»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'perl6: say 3;' or rakudo:, niecza:, std:, or /msg camelia perl6: ... | irclog: irc.perl6.org | UTF-8 is our friend!
Set by sorear on 25 June 2013.
00:05 virtualsue left 00:12 Mouq joined
Mouq r: ｢works already?｣ 00:12
camelia rakudo-parrot 41cd5e, rakudo-jvm 41cd5e, rakudo-moar 41cd5e: OUTPUT«===SORRY!=== Error while compiling /tmp/tmpfile␤Bogus statement␤at /tmp/tmpfile:1␤------> <BOL>⏏｢works already?｣␤ expecting any of:␤ statement list␤ …»
Mouq Oh, I understand that roast commit, ok 00:13
++lue :)
lue Not a very hard fix, just need to update everything, and then spectest the before and after :) 00:16
dalek p/moar-conc: b918e81 | jnthn++ | src/vm/jvm/ (5 files):
Add ReentrantMutex REPR and lock/unlock ops on JVM
 00:19
TimToady I presume we still have to be careful about putting big codepoints into CORE directly
jnthn TimToady: Probably, but (in theory) Moar and JVM aren't vulnerable... 00:20
TimToady Moar deals with it by slowing down all strings instead? :P
jnthn TimToady: Well, Moar used to have a simple immutable string model, then diakopter++ added ropes, which didn't actually work properly and so are flattened away immediately after creation as a hack, which is SLOW. :/ 00:21
TimToady: At this point, I dunno what's going to happen there. The code is macro'd to the point of being a bloody nuisance to debug. :( 00:22
Not one of the bits of the codebase I'm happy with. 00:23
lue jnthn: I suppose there isn't an easy way to switch back to simple immutable in the meantime, is there? 00:24
jnthn lue: Not easily.
I just need to throw some time at understanding what's there and working out a way forward with it. 00:25
dalek kudo/moar-conc: 9afc8c3 | jnthn++ | src/vm/jvm/core/Lock.pm:
Use ReentrantMutex REPR and new ops for Lock.

Means it is no longer tied to the JVM.		 00:30
kudo/moar-conc: 89ff343 | jnthn++ | / (3 files):
Move Lock.pm out of src/vm, now it's VM-generic.
ast: ef25233 | coke++ | S12-class/augment-supersede.t:
add test for RT #117165
 00:36
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...?id=117165
pippo jnthn: does Configure.pl for MoarVM check for presence of libreadline? 00:41
jnthn pippo: Doesn't look like it 00:42
00:43 colomon left
pippo on my box it does not seem detecting it. Is it trying to compile a test program to detect libredline? Does it include linking to libncurses? 00:44
00:44 colomon joined
jnthn pippo: It simply isn't looking for it right now and always using the bundled linenoise instead. 00:45
pippo: So I guess the problem on Linux is with that...
pippo I have used --use-libreadline and added linking to include libncurses and it works! 00:47
jnthn Oh...
pippo I have used perl Configure.pl --use-libreadline and added linking to include libncurses and it works!
jnthn Nice :)
yeah, I see it is in there... 00:48
OK, so it doesn't probe for it, but you can configure it.
pippo yes! but you have to add linking to libncurses afterwards manuallly. At least on my box. 00:49
jnthn Oh. Wouldn't it be libreadline, though? 00:50
pippo both. On my box libreadline requires also libncurses dont know why.
geekosaur because libncurses includes the low level terminal routines 00:51
jnthn geekosaur: Is that typical/normal? So if you link against one you should always link against the other?
geekosaur they dropped the separate libtinfo years ago
jnthn geekosaur: Uh, to phrase it better, if you like to libreadline you should always link to libncurses too? 00:52
geekosaur typical for ncurses systems, yes. if using bsd or original sysv curses (which probably only means commercial unixes) then it'd be libtermcap, libtermlib, or libtinfo
but on a syste, with ncurses, ncurses subsumes the old low level libraries 00:53
[Coke] huh. where is rakudo's STD.pm ?
geekosaur (and on at least some System Vs, libcurses is preferred with libtinfo only there for backward compat and a symlink to libcurses) 00:54
lue [Coke]: src/Perl6/Grammar.nqp ?
pippo jnthn: sorry the configure param is --use-readline and not --use-libreadline. Good night! 00:56
00:56 pippo left
geekosaur so, upshot: except on some (mostly older) commercial unixes, readline uses curses/ncurses for the low level terminal stuff. you may well not even care about supporting those older ones 00:56
jnthn geekosaur: OK, thanks. 00:57
geekosaur since they're at least 10 years behind current standards (I think both POSIX and the last X/Open standard specced curses as holding the former terminfo library)
[Coke] lue++ 00:58
geekosaur heck, I think SVR2 was the last release where terminfo was separate from curses, it was merged by SVR3.1 with only a compat symlink left. does anyone really care about anything pre-SVR4? 00:59
timotimo jnthn: do you think i could do the same jvm-interop-to-REPR-plus-op for "the next thing" if i looked at your code? 01:05
jnthn timotimo: Hmm, depends which one we make the next thing I guess :) 01:06
hoelzro I found some oddness with the compiled URI on MoarVM 01:07
'need URI::DefaultPort; URI::DefaultPort::scheme_port("https")' works fine
*unless* you 'use URI' in that script 01:08
jnthn hoelzro: Urgh. 01:10
hoelzro if I remove the 'need URI::DefaultPort' in URI.pm, it works
I'll try to create a minimal example 01:11
there's no RT ticket for this, is there?
timotimo jnthn: i'd leave that up to you. or you tell me which one would be the easiest ;)
hoelzro I'm guessing that it's something with the %default_port hash being loaded twice? 01:12
timotimo jnthn: on the other hand, i'm going to go full-on sleep-deprivation mode this weekend :P
jnthn timotimo: Does that mean you're going to sleep a lot, or a little? 01:15
timotimo: Semaphore is easiest one I think
timotimo: You can follow same pattern as for lock
timotimo little, probably ;)
01:19 dmol left
timotimo jnthn: how do i best get the initial number of slots into the semaphore? 01:20
01:20 grondilu left
timotimo i don't think i can give allocate more parameters, aye? 01:20
hoelzro jnthn, timotimo: here's a minimal example for the URI module
timotimo oh, you're hunting a crash 01:21
hoelzro rt.perl.org/Ticket/Display.html?id=121298
timotimo ?
hoelzro it's that "cannot invoke null object" issue that I saw with URI
I reduced it to a simple example and attached it to the ticket
timotimo oh, ok
thank you. i don't have the smarts to look into it right now 01:22
hoelzro it looks like two loads messes up a module-level data structure
no rush =)
timotimo jnthn: since set_int is used to preallocate lists to a given size, should allocate set null for the semaphore and set_int should create the actual semaphore with the given number? 01:23
set_elems, actually 01:24
01:24 daniel-s_ left
hoelzro family time & 01:25
01:25 tgt joined, daniel-s_ joined
jnthn timotimo: No, overloading those feels odd 01:26
timotimo: Well, set_int could work in which case you nqp::box_i(1, SemaphoreType) :)
timotimo: I've just been looking at the libuv semaphore API and how we use them in Promise and...hmm. 01:28
timotimo jnthn: not even set_elems? 01:29
jnthn timotimo: set_elems is worse than set_int... :)
I think I can live with set_int being used for it
01:30 daniel-s__ joined
jnthn The problem is that Promise relies on being able to post a big value to the semaphore 01:30
Whihc libuv's API doesn't let you do.
01:30 daniel-s_ left
jnthn pthreads defines a post_sem_multiple 01:33
timotimo why would it post a big value to the semaphore? 01:35
01:36 berekuk joined
jnthn timotimo: When you .result on a Promise that ain't done yet, it tries to acquire the semaphore 01:36
timotimo: It posts a big value 'cus it doesn't know how many times that has been done.
timotimo d'oh %) 01:37
jnthn Anyway, seems on OSX the semaphore API doesn't have an equivalent of that. 01:38
timotimo so really we want an event instead
jnthn "event"? :)
timotimo er
"condition variable"?
the one where any amount of things wait and one signals
jnthn Yeah. That may be the best way.
timotimo so why am i trying to implement a semaphore right now? :)
jnthn Right.
Semaphores are used elsewhere, iirc 01:39
Yeah, thread pool uses them too
timotimo OK
it won't get us further to promises, though, just to 100% ;)
ah well.
jnthn Well, we need the scheduler before we can have promises 01:42
timotimo the ops for the semaphore go into rakudo, rather than nqp?
jnthn No, NQP 01:43
Use moar-conc branch 01:44
timotimo OK
ah yes indeed
i didn't look correctly
dalek p/moar-conc: bd253e0 | jnthn++ | src/vm/moar/QAST/QASTOperationsMAST.nqp:
Map lock and unlock ops on MoarVM.
 01:45
kudo/moar-conc: c8d8475 | jnthn++ | tools/build/Makefile-Moar.in:
Add Lock.pm to Rakudo Moar build.

With this we pass S17-concurrency/lock.t under Moar.
TimToady \o/ here too :) 01:46
01:46 berekuk left
timotimo should i use lock and unlock for semaphores or rather have ops "verhogen" and "proberen"? 01:47
jnthn uh...neither? :) 01:48
timotimo semacquire and semrelease?
jnthn I like those more.
They're not the traditional terms.
timotimo ok
java has "acquire" and "release"
jnthn Yeah.
timotimo i think the ones that are traditional are post and something else
jnthn wait 01:49
timotimo maybe sempost and semwait
but not post and wait
skids Every time I read a pthread manual page I despair at how much gets sacrificed for portability to platforms that don't implement it.
jnthn Could do those. I think acquire and release are better names though.
timotimo without "sem"? 01:50
jnthn with :)
semacquire and semrelease
timotimo thanks 01:51
jnthn The scheduler is a pretty fearsome porting job...
Well, ThreadPoolScheduler
timotimo :( 01:52
TimToady found a workaround for the overeager for bug: gather for @primes xx * -> $factor { ... } 01:53
jnthn o.O
Wow
timotimo: Well, scheduler takes care of a lot of hard things so other places don't have to. 01:54
dalek kudo/moar-conc: 5b0b70d | jnthn++ | tools/build/Makefile-Moar.in:
Add Scheduler role to the build.

Doesn't contain anything JVM-specific already.		 01:55
jnthn timotimo: We'll need semtryacquire also 01:57
timotimo where do i get ExceptionHandling from? o_O
i thought for every symbol there is there should be an import line 01:58
jnthn org.perl6.nqp.runtime.ExceptionHandling
01:58 berekuk joined
timotimo weird, i didn't see that :| 01:58
is it legitimate to call it SemaphoreRepr? 01:59
because java seems to complain about importing Semaphore and then declaring a class called Semaphore :)
jnthn Yeah, it will... :S
I think you'll just have to fully qualify its one 02:00
Oh
timotimo hm, ok
jnthn Actually, on JVM the REPR's string name isn't actually the class name
See REPRRegistry.java
02:00 klapperl_ joined
jnthn So yeah, you can get away with that. 02:00
timotimo right
i saw that
that's why i asked :)
oh, i think i made it work 02:01
jnthn Timers are going to be kinda fun... 02:02
timotimo how should i best handle the InterruptedException? use acquireUninterruptable instead or just dieInternal when i get interrupted?
jnthn The latter is probably more sociable 02:03
timotimo good
oh, i need to box a bool for that 02:04
probably an int instead
02:04 klapperl left
timotimo then i'd have to pass in an Int SMO, right? 02:04
02:07 Mouq left
timotimo or just use BOOTint from the tc? 02:07
02:08 berekuk left 02:09 tgt left, berekuk joined
timotimo jnthn: can you wait another night for the semaphore code? 02:10
jnthn timotimo: Box a bool for what? 02:11
timotimo: Sure; gonna go sleep in a moment
timotimo the return value of tryacquire
02:13 FROGGS_ joined
timotimo gnite! 02:13
dalek p/moar-conc: 9264434 | (Timo Paulssen)++ | src/vm/jvm/runtime/org/perl6/nqp/ (4 files):
beginning of a semaphore instantce
jnthn Oh, just have the op return a long, and mark it as $RT_INT in the ops file
(Compiler.nqp I mean)
02:15 Sqirrel joined
jnthn 'night, #perl6 02:16
TimToady o/
02:17 FROGGS left, Sqirrel_ left 02:20 berekuk left 02:25 rurban left 02:31 colomon left 02:32 colomon joined 02:38 berekuk joined 02:48 berekuk left 02:59 klapperl joined 03:01 colomon left 03:03 klapperl_ left 03:09 kurahaupo_mobile left 03:11 berekuk joined 03:13 mtj_ left 03:18 mtj_ joined 03:22 berekuk left 03:38 kurahaupo_mobile joined 03:43 colomon joined
dalek kudo/nom: 7b43e38 | Mouq++ | src/Perl6/ (2 files):
Abbreviated Pod blocks can't have configurations
 03:50
03:50 Mouq joined 04:05 fridim_ joined 04:15 xenoterracide joined 04:30 kurahaupo_mobile left
lue Running make -j3 spectest, the wallclock seconds in order of parrot, jvm, moar: 13505, 11311, 3723 :) 04:58
05:13 Mouq left 05:34 BenGoldberg left 05:36 hoverboard left 05:46 hoverboard joined 05:51 rurban joined 06:05 Mouq joined 06:08 rhr left, rhr joined 06:09 Mouq left 06:10 raiph left 06:18 darutoko joined
dalek ast: 0265e51 | lue++ | S26-documentation/09-configuration.t:
[S26-documentation] Change abbreviated to paragraph

Abbreviated can't take config options, which is what's being tested here, so change it to paragraph.		 06:32
ast: d97551f | lue++ | S12-class/augment-supersede.t:
Merge branch 'master' of github.com/perl6/roast
06:34 SamuraiJack joined 06:37 daniel-s__ left 06:38 daniel-s__ joined
dalek kudo/nom: 8b435d3 | lue++ | src/Perl6/ (2 files):
Add support for short-form Q strings.

You should now be able to type out your Q-based strings as ｢...｣, similar to '...' and "...".		 06:44
kudo/nom: 44f3569 | lue++ | .gitignore:
Add missing perl6-debug-p to .gitignore
rakudo/nom: bfe22c4 | lue++ | / (2 files):
rakudo/nom: Fix localtest targets
rakudo/nom:
lue ? I think dalek got caught on something. 06:46
dalek ast: 4f3dfd4 | lue++ | S02-literals/quoting-unicode.t:
Unfudge test for ｢...｣
 06:51
07:19 rurban left 07:22 daniel-s__ left 07:26 daniel-s__ joined 07:31 daniel-s__ is now known as daniel-s_ 07:34 hoverboard left 07:48 kst joined 07:53 Mouq joined 07:54 slavik left 07:56 berekuk joined 07:57 Mouq left 08:08 virtualsue joined 08:17 kivutar joined 08:28 sqirrel_ joined 08:31 bjz_ left 08:32 telex left
raydiak colomon++ # your idea that my mysterious test failure might be a precompilation bug was absolutely correct 08:34
08:35 telex joined
raydiak FROGGS_: ^ found the cause of my test failure :P 08:36
FROGGS_ raydiak: that was my guess too... we'd just need to golf it down so we know what breaks when precomp'd 08:39
p: say ｢...｣
camelia rakudo-parrot bfe22c: OUTPUT«...␤»
FROGGS_ that is interesting 08:40
08:41 rindolf joined
raydiak forgot I was running panda install with jvm, and ran panda-test with moar, thus my "why doesn't it break with panda-test" confusion 08:42
FROGGS_ raydiak: that is what my S11 implementation is supposed to solve 08:43
raydiak ?
FROGGS_ you will automatically get the -m/-j/-p suffix when installing a binary/script
so there would be a panda-test-j, that will always delegate to perl6-j
raydiak that sounds handy 08:44
08:50 molaf_ left 08:54 kurahaupo_mobile joined 08:57 kivutar left
FROGGS_ hoelzro / colomon: there? 08:58
raydiak: this might even be your issue: rt.perl.org/Ticket/Display.html?id=121298 08:59
09:00 kurahaupo_mobile left, kurahaupo_mobile joined
raydiak FROGGS_: looking 09:02
09:09 bjz joined
raydiak hmmm...I can check to see if the things being passed as params are defined then, so I know if it's the same problem 09:09
or better yet, I'll try precompiling this test as a jar and running it again...if this works under jvm, it must not be my problem
09:10 bjz_ joined, bjz left
raydiak also broken under jvm, so it could be my issue...still investigating 09:14
eh? when not logged in I can view the ticket but when logged in I get "No permission to display that ticket"? 09:19
FROGGS_ O.o 09:21
raydiak haha I get the weirdest obstacles lately...was just gonna add "also broken under jvm" 09:23
my problem looks different, though...like it's not seeing the overloaded infix:<+> 09:26
09:30 berekuk left
raydiak will finish pinning this down when I can think better...tis late; g'night FROGGS++ and #perl6 o/ 09:36
09:36 berekuk joined
masak antenoon, #perl6 09:37
yoleaux 21 Feb 2014 21:24Z <raiph> masak: I filed #121296; plz lemme know if I could have done better
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...?id=121296
masak raiph: looks very good to me. 09:38
timotimo i'll fix the return type of the tryalloc right away 09:45
dalek p/moar-conc: 326d6dc | (Timo Paulssen)++ | src/vm/jvm/ (2 files):
have semtryacquire return an INT, map sem{{try,}acquire,release}
 09:47
10:04 klapperl_ joined 10:07 klapperl left 10:24 cognominal left 10:29 espadrine joined 10:30 espadrine left
timotimo and now the rakudo part 10:32
10:34 sqirrel_ left 10:36 dmol joined
timotimo jnthn: is it really box_i or box_I? 10:39
timotimo would try, but has to build nqp first
jnthn timotimo: box_i; box_I = bigint. 10:40
timotimo good 10:41
i'll leave the Promise.pm to have the jvm interop for semaphore for now, aye?
jnthn Yeah
We may want to spec and make semaphores available too. 10:42
Not because I want people to use them day to day, just because I want folks to be able to build new concurrency abstractions as the whole industry learns better ways to do things. 10:43
timotimo sure. this is work in a branch so far, so i'll get it to work and then worry about making it perfect
jnthn Sure :)
10:45 _thou left 10:52 virtualsue left 10:55 berekuk left
timotimo we need some adjustments so that we can somehow run the individual spectests with just make without being required for the parrot build to work 10:57
because right now parrot won't touch the Semaphore due to an unknown repr :P
jnthn timotimo: They are marked # jvm right now 10:58
timotimo: I suggest we add # conc
timotimo: And pass --conc to the harness for JVM/Moar.
timotimo that's not what i mean :) 10:59
the **/*.t ruel in the makefile depends on P_HARNESS_WITH_FUDGE or something 11:00
that P_ is what i meant
jnthn: how do you feel about AtomicInteger getting its own REPR?
11:01 berekuk joined
jnthn timotimo: It shouldn't. 11:02
timotimo that's right :)
but how do e differentiate if "make t/spec/foo/bar.t" means "do it with parrot" or whatever?
jnthn Oh, no idea
I never use that :)
timotimo should the method for tryAcquire have that same capitalization in perl6 as it does in java? 11:03
jnthn No
nqp::semtryacquire
timotimo i've given Semaphore a method to do that
jnthn ? 11:07
Oh, you mean at Perl 6 level? 11:08
Should be try_acquire there most probably...
timotimo good
jnthn Though really it doesn't matter what we call it; TimToady++ will spring up with a better name anyway. :D
timotimo This representation can not unbox to a native int - that's what i get for my tryacquire ... huh
oh, i may need to recompile nqp perhaps? 11:09
nope :|
jnthn Unbox? Hm
Don't know we should be trying to do that. 11:10
timotimo yeah, i'm not sure what's causing the error
i'll try to golf it and see if that gives me any insight
./perl6-j -e 'Semaphore.new(1).tryAcquire()' 11:11
that triggers it already
jnthn .tell hoelzro I think rt.perl.org//Public/Bug/Display.html?id=119877 is fixed - please can you check? 11:14
yoleaux jnthn: I'll pass your message to hoelzro.
timotimo should i just push it to see what you think?
11:16 Rotwang joined
jnthn timotimo: Yeah, that may be the easiest way. 11:16
dalek kudo/moar-conc: 51fb614 | (Timo Paulssen)++ | / (3 files):
first (nonworking) draft of Semaphore class
 11:18
timotimo there we go. good luck :)
oh, is that from "int $permits"? 11:19
jnthn timotimo: Hmm, I'm thinking the problem may not be in that code 11:20
Moment
QAST::OperationsJAST.map_classlib_core_op('semtryacquire', $TYPE_OPS, 'semtryacquire', [$RT_INT], $RT_OBJ, :tc); 11:21
There you are.
The thing in the square brackets are the op argument types
And the one to the right is the result type.
timotimo oooooh
haha :)
jnthn So your tryacquire is an i -> o op, not an o -> i one like you wanted :) 11:22
timotimo thanks
jnthn np :)
timotimo i'm kinda groggy >_
>_<
jnthn Maybe it's the low-sleep thing ;) 11:29
timotimo the amount fo sleep was sufficient, i think
but it was a mattress that was much softer than i'm used to and a pillow that was much smaller than i'm used to
well, it was a bit bigger, but it compressed much moar 11:30
jnthn ah
timotimo i should have seen that the stuff in the [ ] is the arguments, because that's a list and a list can also hold multiple things 11:33
unlike a retur nvalue, which is usually one thing
even if it's a list
dalek p/moar-conc: 00e4d36 | (Timo Paulssen)++ | src/vm/jvm/QAST/Compiler.nqp:
confused returnval and argument types
 11:35
timotimo jnthn: the test file scheduler.t doesn't seem to terminate after the 55th test 11:36
jnthn timotimo: On JVM? 11:37
timotimo: Did you fudge it?
timotimo i didn't :(
jnthn oh, it's no fudge marks.
timotimo that would be it.
right, no fudge marks 11:38
jnthn timotimo: Did it terminate before your changes?
It does here...
timotimo didn't test :P
11:38 beastd joined
timotimo takes so long :( 11:40
i really liked the quick turnaround times on moarvm during development
i should probably support acquiring and try-acquiring more than 1 permit, right? 11:41
even though that's not needed yet
11:41 Zaydon joined
jnthn timotimo: No, don't add those 11:42
timotimo: They're not portably available.
timotimo oh, ok
that's good then
jnthn Well, it's not good, but it's how it is...
timotimo it does not terminate without my changes 11:45
jnthn Um. 11:47
jnthn tries it 11:48
uh, once I have a JVM build... 11:49
timotimo that could take a while, aye :) 11:51
jnthn nah, 'tis done 11:52
timotimo the next thing we're going to see are mappings for the semaphore ops in nqp? 11:55
11:55 berekuk left 11:56 virtualsue joined
Ven next thing to do is build a jvm in perl6, I guess :) 11:58
11:58 SamuraiJack left
jnthn timotimo: Feel free to add those, I don't have them here yet. 11:58
Ven you'd get the best perfs ever 11:59
(that really seems like an interesting research project though) 12:00
jnthn timotimo: Hmm. I get the exit hang too when running the test solo, but not under the eval server. Huh... 12:01
12:11 pdcawley joined, pdcawley_ left
timotimo jnthn: huh. 12:17
because the thread the test starts lives past the single program and reaches the 10000 seconds timeout? perhaps? 12:18
jnthn: i'll map them
12:19 kshannon joined
jnthn timotimo: Yeah, but they should be background/daemon threads, meaning they don't block termination. 12:19
timotimo ah, huh 12:20
jnthn timotimo: I'll have to dig a bit deeper on it. 12:21
timotimo: Currently researching/planning for the concurrent blocking queue thing we need for channels and the scheduler.
timotimo aye, that does sound good :) 12:22
12:23 berekuk joined
Timbus hm. i just stumbled into this today: rosettacode.org/wiki/Luhn_test_of_c...ers#Perl_6 12:27
and surely this is wrong: my $s1 = [+] @digits[0,2...@digits.end]; 12:28
the sequence, that is
jnthn timotimo: Yeah. I think I'll go for a 2-lock approach (separate head/tail locks) for now. That's what the JVM one is doing (turns out JVM one's source is public domain :)) 12:30
Timbus m: my @a = <0 9 8 7>; say @a[0,2...@a.end]; 12:31
camelia rakudo-moar bfe22c: OUTPUT«(timeout)»
Timbus m: my @a = <0 9 8 7>; say @a[0,2...*]; 12:33
camelia rakudo-moar bfe22c: OUTPUT«0 8␤»
Timbus oh.
jnthn m: my @a = <0 9 8 7>; say @a[0,2...+@a]; 12:35
camelia rakudo-moar bfe22c: OUTPUT«0 8 (Any)␤»
jnthn m: my @a = <0 9 8 7>; say @a[0,2...^+@a];
camelia rakudo-moar bfe22c: OUTPUT«0 8␤»
jnthn That may be better.
12:38 berekuk_ joined, berekuk left, tgt joined
timotimo i might need to write a test file for semaphores to ensure that the moarvm implementation is even right 12:40
jnthn We need one anyway. 12:41
Timbus m: my @a = <0 9 8 7>; say @a[1,3...^+@a];
camelia rakudo-moar bfe22c: OUTPUT«(timeout)»
jnthn Timbus: The ending condition has to be met exactly 12:42
Can put a code block there with a comparison op
timotimo i'm surprised ...* doesn't actually infiniloop; ... should interpret * manually, should it not? 12:43
m: say (0,2...*)[^50]
camelia rakudo-moar bfe22c: OUTPUT«0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98␤»
timotimo m: say (0,1...20)[(0,2...*)[^10]]
camelia rakudo-moar bfe22c: OUTPUT«0 2 4 6 8 10 12 14 16 18␤»
timotimo interesting
i'm not sure i understand. 12:44
JimmyZ libuv has a good queue.h file but it's without lock 12:45
Timbus at a guess its a whatevercode block inside the [], not a whatever being passed to the seq
timotimo Timbus: i know how the whatevercode generation code looks like :) 12:46
in my mental model, it shouldn't know about that there's a [ ] around it at the point where it encounters the ... operater 12:47
Timbus i think i follow you. after it's been given the size it.. huh. why doesnt it infiniloop 12:48
m: my @a = <0 9 8 7>; say @a[1,3...*]; 12:49
camelia rakudo-moar bfe22c: OUTPUT«9 7␤»
Timbus well im going to use that then. because: @a[1,3...^*>=@a] this is bad and i feel bad 12:50
jnthn JimmyZ: I don't see anything showing that's threadsafe, though. Plus it's not, afaict, public API. 12:58
JimmyZ oh, yeah, I didn't realize it's not public API 13:01
btw, will we need rwlock? 13:02
jnthn Don't think so.
13:10 kaare_ left 13:25 rindolf left 13:30 hugme left, virtualsue left 13:31 hugme joined, ChanServ sets mode: +v hugme
dalek p/moar-conc: cd2f5c7 | (Timo Paulssen)++ | src/vm/moar/QAST/QASTOperationsMAST.nqp:
map semacquire, semtryacquire and semrelease on moar
 13:34
hoelzro good * #perl6 13:38
yoleaux 11:14Z <jnthn> hoelzro: I think rt.perl.org//Public/Bug/Display.html?id=119877 is fixed - please can you check?
hoelzro jnthn: thanks, I'll check 13:39
timotimo jnthn: would a test file for semaphores help you/us? should it be a spectest or an nqp test? 13:42
jnthn timotimo: Since we're adding a user-visible Semaphore class, should be spectest. 13:45
timotimo: It warrants an S17 addition to. 13:46
*too
hoelzro timotimo, jnthn: did either of you get a chance to look at that multiple import moar bug I reported last night? 13:47
rt.perl.org/Ticket/Display.html?id=121298
jnthn hoelzro: No, but got it open in a browser tab here. It's on my "to do soon" list 13:50
hoelzro ok, cool 13:51
it was late when I reported it, so I wanted to make sure one of you had seen it =)
timotimo hoelzro: i did not get to it :( 13:53
but thank you kindly for golfing it
hoelzro np, it was interesting
timotimo i'll write a bit for the semaphore 13:55
the same kind of "use higher level instead if you can, but if you really, really, really need it, take this"
actually, time to attend a presentation 13:56
jnthn "Typically you should acquire a higher level solution; this is just a release valve." :P
timotimo :) 14:02
14:06 hoverboard joined
timotimo puts lock and semaphore under a common section 14:06
jnthn wfm 14:07
14:11 xinming joined
timotimo do i need a code example to spice up the text or will an explanation of the operations suffice? 14:11
i'll also write that acquiring and releasing multiple permits at once is not supported on all backends, so will not be offered 14:12
JimmyZ Do we need AtomicInteger repr too? 14:14
dalek ecs: 7f3f23c | (Timo Paulssen)++ | S17-concurrency.pod:
semaphore spec draft
timotimo JimmyZ: i asked it already and got no answer :P 14:15
JimmyZ :/
timotimo jnthn: there seems to be public interest in this :P
jnthn 10:59 <timotimo> jnthn: how do you feel about AtomicInteger getting its own REPR? 14:16
11:00 <jnthn> timotimo: It shouldn't. 14:17
:P
14:17 tgt left
timotimo oh 14:17
thank you! i didn't see that
jnthn ;)
We'll get rid of it for now by doing a lock, and later we'll add CAS primitives.
JimmyZ I didn't see the Q&A :P
jnthn np :) 14:18
14:19 beastd|2 joined
JimmyZ I'm just thinking how to do CAS ... 14:19
14:19 [Sno]_ joined
jnthn JimmyZ: Needs rather careful design, as it needs to co-opt REPRs too... 14:20
14:20 beastd|3 joined
timotimo we'll be able to atomically CAS any P6opaque? 14:21
hoelzro jnthn: thanks, it's fixed
jnthn timotimo: Not sure at present. I need to look at where the boundaries lie on JVM. 14:22
14:23 kst` joined, FROGGS__ joined 14:24 takesako___ joined 14:30 Psyche^_ left 15:12 ilogger2 joined 15:13 xfix left, Ulti_ left, bjz left, nwc10 left, cooper left, sftp left, mattp__ left, Ven left, Zaydon_ left, Alina-malina left, Exodist left, bonsaikitten left, PerlPilot left, Sqirrel_ left, avuserow3 left, flussenc1 left, darutoko- left, kst` left, kshannon left, pdcawley left, dmol left, klapperl_ left, dagurval left, Timbus left, dalek left, tadzik left, risou_awy left, integral left, daniel-s__ left, skids1 left, rhr left, LordVorp left, takesako___ left, FROGGS__ left, xenoterracide left, fridim_ left, pecastro left, Pleiades` left, Gothmog_ left, nebuchadnezzar left, felipe left, clsn left, grep0r left, timotimo left, tokuhirom left, hoelzro left 15:17 Vlavv joined, mtk joined 15:19 jnthn joined 15:20 bakedb joined, lizmat joined, ilbot3 joined, Nei joined 15:24 cxreg joined 15:26 silug__ joined, hugme joined 15:28 ilbot3 left 15:29 cosimo joined, robinsmidsrod joined, pochi joined 15:33 [Coke] joined 15:34 [particle] joined 15:47 flussence joined, cxreg left 15:51 Nei left, jnthn left, sunnavy joined 15:52 flussence left, cosimo left, perigrin joined, silug__ left 15:53 itz joined 16:13 ilogger2 joined, ChanServ sets mode: +v ilogger2, colomon joined, gfldex joined
tadzik fwiw, the Threads module has some example programs written with Semaphores 16:13
16:13 Tene joined, Tene left, Tene joined 16:14 Rotwang1 joined 16:18 cxreg joined, amkrankruleuen joined 16:19 japhb_ left, rurban_ left 16:20 silug__ joined, Nei_ joined, ruoso left, lestrrat left, japhb_ joined 16:21 simcop2387 left, pnu_ left, hoelzro left, bakedb_ joined, Nei_ is now known as Nei 16:22 rurban_ joined, hoelzro joined, lestrrat joined, cxreg left 16:23 raydiak__ joined 16:24 ashleydev left, ribasushi left, Grrrr left, vendethiel joined, hugme joined 16:25 gfldex_ joined, retupmoc1 joined 16:26 apejens joined 16:27 kurahaupo_mobile joined, berekuk joined 16:28 clkao joined 16:29 cooper- joined 16:30 lestrrat left, gfldex_ left 16:31 Zaydon joined, timotimo left, cooper- left, Grrrr joined 16:32 apejens left, Nei_ joined, ashleydev joined, lestrrat joined, Nei left, Nei_ is now known as Nei, rurban1 joined, itz_ joined 16:33 itz__ joined, sunnavy joined 16:34 amkrankruleuen left, amkrankruleuen joined 23:41 ilogger2 joined, ChanServ sets mode: +v ilogger2
Mouq lue: What I mean is that <?{}> can put the AST of its contents into the tree. Quantifiers work by creating a partial AST node with how many repetitions it allows as named params, and then whatever it's quantifying injects itself into that node 23:43
lue: So the way things are now, AFAIK you'd have to get **{}'s compile time value 23:44
or something
:/
lue Hrm, compile time won't cut it, 'cos you could always use bare ** for those :) 23:45
Mouq r: =for pod :test(1+12)␤content␤␤say $=pod.perl
camelia rakudo-parrot b2b333, rakudo-jvm b2b333, rakudo-moar b2b333: OUTPUT«Array.new(Pod::Block::Named.new(name => "pod", config => ("test" => "1+12").hash, content => Array.new(Pod::Block::Para.new(config => ().hash, content => Array.new("content")))))␤»
Mouq lue: Yeah, I didn't think so 23:46
lue: So something would have to be restructured
23:46 erkan joined, bjz joined, timotimo joined, pnu_ joined, ruoso_ joined, sftp_ joined, Celelibi joined, __sri joined, zamolxes joined
lue I wonder if **{} would require a retooling of how quantifiers work, or a separate node type 23:46
Mouq also see 23:47
r: say "a b c" ~~ /:s(True) \w +/
camelia rakudo-parrot b2b333, rakudo-jvm b2b333, rakudo-moar b2b333: OUTPUT«Nil␤»
Mouq r: say "a b c" ~~ /:s \w +/ 23:48
camelia rakudo-parrot b2b333, rakudo-jvm b2b333, rakudo-moar b2b333: OUTPUT«｢a b c｣␤␤»
Mouq (although that's partially an issue with the way the grammar handles :s) 23:49
n: say "a b c" ~~ /:s(True) \w +/ 23:51
camelia niecza v24-109-g48a8de3: OUTPUT«===SORRY!===␤␤Action method mod_arg not yet implemented at /tmp/Lpz5NyMx2P line 1:␤------> say "a b c" ~~ /:s(True)⏏ \w +/␤␤Unhandled exception: Cannot parse number: (True)␤ at /home/p6eval/niecza/bo…»
Mouq std: say "a b c" ~~ /:s(True) \w +/
camelia std 09dda5b: OUTPUT«ok 00:01 126m␤»
23:56 bnebheni joined 23:58 bnebheni left, erkan left, bjz left, timotimo left, pnu_ left, ruoso_ left, sftp_ left, Celelibi left, __sri left, zamolxes left

Powered by Cro and the Raku® Programming Language.

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

Thank you!