psch Zoffix: my RT email is the same as my github 00:21
Zoffix: re the reduce bit: in the grammar it looks easy enough to throw a typed ex; we have an explicit <!> there which could likely be a self.typed_panic 00:22
Zoffix: the whole "can't have space" seems to be just an artifact of how infixish is implemented from what i can see right now -- it is kinda late though, and i didn't look too closely 00:24
...nevermind there's a look-ahead that's looking for \S+
the typed_panic might still work...
Zoffix psch: you should have bugadmin rights now 00:28
But do we have the <!> there so that Array constructor would be matching instead? So if you make it typed panic, this would begin to throw: `say [[&({($^a.uc) ~ ($^b.lc)})]], "foo", "BAR", "baz"` 00:30
psch oh duh, of course 00:31
...it *is* late >_>
right, <!> means "stop this rule," not "fail parsing"
Zoffix & 00:32
TimToady hmm, so how come my $*OUT.flush isn't flushing? 01:26
lizmat Files=1233, Tests=76126, 305 wallclock secs (14.75 usr 5.09 sys + 2091.00 cusr 202.29 csys = 2313.13 CPU) 08:09
yoleaux 00:09Z <timotimo> lizmat: almost all the work from that commit was actually done by nine, i assume he's jusf flattened the commits and it somehow kept me as the author?
lizmat and another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2018/01/29/...s-modules/ 08:32
(oops, forgot to clickbait) 08:33
m: LEAVE say now - ENTER now # whee! 09:11
evalable6 (exit code 1) ===SORRY!===
Cannot reference undeclared local 'enter_result__1'
Zoffix TimToady: is it the default $*OUT? It's not buffered by default. 09:24
Zoffix would've expected `$*OUT.out-buffer = 10000; print "meow"; sleep` to buffer, but it doesn't :/ 09:26
Ahhh.. The first one isn't buffered. Now I remember 09:28
`$*OUT.out-buffer = 10000; print "meow"; print "foos"; $*OUT.flush; sleep` works for me. the `"foos"` gets buffered and flush does flush it 09:29
Geth: what was up with that, brother? 09:33
ZOFVM: Files=1287, Tests=153130, 143 wallclock secs (20.42 usr 3.15 sys + 3037.94 cusr 154.28 csys = 3215.79 CPU)
Geth nqp: aaf5b7865c | (Zoffix Znet)++ | tools/build/MOAR_REVISION
[MoarVM Bump] Brings 18 commits

MoarVM bump brought: github.com/MoarVM/MoarVM/compare/2...9-g783a4f0 783a4f0 don't output PHI in expr jit log 8f194be better error in boxed ref for repr_id mismatch c437b79 Merge pull request #790 from MasterDuke17/add_even_more_jit_templates 3055deb Add tiles for binary XOR and NOT ... (14 more lines)
09:34
¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...9-g783a4f0
rakudo: 4d2b54ebe8 | (Zoffix Znet)++ | tools/build/NQP_REVISION
[NQP Bump] aaf5b78 [MoarVM Bump] Brings 18 commits

NQP bump brought: github.com/perl6/nqp/compare/2018....0-gaaf5b78
MoarVM bump brought: github.com/MoarVM/MoarVM/compare/2...9-g783a4f0 783a4f0 don't output PHI in expr jit log 8f194be better error in boxed ref for repr_id mismatch ... (16 more lines)
09:35
rakudo: version bump brought these changes: github.com/perl6/nqp/compare/2018....0-gaaf5b78
58de239cc9 | (Zoffix Znet)++ | src/Perl6/World.nqp

Phixes github.com/rakudo/rakudo/issues/1455
The combination of phasers results in ENTER trying to access an out of scope local var. Fix by using a lexical var instead when constructing the QAST. The static optimizer lowers it to a local for us, whenever that's possible.
lizmat github.com/rakudo/rakudo/issues/1455 # LEAVE say now - ENTER now 09:38
fwiw, these commits don't seem to affect test-t much 09:39
Zoffix grabs it 09:40
jnthn About $*OUT and buffering: it's not buffered if we detect that it's a TTY 10:07
Otherwise it is
Zoffix jnthn: how come some blocks do this thing where they declare a lexical var and a local param and then bind the param to the lexical. Is that beneficial over than just declaring a lexical param? gist.github.com/zoffixznet/8091ec4...-p6-L6-L14 10:10
That's from a dump of `1 andthen 2 andthen 3` 10:11
jnthn Yes
Generally, we want it in a normal register while we enforce all the various checks 10:12
(type, concreteness, etc.)
And only then to bind it into a lexical 10:13
Zoffix Ah. Thanks.
jnthn Reason being that lexicals are more difficult for something like spesh to reason about
Because it has to assume (without doing further, more costly/difficult proofs) they're visible and changeable from something outside of the code it's reasoning about. 10:14
Zoffix Makes sense.
AlexDaniel squashable6: next 10:22
squashable6 AlexDaniel, ⚠🍕 Next SQUASHathon in 2 days and ≈23 hours (2018-02-03 UTC-12⌁UTC+14). See github.com/rakudo/rakudo/wiki/Mont...Squash-Day
Zoffix nine: camelia seems ded. I see the script is running, but can't ping irc.freenode.net from its box 10:39
13 packets transmitted, 0 received, 100% packet loss, time 12096ms 10:40
ZOFFLOP: t/spec/S17-procasync/basic.t (test #39; status code 1) 10:43
roast: 8e4f5edff7 | (Zoffix Znet)++ | S04-phasers/enter-leave.t
Test nested ENTER/LEAVE does not crash

Closes github.com/rakudo/rakudo/issues/1455 Rakudo fix: github.com/rakudo/rakudo/commit/58de239cc9
10:47
jnthn Ah, it was generating a local. Interesting.
Zoffix :)
dogbert17 m: for ^2 { for ^2 { say "{ (state $a)++ }" } } 11:22
evalable6 0
0
0
0
dogbert17 m: for ^2 { for ^2 { say (state $a)++ } } 11:23
evalable6 0
1
0
1
|Tux| Rakudo version 2018.01-58-g58de239cc - MoarVM version 2018.01-49-g783a4f07c
csv-ip5xs0.957 - 0.958
csv-ip5xs-2010.772 - 11.100
csv-parser11.105 - 11.561
csv-test-xs-200.445 - 0.475
test9.157 - 9.240
test-t2.515 - 2.548
test-t --race1.054 - 1.115
test-t-2045.186 - 46.055
test-t-20 --race15.294 - 15.426
11:43
Zoffix m: use nqp; my $out := none True; my $p := 0; if $p == 42 or $out ~~ Failure and not $out.handled {}
evalable6 (exit code 1) No such method 'handled' for invocant of type 'Bool'
in block <unit> at /tmp/0UxH6Z07r2 line 1
Zoffix I wonder how many bugs in core we have due to that ^ :)
The assumption is a shortcurcuit would happen when condition is false, but the condition is true if a `none` junction is used with a false condition
There's one in REPL. 11:45
m: 'meow'.say with none Any 11:46
evalable6 meow
dogbert17 Zoffix: thx for closing RT#127135 11:48
synopsebot RT#127135 [resolved]: rt.perl.org/Ticket/Display.html?id=127135 [BUG] Infinite Loop with missing second arg on .subst-mutate
Zoffix Filed a code-review issue R#1456 11:53
synopsebot R#1456 [open]: github.com/rakudo/rakudo/issues/1456 [META] Review codebase for bugs with `none` Junctions
Zoffix Was my `where` clause optimizations ripped out? I thought I did something with `-> $foo where Bar {}` but doesn't seem to be applied any more 12:00
They were probably buggy when Junctions where involved
Ah, nm, still there: github.com/rakudo/rakudo/blob/mast...2181-L2222 doesn't seem to be applied when I would've expected it to 12:01
Yeah, I fucked it up 12:04
c: 2018.01,HEAD my $out := one True; -> $ where Bool|Int { say "in" }($out)
committable6 Zoffix, ¦2018.01: «in␤» ¦HEAD(58de239): «Internal error: inconsistent bind result␤ in block <unit> at /tmp/88QCzhz6Nw line 1␤␤ «exit code = 1»»
Zoffix (post-release branch)++ :) 12:05
Geth nqp: 09e59a12ae | pmurias++ | src/vm/js/nqp-runtime/runtime.js
[js] Make matching with :i and :m handle graphemes better
12:09
nqp: 9bce90bb80 | pmurias++ | src/vm/js/nqp-runtime/unicode-data/NumericValue.trie
Add missing file
nqp: 84a8a97e10 | pmurias++ | 6 files
[js] Move preprocessed UCD data into a seperate npm module

Keep less generated stuff in the repo
rakudo: 7f07f68fdf | (Zoffix Znet)++ | 2 files
Fix incorrect routine name in nodemap

When throwing X::Cannot::Lazy
12:26
rakudo: 2c36ab2ef9 | (Zoffix Znet)++ | 2 files
Fix wrong error in REPL with none Junctions

Fixes RT#132780: rt.perl.org/Ticket/Display.html?id=132780
When the return of a statement in REPL is a `none` Junction, it'll go through the `$out ~~ Failure` and actually result in True when none of the objects in the `none` Junction are a Failure, so we end up calling .handled on the wrong object (subject of Issue
  github.com/rakudo/rakudo/issues/1456 )
Fix by using `nqp::istype` instead that doesn't descend into Junctions.
12:30
synopsebot RT#132780 [new]: rt.perl.org/Ticket/Display.html?id=132780 Issue with one and none Junctins when passed as parameter
nqp/master: 4 commits pushed by pmurias++ 12:48
roast: 115be05bb9 | (Zoffix Znet)++ | S02-literals/quoting.t
Work around highlights explosion in Atom

Issue github.com/perl6/atom-language-perl6/issues/78
13:26
Zoffix Looking at rt.perl.org/Ticket/Display.html?id...et-history In `qq` the `\Y would throw; in `q` it would become `Y`. In `Q:b` `\Y` stays as `\Y`, and `\$foo` stays as `\$foo`. But what's the behaviour if we do `Q:b:s`? The `\$foo` would need to become `$foo` to match user's expectations, I think. It seems like "unrecognized backslash sequence" should be a property of `b1` tweak rather than 13:48
`qq` quoter (and it'd attain it since it `does` b1)
Currently spectesting that change, to see if there's any failign tests.
so in Q:b `\Y` would throw 13:51
ZOFFLOP: t/spec/S10-packages/basic.rakudo.moar (test 83; status 1) 14:05
ZOFFLOP: t/spec/S12-attributes/class.t (no plan; status 1)
ZOFFLOP: t/spec/S17-procasync/basic.rakudo.moar (test 39: status 1)
stresstest is clean and it makes sense to me for these to be on :b so I'm pushing :)
dogbert17 Zoffix++ 14:23
Geth rakudo: deffe54b8d | (Zoffix Znet)++ | src/Perl6/Grammar.nqp
Change semantics of :b quoter tweak

Fixes 127226: rt.perl.org/Ticket/Display.html?id=127226
Make it more aggressive and behave like backslashes behave in `qq` quoter: throwing on any unrecognized backslash sequences and giving the \W itself when an unknown \W is backslashed.
This makes Q:b combined with other tweaks, e.g. :s, behave more consistently in that both `Q:b/\$foo/` and `Q:b:s/\$foo/` would result in string `$foo`.
14:29
roast: 284a97d31d | (Zoffix Znet)++ | S02-literals/quoting.t
Spec Q:b more thoroughly

RT#127226: rt.perl.org/Ticket/Display.html?id=127226 Rakudo impl: github.com/rakudo/rakudo/commit/deffe54b8d
synopsebot RT#127226 [open]: rt.perl.org/Ticket/Display.html?id=127226 QUOTE does not remove backslash with adverb :b
Geth roast: 9721b31af4 | (Zoffix Znet)++ | S02-literals/quoting.t
Test :b lets one escape delimiters
15:10
lizmat travel& 16:58
stmuk pl6anet.org/drop/rakudo-star-2018.0...0(JIT).msi 17:26
macOS build to follow .. I'm trying to persuade virtualsue to do it
32 bit windows msi donations gratefully received :) 17:27
nine ZofBot: camelia is back but I do not have a clue about why it couldn't reach the network. Packets were going out and were properly NATed. They came back but never made it onto the virtual network. They also did not appear in iptables even with LOG targets before any of the rules. I installed (kernel) updates and rebooted the host. 18:30
ZofBot nine, IIRC my CLA approval took around 6 weeks?
nine Zoffix: camelia is back but I do not have a clue about why it couldn't reach the network. Packets were going out and were properly NATed. They came back but never made it onto the virtual network. They also did not appear in iptables even with LOG targets before any of the rules. I installed (kernel) updates and rebooted the host.
moritz: ^^^
moritz nine: thanks 18:57
TimToady since rakudo does not yet detect whether stdin is a socket, I'm using NativeCall with recv: 19:02
my ssize_t sub recv(int32 $sockfd, buf8 $buf, size_t $len, int32 $flags) is native {*}
but this seems to leak somehow: 19:03
whenever Supply.from-list(gather while recv(0,$buf,bufsize,0) -> $bytes { take $buf.subbuf(0,$bytes).decode }) -> $_ is copy { ... }
(inside a react)
is there anything in there that would imply a .cache?
or is there a shorter way to write that? 19:04
anyway, that whenever seems to grow more than half a gig overnight 19:07
(processing syslog events from my DSL modem so I can do caller id :)
((which is a really cool way to do it, 'cuz you get your caller id before it even rings the first ring, instead of waiting till after the first ring)) 19:08
also, the corresponding P5 program doesn't leak, and there's nothing in my loop that is saving anything longterm 19:09
so it's gotta be the whenever leaking, I think 19:10
(this is running on 2018.01 star, btw)
on the $*OUT.flush thing, yes, I know that stderr is unbuffered, and stdout is buffered by default. I'm complaining that a .flush wasn't overriding the built-in buffering like it should. 19:12
AlexDaniel didn't know ZofBot had a CLA approved…
quotable6: uptime 19:13
quotable6 AlexDaniel, 22 hours, 49 minutes, and 44 seconds, 443.957031MiB maxrss. This is Rakudo version 2018.01-29-ga2499c90f built on MoarVM version 2018.01 implementing Perl 6.c.
TimToady ZofBot is often delusional...
AlexDaniel fwiw whateverables are still leaking noticeably 19:15
but it seems to be a bit better after 2018.01
TimToady can't seem to reproduce the $*OUT.flush thing in a small test case though, so maybe it was cuz it was running in an older star at the time 19:31
jnthn A .flush does flush the handle's write buffer, so not quite sure what you were observing 19:37
I wonder if .from-list somehow manages to .cache 19:38
Don't have a free moment to look right now, alas 19:39
TimToady it's probably not the whenever, now that I monitor the growth more closely 19:50
it only grows when it's actually processing a packet from the voice daemon, not when it's throwing out the packets it doesn't care about 19:51
so it's something deeper in my loop that's hanging onto data somehow
so nevermind for now 19:52
nine Is there any way to pass an nqp::list to a Perl 6 function without any implicit conversion to List? 20:08
jnthn TimToady: ok 20:13
nine: No, but you can make an IterationBuffer instead of an nqp::list (both are VMArray REPR, but the former is a Perl 6 type) 20:14
nine jnthn: I can make one but I can't make nqp::split give me one :) So I'll just have to pass a List instead 20:16
gfldex Zoffix++ # for closing my bugreports 20:48
Geth nqp: tbrowder++ created pull request #403:
convert tabs to 4 spaces, correct obvious code misalignment
23:16
AlexDaniel tbrowder: what do you use to review the diffs? 23:17
stuff is still misaligned and I'm wondering what's preventing you to see that 23:18
one of the many examples: github.com/perl6/nqp/pull/403/file...977efcR624
at least 5 lines like that 23:19
if you look at the stuff on github, then I *guess* you should be able to see it? 23:20
but it depends on what you use view the page I guess
to* 23:21
fwiw if you're using a screen reader or something, let me know and I'll stop assuming that you should be able to see it 23:23
tbrowder well, i can see the few on github, but it’s easy to miss in my emacs editor’s window. i’ll close again and resubmit, without updating master, but referencing the closed pr—sorry. 23:26
AlexDaniel tbrowder: why not --amend it instead? 23:28
you change the files, then you `git add` them
then instead of doing a new commit with ``git commit`` you do ``git commit --amend``
and this modifies your last commit
when you push that, it won't accept it (because you're rewriting history), so you'll have to ``git push --force``, but that's OK as long as you're not doing that in the actual repo 23:29
tbrowder well, it looks ugly the way...
AlexDaniel not uglier than creating a new pull request every time you need to change some small detail? 23:30
also, to review the diff, use something like ``git show`` to see the last commit 23:31
alternatively, ``git commit --staged`` if you didn't commit something yet
(but git added it)
if not added and not committed, then just ``git diff`` should do the trick 23:32
tbrowder yep, good ideas, i really don’t know git in depth, obviously. if it’s ok, git-wise, and dev-wise, could i just make the changes with the web editor? 23:59