Think twice before running "make install" for Pugs | moritz.faui2k3.org/irclog/ | pugscode.org | sial.org/pbot/perl6 | ?eval [~] <m oo se> | We do Haskell, too | > reverse (show (scanl (*) 1 [1..] !! 4)) | "Perl 6 Today" video from YAPC::Asia: xrl.us/v6op
Set by agentzh on 6 May 2007.
00:15 jisom_ joined
Aankhen`` Man, SVN is bloody crazy. 00:43
I `svn up` a file, and it's all fine.
I make one change to it, and suddenly when I view the diff, it's doing a diff against some ancient version.
(No updates in the meantime or anything.) 00:44
Tene Well *obviously* you should use $other instead! 00:46
svnbot6 r16261 | Aankhen++ | util/smartlinks.pl:
r16261 | Aankhen++ | * added skip links for non-JS users.
Aankhen`` Tene: Hmph. 00:47
Ah, it was some weird problem with .svn being out-of-sync or something like that. 00:50
svnbot6 r16262 | Aankhen++ | util/t/01-smartlinks.t: 00:55
r16262 | Aankhen++ | * updated &fix_randomness for the updated classnames.
Aankhen`` How very unnecessarily redundant. <G> 00:58
01:28 weinig joined 01:35 ProperNoun joined
dduncan I have a style question: in practice, do people usually write their SQL keywords all in uppercase, all in lowercase, or a mixture thereof? 01:58
Most examples I see are all uppercase, but does anyone actively prefer alternatives? 01:59
oops, wrong channel 02:00
ignore that please
02:20 Khisanth joined
svnbot6 r16263 | Aankhen++ | * util/smokeserv/smokeserv-client.pl: only show 'consider installing...' messages if the modules did not load. 02:29
Aankhen`` With a precompiled Test.pm, on Debian-under-VMWare with 1,024 MB of memory and two processors (single dual-core), a full smoke took 24.68 minutes. 02:35
Seems pretty fast to me. :-D
02:47 Khisanth joined 02:57 the_dormant joined 03:02 blaze-x joined 03:04 ProperNoun joined, trippeh joined, [particle] joined, dduncan joined, ozo joined, lisppaste3 joined, lumi joined, rashakil joined, stef_ joined, polettix joined, theorbtwo joined, avar joined, f0rth joined, meppl joined, sri_ joined, frobnitz joined, baest joined, jjore-w joined, jvoorhis joined, Grrrr joined, nnunley joined, Eidolos joined, Juerd joined, scw joined, zamolxes joined, spinclad joined, yves_ joined, tokuhirom_ joined, autark joined, charsbar joined, mr_ank joined, nelhage joined, specbot6 joined, buubot joined, fridim joined, jql joined 03:12 lidden joined, bas joined, catch23_ joined, rhr joined, JarJarBinks joined, gnuvince joined, ajs_ joined, diakopter joined, audreyt joined, Kattana joined, stevan__ joined, mugwump joined, Teratogen joined, awwaiid joined, shady joined, ofer joined, rlb3_work joined, cj joined, japhb joined, ruz joined, araujo joined, jikuty joined, cognominal joined, turrepurre joined, idiotmax joined, pbuetow joined, Belaf_ joined, tewk joined, Entonian joined, Khisanth joined, blaze-x joined, trippeh joined, [particle] joined, dduncan joined, ozo joined, lisppaste3 joined, lumi joined, rashakil joined, stef_ joined, polettix joined, theorbtwo joined, avar joined, f0rth joined, meppl joined, sri_ joined, frobnitz joined, baest joined, jjore-w joined, jvoorhis joined, Grrrr joined, nnunley joined, Eidolos joined, Juerd joined, scw joined, zamolxes joined, spinclad joined, yves_ joined, tokuhirom_ joined, autark joined, charsbar joined, mr_ank joined, nelhage joined, specbot6 joined, buubot joined, fridim joined, jql joined 03:13 Southen joined, ingy joined, wolverian joined, BooK joined, gugod joined, mj41 joined, TimToady joined, bennett joined, nekokak joined, Loro_ joined, pjcj joined, statico joined, wilx joined, clkao joined, offby1 joined, evalbot_r16247 joined, drbean joined, Lunchy joined, dvorak joined, LCamel joined, Maddingue joined, shachaf joined, Tene joined, drupek12 joined, cernd joined, Daveman joined, takanori joined, mtve joined, nothingmuch joined, gaal joined, perlbot joined, Caelum joined, masak joined, rgs joined, GeJ joined, agentzh joined, SamB joined, kcwu joined, allbery_b joined, Yaakov joined, svnbot6 joined, takesako___ joined, orafu joined, buu joined, cls_bsd joined, pasteling joined, arguile joined, TreyHarris joined, QtPlatypus joined, Gothmog_ joined, kolibrie joined
agentzh "0 ōæ½xA1ōæ½xCC, 0 ōæ½xA1ōæ½xC1" looks odd to me. it seems that the .t file crashed or did not use Test.pm. 03:15
sigh. my client doesn't send UTF-8. 03:16
moritz: it's GBK :) 03:17
wow, 50.72 min smoking period on feather for the very first time! it was 130+ min or even 200+ min in the last few weeks! 03:20
feather++
Juerd++ 03:21
03:33 electrogeek joined 03:39 amnesiac joined
agentzh yay! util/t/01-smartlinks.t passes on windows! 03:45
Aankhen``: the tarball is now updated on feather. 03:47
svnbot6 r16264 | agentz++ | [util/t/01-smartlinks.t]
r16264 | agentz++ | - fixed the \r issue on Win32.
r16265 | agentz++ | 01-smartlinks.t - minor cleanup 03:53
Aankhen`` agentzh: ExcellentƩ. 03:54
agentzh :)
Aankhen`` That's weird. Did I introduce new issues?
agentzh ?
Aankhen`` In 01-smartlinks.
t
Hmm, guess not.
agentzh it's failing on your side? 03:55
Aankhen`` No, I was wondering about the cleanup.
But looking at the changes it seems fine. :-)_
s/_//
agentzh :)
Aankhen`` I'm thinking about working on smoke-server.pl
agentzh what features are you going to add? 03:56
Aankhen`` Not going to add, just fiddle with the HTML a bit.
I'm not really the feature adding type, heh.
agentzh Aankhen``++ # the HTML guy
Aankhen`` :-D 03:57
agentzh feather is still quite slow for interactive use. not suitable for coding and manual testing. 03:59
04:00 REPLeffect joined
svnbot6 r16266 | agentz++ | smartlinks.pl - added hyper-links "[Index of Synopses]" 04:23
r16266 | agentz++ | and "[Top]" to every synopsis page except S26.
r16267 | agentz++ | smartlinks.pl - added proper footer for S26.html too 04:29
r16268 | agentz++ | smartlinks.pl - added the missing #__top anchor link to S26.html. Perl6::Perldoc::To::Xhtml-- :) 04:34
05:06 BooK_ joined
svnbot6 r16269 | agentz++ | gen-smartlinks-data - excludes S26.html from the tarball 05:06
agentzh is adding line-anchors support to smartlinks.pl 05:18
05:23 offby1 joined 05:26 weinig joined 05:46 REPLeffect joined
agentzh can't reach feather's http... 06:01
and can't commit to the pugs repos. alas. 06:07
moritz: line anchor support has landed. i'll commit it as soon as feather is back. 06:08
Aankhen`` Heh, I'll finally understand what it means. :-P 06:09
agentzh line anchors?
Aankhen`` Yeah.
Do you mean linking stuff like S02:273 to the exact line?
agentzh right.
Aankhen`` Ahh. 06:10
That wasn't so hard.
I wonder why it made my head explode earlier.
agentzh we are talking about .../S04.html#_line_117
Aankhen`` Right.
Ah, would be useful to have XHTML 2's `l` element there.
agentzh what's that?
Aankhen`` You could think of it as a more useful translation of the `br` element. It's intended for marking up individual lines. 06:11
<l>Line 1</l><l>Line 2</l> # among other things, they'd have a linebreak in between.
agentzh ah 06:12
how to jump to a particular line then?
Aankhen`` Oh, you'd give all the `l` elements themselves IDs. 06:14
<l id="_line_1">...</l><l id="_line_2">...</l>
agentzh okay
Aankhen`` What are you placing the anchors on, BTW? 06:15
agentzh currently i'm using <a name="_line_1"></a> <a name="_line_2"></a>
06:15 kunwon1 joined
Aankhen`` Mmm. 06:15
agentzh it works at least :) 06:16
Aankhen`` Well, it would probably be better semantics, if you feel like doing it, to make it <span id="_line_1">...</span> and so on.
Oh, sure, getting it to work is the most important part. ;-)
I'm just very big on "refining" other people's work and not really doing much on my own. :-P
agentzh hopefully you'll refine the line anchor stuff after i commit it. 06:17
use of <span> looks tricky to implement since it wraps things.
i tried very hard to ensure those anchors didn't affect the appearance. 06:19
Aankhen`` I'll take a whack at it. 06:23
agentzh summons feather. 06:24
Aankhen`` Go agentzh! ^_^
agentzh weird. apache is down? ssh works, ping works, but http doesn't :( 06:25
Aankhen`` I was just thinking about it, and I suddenly realized that if I had to choose the one most important thing Perl 6 brings to the table, it wouldn't be any syntactic feature or programming style or anything of the sort. Rather, it would be the sense of opening up the language. 06:40
agentzh totally agree. just consider other languages like C# and Java... 06:42
Aankhen`` Absolutely. 06:43
But even more "flexible" and "dynamic" languages, like Python, Ruby or *shudders* PHPā€¦ they've got nothing on Perl 6.
I suppose Perl 6 will have nothing on Lisp. *giggles* 06:44
QtPlatypus Aankhen``: I secretly suspect that Perl6 is a lisp. 06:55
Aankhen`` LOL. 06:56
agentzh i restarted apache on feather. 07:02
Aankhen`` Kewlies.
agentzh httpd was not running.
Aankhen`` Ah.
I wonder how come?
agentzh not sure if it's intentional or not.
maybe our admin brought it down for maintainance? 07:03
Aankhen`` shrugs.
07:04 Belaf_ is now known as Belaf
agentzh well, at least i can commit now :) 07:05
svnbot6 r16270 | agentz++ | [smartlinks.pl]
r16270 | agentz++ | * added line anchor support via the --line-anchor option
r16270 | agentz++ | (default off)
r16270 | agentz++ | Examples:
r16270 | agentz++ | S04.html#_line_117
r16270 | agentz++ | S01.html#_line_1
r16270 | agentz++ | The line numbers are continuous and starts from 1.
Aankhen`` Yay!
agentzh (root access)++
Aankhen`` agentzh++ # kudos :-)
agentzh :)
Hmm, 0 ōæ½xA1ōæ½xCC, 0 ōæ½xA1ōæ½xC1 should be "N/A" 07:11
or it's rather confusing.
07:12 the_dormant joined
pasteling "rindolf" at 88.152.120.36 pasted "Problem with Parrot" (41 lines, 3.7K) at sial.org/pbot/24857 07:12
agentzh rindolf: try #parrot? 07:13
07:15 rindolf joined
svnbot6 r16271 | agentz++ | [docs/feather/script/update-syn] 07:15
r16271 | agentz++ | * added the --line-anchor option for smartlinks.pl
07:20 weinig is now known as weinig|zZz
Aankhen`` Ahhā€¦ just went over S26 in full. Pod++ # more of what I was talking about with regards to "opening up" the language 07:23
agentzh okay, synopses on feather have been updated with line anchors: perlcabal.org/syn/S04#_line_53 07:27
lambdabot Title: S04
agentzh feedback and bug reports welcome :) 07:28
Aankhen`` Neato.
I take it the IRC logs will eventually use them as well?
Er.
Yeah, since that was the point.
Aankhen`` shuts up.
agentzh @tell moritz you can now link to a particular line from the irc logs. see r16270 07:29
lambdabot Consider it noted.
agentzh Aankhen``: absolutely 07:30
the anchors' names shouldn't be lead by underscores? 07:33
it jumps twice, hmm 07:34
odd... 07:35
it works fine in IE, but firefox tends to take another jump away from the right position... 07:37
07:40 Belaf_ joined
svnbot6 r16272 | Darren_Duncan++ | ext/QDRDBMS/ : added more methods to AST.pm ; expanded SeeAlso.pod 07:40
agentzh The "Go" button of firefox works while the refresh button does not :( 07:42
dduncan regarding those line-specific anchors, I suspect it may be more maintainable if they count lines relative to (following) a particular section heading ... that way, as the Synopsis continue to be edited, the links are less likely to break horribly
at least assuming the current numbers are absolute for the file 07:43
agentzh dduncan: but TimToady tends to get line numbers from his vim.
dduncan: it's mainly for linking from "S04:123" in irc logs rather than from the test suite via smartlinks. 07:44
dduncan okay
agentzh :)
dduncan in that case, since they are just used for short periods, that should work well
agentzh *nod*
heh, i like my red in moritz's logs :) 07:51
(although not as bright as audreyt's) 07:52
svnbot6 r16273 | agentz++ | smartlinks.pl - replaces the confusing "0 passed, 0 failed" info to "N/A" 07:59
agentzh s/to/with/ 08:00
08:03 iblechbot joined 08:04 evalbot_r16247 is now known as evalbot_r16272 08:15 evalbot_r16272 is now known as evalbot_r16273
agentzh ?eval 'a' ~~ /$<b> := ($<alpha>)/; $<b><alpha>.say 08:19
evalbot_r16273 *** Cannot parse regex: $<b> := ($<alpha>)ā¤*** Error: ā¤Error: Can't modify constant item: VUndef
agentzh ?eval 'a' ~~ /$<b> := (<alpha>)/; $<b><alpha>.say
evalbot_r16273 *** Cannot parse regex: $<b> := (<alpha>)ā¤*** Error: ā¤Error: Can't modify constant item: VUndef
08:39 Belaf_ is now known as Belaf 08:42 franck__ joined 08:47 dduncan left 08:51 ludan joined 08:57 bernhard joined 09:10 elmex joined
moritz agentzh++ I'm gonna implement linking to synopsis lines now ;) 09:19
lambdabot moritz: You have 1 new message. '/msg lambdabot @messages' to read it.
moritz @moosages
lambdabot agentzh said 1h 50m 14s ago: you can now link to a particular line from the irc logs. see r16270
jrockway heh 09:20
@messages
lambdabot You don't have any new messages.
jrockway @tell jrockway that he doesn't have any messages
lambdabot You can tell yourself!
jrockway anyway :)
moritz @tell jrockway lambdabot is amazing ;) 09:21
lambdabot Consider it noted.
jrockway @tell moritz lambdabot++
lambdabot Consider it noted.
jrockway he seems to have fallen off #haskell on irc.perl.org though
lambdabot jrockway: You have 1 new message. '/msg lambdabot @messages' to read it.
jrockway @moosages
lambdabot moritz said 37s ago: lambdabot is amazing ;)
moritz @moosages
lambdabot jrockway said 27s ago: lambdabot++
perlbot What kind of idiot karmas himself? Your kind of idiot!
moritz *g* 09:22
agentzh bots are interacting with each other!
cool
jrockway it's more fun when they reply to each other
lumi It's best when you can get them looping! A Haskell/Perl quine possibly 09:23
svnbot6 r16274 | moritz++ | [irclog] added line links to synopsis 09:32
agentzh yay, moritz++
moritz moritz.faui2k3.org/irclog/out.pl?ch...13#id_l231 as an example
lambdabot Title: IRC log for #perl6, tinyurl.com/2h3hf6
agentzh S04:53 09:34
moritz damned 09:35
agentzh ;)
moritz ah, I asserted whitespaces around it 09:36
stupid idea
I should just use word boundaries
agentzh nod
see "S04:53"
supper & 09:38
svnbot6 r16275 | moritz++ | [irclog] synopsis links don't need to be surrounded by whitespace anymore
09:38 mayweed joined
moritz and btw links to line number <10 don't work, because I don't think it will happen very much 09:38
or should I allow it nevertheless? 09:39
ok, it's perl, I allow it ;)
mayweed does anyone know when Trac will be re-enable? 09:41
moritz my guess is "when feathers load is acceptable low"
mayweed: which is propably not the kind of answer you want ;)
09:42 agentzh left
mayweed moritz: feathers load will never be "low" wont it? 09:43
moritz I don't know, I don'T even have an account 09:44
but Juerd said something about moving it somewhere
svnbot6 r16276 | moritz++ | [irclog] minor fixes 09:47
10:06 ludan joined
Juerd Trac will be re-enabled when feather's performance problem is solved, indeed 10:06
A linux expert friend is going to look at it tomorrow. I hope he'll find something that can be improved 10:07
moritz I think you should forbid most of trac's pages for robots 10:09
Juerd Depends on how well or bad performance gets
I disabled trac because it easily spent a few minutes per request 10:10
And no browser would still be connected to the server at the point where it started outputting
Juerd hasn't considered for a second to profile and/or optimize trac 10:11
Python... 10:12
moritz if most web app were written in perl the world would be a better place 10:13
at least the internet, that is ;)
integral The problem is something deeper than feather just being very heavily loaded with pugs and svk processes? 10:14
jrockway svn::web? 10:36
along with rt for tickets and kt::timeline for... a timeline (if you like that)
and as for wikis, there are plenty of perl ones 10:37
kwiwi, mojomojo etc.
trac--
10:52 agentzh joined
agentzh Aankhen``: Oh...the Javascript code executed at load time changes the page layout so much that line anchors have little effect afterwards. :( 10:55
11:41 ofer joined
moritz I think the problem is that the test sections are visible by default 11:53
It should be fixed by setting them to display:none by default 11:54
agentzh moritz: no, "display:none" only makes the deviation much smaller, but still inexact. :( 11:55
i've already tried that.
moritz but less error is better than more error
agentzh disabling javascript completely "fixes" it.
moritz: i'm looking for a true fix.
moritz agentzh: fix the browsers *SCNR* 11:56
agentzh i think we should do some kind of repositioning at the end of load-time js handler. 11:57
not sure if location.hash='#_line_XX' is a solution.
it loops, it seems. 11:58
moritz whats .hash?
agentzh www.thescripts.com/forum/thread168192.html
lambdabot Title: javascript anchor help - JavaScript
12:06 audreyt joined
agentzh programming in javascript is a pain... 12:06
moritz yes, indeed 12:07
I implemented an undo/redo function for my online sudokus - I can still feel the pain ;) 12:09
and the functionality to use the arrow keys to navigate between different input boxes... 12:10
I haven't managed it to get that working in all browsers
12:10 ico joined
agentzh i know that feeling... 12:10
okay, confirmed that location.hash works in IE and firefox. 12:18
@tell Aankhen`` i just don't understand why the window load event handler runs so many times while loading. 12:19
lambdabot Consider it noted.
svnbot6 r16277 | agentz++ | [smartlinks.pl] 12:39
r16277 | agentz++ | * explicitly reset location.hash in the JavaScript code
r16277 | agentz++ | if there's an anchor in the url.
r16277 | agentz++ | now urls like S04.html#_line_53 should work precisely in
r16277 | agentz++ | both IE and firefox. reports for other flavors of
r16277 | agentz++ | browsers will be very appreciated. :)
agentzh now perlcabal.org/syn/S04.html#_line_53 should locate itself exactly at the beginning of "Except for such formal parameter" 12:50
lambdabot Title: S04
agentzh if not, tell me your the name of your browser :) 12:51
s/your the/the/
moritz agentzh: works perfectly in konqueror, but not in iceweasel 2.0
agentzh moritz: thanks :)
moritz agentzh: when I enter the URL with anchor and hit enter, the anchor is loacted wrongly... 12:54
agentzh in iceweasel only?
moritz agentzh: but when I hit enter again in the URL bar, it works
in konqueror it works without problems
agentzh gotcha. 12:55
google doesn't give me any clue regarding iceweasel's oddity...
moritz agentzh: you know that iceweasel is nearly firefox? 12:56
it's just rebranded + debian patches
agentzh moritz: yes, i'm aware of.
that
my win32 port of firefox works fine.
moritz and links from other pages work as well 12:57
so it's just an oddity that you can safely ignore ;)
agentzh i'd make some small changes to see if the problem can go away.
agentzh pushes feather forward very hard. 13:06
svnbot6 r16278 | agentz++ | [smartlinks.pl]
r16278 | agentz++ | now we use location.hash = "#_line_XX" instead of
r16278 | agentz++ | location.hash = location.hash
r16278 | agentz++ | hopefully this is why iceweasel gives up repositioning.
agentzh updating the synopses on feather...it's slow...
moritz agentzh: tell me when it's done ;) 13:07
agentzh moritz: sure!
moritz damned caching, atm I can't test it 13:10
agentzh caching? 13:11
moritz browser cache 13:12
agentzh refreshing doesn't help?
moritz no, because after refreshing it worked previously as well ;)
agentzh oh 13:13
agentzh misses the fast feather in those good old days. 13:15
although our test suite has now doubled, it shouldn't take years to process its smartlinks... 13:16
moritz perhaps we should rename it to "plumb"
agentzh lol 13:17
13:17 rindolf joined
agentzh moritz: S04 has been updated (but not others) 13:22
moritz *testing* 13:25
agentzh moritz: make sure you see the line "if ... location.hash = RegExp.lastMatch;" in the HTML source in case your browser cache is too smart. 13:26
s/smart/lazy/
moritz it does show up, and it works 13:27
yay
agentzh hooray
13:31 Limbic_Region joined
agentzh it's quite amusing to test the SXX:XXX links given by TimToady in the following log page 13:36
moritz.faui2k3.org/irclog/out.pl?ch...2007-05-10
lambdabot Title: IRC log for #perl6, tinyurl.com/2dbbj4
moritz ;) 13:37
agentzh S02:2775 is deviated by 1 line due to the =item * stuff while translating to HTML.
i think it's acceptable. 13:38
moritz yes
and it will drift anyway as lines will be inserted and deleted
agentzh yup
moritz so it should be viewed as a rough estimate rather than an exact pointer
agentzh aye
S26.html current contains no line anchors because it uses a different POD parser (i.e. Perl6::Perldoc::Parser). 13:41
*currently
13:41 edenc joined
moritz we could ask damian to include line numbers 13:42
agentzh right
and also API for updating the DOM tree 13:43
since i need to insert placeholders like _LINE_32_ into the POD DOM.
13:43 zamolxes left
moritz agentzh: can you write the email? I'm not so involved in the topic of POD parsing.. 13:45
agentzh or...write a pod6 parser myself.
it needn't be complete. the current pod5 parser in smartlinks.pl is only ~30 lines or so.
moritz: sorry, lack the tuits right now. have a big deadline for my AI exam. 13:46
moritz agentzh: don't bother, it's not a pressing issue at all ;)
agentzh k
moritz agentzh: and good luck with your exam!
agentzh thanks! 13:47
agentzh wanders off to review his AI textbook &
13:47 agentzh left
moritz ?eval sub prefix:<!> (Num $x) is tighter(&infix:<**>) { $x * 3 } !1**2 13:55
13:56 evalbot_r16273 is now known as evalbot_r16278
evalbot_r16278 3 13:56
moritz we need more tests ;)
rindolf Hi moritz 13:57
moritz hi rindolf 13:58
14:15 buetow joined 14:26 chris2 joined
moritz pokes svnbot6 14:30
svnbot6 r16279 | moritz++ | added tests for "is tighter", "is looser", "is equiv" in
r16279 | moritz++ | t/operators/precedence.t
moritz hey, it worked ;)
should these tests be in the operators/precedence.t or in blocks/sub_parameter_traits.t? 14:31
wait, the are no parameter traits 14:32
rhr moritz: in the test, you have postfix:<!>, but you use it as a prefix op 14:39
moritz rhr: damned, you're right 14:40
svnbot6 r16280 | moritz++ | t/operators/precendence.t: fixed braino s/postfix/prefix/, rhr++ 14:42
r16281 | moritz++ | perl6.vim: added subroutine traits and 'infix', 'postfix' etc. 14:45
15:22 sri_ joined 15:26 riffraff joined 15:28 iblechbot joined 15:35 DarkWolf84 joined 15:46 rff joined 15:51 [particle1 joined 15:55 drupek12 joined 15:58 the_dormant joined 16:13 penk joined 16:14 ggoebel joined 16:19 franck__ joined 16:21 [particle] joined 16:24 Casan joined 16:25 marmic joined 16:27 larsen_ joined 16:46 amnesiac joined 17:00 weinig|zZz joined 17:02 weinig|zZz is now known as weinig, the_dormant joined
diakopter summons audreyt auspisciously. 17:07
17:09 macroron joined
moritz what should we do with a wannbe-conjurer? burn him? *g* 17:12
17:22 lumi joined 17:29 REPLeffect joined 17:34 mako132_ joined 17:42 [particle] joined 17:45 Talaman72 joined 17:50 Psyche^ joined 17:53 [particle] joined 18:00 [particle1 joined 18:05 Psyche^ is now known as Patterner 18:29 [particle] joined 18:43 franck__ joined 19:02 [particle] joined 19:07 jisom joined 19:12 offby1` joined 19:27 [particle1 joined 19:32 wolverian joined
diakopter moritz: :P 19:34
19:34 wolverian joined 19:36 diakopter joined 19:37 the_dormant joined
wolverian Juerd, any progress on the load? 19:37
diakopter err, 'auspiciously' I meant 19:38
wolverian Juerd, a friend is suggesting to look at whether the io system is borking up
19:39 offby1` is now known as offby1
Juerd wolverian: No progress 19:41
wolverian: How? I've tried vmstat, but that looks normal
wolverian so you know what it should look like? hm. no idea then. 19:43
(the problem with vmstat is that you need a reference figure to compare against)
Juerd, have you looked for rootkits? :/ 19:50
Juerd wolverian: Hm.
No.
Hadn't even thought of it 19:51
But it's of course very much possible, with so many local users
My god, chkrootkit is slow )
:)
wolverian a good rootkit wouldn't even show up in top
heh
Juerd I know. I didn't think of it this time because I kind of trust these dozens of people I never met before, better than I trust my own clients.
wolverian well, it could have been a service hole too.. though I trust debian as well. 19:52
(but not blindly)
Juerd chkrootkit found nothing
wolverian hm. 19:54
Juerd, is the load high even if you kill the various bots and services? 19:55
Juerd Yes, as soon as it starts doing something substantial 19:56
Like a backup, distupgrade, heavy CGI thing like jifty
compiling, loading irssi (but when loaded, it's fine)
wolverian have you looked at top with threads enabled? 19:57
(press H)
that is a lot of pugs threads..
Juerd Doesn't reveal anything special in my view
Sure, but the pugs things aren't very active
A machine that's mostly 80% idle shouldn't have a load of 5 19:58
I/O is mostly idle too
19:58 Aankhen`` joined
wolverian hm 19:58
when I look at it, pugs jumps to the top every five seconds or so
with 100% cpu
Juerd Yes, but that's not all day 19:59
And still with 100% cpu, I don't expect such a high load
wolverian ah, okay.
true enough
Juerd I do see that AGAIN, people don't nice their non-interactive processes
wolverian hm. 20:00
20:01 evalbot_r16278 is now known as evalbot_r16281
Juerd And I've specifically asked one of these people several time 20:01
20:02 crashmatrix joined
wolverian I guess I want to see some cpu and io graphs 20:03
processes waiting for io
Juerd Do you know how to get those graphs?
wolverian gnome-system-monitor... ;) 20:04
wolverian is terribly noob at server administration
I wonder if dstat would help 20:14
Juerd gnome-system-monitor doesn't say which processes are waiting for io 20:19
wolverian that's true. sorry. 20:20
I was just thinking of graphs there
Juerd Do you know what dstat's "csw" means? 20:23
20:24 justatheory joined 20:25 bonesss joined
wolverian Juerd, no. 20:25
20:25 justatheory joined
wolverian Juerd, context switches, proabbly. 20:28
probably
Juerd Ah 20:29
Is ~450 okay? :P
Juerd has no idea how to read these values
wolverian my desktop has 1452
wolverian neither
Juerd I like dstat though 20:30
It's a nicer tool than vmstat and iostat 20:31
wolverian yes
it doesn't show anything suspicious here though
I wish this was solaris
Juerd I have ~850 csw on my laptop
wolverian it has proper io accounting
Juerd That by itself isn't a good reason to run solaris 20:32
I don't think solaris would make a great platform for feather :)
wolverian right, but I wish linux had its iostat :)
I can't think of anything else 20:33
either it's a rootkit or it's processes waiting for io 20:34
since the cpu is idle
Juerd Nothing is actually doing any significant io though
wolverian I want to ask "how much time is spent waiting for the spindles to move"
unsurprisingly I can't ask that in linux
that's a bit pessimistic 20:35
the io stats just suck.
hm, the uptime is now down to 0.2 20:36
er.. load
aaand now it's back to over 1. oh well. 20:37
a load of pugses lifted it again :)
(a plural of a plural there)
and I guess I should have said s,load,pack,
PerlJam Why the microscopic attention to the load average? 20:40
wolverian because it's so high even when the machine is supposedly idle
I guess one could use sa(8) to look at total accounting information
PerlJam Even a mostly idle machine can have periodic fluxes in load average. 20:41
load average is sort of like blood pressure. It's only a problem if it's persistently high. 20:42
wolverian I understand that's exactly why we're looking into it. because it's persistently high.
unfortunately I'm not very good at it, so please help if you know how :)
PerlJam clearly not if it drops and then goes back up.
up and down is normal. up and up and up is bad. 20:43
wolverian well, ask Juerd. maybe he did something that fixed it now. :)
Juerd PerlJam: Because it's constantly much higher than I'd expect, and many people complain about performance.
PerlJam: In fact, it's now under 1, for the first time in a week I guess 20:44
wolverian yeah.
Juerd I've slayed agentzh, who had a pugs compile running unniced, while not being logged in.
But I can't imagine that was the problem, a single compile
And load's at 4.77 again now.
PerlJam Hmm.
wolverian all I see in top are andara's and audrey's pugs 20:45
I don't know how to read the thread information, but a screenful of them all at 99% looks scary :)
PerlJam top is a very blunt tool for diagnosing load problems.
(unfortunately I don't know of a finer tool right off)
Juerd Yes.
PerlJam I wonder what pugs does while idling? 20:46
Perhaps there are some Haskell-related GC going on periodically.
Juerd PerlJam: Whoa, numerous ALRMs apparently 20:47
wolverian ow.
PerlJam strange
Juerd Yep, constant ALRM and INTs
For every idle pugs process
igreturn() = ? (mask now []) 20:48
futex(0x8946660, FUTEX_WAIT, 33503, NULL) = -1 EINTR (Interrupted system call)
--- SIGALRM (Alarm clock) @ 0 (0) ---
repeated
I don't know which one comes first.
PerlJam the trick is figuring out if those are endemic to long-running ghc processes or just pugs 20:49
Juerd What is endemic? 20:50
Hm, things are fast now
wolverian e.g. an endemic disease; related to a specific people or locale
Juerd So it may indeed have been agentzh's processes
wolverian or process ;)
PerlJam um ... endemic is like characteristic
wolverian how about andara's and audrey's? 20:51
do they behave similarly?
Juerd Yes
20:51 lichtkind joined
Juerd Though they do different things while the signals come in 20:51
andara's pugses do ioctl()s, audrey's do futex()s 20:52
But both get a constant flood of ALRMs
That's a lot of signals.
wolverian ghci does the same thing 20:53
Juerd I don't know if it's bad
PerlJam it may not be bad per se, but it might be bad for feather
Juerd At some level I think an OS should just deal with it.
wolverian well, perl -d0 doesn't do it ;) 20:54
Juerd That waits for user input 20:55
wolverian so does ghci
Juerd So it probably select()s
PerlJam hmm. ghc may have a poor IO model
lichtkind moritz: hi grad wach? 20:56
moritz lichtkind: ja, aber kaum zeit
Juerd Deutsch ist schwer :) 20:57
lichtkind schade wollt aber eh nur kurz was loswerden
wegen dem tutorial kam grad nach hause
moritz: hast schon angefangen? 20:58
moritz lichtkind: womit?
Juerd pugs processes get approx 18 ALRM signals per second
lichtkind moritz: dein tutorial 20:59
moritz lichtkind: ja klar, ist schon auf meiner seite
Juerd Heh, would this still be the old 18.2 thing? 21:00
wolverian 18.2?
Juerd Yeah, something about 18.2 ticks in a cpu second
When I still coded in BASIC, this was a big thing, because it meant the finest granularity in sleeps was 55 ms
?eval 1/18.2
evalbot_r16281 5/91 21:01
wolverian yes, that's the basic timer granurality
Juerd Sigh
wolverian ..granularity
Juerd It can't be a coincidence that these pugs processes get approx 18 sigalrms per second, right? :)
That's as fast as it gets probably
PerlJam It's probably trying to generate more than 18/sec but that's all that can be processed.
right. Not a coincidence
Juerd That sounds like a very, very tight event loop...
wolverian event loops are supposed to wait for io... 21:02
surely..? :)
PerlJam Tell audrey and I'm sure she'll know exactly what to change to make it better ;-)
Juerd audreyt: ping 21:04
I wonder if she'll pong
PerlJam wanders off to prepare for a 3 hour journey
lichtkind moritz wollt halt eigentlich nur sagen das ich mein tutorial leicht umarbeiten will das es nicht unbedingt perl5 vorausgestzt wird aber es wird halt kein kurs programmieren lernen mit perl6 sondern schon eher ein komprimiertes lernen
audreyt it's the GHC microkernel
it has its own scheduler
which is preemptive
the ALRM signals are entirely normal
Juerd audreyt: Could they be the source of feather's bad performance? 21:05
I don't know what signals do to a system 21:06
mugwump ?eval 4.772727 / 65536 / 4
evalbot_r16281 4772727/262144000000
wolverian haha.
lichtkind audreyt++ for her last talk
Juerd But I can imagine that a dozen pugses all doing 18 signals per second, might have some effect
mugwump ?eval 4772727 / 65536 / 4
evalbot_r16281 4772727/262144
mugwump ?eval (4772727 / 65536 / 4)+0.0
evalbot_r16281 4772727/262144 21:07
audreyt Juerd: all GHC applications do that... I think it's fine, in particlar on linux. in Israel we suspected the same thing but avar(?)'s debugging showed it's false alarm
Juerd Okay
Nice pun btw :)
audreyt entirely unintentional 21:08
Juerd Still nice :)
And it's nice to see you're still here; I'm not active much, but see you much less than before when I am. How are you? 21:09
audreyt been hacking away at the local branch before talking about it :)
Juerd And how are you personally? :)
audreyt before today it was 8543 test cases 6500ish ok 21:10
5 minutes ago it was 11436 test cases: 10006 ok, 1430 failed, 1024 todo, 770 skipped and 23 unexpectedly succeeded
moritz lichtkind: ich hab halt drei verschiedene Tutorials, unterschiedliche Zilgruppen haben...
audreyt still, the smoke clusters has it at 18603 test cases: 18085 ok
so still some ways to go before I can commitback
offby1 ave
audreyt (it's the pads+MO branch)
wolverian shiny.
Juerd Did you hack all day between before-today and 5 minutes ago? :)
moritz lichtkind: und ehrlich gesagt mir gefƤllt das layout des wikis Ć¼berhaupt nicht, deshalb habe ich keine groƟe Lust, darin mitzuhelfen... ein paar fixes vielleicht, aber kein groƟen neuen content ;) 21:11
audreyt Juerd: something close to it yes.
lichtkind moritz: seh ich grad was ich nicht unbedingt ideal halte aber vielleicht kann ich dir trotzdmem bei der seite helfen
Juerd audreyt++
audreyt and the night before, actually
moritz lichtkind: hilfe ist immer wilkommen... wenn du willst geb ich dir zugang zum svn
lichtkind moritz: wir kƶnnen sie ja auch bei dir reinstellen
audreyt it's good that smoke tests doesn't lie ;) because it forced several rewrites ;)
Juerd lichtkind, moritz: Could you please use a query or speak English here? 21:12
moritz Juerd: sorry, yes
lichtkind: are you in #perlde on irc.perl.org?
Juerd stupidly wrote that sentence in German first...
lichtkind moritz: no
Moritz i first finish second part of my perl6 article for foo magazine 21:13
moritz lichtkind: I'd like to discuss it at some other time anyway, perhaps I'll send you a mail, or ping you sometimes next week
Juerd himself is bad at context switching, specifically between languages
moritz ok, so no need for hurry
lichtkind k
Juerd I killed 10 or so idle pugses 21:17
And although I like to believe that the sigalrms don't do much, it did make performance better 21:18
audreyt pugs's own event loop may be cpu consuming a lot. 21:19
if it's stuck in some while loop or so
Aankhen`` Okay, first draft of Perl6::Perldoc::To::Html is done. :-D Now to begin the debugging cycle.
lambdabot Aankhen``: You have 1 new message. '/msg lambdabot @messages' to read it.
Aankhen`` @messages 21:20
lambdabot agentzh said 9h 5s ago: i just don't understand why the window load event handler runs so many times while loading.
Juerd audreyt: it's not cpu or io, at least not measurable by the usual tools (top, iostat)
Aankhen`` @tell agentzh That's strange, the `load` handler should only be running once.
lambdabot Consider it noted.
wolverian Aankhen``, what's wrong with To::Xhtml? :)
audreyt nod. ok then.
Aankhen`` wolverian: It's XHTML. :-) 21:21
Juerd audreyt: I've re-enabled trac
audreyt k
wolverian Aankhen``, ...that's bad? ;)
Aankhen`` wolverian: In this context, it's pointless.
Juerd audreyt: Let's see how this goes. Currently it returns pages within a few seconds. That's better than a few minutes, at which it was before
wolverian Aankhen``, hm. why?
Aankhen`` wolverian: XHTML = 0 benefits, larger code size, and treated as tag soup by the browser since it's being sent as text/html.
wolverian Aankhen``, oh. holy war, then. :) 21:22
Aankhen`` LOL.
Juerd XHTML is useful only if you're combining it with other XML stuff.
Like SVG
Or XSL
wolverian or mathml.
Juerd Right
Aankhen`` Well, ::To::Xhtml is fine, but to satisfy myself I'm writing ::To::Html.
Aye. 21:23
Juerd But it's utterly useless just as an html replacement.
Aankhen`` Yup.
Juerd In fact, it's downright annoying to me.
Aankhen`` Same here.
Hence ::To::Html
lichtkind moritz: maybe i start with some blog posts there
Aankhen`` Okay, shower time.
Aankhen`` wanders off.
wolverian I have high hopes for xhtml2, which is remarkably sane, but it goes the way of xhtml1 because IE won't implement it. oh well.
Juerd I write my html by hand, and don't like to use a gazillion quotes where they were previously unneeded
Quotes around integers feel wrong anyway.
wolverian: Let's first have hope for xml2 21:24
Because xml is the greatest downside of xhtml.
avar xhtml is not pointless, you can parse it with xml parsers which are a dime a dozen, instead of a much more complex sgml parser
Juerd Many of its issues are irrepairable because it depends on xml1
avar: I implied that
wolverian what's wrong with xml1? 21:25
Juerd avar: But the sgml parser already exists, and there's no noticeable difference in speed for the user.
wolverian: Syntax.
wolverian: Mostly, to me personally, quotes and closing tags.
wolverian Juerd, yeah. vim's xhtml mode helps _a lot_. :)
and running the file through gnu tidy ;)
(xhtml mode ne html mode) 21:26
(ne xml mode)
Juerd wolverian: How does one enable that?
wolverian :set filetype=xhtml
Juerd What does it do except colour?
Khisanth and what does it do?
wolverian e.g. <foo>> inserts <foo>\nindentation\n</foo>
I think it also adds omnicompletion, but not sure. 21:27
Juerd Doesn't for me.
wolverian maybe I got it from vim.org
Khisanth doesn't do it for me either
Juerd I just write html and when I need it to be xhtml, i postprocess it.
wolverian yeah. it's not a core script.
Juerd width="123" sucks. 21:28
width=123 feels much better both to mind and hands.
wolverian www.vim.org/scripts/script.php?script_id=1397
lambdabot Title: xml.vim - helps editing xml (and [x]html, sgml, xslt) files : vim online
21:33 bonesss joined
lichtkind tscha 21:34
Juerd wolverian: <img></img> isn't very useful though 21:35
And I still need to quote integers myself 21:36
wolverian it shouldn't do that if you type <img />, surely. and yeah. that's more of gnu tidy's job. 21:38
I don't need to write attributes with numbers much
Juerd Why should *I* have to type that /?
Wasn't the cool thing about DTDs that they could indicate EMPTY? 21:39
XML is only user friendly with a full blown schema-and-dtd-supporting xml editor, which vim is not
moritz xml is mainly intended to be written and parsed by computers, not humans 21:41
Juerd It's utterly verbose for a computer language. 21:42
So I don't believe this.
And, I want to type my own html 21:43
Because all software sucks at doing it.
moritz same for me
wolverian well, I agree. 21:45
I don't write much (x)html, which is why I don't feel the pain so much, I suppose.
21:46 Talaman72 joined
moritz well, nearly all of my html is xhtml 1.0 strict, and it's just a matter of getting used to it 21:47
21:48 Talaman72 left 22:05 rashakil_ joined
svnbot6 r16282 | moritz++ | [irclog]: fixed revision links at start of lines 22:18
22:18 fridim joined 22:33 fed joined 22:57 DarkWolf84 joined 23:24 SubStack joined

Powered by Cro and the RakuĀ® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!