bartolin .tell MasterDuke regarding the compilation of the setting on jvm: I *think* this workaround I put in a while ago really doesn't help making it faster: github.com/perl6/nqp/commit/e73c94f69c (I vaguely remember comparing compile times before and after this patch) 05:30
yoleaux bartolin: I'll pass your message to MasterDuke.
moritz Zoffix++ # ZofBot GCT 05:51
lizmat Files=1239, Tests=76308, 318 wallclock secs (15.69 usr 5.53 sys + 2191.43 cusr 217.75 csys = 2430.40 CPU) 09:40
weekly: github.com/scriptkitties/perl6-dis.../issues/37 09:41
notable6 lizmat, Noted!
AlexDaniel reportable6: 2018-04-16T00:00:00 2018-04-23T00:00:00 11:13
reportable6 AlexDaniel, OK, working on it! This may take up to 40 seconds
AlexDaniel, gist.github.com/7de3eeed5906f78022...d05011b558 11:14
AlexDaniel weekly: reportable gist.github.com/7de3eeed5906f78022...d05011b558 11:15
notable6 AlexDaniel, Noted!
Zoffix With possible approval of CaR Grant on the horizon. Reminder to look over the Work Proposals for Constants and Rationals: github.com/rakudo/rakudo/blob/mast...8-02-10.md and github.com/rakudo/rakudo/blob/mast...tionals.md Speak now or forever hold your silence :) 12:04
dogbert2_ Zoffix: have you spoken to Larry about it? 12:56
Zoffix dogbert2_: yeah, and after some original comments, I ended up with MidRat type and botmessaged him that I plan to add it. I think he saw those messages but didn't leave any response, so I figured he doesn't have anything against the proposal 12:57
That's what the revisions of Rat proposals were about; modifications for TimToady's original comments and the last revision was my changing away from RatStr-as-MidRat because it's not really working out 13:00
dogbert2_ Zoffix: are you still happy with what you want to do or have you come up with something different ideas in the meantime, i.e. since the latest draft was written? 13:02
s/something/some/
Zoffix It's the best version of what I come up with, but I'm not happy with having 3 different Rational types.
I can foresee the winging "Perl 6 has bazillion Numerics! It's so hard" 13:04
Well, 5 different Rational types if you include the allomorphs: Rat, RatStr, MidRat, MidRatStr, FatRat
dogbert2_ someone will say that the suggestions are irRational :-)
Zoffix And I'm slighly bugged that one of the supposed uses of MidRat is to be able to provide high-precision, non-infectous pi and e constants.... but... MidRat is Rational and pi and e are not rational numbers :) 13:06
dogbert2_ m: 355 / 113
camelia WARNINGS for <tmp>:
Useless use of "/" in expression "355 / 113" in sink context (line 1)
dogbert2_ m: say 355 / 113
camelia 3.141593
Zoffix m: say 314159265358979323846264338327950288419716939937510582097494459230781640628620899862803482534211706798214808651328230664709384460955058223172/10**140 13:08
camelia 3.141592653589793
dogbert2_ would it be possible to start with the things which does not involve the MidRat, so that you get some extra time to ponder about them? 13:09
btw: when will 2918.04 be out? 13:14
grr, 2018.04
Zoffix In about 900 years :)
dogbert2_ :)
Zoffix I think AlexDaniel wants to wait for JVM build to get fixed. 13:15
AlexDaniel yes, but we also don't have a moarvm release yet 13:16
Zoffix AlexDaniel: should it be cut though? I suspect samcv is waiting for a go-ahead from you.
samcv Zoffix: i will release today
was busy this weekend so didnt have time to do it then 13:17
AlexDaniel well, at least the changelog should be in place I guess, doing the actual release is much less of a problem 13:18
dogbert2_ so there's still time to dig up a new blocker :) 13:19
AlexDaniel heh, yes. Which is nice
Zoffix I think the proto stuff should be reverted.
AlexDaniel Zoffix: here's a crazy ideaā€¦ can't we just revert onlyā†’multi changes also? That would throw us like 2 months back, but at least no regressions to existing code 13:21
Zoffix It's a bit harder, because IIRC lizmat also made a lot of optimizations in the process that in some way rely on multiness.
But I was thinking of fixing the .count issue by taking care of the things that care about count. 13:22
Rather than messing with all the protos and requiring that users know all of our routines or declare explicit protos all the time.
Well, I guess it's a bit of a case at the moment too; 13:23
AlexDaniel rereads R#1739
synopsebot R#1739 [closed]: github.com/rakudo/rakudo/issues/1739 [āš  blocker āš ] Unintended consequences of converting routines from `only` to `multi` 13:24
Zoffix c: 2018.03 multi copy ($table, $other-table {}
committable6 Zoffix, Ā¦2018.03: Ā«04===SORRY!04=== Error while compiling /tmp/XXDSmnytelā¤Missing blockā¤at /tmp/XXDSmnytel:1ā¤------> 03multi copy ($table, $other-table 08ā04{}ā¤ Ā«exit code = 1Ā»Ā»
Zoffix c: 2018.03 multi copy ($table, $other-table) {}
committable6 Zoffix, Ā¦2018.03: Ā«Ā»
Zoffix c: 2018.03 multi copy ($table, $other-table) {}; copy 42, 100 13:25
committable6 Zoffix, Ā¦2018.03: Ā«Failed to copy '/home/bisectable/git/whateverable/42' to '/home/bisectable/git/whateverable/100': Failed to copy file: no such file or directoryā¤ in block <unit> at /tmp/GhgUrCPJTZ line 1ā¤ā¤ Ā«exit code = 1Ā»Ā»
Zoffix c: 2018.03 .say for &copy.candidatesĀ».signature 13:26
committable6 Zoffix, Ā¦2018.03: Ā«($from, $to, :$createonly)ā¤Ā»
Zoffix Looks like coercers don't show up in Signature's gist. The $from/$to are actually `IO()`
stmuk_ releasable6: status 13:27
releasable6 stmuk_, Next release will happen when it's ready. 0 blockers. 262 out of 263 commits logged
stmuk_, Details: gist.github.com/88404ff2eb46e2c876...9ecdb45092
Zoffix c: 2018.03 multi copy (IO() $table, IO() $other-table) {}; copy 42, 100
committable6 Zoffix, Ā¦2018.03: Ā«Failed to copy '/home/bisectable/git/whateverable/42' to '/home/bisectable/git/whateverable/100': Failed to copy file: no such file or directoryā¤ in block <unit> at /tmp/Sby1h_Lp5j line 1ā¤ā¤ Ā«exit code = 1Ā»Ā»
Zoffix Don't even know why the core multi gets chosen instead of complaining there's an ambiguity :S 13:28
But in either case, that kinda shows that even before the proto changes it was easy for users to step onto the toes of core multies, so I guess my main argument against proto changes doesn't quite stand up. 13:29
In fact, the new version is better because if proto sig doesn't match the user would get an error saying they're actually trying to add a multi for a routine that already exists in core.
s/I think the proto stuff should .+//; # I guess it's fine afterall 13:34
AlexDaniel m: proto z($) {*}; multi z($, $) {say ā€˜impossibruā€™}; &z.candidates[0](42, 50) 13:41
camelia impossibru
[Tux] Rakudo version 2018.03-263-gd4a6b92f3 - MoarVM version 2018.03-126-g876aa90ee
csv-ip5xs0.878 - 0.888
csv-ip5xs-208.768 - 8.848
csv-parser37.264 - 39.980
csv-test-xs-200.418 - 0.510
test9.147 - 10.320
test-t2.473 - 2.589
test-t --race1.003 - 1.029
test-t-2045.376 - 45.444
test-t-20 --race14.560 - 15.598
AlexDaniel m: proto z($) {*}; multi z($, $) {say ā€˜impossibruā€™}; say &z.cando(\($,$))[0](42, 50) 13:42
camelia impossibru
True
Zoffix JVM builds for me....or at least, doesn't crash in the parse stage 13:49
(google killed my VM before the build had a chance to complete entirely...)
AlexDaniel O_o 13:53
timotimo it's the "we charge you less money, but we're free to kill your vm at any time" pricing model 13:55
[Coke] needs more coffee. 13:56
Zoffix Yeah, it's the "you get whatever lands on the floor" pricing model :) 13:57
Oh, it fails when parsing 6.d setting.. :( 14:00
or whatever it's parsing after finishing parsing the first time 14:01
Reproed the crash: gist.github.com/zoffixznet/136495e...292e6aefe2 14:02
Makefile:906: warning: overriding recipe for target 'check_nqp_version' 14:03
Makefile:424: warning: ignoring old recipe for target 'check_nqp_version'
I'm guessing that's normal?
stmuk_ releasable6: are we there yet? 14:04
releasable6 stmuk_, I cannot recognize this command. See wiki for some examples: github.com/perl6/whateverable/wiki/Releasable
AlexDaniel releasable6: status 14:06
releasable6 AlexDaniel, Next release will happen when it's ready. 1 blocker. 262 out of 263 commits logged
AlexDaniel, Details: gist.github.com/11f54d5db189e4b3ee...19bc2a1c88
Zoffix How can I print the type of nomType here? github.com/rakudo/rakudo/blob/mast....java#L440 14:35
I tried a bunch of stuff but best I got is "org.perl6.nqp.sixmodel.TypeObject@10b3df77" 14:36
timotimo does the jvm backend have something like moarvm's debugname? 14:37
Zoffix or is there an nqp op that prints the name? 14:42
'cause Ops.isnull(nomType) worked and says 0 14:43
but Ops.how(nomType).name(nomType) and Ops.how(nomType.st.WHAT).name(nomType.st.WHAT) tell me .how can't be applied to that
Zoffix tries Ops.what
nope :/ method what in class Ops cannot be applied to given types 14:44
Ops.typeName( might be it 14:45
nope :) 14:46
timotimo use getdebugtypename 14:48
hm, is there a get for the set?
Zoffix oh yeah public String debugName; 14:49
timotimo OK, you need the .st.debugName of the object
Zoffix "Associative" 14:51
ZofBot: win!
I guess it's not composed by that time yet so that's why it doesn't match Mu or something?
Well, I don't see a problem with making trait_mod protos a (|); it's not like someone's gonna .sort with &trait_mod 14:54
still crashes, now with prefix:<-> :-( 15:06
Zoffix rocks out at www.youtube.com/watch?v=VCxGU23yUGQ 15:26
Somehow github.com/rakudo/rakudo/blob/mast....java#L899 bindee ends up being null and that's when everything blows up 16:15
and not Java null, but I'm guessing nqp::null; 'cause bindee != null, but (later derived from bindee) decontValue.st.debugName is 'null'
m: use nqp; say nqp::istype(nqp::null, Mu) 16:16
camelia 0
Geth nqp: 6c982a88f1 | (Will "Coke" Coleda)++ | t/nqp/019-file-ops.t
This test is now passing on OS X.

Closes #274		 16:24
synopsebot NQP#274 [open]: github.com/perl6/nqp/issues/274 t/nqp/19-file-ops.t failure in pre-2016.01
Zoffix Well, fuck me. 16:40
Stage classfile : java.lang.OutOfMemoryError: Java heap space
bartolin: yo, you around? How do I increase java heap space? I need to stick JAVA_OPTS somwhere? 16:41
Zoffix tries it as an env var
From our README: "If you get an out of memory error building rakudo on the JVM, you may need to modify your NQP runner to limit memory use. e.g. edit the nqp-j / nqp-j.bat executable (found wherever you installed to, or in the install/bin directory) to include -Xms500m -Xmx2g as options passed to java." 16:53
So.... why not make the runner stick those options in in the first place?
stmuk_ maybe it makes the build slower? 17:03
Zoffix I see 17:04
robertle from my java experience I would say the Xms is not required to make the error go away, and will only help very marginally with build time. the Xmx will not make the build slower (in the presence of enough ram), but will make it use more memory. adding -Xmx2g to the build flags all the time seems safe to me unless the machine has less than 2GB ram, in which case it's probably doomed anyway 17:14
Geth nqp: 05c18bc856 | usev6++ | tools/build/install-jvm-runner.pl.in
[JVM] Increase Xmx for nqp-j to 1 GB

With the old value (512 MB) build failures for rakudo-j started to happen.		 17:38
Zoffix bartolin++
bartolin I hope that helps.
Zoffix Thanks. 17:39
And I may have a cludgy fix for the blocker.
bartolin Zoffix++ 17:41
AlexDaniel Zoffix++ indeed 17:51
Zoffix bartolin: have you been getting "Type check failed in assignment to $!CWD; expected Str but got IO::Path (IO::Path.new("/home/c...)" from ./perl6-j tools/build/install-core-dist.pl /home/cpan/J/install/share/perl6 before the recent breakage?
somehow IO::Path.BUILD is called with IO::Path $!CWD even tho all the .new methods coerce $!CWD to Str :/ 17:56
j: ".".IO.WHAT.new(".", :CWD(".".IO)); 18:01
camelia ( no output ) 18:02
Zoffix j: ".".IO.WHAT.new(".", :CWD(".".IO));
camelia ( no output )
Zoffix j: ".".IO.WHAT.new(".", :CWD(".".IO)); say "pass"
camelia pass
Zoffix Booo.
Yeah, so it's the result of my "fix" :)
Zoffix switches to different debugging music and tries again :) www.youtube.com/watch?v=qwJj2EpC8vg 18:03
bartolin Zoffix: sorry, was afk. but it looks like you figured it out yourself ... 18:12
Zoffix yeah :) 18:13
[Coke] just got "Stage parse : Nominal type check failed for parameter '<anon>'" on the java build. is this the current error? 18:48
Zoffix Yes
[Coke] (I'm on d4a6b92f3)
danke. 18:49
Zoffix Man. somehow "SixModelObject bindee = gcx.Hash.st.REPR.allocate(tc, gcx.Hash.st);" ends up being with .st.debugName of "null". I dunno if it's actually null or just the debug name is null :/ 19:16
well, Ops.istype_nodecont(decontValue, gcx.Mu, tc) gives false, so looks like it doesn't even typecheck with Mu 19:18
wait wat 19:22
I think I got too much print formats; it is actually Mu 19:23
aha. it typechecks as Hash, but it doesn't want to typecheck as Associative (at least that's what nomType is), but if I try to typecheck it against gcx.Associative, I get a null pointer exception 19:29
man, time to give up 19:50
Unless there's "is compiling core" setting or something 19:51
Then just fudge it to type-check check for Hash rather than Associative while in core.
It's probably the same cause that causes some Callables to not typecheck as Callable because they were made too soon.
samcv either i'm doing something i have no clue what, or this is an issue 20:11
i run ./apply-patchs.p6 '/usr/src/linux'
and it tells me that it expected 2 arguments but got 3
./apply-patchs.p6 '/usr/srclinux' works fine. MAIN's signature is sub MAIN (Str $linux-source-dir) {
tried with bash and with zsh
oh facepalm. i think it's my code, not MAIN. i need some tea 20:13
lizmat buggable: weekly 20:44
buggable6: weekly
notable6: weekly
notable6 lizmat, 2 notes: 2018-04-23T09:41:53Z <lizmat>: github.com/scriptkitties/perl6-dis.../issues/37 ; 2018-04-23T11:15:56Z <AlexDaniel>: reportable gist.github.com/7de3eeed5906f78022...d05011b558
MasterDuke hm. any reason these .new's github.com/rakudo/rakudo/blob/mast...m6#L35-L43 couldn't be collapsed down to `multi method new(IO::Path: Str(Cool) $path, :$SPEC = $*SPEC, Str() :$CWD) { self.bless(:$path, :$SPEC, :$CWD); }` ? 21:02
yoleaux 05:30Z <bartolin> MasterDuke: regarding the compilation of the setting on jvm: I *think* this workaround I put in a while ago really doesn't help making it faster: github.com/perl6/nqp/commit/e73c94f69c (I vaguely remember comparing compile times before and after this patch)
timotimo Zoffix: at least moarvm has a type called VMNull, which is there only so that doing stuff with nulls is only explosive in a "throw an exception" way, not a "cause a segfault" way; probably the same is true for the jvm backend 21:25
lizmat and another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2018/04/23/...u-station/ 21:26
timotimo oh, sweet! bduggan's talk video recording is online \o/ 21:28
Zoffix MasterDuke: as I remember it, those were micro optimizations. Don't know if they still offer anything of value. 21:42
timotimo there's much more cool stuff in there than i had expected! 21:52
bduggan++
MasterDuke Zoffix: fwiw, tons of 'Use of uninitialized value of type Any in string context...' in tools/build/create-moar-runner.pl, tools/build/install-core-dist.pl, tools/build/upgrade-repository.pl, etc if you make the change i asked about 23:02
.tell bartolin i don't have before/after timings either, but it was pretty slow before that commit anyway
yoleaux MasterDuke: I'll pass your message to bartolin.
MasterDuke i installed VisualVM, trying to get it working well with the build 23:03
Zoffix MasterDuke: probably because Str() :$CWD should be Str() :$CWD = $*CWD 23:09

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!