Chip needs help!!! www.geeksunite.net | pugscode.org/ <Overview Journal Logs> | pugs.kwiki.org | paste: sial.org/pbot/perl6
Set by theorbtwo on 10 July 2005.
svnbot6 r5457, nothingmuch++ | geoffb++'s next bug test case, and a similar test case for redo. 00:00
putter Do we have NEXT blocks yet? 00:01
nothingmuch not that I know of
that will require unification
we should generalize, actually
the next constructor takes the action for next
but also performs NEXT { } blocks on the sub body
and does the continuation magic
i'll discuss with autrijus tomorrow 00:02
putter ok.
can one make $i rw in for 1 .. 10 -> $i { ... } ? 00:15
nothingmuch not in a range 00:16
what are you mutating?
you can declare is copy
putter I went to transliterate a ruby redo test, and it didnt quite work... 00:18
for i in 1..10; sum += i; i -= 1; if i > 0; redo; end end assert_equal(220, sum)
Juerd "i -= 1; if i > 0; redo;" 00:24
That's just downright stupid, with all respect for the original author.
coral interesting way to do a while loop
er, for
Juerd Just use i > 1
And that's not even looking at the horrors of changing what was a constant. 00:25
coral there's more than one way to do it, but only one will look satisfactory to any single person.
putter ok, with -> $i is copy {}, it runs, but loops indefinitely. It looks like redo takes us back before the assignment to $i. perl5 otoh has the same behavior as ruby.
for my $i (1..4){$i--;print $i;redo if $i > 0;print "\n"} works.
Juerd putter: I think the original code needs a change first.
putter Juerd: ?? 00:27
Juerd I'm amazed that it works in Perl 5.
Apparently ranges do result in writable values, and 1..2 isn't expanded until runtime, so no literals are changed. 00:28
02:28 < Juerd> That's just downright stupid, with all respect for the original author.
I take that back. The code was misleading, not wrong.
nothingmuch geoffb: you're off the hook, I did all three
svnbot6 r5458, nothingmuch++ | Add tests for redo and next under all loop (;;), for, and while (every loop 00:29
r5458, nothingmuch++ | type is broken in one way or another - tests todo'd)
Juerd Still, a range should theoretically of course return immutable values
putter perlfunc shows redo returning to while (EXPR) { #### HERE .... }.
Juerd Whatever 1..3 returns should be equal in semantics to 1,2,3 00:31
I'd consider perl 5's behaviour a bug.
putter Juerd: it does... which is why the for mumble -> $i is copy {mumble} was needed... no?
Juerd Yes.
nothingmuch good night everyone 00:32
putter Good night. I'll add a test for "returning to too early redo". Fun. tnx 00:33
nothingmuch returning to too early redo?
putter err, backlog?
this, 00:34
nothingmuch oh
redo redoes the cond?
putter for 1..2 -> $i is copy { $i--; redo if $i > 0; }
nonterminates
nothingmuch appearantly it redoes the increment clause of loop(;;){}
which is not good 00:35
see last test in t/statements/next.t
putter looking...
nothingmuch $j is not 4 by the time it's done, but 6
i'm not sure if it's because it does the butt end of the loop 2 times more than it ought to
anywho, *poof* 00:39
svnbot6 r5459, dudley++ | Removed the extraneous quotes that were causing the test for [...] on user defined ops to fail
putter nothingmuch vanishes in a cloud of smoke. 'night 00:40
kelan seen kelan 00:41
perlbot: seen kelan 00:42
putter does this look plausible? my $i; while (do{$i++ < 10}) {say $i} 00:53
kelan why have the do-block? 00:54
putter the real example is something like do{$x=$i; $i++ < 10}... but bad things happen... 00:58
(nonterminates, and $x is not set) 00:59
svnbot6 r5460, putter++ | Additional redo.t tests. 01:12
putter Hmm... in p6, how does one next/last/redo to non-inner loops? p5's OUTER:{... while ... { last OUTER;}} 01:17
autrijus that should continue to work (even tho we don't yet support it) 01:20
putter ah, tnx.
autrijus leave :label<OUTER>;
is what that really means
putter tnx. I started to transliterate p5's loopctl.t, but it's too dependent on labels and NEXT blocks. ah well. 01:24
autrijus for some reason leave() works in PIR :) 01:29
it should be much easier to work with once PIL evaluator is there 01:30
putter right :). 01:31
svnbot6 r5461, fglock++ | added bug list; fixed docs 01:37
autrijus dear larry. you have again inflicted your whim on me. :-/ 01:41
putter ? 01:42
autrijus so this week I need to, at compile time, detect if $?SELF and $_ is the same thing
and if yes, raise a fatal exception on .method
er
I mean, if not
putter ;) 01:43
autrijus and force dialects with "use self"
and it's only applicable this week 01:44
mugwump why not just defer such behaviour until some indeterminable time in the future when everything is self hosting and mutating the grammar isn't a PITA
putter hmm, perhaps that's not the highest thing on the priority list...?
autrijus esp. seeing how existing ./foo needs to be rewritten
putter: yeah, but I feel uncomfortable having a mass of test suite actively disagree with ruling 01:45
or do you think it's sane to say that pugs has `use self "./"` always in scope, until we really have a compile time and PIL evaluation, at which time self.pm can actually be implemented with P6? 01:47
mugwump sure. why not do something really interesting, like Proxy objects :) 01:48
autrijus mugwump: there's already ScalarProxy IType
just hook it to Proxy::new and you're set
mugwump Oh cool... and I was figuring that the Array and Hash proxies wouldn't need much, probably just connections with the hash/array de-ref methods 01:49
ie, the post_circumfix:Ā«{}Ā» methods,
and post_circumfix:Ā«[]Ā» of course. 01:50
putter I do think it sane. Hmm... so let me reflect on why...
mugwump everything else is simply an object method on the hash/array
putter Timeframe is one issue. A hack for the day would be a no-brainer. A forever divergence from p6 spec would be problematic. Someplace in between discomfort grows. 01:53
A possible metric is the usual "the next week of development time can buy x or y. which?" 01:54
I started a "ruby on p6" port. The roadblocks were pugsbugs, and objects being largely unimplemented. Speed was secondary. 01:57
mugwump autrijus: I'm interested in playing with the IScalarProxy type, I notice that there is an associated class Scalar::Proxy, and I can call .new() on it; what else can I do with it? 01:59
autrijus mugwump: not much yet
mugwump the symbols "scalar_fetch = fst", what does the "fst" refer to? 02:00
(in the relevant definition in src/Pugs/Types/Scalar.hs)
I see a proxyScalar in Pugs/AST/Internals.hs, too 02:02
autrijus proxyScalar is the one you're looking for 02:03
it takes two thunks, one for fetch and one for store
mugwump and yields an IScalar 02:04
autrijus and return a IVar VScalar -- suitable to pass into IScalar and maybe VRef
mugwump ok, so I could make a primitive that returns such an item, and then make the Scalar::Proxy::BUILD call that? 02:06
putter Hmm, what is the motivation behind the current focus on PIR? 02:07
mugwump putter: AIUI, to simplify the task of getting to a self-hosting point
putter so a path of getting working: pir backend, hs frontend, p6 meta, p6 frontend? 02:13
mugwump Hmm, my comment wasn't quite right is it. that's a secondary goal, the first is simply to make it easier to compile to backends by reducing the number of distinct operations that need to be translated/tested to the target language/platform. this should aid in the self hosting goal. 02:14
putter so -BPIR is primarily being used as a motivating case for a kernel-languagey restructing of pugs? 02:16
autrijus putter: PIR codegen is merely a motivating case of PIL compilation 02:18
putter: the most pressing thing for me is to switch to PIL evaluator in haskell side so our interpreter won't differ from compiler
porting the metamodel to Haskell goes with that. 02:19
after those two, we'll have reliable way to write codegen in P6
for example codegen into P5 02:20
and _then_ we can port compiler to P6
I need to draw some graph to explain all this.
(btw, my reply to larry is posted to p6[cl]. feedback welcome)
geoffb nothingmuch++ for doing all the loop tests -- I just got slammed this afternoon with work stuff, sorry about that 02:31
putter so the top level objective is: replace the current hs interpreter backend because... something. (a clean backend is a prerequisite for self hosting, and simplifies pursuing it?) 02:32
that should have been, 02:33
so the top level objective is: replace the current hs interpreter backend because... something? (a clean backend is a prerequisite for self hosting, and simplifies pursuing it?)
geoffb that was different how?
oh, s/./?/ in one spot
sheesh
mugwump putter: for consistency of behaviour between interpretation and compiled execution
putter and compiled execution is needed because...X? 02:34
autrijus putter: currently there is no compile phase 02:35
mugwump lots of reasons. to run outside of pugs, for instance on parrot. to convert to other code forms. to boldly run where no dynamic code has run before
autrijus so we don't get strict, warnings
pragmas 02:36
and "1 = 1" does not fail until runtime
there's no semantic analysis.
treatment of class literals depend on order of introduction.
I think those are reasons enough
I hope you agree :)
hm, the first paragraph in the perl 6 faq is bad. 02:37
autrijus mails allison to see if it can be taken down.
geoffb (Note: this is NOT intended as pressure, I am merely inquiring): autrijus, I've been holding off learning Pugs' Haskell guts on the assumption that the PIL change would pretty much put the internals in a blender. A) Is this true, and B) when abouts is the planned change? 02:38
autrijus geoffb: this is not true; PIL evaluation will be a refactoring of Eval.hs, not rewrite
the planned change is supposed to start happening as soon as 6.2.8 is out of the door.
putter Oh, I don't disagree with any of this. I've just been aware for a long time that I couldn't write the one paragrah pithily insightful description of what we are doing, when, and why. So I'm taking this opportunity to explore it. Can write up result.
autrijus ooh, please do 02:39
I see, you're the interviewer here
putter Yes :)
good description
so the primary objective is to obtain a compile time? 02:40
err, compiler opportunity?
autrijus right.
to not run the parse tree. 02:41
to allow a semantic analysis of parser tree
geoffb OK, given the above response, I'm thinking of diving into src/ sooner rather than later. What is available to help me understand what does what? 02:44
svnbot6 r5462, geoffb++ | hack.pod updates: split ext/Span out of ext/Set-Infinite, and reword the set-related module descriptions to be parallel
mugwump geoffb: PA01?
autrijus geoffb: PS02
er
geoffb If the answer is "not bloody much", I may end up writing something, but no promises yet
autrijus PA02
you can augment PA02 :) 02:45
geoffb Is that up to date enough not to lose me?
autrijus it's up to date enough.
it is also nothing much
geoffb OK
autrijus but I'm thinking about renaming it into PA02
er
PA03
and leave PA02 as the design thing that putter was interviewing me with 02:46
i.e. the grand scheme of things
geoffb
.oO( I must be a masochist with the amount of work I pulled onto my pile in the last few days )
autrijus geoffb: arbeit macht frei
geoffb wah? 02:47
autrijus "Arbeit Macht Frei" (Work Brings Freedom) was the sign over the gates of Auschwitz.
geoffb gets chills
autrijus it's intended to have that effect :)
autrijus hacks on with pirsmoke -- that was (and still is) an insane amount of work too :) 02:48
geoffb I am reminded of freshman CS, in which the prof gave extra credit on a midterm if you answered the question "What is the best novel ever?" with "The Dispossessed" 02:49
autrijus lol 02:50
putter so the top level objective is to do compiler thingys, rather than running the parse tree? this requires an alternate tree? PIL? Eval.hs is refactored to... run this tree rather than parse tree?
autrijus putter: yes and yes. 02:51
theoretically we can go the Perl5/Ruby route and augment the parse tree instead of inventing a new minimal language
but that makes the life of codegen immeasurably more difficult.
and nonlocal optimization virtually impossible
putter kernel languages++
autrijus so we introduce PIL as a core calculus of perl6
yup
geoffb Does PIL though contain enough information to make micro-optimizations relatively easy? As I recall, P5 does a lot of that sort of thing 02:52
autrijus really? 02:53
it does constant folding
static lexical lookup
that's all 02:54
I think
oh and void map, and reverse sort
sure, all those will be possible
geoffb (IOW, if everything that loops gets completely abstracted to a single overarching continuation concept, do you then lack the necessary info to do stuff like turn while 1 {} into something really damn simple and fast?)
autrijus do note that it's just a call to &statement:<while>. 02:55
geoffb ok, fair enough
AH!
autrijus so... the answer is voidly true :)
geoffb OK, not having actually looked at PIL, I imagined that you had made something between a turing machine and PIR in complexity -- extremely abstract
turing machine with continuations? :-) 02:56
autrijus PIR is between turing machine and PIL.
PIL is basically scheme with some more calling convention.
putter so work on Eval/PIL will begin shorty, and work on PIL-to-PIR is in progress? Why PIL-to-PIR before Eval/PIL? The motivation for PIL->PIR is... something. (reality check on PIL design?)
autrijus putter: PIL->PIR is keeping us honest.
it's trivial to make PIL bloaty and codegen unfriendly
by shuffling a lot hidden complexity to haskell runtiem
switch to another runtime forces us to define all operational semantics 02:57
geoffb autrijus++ # psychological engineering on himself
autrijus because haskell runtime and pir runtime (and p5 runtime) cannot agree
so all semantic must be preserved at PIL level
geoffb bah, must go again.
autrijus otherwise I imagine Eval/PIL producing something that's hard to codegen for. 02:58
s/for/from/
(if we started from that first)
so yes, "reality check" is pretty accurate
putter: please commit whatever you have written down to docs/notes/ :) 02:59
putter so pir, rather than p5, because...? (same reason?)
will do. :)
autrijus because p5 is too smart. 03:01
similar to our existing hs runtime
there's too much logic in p5 code
meppl what a pitty, that im drunken again - bye
autrijus even in a single method call
meppl: see ya
so it does not keep us honest the same way as PIR does
otoh, once PIL is more solid, p5 runtime will have a lot of practical value 03:02
because p5 is known to segfault less.
and DBI is written in p5 not pir.
so it makes sense to pursue Eval/PIL with PIL->PIR until we have a good idea of what PIL's final form will be 03:03
then work on PIL->P5 in earnest
and maybe PIL->JS etc.
stevan (PIL->JS)++
hola autrijus :)
autrijus greetings stevan :)
autrijus goes to find food and repair PIR 03:04
putter: please type away more questions if any -- I'll reply when I'm back
stevan autrijus: I am thinking of implementing Scalar, Array and Hash in the metamodel, but using tie() structures for now
autrijus stevan: sure, tie() structure is what I'm using in hs anyway
stevan when you return from food, let me know you opinion of the sanity of that idea
autrijus same interface
so sure, sane
stevan yes, this was part of the inspiration 03:05
ok
this way I can do typed containers and ro/rw traits
and get that out of the metamodel
putter so the motivation for doing PIL->P5 once PIL has a finalish form is... something?
stevan then we can compile Perl 6 to Perl 5 :)
putter :) 03:06
autrijus the motivation is that we can write "use perl5:DBI;" and expect it to work in both hs evaluator and p5 evaluator. 03:08
and have it work far more reliably in the latter.
I prophesy that once that happens we'll enter the "hack hack hack" phase of imaginary perl6 timelinee
and out of the current "slowly gaining more respect" phase 03:09
stevan is very happy that Larry is talking meta-model stuff :)
hack++ hack++ hack++ :)
autrijus bbiab. :) & 03:10
putter so at some time X? weeks from now, Eval.hs has been refactored. the primary objective then becomes... what? (PIL->P5?)
stevan I think the PIL>p5 effort will happen simultaneously with the PIL in Haskell effort 03:11
especially since PIL is soooo much cleaner in Haskell than it is in Perl 03:12
but that is not what I am working on, so I can only guess :)
autrijus stevan: PIL in Haskell is Eval/PIL is Eval.hs refactoring 03:19
PIL->P5 will probably happen simultaneously with PIL->PIR and other PIL->* as people find time to work on it
I'm equally interested in P5 and PIR targetting, but P5 brings more instant gratification 03:20
mugwump so it's not better to go PIL -> PIR -> P5 ?
or does that miss out on too many good integration features
autrijus mugwump: PIR on P5 is an interesting subject. I suspect it will run very slowly
unless you of course link with libparrot and pmcs 03:21
if so, that is valid approach (and the one allison originally proposes) but I'm not personally interested
if we have have a single runloop, why settle with two? 03:22
and some core parrot opcodes, eg. bsr/ret and tailcc 03:23
and its calling convention
cannot be directly expressed in p5
they must be emulated if youa re targetting pure p5 (without linking libparrot)
and I suspect it will be slower than acceptable
but people are welcome to try :)
putter so the comming thing in pugs is PIL? PIL->PIR provides honesty in PIL design? PIL->P5 provides reliability and visibility? PIR->hs/Eval provides... something? 03:38
svnbot6 r5463, autrijus++ | * The Alligator in 0.2.2 prevents the perfectly good 03-equal
r5463, autrijus++ | sanity test from working. Got no choice but to remove
r5463, autrijus++ | two lines that prints nothing; need reinvestigate with leo-ctx5.
r5464, autrijus++ | * remove dead code in Bind.hs.
autrijus putter: provides clear definition of semantics.
putter: taking advantage of haskell's uncanny ability of serving as specification that happens to run. 03:39
so it is the reference evaluator.
until, some day, we finish PIL->Perl6. 03:40
which would be the last milestone of pugs
(of course, the compiler, emitter and metamodel + object space should be all in P6 in the last milestone) 03:41
but there are two levels of PIL->P6; one is merely a pretty printer 03:42
which will not need to implement any semantic
the "last milestone" one is however different; it's basically an implementation of a perl 6 virtual machine on top of perl 6
it's there so we can replace the haskell one as the reference implementation 03:43
alternatively, if people all agree that parrot runtime should be the reference one
then that step would not be needed
hence the "if needed" clause in the pugs roadmap.
putter so for X in hs,pir,p5, what does a success for PIR->X look like? (X-smoke?) 03:45
autrijus X-smoke. 03:47
putter an "X-smoke no worse than hs-smoke"? 03:48
autrijus right 03:49
the success will be quantifiable
it's not a 0% - 100% thing
er
0% / 100% thing
putter so what are the parts to an PIR->X? (an emitter? an X-Prelude.p6? a X-native runtime? other?) 03:52
autrijus assuming you mean PIL->X
putter oh, yes... 03:53
autrijus the parts are: CodeGen from PIL to X-tree; an emitter from X-tree to X-bytecode; a X runtime.
Parser->Compiler->CodeGen->Emit->Runtime 03:54
putter which parts are written in hs, p6, X? 03:55
autrijus only the Runtime is in X. 03:56
currently all other parts are in hs
to be replaced by p6 gradually.
X does not enter the picture.
except for the Runtime
which includes object space, primitives, and maybe more emulation foo 03:57
oh, replace X-bytecode with X-code 03:58
we don't quite care whether it's bytecode of other kind of concrete representation
putter before broadening focus, anything else which should be said about the backends? 03:59
so what else is going on as these backends are being developed? 04:01
(tests? frontend? docs? somethings?) 04:02
rephrased, n weeks from now, three greenish smokes, what else would you like to have seen happen? 04:07
masak yawns 04:16
morning 04:17
QtPlatypus asks a question for confermation "At the moment in pugs where is no way to manipulate raw bytes, everything gets utf-8ed." 04:24
brentdax Hmm. 04:26
autrijus QtPlatypus: that is correct.
QtPlatypus: mostly due to the fact that I'm not sure how raw bytes are to be presented in p6. 04:27
putter: tests are shared. other frontends -- i.e. *->PIL may be interesting, but not of active interest to me
QtPlatypus And strings that corraspond to illegal Unicode or utf-8 cause pugs to get indigestion.
autrijus putter: Perl5->PIL is indeed interesting, as is Python->PIL etc 04:28
brentdax Perl 5 interop with DBI. $sth.execute(...) says "no compatible sub". $sth.can("execute").($sth) executes happily. This only happens in a module which is part of a large program--when I write a little test script, it all works fine. 04:29
Other DBI::st methods (like .fetch) seem to work fine. 04:32
Any thoughts?
autrijus hrm, not sure. 04:42
mugwump reducing to a simple test case can really be a bitch sometimes aye!
autrijus hear hear. 04:47
autrijus hopes allison will reply that punie is targetting perl6 :) 04:48
obra ? 04:50
autrijus obra: # www.nntp.perl.org/group/perl.perl6.compiler/931
putter weeee... putter finishes pass at writeup. looks at clock, and concludes discussion of frontend can wait 'till another day. 05:16
autrijus ooh ooh mesee mesee
commit to docs/notes/ ?
putter ok... 05:17
autrijus weird, the feature motd looks upside down 05:18
Ā”Ę®uĖ™9Ę®É¹ĒdĖ™É¹ĒÉ„Ź‡ÉĒÉŸ oŹ‡ ĒÉÆoɔʮĒM
s/feature/feather/
putter 5465. docs/notes/plan Too tired to think of a better name.:/ 05:22
svnbot6 r5465, putter++ | Added docs/notes/plan - some notes on the pugs development plan.
autrijus hm, evalbot is offline 05:23
and borken
I wonder why
putter++ putter++
lunch... bbiab &
putter It is not the pithy paragraph I originally envisioned. Who was it that said, sorry for the length of my note - I didn't have time to write a short one. 05:24
Any/everyone, feel free to pithify. ;) 05:25
Thanks autrijus, was fun.
good night all &
autrijus nite! 05:26
&
putter & 05:27
geoffb If noone else is working on refactoring putter's note, I'll take a crack at it . . . 05:38
geoffb begins hacking on it 05:42
masak putter: (in case you're backlogging) it was supposedly oscar wilde who said "I'm sorry for the length of this letter but I didn't have 05:46
time to write a shorter one."
gaal i think not. voltaire perhaps; checking. 05:47
ah: pascal. 05:53
06:16 spinclad_ is now known as spinclad
svnbot6 r5466, gaal++ | * `require Bar; require Bar` only loads Bar once 06:43
geoffb is having fun rewriting docs/notes/plan
I know developers are supposed to hate documentation, but I just don't. :-) 06:44
gaal docs can be fun given the right *.
pugs often has the right * to it. 06:45
06:45 castaway_ is now known as castaway
geoffb gaal, nodnod 06:46
allo, castaway
masak gaal: you're right about pascal, the quote is "I would have written a shorter letter, but I did not have the time." according to wikiquote 06:53
gaal word. 06:54
castaway g'mornin gaal 06:56
gaal hey castaway!
castaway umm, mornin geoffb .. /me rubs eyes
geoffb :-)
gaal offers coffee
geoffb offers tea
castaway tea :9 06:57
see, you must be a brit :)
gaal bbiab &
geoffb castaway, :-) 06:58
castaway mm, you like writing docs, eh?
I wouldnt spread that about were I you :)
geoffb I can always say no -- that's the beauty of the bazaar. :-) 06:59
castaway ,) 07:00
castaway has bundles that needs writing..
geoffb Well, if you'd like to spring for my consulting fee . . . :-) 07:04
castaway heh
gaal okay, here's something sad: i was experimenting with precompiling Test.pm 07:05
but it causes pugs to segfault :(
geoffb bleah 07:06
gaal inconsistently :(
geoffb double bleah
theorbtw1 Segfault? 07:13
07:13 theorbtw1 is now known as theorbtwo
theorbtwo With precompile to AST, or precompile to something more parroty? 07:14
gaal to2: apparently the segfault was just an optimized-unoptimized artefact.
looking at other problems now
theorbtwo Ah, that works for me. 07:15
Hm, I should look into the gcc4/ghc problems, possibly, since debian unsable now defaults to gcc4.
gaal on my fast, machine, ghc peaks at 120mb. i want it to use *more* ram if it needs it. how do i do that? 07:19
i'm not sure if it gets there because of a limit or bacause that's what it needs. 07:21
oh, i just said that and now it's at 175mb. heh. 07:22
geoffb gaal, opt or unopt? A couple days ago I had unopt peak at 140 MB ghc + 130 MB ld
Limbic_Region has seen a peak of 750 MB each for ghc and ld with optimized builds, but that was before PCPrelude. 07:23
gaal optimized. Pugs.Run with large includes.
on the contrary, PCPrelude increases that figure.
geoffb Though Limbic_Region may have been compiling on Win32, I'm not sure.
gaal well not *that* figure
it increases build-time ram usage.
well, it eventually peaked at just under 200mb. hmmm. 07:24
geoffb Build time RAM was what I was referring to. :-)
gaal yeah - meant i never saw it anywhere near 750mb 07:25
geoffb gaal, *nix or Win32?
(My numbers are on Debian testing) 07:26
gaal win32. on linux my box only has 256mb...
so i never make optimized there.
geoffb yeah, opt make OOM kills my linux box (192 MB) 07:27
gaal brb
autrijus, ping 07:35
07:37 Aankh|Clone is now known as Aankhen``
autrijus gaal, pong 07:39
gaal hey. i'm getting something weird precompiling Test, and i remember something about END blocks...
autrijus right. :)
gaal i've basically concatenated Prelude.pm with Test.pm
autrijus @*END would be filtered out 07:40
gaal but now i'm getting errors about "undeclared $testing_started"
autrijus yes, because the compilation does not preserve lexical context
gaal in &test_ends
autrijus so you don't get visibility
you can workaround using explicitly qualified vars
gaal but some other classes in the prelude do have that: eg Time::Local
oh 07:41
Aankhen`` Morning.
gaal no it doesn't :)
geoffb morning, Aankhen``
autrijus it doesn't. :)
it's an outstanding problem
maybe we can workaround it for this release
by using qualified vars in Test
what do you think?
gaal errr wait, File does.
oh and it does indeed fully qualify them. 07:42
autrijus it does not.
and that means the three 'my' should really be 'our'
the 0 worked by chance.
because undef happens to be 0.
gaal it says the crazy $File::SEEK_START
ahh :)
okay :)
Aankhen`` geoffb >> What's shakin'?
geoffb Rewriting docs/notes/plan 07:43
Aankhen`` Er... no commits since last night? o_O
autrijus but I suggest you just use qual now
geoffb (almost finished with that)
Aankhen`` Oh, that was my browser messing around.
autrijus $File::SEEK_START = 0;
etc
so what about Test? do you think we qual it for now?
gaal autrijus: sure. but (a) do we just not declare the vars, like p5?
autrijus that can do
gaal (b) i hope TEst doesn't make use of the lexicality of vars, eg. with closures 07:44
autrijus would you like to take a look at it?
gaal i'll happily experiment.
autrijus please do.
gaal++
btw if you have parrot
gaal yes
external executalble though. 07:45
autrijus you can easily test that with ./pugs -BPIR -MTest -e 'ok 1'
that's fine
it should exhibit the same bug
for exactly the same reason
gaal right, i'm on it.
autrijus gaal++
Aankhen`` I have a question: since Prelude is precompiled, does that mean it makes more sense for builtins to be implemented there from now since they eventually need to be written in Perl 6 anyway? Or will that be done later?
autrijus Aankhen``: yes, it makes total sense now 07:46
I said that in my journal
new prims should be in Prelude.pm preferably
I don't see a need now to completely redo Prim.hs
gaal we need a way to tag them as pure perl imo.
Aankhen`` I just wanted to clarify "I expect" vs. "I think they should be". ;-)
autrijus ok. I expect, and I think they should be 07:47
gaal be it a trait or a documentation convention.
Aankhen`` OK.
geoffb OK, docs/notes/plan updated. autrijus, can you take a look and see if you disagree with anything? 07:51
Aankhen`` $I.ve(got => "the power"); 07:55
svnbot6 r5467, geoffb++ | rewrite of putter++'s interview/brain dump with autrijus on PIL and Pugs as a compiler
gaal Aankhen``, perl4 is clearly superior to p6 as your example demonstrates. (you can spell that "I've" in p4) 07:56
Aankhen`` BAH, I SAY 07:57
geoffb You can spell it in P5, too, but people will look at you funny
gaal true.
theorbtwo macro <'> {'::'
Er, wait, not quite.
geoffb That could have interesting effects on quoting . . . . 07:58
Aankhen`` macro infix:<'> is parsed(/<null>/) { "::" } 07:59
gaal even though precompiling Test is expected to give another factor of two speedup in smokes, once it works, it involves a significant build time hit.
autrijus geoffb: I committed chgs. I'll brb 08:08
svnbot6 r5468, autrijus++ | * minor fixes to geofb++'s hackery to notes/plan
autrijus geoffb++
gaal does it make sense that i have to fully qualify sub names?
&proclaim isn't being found.
it's declared after most of the subs that call it. 08:09
geoffb pulls
Cool, thanks autrijus. I knew the overview chart would need fixing, but I figured a first pass would help. 08:12
gaal geoffb++ putter++ # helpful docs 08:13
geoffb thanks, gaal
autrijus geoffb: it helped a lot! I think I'll just build PA02 out from that and hw2005 08:27
gaal i have a smoke running on precompiled Test. a little soon to say if it's faster though. 08:28
geoffb cool beans, autrijus
QtPlatypus can't work out why grep expects Vcode 08:32
svnbot6 r5469, autrijus++ | * acknowledge zeroderef and context bugs as :todo<bug>.
QtPlatypus Should the following work? 08:33
grep /a+/,"a","b"
gaal i think not
QtPlatypus gaal: Why not?
gaal iirc only grep CODE LIST is in now?
geoffb is several hours overdue for bed . . . time to attempt sleep
QtPlatypus nods. 08:34
geoffb mumbles "It would be spiffy keen if someone were to fix the while+next and related bugs . . . ."
gaal apparently i'm wrong though... S29 mentions "MatchTest $test"
but i don't know what MatchTest is. 08:35
autrijus it's Item or Junction.
but for all practical purpose we can limit it to Rule 08:36
I wonder who added it there, but it makes sense
gaal hmm, since the reference implementation does a smartmatch, why not put an Any there?
autrijus that would do too 08:37
although Any would include pairs
grep :x<y>, "a", "b";
so I think Item|Junction is indeed correct
QtPlatypus: make it a :todo<feature> 08:38
svnbot6 r5470, autrijus++ | * more test cleanups.
autrijus QtPlatypus: if you would like you can also implement it :)
modify op2Grep to do a smartmatch
QtPlatypus nods. 08:41
gaal autrijus, any idea why i needed to fully qualify sub names in Test too?
autrijus gaal: namespace bugs 08:42
gaal: is it PIR or precomp?
gaal precomp
autrijus I bet the namespace isn't propagated
so it looks at main::proclaim
because it's exported
and called inmain
gaal smoke ended, let's look at the times.
autrijus I need to fix export.t first too
gaal duration: 1350 08:43
that's measly :-(
precomp prelude alone gave ~1400
but that must meant my experiment is flawed. 08:44
svnbot6 r5471, autrijus++ | * more deref victims.
gaal looks as if Test.pm is compiled again? /me moves away Test.pm
drats, if *did*. but why? i added an %*INC guard. must have had a bug there. 08:47
sysinternals.com++ 08:49
aha! %*INC was populated with "Test.pm" instead of "Test". that's probably wrong, isn't it? 08:51
autrijus it agrees with p6. 08:52
er, p5.
$ perl -MTest -e 'die %INC->{"Test.pm"}'
/usr/local/lib/perl5/5.8.6/Test.pm at -e line 1.
autrijus gleefully abuses one of the p5 bugs
(the %hash->{key} syntax) 08:53
oh wait, it's a feature not a bug.
gaal okay, /me gleefully changes his most hacky line yet to %*INC<Test.pm> = '<precompiled>';
it's a deprecated feature
autrijus yup.
#perl6: Powered By Gleemax 08:54
gaal with no justification whatsoever.
obviously likeable. :)
autrijus, what's the advantage of keeping the keys of %INC in filesystem format? classname format looks cleaner to me 08:58
du -b src/Pugs/PreludePC.hs => 1763898 # eep 08:59
autrijus gaal: I suspect it's due to require's desugared form. 09:01
require 'Test.pm';
require 'Test'; # won't work
gaal yeah
autrijus because you my be doing
require 'heya.gaal' 09:02
gaal but we don't like that, do we? :)
autrijus and then what will its %INC entry be?
I do not like it but I see its use.
gaal why not make require accept only class names? the fs mapping is a runtime detail. 09:04
okay 110 seconds is much too fast.
autrijus sometime I may really have some perl code is "config.ini" 09:05
or "foo.conf"
or "heya.gaal"
kungfuftr or blah.pl 09:06
autrijus yes.
gaal or evil.com
theorbtwo Or 123341232.dump
autrijus to not allow that is, to quote tim oreilly, "forgetting our working-class roots"
;)
gaal so write your own loader, or else we need a primitive
"working class" 09:07
require
and a higher level one.
autrijus but what will %INC be?
gaal what does the key of p5 %INC mean anyway? it's some sort of key to the module, not to a class 09:08
because MyMod can contain many classes
broquaint %INC key = join('/', split '::', $name) . ".pm";
autrijus the key of p5 is filenames used in require.
gaal i meant what does it conceptually mean 09:09
it's the name of a unit of code
autrijus it's the name passed to require.
theorbtwo It's the file (under @INC) that is being required.
broquaint It's the name of the file the represents a unit of code.
autrijus that's what it conceptually means.
$ perl -e 'require "/home/autrijus/work/pugs/z"; die keys %INC'
/home/autrijus/work/pugs/z
$ perl -e 'require "z"; die keys %INC'
z at -e line 1.
gaal i suppose i'm saying that there's no reason to use "filenames under @INC". 09:10
theorbtwo But there is, because otherwise you can't record things that are only filenames. 09:11
autrijus require 'auto/Tk/Widget.al';
require 'open2.pl'
gaal i mean i want to widen it, not narrow it
theorbtwo Have it either be a filename or a somethingelse? 09:12
gaal URI:code.com/yourmod
theorbtwo But then you have to worry if this filename and this somethingelse are really the same thing.
broquaint As long as it's consistent.
autrijus you mentioned that %INC for Test.pm should key on 'Test'
gaal Class:Test
broquaint Consistent!
autrijus class:Test -- it's fine by me.
p6l p6l :)
(but it's really not test, and not module, and not class)
it's rather a file. 09:13
broquaint It should key on an INC object, or some such.
autrijus require does not deal with packages etc.
only "use" does
I'm fine to have "use Test" record Test.pm in %INC
gaal the *actual* file is in the value of %INC (as i know you know :)(
autrijus as well as module:Test
gaal okay, i'll get back to this, i want to figure out this too-fast smoke business first. 09:14
what might be causing this output? 09:15
pugs.exe: 1
autrijus not sure 09:16
it's an user error using 'error' call
gaal autrijus, can you review my %*INC patch (opRequire) from earlier today? 09:17
i'm returning VInt 1 if a module has already been used
so i fear this may be the same 1.
masak where can one read more about the perl6 semantics of BEGIN et al.? 09:24
svnbot6 r5472, autrijus++ | * remove dead code.
autrijus gaal: url? 09:25
masak: hw2005 has some writeup
gaal it's committed as r5466 09:26
# rt.openfoundry.org/Foundry/Project/...;rev2=5466 09:27
autrijus gaal: worksforme 09:29
gaal it worked for me too, but i though it might have been triggering those errors with the precomp Test 09:30
masak autrijus: thx 09:31
autrijus masak: np 09:32
gaal: hrm, if you change it to return vint 2, see if it fails with 2
I doubt it is the case
gaal doing just that :) but build is slowww
autrijus unoptimised too? 09:33
gaal i got segfaults trying to mix them
and i'm shooting for an eventual benchmark of the smoke, so i did need an optimized pugs... 09:34
autrijus nodnod 09:36
gaal like i said, Pugs.Run is so expensive to compile now I'm not sure it's even worth it to precom Test. :( 09:38
autrijus it makes more sense if we produce Test.so.
gaal yeah 09:39
autrijus which is actually not very hard
gaal but then we get into hairy platform dependencies.
autrijus only to make it more portable, Test.pil is probably what we need to shoot for
heh, jinx
gaal yup
:)
autrijus and worse case we can just derive Read
and read in the PIL expressions
gaal won't they be very large? PreludePC.hs is huge already. 09:40
autrijus that is likely Fast Enough. if not, we can use DriFT to derive Binary Read
akin to Storable's thaw
gaal jinx again, only you know the stuff better
autrijus but I think that's a good idea overall :)
probably not push Test.pm precomp for 6.2.8 then
but please still commit the crippled -- I mean workaround -- Test.pm
I'd like to look at it and see how much I can preserve for pirsmoke 09:41
gaal what do you know, it *is* my %*INC guard.
autrijus ooh.
(bbiab)
gaal why would returning a value trigger an error?
autrijus: perlcabal.org/~gaal/Prelude.pm 09:43
i was in q&d mode, so this has the modified Test tacked on to the Prelude.
autrijus can you sep it out and commit it back to ext? 09:44
ext/Test/lib/Test.pm
that makes it easier for me to diff and work with chunks
gaal if this is deemed worthy, i can write an evil mangler^W^Wnefarious mutilator to generate a unified Prelude.pm from Test.pm and the real prelude
sure.
autrijus hm, sort of like perl5's prelink build option 09:45
for platforms without dyld
I think it should not be the default, but having it available is a good thing 09:46
(exactly the same as p5)
gaal old wheels, new cars
s/cars/ships/
autrijus yup :) 09:47
gaal who needs wheels on a ship?
autrijus exactly
gaal works on getting the hawser in shape
autrijus there is a pirate joke in there somewhere but I will refrain from uttering it 09:48
gaal arrr.
there was a problem saying "use" from within a lexical scope, right? was it fixed? 09:49
because i'd omitted 'use v6' in my version out of paranoia
autrijus I don't think that matters. 09:51
gaal r5473. 09:52
brb 09:54
svnbot6 r5473, gaal++ | prelodability mutilations
autrijus I think I finally fixed export.t
testing
gaal [off-topic] btw, autrijus, when does par use /tmp/par-SYSTEM for a cache? people showed me a case at $work where cgis that used par (don't ask) used that. 09:59
autrijus SYSTEM is usd on win9x and other cases where I can't determine $USER. 10:00
you can turn that behaviour off with pp -C or runtime with PAR_GLOBAL_CLEAN
gaal when might that happen on unix? when /etc/passwd was modified under your feet? 10:01
autrijus huh. 10:02
$ENV{USER} has nothing to do with /etc/passwd
gaal oh, env! all right.
thanks. :)
autrijus np :)
autrijus lols at Installing /usr/local/lib/perl6/site_perl/mach/CORE/pugs/pge/parrot.core 10:06
svnbot6 r5474, autrijus++ | * Allow exporting into non-main packages; export.t finally fixed. :)
nothingmuch morning 10:14
svnbot6 r5475, autrijus++ | * more TODOification 10:15
r5476, autrijus++ | * fix typo in test.
nothingmuch autrijus: i need help with continuations, and your MMD
autrijus my MMD? 10:16
nothingmuch the current implementation in puts
pugs
autrijus oh. the current implementation is not at all scientific 10:17
what help do yo need with cont?
nothingmuch okay, 10:18
look in t/statements/{next,redo}.t
we have some horrible bugs
next and redo behave like calls instead of gotos: the statements right after them in the dynamic scope are returned to
i've fixed next in while 10:19
but it's pretty ugly
i think we should generalize them so there is a 'generate a sym that wil goto this code', a bit more specific than genSymCC
autrijus ah. I can see why.
yes. 10:20
svnbot6 r5477, autrijus++ | * now export is working, fix the misimporting test
r5478, autrijus++ | * instead of :todo(1), please use :todo<bug>
autrijus that'd be the right direction
nothingmuch also look at Monads.hs and tell me if my fix is valid
autrijus it is valid. 10:23
nothingmuch yumm
I had no idea what it was doing, so i was guessing 'undef' is the expression evaluated to make a return value for 'last' 10:24
autrijus yes. 10:25
"esc undef" is basically shifting out to outer dynamic scope
and push undef as the reply 10:26
the inner dynamic scope is then thrown away
nothingmuch okay
goody
autrijus a dynamic scope is deliminated by resetT
nothingmuch could you briefly explain Cont in haskell? 10:29
"dynamic scope" means the caller continuation stack maintained by the Cont monad?
autrijus yes... it is unrelated to envCaller 10:30
and the "callCC" works inside a stack frame
nothingmuch we just sync them because it's convenient?
autrijus yeah
callCC snapshots the flow in a stack frame
nothingmuch callCC replaces the last stackframe with a certain action?
autrijus and when invoked it resumes execution from there
yes 10:31
nothingmuch when who is invoked?
autrijus when the cc is invoked
look at its type
((a -> m b) -> m a) -> m a
it is a "Eval a"
that demands one argument 10:32
the argument is a code that gets the CC
if it finishes normally without invoking the CC
then the resulting (m a) is the final (m a)
if it invokes CC with a
-- which may be done in any position -- it does not ever return anyway
(hence the m b)
nothingmuch okay
autrijus then it escapes out with the a as the final (m a) 10:33
nothingmuch and (a -> m b) is the basically the the point to continue on
that is the thing right after $?CALLER::POSISITION if there was one in haskell?
autrijus yes I think that is correct.
nothingmuch ok, how would you refactor it? 10:39
autrijus sorry; I'm not currently thinking about it. 10:40
I just got pirsmoke working again
autrijus is firmly in relenging mode
so I'd be happy to answer questions, but more high level thinking needs to wait until 6.2.8
(I apologize for the inconvenience) 10:41
gaal autrijus, i'm not working on Test.pm now (out of ideas re: the error). are we punting on that?
autrijus gaal: yes, punting on precomp Test
gaal ok
autrijus I doubt the gain anyway
gaal me too
autrijus good we agree :)
nothingmuch autrijus: so should I just repeat what I did? 10:42
autrijus nothingmuch: that would be good
svnbot6 r5479, autrijus++ | * more unTODO and TODOify
r5480, autrijus++ | * `make pirsmoke` revived.
nothingmuch okay
gaal i just hoped to be able to *measure* it :) but ther's no rush.
nothingmuch can I do what I did in Monads outside of Monads?
how is line 380 of Eval.hs (genSymCC "&next") different from any &last? 10:44
autrijus you can do it in Eval.hs directly 10:45
line 380 captures the point where the next line is runBody 10:46
so people invoking symNext will then run that.
hence, next
nothingmuch oh, i see
theorbtwo It doesn't, it just happens at a different time, IOW. 10:47
nothingmuch so how does &last get defined?
i.e., how is there a symbol
theorbtwo Which is good, because next doesn't differ from last, except where they go.
nothingmuch if it happens later?
or is 'runBody res sub'' not executed at all?
oh, whoops 10:48
autrijus I hoped I don't have to explain that :)
(the genSymCC for &last is the end of the reduction.)
nothingmuch i was reading it incorrectly, nevermind
i got it now
autrijus cool
nothingmuch tries to see how todo might be written 10:49
s/todo/redo/
autrijus I need to run to dinner. down to 8 failed testfiles now 10:51
(in the main suite)
gaal wow
autrijus hopes to crush them during dinner
Tree is still burning, as is Date and DateTime 10:52
spare tuits to look at them welcome
nothingmuch autrijus: i looked at date/datetime
the problem is that MMD is selecting the wrong BUILD 10:53
or at least a problem
i tried to hack your MMD
but didn't get far
autrijus if they are beyond repair for the moment, add them to MANIFEST.SKIP please
nothingmuch i don't think so
theorbtwo Better: Add a test for the underlying problem, and put them both in MANIFEST.SKIP.
autrijus right.
nothingmuch the case is that BUILD(Str $string) is somehow bound to new(epoch => 123); 10:54
autrijus er, the test for underlying problem can be a dies_ok
nothingmuch even though there is a BUILD (Int|Real +$epoch) {
gaal note that i hacked on Test.pm - I may have introduced a bug
nothingmuch when Str is commented out, then submethod BUILD () is used instead. 10:55
gaal unlikely since 7000 tests still pass, but still :)
nothingmuch gaal: what did yo8u hack?
gaal workarounds to allow it to be precompiled 10:56
jaggers guys sorry for the interruption , i need a bit of help
autrijus Time::Local can't numify anymore? hrm
gaal they can be reverted, in a pinch -- we're not precompiling for this release.
autrijus it should have numified to the original time() no? or am I mistaken
jaggers i ve been to trying to connect to DALnet , but not successful , i was able to connect thru win XP , but not able to connect thru fedora 4 10:57
nothingmuch autrijus: anyway, should (epoch => 123) bind to (Str $string
gaal autrijus, did Time::Local ever numify?
nothingmuch ) at all?
autrijus gaal: hrm, t/builtins/io/time.t tested something weird
svnbot6 r5481, autrijus++ | * fix borken test due to nonnumiability of Time::Local 10:58
autrijus nothingmuch: no, it should decidedly fail
ok, two more fails bite the dust.
really need to run for dinner now. I'll be back in (say) a hour
autrijus waves &
gaal jaggers, wrong channel?
jaggers i dunno 10:59
Aankhen`` gaal >> You can't just use a regular, older version of Test.pm?
jaggers this is wat is keeps saying GAAL
* Looking up irc.dal.net
* Connecting to irc.dal.net (194.68.45.50) port 6667...
* Connection failed. Error: Connection timed out
Cycling to next server in DALnet...
* Disconnected ().
gaal jaggers: this is for perl6 development, i don't think we can help you
jaggers am sorry gaal 11:00
gaal Aankhen``: sure, i was saying, we can revert if necessary
Aankhen`` Ah.
Right, I missed that. :-)
nothingmuch autrijus: how do i comkpose the return of genSymPrim and genSymCC? symRedo . symNext doesn't work anymore
gaal jaggers, no problem
nothingmuch do I need to call updateSubPad N times?
Juerd wonders what lwall has been smoking. 11:01
lightstep blue sugar? 11:02
wilx DALnet isn't dead? 11:03
svnbot6 r5482, autrijus++ | * restore PUGS_VERSION.
jaggers no i cud connect to DALnet thru win XP , while iam not able to connect from fedroa 4 11:04
Aankhen`` Juerd >> What prompted that wondering?
Juerd Aankhen``: .foo will be forbidden whenever a topic is used in a method.
that is: method foo { given $something { .foo } } will be a compile time error 11:05
Aankhen`` Ah, that threw me off too.
Juerd See also my post, of which the subject matches /WTF/
Aankhen`` That should be interesting. :-)
Juerd I'm amazed and very upset at the same time. This time, I really did lose confidence in Larry.
I had anticipated a lot, but not this. 11:06
nothingmuch oh man!
that sucks 11:07
nothingmuch wrote a lot of code that uses ./foo and .foo to disambiguate with *HEAVY* (read 2-3 layer nested) topicalization
and it was nice code to write
not confusing
Aankhen`` I used it a lot too... not 2-3 nested layers, but enough that this sucks. 11:08
svnbot6 r5483, iblech++ | * Finished the move of pugsbugs/export.t to packages/export.t and added another test.
r5483, iblech++ | * Usual svn props added to new t/statements/redo.t.
Aankhen`` Juerd++ # for writing it up :-) 11:09
Juerd Please let your opinion be heard on the mailinglist as well
If only an AOL-ese response. 11:10
nothingmuch done 11:11
Juerd goes to collect some cash from a client 11:12
His time is up.
Bye 11:13
nothingmuch good luck
Aankhen`` See ya. 11:14
Aankhen`` 's "me too" post has been sent. 11:15
lightstep what's with the tendency to preform monadic actions in procedures? like genSymPrim doesn't return m(Pad->Pad), or ifListContext doesn't return (m Bool) ? 11:16
what about separation of concerns? 11:17
nothingmuch lightstep: please separate concerns... that code is confusing me =)
svnbot6 r5484, Aankhen++ | * t/data_types/enum.t: added a few tests, modernised some of the older tests. 11:32
r5485, Aankhen++ | * t/data_types/enum.t: added 'hex' enum.
nothingmuch autrijus: 11:34
eep
gaal perlcabal.org/~gaal/smoke_xp.html # r5466 11:36
theorbtwo wonders what bonus: 0 means in the tests.yml file. 11:42
svnbot6 r5486, Aankhen++ | * t/data_types/enum.t: fixed a few mistakes in the previous commit related to quoted numbers.
gaal more tests run than planned, iirc 11:43
nothingmuch autrijus: help
pasteling "nothingmuch" at 212.143.91.217 pasted "code that makes infinite loops" (27 lines, 1.2K) at sial.org/pbot/11709 11:45
nothingmuch my impl of for is looping into the redo continuation infinitely
why is that?
svnbot6 r5487, iblech++ | pugs::run -- Added entry about %*ENV<PUGS_COMPILE_PRELUDE>. autrijus: Please 11:54
r5487, iblech++ | fill in the XXX, thanks :)
r5488, nothingmuch++ | Ugly fix for redo in for loops 12:01
nothingmuch genSymCC needs to be redone 12:02
it's no good when there are more than 2 escapes involved
iblech--; # made my recompilation cycle slow 12:06
Limbic_Region shakes his head in disbelief 12:08
nothingmuch lightste1:
Limbic_Region: ?
Limbic_Region nothingmuch - did you follow the GHC perl.exe conversation I was having yesterday? 12:09
nothingmuch no
fill me in?
Limbic_Region long story short then
ghc ships with a perl binary at the root install directory, windows is broken with regards to . and PATH so if you are there you get GHC's perl, if you remove it you won't know it is a problem right away and even then the error message is too cryptic to know what the problem was 12:10
so I recommended 3 things that could make it more user friendly - rename the binary, improve the error message, have it look at %PATH as a backup 12:11
nothingmuch ghc ships with perl? wtf?
Limbic_Region well - after a long series of emails trying to convince each other of the validity of our respective perspectives
theorbtwo Probably because of the evil glasgow haskell compiler, what ghc uses to bootstrap with. 12:12
nothingmuch heh
Limbic_Region I got an email from a guy at microsoft saying that Windows is broken WRT to . and PATH and that renaming the exe and documenting the fact that changing the root install directory is bad would be a good thing
nothingmuch - if you are really interested in why GHC ships with a perl binary - grep yesterday's logs - autrijus linked to relavent sources IIRC 12:13
Limbic_Region was just shaking his head in disbelief of what he saw coming from a microsoft guy - in writing - from his microsoft account
nothingmuch pok^H^H^Hk^H^Hnah
i can't type today
that was the weirdes typo ever
lightste1 he's at MS research, not in MS evil empire 12:14
Limbic_Region simon marlow not the other simon - but I suspect you are still correct lightste1 12:15
12:15 lightste1 is now known as lightstep
nothingmuch fixses tricksy bugses 12:20
anyone feel like untodoing? 12:22
Aankhen`` Are nested `gather/take`s allowed? 12:23
svnbot6 r5489, nothingmuch++ | Same ugly fix for next/redo in loop(;;){} loops
Aankhen`` tries it.
nothingmuch wait, more on the way
theorbtwo Should be allowed, I think.
Might not presently work, though.
nothingmuch Aankhen``, theorbtwo: i think it does work
luqui and I used it in the pil -> p5 12:24
Aankhen`` Doesn't seem to.
Hrm.
Pugs is fabricating array elements.
pugs> gather { take "a"; take (gather { take "JUMPED IN!"; take "AGAIN!" }); tak
e "c"; };
('a', 'b', 'c', 'd')
nothingmuch okay, then it did work ;-)
Aankhen`` Am I missing something?
nothingmuch oh wait, gather take, not given when 12:25
nothingmuch idiot
Aankhen`` `gather { take "foo" }` is like `my @ret = (); push @ret, "foo";`, right?
svnbot6 r5490, nothingmuch++ | Fix redo test and implementation of next for loop(;;){}
r5491, nothingmuch++ | redo fixed for while loops... code needs cleanup
QtPlatypus Aankhen``: Yes. 12:26
theorbtwo Roughly, but with extra added funky lazy goodness.
QtPlatypus theorbtwo: In a lazy context.
Aankhen`` Indeed.
pugs> gather { take "a"; take "b"; } 12:27
('a', 'b')
svnbot6 r5492, autrijus++ | * Fix the broken handling of %INC -- the blind `VInt 1` 12:38
r5492, autrijus++ | broke `use Foo; use Foo;` because the second one does
r5492, autrijus++ | not return a VControl ControlEnv as it should.
r5493, autrijus++ | * identity test for functions should now pass; Eq and Ord
r5493, autrijus++ | for function and dynamics redefined to base on 'show'.
r5494, autrijus++ | * more TODOify.
autrijus two left: export, examples 12:40
rafl is working on having lib/* installing to sane paths
(maybe list them as subdirs?)
I'm still fixing export #2 and examples/quicksort; after that I'll clean up modules. if lib/ installation is done by then we'll enter preflight. 12:41
nothingmuch autrijus: before preflight, i'd like you to look at Date vs MMD with me 12:42
i think the problem is simple
but I don't know
either way, all next/last/redo statements are contized now
autrijus nothingmuch++ # wonderful, seriously 12:44
svnbot6 r5495, autrijus++ | * align signature_matching's todoity with reality
autrijus nothingmuch: can you reduce the problem to something in the main test suite?
nothingmuch i'm not 100% sure
i'm having trouble actually diagnosing the dispatch problem 12:45
autrijus please try anyway; I'll have tuit to look at that after I fix export and quicksort
nothingmuch okay
autrijus mm backlash in p6l 12:46
autrijus steps back very slowly
nothingmuch the .foo forbidden thread? 12:47
autrijus yup 12:48
broquaint Ignore p6l, it's just noise.
autrijus broquaint: sometimes noise makes into synopsis :)
nothingmuch food! 12:49
broquaint Good idea!
broquaint away & # insert pie into piehole 12:50
castaway bah, me hungry 12:51
Limbic_Region looks around for bah - doesn't see them and says nice to meet you hungry - me L~R 12:54
castaway pokes Limbic_Region 12:55
nothingmuch swings from a vine, and saves Limbic_Region from hungry's poking
castaway this getting a visa thing is turning out to be more complicated than we thought :(
nothingmuch me tarazan, you jean husband
Limbic_Region rewards nothingmuch's gallantry with a bananna 12:56
nothingmuch is offended by Limbic_Region's obcenety 12:57
Limbic_Region WRT the .method acting on $self thread - is the OO code currently written still valid
I really should have been following that thread closer 12:59
autrijus Limbic_Region: glad to hear spj resolved the bug 13:09
calling it ghc-perl.exe would be good enough
or was it jeffacake?
*JaffaCake 13:10
svnbot6 r5496, autrijus++ | * JeffaCake notes that there is now Linux x86_64 GHC snapshots, 13:15
r5496, autrijus++ | and we should reflect this in INSTALL.
Limbic_Region it was spj that noted clearer docs and renaming the exe were no effort things that could make it easier on the next person 13:26
autrijus Limbic_Region: cool.
Limbic_Region I feel really good about the exchange since the original reaction was - you really shouldn't be doing that and I can't think of a good reason why anyone else would 13:27
autrijus - fwiw, there is a "should I develop in p5 or wait till p6" thread at the monastery - I can post link if interested 13:28
autrijus sure, url?
stevan autrijus: is Tree still burning?
autrijus stevan: yes
stevan morning BTW :)
autrijus: I think I did that, so I will go fix :) 13:29
autrijus please do :) it's something about Tree::BUILD
stevan ah
ok
I didnt do *that*, but I will fix none the less :)
Limbic_Region perlmonks.org/index.pl?node_id=474125 13:30
stevan wonders why more people arent using Pugs in production :) 13:31
integral too slow 13:32
autrijus indeed.
stevan bah! good things come to those who wait :) 13:33
autrijus we really need PIL->P5 to convince the production crowd of adopters
QtPlatypus Keeps changing.
integral and I don't trust parrot, and PIR output isn't good enough yet
Limbic_Region I would be less concerned about the slowness factor than the stability
autrijus QtPlatypus: strangely though, there are very few regressions in each release
Limbic_Region having correct code that is only going to get faster wouldn't be a big concern for me 13:34
autrijus but "keeps getting new and/or previously unspecced stuff" indeed
Limbic_Region but having code that may be b0rk in two days because the wind changes directions would really p!$$ me off
I would likely have fun with refactoring previously stable code after new and exiting features were added 13:35
autrijus nodnod.
Limbic_Region but having to deal with all the changes all at once would just be too much
lumi The Lilypond folk have a version porter 13:36
gaal interesting thing, with the %*INC bug. i thought the typechecker would refuse to accept such things :) 13:44
autrijus gaal: it's not a typechecker thing
gaal: VControl ControlEnv is still a Val.
gaal well, i suppose that means i can get back to experimenting with precomp Test. i have a 4596 smoking now, though i did forget to look at the build start time. 13:45
autrijus gaal: the smoke report has a "duration"key. 13:46
it's listed at top of html
I should know, I asked for that :)
gaal yes, i know (i added it) but it only counts the smoke itself.
Limbic_Region hmmm - given my limited understanding of Win32 architecture - I find this behavior a bit odd
autrijus heh, right, and I was about to say you implemented it iirc
Limbic_Region I am in the middle of building Pugs and I decide I want to abort, svn up, and make clean 13:47
I hit ctrl-C and get the prompt back
gaal i'll time the second pass.
Limbic_Region I type exit - and it just sits there
I check the process table and GHC is still running
gaal l~r: similar things happen to me too
Limbic_Region shouldn't GHC have died when I killed the make process since Win32 doesn't have a real fork/exect ?
autrijus it means it uses CreateProcess
but failed to capture SIGINT 13:48
Limbic_Region which for all intents and purposes is the same thing?
autrijus I think it's a GHC issue.
Limbic_Region heh - not going there autrijus
autrijus JaffaCake is still in #haskell -- maybe ask him there :)
Limbic_Region I am 1 and 0 13:49
autrijus (== simon marlow)
gaal what i've seen is that ^Cing the tests sometimes takes some time to kick in.
svnbot6 r5497, Stevan++ | Tree - putting out the fire (i think)
Limbic_Region wants to retire a winner
autrijus lol
Limbic_Region anyway - I have started another build on trunk so I will be AFK for about 45 mins
Limbic_Region AFK &
autrijus ok, quicksort is the only remaining main suite failure 13:50
svnbot6 r5498, autrijus++ | * export.t -- putting out the fire (I think) 13:54
r5499, Stevan++ | Tree - pouring more water on the fire .oO( <smoke> )
castaway laughs 13:55
gaal are you getting failures on pugsrun/09- and 10- ? win32 13:59
autrijus iblech++ # creative use of pzero to parse `multi foo { }`
svnbot6 r5500, Stevan++ | Tree - .oO( *cough* *cough* )Oo. 14:00
r5501, Stevan++ | Tree - only *you* can prevent forest fires :) 14:05
r5502, autrijus++ | * reflect the `multi foo` shorthand for `multi sub foo` in quicksort
r5502, autrijus++ | and in ChangeLog
r5503, autrijus++ | * add minimal test case for the quicksort testfail in multi_sub.t.
r5504, autrijus++ | * avoid duplicate #define in Pugs.Version
r5505, rafl++ | Added Makefile.PL for Inline::Pugs.
14:06 wilx` is now known as iwlx, iwlx is now known as wilx
stevan autrijus: the Tree fire is out 14:07
geoffb Morning, all . . .
stevan looked like it had to do with the BUILD MMD
autrijus wonderful.
stevan morning geoffb :)
autrijus Perl::MetaModel is still burning 14:08
rafl_ Hm. I try to rename lib/Inline to lib/Inline-Pugs, but svk gives me: Merging back to mirror source svn.openfoundry.org/pugs.
Commit message saved in svk-commitTkN4a.tmp.
Why?
autrijus I'm rapidly thinking we should move docs to notes and ditch all code in ext/
stevan autrijus: that's mugwumps
gaal pcTest build phase near completion.
stevan I dont think it is meant to be executable
autrijus unless mugwump is willing to maintain the p6 equiv to the p5 one in lib
in which case it should be made to pass tests
stevan autrijus: we can move lib/ into docs/
as they are really just notes IMO
autrijus mugwump: ideas? 14:09
rafl_: hm? curious.
rafl_: but that'd be wrong still
stevan also questions the sanity of the contents of docs/* now
rafl_ autrijus: Why?
autrijus better to mkdir lib/Inline-Pugs/lib/Inline/
and then move Pugs.pm there
rafl_ autrijus: OK. 14:10
nothingmuch moose! 14:12
svnbot6 r5506, Stevan++ | Tree - oh no! it still burning <stomp> <stomp>
nothingmuch stevan: i'm stuck with haskell, so wanna do sigs?
clkao rafl_: hmm? why would you see that merge back thing when you do mv? or you mean duiring the commmit? 14:13
castaway grins at the svn messages
rafl_ clkao: Yes.
stevan nothingmuch: sure, but I will want your input 14:14
autrijus also, we have conflicting Date.pm in DateTime and Date
stevan castaway: :)
nothingmuch no, i mean want me to do thjem?
and help out a bit?
MMD in pugs is a bit too hairy for me
autrijus which means we clearly need to work on multiname library support
rafl_ Hm. I don't get it. How can I reset my local svk copy to that version on the server? 14:15
nothingmuch rafl_: rm -rf...
autrijus but seeing Date.pm is all stub, maybe I'll just MANIFEST.SKIP it for the release.
stevan nothingmuch: ok, but I have to do $work today too
autrijus (Date.pm in ext/Date/, that is)
nothingmuch stevan: ok
autrijus mugwump: ideas?
nothingmuch i think we should namespace it a bit like CPAN 14:17
QtPlatypus How so nothingmuch ?
svnbot6 r5507, autrijus++ | * DateTime - remove the erroneous use of colon
nothingmuch lib/AUTHOR/Module-Name-0.01/Module/Name.pm
autrijus: now that you've removed the colon from datetime, there are MMD match errors
with AUTHOR not being only uppercase 14:18
and not being limited to 8 chars
maybe lib/AUTHOR/Module-Name-0.01/lib/Module/Name.pm so that it's a bit more extensible 14:19
btw, since we're in preflight, smoke report just finished
Khisanth hmm
gaal we're in preflight? 14:20
nothingmuch it sounds like it
gaal i'll have pcTest smoke results up in a bit.
nothingmuch burning trees, everyone wading through bugs
seen geoffb 14:21
castaway geoffb went to bed earlier (several hours ago?)
nothingmuch perlbot: seen geoffb?
ah, ok
Limbic_Region seen geoffb
nothingmuch Limbic_Region: jabbot isn't here
Limbic_Region ahh - perlbot doesn't keep track 14:22
gaal i'm tempted to add a progress indicator to the harness :)
geoffb Um, right here
stevan I just saw geoffb about 20 min ago
geoffb what's up guys?
stevan there he is,... get em!!!
nothingmuch geoffb: all of your bugs are belong to past
do you have any more tests?
gaal stevan++ # nothing special, i just like your attitude
QtPlatypus geoffb was last seen 15 minits ago. 14:23
castaway scratches head.. 8short sleeper?)
stevan gaal++ # coffee is goooooddddd
geoffb nothingmuch, they're all fixed? Schweet
nothingmuch gaal: while you're at it, make the parallel harness truely async
geoffb castaway, yeah, only one sleep cycle last night, instead of two . . . :-/
gaal suuuure
nothingmuch parent prints test file names (padded) to pipe
autrijus we're not yet in preflight because I'm quickly running out of sanity :)
gaal nothingmuch: if it were written in p6, i'd consider it
nothingmuch child processes sysread them, and unpack("Z*")
gaal but it has to work portably. 14:24
geoffb nothingmuch, I'm still reading backlog ATM . . . and last night I put all my energy into docs/notes/plan before falling into bed
gaal okay, the results are in:
pcTest is indeed much faster to run (duh!)
autrijus question is how much and whether it makes up for the comp time
gaal 782 sec instead of 1368
nothingmuch autrijus: are you wokring on datetime?
gaal however, the 2nd compilation phase is about five minutes on my machine 14:25
autrijus nothingmuch: no, I just saved it from parsefail
nothingmuch autrijus: are you planning on working on it?
autrijus gaal: so that's still 5 mins saved
nothingmuch: my plan doesn't matter... my body will force me into blissful sleep in 10 minutes 14:26
so not until I wake up, no
gaal and of course, we already pay some of those 5 minutes on account of the pcPrelude
i can get the exact delta if it's interesting.
autrijus gaal: if it can be made optional, I think definitely check it in
geoffb gaal, but pcPrelude itself already saves net time, yes? 14:27
autrijus "it" being the evil mangler thing you referred to
gaal it can of course be made so, but i don't want to bloat the build system more
geoffb, sure it does.
geoffb I just didn't understand your "we already pay some of those 5 minutes" comment 14:28
gaal i like evil manglers, but i don't like having too many knobs
svnbot6 r5508, rafl++ | * Moved lib/Inline to lib/Inline-Pugs.
r5509, autrijus++ | * untodo the now passing next/redo tests, nothingmuch++
autrijus gaal: bah, the build system can survive another bloat. :) but it's ultimately your call
geoffb: the Run.hs 5min build includes both Prelude.pm and Test.pm
rafl_ Can u push the Version of Perl6/Pugs.pm to 6.002008?
gaal here's my proposal:
autrijus rafl_: you can do that.
geoffb AH, gotcha, thanks autrijus
gaal put all those build options in a yaml file
including whether to optimize or not
geoffb clearly short on sanity from lack of sleep
gaal and pcPrelude or nor 14:29
nothingmuch and ghc heap size
gaal and all other things
and nothingmuch heap size
errr
how about that?
nothingmuch i like it
geoffb All in favor say aye? Aye.
nothingmuch aye
autrijus I don't particularly like to depend on YAML.pm :) 14:30
but I can see a minimal yaml parser in two lines
so, aye.
gaal split /:/ :-)
autrijus but please leave nothingmuch's heap size out of it
nothingmuch what's bad about a YAML dependency? 14:31
gaal lol
i had a feeling you'd say that
stevan we already have a YAML dep for the smoker
autrijus nothingmuch: it will increase our dependency list by +Inf times
(it's currently 0)
gaal err, not so, autrijus, we already need a newish MM 14:32
nothingmuch yaml's meta.yml:
build_requires:
Test::More: 0.54
requires:
perl: 5.005003
geoffb dependency list 0? bwa-ha-ha-ha-haaaa . . .
geoffb remembers the day spent finding and building various haskell dependencies for pugs . . . . 14:33
autrijus gaal: er, that means 5.6.2 or 5.8.1
gaal: hardly newish
stevan ... "I get by with a little help from my friends... gonna try with a little help from my friends"
autrijus geoffb: CPAN dependency list, silly :)
gaal well, fwiw, that means you can't build on a vanilla msys.
geoffb autrijus, I knew that, it just hit my funny bone
gaal not that i know of anyone who still does that
autrijus but anyway. I'd like to avoid depending on a noncore 5.8.1 module as much as possible. 14:34
gaal but maybe that proves my point
anyway:
nothingmuch YAML has almost no deps
gaal i'm fine with parsing yaml ourselves
minimally
kungfuftr JSON!
autrijus cool. 14:35
gaal ?
autrijus gaal++ # make it so then :)
JSON is a weird YAML dialect that is somehow welcomed by javascript folks
gaal okay, i think i'll phone gabor and pick up the wiki router for the hackathon first.
rafl_ Hm. What about P5-PIL-Run? What is it needed? Should it be installed? Looks suspicious to me. 14:36
kungfuftr JSON isn't YAML
clkao rafl_: ah, you just hit a bug which is fixed in 1.01 i guess.
autrijus actually, all valid JSON is valid YAML, iirc.
rafl_ clkao: What bug?
geoffb hmmm . . . why not just include YAML in inc/?
stevan rafl_: that should not be installed, and neither should Perl6-MetaModel
kungfuftr autrijus: JSON can be a single line
autrijus so can yaml.
kungfuftr and be a huge data structure
rafl_ clkao: Oh, you mean the svk thing? Yes. Maybe. It works now.
clkao rafl_: mv tries to do funky things
autrijus ditto.
kungfuftr *blink*
rafl_ stevan: OK.
kungfuftr 2 secs
gaal it doesn't say "yaml" anywhere on www.crockford.com/JSON/index.html -- but i don't know if that's the official page 14:37
kungfuftr www.crockford.com/JSON/index.html ?
autrijus kungfuftr: and YAML.pm is horribly outdated and broken. :)
yes
gaal kff: 1st google hit
stevan rafl_: FYI, P5-PIL-Run is the start of the Perl 5 runtime for PIL/Perl6
autrijus kungfuftr: redhanded.hobix.com/inspect/yamlIsJson.html
kungfuftr ah!
autrijus all other languages get to use a nonbroken YAML implementation 14:38
except for Perl5 :)
svnbot6 r5510, rafl++ | * Moved lib/Perl6 to lib/Perl6-Pugs.
r5510, rafl++ | * Bumped up version of Pugs.pm to 6.002008.
r5510, rafl++ | * Moved lib/pugs* to lib/Perl6/Pugs.
r5510, rafl++ | * Added Makefile.PL for lib/Perl6/Pugs
r5511, autrijus++ | * more TODO juggling in light of the namespace fixes.
kungfuftr so you can use the JSON parser instead? 14:39
gaal is yaml.pm core in any versio of p5? 14:40
autrijus gaal: no.
kungfuftr: no... Syck can parse JSON; the reverse isn't neccessarily true
geoffb So, at the risk of repeating myself, why not just have a YAML parser in inc/? 14:41
Rather than writing from scratch?
kungfuftr autrijus: bah... soz... head's away with it... too warm here
autrijus geoffb: I guess 62k isn't much
it's pushing it a bit 14:42
and if all we are using is a single hash with no nestedness, then it's quite wasteful
geoffb It get 54 MB for the source tree in total . . . .
autrijus but should we ever need nested data or anything beyond a simple regex parser
geoffb fair enough
autrijus then sure, inc/YAML.pm it is. 14:43
er what? 54mb?
geoffb Oh, duh, minus 12.7 MB -- I still had a blib
autrijus er what?
are you counting .o files?
gaal geoffb, that's after a build, surely?
geoffb hold, on, making realclean
autrijus 5mb here. 14:44
geoffb oh, phew, only 10.5 MB after that
OK, so why am I still off by factor of 2?
autrijus I have no idea.
rafl_ Why do we have use lib 'lib' in Makefile.PL? I can't see anything from lib being used there. 14:45
integral .svn?
gaal df -k if you're on hpux :-P
geoffb integral, using svk
gaal (*du
geoffb I did du -cskx * |sort -n
autrijus rafl_: I didn't remember. try taking it off and see if it works still
gaal bbiab &
autrijus if yes take it out
rafl_ autrijus: Did that. It does. 14:46
autrijus anyway... 62k isn't much for a 5mb tree. iff it's needed, throw it to inc/.
rafl_: take it out then.
rafl_ autrijus: I replaced that with use lib 'util'. I had some problems in inc/Module/Install/Pugs.pm to use PugsConfig. 14:47
autrijus sure.
I can't stay up any longer. hack on, and see you tomorrow :)
theorbtwo Do you have .svn dirs in there?
autrijus waves & 14:48
geoffb bye, autrijus
svnbot6 r5512, bsmith++ | Makefile.PL - get version from lib/Perl6-Pugs/lib/Perl6/Pugs.pm and
r5512, bsmith++ | use lib 'util'.
rafl_ Hm. Is he in here? I'm also already working on that. 14:50
integral oh, yep I am
I've just found some more places too
rafl_ So, please wait a moment. I'm just trying if everything is working and will commit it then, OK? 14:51
integral ok, sorry. I was just trying to build to check my changes with HEAD
rafl_ Oh, OK. 14:52
What other places did you found? util/config_h.pl and Makefile.PL is everything I needed to change. 14:53
integral that's it, it's building fine atm
nothingmuch make: *** No rule to make target `lib/Perl6/Pugs.pm', needed by `src/Pugs/pugs_config.h'. Stop. 14:54
rafl_ nothingmuch: integral fixed that in 5512 I think.
integral that's the one I missed 14:55
nothingmuch i just synched
integral rafl_: are you going to commit the rest of the changes?
rafl_ integral: Yes. I'll do that now.
integral rafl_++ 14:56
svnbot6 r5513, rafl++ | * Build lib/ recursively, just like ext/. 15:02
r5513, rafl++ | * s#lib/Perl6/Pugs.pm#lib/Perl6-Pugs/$&# Makefile.PL util/config_h.pl
r5514, rafl++ | * Adjusted path for pugs.pod in debian/rules.
rafl_ Is build_subdirs also running the test for each subdir? 15:19
Who want's to fix utils/src_to_blib.pl? It installs everything, including temporary and object files and stuff. That's not really needed, is it? 15:30
What creates the contents of blib/ during the build? 15:37
Also p6 stuff seems to be put into blib/. Shoudn't that go into blib6? 15:38
gaal rafl_: (i'm not an expert on MM, but) see PM_TO_BLIB in Makefile
rafl_ Hm. Can I prevent Module::Install to look into lib to find something to install? 15:40
gaal sorry, dunno.
rafl_ ingy, autrijus: Do you know how to prevent Module::Install to look into lib/ to find its stuff to install? 15:41
(They're listed in AUTHORS, so I assume they know about that. 15:42
gaal at least one of them is asleep atm :)
svnbot6 r5515, putter++ | docs/notes/plan: added link to original form, and a note re it not really being an interview.
putter geoffb: Thanks for the rewrite. I appreciate the help. I have some mixed feelings about the result. Some information lost, some mistakes added. Better grammar, but also less dense. But nice things added. Anyway, so I added a link to the original, so folks have both, and a note on the liberties I took, which would have been inappropriate if it had really been an interview. Thanks again. 15:50
svnbot6 r5516, putter++ | notes/plan: added a note of context at top. As, being linked from autrijus's journal, many will see it rsn. 15:58
putter masak: "I am sorry for the length of my letter, 16:00
but I had not the time to write a short one."
[Blaise Pascal [1]]
www.vendian.org/envelope/dir1/cant_explain.html 16:01
tnx :)
geoffb++ 16:05
gaal what can i assume about make? is there a minimum standard for rules that { gmake, nmake, bsd make } agree to? i just want "if". 16:07
putter And, I forgot last night to say, thanks to mugwump++, geoffb++, stevan++, and anyone I missed, for your help collecting ideas for the note.
gaal: perl? 16:08
gaal hmm
i might actually do that.
how long can an environment variable typically be, on the worst of OSes? 16:10
is 100 chars still safe? 16:11
150? 250?
putter 10? ;)
lightstep enough to store PATH
gaal :-)
Limbic_Region Failed 6/355 test scripts, 98.31% okay. -2/6342 subtests failed, 100.03% okay.
lightstep that is, dozens of chars
Limbic_Region -2 tests failed - 100.03% ok
got to love those numbers
;-)
gaal how many dozens is the crux of the matter 16:12
dudley gaal: I think VMS has a limit of 255...
gaal i think i can live with that.
100 will probably be okay for what i need too, but 50 is iffy. 16:13
the idea: config.yml has key/value pairs. you set those with an editor or a Configure script (not included). 16:14
the pro is that there's a standard way to override those w/o going to the editor again 16:15
env PUGS_OPT=key:val make
(usually you don't need that at all) 16:17
putter re "standard way", but note that it is os specific. so ok for users, but needs to be wrapped by perl in makefiles. 16:19
gaal by os specific you mean that there's no env on windows, or something else? 16:20
dudley There's no env on VMS, either. I get the distinct feeling that I'm going to end up trying to port Perl6 to VMS someday... 16:24
gaal dudley: how do you pass args to make on vms? just command line? 16:25
dudley Gnu make barely works on vms. MMK is the standard build tool. 16:26
gaal the drive here is primarily to put all build preferences in one place; making that overridable from either the command line or the environment is very important but not crucial. anyway a vms is not for a while is it? :) 16:27
*a vms port
dudley Yeah, i was mostly joking.
I do need it to run on VMS eventually, though. 16:28
gaal beh, use ghc to crosscompile.
you didn't say you needed it to *build* there :)
dudley :-) 16:29
putter JeffaCake/autrijus re Linux x86_64 GHC snapshots... but do they actually work? the two I tried failed to make. (just a cautionary note while backlogging...) 16:30
dudley wanders off to find a bookstore &
putter & 16:31
Re inconvenience of changing p6 spec, I can't find the quote now, but Denis Richie(?) wrote how he realized having whitespace be significant (or space vs tab??) was a mistake a couple of weeks(?) after doing it (in 197x(?)), but at that point he had too many users (order-10 (?)) to fix it. 16:36
gaal Stuart Feldman, and it's a great story, but i can't find the quote either. 16:40
www.faqs.org/docs/artu/ch15s04.html 16:42
" And then a few weeks later I had a user population of about a dozen, most of them friends, and I didn't want to screw up my embedded base. The rest, sadly, is history." 16:43
putter :) thanks! 16:46
cognominal tab should not even be a character, more like a mecanographic device 16:47
gaal likes the librarian job 16:48
# www.livejournal.com/users/gaal/2002/03/14/
putter gaal: knobs, please knobs. (backloging) 5 min is deadly in a development cycle. way too much for "does it work _now_?". but precompiled speed great for "ok, think so, run make test". knobs.
gaal brb, $work phone 16:49
putter re OED balcony service :) :)... now if only it wasn't 1/4 way round the planet. 16:51
revdiablo hmm, is anyone else using {''} instead of <> with nested hashrefs, just to maintain symmetry? I mean, %hash{$foo}{$bar}<baz> just looks ugly to me. 16:53
gaal putter: i'm aiming for Standard Knobbage
so instead of vi SOMETHING ; env THIS make THAT 16:54
you either vi config.yml and make away,
or override via env 16:55
but hopefully the names stay the same
of course you should be able to change anything, but right now if you grep util for ENV you'll see there are too many disparate entities
re: the look-up service, i regret to say that although i'd love to, i cannot take email requests at this moment. 16:56
putter ;)
gaal revdiablo: %hash<<$foo>><<$bar>><<baz>> ? :-) 16:57
prolly not a good idea though.
revdiablo heh
gaal since they may contain spaces.
putter Have an OED around here someplace. Misplaced it. One volume, with teeny tiny print. Wish I had a monitor with print like that.
gaal well, i bet you didn't lose the magnifying glass that came with it -- those are huge 16:58
i actually have the full version here. snatched it in an inexplicable bargain.
they sold for a while at an insane discount, but the real insane thing was that they charged shipping as if it were one volume. 16:59
the whole deal was for under $400 (instead of $3000).
geoffb putter, noticed this window was blue, so have just read what you said to me (and nothing else in backlog yet) . . . sorry to have messed up your stuff, I thought I had captured most of what you had included. Sigh. 17:01
putter Oh hey, no problem. Appreciate the work. Especially since past experience has shown folks find my draft design space notes dense to the point of being incomprehensible. 17:03
gaal textwork is hard. team textwork is also hard. /me returns to configuration matters which are of course easy 17:04
geoffb Funny thing is, I debated just leaving the original intact at the bottom, and finally decided against that, because "Curious people can always look at the change history." putter had the better option -- drop the old text but add a link to the old revision. I'll remember that for next time. :-) 17:07
putter You did (capture most). Was good. With added link to old, readers in better state. gaal: :) 17:08
ah. interesting.
geoffb putter, also you mentioned earlier that I added some mistakes -- but a diff doesn't show you edited content stuff. Is there something that still needs to be changed?
putter let me see... 17:09
yeah... sigh. 17:10
do we have stats on how many people read autrijus's journal? 17:11
Limbic_Region isn't sure that's possible 17:12
geoffb Limbic_Region, a lower bound certainly is . . . .
(and I suspect that lower bound wouldn't be more than an order of magnitude off at worst) 17:13
gaal putter: no way to do that. it's syndicated.
Limbic_Region for non-anonymous readers - see use.perl.org/~autrijus/fans
gaal at least 8 users on LJ :)
putter ;)
gaal plus probably the planetperl aggregators? 17:14
Limbic_Region as I said - not sure its possible to know
gaal i'm sure it isn't. :)
geoffb assumed most of the other sites would be links, not wholesale syndication . . . bah 17:15
syndication--
Limbic_Region even if it weren't - it is still only an approximation to parse the logs looking for unique IPs
you have proxies that make many people look like 1
and with DHCP - 1 can look like many 17:16
geoffb Yeah, that's why I said "lower bound within an order of magnitude" . . .
and you wouldn't do unique IPs. For a daily journal, it would better be hits.
Limbic_Region ick - I check his journal several times a day 17:17
geoffb crazy bastard
Limbic_Region why?
I don't have an RSS feed so it is the only way to know if there are comments or updates
geoffb Can't use an RSS feed, you mean? I know use.perl generates various RSS feeds . . . . 17:18
Limbic_Region no
geoffb Comments I can see being a problem 17:19
Limbic_Region I don't have = I don't use
gaal bbiab
Limbic_Region see - I don't ever get anything done because I am too busy keeping up with what everyone else is doing
literally
geoffb I think it's time to get away from the keyboard. I'm just not mentally all there today. 17:20
Limbic_Region TTFN geoffb - enjoy
putter geoffb: me too. worry about plan later. ? 17:31
Aankhen`` (backlogging) Limbic_Region: Re ^C-ing the build process: try repeating it a few times. A similar thing happens to me when installing a module from CPAN: if I ^C it, it drops me back to the regular DOS prompt, but the commands are interpreted as if by the `cpan` shell, and I have to keep pressing ^C for a while. 17:42
`cpan` is actually a lot worse, since I'm running it, it runs `nmake`, and then `nmake` runs `cl` or whatever. 17:43
gaal alias kill kill -9 :-) 17:46
Aankhen`` Wow... Pugs really starts up fast now. o_O 18:13
gaal aankhen``: are you using the precomp test as well? i hand out free sneak previews if you want. :) 18:14
Aankhen`` No, I don't run `nmake test`. :-P
I was just running the interactive shell. 18:15
gaal that does make things faster. :)
Aankhen`` Hehe.
putter nothingmuch.woobling.org/pugs_test_status/ smoke has gone bogus. 27 test cases. 18:19
Aankhen`` putter++ geoffb++ # docs/notes/plan
svnbot6 r5517, iblech++ | * Usual svn props.
r5517, iblech++ | * Minor typo and grammatic fixes: INSTALL, gen_prelude.pl,
r5517, iblech++ | lib/Inline-Pugs/Makefile.PL, docs/notes/plan, ChangeLog.
r5517, iblech++ | * Renamed docs/other/SHOT_IN_THE_FOOT to docs/other/shot_in_the_foot, as all
r5517, iblech++ | other files in docs/other/ are all-lowercase, too (consistency).
r5518, Stevan++ | Perl6::MetaModel - changed all croak()s to confess()s (mmmm stack trace :); added Perl6::Container::Scalar as a (tied) stand in for real container types; added tests for it as well; next step is to actually use it in the MetaModel; 18:26
putter I thought make pirtest was working? With 6.4.1.yesterday, 0.2.2, and pugs head, my sanity is all dubious. 18:27
And everything else is Can't locate Test in @*INC.
Ovid OK, I think I'm a little irritated with Pugs right now. Specifically, I went ahead and did a "sudo make install" and just hosed my Perl installation due to many Perl 6 modules overwriting Perl 5 modules. Why did Pugs do that? Is this something I was just supposed to "know"? 18:33
putter Here is a project for someone with bits on feather - set up a pair of smoke and pirsmoke. Noone seems to be running a pirsmoke yet, and we'll need it, and on feather perhaps multple people keep smoking going reliably. 18:34
Ovid: thats _very_ not the right thing. I suspect no one is really testing make install.
gaal putter: i'll do that once i get tuits. 18:35
putter gaal++
Ovid Ah, jeez. I'm seriously screwed. I've got a ton of work to do and I have no idea how long it's going to take to rebuild my Perl install.
gaal Ovid: definitely a bug.
Ovid When I did the install, I got lines like Installing /usr/local/lib/perl5/site_perl/5.8.6/HTTP/Headers.pm 18:36
That looks normal, but that's a Perl 6 module which overwrote my Perl 5. :(
stevan anyone have any idea how to return a tied scalar from a subroutine and how have it do return tied($scalar)->FETCH ??? 18:39
svnbot6 r5519, Aankhen++ | * edited docs/notes/plan.
stevan (this is a Perl5 question relating to the perl6 metamodel :)
Khisanth hmm
Ovid: you didn't get a /usr/lib/perl6/ ? 18:40
Ovid No.
Khisanth hmm which version? 18:41
Ovid I'm going through the module list right now to see if I can automate the fix or if I have to wipe out Perl and start over (not very easy)
Khisanth create a bundle and force install?
Ovid Khisanth: it's overwriting 5.8.6: Installing /usr/local/lib/perl5/site_perl/5.8.6/Test/Builder.pm 18:42
Hmm, a bundle might work.
A bundle won't work. Core modules that cpan and cpanp require have been overwritten. Even 'lib' was blown away :( 18:50
svnbot6 r5520, putter++ | Its quite a kludge, but this will stop anyone else from getting burned by make install until we figure out whats going on.
Aankhen`` I use `nmake install`; modules seem to install alright into the PREFIX directory. 18:51
gaal putter++ # stopgap
Aankhen`` Then again... EVERYTHING installs into the PREFIX directory, including lib/ (as-is). 18:52
putter I added an extra install: target to Makefile.PL. Intended to say "sorry, disabled", instead it causes a : vs :: conflict error. But that's hopefully good enough.
Ovid Aankhen: Regardless, this much destruction should not happen.
Aankhen`` Indeed...
putter Ok, my Makefile.PL patch did make make install safe... by completely breaking the Makefile. Regressed. So we still need a stopgap... 18:55
Khisanth Ovid: so which version of pugs did you install? 18:56
Ovid Fresh from svn.
Version: 6.2.7 (r5516) 18:57
svnbot6 r5521, putter++ | Makefile.PL patch broke the build. Removed. Deadly make install is back.
Khisanth oh... I guess someone broke the Makefile between 6.2.2 and now
gaal perl -i.bak -pe 's/^(install.*)/$1\tfalse/' the generated Makefile?
missing an \n at the end of the substitution there. 18:58
stevan eval? class Foo { has $.bar is ro; submethod BUILD ($.bar) {} }; my $foo = Foo.new(bar => 42);
oop :) 18:59
gaal evalbot isn't up, stevan.
stevan bah!
ok
thanks gaal :)
putter Khisanth: so make install was doing the right thing, whatever that is, at some point in the past? 19:01
gaal has only ever made install once. but it worked then! 19:02
svnbot6 r5522, putter++ | Added purely advisory "do NOT use make install" to Makefile.PL. Still dangerous.
putter I've added a purely advisory "dont run make install!" after the *** Enter make to build Pugs. But we really should disable it until its fixed. 19:03
Khisanth putter: I installed 6.2.2 through CPANPLUS and it seems to be in /perl6 as expected
gaal it was months ago, so that doesn't help tracebacking much.
putter looking... 19:04
gaal (it == the version i installed, not yours, Khisanth)
putter Ovid: thank you. if that had gone out a release... "shudder". 19:05
Ovid No problem.
Khisanth hrm
so safe to assume 6.2.6 didn't have this problem?
Ovid I don't know. I haven't touched Pugs in a while as I've been stuck on a huge contract and now am prepping for OSCON. This was just a lark. 19:07
dudley I did a make install within the last week, and it went in /perl6
svnbot6 r5523, Aankhen++ | HTTP::Cookies:
r5523, Aankhen++ | * modified &scan to pass rw parameters to callback.
r5523, Aankhen++ | * implemented &as_string.
putter rafl: ping 19:09
19:11 Ovid is now known as Ovid_afk
putter Does svn have a line-by-line revision attribution? Colorized? 19:13
integral svn blame?
gaal svn praise
putter thanks! 19:14
eric256 hey. if i wanted to add something to Prelude.pm .... is there somewhere i should ask first? BTW i want to add a .shuffle method to arrays, seems like something that gets done enough to put in prelude. 19:15
gaal eric, yes, propose it on p6-l, and have it added to S29. 19:16
eric256 suppose a local working version would go a long way towards getting the go ahead?
gaal sure, though the most important things are probably the function signature and an argument for the algorithm you've chosen to implement this in. 19:17
plus why this should be in the prelude, of course.
eric256 in prelude becuase its cheap enough to add, and i'm pretty sure (thought possibly wrong) that fisher_yates could be considered the standard, so why make people keep coding it? ... /me realizes i don't need to argue for it hear, just testing it out 19:18
gaal read around about shuffles - i seem to recall a discussion that compared a few. fisher-yates wasn't the only interesting one there.
unfortunately i have no idea where i did see it, except that maybe it was in some functional programming forum or faq or archive. 19:19
eric256 mhm okay. i'll get the bones laid out and working, then look into better solutions. 19:21
make will recompile Prelude.pm right?
gaal sure.
eric256 does this mean perl can now cache compiled versions of modules for quiker load times?
gaal when developing this, though, you may find it faster not to rebuild pugs every time 19:22
set PUGS_BYPASS_PRELUDE=1 in your env, and use -I src/perl6 -MPrelude
eric256 will it know to use the edited Prelude.pm ? or should i just code it in a seperate file at first, then merge when working?
or that would work. ;)
thanks
gaal np :) 19:23
re: caching arbitrary modules, we're not there yet, but backlog a bit, there's been discussion about it today.
svnbot6 r5524, Aankhen++ | * HTTP::Cookies: implemented :host, :uri_path and :normalize_path.
putter eric256: it sounds like something which to start with should simply go in a module. then if gets made part of Array, it can be moved to the Prelude, or to some Array.pm if that's been spun of by then. 19:28
Bah, makefiles. Lets use rake. ;) 19:29
eric256: one of the nice things in p6, is one can simply do module SomeArrayExtensions; method Array::foo().... use SomeArrayExtensions; 19:32
eric256 yes. but a whole module for array shuffling seems silly. either way i can't seem to get this working anyway..../me attacks from a different angle. 19:33
Aankhen`` I think I'll leave the last two methods from HTTP::Cookies for another time. :-P 19:34
gaal be sure to "is rw" :-)
svnbot6 r5525, Aankhen++ | * HTTP::Cookies: implemented &set_cookie and &set_cookie_ok.
Aankhen`` wanders off.
eric256 hmmm....number of elements in array is?? not .length or .elements 19:37
gaal .elems 19:38
eric256 hehe. shortend the wrong one. thanks gaal....i've been away for too long, its all coming back slowly though
gaal t/data_types may come in handy, eric256. 19:39
eric256 thanks i'll go look around in it.
gaal np. dinner &
eric256 things long forms should work too...but that would probably be wastefull namespace pollution
rafl_ putter: pong 19:42
Limbic_Region wonders WTF this message nmake is spitting out is all about 19:56
eric256 i wonder that pretty much nmake says anything 19:58
nothingmuch uh, wtf?! 19:59
nothingmuch.woobling.org/pugs_test_status/
gaal_ look on the bright side, nice duration. 20:00
eric256 hmm. i did what gaal said to load perlude.pm at run time, edited, got it working, then recompiled and tried and the changes appears gone.. (change = added Array::shuffle sub) 20:01
gaal_ does your pugs have other prelude functions? 20:02
eric256 lemme see. 20:05
"hello".as(""); prints undef in the interactive shell
gaal_ what does open("AUTHORS").readline.say say?
eric256 a whole bunch of authors names 20:06
gaal_ well, you do have *a* prelude
are you sure your prelude.pm still has your code? 20:07
if yes, try incrementally making cleaner and cleaner.
eric256 yep
gaal_ start by just rm pugs ; make 20:08
putter rafl_: make install overwrote someone's p5 lib/ with p6 stuff. yipes. saw you had been working with lib/ handling and wondered if you had any ideas. after a long pause, I'm again looking into it...
gaal_ then delete src/Pugs/PreludePC.o and .hi
then make clean
etc.
eric256 okay. i'll see if i can find when it starts working agian. (i'm guessing that is what you are looking for) 20:09
gaal_ yup
i'm however low on wakies so i'm not here for long 20:10
eric256 online a PreludePC.hs file... is that the one you meant? 20:12
Prelude.(hi|hs|o) are there as well
s/online/only/
ls
gaal_ uh, sorry, make that Run.hi and .o
eric256 hopes that deleting the .hs file isn't bad? 20:13
gaal_ nope, svn up to restore it 20:14
rafl_ putter: Right, I'm working on that.
eric256 hmm svn says that file isn't under svn control....now what did i break... 20:15
gaal_ which one, ppc? oh, right, that's autogenerated so you really did no harm :)
make again.
eric256 NMAKE : fatal error U1073: don't know how to make 'src\Pugs\PreludePC.hs'
eric256 is no longer allow to play in the src directory 20:16
gaal_ uh, is this a recent r?
eric256 yep. just updated an hour ago...unless the update didn't take.
gaal_ nmake clean and try again. 20:17
i gotta go now, good luck. :)
eric256 thanks for all the help so far
gaal_ np :)
svnbot6 r5526, gaal++ | mini-YAML parser, to be used by the build system to handle prefs
eric256 great...one of the commands from nmake clean is too long 20:18
gaal_ you need a newer make
forum2.org/gaal/pugs/nmake.exe should do the trick.
night :) &
Limbic_Region eric256 gaal_'s nmake.exe might break about a missing .dll - it is also freely downloadable off the net 20:19
putter & 20:20
Limbic_Region msvcr71.dll in case it matters
Limbic_Region calls it a day
Aankhen`` G'night.
putter Everyone is leaving... :(
eric256 yea. that at leans finished the clean correctly...... and now SVN update found tons of changed files....odd 20:24
brentdax Is there a reason for the warning against running 'make install'? 20:39
putter It smashed someone's p5 lib/ with p6 files. 20:40
Cause unknown. rafl_ is looking into it. 20:41
putter wonders if there might be a connection with make test and smoke being br0ken.
Is anyone looking into what happened to use()? 20:42
brentdax I see. This makes my life a tad difficult, but I'll find a way to manage... 20:43
putter t/01-sanity/06-use.t is failing...
dudley putter: That one's passing here... 20:45
putter wishes there was log file with one test summary line per revision. Good for detecting regressions, good for figuring out when/why the world broke, etc.
stevan putter: I think nothingmuch is working on just such a thing 20:46
dudley has wished for the same thing
stevan on top of Test::TAP::HTMLMatrix
dudley ah, good, cause tuits are in especially short supply for me right now
putter dudley: I'm failing with Can't locate lib.pm in @*INC. Clean build of head.
After make;make test, there is no blib6/lib. !?! 20:48
dudley putter: I've got a dirty build running make test right now
putter @*INC contains /usr/..., so if you ever did a make install in the past, you may be running on those files. 20:49
Do you have a blib6/lib?
dudley Nope, you're right 20:50
svnbot6 r5527, cdpruden++ | Typo fix; missing pipe in regexp
putter rafl_: note blib6/lib is nonexistent after make on pugs head. fyi. 20:51
Just to save me time in binary search of revision space, what is the latest revision anyone has which _does_ have a ./blib6/lib directory? 21:04
Thanks.
r5480 is ok. 21:05
stevan can I get a Perl6 sanity check? 21:10
class Foo { has $.bar is ro; }
the "is ro" should only relate to autogenerated accessors right?
otherwise you could never assign to $.bar 21:11
because it would make the underlying scalar container read only
putter gaal_: re knobs, how about a 'make quickly' target which is make unoptimized and dont precompile the prelude? 21:12
stevan: would you believe 'is ro' doesnt appear in (an old copy of) the bible? 21:14
stevan yup 21:15
the bible is unclear on this
putter better yet, \bro\b doesn't appear.
stevan but I have a vague recollection of Larry and Autrijus talking about this at the hackathon
putter: "is ro" is the default for class accessors 21:16
putter wait, doesnt "is rw" just mean create r and w accessors? so "is ro" == "is r".
stevan and then generates an accessor only
putter ah. 21:17
theorbtwo Depends on where you mean, I think.
stevan but "is rw" makes the accessor and the mutator
theorbtwo Do you mean on vars with class scope?
Or do you mean on arguments?
stevan theorbtwo: I am speaking only about "has $.bar is rw" in class definitions 21:18
theorbtwo Ah.
stevan has $.bar is ro; should make a rw scalar container but only generate a read accessor
theorbtwo Ja. 21:19
stevan ok
theorbtwo (I think... and it was my RFC, dammit, not that it means anything.)
stevan so the "is <trait>" on 'has' attributes does not apply to the underlying container
theorbtwo: I am trying to fake basic container types in the metamodel with tie(), this is why I am asking 21:20
theorbtwo Ah.
putter autrijus's post-hackathon p6l post had some discussion of "is constant", etc...
theorbtwo Yeah, that's right.
stevan I had passed the "ro" on, and things got really ugly :) 21:21
cause I could no longer assign to my class attributes
theorbtwo is constant is more general then is ro, but similar.
putter Including "In order to pass by read-only reference"...
theorbtwo Right. Inside the class, it doesn't matter what accessors there are -- you are assigning directly to the var, not using an accessor... I think.
my $foo is constant = 42; 21:22
($foo can never be assigned to again.)
has $.foo is ro = 42;
stevan yes
putter r5500 is ok.
theorbtwo $.foo can be assigned to again, but only by class methods.
stevan yup
that is what makes sense to me
theorbtwo To me too. 21:23
stevan so what parts in the class declaration does the underlying scalar container need to know about
the type?
has Dog $.fido;
theorbtwo I think, BTW, the default isn't ro, but no access.
stevan theorbtwo: I thought it was ro 21:24
oh well easily changed in the metamodel :)
theorbtwo I could well be wrong.
stevan has Dog $.fido handles "wag";
theorbtwo I think everything but a few blacklisted things get passed on.
stevan the "handles" part should be handled by the class not the container
thats delegation 21:25
all I can think of is passing on the Type
theorbtwo Huh? Never noticed the existance of that... were is is it documented?
putter r5510 is broken, but differently.
stevan SA12
look for "Delegation"
theorbtwo And typeish things -- is shaped, is newlined (or whatever it's called...)
stevan has $.bar is constant = 42; 21:26
that would be a different kind of container
a ScalarConstant
or somesuch
theorbtwo Da.
Or a Constant with a Scalar inside it.
stevan ok, I am not worrying about is shaped or any of that non-sense
theorbtwo Handle what you know how to handle, pass the rest on? 21:27
stevan ok, so my Scalar container will really just deal with type checking (if a type is added to it)
theorbtwo Sounds about right. 21:28
stevan theorbtwo: I am mostly trying to figure out *what* I need to handle... the how can come later
cool
danke theorbtwo :)
clkao 8use.perl.org/~Ovid/journal/ 21:34
LOL
Limbic_Region is just reading that now 21:35
clkao u t/operators/eq.....................................pugs: *** Can't locate Test.pm in @*INC (@*INC contains: /Users/clkao/work/pugs/blib6/lib /System/Library/Perl6/darwin-thread-multi-2level /System/Library/Perl6 /Library/Perl6/darwin-thread-multi-2level /Library/Perl6 .). 21:41
wtf?
thisw is a clean build on macosx
putter pugs is broken at the moment. broke between r5505 (ok) and r5515 (not). 21:43
narrowing it down...
nothingmuch is working on a test log so we dont have to play this silly game. nothingmuch++ 21:46
nothingmuch putter: if you have a full TAP output or yml file of both, i can do sometthing
svnbot6 r5528, Stevan++ | Perl6-MetaModel : the Scalar container only handles types now; made some minor modification to the way attributes are stored so that it handles tied scalars correctly; various other small changes, nothing big;
nothingmuch like, right now
stevan nothingmuch: hola 21:47
nothingmuch uh, whoe ees thiis naathingmuch you speek off? 21:48
putter nothingmuch: huh?
nothingmuch putter: i have a working diffing TTH
if you have the data to shove in it I can graph it for you
putter Ah, no. Nifty thing, and I look forward to using it. But at the moment, 21:49
nothingmuch what I didn't get around to is rounding corners
putter what I'd like is a file of lines like 'r5500 - 26 test cases: 26 ok, 1 failed, 2 todo, 2 skipped and 0 unexpectedly succeeded' so I dont have to search through revision space. 21:50
nothingmuch oh, that' easy
putter :)
nothingmuch generate Test::TAP::Models somehow (call analyze_lines?)
then simply print each ones ->summary
putter r5505 ok, r5509 and r5510 broken differently, r5515 ok. proceeding... 21:51
err, no. r5515 br0ken.
Err, no again. r5509 ok. r5510 broken differently. r5515 broken. 21:52
:) 21:53
rafl_: ping? 22:00
nothingmuch use.perl.org/~Ovid/journal/25550 <-- is he asking for ltu? 22:01
stevan nothingmuch: he asked for it over on LtU too 22:02
nothingmuch hah 22:08
so he really wants LtU in chrome ;-) 22:09
nothingmuch should read more ltu
there, rss feed
putter ok, r5509 was working. r5510 wasnt. r5511 is irrelevant. r5512 also broken. r5513 has the current form of brokenness.
rafl_: ping?
rafl: ping? 22:10
Ok, to sum up the State of the Build: r551{0,2,3} broke make test. It's still broken. Around the same time make install smashed Ovid use.perl.org/~Ovid/journal/25660 "Vicious Pugs :(". 22:16
They may or may not be connected, but...
Someone needs to go through r551{0,2,3} and fix or revert. 22:17
I'll do it when I get back from dinner if someone hasn't beaten me too it.
Unfortunate. 22:19
&
mugwump 5510 added Perl6::MakeMaker (see lists.utsl.gen.nz/pipermail/pugs-co...2842.html) 22:42
that might have been a rename
lists.utsl.gen.nz/pipermail/pugs-co...02844.html is probably the culprit 22:43
pjcj catches up on today's p6l excitement 22:45
clkao pjcj: hey, i thought you are busy on work or fixing devel::cover 22:57
pjcj Unfortunately, I am at work, yes. 23:00
but I was just waiting for a db to install
wolverian chronoa.dy.fi/~wolverian/nuclear.p6 # does this look sane? particularly, does role inheritance work like that? 23:14
stevan nothingmuch: ping 23:33
nothingmuch stevan: sorry 23:57
svk merge got stuck because i forgot to quit filemerge
it's sooo slooow 23:59

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!