The Return of the Journal : pugs.blogs.com/ | pugscode.org | pugs.kwiki.org | paste: paste.lisp.org/new/perl6 or sial.org/pbot/perl6
Set by GammaRay on 31 December 2005.
00:01 stevan joined
putter xinming: linux/fedora 00:03
Juerd putter: Do you like Fedora? 00:07
00:13 stevan joined 00:14 Khisanth joined
putter Juerd: mixed feelings. though with dual caveats of 1- I've been using it / redhat for a long time, so it becomes unclear whether remembered failings reflect current state, and 2- I'm running a rev+ back from edge, so no doubt things have improved. 00:21
Juerd putter: Have you ever tried Debian Sid or (k)Ubuntu? 00:27
afk # z 00:28
putter if I was currently spread across multiple boxes, I'd probably do debian stable on one for services, and debian "I dont care if it breaks every month" on another, for development. fedora manages to be a somewhat cruft compromise between the two. gentoo isnt an option because it makes "state of the tool chain" too unceratain.
I've done ubuntu for family, but not myself. mepis also. no experience with sid. 00:29
Juerd Debian Sid doesn't break every month. It usually breaks notably once a year, and I think it's very much worth the little struggle.
putter hmm. what do you think of as the biggest downside of sid? 00:30
wolverian the ubuntu server distribution isn't bad either.
Juerd putter: That you MUST keep up with it, for a box that hasn't seen any updates in 2 years is hard to get back on track.
mugwump Mark Shuttleworth gave some great arguments why you shouldn't care which debian variant you use at Linux.conf.aui 00:31
Juerd mugwump: Well, he forgot some practical points.
mugwump or, you shouldn't care that you're using a variant, anyway
Juerd But in general, he's very right.
wolverian: I've yet to use Ubuntu for a server. I find it a bit scary, because I'm not used to it and don't know its quality. 00:32
mugwump ubuntu is actually capable of libc upgrades within a week or two rather than the 6 months+ time it takes Debian
or gcc upgrades, etc, where lots of packages are affected
Juerd wolverian: Also I don't really like the idea of releases with a server, but that's because it makes me think of Redhat, so it's a bad argument.
mugwump: Hardly an issue for users. 00:33
I'm going to bed now
Goodnight
putter remembers irrecoverably smashing a mepis by doing apt-get update, which at the time turned out to be documented as a "we have non-standard packages - you _will_ die" no-no. perhaps choice of variant matters.
mugwump waves
putter good night Juerd & 00:34
putter reconsiders upgrading fedora this week...
mugwump: re libc, that sounds like a receipt for never being certain what's causing your language runtimes to be flakey. ;) 00:36
mugwump what I mean is, say you want to upgrade the libc in unstable for the next release. Canonical can get 3 devs to spend 1 week to do the whole thing. Debian gets every package maintainer to do it individually. 00:37
putter ahh.
mugwump Say it's a new libc or gcc, the problems faced will be similar for all packages being recompiled, for instance extra compile warnings etc. 00:38
svnbot6 r8839 | audreyt++ | * Change the message text in Makefile.PL for people with GCC 4.0 01:24
r8839 | audreyt++ | on Tiger and make it clear that an upgrade to GHC 6.4.1 is preferred.
wolverian mm.. foo($bar) ā‡’ grep ā†’ $x { baz $x - 2 } ā‡’ say; # sexy 01:25
audreyt surely you mean ==> ? 01:26
wolverian it's a unicode wide character that I aliased to ==>
audreyt ahh.
clever
xinming audreyt: pugs will encounter the seg fault. :-/ 01:27
07:59 ilogger2 joined 08:10 nothingmuch_ joined
svnbot6 r8840 | audreyt++ | * It turns out that autoflush of stdin/stdout is sufficient to 08:10
r8840 | audreyt++ | avoid infinite loop in readline's signal handling. Therefore
r8840 | audreyt++ | we revert the sigwinch/signal disabling as it causes xinming's
r8840 | audreyt++ | debian system segfault.
nothingmuch_ weeeeeeeeeeeeeird
we have a machine
we have two users 08:11
one user runs a process fast
the other does not
(load goes up to 0.03)
lsof lists nothing odd in terms of data files that could be slow
benchmarking reads/writes is also weird
i think i need to strace it
08:16 geoffb joined 08:20 BakedB joined 08:46 nothingmuch_ is now known as nothingmuch 09:00 drbean_ joined 09:09 kane_ joined
svnbot6 r8841 | audreyt++ | * remaining of TODOization, forgot to commit yesterday. 09:46
09:58 dada joined 10:40 r0nny joined 10:50 drbean joined 10:54 G2 joined 11:23 pdcawley joined 11:25 nothingmuch joined
svnbot6 r8842 | audreyt++ | * Fix the dreaded expression inferencing bug that made 12:11
r8842 | audreyt++ | $.foo.bar()
r8842 | audreyt++ | fail to dispatch on the runtime type of $.foo.
r8842 | audreyt++ | * Consequently, DFA.t now passes all its tests.
12:13 feng123 joined 12:39 rantanplan_ joined
svnbot6 r8843 | audreyt++ | * Tree.pm: repair the use of private methods. 12:45
12:49 r011 joined
svnbot6 r8844 | audreyt++ | * HTTP::Headers::Util: Reword the destructive and not-quite-working 12:54
r8844 | audreyt++ | algorithm with a cleaner gather/take.
13:00 r011 left 13:03 chris2 joined 13:12 iblechbot joined
svnbot6 r8845 | audreyt++ | * Because named arguments cannot be bound positionally anymore, 13:15
r8845 | audreyt++ | Test.pm's 1-basic.t no longer treats
r8845 | audreyt++ | is(1, 2, 'feature')
r8845 | audreyt++ | as a synonym of
r8845 | audreyt++ | is(1, 2, :todo<feature>)
r8846 | audreyt++ | * more test triaging: removal of flunk() and TODOization. 13:21
13:33 kolibrie joined 13:37 orafu joined
audreyt some help on triaging the remaining module test failures (Perl6::Value::List, Span, Set::Infinite) would be appreciated 14:09
svnbot6 r8847 | audreyt++ | * Set::Infinite - some workarounds for the chained method call bug;
r8847 | audreyt++ | thinking about how to solve it properly...
audreyt for people with embedded parrot segfaulting before, please re-test with r8848. 14:11
svnbot6 r8848 | audreyt++ | * disable runcore selection for parrot as it causes too many
r8848 | audreyt++ | unidentifiable segfaults.
Juerd I'll upgrade feather tonight 14:22
It's quite expensive memory, but I have it around and no direct purpose for it, so perhaps it'll go down again later to have it replaced with cheaper memory :) 14:23
audreyt sure :)
theorbtwo How much RAM is feather going to have? 14:24
Juerd theorbtwo: It has 2*512 now, it'll have 2*1024 later.
theorbtwo: And that't the max for this board
At least it'll accommodate two more builds in memory :P
Without swapping
Do we know why the compilation process is *so* memory hungry, by the way? 14:25
theorbtwo Wow.
Juerd theorbtwo: Wow re what?
integral it's the big files like Pugs.Run afaik. 14:26
I believe it's because the prelude is a big haskell structure, so it's being partially evaluated etc by the compiler 14:27
theorbtwo 2GB is a lot of ram to me.
audreyt integral: yeah, it's controlled by precompile_prelude: true in config.yml
Juerd integral: Is GHC inefficient, or because it's used wrong?
audreyt integral: at this moment it's possible already to turn the parse tree into YAML and load back
ayrnieu theorbtwo - I sympathize -- I'm still blown away by my quick succession of 1G RAM machines :-)
Juerd i.e. what do we blame? :)
audreyt Juerd: me, as usual
integral audreyt: maybe we should get that done then :)
10000 line haskell files are never good ideas ;) 14:28
Juerd audreyt: That won't do. You don't compile manually, do you? :)
audreyt Juerd: I do
integral: indeed
integral hmm, so this would take a new script to kick off the precompiling, and a new template to stuff the YAML into
theorbtwo integral: Planning on fixing after pending release, I think. Possibly better to release with precompiled prelude off? 14:29
integral theorbtwo: not my call, I don't do releases :)
Juerd audreyt: Hmmm... Do you have that much RAM? :) 14:30
audreyt: Amazing :)
audreyt 768mb :)
Juerd That explains a lot.
audreyt theorbtwo: with precomp off, "make test" takes longer 14:33
14:33 Qiang joined
audreyt and test time > compilation time 14:33
theorbtwo Only if you do a full test run every time you compile. 14:34
Juerd audreyt: That's certainly true 14:36
audreyt but incremental compile, as long as you don't touch Pugs.AST etc, won't recomp Run.hs 14:37
Juerd theorbtwo: What other tests are there?
audreyt: But automated compiles have to make clean first, because once every few months, it's needed.
integral at one point didn't we have optimizations turned off on Run.hs specifically? 14:38
Juerd I'd love to have something like ccache, for ghc
audreyt integral: turns out with ghc 6.4.1, that takes _more_ memory
Juerd: ghcuses ccache
integral audreyt: eww :)
Juerd audreyt: Then I want even more layers :)
integral presumably we do a mv-if-diff type thing?
audreyt integral: hm? 14:39
ah, good idea
integral ie generate the prelude to a tmp file, and only replace it if it's different
audreyt *nod* want to take a stab? :) 14:42
(it's getting late here so my typing gets slow) 14:43
stevan hello all :)
integral if someone gives me a hint how to do cmp(1) in a way that'll work on windows
audreyt stevan: hey 14:44
14:45 rantanplan_ joined
audreyt integral: perldoc File::Compare 14:45
integral thanks!
audreyt :)
integral oh, I don't think this is going to work because we have to build with the null prelude first 14:49
audreyt mm
theorbtwo So?
integral well there's no point doing the mv-if-diff thing to the real precomp-prelude since we don't have an object for it anymore
theorbtwo D'oh. 14:50
Yes, very good point.
integral and we can't just save the old Run.o since we don't know if there's another reason apart from prelude changes that it would need recompiled
theorbtwo Right. 14:51
integral hmm, so which YAML target do I use? Parse-YAML? 14:52
audreyt yup 14:55
nothingmuch btw, audreyt: is there any plan to make e.g. 'use YAML qw/-Backend=Syck/' and the same for JSON? 14:58
15:00 vel joined, putter joined
putter sigh. 15:00
15:00 vytautas joined
audreyt putter: ? 15:01
nothingmuch: yes there is plan. no there is no code. yes you can commit :)
nothingmuch might be tempted to steal the XML::SAX code
so that you can set up a per-site preference for YAML backend
but code with specific requirements can explicitly create e.g. a YAML::Syck dumper handle 15:02
www.quietrevolution.co.uk/
putter audreyt: re reverting readline: xinming's segfault is not a readline regression. he tried using gcc 4 with ghc 6.4.0, a combination known not to work. this was explained to him. it's in INSTALL. it's not a regression connected with the readline change. when he raised it again with you, I considered flagging it, but thought that might be impolite. sigh. 15:04
nothingmuch ponders Module::Backend::Provider and Module::Backend::Selector 15:07
audreyt putter: nod, though that did prompt me to flag off sigwinch/sighandler and replace it with autoflushing stdio
nothingmuch and then refactoring YAML, JSON, XML::SAX etc to use these
audreyt putter: and so far there's no mysterious segfaulting/backspace problems with threaded builds here
putter: so I think it may make more sense this way, as dduncan can keep using the NetBSD readline 15:08
putter: but thanks for the headup
15:14 BeoTW joined
putter k 15:15
could you give a one-line summary of current state?
putter svn up's
svnbot6 r8849 | audreyt++ | * unTODO unexpected successes and fix syntax mishaps. 15:17
integral is there a single document describing how the build system fits together? Like the responsibilities split between Makefile.PL, Setup.hs and the bodged perl scripts
audreyt one-line summary: "I'm about to finish all non-module test triaging (yesterday I missed those that parsefailed without failing any subtests), and would appreciate help on triaging the remaining module failures" 15:19
svnbot6 r8850 | audreyt++ | * fix two more parsefails: "use Perl-6" now works, and BEGIN-time binding
r8850 | audreyt++ | a block to &infix:<operator> can manufacture an infix operator.
r8850 | audreyt++ | (previously it required a "sub{}".)
putter tnx :) 15:20
audreyt integral: I don't think there's one, no. simply put, Makefile.PL compiles a Setup.hs, and setup util/build_pugs.pl flags so it can call to ./Setup
it does so by writing a Pugs.cabal file; Setup simply responds to it
integral ah, okay. It's not exactly what I was expecting but it's close enough :)
audreyt that builds libHSPugs; afterwards it builds ./pugs.exe
putter integral: capture that paragraph somewhere? top of Makefile.PL? 15:21
integral sounds a good idea! 15:22
putter :) you or me? 15:24
integral can you? I've got messy checkouts everywhere :)
putter doing...
"Simply put, Makefile.PL compiles a Setup.hs, and setup util/build_pugs.pl flags so it can call to ./Setup. It does so by writing a Pugs.cabal file; Setup simply responds to it. That builds libHSPugs; afterwards it builds ./pugs.exe." I didn't want to edit, for fear of altering meaning. 15:29
15:31 arguile joined
audreyt er, that means my meaning wasn't clear 15:31
svnbot6 r8851 | audreyt++ | * parse (but not handle) the "handles" form for subroutines and methods.
r8851 | audreyt++ | * fix syntax errors in delegation.t and lists.t.
audreyt 1. Makefile.PL calls GHC to compile Setup.hs to ./Setup.exe
2. Makefile.PL writes a Makefile that would call util/build_pugs.pl upon "make"
3. build_pugs.pl writes a Pugs.cabal file based on the flags given in the Makefile.
4. build_pugs.pl calls ./Setup.exe.
5. Setup.exe builds dist/build/libHSPugs-6.XXX.YYY.a and returns to build_pugs.pl 15:32
6. build_pugs.pl calls GHC to build ./pugs.exe.
(by simply linking in the freshly built libHSPugs package.)
(with src/Main.hs.) 15:33
putter great. :) thanks.
audreyt :) 15:34
15:36 stevan_ joined
putter "by simply linking the freshly built libHSPugs package with src/Main.hs", yes? 15:37
audreyt yes.
15:37 Eimi joined 15:38 hexmode joined
putter in 15:41
svnbot6 r8852 | putter++ | Makefile.PL: added a "Build System Overview" pod. Makefile.PL is otherwise unchanged.
putter audreyt++ integral++ 15:42
15:43 rodi joined
audreyt woot 15:44
putter++
theorbtwo putter++ 15:45
integral audreyt: hmm, are you sure Parse-YAML is ready for the precompiled prelude? It's got output of envGlob commented out at the moment
audreyt integral: that... needs to change 15:46
integral sees what blows up when he uncomments it
putter actually, one question I've had about the build system is at what point the PUGS_EMBED and PARROT_PATH variables no longer need to be set properly. Just for Makefile.PL? for make too? 15:47
audreyt putter: I think they both are meant for Makefile.PL 15:49
putter tnx 15:51
putter has a fuzzy impression of a multi-day shakedown after each change to the precompiled prelude build system... 15:53
are we nominally targetting a Feb 1 release?
integral is putting in all the yaml code as additional to, rather than replacing 15:54
putter ah :)
integral I'll throw in a config.yml flag to control build_pugs, at the moment you just pass --yaml to gen_prelude.pl 15:55
audreyt putter: we are 15:56
svnbot6 r8853 | audreyt++ | * infix macros no longer parsefails
audreyt which means preflight tomorrow once I wake up
integral what's the difference between "askGlobal" and "asks envGlobal" in the Eval monad? 15:58
svnbot6 r8854 | audreyt++ | * stdrules.t: we don't support charnames.pm, so this test 15:59
r8854 | audreyt++ | was testing something else entirely. Fixed it by replacing
r8854 | audreyt++ | \c[NAMED UNICODE CHARACTER] with \x[HHHH].
15:59 justatheory joined
audreyt ok. I think as of r8855, really all the main subtest failures are triaged. 15:59
but I might have missed some. another round of smoke would be helpful tomorrow
putter one capability which may or may not be of interest in future, is ghc being able to dump a post-optimization ast to a file, and then take it on the command line just like a .hs file. but I have no idea how robust it is, file size, resource impact on compilation, etc. just a random observation.
audreyt integral: envGlobal points to a TVar. 16:00
svnbot6 r8855 | audreyt++ | * oops; add a missing right parens. also unTODO two more subtests.
audreyt integral: so asks envGlobal gives a TVar, but askGlobal reads it for you
putter: it's part of GHC 6.6 as libGHC, I think
integral ah!
audreyt putter: libGHC lets you manipulate all the ghci stuff, including bytecode loading and compilation etc -- much richer than hs-plugins as you have control over the entire compiler 16:01
putter: there's also a lot of 6.6 features that I'm itching to use (or was hacked in specifically for pugs)
putter: but those would have to wait :)
ok, I think I'm too exhausted to journal tonight. 16:03
audreyt presses the "shift" key
g'nite, see y'all tomorrow :) 16:04
&
putter g'night &
hmm A05 and E05 both have \c[NAME NAME NAME], but not S05?
integral night! 16:05
putter audreyt: re \c[NAMED UNICODE CHARACTER], A05 and E05 both have such, but apparently not S05. But that's why they are there. 16:07
audreyt putter: pugs don't code to [AE]05 though. 16:08
16:08 simon2 joined 16:09 simon2 left
theorbtwo Mentioned fairly often by $Larry on various MLs... 16:09
audreyt but then, E07 also mentioned \c[2026] 16:10
which is spelled \d[2026] nowadays
in any case, it should below to S02 16:11
after "Characters indexed"
s/below/belong/
a patch to p6l may make sense
putter audreyt: yeah. I believe it came over from damian's perl6::rules t/. 16:13
audreyt ok. 16:15
I've p6l'ed it.
putter sorry for keeping you awake ;) 16:17
audreyt 'sall good :) 16:18
16:23 elmex joined
putter starts 8855 smoke 16:33
16:40 cdpruden joined
svnbot6 r8856 | audreyt++ | * fix infinite loop in 20_win32_test.t (and File::Spec::Win32) 16:40
r8856 | audreyt++ | so make smoke can continue.
audreyt putter: you want 8856 smoke :) 16:42
svnbot6 r8857 | audreyt++ | * reword Algorithm::TokenBucket using postfix if.
16:43 estel joined
svnbot6 r8858 | bsmith++ | Fixed the ghci makefile target by just making it depend on the pugs target. 17:03
r8858 | bsmith++ | I thought this would be more reliable than duplicated a build_pugs.pl line yet again.
ingy audreyt: ping 17:08
audreyt++
Module::Install rocks
audreyt: I autobundled Test-Base into YAML 17:09
now YAML has no deps 17:10
but the thing was I needed 6 or 7 lines of Makefile.PL
what I really wanted was":
use_test_base;
audreyt so define it in Module::Install::TestBase 17:11
ingy I realized that I can distribute Module::Install::TestBase in Test-Base to do just that!!
audreyt and use it locally
yeah
putter ingy: audreyt is (hopefully) catching Zzz's (oh, never mind)
ingy and now everyone can do it 17:12
audreyt that's the whole point of Module::Install over CPAN::MakeMaker :)
rafl ingy: Couldn't you do that before I packaged the whole dependency chain for Debian? :-)
putter rafl's sisyphean labors continue... :) 17:15
17:16 nothingmuch joined 17:19 Shachaf joined
putter re the macro thread currently on p6l, it would seem an ast could present a "I am lisp tree of lists" facade, no? Which would both let us use the "n decades of effort" CL and scheme macro tech, and reduce the importance of having an unchanging ast. an experiment for piln perhaps. 17:27
integral that sounds a good idea! 17:28
svnbot6 r8859 | bsmith++ | * "fixed" make haddock by adding #ifndef HADDOCKs for the preludepc include in Pugs.Run, and around some uses of fancy features in Text.Parser.Kwid. 17:29
putter And re default vs trait, why not just use type? macro foo (Sexp $ast) {...} and coerce:as does the right thing.
integral: :) 17:30
audreyt (please abort smoke and restart at r8860 -- sorry for the misfix.) 17:34
svnbot6 r8860 | audreyt++ | * redo.t was looping infinitely due to a misplaced
r8860 | audreyt++ | precidence parsing for infix primitives, placing "=="
r8860 | audreyt++ | as lower than "and". Fixed.
audreyt (was introduced in r8856)
rafl audreyt: Any reason why PAR 0.91 isn't released yet?
audreyt rafl: the reason is that Pugs 6.2.11 isn't released yet 17:35
(either that, or tsee hadn't mastered the art of building parballs on win32 yet, or maybe he has other reasons) 17:36
tsee is now the releng person for PAR, and I wouldn't be able to look at it before Pugs 6.2.11, so maybe ask tsee?
(he's in irc.perl.org #perl)
(ditto for Module::Install and Alias) 17:37
putter: that listying ast idea is lovely 17:38
putter: we already pretend a string facade on a ast, so why not sexp
rafl audreyt: Will do so. Thanks! 17:41
audreyt np :)
putter well, re don't really do a string facade, but rather convert and then reparse. we could certainly play the a similar game with sexp instead of string. but the ast is rich, and sexp's are notably poor at containing "extra" bits. so the challenge is to have a sexp which isnt just as complex as the ast. I was wonder if 17:47
the ast could merely pretend to be a simple sexp, eg, with apparently untyped argument symbols, but preserve the information as the macro moves things around. new macro code would of course need the parse treatment. I'm not sure it buys you enough to be worth it. it would mean the macro could examine type info selectively (ie, it looks like a (define (f x) ...) but you can also match x with say (typedvar name type)). 17:51
17:54 nothingmuch joined
putter an alternative would of course be to have various flavors of sexp. bare-scheme-like-sexp, sexp-with-type-info, etc. macros which dont care about examining type info could use the simple version, macros which do would just have to deal with it everywhere. 17:55
audreyt that sounds like... objects. 17:56
;)
putter lol
yeah. but there's no reason something which looks like a mere symbol couldn't have... hidden talents. ;) 17:57
s/but//
audreyt chuckles, waves, and sleeps (this time for real) 18:02
leo_ audreyt: parrotsketch :-)
putter but I'm not sure how one toggles "ok, all formal-argument-symbols now pretend to be lists!" in a non-procedural macor system, pattern matching, like syntax-case. hmm... 18:03
good night audreyt & putter says with just a hint of disbelief in his voice ;)
18:03 G2 joined
putter maybe just magic phrases like "typevar". in procedural (list? formal-var) depends on context, in pattern, magic (de)constructors. 18:06
s/in pattern,/in pattern, use/ 18:07
Juerd: re cafepress t-shirt: searching for "perl" and "perl6" does not find the "s:g/5/6/;" shirt. perhaps some keywords should be added? I also note "perl6" turns up nothing. and neither turns up audreyt's t's. "pugs" might, but with 200+ pages... (only 120+ if one limits it to clothing, but still) I haven't checked. 18:16
Juerd Wha
What the
I switch on the screen, and there's a fresh message from you 18:17
It's a bit scary
Now I will read the message.
putter lol
Juerd I think that re the shirt, cafepress should be avoided.
If you want it findable, link to it
perl6 is the shop name. 18:18
Perhaps audrey can link to it on pugscode.org
putter yes. but someone who is at cafepress and searches for "perl" will find lots of perl swag, but not ours. 18:19
which seems suboptimal perhaps
but not worth worrying about. i just mention it in case it was an oversight. 18:20
Juerd I'm not intentionally making it hard to find
But I am intentionally not making it easy to find 18:21
Because I think such things should manage themselves. If people think the shirt is neat, they'll talk about it and it'll get found through that. If they do not, then apparently the shirt sucks, and would only pollute the search results.
18:22 nnunley joined
Juerd Also, the shirt is an inside joke, so hardly a PR thing. 18:22
putter can one use a pugs with a precompiled Test.pm when smoking pil2js? I'm not sure I understand the concequences of the choice, if any. 18:25
18:32 vel joined
robkinyon sometimes the best pr is an inside joke 18:51
svnbot6 r8861 | audreyt++ | * fp.pm, string_returning_macros.t, my.t:
r8861 | audreyt++ | fix remaining regressions and parsefails.
r8861 | audreyt++ | (Set-Infinite, Span, Perl6-Container-Array and
r8861 | audreyt++ | Perl6-Value-List still needs help.)
lypanov giggles 18:52
lypanov remembers several hours back audreyt saying she was heading to bed ;)
putter well, for whatever it's worth, there's a smoke of r8860 up. the rules falures indicate embedded parrot still segfaulting on me. 19:07
I'll do a non-embedded p8861 now... 19:08
lypanov: yeah. on the one hand, when you're on a roll, its good to go with it. on the other hand, it means your less productive the next day. a well rested audreyt is amazing to behold. fortunately for us poor mortals, it doesn't happen that often. :/ 19:12
s/your/you're/
lypanov putter: /me grins. agreed
putter tries to picture pugs development where each day one woke up to a well rested audreyt chugging away. that would be... more than a little bit mind-blowing. ;) 19:15
audreyt: enough hints for now? ;)
lypanov finds her sleeping pattern fairly sane 19:18
larvae state for several days then 14 hour bursts
lypanov wishes he /didn't/ have to conform to standard 10-6 days 19:19
kattana_ "No rest for the wicked", its hard to sleep with voices whispering in your head. 19:25
19:33 _bernhard joined 19:38 nnunley joined 19:59 vytautas left 20:05 dduncan joined
svnbot6 r8862 | iblech++ | * Usual svn props. 20:13
r8862 | iblech++ | * EOLs at EOFs.
dduncan so, should each ext/ have its own ChangeLog, or their changes be in Pugs' main changelog? 20:16
I see some ext/ have changelogs and some don't
svnbot6 r8863 | iblech++ | ext/Span: 20:19
r8863 | iblech++ | * Several tests in span.t and unicode.t flunk()ed with "this tests is known to
r8863 | iblech++ | hang", but this is no longer the case; with the flunk()s removed and the real
r8863 | iblech++ | tests uncommented, the affected subtests pass just fine.
r8863 | iblech++ | * @array_of_spans[$index].stringify does find Span's stringify method (the
r8863 | iblech++ | usual my $x = @array_of_spans[$index]; ...$x.stringify... works around this
r8863 | iblech++ | bug, but I did *not* add this workaround as audreyt++'s r8842 indicates that
r8863 | iblech++ | the bug /should/ be fixed).
dduncan I would expect the failures I saw yesterday with Test-Builder would be fixed by r8842 20:21
and several others besides
as soon as I've rebuild, I'll continue checking ext/
Khisanth slowly gains enough grip to strangle Haskell 20:26
integral the number 1 curse of haskell is variable names 20:29
svnbot6 r8864 | iblech++ | Pugs.Version: The copyright information still showed 2005, not "2005, 2006". 20:30
r8864 | iblech++ | Also, audreyt, have a look at Pugs.Version.disclaimer (still from r5444):
r8864 | iblech++ | Pugs.Version.disclaimer doesn't mention the Artistic License at all.
gaal integral: you mean that they are named variables when they aren't? :)
rehi btw :)
integral I mean if someone misleadingly names a variable you become more confused :)
gaal ah, because there are no sigils? 20:31
integral yeah, and reading the code is hard because you have to hold so much information in your head I find 20:32
20:32 Aleph__ joined
gaal well, that's what short functions and referential transparency are for :) 20:32
integral yes, but Pugs seems to like to grow multipage functions ;)
gaal audrey mentioned that's one of the reasons for rewriting the runcore. :) 20:33
integral agrees 20:34
Juerd feather goes down in less than 5 minutes 20:41
integral hmm, autrijus seems to have a darcs running, Juerd 20:45
oh, it's gone n/m
20:50 Aleph__ left
Juerd Damnit. 20:51
Feather doesn't eat the new RAM. 20:52
Three beeps.
integral eep. (eep. eep.)
Juerd I don't get it. Same brand, same speed, same latency, just double the size.
integral dust? 20:53
Juerd No dust as far as I can see.
It's fscking now. At least there's some value to my visit :| 20:55
integral hates the propensity of journaled filesystems not to fsck frequently
Juerd What's propensity? 20:56
integral tendency
dduncan now starting to smoke r8864 20:57
Juerd This stuff fscks every 250 days. That's 2 bootups, on average :)
That's often enough
dduncan oh also, one of the recent shell updates messed things up again 20:58
running the pugs shell results in incessant beeps
while the backspace key still deletes, hitting it once removes the whole line like the key was held down 20:59
this would have been after update 8839
perhaps 8840 did it 21:00
putter rehi. smoke of r8861 up.
tewk 21:01
21:01 Khisanth joined
dduncan tewk, your encoding's incompatible 21:02
try unicode/utf8
so I can't see what you said 21:03
but looking at the diff, r8840 only removed the signal lines; that shouldn't cause what I'm seeing 21:05
tewk just like so seend meaningless junk to #perl6 such as ls -al, exit, passwd: *****, or just a space followed by enter :) 21:06
Juerd This memory should work according to the manual too.
tewk launching a darwin smoke
dduncan okay tewk, I can read what you said now
putter iblech: is there any incompatibility between a pugs with precompiled prelude and a smoke-js? 21:09
putter smokes js
Juerd * 21:10
hate
*
s/\n//g;
putter Juerd: what about only using one or the other memory card (assuming they dont have to go in as pairs). perhaps one of the two is fried.
Juerd putter: I've already tried all possible configurations 21:11
One dimm in either bank, two dimms in either order 21:12
Combined new with old
putter :(
integral the old still works hopefully...
Juerd integral: Yes 21:13
Hey, it's booting now
Tried another combination of new/old
It's hard to keep track of what I tried exactly.
Though I had all possibilities covered 21:14
It works because it only uses the old one :(
It just ignores the 1024 one
Leaving only 512
I hate hardware
integral :-( oh well, hopefully we'll bin the precomp prelude soon
putter audreyt: It might be worth adding a note to README's license information that... something like "developers should expect that pugs may be distributed under other terms in future". or something. basically so you can switch to GPL3, or Artistic, or whatever Perl6 ends up with, etc, without the linux kernel issue of "we'd have to get permission from all the contributors to switch". 21:15
dduncan I wholeheartedly agree
theorbtwo The thing to do, then, is point that out before you take patches from anybody, or give out committer bits. 21:16
Juerd putter: Then still you would need permission from all previous contributors.
theorbtwo ...and that.
Juerd putter: I think it's more legal-proof to appoint one person (Audrey, I'd say) as the one who gets to decide. 21:17
So that that is clear upfront.
dduncan on another matter, OSCON 2006 is now asking for speaking proposals, due Feb 13th ... any of you going and/or want to talk?
Juerd Still, permission from all previous contributors is needed in any case.
dduncan but about pugs 21:18
I'm fine if audrey is just assigned copyright for the pugs core
or a certain small number of people
theorbtwo So am I.
dduncan that did most of the work
doing this does make legal matters easier in general 21:19
I say "certain small" since there are some others that did a lot of work, like spearheading the metamodel and pil and such stuff
Juerd Any suggestions perhaps on how to get feather to accept the new ram? 21:23
Hmmm 21:27
Hmmmmmm
I may have another option
I have another box here, that is officially just a test box 21:28
And every user knows it may go down at any point, without notification...
Sounds quite attractive
integral and good user training too. They've got to learn that test means test sometime ;-)
Juerd Hehe 21:29
I'm just considering swapping the two boxes' hard drives
Both have a P4 2.4 21:30
21:30 lisppaste3 joined
Juerd Oh, no users even currently 21:31
Even better
And, it's a P4 2.8
So it'll be a CPU upgrade too
):
eh
:)
Or maybe even >:)
21:33 estel left 21:48 rafl joined
Juerd Damn, that box also doesn't like it. 21:48
21:48 rafl joined
Juerd Though the memory certainly does work in other boxes 21:48
What the hell is going on.
integral not a bios version thing? 21:49
Juerd The changelogs don't say anything about memory
I've never been good at giving up 21:56
So I'm going to try another combination of parts
Juerd enjoys the free delicious hot chocolate milk 21:57
rafl Juerd: Thank you!
Juerd: I finally got enough motivation to set up ctrlproxy on another box :-)
21:58 rafl joined
Juerd What's ctrlproxy, and has it anything to do with your connection cycling? 21:59
rafl Juerd: It's an irc proxy that I need to use because perldition.org's ISP firewalls 6667. The cycling is caused by me being to stupid to configure it the right way in the first turn. 22:00
Juerd: Somehow it's still not working correctly. It says that it joined #parrot in irc.perl.org, but doesn't show me a #parrot window. Can someone tell me if I'm joined to that channel? 22:01
Juerd I hate motherboard swapping. The things have too many screws :(
rafl Therefor I don't use cases. ;-) 22:02
Juerd That'd be rather annoying in a rack 22:04
rafl That's also annoying at home, but it's less work for me. At work I have slaves that screw the machines there up and down if I want them to ;-) 22:05
putter mmm, minions... 22:06
?eval 3 22:08
svnbot has been gone for a while. looks like evalbot is too. 22:09
Juerd: electric screwdrivers? 22:10
Juerd putter: Didn't carry any because I thought I'd be done in 5 minutes. 22:12
Does a fan that came with a 2.4 GHz Pentium 4 properly cool a 2.8?
22:15 rafl joined 22:22 rafl joined
putter Juerd: ? 22:22
rafl OK, I'm hopefully done now. 22:23
putter as in, a bit of surfing didn't turn up anything.
Juerd putter: I've scrambled around parts from three systems and just the other second, saw the first signs of a booting system
The less fun thing is that after this I need to reassemble the other two boxes from the parts I have left 22:24
22:25 jiing is now known as Gniij 22:29 rafl joined
putter hmm. eval_yaml.t has two tests. they work under ./pugs -Iblib6/lib ..., but the second fails under smoke. but if I reverse the order of the two tests, it works. 22:29
Juerd So, feather's now theoretically faster 22:34
But practically slower.
I'll just sling it online and see how well it copes 22:35
It may need a newer kernel
22:39 hexmode joined
putter hopes Juerd is having fun... 22:44
22:45 nothingmuch joined
Juerd putter: Not really 22:46
feather is back in the rack
Don't stress test it yet, please :)
It's not coming up. Fuck. 22:49
Oh, now it finally pongs
First thing I'll do is upgrade the kernel, I think
putter notes that "return to known working state" seems to be fading into the distance 22:51
buu Does anyone know who owns 'evalbot' ?
rafl Juerd: The escape sequence in front of 'Please log out again' is broken. :-)
buu: Currently leo or luqi run's it, afaik. 22:52
buu rafl: Hrm. It's named evalbot_xxx
I was looking for 'evalbot'
rafl buu: That's the revision of pugs that it runs.
integral leo said he'd restart it when feather came back. But I think luqui runs it sometimes, and sometimes autrijus does too =)
buu Apparently it's in use, but not registered.
integral: No no, the nick, not the bot. 22:53
rafl integral: And sometimes I do as well. :-)
integral heh
Juerd rafl: fixed
integral buu: it doesn't use a single nick, it's not compatible with nickserv
buu integral: You seem to be confuse
rafl Juerd: Great. Now I can logout again. :-)
buu integral: I said "Who owns the nickname evalbot"
integral: not "how does the bot some times named evalbot_xxx get run" 22:54
integral oh, sorry. I should have said "who cares? why?"
Juerd rafl: Or watch the box compile :)
buu integral: Because I want it =[
SamB ?eval 1
buu eval: 1
p5evalbot buu: Return: 1
buu =/
integral buu: umm, nickserv says that "Evalbot" isn't registered.
Juerd disk i/o is really slow at the moment
As if DMA is disabled, though it's S-ATA
buu integral: Yeah, but it's inuse
integral buu: msg it!
putter buu: ahem: (17:44:38) buu: Does anyone know who owns 'evalbot' ?
rafl Juerd: It doesn't even compile yet.. boring! 22:55
Juerd rafl: :)
buu putter: Yes?
integral clearly not talking about the nick!
Juerd rafl: Shall I do it in screen so you can watch? :)
rafl Juerd: Sure, why not?
Juerd Screen started. It's really slowly processing bash_completion now.
Feel free to attach
Just don't touch the keyboard :)
22:56 Daveman joined
buu integral: I so was. 22:57
rafl Juerd: Hm.. how does that work if the screen isn't owned by me?
Juerd rafl: Be root? 22:59
> I came across your Web site, and your writeup on setting up a RAID1 on
> Debian (juerd.nl/site.plp/debianraid), and would like to know
> if you can modify your script for me to handle the following: 4 SCSI
> each with 36GB, and 2 SCSI each with 9.1 GB. Kindly send me a perl
> script that will do RAID on the above system.
Of course I will gladly change the script for your specific application.
My hourly rate for custom programming is EUR 75.00; International
clients pay in advance.
integral hah, hehe
rafl Juerd: I still can't read /dev/pts/whatever then. 23:00
Juerd: But screen knows about multiuser sessions.
Juerd rafl: Why can't you read it? it's 620, root:tty
rafl Juerd: Oh.. it's root:tty.. tried it as juerd. 23:01
Juerd sudo screen -x
dduncan Juerd, that's quite a chunk of change
how many hours is a typical job? 23:02
rafl Juerd: Yes. Thought you started the screen as juerd.
Juerd: Configure your kernel!
Juerd :)
putter dduncan: dont forget you have to handle taxes out of it yourself. 23:03
dduncan yes
even so, currently the most I've ever charged someone is CDN$40/hr, for tasks under 10 hours, and $28 for tasks over 10 hours 23:04
and even that, some people consider high
rafl Juerd: No ipv6? :-)
putter goes to find a CDN converter... 23:05
dduncan I think your figure would be over 100 cdn/hr
putter CDN$40 for consulting would be _way_ low around Boston. 23:06
integral dduncan: you seem really cheap
dduncan perhaps ... 23:07
getting more would be nice, but currently my only paid work is from locals
also, my last employment position was about cdn$15.50/hr, up to 3 years ago 23:08
kattana_ makes note to outsource the actual work to dduncan
dduncan which is low by local standards, but its where I was moving up from
well, I am rather starved for paid work at the moment, so I'd probably be grateful 23:09
partly this is because I was focusing on projects that don't pay anything right now
... ah, the smoke's done ...
kattana_ if its that bad maybe you should move somewhere with more tech
rafl Juerd: Huh? IB in included in the kernel nowadays? 23:10
dduncan ironically, Victoria BC is a big tech center
maybe part of this is that what I've been paid for isn't too specialized yet ... typically a few random php things people want help with
integral IB == Infiniband? 23:11
Juerd rafl: "IB"?
rafl Juerd: Infiniband
Juerd I have no idea what that may be
And no, no ipv6 :)
Rogue addresses are a security risk
putter salary.com/ claims the median salary of an entry level programmer in boston is $56k. 23:12
Juerd When I do ipv6 I will do it properly
dduncan smoke of r8864 is uploaded for darwin/haskell
Juerd And only after careful consideration
putter re, smoke, awesome.
dduncan shows 98%
hm, maybe 6.2.11 can go out today after all 23:13
rafl Juerd: IB is ethernet, but fast and without ethernet. I thought the drivers were proprietary.
Juerd: I use it at work and was used to patch my kernel.
dduncan it appears that the only red areas in the smoke are in 4 ext/ items 23:14
I'll look at them in a couple hours ...
I have to go out for yard work while there's still daylight
kattana_ victoria has been a tech hub historicly, but thats history now, tdot and mtl especialy are the growing areas, lots of game and design stuff in qc in general, web dev as well since the big telcoms are in the area 23:15
putter t/magicals/env.t
dduncan and yet, IT is one of the biggest industries here, tourism being the other one 23:16
used to be natural resources, but that isn't so hot
unfortunately, there's a lot of dot-net around here, and I like to avoid that
rafl Juerd: entertaining! 23:17
dduncan I don't move elsewhere partly because the climate is good here
Juerd Should have extracted the kernel into tmpfs rather than this badly performing disk setup
dduncan and I expect that a lot of work could be done by remote
Juerd With 2 GB of RAM, it should fit :P
dduncan anyway, back in a bit
rhesa dduncan, i have outsourced some small jobs to robkinyon
kattana_ s/good/wet/
Juerd rafl: Now look what you did 23:18
fs/bio.c: In function 'bio_alloc_bioset':
fs/bio.c:167: warning: 'idx' may be used uninitialized in this function
That happened when you typed "sure" :)
rhesa he's in the states, while i am in norway :)
dduncan rhesa, I hear you
Juerd I don't get english.
kattana_ outsourcing to the states? how odd
Juerd "I read you" when you hear someone
And "I hear you" when you read what they type
rafl Juerd: Stop the build and I'll fix it.. 23:19
dduncan well, if you have other needs, let me know for consideratoin
rhesa kattana_: most of my regular work is in the states anyway
kattana_ "I feel you" when you're a gangsta
putter the ext/ failures match my micro-smoke or 8868 linux.
Juerd rafl: Nevermind ;) 23:20
rafl :-P
Juerd I'll rebuild the other two boxes in the meantime...
Instead of chat until the kernel is finished and then rebuild them :) 23:21
rafl Juerd: Why? Don't you like night shifts? 23:23
Juerd I don't like spending them in a data centre with people who speak a language I don't understand 23:24
They are having hardware problems, I was having hardware problems. Communication could have helped us both.
rafl OK.. I don't like being in data centres at all. 23:25
putter dduncan: re your 3 t/magicals/env.t failures, I dont see anything similar on either past darwin smokes, or smokes on other platforms. ???
rafl Way too much noise..
Juerd rafl: I like this particular one. They have a relatively silent work floor with free hot coco
(or tea, or coffee, I just think the chocolate milk is splendid.) 23:26
rafl Juerd: Is that the place where we'll sleep during fosdem and where you wanted to show me feather?
Juerd The latter 23:27
I wish we could sleep here instead of my office; it's much closer to Amsterdam
rafl That's enough.. free choco milk, yay!
Juerd lol
rafl Juerd: We could tell them that we need to guard feather?
Juerd Why tell anyone anything? 23:28
rafl So we could sleep there.
Juerd Haha
No, that won't work
rafl :-(
I think I'll leave now.. typing hurts. My fingers are wounded :-/ 23:30
Juerd Good luck 23:31
Who'll keep me company now?
Both webmind and rafl are going to bed
rafl Start evalbot?
Juerd Hehe 23:32
Shit. 23:34
One of the boxes doesn't boot because $colleague forgot to compile in support for sii sata controllers 23:35
That's against our standard set of drivers :)
rafl Go home and let him fix that box tomorrow! 23:36
Juerd Hehehe
No, I do stick to the rules
If you break a box, you fix it :)
Now if I could just find a bootable cd
@#%$#^^$#
I came here unprepared, unpacked.
rafl Juerd: But he broke it.. 23:38
Juerd I see a box of ubuntu cd's in a closed office room 23:40
How annoyingly frustrating.
putter ok, the Span file failures are all the same - failure to find it's method stringify. @a[0].ref is Span. but @a[0].stringify() doesnt.
rafl putter: iblech commited something concerning this issue today. Take a look at the commit log. 23:42
putter will do. it's the access via array which is key. my $x = @a[0]; $x.stringify works fine. 23:43
sigh. ok. known problem. the Span failures need core hacking. 23:45
colabti.de/irclogger/irclogger_log/...l=248#l395
Juerd Broke a fucking s-ata connector on a mainboard 23:47
Triple damnit.
WHY do they make the cables so that they withstand almost everything, including chewing, but then invent such fragile connectors? 23:48
And when will that kernel compile be ready, damnit 23:49
putter Juerd: so you always know where the problem is. if connectors were more robust, then sometimes it _wouldn't_ be the connector that failed. that would be very confusing. 23:51
Juerd putter: ...
But now things break
I don't mind cable connectors breaking
I do mind headers on mainboards and harddrives breaking
"We took a good look at the USB design and copied the good parts" 23:52
No, morons, USB is rugged compared to this flimsy shit
rhesa they didn't say where they copied the good parts to...
putter what is @.foo ? .foo ./foo but @.foo??? 23:53
*chuckle*
Juerd putter: has @.foo 23:54
putter: has $.foo
putter: Does that ring a bell yet?
rhesa class attributes?
rafl Juerd: re "when will that kernel compile be ready": Do it like I do.. get an instrument (like a bass) and practice when something compiles. That's the only way I survive those tons of pugs builds I need to do before I have some good Debian packages. :-)
Juerd I'm in a data centre. 23:55
rafl So what?
putter Juerd: ah. more like a wood-on-wood thunking noise, but yes. ;) thanks
rhesa even better: no neighbours complaining ;)
rhesa sympathises with Juerd
Juerd There goes feather with the new kernel 23:56
putter rhesa: yes, tnx
rafl Juerd: And my screen session dies.. :-(
Juerd rafl: No instruments here :(
rafl: That was to be expected
rafl Juerd: Doesn't feather have a serial console?
Juerd No
rafl Juerd: Buy one and take it with you!
:-(
Juerd Come up, feather 23:58
Please don't fail on me now
Pongs!
Okay, now boot sshd
Thanks for your speedy reply to my e-mail. Unfortunately, I simply don't have the 23:59
kind of money you are charging for your services.
Again, thanks so much for your reply, and best wishes.
23:59 azuroth joined, stennie 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!