Geth rakudo/nom: 01c2a22a66 | (Zoffix Znet)++ (committed using GitHub Web editor) | CREDITS
Add `pmurias` to credits to fix them appearing twice
 01:34
Zoffix "fatal: write failure on 'stdout': Connection reset by peer" :/ 02:20
ah, it's from a Proc, never mind.
Geth rakudo/nom: 9d63eccf7f | (Zoffix Znet)++ | tools/contributors.pl6
Make contributors tool skip point releases

Since point releases often include just the bug-fix commits, if we go by last commit to VERSION as last date of release, we end up missing all the commits that happened between last non-point release and the latest point release.
Fix by looking up last non-point-release tag and using its date as the date of last release.		 02:27
rakudo/nom: 2354cbddd6 | (Zoffix Znet)++ | tools/contributors.pl6
Simplify tag-searching code a bit
 02:32
rakudo/nom: 2aa6bc045b | (Zoffix Znet)++ | tools/contributors.pl6
Simplify tag-search code even more
 02:35
rakudo/nom: 180041adc5 | (Zoffix Znet)++ | tools/contributors.pl6
Fix comment
 02:36
rakudo: the-eater++ created pull request #1085:
Add eater		 07:36
eater why the ++ behind everyone's name? 07:37
for the lingering karma bots?
Geth rakudo/nom: 1946aee247 | eater++ (committed using GitHub Web editor) | CREDITS
Add eater
 07:41
rakudo/nom: 6064e0e1a2 | niner++ (committed using GitHub Web editor) | CREDITS
Merge pull request #1085 from the-eater/patch-1

Add eater
eater \o/ 07:45
masak eater: yes/no 08:02
eater: the surface reason is to distribute karma to contributors
eater: but the deep reason is to set up a community where contributions are the quickest way to karma (but where the karma points/gamification are just a trick to make that happen, not something important in itself) 08:03
eater aka positive reinforcement? :) 08:04
masak just so
once every year or so, we check people's scores. mostly just for fun. some people are in the thousands.
I was once double-minus'd by moritz++, which I'm duty-bound to mention in conversations like this :P
eater D: 08:05
moritz some things will never be forgotten!
moritz doesn't remember the trigger though 08:06
masak moritz: trying to reproduce the one-liner now
moritz: seems we've had some regressions thereabouts :P
this is how we find bugs! :P
eater I see alot of positive reinforcement in perl6 tho, which is nice :)
masak m: my $lang = "perl"; $lang++; say $lang
camelia perm
masak m: my $lang = "perl"; $lang++ for ^2; say $lang
camelia pern
eater say "eater".succ 08:08
evalable6 eates
eater :D
masak ah, here we go
m: my $lang = "perl"; $lang++ for ^45_565; say $lang
camelia ruby
masak spreads his arms and waits for moritz 08:09
moritz hugs masak
masak ...we've changed as a community :P
eater :') 08:10
masak this one works nowadays, too
m: my $lang = "perl"; $lang++ for ^46_002; say $lang
camelia rust
masak oh, and I hadn't discovered any regressions. I just wrote `for 45_565` first without the ^. PEBCAK.
moritz irclog.perlgeek.de/perl6/2007-07-17#i_64543 08:11
masak man, almost 10 years ago
moritz: I find it funny that I didn't make a big deal about it at all *at the time* :D 08:12
moritz: but it's later been promoted into "the time moritz double-minus'd me" :P
with a wink and a "no offence", no less
moritz must've been quite the traumatic experience :-) 08:14
eater m: my $lang = "perl"; $lang-- for ^108_067; say $lang 08:15
camelia java
masak moritz: "better have a thick skin in order to be part of our community" :P :P :P 08:16
eater++ :P
moritz: man, now I'm picturing someone being extremely hurt by what I just wrote and mentioning it to me ten years from now. 08:17
moritz masak: the emotional stress you're putting me through... :-) 08:18
masak huh, I seem to have a speech tick where I prefix my utterances with "man, " -- maybe I should write "humanoid, " instead
but that's a bit excluding wrt our bots 08:19
"entity, "
moritz or write "*, " instead
masak "splat, " 08:20
actually, referring to someone as "star" sounds like it has good connotations
"greetings, my stars"
m: my $lang = "perl"; $lang++ for ^13_568; say $lang 08:21
camelia pyth
masak m: my $lang = "perl"; $lang++ for ^64_592; say $lang 08:22
camelia swft
masak hm, what am I doing wrong here? I expected this to work: 08:25
m: multi infix:<**>(&fn, Int $n) { -> $val is copy { $val = fn($val) for ^$n; $val } }; say (&postfix:<++> ** 45_565)("perl")
camelia perl
masak ooh! 08:26
m: multi infix:<**>(&fn, Int $n) { -> $val is copy { $val = fn($val) for ^$n; $val } }; say (&prefix:<++> ** 45_565)("perl")
camelia ruby
moritz is surprisingly incompetent at building Debian packages. 3 instances of "worked on my machine, failed on build server" in a row 08:27
4 now :( 08:28
masak m: multi infix:<**>(&fn, Int $n) { -> $val is copy { $val.=&fn for ^$n; $val } }; say (&prefix:<++> ** 45_565)("perl") 08:29
camelia ruby
masak m: multi infix:<**>(&fn, Int $n) { -> $val is copy { $val.=&fn for ^$n; $val } }; say (*.succ ** 45_565)("perl") 08:41
camelia Cannot convert string to number: base-10 number must begin with valid digits or '.' in '3⏏5perm' (indicated by ⏏)
in block <unit> at <tmp> line 1

Actually thrown at:
in block <unit> at <tmp> line 1
masak oh, aha
it creates a WhateverCode of the whole expression, I guess 08:42
lizmat Files=1192, Tests=58608, 204 wallclock secs (12.41 usr 4.82 sys + 1212.94 cusr 114.51 csys = 1344.68 CPU) 08:55
jnthn haha, now I don't have to be jealous of spectest in 204 wallclock :P 09:01
The machine I'm sat on now can do them in 106 wallclock :)
lizmat hmmm 09:12
lizmat is jealous
is it a notebook ?
jnthn No :) 09:13
6-core Xeon 09:14
eater jnthn: how many Ghz? 09:17
jnthn 3.6 base, 4.0 turbo 09:19
eater hmm, mine 3.5 base, turbo 3.9 09:20
but has 8 cores :D
jnthn I thought about the 8 core one, but it was quite a bit more pricey than the 6 core one 09:21
eater jnthn: which one do you have?
jnthn E5-1650 v4
eater I have the E3-1241 v3
ah 12 threads?
jnthn Yeah
6 cores, 12 virtual 09:22
eater ah
yeah I have 4 cores, 8 virt
jnthn Ah, I thought you meant 8 cores :)
nine 106 seconds on a Ryzen 1800X. With the hope of becoming faster with a future BIOS update as the memory is not yet running at its rated speed. 09:23
eater nah, when I'm talking about cores Im mostly talking about threads
nine: :D
I wish I had the money to buy one
nine I guess a 1700X is not much slower and costs only ~ 320 Euros 09:24
eater haha
I need to replace my mobo and ram, it's a joke of 500e 09:25
nine yeah, there's that
eater currently running my desktop on a Phenom II X4 09:26
Geth rakudo/refactor-socket-encoding: 7508a1b3de | (Jonathan Worthington)++ | 2 files
Switch socket .get to use streaming decoder.

Therefore hopefully removing the last place that the VM-provided char I/O on sockets is used.		 09:43
jnthn That wasn't all that bad
Doing it for file handles will probably be more effort. 09:44
Geth nqp: 7da8da68ad | (Jonathan Worthington)++ | src/vm/moar/QAST/QASTOperationsMAST.nqp
Remove deprecated async str I/O ops.

They've been marked deprecated in MoarVM for a while, and not used in Rakudo for just as long. This paves the way for removing the code that is associated with them in MoarVM.		 10:04
Zoffix gives up backlogging 10:38
jnthn Zoffix: Here, of 10:39
*or #perl6?
jnthn gave backlogging #perl6 up long ago, but still kinda attempts here :)
Zoffix Here 10:40
I gave up #perl6 long time ago, though I see today it's pretty dead :)
eater dont have a bouncer Zoffix? 10:47
Zoffix eater: "bouncer"? what's that? 10:49
eater Zoffix: wiki.znc.in/ZNC
Zoffix eater: I have an easier solution: weechat running in screen on my server... 10:51
eater: but that doesn't solve the problem of having to read enormous amounts of text to keep up with the channel traffic :)
eater doesn't work well for me, since I need to be in different servers on different clients 10:52
yeah true
Zoffix Weird. Noticing some "duplicated" commits. Same message, date, and changes, but different shas. e.g. github.com/rakudo/rakudo/commit/3c...4cc43a0c67 and github.com/rakudo/rakudo/commit/01...0f34cc9e21 10:58
one is pointing to 2017.04.3 tag another to nom :\
Since people were bragging about their spectest times... I got everyone beat: Files=1192, Tests=58608, 78 wallclock secs (14.89 usr 2.44 sys + 1577.35 cusr 105.72 csys = 1700.40 CPU) 11:00
jnthn On your everyday work machine? Or on the awesome google cloud VM you fire up for fast spectests? :) 11:01
m: say 78 / 106
camelia 0.735849
Zoffix :D my awesome google cloud VM :) 11:02
jnthn Ah, OK :) 11:03
The 106 is on the machine sat next to me :) 11:04
Zoffix :)
jnthn A bit surprised that a it manages it in almost 3/4 of the time a 24-core of whatever monster does
*or
Righty, lunch time :) 11:05
bbl
Zoffix "Model name: Intel(R) Xeon(R) CPU @ 2.30GHz"
m: say 24/12
camelia 2
Zoffix Well, half the cores but more hurts is how I'm guessing:) 11:06
... ... did... did I just used a calculator to solve 24/12? ... 11:08
Zoffix blames it on too much blood in coffee veins
dogbert11 Zoffix, how long does a spectest take for you if you set TEST_JOBS to 16 or thereabouts? 11:12
Zoffix As opposed to current value of 30? 11:13
1 sec; doing bumps to get latest and greatest for release testing.
dogbert11 will be interesting 11:14
Zoffix BTW, the other day you hypothesized about too many open files being the reason for my test flops... unlikely as turns out my default limit is 65536 11:16
Wonder what these travis failures are about :S travis-ci.org/zoffixznet/perl6-Tes...tification 11:20
I do have `zef --debug install .` so I'd expect more output on proper failures
timotimo it doesn't use that 11:23
that part of the output is from rakudobrew build-zef
Zoffix ah
ZOFVM: Files=1242, Tests=135584, 117 wallclock secs (23.13 usr 3.21 sys + 2425.86 cusr 135.76 csys = 2587.96 CPU) 11:29
Geth nqp: de324a2d0f | (Zoffix Znet)++ | tools/build/MOAR_REVISION
Bump MoarVM

MoarVM bump brought commits:
  github.com/MoarVM/MoarVM/compare/2...8-g5f23324
5f23324 Remove asyncreadchars op and supporting code. 9d3ede0 Remove asyncwritestrto op and supporting code. 863e920 Remove asyncwritestr op and supporting code. 22fb8e4 don't crash on reading a closed dir handle
¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...8-g5f23324
rakudo/nom: 902fac5cec | (Zoffix Znet)++ | tools/build/NQP_REVISION
Bump NQP

NQP bump brought commits:
  github.com/perl6/nqp/compare/2017....8-gde324a2
de324a2 Bump MoarVM 7da8da6 Remove deprecated async str I/O ops. ... (25 more lines)		 11:30
¦ rakudo/nom: version bump brought these changes: github.com/perl6/nqp/compare/2017....8-gde324a2
nqp: 9c4d4586c0 | (Zoffix Znet)++ | t/nqp/019-file-ops.t
Test no SEGV when reading from closed dir handle

RT#131301: rt.perl.org/Ticket/Display.html?id=131301 Fix:
  github.com/MoarVM/MoarVM/commit/22...6815c8e526
  github.com/perl6/nqp/commit/de324a...57e3631f9e		 11:35
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=131301
Zoffix ZOFFLOP: t/spec/S17-promise/nonblocking-await.t 11:40
... make it twice 11:41
huh, weird. both harness6 and prove seem to be perfectly fine with nested subtests not correctly nested: gist.github.com/zoffixznet/3af63ed...677f8cdc95 11:43
TIL 11:44
almost makes me feel like I shouldn't bother fixing this bug in Testo :P
dogbert11: with TEST_JOBS=16: Files=1192, Tests=58608, 95 wallclock secs (13.65 usr 2.49 sys + 1126.09 cusr 92.65 csys = 1234.88 CPU)
dogbert11 Zoffix: isn't that better than before? 11:56
Zoffix m: say 95-78 11:57
camelia 17
Zoffix dogbert11: nope, 17s slower
dogbert11 was speculating about the default thread scheduler only having 16 threads and that adding more jobs might not be optimal
Zoffix this is harnes5 it don't use Perl 6 to schedule tests 11:58
dogbert11 is always using harness6 12:00
even though it's a bit slower than harness5 12:01
Zoffix It's twice as slow for me
dogbert11 so 'halving' the number of threads mad you lose ~20%
s/mad/made/ 12:02
perhaps a number around 20 is optimal from cost/performance standpoint
btw, there's a known bug in t/spec/S17-promise/nonblocking-await.t 12:03
i.e. github.com/MoarVM/MoarVM/issues/554 12:04
timotimo the second output is useless, that's just the global destruction not being careful about threads 12:05
so i wonder if the parametric 6model stuff is not threadsafe, and perhaps should be
dogbert11 timotimo: what kind of output would you like to have? 12:10
timotimo hmm. maybe the cross thread write log could help 12:11
if you get this segfault in gdb again, could you print out parts of the obj argument to MVM_repr_elems? 12:13
i wonder what this "<main_arena+88>" is supposed to tell us 12:14
hm, i wonder if this object here is corrupted 12:18
its REPR has a name of ""
dogbert11 let me give it a shot ...
timotimo what's the magic gdb command for "re-run until segfault" 12:21
actually, i'm not sure the crossthreadwrite log will even notice this, as it's happening from C code 12:22
brrt just 'run' 12:23
timotimo brrt: that doesn't re-run if it exits normally
brrt oh, right
timotimo i have to arrow-key-up + return a bunch of times
brrt that i don't know
timotimo ha! there is the segfault
it does look like a corrupt object's being put into that paramet.ric.lookup array 12:27
aha 12:28
finish_parameterizing has a "XXX handle possible race" comment
doesn't seem to break any more with a mutex added to it 12:34
dogbert11 have you already fixed it? 12:36
timotimo it may not be the fix jnthn wanted :) 12:37
dogbert11 :) let's see what he says 12:38
too late I guess, but now I have the SEGV in gdb 12:40
turns out that there are two tests which fails intermittently, test 8 and 11 12:41
timotimo pushed
dogbert11 when test 8 fails all I get is: 12:42
Unhandled exception in code scheduled on thread 3
This continuation has already been invoked
timotimo i didn't see that yet
i'll switch over to dayjob work 12:43
robertle jnthn: I was thinking a bit more about our discussion yesterday regarding "non-blocking await". this would be super cool, but i am wondering whether the fact that you are potentially running under another thread after await(), and the fact that this depends on whether the thread comes fomr a threadpool or is created explicitely is safe 13:07
either swapping the thread underneath your nose is always safe, in which case you can do it for all threads 13:08
or it isn't safe in all cases, in which that distinction doesn't really help due to composition problems:
if I write code in my module and do await, then I can't possibly know whether whoever is calling my module uses the main thread, starts one explcicitely or uses a pool 13:09
the place where the threading decision is made is not where the await() happens
so I call MyModule::some-stuff() and it works fine, and I then do that in a start { } block and weird shit happens *within* the module because it relies on the thread not being swapped 13:10
jnthn As I said yesterday, the cases where it matters will be when you're dealing with stuff *outside* of Perl 6 that cares about it, which essentially means "doing native calls". 13:12
robertle ah, ok I missed that part. so are you saying I call some native stuff, which uses TLS, then do await, then call native again? 13:13
which relies on being in the same thread
jnthn Right, then you could have an issue. But...if you're scheduling stuff on the thread pool you already don't know what thread it will be on.
So that problem already exists today. 13:14
robertle but currently you know that a sequence of statements will be run on the same one, and that's what I am wondering about
jnthn Right, that's what will change
robertle I think I just don't understand the use case where it would ever matter if the thread changes... 13:15
jnthn You'd pretty much have to be binding a library that uses thread local storage
A much more common situation is that you're binding a library that's not threadsafe and exposes an event loop 13:16
In which case you run a real Thread
robertle ok, the latter makes a lot of sense! regarding TLS, I am not sure whether the distinction between pool or not helps: 13:17
because your code would call native-stuff-with-tls-A() and later native-stuff-with-tls-B() and the problem is that these two need to happen in the same thread
so if you do that in your main code, you are safe because you can choose what sort of thread to run it under 13:18
but if your code that does these calls is in a module, then you may not...
but I do get your point that this only matters if you do something in native code, and that does help a lot because if you do then you need to be quite careful anyway... 13:19
jnthn Well, yes. I think it's on the people writing native library bindings to understand this stuff and make bindings that will not be fragile to such cases 13:20
And even then, in reality, the problem is more typically that the library is just not thread safe at all, or only is if you follow certain rules.
Zoffix .tell lizmat These commits [c1bd844][0e0ac2fb][0338ce46][14e09532][c61c7f88] fix RT #131241 for [Bag/Mix]Hash, but it appears the same problem exists in SetHash as well: `m: my $b = <a b b c c c>.SetHash; $_ = -1 for $b.values; dd $b` retains all the keys. Are you able to fix that case as well? 13:21
yoleaux Zoffix: I'll pass your message to lizmat.
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=131241
robertle sure. I also wonder whether the assumption that the thread does not change when calling into stuff is so deep in people's minds that they could get quite confused... 13:22
Zoffix .tell lizmat from changelog, the affected methods are .values/.iterator/.pairs/.kv
yoleaux Zoffix: I'll pass your message to lizmat.
robertle but this is super-powerful and exciting! I do look at systems that receive http calls and then make calls into other systems with retries and standoffs, collate the results and feed them back. these systems would benefit massively from this... 13:23
jnthn I don't think there's any silver bullets, but I think the improved scalability and lower risk of deadlocks from pool exhaustion will massively outweight the very small number of surprises that may occur. 13:25
C 13:26
oops
C# does the same thing, fwiw, and I can't think of a case where I got bitten. 13:27
Of course I understood the rules so could avoid getting bitten, but even then I struggle to recall needing to work around it.
robertle cool, really looking forward to this! 13:35
Zoffix FWIW non-blocking await is already available if you use `use v6.d.PREVIEW` pragma 13:36
m: await ^30 .map: {start await start sleep 1}; say now - INIT now 13:39
m: use v6.d.PREVIEW; await ^30 .map: {start await start sleep 1}; say "6.d: ", now - INIT now 13:40
camelia (timeout)
6.d: 2.0387999
jnthn It'll work even better if you use await Promise.in(1) :)
Zoffix m: use v6.d.PREVIEW; await ^16 .map: {start await Promise.in: 1}; say now - INIT now
camelia 1.0214228
Zoffix m: await ^16 .map: {start await Promise.in: 1}; say now - INIT now
cool. 13:41
camelia (timeout)
Zoffix m: use v6.d.PREVIEW; await ^3000 .map: {start await Promise.in: 1}; say "6.d: ", now - INIT now
camelia 6.d: 1.6653901
Zoffix <3 13:42
.ask pmurias I see the commits[^1] that look to be about removing too liberal str->num coersions, but this code still coerses them without error. Do we need to do more work to make this stuff fail to coerce on RakudoMoar? `m: my num $x = my str $ = '6 cute'; my num $y = my str $ = '!'; dd [$x, $y ];` [1] github.com/perl6/nqp/compare/2017....2-gc60df1e 13:51
yoleaux Zoffix: I'll pass your message to pmurias.
Zoffix Future-Zoffix: if we're adding ^ that to changelog, add: + Fixed native str to native int/num coersion accepting non-numeric strings [c776c087]
.tell lizmat seems there's inconsistency in behaviour introduced in 31be512 and 08b5c10 ; BagHash.pick throws when given NaN or -Inf $count, while BagHash.pickpairs equates those value to $count = 0 14:13
yoleaux Zoffix: I'll pass your message to lizmat.
Zoffix .tell lizmat just noticed: also for negative values: .pick treats them as Inf, while .pick as 0 14:14
yoleaux Zoffix: I'll pass your message to lizmat.
Zoffix yoleaux: help 14:15
.help
yoleaux Zoffix: I'm yoleaux. Type .commands to see what I can do, or see dpk.io/yoleaux for a quick guide.
Zoffix .in 24h Zoffix: ensure these (and and few lines up/below) issues are resolved/filed before release: irclog.perlgeek.de/perl6-dev/2017-...i_14607286 14:17
yoleaux Zoffix: I'll remind you on 20 May 2017 14:17Z
dogbert11 jnthn: did you see timo's recent MoarVM fixes? he was slightly worried that you might not find them 100% agreeable. 14:22
jnthn Yeah, I left a comment :) 14:24
dogbert11 :) 14:30
Zoffix .tell AlexDaniel something's wrong with commitable. I was doing a bunch of runs for `c: 2016.12 for ^1000_000 { }; say now - INIT now` and after doing a few it stopped and even after restarting it it no longer responds 14:46
yoleaux Zoffix: I'll pass your message to AlexDaniel.
Zoffix BTW: the reason I was doing ^ that is I noticed on 2016.12 Rakudo `my $x = 42.abs; my $y = 2.abs; for ^1000_000 { $ = $x }; say now - INIT now` is 2x faster than on HEAD :/ 14:47
m: for ^1000_000 { }; say now - INIT now
camelia 0.47756551
Zoffix star: for ^1000_000 { }; say now - INIT now
camelia 0.188090
eater oh
Zoffix m: say .47756551/.0188090
camelia 25.39026583
Zoffix m: say .47756551/.188090 14:48
camelia 2.539026583
Zoffix .ask lizmat `m: for ^1000_000 { }; say now - INIT now` is 2.5x faster on 2016.10 release than on current HEAD. You're the expert on this stuff... is that due to bugs fixed or something that's potentially fixable? 14:49
yoleaux Zoffix: I'll pass your message to lizmat.
Zoffix AlexDaniel: you have a robo message 15:05
AlexDaniel . 15:06
yoleaux 14:46Z <Zoffix> AlexDaniel: something's wrong with commitable. I was doing a bunch of runs for `c: 2016.12 for ^1000_000 { }; say now - INIT now` and after doing a few it stopped and even after restarting it it no longer responds
AlexDaniel :O
AlexDaniel slaps commitable 15:07
oh wait, it's not here
AlexDaniel slaps committable6 15:08
c: 2014.01 say 42
committable6 AlexDaniel, ¦2014.01: «42»
AlexDaniel c: 2016.12 say 42 15:09
committable6 AlexDaniel, ¦2016.12: «42»
Zoffix \o/
AlexDaniel why the f does it happen though
I mean, it failed to delete extracted build
so when you tried it again it was like “wait, I see it's already there, I think another bot is using it right now… it'll probably delete this build soon so I'll just wait here for my turn” 15:10
committable6: why did you leave it there in the first place? 15:11
committable6 AlexDaniel, ¦why: «Cannot find this revision (did you mean “all”?)»
AlexDaniel e: say 42
evalable6 42
AlexDaniel /o\ 15:12
Zoffix: in case it comes up again, the solution is: rm -rf /tmp/whateverable/rakudo-moar/b2a3441749878e338b0861b14b3b9433cc902f42/ 15:15
Zoffix will try to remember that 15:16
AlexDaniel: is it always the same hash?
Zoffix giggles at title of github.com/rakudo/rakudo/commit/70...9de7a700da
AlexDaniel Zoffix: for 2016.12, yes
Zoffix ZofBot: the IO gran! She's old, but tech savvy!
ZofBot Zoffix, ␤ in block <unit> at <tmp> line 1␤␤» m: class { multi method x ($l = *, :$close) {say "here"}; multi method x (:$close) { say "there" } }
Zoffix Oh, turns out I already have that in my notes.... 15:17
<AlexDaniel> now, the problem we have right now is that some of them may leave some crap behind
<AlexDaniel> especially if it segfaulted
<AlexDaniel> so you'd probably need to do something like this: rm -rf /tmp/whateverable/rakudo-moar/*
AlexDaniel /* is a bit too radical given that right now we are building stuff :)
Zoffix AlexDaniel: so I can just rm -fr /* ? No need to look for a hash?
Ah. OK
AlexDaniel so it will have some folders where new builds are actually being… built :) 15:18
Zoffix reminds eater to write docs for Proc::Async.ready 15:22
Zoffix wants a `make` target in Rakudo that does Step 12 of release guide: clone zef, setup paths, and install Inline::Perl5 15:36
jnthn NeuralAnomaly: status 15:41
NeuralAnomaly jnthn, [✘] Next release is today. Since last release, there are 37 new still-open tickets (0 unreviewed and 2 blockers) and 0 unreviewed commits. See perl6.fail/release/stats for details
jnthn Will do the MoarVM release this evening 15:42
AlexDaniel (if anybody wants to write a workaround for whateverable issue, here is the ticket: github.com/perl6/whateverable/issues/144 ) 15:47
Geth rakudo/nom: 164c0b5c2d | (Zoffix Znet)++ | docs/ChangeLog
Log all commits to date

Adds commits: 01dd2138 0338ce46 05479793 06d8800e 07feca67 08a8075f 08b5c101 09506dc8 0bd39de2 0c46aff2 0e0ac2fb 0f21f511 12c50b63 12cec7a7 134efd83 13924397 14568496 146f3a39 14e09532 1562da07 18706852 1a920dcc 1b0e41f9 1c21974d 1c4d8451 1cf7ccba 1d6a0023 1e58925c 1ed76a90 204ea59b 226cd8b6 22bd2bbd ... (20 more lines)		 15:48
Zoffix ????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????
REMINDER: Rakudo 2017.05 will be released this Saturday. Please check
docs/ChangeLog to ensure your work has been correctly recorded.
????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????
timotimo whoa! 15:50
life comes at you fast, bro
AlexDaniel yeah, I just blinked and there's a new release coming 15:54
didn't we have all these release issues yesterday? :P 15:55
timotimo and how
Zoffix .tell ugexe latest `zef --debug test` puts an extra blank line after each test. The issue got introduced somewhere in github.com/ugexe/zef/compare/23e80f6...HEAD 16:03
yoleaux Zoffix: I'll pass your message to ugexe.
ugexe github.com/ugexe/zef/blob/master/l...m6#L26-L36 probably this adding the extra \n (the .emit basically goes to a .tap: { .say }) 16:08
yoleaux 16:03Z <Zoffix> ugexe: latest `zef --debug test` puts an extra blank line after each test. The issue got introduced somewhere in github.com/ugexe/zef/compare/23e80f6...HEAD
Zoffix huggable: modules :is: 100 Most Popular (by github stars) Modules: perl6 -MWWW -e 'jget("modules.perl6.org/.json")<dists&....grep(none <p6doc panda>).head(100).say' 16:10
huggable Zoffix, Added modules as 100 Most Popular (by github stars) Modules: perl6 -MWWW -e 'jget("modules.perl6.org/.json")<dists&....grep(none <p6doc panda>).head(100).say'
Zoffix Gonna use ^ that list as a yard-stick for release testing for now... until we get a bot doing this daily
ugexe mayve check last commit/update too 16:13
Zoffix "===> Failed to find dependencies:"... OK, Plan B :) 16:19
github.com/ugexe/zef/blob/master/l...m6#L26-L36 got an `$out` but it never uses it... 16:28
ugexe ah, yeah i moved that stuff to Zef::Report so not needed 16:30
Zoffix this solves the issue and passes t/ tests: github.com/ugexe/zef/pull/163 16:34
And xt/ tests seem to be wonky even without this patch 16:35
"not ok 3 - Ecosystems => cpan" and unhandled failures in destroy
full output: gist.github.com/zoffixznet/5bd101e...7e9ea4bf6b
ugexe eh probably cause i just emptied the cpan dist list 16:36
unhandled failures are from deleting temp files i think... they always happen
so is ok 16:37
dunno if .chomp is right, since print doesnt have to end with a \n 16:38
e,g, probbly should do .tap: {.print} instead of .say 16:39
evalable6 ugexe, Full output: gist.github.com/d5bed44d0003be4fc6...e1fc6f5c31
(exit code 1) 04===SORRY!04=== Error while compiling /tmp/TT0gKnM6oR
Strange…
Zoffix huggable: modules 16:44
huggable Zoffix, 100 Most Popular (by github stars and recently updated) Modules: perl6 -MWWW -e 'jget("modules.perl6.org/.json")<dists&...dated>) after Date.new: "2017-01-01"}).head(100)»<name>.grep(none |<p6doc panda v5>, /^ "Inline::"/).put'
Zoffix Well, I'm up to that ^ but still doesn't want to zef --serial install install them :/ Dies with "Failed to find some required dependencies", but doesn't tell which and I grepped the ecosystem for all the modules it says " Searching for missing dependencies:" and they're all in ecosystem :/ 16:45
Changed say to print in tap: github.com/ugexe/zef/pull/164 16:49
re-tried the 100-module install... seems to be moving now... 16:51
"No extracting backend available" ... ok... Plan E... 16:57
""source-url" : "github.com/araraloren/Getopt-Kinok...2.9.zip"," 16:58
ugexe works for me, but I have the unzip command available 16:59
japhb ugexe: Does it only try that particular command? 17:01
ugexe depends 17:02
on windows there is a powershell unzip
so if the unzip command wasnt available it would make it to that
probably should just create a zip version of github.com/ugexe/zef/blob/master/r...erl5tar.pl 17:03
Zoffix Yeah, it works if I have unzip installed, but I didn't by default 17:13
ugexe hmm maybe your first PR with .chomp is the technically correct one. I will have to test with other backends to see which pr gets more consistent output, but thanks 17:28
for once windows comes out ahead 17:29
[Coke] one thing we should probably do is start bumping nqp/moarvm releases in advance of the release instead of waiting. Most of the time it's fine, but seems like nom HEAD should have the updates asap (which isn't immediately) 17:56
Zoffix [Coke]: I genereally do the bumps when I do my pre-release testing, like today 18:03
we can have a bot auto-bump daily too, I guess
timotimo then we'll have to be much more branchy in nqp and moarvm 18:11
Zoffix kinda thought that was already the case 18:15
i.e. nqp/moar are kept in the same "spectest is clean" state as nom
rather than having some partial work in it
jnthn We generally do try to keep "master is shippable" discipline 19:01
timotimo not always, though. like when i was dumb enough to merge telemeh before it was readý 19:04
jnthn Well, yes, it only works as well as people committing make it work :) 19:09
Zoffix bah... my super twittable trick of `but`ing a Promise to attach a tag to it failed :( 19:25
m: my @s = start { sleep 1; say "one"; }, start { sleep 2; say "two" }; await Promise.anyof: @s;
camelia one
Zoffix m: my @s = (start { sleep 1; say "one"; } but role {}), (start { sleep 2; say "two" } but role {}); await Promise.anyof: @s;
.anyof seems not to like butted promises
s: Promise, 'anyof', \(start {})
SourceBaby Zoffix, Sauce is at github.com/rakudo/rakudo/blob/164c...se.pm#L231
camelia (timeout)one
two
jnthn I'm not surprised 19:27
but does a clone
timotimo hah, of course
jnthn does would be more suitable
Zoffix Ohh, right
m: my @s = (start { sleep 1; say "one"; } does role {}), (start { sleep 2; say "two" } does role {}); await Promise.anyof: @s; 19:28
camelia one
Zoffix yey
jnthn++
man, forgot role is not a closure 19:43
eater can I note that the docs around Proc::Async.start are very confusing 19:44
Zoffix m: my @s = ^1 .map: -> $m { start { sleep 1; say "one"; } does role { method Name { $m } } }; await Promise.allof: @s; @s».Name.say 19:45
camelia one
[0]
Zoffix eater: well-volunteered! :) BTW, reminder to add Proc::Async.ready docs. Release soon will ship and it's in the changelog and people will look it up 19:46
eater Zoffix: ye :p that's why Im noting that
building and checking now if it's any good
Zoffix m: my @s = <foo bar ber>.map: -> $m { start { } does role { method Name { $m } } }; await Promise.allof: @s; @s».Name.say
camelia [ber ber ber]
Zoffix m: role Foo[$m] { method Name { $m } }; my @s = <foo bar ber>.map: -> $m { start { } does Foo[$m] }; await Promise.allof: @s; @s».Name.say 19:51
camelia [foo bar ber]
Zoffix win
eater m: my $d = Proc::Async.new(<bash -c>, "exit 1"); $d.ready.then({ say "Kept! "; }); try sink await $d.start; say $d.ready.status 19:53
camelia Proc::Async is disallowed in restricted setting
in sub restricted at src/RESTRICTED.setting line 1
in method new at src/RESTRICTED.setting line 32
in block <unit> at <tmp> line 1
eater :(
anyway this results in "Kept! ␤Broken" 19:54
A promise can be broken after it has been Kept?
Zoffix eater: .then starts a new promise that runs after the first one finished, regardless of whether it was Kept or Brokenb 19:55
m: await start { die }.then: {say "Live!"}
camelia Live!
Zoffix m: await start { die }.then: { say .status } 19:56
camelia Broken
Zoffix m: await start { 42 }.then: { say .status }
camelia Kept
eater ah
thanks Zoffix :)
or wait! 19:58
eater++
eh
Zoffix++
[Tux] This is Rakudo version 2017.04.3-291-g164c0b5c2 built on MoarVM version 2017.04-68-g5f233249 20:13
csv-ip5xs 2.491
test 12.723
test-t 4.258 - 4.289
csv-parser 12.727
japhb huggable: speed 20:19
huggable japhb, tux.nl/Talks/CSV6/speed4.html
japhb buggable: speed
buggable japhb, ↑↑▅▄▄▇█▄▄▇▆▃▄▃▂▄▃▃▃▃▃▂▂▂▂▂▂▂▂↑▂▁▁▃▁▁▂▁▁▂▁▁▃▃▃▃▃▃▂▂ data for 2017-04-30–2017-05-19; range: 4.194s–6.181s; 25% faster
Zoffix buggable: speed 20 20:20
buggable Zoffix, ▂▁▂▃▁▁▂▁▁▂▁▁▄▃▃▃▃▄▂▃ data for 2017-05-12–2017-05-19; range: 4.194s–4.352s; 1% slower
jnthn Yes, .then is called even on a Broken Promise
The idea being that you'll usually call .result in there
Zoffix
.oO( there are lies, damned lies, and buggable... )
jnthn And that will rethrow the exception
Geth rakudo/nom: a61746feda | (Zoffix Znet)++ | src/core/Promise.pm
Make Promise subclass-friendly

Instead of hardcoded `Promise`, use `self` when instantiating new Promises, that way if we're a subclass, we instantiate a sublcass instead of the original `Promise`.		 20:28
Zoffix grrr... someone's grinding concrete upstairs... What a pleasant environment to hack in >_< 20:36
5pm before a long weekend... go home... 20:37
m: await Promise.in(-0.001).then: {say "done"} 20:38
camelia (timeout)
Geth roast: 7e9451614b | usev6++ | 2 files
[JVM] X::Parameter::InvalidConcreteness NYI
roast: 225d12ba80 | usev6++ | 7 files
[JVM] Fudge some newly added tests
roast: 105e18043b | usev6++ | S16-io/comb.t
[JVM] Two new failing tests, one passer for comb
Zoffix ZofBot: engage Anti-Concrete-Grinding-Device! Blast it, baby: www.youtube.com/watch?v=mxDiKSHqBIc 20:40
ZofBot Zoffix, new(2, "2")),)
Zoffix hm, this one's better: www.youtube.com/watch?v=7EvOTkEMlug 20:42
Geth rakudo/nom: a7c23aa214 | (Zoffix Znet)++ | src/core/Promise.pm
Make .then on completed Promise subclass-friendly
 21:04
roast: 926f20d109 | (Zoffix Znet)++ | S17-promise/basic.t
Test Promise is subclass-friendly

Rakudo fix: github.com/rakudo/rakudo/commit/a61746feda
   github.com/rakudo/rakudo/commit/a7c23aa214
lizmat . 21:34
yoleaux 13:21Z <Zoffix> lizmat: These commits [c1bd844][0e0ac2fb][0338ce46][14e09532][c61c7f88] fix RT #131241 for [Bag/Mix]Hash, but it appears the same problem exists in SetHash as well: `m: my $b = <a b b c c c>.SetHash; $_ = -1 for $b.values; dd $b` retains all the keys. Are you able to fix that case as well?
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=131241
yoleaux 13:22Z <Zoffix> lizmat: from changelog, the affected methods are .values/.iterator/.pairs/.kv
14:13Z <Zoffix> lizmat: seems there's inconsistency in behaviour introduced in 31be512 and 08b5c10 ; BagHash.pick throws when given NaN or -Inf $count, while BagHash.pickpairs equates those value to $count = 0
14:14Z <Zoffix> lizmat: just noticed: also for negative values: .pick treats them as Inf, while .pick as 0
14:49Z <Zoffix> lizmat: `m: for ^1000_000 { }; say now - INIT now` is 2.5x faster on 2016.10 release than on current HEAD. You're the expert on this stuff... is that due to bugs fixed or something that's potentially fixable?
Zoffix :}
lizmat just back from a Guardians of the Galaxy / Alien Covenant marathon 21:35
fwiw, I think the SetHash behaviour has been around a long time already 21:37
and I'm not clearminded enough atm to attempt to fix that (although I think I know how) 21:38
how much time do I have before release ?
Zoffix m: say (DateTime.new('2017-05-20T11:00:00-04:00') - DateTime.now.in-timezone(-4*3600)) / 3600 21:41
camelia 17.3136951431377
Zoffix lizmat: that many hours, but I'm flexible.
lizmat well, I'm gonna need at least 9 hours of sleep 21:42
MasterDuke Zoffix, lizmat: graph for `for 1000_000 { }` gist.github.com/Whateverable/36a4c...aaf82c39c7 21:46
Zoffix cool
c: c069f45 for ^1000_000 { }; say now - INIT now 21:48
committable6 Zoffix, ¦c069f45: «0.5158494»
Zoffix c: c069f45~1 for ^1000_000 { }; say now - INIT now
committable6 Zoffix, ¦c069f45~1: «0.5253575»
Zoffix c: c069f45~10 for ^1000_000 { }; say now - INIT now
committable6 Zoffix, ¦c069f45~10: «0.4419105»
lizmat $ 6 'for 1000000 {}; say now - INIT now' 21:49
0.0015960
Is my machine so much faster ?
MasterDuke c: bfbf348,c069f45 for ^1000_000 { }; say now - INIT now
committable6 MasterDuke, ¦bfbf348: «0.171921» ¦c069f45: «0.5364211»
MasterDuke lizmat: missing a `^`
japhb lizmat: You forgot ^
lizmat ah, duh 21:50
see, I'm in no shape to be programming :-) 21:51
Zoffix :)
c: c069f45,87f61d9,18e6f0d,f2b97b0,6c92994,ab3162c,7384939,9f15a4d,bc53f67,97359ae,951a441,9493ffb,d69f375,b9d9279,0152316,9ed4449,f8b3469,5401a1a,f94cb21,4366603,f22170f,a2d69a0,ee7c1bb,3e28b1e,87d40ab,ac37a20,084cae1,9086216 for ^1000_000 {}; say now - INIT now 21:52
lizmat reverting c069f4598b0a8a51a9 to see if that makes a difference, don't think it will but...
committable6 Zoffix, gist.github.com/8fa9999eb3502a7c79...836f88a2d3
Zoffix c: 5401a1a for ^1000_000 { }; say now - INIT now 21:53
committable6 Zoffix, ¦5401a1a: «0.4943332»
Zoffix c: 5401a1a~1 for ^1000_000 { }; say now - INIT now
committable6 Zoffix, ¦5401a1a~1: «0.1777426»
Zoffix Looks to be due to 5401a1a
lizmat points to jnthn :-) and indeed looks like we lost the ^N optimization again 21:54
jnthn Again? Urgh
lizmat decides she is too tired to look at the Baggy / Setty stuff now
will check in the morn 21:55
nighty night!
jnthn That opt is fragile as heck...
'night lizmat
Zoffix Filed as rt.perl.org/Ticket/Display.html?id=131330 21:56
MasterDuke++ nice bots :)
MasterDuke thanks 21:57
Zoffix m: ^2 .grep: {try 1 after 0} 22:23
camelia WARNINGS for <tmp>:
Useless use of "after" in expression "1 after 0" in sink context (line 1)
Zoffix Filed as rt.perl.org/Ticket/Display.html?id=131331 22:24
hm. looks like something broke my `M` module I made in Feb :/ travis-ci.org/zoffixznet/perl6-M 22:31
18% of top 100 modules fail :( Tested with my zefyr.p6 tool: gist.github.com/zoffixznet/5909465...7fb4915598 22:34
s: [], 'head', \()
SourceBaby Zoffix, Sauce is at github.com/rakudo/rakudo/blob/a7c2...s.pm#L1852
Zoffix I think it needs `is nodal` 22:35
Failures in top 100: Bailador Crust November Digest Frinfon Slang::SQL List::Utils XML::Parser::Tiny ANTLR4 Perl6::Maven GraphQL MagickWand WebSocket IO::Prompter BioInfo Flower Plosurin MongoDB 22:36
Zoffix builds 2017.03 to see if numbers change 22:37
Actually strike Digest from the list. I know a fix for it 22:40
ZofBot: ♫ everyone's asleep ♫ it's just youuuuu and meeee ♩ tood-todoooo ♩ tad-tadaaaaa 22:44
ZofBot Zoffix, 10 (Debian) Content-Length: 301 Connection: close Content-Type: text/html; charset=iso-8859-1 <!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 2
Zoffix ZofBot: HTML 2?! What kind of places do you visit! BAD ROBOT! NO COOKIE! 22:45
ZofBot Zoffix, (This message differs from what "fail", "warn", and "die" would say by default, since the latter operators typically point out bad data or programming rather than just an incomplete design
Zoffix holy crap... I'm having trouble even counting the number of spectests that fail on Windows :| gist.github.com/zoffixznet/856593c...16ffe14809 22:50
Ah, ok. Good news. Many of them are due to apparently-busted `make-temp-file` routine 22:51
or crap.. bad news, make-temp-file is likely busted due to some IO unholiness :S "Failed to open file C:\C:\Users\zoffi\AppData\Local\Temp\perl6_roast__e_line1_0_1603368517871361495234412: no such file or directory" 22:53
'C:\' x 2 :/
Well. Glad I ran the tests on Win32. 'cause I can fix this before release \o/ 22:54
MasterDuke yay for testing
Zoffix hm; ok. it's `.resolve` stuff. Won't fix for release. Was broken since Christmas 22:59
well, `WWW` that zefyr.p6 needs won't install on 2017.03 :| 23:00
Zoffix tries an earlier-still release 23:01
Geth roast: b5b975b7d8 | (Zoffix Znet)++ | packages/Test/Util.pm
Temporarily remove .resolve for Windows in make-temp-*

Currently Rakudo's IO::Path.resolve isn't Windows-friendly; should might be fixed in a month or so.		 23:12
Zoffix Hm... JSON::Unmarshal appears to fail on Rakudos < 2017.04 or so. And that module is used by META6 23:25
Zoffix feels stupid making Test::META a mandatory, user-facing test now

Powered by Cro and the Raku® Programming Language.

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

Thank you!