
11:59 sri btw. i'm starting to like haproxy
12:01 janus uhm... to_xml is strange here
12:01 janus <table cellpadding cellspacing width="865" /> <tr />
12:02 sri moritz: or you could try to debug fastcgi
12:02 sri MOJO_SERVER_DEBUG=1 will dump fcgi packets to the log file
12:03 janus gah, local version override crap
12:03 moritz sri: will see if I'll get around to it - need to go out with the small one first
12:04 sri if the bug was on our side you would be lucky though, mod_fastcgi and mod_fcgid are known to be rather fragile
12:20 Kulag left #mojo
12:20 Kulag joined #mojo
12:36 janus $ perl -Mojo -le 'g("http://www.worldatlas.com/cntycont.htm")->dom->find(qq|table[width="865"​]>tr>td|)->each(sub{print$_[0]})'
12:36 janus $
12:37 janus given the html/xml contains something that matches, shouldn't it print all td's content?
12:38 janus tried varies things now, using the latest and greatest but still fail to access td's, whereas it works so far in all isolation efforts :(
12:39 janus s/es/ous/
12:39 janus as in, sorry for the lack of a better test
12:42 sri janus: http://validator.w3.org/check?uri=http%3A%2F%2Fwww​.worldatlas.com%2Fcntycont.htm&amp;charset=%28dete​ct+automatically%29&amp;doctype=Inline&amp;group=0
12:42 sri broken html
12:42 purl broken html is just broken.
12:42 sri always check the html is actually valid before reporting a bug
12:44 yko left #mojo
12:45 yko joined #mojo
12:45 janus doh, crap
12:45 yko janus: i believe something wrong with html there
12:46 sri 332 things actually
12:48 janus libxml to the rescue once again :)
12:51 marcus sri: I thought you were supposed to be a liberal! :/
12:51 sri marcus: ?
12:55 marcus sri: 14:42 sri: always check the html is actually valid before reporting a bug.
12:55 marcus :)
12:55 * sri doesn't follow
12:56 marcus sri: I thought you tried to parse html like browsers do, not just according to spec.
12:56 sri there is an infinite number of ways to break html, do you seriously suggest we consider every single one a bug in Mojo::DOM?
12:57 sri i never said i would reject a patch, but i'm surely not looking for a lifelong quest
12:57 marcus only if it includes a failing test obviously =)
12:57 sri i'll forward those to you then ;)
12:58 * marcus goes to try to find something to eat in manhattan
12:58 sri feature requests are fine, but don't suggest they get the bug treatment
12:58 sri s/suggest/expect/
12:59 marcus I would rather say it's a veeeery minor bug.
12:59 marcus but whatever :)
12:59 sri based on what grounds?
13:00 sri what would be the final goal?
13:00 marcus to be able to scrape any page that renders well in ie6? :)
13:01 sri then perl would be the wrong language
13:01 sri if you want a headless browser, you should use a headless browser
13:34 yko left #mojo
13:36 yko joined #mojo
13:51 yko left #mojo
13:51 yko joined #mojo
13:51 GitHub63 joined #mojo
13:51 GitHub63 mojo: master Sebastian Riedel * b6de1a3 (3 files in 3 dirs): improved Mojo::DOM to ignore useless end tags - http://bit.ly/ft8Amo
13:51 GitHub63 left #mojo
14:08 yko left #mojo
14:12 yko joined #mojo
14:19 perlrocks Twitter: "I found what I was looking for. https: / / github.com / sharifulin / mojolicious-plugin-mail" (ja) --hoyo1111 http://twitter.com/hoyo111​1/status/56722840979910656
14:39 GitHub13 joined #mojo
14:39 GitHub13 mojo: master Sebastian Riedel * d06e58e (1 files in 1 dirs): cleanup - http://bit.ly/f32iz2
14:39 GitHub13 left #mojo
14:50 sri https://github.com/kraih/mojo/issues#issue/111
14:50 sri this is how it's done
14:52 perlrocks Twitter: "Opened my first wishlist issue for #mojolicious today, any takers? :) http://t.co/MUrYMZo #perl" --kraih http://twitter.com/kraih/status/56731208058544128
15:09 GitHub194 joined #mojo
15:09 GitHub194 mojo: master Sebastian Riedel * c319f32 (2 files in 2 dirs): pod improvements - http://bit.ly/fr91fP
15:09 GitHub194 left #mojo
15:12 sri marcus: i expect you to close that issue! :D
15:35 uniejo left #mojo
15:55 marcus sri: WHAT? NO FAILING TEST? :-O
15:55 marcus :D
16:03 perlrocks Twitter: "Mojolicious:: Plugin:: Mail and Email:: Sender and String:: Random knew. All CPAN. CPAN Perl people have found little reason to believer deified." (ja) --hoyo1111 http://twitter.com/hoyo111​1/status/56749026070102016
16:59 daviddelikat1 joined #mojo
17:03 daviddelikat left #mojo
17:09 alnewkirk cewl
17:16 alnewkirk left #mojo
17:17 megatron left #mojo
17:23 alnewkirk joined #mojo
19:17 janus mhm, if you're strict with dom parsing it should tell the user imho when things fail
19:20 janus oh well, heads-- we go :)
19:23 sri janus: how exactly do you know things fail?
19:25 sri janus: and what makes you assume Mojo::DOM is strict?
20:30 dku joined #mojo
20:30 GitHub93 joined #mojo
20:30 GitHub93 mojo: master Sebastian Riedel * ada3863 (2 files in 2 dirs): cleanup - http://bit.ly/dNVsHO
20:30 GitHub93 left #mojo
21:27 trone_ joined #mojo
21:32 trone left #mojo
22:18 GitHub177 joined #mojo
22:18 GitHub177 mojo: master Sebastian Riedel * ac46ac8 (2 files in 2 dirs): fixed small Mojo::DOM bug - http://bit.ly/iel3c2
22:18 GitHub177 left #mojo
23:11 zoo-zed joined #mojo
23:17 daviddelikat joined #mojo
23:23 daviddelikat1 left #mojo
23:42 daviddelikat1 joined #mojo
23:45 daviddelikat left #mojo

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