Zoffix c: bad26 say 'This commit is pretty bad' 00:03
committable6 Zoffix, ¦bad26: «This commit is pretty bad»
Zoffix ^_^
AlexDaniel- ouch. This should probably complain 00:04
github for example refuses to show it github.com/rakudo/rakudo/commit/bad26
Zoffix it is a valid commit tho :) 00:05
`git show` also shows it :)
AlexDaniel- right
fwiw you can craft a commit with as many zeros as you … um… let your computer run :) preferably thousands of years 00:06
but 8 zeros shouldn't be that big of a deal maybe
Zoffix ZofBot: and that's why we can't have nice things! 00:15
ZofBot Zoffix, ) Note, however, that it does no flattening of its list items: @x = lol(item (1,2),(3,4)) # @x eqv LoL( (1,2), (3,4) ) * The "list" contextualizer list foo() Forces the subsequent expression to be evaluated in list context
mst Zoffix: what have you done to git? 00:17
Zoffix mst: whatumean?
mst Zoffix: bad26? 00:18
Zoffix mst: It's one of the commits. bad2657d4f37e4ae73adfc931a769e64ee7588c1
mst ok but if github won't show it I'm curious how you broke it
Zoffix I'm guessing github just doesn't accept shas shorter than N, while git doesn't mind 00:19
Zoffix goes to bed 00:20
mst ah, right, yeah, maybe it can't handle that bit
AlexDaniel- and I guess they do it so that people won't be using links idiotic links with too short shas
s/links//
IMO they already go too far by allowing 6 00:21
mst six is the common way to shorthand a commit though 00:34
AlexDaniel- not if you want to post that as a link somewhere 00:38
because a few moments later it is suddenly pointing to another commit…
or does it?
mst the git binary tells you there's more than one and makes you think about it
MasterDuke ooo, now making real progress, INTERPOLATE seems to be ~25% faster 01:17
Geth nqp: 31a40557c2 | usev6++ | src/vm/jvm/runtime/org/perl6/nqp/sixmodel/reprs/P6bigintInstance.java
[jvm] Avoid failure in getattr

An exception was raised in getattr, because we tried to set the first field found (which was 'private static final').
A better fix might be to change the logic in getattr, but this helps for now.		 06:16
travis-ci NQP build passed. usev6 '[jvm] Avoid failure in getattr 06:33
travis-ci.org/perl6/nqp/builds/286879401 github.com/perl6/nqp/compare/e9bca...a40557c2c9
[TuxCM] This is Rakudo version 2017.09-380-g4b8a0ef65 built on MoarVM version 2017.09.1-594-gb9d3f6da 07:23
csv-ip5xs 1.159 - 1.161
test 9.572 - 9.632
test-t 3.093 - 3.150
csv-parser 0.831 - 0.894
timotimo it hasn't been this close to below 3 has it? 07:37
ah, *now* i get integers with json::fast and enums 08:05
i guess ... the reboot fixed it?
(but i'd like the enum pieces to show up as strings instead) 08:08
jnthn m: ('hello', 1, 22/7, 42, 'world').classify: { .Str.chars }; 10:42
camelia ( no output )
jnthn m: ('hello', 1, 22/7, 42, 'world').classify: { .Str.chars }.^name.say
camelia Block
Cannot resolve caller classify-list(Hash[Any,Any]: Bool, List, :as(Callable)); none of these signatures match:
(Hash $: &test, \list, :&as, *%_)
(Hash $: %test, |c is raw)
(Hash $: @test, |c is raw)
(Hash $: &test, **@list…
jnthn m: ('hello', 1, 22/7, 42, 'world').classify({ .Str.chars }).^name.say 10:43
camelia Hash[Any,Any]
jnthn m: ('hello', 1, 22/7, 42, 'world').categorize({ .Str.chars }).^name.say
camelia Hash[Any,Any]
jnthn m: ('hello').categorize({ .Str.chars }).^name.say
camelia Hash[Any,Any]
jnthn m: ('hello').classify({ .Str.chars }).^name.say
camelia Hash[Any,Any]
Geth nqp: 4d312edbfb | pmurias++ | 2 files
[js] Fix typed access to multidims and make positional refs use typed acess
 11:25
nqp: fb2f34a486 | pmurias++ | t/nqp/102-multidim.t
Test access of uninitialized elements of multidims with nqp::atpos_* ops
Zoffix buggable: speed 100 12:34
buggable Zoffix, ▅▅▆↑▇↑▆▅▅▅▅▆▅▅▄▄▅▇▆▆↑↑▅▅▅▅▄▄▅▅▄▅▄▄▅▄▄▆▄▅▄▄▄▄▄▅▅▆▄▆▄▃▆↑█▆▃▃▂▂▃▂▃▂▂▂▄▅▃▄▃▃▆▃▂▂▂▂▃▂▃▃▃▃▂▂▃▃▂▃▂▂▂▁▂▂▄▂▁▂ dates: 2017-09-08–2017-10-12; range: 3.093s–4.237s; speed: 15% faster (widths: 10/3)
Zoffix buggable: speed 80
buggable Zoffix, ↑↑▇▇▆▇▆▅▆▆▆▆▅▅▆▆▅█▆▆▅▅▅▆▆▆▇▇▆▇▆▄█↑↑█▄▃▂▃▃▂▃▃▃▃▅▇▄▅▃▄█▄▂▂▂▃▃▃▃▄▄▃▃▃▄▄▂▄▂▂▃▂▂▂▆▃▁▂ dates: 2017-09-16–2017-10-12; range: 3.093s–4.237s; speed: 14% faster (widths: 10/3)
Zoffix Made it sample averages from both ends, so it's less sensitive about which commit the endpoint is at. 12:35
(widths: 10/3) => average 10 commits from the past end and 3 commits from the recent end
c: 2017.04 { temp $*OUT = "/tmp/czxczxczxc".IO.open: :w; for ^10000 { start say $_ } } 12:40
committable6 Zoffix, ¦2017.04: «»
Zoffix c: 2017.04 { temp $*OUT = "/tmp/czxczxczxc".IO.open: :w; for ^10000 { start say $_ } }
committable6 Zoffix, ¦2017.04: «»
Zoffix c: 2017.04 { temp $*OUT = "/tmp/czxczxczxc".IO.open: :w; for ^10000 { start say $_ } }
committable6 Zoffix, ¦2017.04: «9995»
Zoffix ^ wtf? Why does that one end up on STDOUT instead of the handle I po... oh ok neverfmind 12:41
There's no await
m: { temp $*OUT = "/tmp/czxczxczxc".IO.open: :w; start { sleep 1; say 42; } }; sleep 2 12:42
camelia 42
Zoffix Tho wonder if this should be concidered a bug or not
m: my $*x = 42; { my $*x = 70; start { sleep 1; say $*x; } }; sleep 2 12:43
camelia 70
Zoffix m: my $*x = 42; { temp $*x = 70; start { sleep 1; say $*x; } }; sleep 2 12:44
camelia 42
AlexDaniel- Zoffix: hey, I've seen that you tried to use quotable6 a couple of times without much success. It's a little bit better now, maybe give it a try
quotable6: squashathon
quotable6 AlexDaniel-, OK, working on it! This may take up to three minutes (4412528 messages to process)
Zoffix doesn't remember using it
AlexDaniel- well, a few weeks ago :)
quotable6 AlexDaniel-, Found nothing! 12:45
AlexDaniel- WHAT?
AlexDaniel- facepalms
ZofBot: can you teach your little fella? 12:47
ZofBot AlexDaniel-, For security reasons, many operating systems limit this functionality to the superuser
AlexDaniel- quotable6: squash 12:49
quotable6 AlexDaniel-, OK, working on it! This may take up to three minutes (4412528 messages to process)
jnthn start blocks don't make a complete clone of the dynamic scope, so they'll see any mutations in it 12:50
They just have a reference to the context they were invoked in
quotable6 AlexDaniel-, gist.github.com/a59458684f0ae2cb0f...765643430a 12:51
AlexDaniel- ↑ I know it's a bit LTA, but enough to be useful 12:52
Zoffix jnthn: I just filed it as rt.perl.org/m/ticket/show?id=132275 lol
AlexDaniel- except that I have no idea why it failed to match “squashathon”
Zoffix jnthn: so should it be closed?
AlexDaniel- good night everybody o/
jnthn o/ AlexDaniel-
Zoffix: Yeah, I mean, what would a fix look like?
Zoffix shrugs and closes :) 12:53
jnthn You'd basically be asking for it to walk down the entire call stack on each start block, identify every dynamic variable, and clone it, to some depth or other
Hmm... 12:56
Just looking at Seq
Oh, maybe it's OK but odd 12:57
Zoffix What is it?
jnthn I'm guessing Iterator.count-only and Iterator.bool-only, if defined, must be able to produce the result wihtout actually doing any kind of iteration work?
But even then it's a tad weird 12:58
There's code like this:
nqp::if(
nqp::can($!iter,'bool-only'),
$!iter.bool-only,
nqp::if(
nqp::can($!iter,'count-only'),
?$!iter.count-only,
self.cache.Bool
)
),
The optimized paths aren't equivalent to self.cache.Bool though
Because self.cache takes the iterator
And the optimized paths don't 12:59
So you can get inconsistent semantics
Depending on if the Seq just happens to have an iterator under it that supports count-only or bool-only
But also the optimization *relies* on keeping the iterator around :/
Zoffix Yeah 13:00
But you can think of it that part of optimization is you don't have to cache stuff for iterators that can do bool-only and count-only
jnthn But it makes "This sequence has already been consumed" behavior inconsistent 13:01
my \s = get-some-seq(); say s.elems; for s { ... }
If get-some-seq() returns something that has the count-only optimization then this code accidentally works 13:02
Then you do some seemingly harmless refactor inside of get-some-seq (which may come from a module you don't control)
(Or you might control the module but not the downstream code)
And then boom, breakage.
For code that never should have worked in the first place anyway
So the optimization violates the Seq contract 13:03
Zoffix No, it works in all the cases. If optimization exists, `.elems` won't steal iterator and `for` will use it. If optimization doesn't exist, .elems will cache the sec and for will use the cached iterator 13:04
jnthn We...used the cached iterator?
Hm, we do
Zoffix m: my $s := Seq.new: class :: does Iterator { method zool-only { say 42 }; method pull-one { $++ and IterationEnd } }.new; say $s.elems; .say for $s
camelia 1
0
Zoffix m: my $s := Seq.new: class :: does Iterator { method bool-only { say 42 }; method pull-one { $++ and IterationEnd } }.new; say $s.elems; .say for $s
camelia 1
0
jnthn Did that change at some point? I'm sure I didn't have it that way in the first place 13:05
But yeah, then it is OK
Zoffix Yeah, I changed it a few months ago
jnthn OK, I guess I can live with it
Zoffix Basically things that want an iterator when they can't find it also check if our Seq is already cached and use that
jnthn It's DWIM-y I guess 13:06
Zoffix thought old version was a bug :o
:)
jnthn No, I had it that way so that if you were dealing with a Seq that you were expecting to be just iterating and throwing things away, then you'd be loudly told about it 13:07
Rather than surprise memory retention
Zoffix Ah. From my POV it was: why doesn't it just use the already-cached version instead of complaining at me ^_^ 13:09
jnthn But it's one of those things I can argue either way. On the one hand, it's nice if the user doesn't have to care and write .cache in some more places. On the other hand, it meant you had to be explicit about the use of the caching
Zoffix Some things implicitly cached tho (like .elems)
jnthn So you could more easily spot O(n) vs O(1) memory use
True 13:10
Though a relatively small set of them (those that are involved in indexing)
I guess by this point we can't change it back without breaking a bunch of code that now relies on things, though 13:11
s/things/the new semantics/ 13:12
At least elems and friends aren't as inconsistent as I thought, though
And I guess we can rely on having better allocation/memory profiling tools to help users spot their caching issues
Zoffix We could, since it's not part of any language release yet. Though IME I haven't ran into "seq already consumed" errors recently and haven't seen many users complain about iterated Seqs. The complain is nice when you know what you're doing and can fix it, but can be quite annoying for people who haven't mastered Seq vs. List vs. Array vs Slip; especially since many methods return Seqs. 13:14
jnthn *nod*
I think we can leave it as it is 13:15
Zoffix \o/
jnthn I just noticed this stuff 'cus I'm pulling a Sequence role out of Seq, which HyperSeq and RaceSeq will also do
And was thinking what does/doesn't go into said role
Zoffix
.oO( teach fudger to handle stress tests: you mark a particular test as stress test so it only runs during `make stresstest` and not `make spectest` )
 13:43
'cause some of the tests just cover the bugs and don't really need to be part of spectest, but creating a whole nuther file just for one test is kinda meh
.ask AlexDaniel- OOC, is there a CLI interface to commitable? Kinda a waste of time to be going to view at gists all the time to read output, while trying to make some code that covers a bug. Looking for something like `./commitable-cli.p6 2017,HEAD file-with-code.txt` or `./commitable-cli.p6 2017,HEAD -e 'some code'` and it'd just dump all the output into terminal 13:56
yoleaux Zoffix: I'll pass your message to AlexDaniel-.
Geth roast: ccdbc68336 | (Zoffix Znet)++ | MISC/bug-coverage-stress.t
Add MISC/bug-coverage-stress.t

For an oddball bugs to cover and to run only under stresstest		 13:57
rakudo/nom: 37689a3e57 | (Zoffix Znet)++ | t/spectest.data
Add MISC/bug-coverage-stress.t to list of files to run
 13:58
Zoffix Oh oops.. Add MISC/bug-coverage-stress.t ... also include tests to cover mutex deadlock for RT#132042
synopsebot RT#132042 [open]: rt.perl.org/Ticket/Display.html?id=132042 [TESTNEEDED] [CONC] Concurrently calling `say` can cause Rakudo to hang
Geth roast: 57b039fb48 | (Zoffix Znet)++ | MISC/bug-coverage-stress.t
List RT number in test
 13:59
Zoffix & 14:01
So I got myself a helper that does some HTML work... He made a dozen <p> elems in a row, each with class="bluefont". That's bad, but not funny... The funny part is in CSS it's defined as: `.bluefont { color: black; font-size: 22px; }` 14:19
timotimo .o( naming things and off-by-one level of nesting ) 14:23
Zoffix :) 14:24
Zoffix looks at github.com/rakudo/rakudo/pull/1160/files 15:52
Oh, nm.
Confused "no buffering on TTY handles" vs "no buffering on stdout"
AlexDaniel-: IMO this feels like pointless waffle: github.com/perl6/doc/commit/700842...448bfa7aaf The docs site documents the language, not an implementation and the bug has now been fixed, and I've no idea if that comment was reverted or not 16:09
without going to docs site and trying to find it
I thought you could merge STDOUT/STDERR with Proc::Async... looks like the info's gone from the docs, even though the code has $!merge_blah blah in it 16:11
jnthn It's just .Supply
Zoffix Thanks 16:12
jnthn $ perl6-m -e 'my $p = Proc::Async.new("perl6", "-e", "say 42; note 43"); react { whenever $p.Supply { print "Merged got $_" }; $p.start }' 16:13
Merged got 42
Seems to work :)
Merged got 43
Zoffix Looks like some of the output gets lost tho. Or maybe Proc::Async.kill ain't resulting in the signal I'm expecting :/ 16:16
m: gist.github.com/zoffixznet/4914127...34b466610f
camelia 1..1
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 4

# Looks like you planned 1 test, but ran 0
Zoffix bah
c: HEAD gist.githubusercontent.com/zoffixz...46a7/p6.p6
e: HEAD gist.githubusercontent.com/zoffixz...46a7/p6.p6 16:17
evalable6 (exit code 1) 04===SORRY!04=== Error while compiling /tmp/N4AA_yPQeN
Confuse…
Zoffix, Full output: gist.github.com/2c50a327cd2226acee...9eae412cf4
Zoffix e: gist.githubusercontent.com/zoffixz...46a7/p6.p6
evalable6 Zoffix, Successfully fetched the code from the provided URL.
(exit code 1) 1..1
not ok 1 - Supply.merge on signals does not crash
# Failed test 'Supply.merge on signals does not crash'
# at /tmp/5A_ipjDGOA line 14
# expected: "pass"
# got: ""
# Looks like you failed 1 test of 1
Zoffix `got: ""` even tho there's some output produced both before react block and within it when signal is tapped 16:18
jnthn Is that unique to the merged version? 16:19
Or does .stdout do the same?
Zoffix .stdout does the same 16:20
jnthn Hm, I don't see anything silly in Proc::Async or down in the VM that'd be closing the handles too soon on a kill being sent 16:23
As in, it all looks like it just sends the signal and that's it
Zoffix Looks to be Proc::Async.kill that's the culprit. If I kill from the shell, instead with .kill, then it gets all the output 16:25
s/instead with .kill/instead of .kill/;
Zoffix files
ohhh 16:29
Never mind no bug. It just kills it before the Proc has a chance to start and sign up for signal handling 16:30
jnthn m: say 2.97 / 0.71 16:36
camelia 4.183099
Geth roast: 016aa10f8e | (Zoffix Znet)++ | MISC/bug-coverage-stress.t
Test Supply.merge on signals does not crash

RT#132016: rt.perl.org/Ticket/Display.html?id=132016 Rakudo fixes: github.com/rakudo/rakudo/commit/1f411693bd
   github.com/rakudo/rakudo/commit/13b6a33c20		 16:40
jnthn Current design/implementation sketching for hyper/race: gist.github.com/jnthn/754292747161...207d077d74
synopsebot RT#132016 [open]: rt.perl.org/Ticket/Display.html?id=132016 [TESTNEEDED] Supply.merge and signals ( signal(SIGTERM).merge(signal(SIGINT)) )
Zoffix huggable: hyper :is: New hyper/race implementation exploration: gist.github.com/jnthn/754292747161...207d077d74 16:41
huggable Zoffix, Added hyper as New hyper/race implementation exploration: gist.github.com/jnthn/754292747161...207d077d74
jnthn Will continue to fill it out 16:42
And at some point, move it into CORE.setting 16:44
Given the current hyper/race are so busted, I guess the threshold for inclusion isn't too high :P 16:45
But want to get a bit further first 16:46
Geth nqp: c4e1d6c984 | usev6++ | src/vm/jvm/runtime/org/perl6/nqp/io/AsyncProcessHandle.java
[jvm] Don't wait for child when program ends

With a user thread the following program sleeps for another 18 seconds after saying bye:
   my $proc = Proc::Async.new('sleep', '20');
   my $promise = $proc.start;
... (6 more lines)		 16:47
nqp: 7528e8104d | (Jonathan Worthington)++ (committed using GitHub Web editor) | src/vm/jvm/runtime/org/perl6/nqp/io/AsyncProcessHandle.java
Merge pull request #373 from usev6/jvm_procasync

  [jvm] Don't wait for child when program ends
Zoffix .tell AlexDaniel- can you revert this now? The ticket is fixed github.com/perl6/doc/commit/700842...448bfa7aaf 16:52
yoleaux 16:43Z <[Coke]> Zoffix: Proc::Q fails its tests here.
Zoffix: I'll pass your message to AlexDaniel-.
Geth nqp: usev6++ created pull request #374:
Jvm killprocasync		 17:51
AlexDaniel- . 18:29
yoleaux 13:56Z <Zoffix> AlexDaniel-: OOC, is there a CLI interface to commitable? Kinda a waste of time to be going to view at gists all the time to read output, while trying to make some code that covers a bug. Looking for something like `./commitable-cli.p6 2017,HEAD file-with-code.txt` or `./commitable-cli.p6 2017,HEAD -e 'some code'` and it'd just dump all the output into terminal
16:52Z <Zoffix> AlexDaniel-: can you revert this now? The ticket is fixed github.com/perl6/doc/commit/700842...448bfa7aaf
AlexDaniel .
Geth roast: c9d52eea5e | (Zoffix Znet)++ | MISC/bug-coverage-stress.t
Remove useless multiplication
 18:44
roast/6.c-errata: 8857362ba5 | (Zoffix Znet)++ (committed using GitHub Web editor) | 26 files
Use `v6.c` instead of `v6` so we get right language version (#336)
Zoffix This "removing content from docs site because there's a bug in an implementation X" feels counter-productive to me. You're deliberately hiding features of the language making potential users re-invent the wheel, instead of potentially volunteering to fix the bug. 18:50
AlexDaniel- Zoffix: ok, so, commit reverted, but I also added a note mentioning brokenness on rakudo versions before 2017.09.
I'm not saying that we should remove content, but if we know that things are broken why can't we be honest about it 18:51
Zoffix We are honest about it: fail.rakudo.party/ 18:52
If I haven't read through that RT I wouldn't even know the feature is commented out. 18:53
And Proc::Async.Supply was fixed months ago.
AlexDaniel- I didn't revert the commit right away because there was no rakudo release at the time
Zoffix It's about separation of concerns. The docs site documents the language. It's not a bug tracker.
AlexDaniel- there was a TODO left in the docs
ahhh… Look, I totally understand that 18:54
timotimo jnthn: i can't run your gist, does it have to be put into rakudo to work? it complains about Sequence being an invalid typename, and just replacing it with Seq won't fly
AlexDaniel- but expecting people to go through 1000+ bugs to know what they can and cannot use is unrealistic
and unfair
Zoffix But it's fair to force them to re-implement features the language provides? 18:55
AlexDaniel- as a user, I want the docs to be practical. I don't need some theoretical info about some imaginary features that have no (or broken) implementation 18:56
Zoffix Re-implementing a feature from scratch instead of fixing a possibly easy bug is not my definition of "practical" 18:57
And faffing around creating and reverting a million commits just takes away more time from fixing said bug
s
Same as a bunch of RTs created for fudged tests when the fudges are already known. 18:58
ZofBot: complain all the things!
ZofBot Zoffix, The "redo" operator, as a variant of "goto", directly transfers control to the first statement of the lexotically enclosed loop
AlexDaniel- Zoffix: you're not wrong, it's just that we can do better 18:59
Zoffix GitHub now has dependency graphs.. neat github.com/rakudo/rakudo/network/dependencies
I mean help.github.com/articles/listing-t...epends-on/
AlexDaniel- and by “better” I mean for the user, and you're trying to solve a problem for yourself :) 19:00
Zoffix Oh "You can see your project's Ruby and JavaScript dependencies in the dependency graph."
Seems oddly specific. I guess that's just for start and they'll add more langs later 19:01
AlexDaniel- can we somehow add support for META6.json?
jnthn timotimo: Needs a commit introducing Sequence, which is in a branch that I probably forgot to push 19:02
timotimo i see! 19:04
AlexDaniel- c: 2017.07 gist.githubusercontent.com/AlexDan...t131763.p6 19:08
committable6 AlexDaniel-, Successfully fetched the code from the provided URL.
synopsebot RT#131763 [new]: rt.perl.org/Ticket/Display.html?id=131763 [ANNOYING] Merged output of Proc::Async does not work (whenever $proc { … })
committable6 AlexDaniel-, ¦2017.07: ««timed out after 10 seconds» «exit signal = SIGHUP (1)»»
AlexDaniel- bisect: old=2017.07 gist.githubusercontent.com/AlexDan...t131763.p6
bisectable6 AlexDaniel-, Successfully fetched the code from the provided URL.
AlexDaniel-, Bisecting by exit signal (old=2017.07 new=37689a3). Old exit signal: 1 (SIGHUP) 19:09
AlexDaniel-, bisect log: gist.github.com/e62286d91420b51fda...47347dd457
AlexDaniel-, (2017-09-13) github.com/rakudo/rakudo/commit/d2...d43ad42da2
AlexDaniel- huh 19:10
c: d2eb7423127 gist.githubusercontent.com/AlexDan...t131763.p6
committable6 AlexDaniel-, Successfully fetched the code from the provided URL.
AlexDaniel-, ¦d2eb742: ««timed out after 10 seconds» «exit signal = SIGHUP (1)»»
AlexDaniel- no, that's not it
AlexDaniel- moves to #whateverable
Zoffix AlexDaniel-: They are already tests covering this bug: github.com/perl6/roast/commit/ae3e...6df346a457 19:11
s/They/There/; 19:12
AlexDaniel- I don't think it does 19:13
or… maybe it does
IIRC you needed at least 20 iterations to make it fail
so it's a good candidate for stress testing 19:14
“Merge branch 'better-sched' into nom” – I think this is what fixed it 19:16
Zoffix ZofBot: bug reporters should report their versions 19:22
ZofBot Zoffix, ) Binding within an "else" automatically binds the value tested by the previous "if" or "elsif", which, while known to be false, might nevertheless be an *interesting* value of false
timotimo i wonder what ) Binding is 19:23
Zoffix m: if Failure.new {} else { $^meows.say } 19:24
camelia (HANDLED) Failed
AlexDaniel- Zoffix: OK I retagged RT #131763 and left a comment there 19:25
synopsebot RT#131763 [new]: rt.perl.org/Ticket/Display.html?id=131763 Merged output of Proc::Async does not work (whenever $proc { … })
AlexDaniel- Zoffix: any other problematic tickets?
Zoffix AlexDaniel-: I can't find a commit where ^ that problem exists
AlexDaniel- OK, I'll take a look 19:26
Zoffix: ticket id?
Zoffix AlexDaniel- the RT you retagged
The Proc thing 19:27
AlexDaniel- Zoffix: ah
Zoffix: ce12e480316
Zoffix Thanks
AlexDaniel- c: ce12e480316 gist.githubusercontent.com/AlexDan...t131763.p6 19:28
committable6 AlexDaniel-, Successfully fetched the code from the provided URL.
AlexDaniel-, ¦ce12e48: ««timed out after 10 seconds» «exit signal = SIGHUP (1)»»
AlexDaniel- it's one of the parents of the “better-sched” merge commit
(timeout *is* the buggy behavior, it gets stuck forever) 19:29
Zoffix Ah ok
[Coke] Speaking of Diwali, I am apparently joining folks in the office in wearing a Kurta next week. (need to pick one up.) 19:37
Zoffix I thought we were renaming Diwali to something or other 19:39
6.d Dismorphia :)
github.com/perl6/6.d-prep/tree/mas...ODO#naming 19:40
There was some bikeshedding on some commit somehwere
timotimo mighty dismorphing power perlers
ugexe metal gear style: Diwali Dismorphia 19:42
timotimo Compilers of the Perltriots 19:43
Zoffix :)
:S 19:45
Geth didn't announce my commit and if I look at it on GitHub it says it was made "11 minutes from now": github.com/perl6/roast/commit/e821...c07792e13c
Oh, I guess the clock is messed up on this box 19:46
12 minutes into the future
ZofBot: I think I entered into a pocket of time radiation and traveled into the future a bit... weeeeee 19:48
ZofBot Zoffix, weird
Zoffix Ya I kno
AlexDaniel- Zoffix: as for the command-line version of committable, there's no such thing *yet*. My goal was to make the bots runnable locally (in some kind of non-irc mode), where all required builds are pulled from the server on demand. This will only work for 64-bit linux of course, but for many purposes it should be good enough. 20:41
no progress on that yet :(
I want to get trisectable (rakudo-nqp-moar) to work properly first 20:42
and instead of that right now I'm trying to revive regressionable
fwiw, volunteers welcome. I can provide hints on how these things can be done 20:44
pmurias Zoffix: in the old time the synopsis (which served as docs back in the day) had something called smartlinks (so you could see a test for the feature in the docs) 21:12
Zoffix: the smartlinks had a status which show if they pass or fail 21:13
Zoffix Considering this feature is dead, it's kind'f indicative that it's not maintainable. 21:14
AlexDaniel-: I was thinking even if it were possible to do on the commitable server. Give the IRC::Plugins a faked out IRC::Client::Message and reply to terminal 21:15
Not really important, I guess. I won't need it after the ~40 testneeded RTs are closed
pmurias Zoffix: it was really cool, it would be great to have something like that for user modules
which is an extra reason not to display module docs on the evil github 21:16
AlexDaniel- Zoffix: thinking about it, there's an IRC server running on the same server (used for all whateverable tests). I guess you can run any bot with TESTABLE=1 env variable, then join a corresponding channel using some cli irc client, and then you'll be able to see the content of all gists in /tmp/whateverable/tist 21:20
a bit convoluted, but let me know if you want to go with this
Zoffix calls it a day 21:21
ZofBot: good night, sweet prince!
ZofBot Zoffix, (You could write a macro that violates this, but resist the urge
AlexDaniel- lol… these answers are very on point
maybe
MasterDuke jnthn: i tried increasing MVM_INTERN_ARITY_LIMIT from 8 to 9 and 16, but that causes segfaults; however, i didn't re-build nqp and rakudo 21:23
i was able to work around the arity limit by combining two parameters into one 21:24
but do you think that limit is going to stay the same? 21:25
Zoffix geth is ded "We couldn’t deliver this payload: Couldn't connect to server" 21:31
And I tried killing geth so it respawns and `service restart geth` and it restarted, but error is still there. And been there for an hour already
$ telnet geth.niner.name 21:33
Trying 78.47.60.231...
telnet: Unable to connect to remote host: Connection refused
something's ain't right
(same if I try geth's port) 21:34
nine: are you around by any chance? Any idea how to get geth back? 21:38
tis the right IP if I go to www.whatismyip.com/ from the box 21:41
(amusingly, site itself rejected me 'cause I used HTTP1.0, but the error page includes IP details :P) 21:42
moritz: do you know how to fix this? For some reason services can't connect to the box with geth on it 21:53
or timotimo 21:54
No errors in geth when sending hooks... Seems it's not getting through to the box 21:56
telnet 78.47.60.231 8888 from geth box don't work but telnet localhost 8888 does 21:57
jdv79 who telnets anymore 21:58
Zoffix people debugging connections
lol 22:00
root's password:
Where did you learn to type?
I love nine's box. It's so polishly setup.
Zoffix gives up 22:02
No idea how this box is set up to reach outside world.
But I commited this, in case anyone wanted to double-check: "Replace cryptic code with direct <.ws> call": github.com/rakudo/rakudo/commit/1a...88f2d61363 22:05
Zoffix &
AlexDaniel- samcv: I think you're the first (and only) person in the whole IRC history who actually attempted to run ｢rm -rf /｣ on the evalbot :D 22:34
and I'm quite surprised
a lot of people attempted to do delete specific files like irclog.perlgeek.de/perl6/2016-01-23#i_11928519 22:35
but not with “-rf /”
private messages excluded, of course
gist.github.com/AlexDaniel/515ecdf...3a54e3d3b3 22:37
that first one probably counts also
jnthn MasterDuke: You may have to rebuild the extops you bump that; I see no reason we couldn't increase it to 16 if we actually have real-world use cases 22:45
timotimo do we count nameds douple in the figure we're using for size limits? 22:46
jnthn yeah
iir
c
timotimo i can imagine having like an invocant and 7 nameds 22:47
jnthn Yeah
MasterDuke jnthn: this is what i did to work around the limit for INTERPOLATE: github.com/MasterDuke17/rakudo/com...b18085f950 22:54
is there any great cost to increasing the limit?
Zoffix m: $*OUT.^lookup("open").signature.params.elems.say 22:55
camelia 21
Zoffix yikes
timotimo MasterDuke: looking up %*RX over and over could be costly, but this code only runs rarely, doesn't it? 22:56
jnthn MasterDuke: I don't think so 22:57
Zoffix: This is about callsites
Zoffix: Which is the argument side
It'd be odd to pass all 21 of them :P
And all the limits on opt stuff are on arguments, not parameters 22:58
If open is really hot enough, spesh is typically able to eliminate branches in the specialization
MasterDuke timotimo: only when interpolating into regexes. fwiw, initial testing of my branch (after that last commit to reduce the number of arguments by one) shows `./perl6-m -e 'my @l = "sm.sql".IO.lines; my $s = "Perl6"; my $t = now; my @m = @l.grep(/ $s /); say @m.elems; say now - $t'` ~25% faster (10k lines in the file) 23:02
timotimo wow, 25% is a great win! 23:03
how big's the change without the latest commit, ooc?
MasterDuke 3x slower
but i'll try increasing MVM_INTERN_ARITY_LIMIT to 16 and then rebuilding nqp and rakudo and then measure again 23:04
timotimo waitwhat 23:06
3x slower o_O
MasterDuke because i converted INTERPOLATE into multis, but the method lookup wasn't being cached 23:09
this is what you and jnthn were helping me debug yesterday
timotimo aaah
MasterDuke turns out it's because the callsites weren't being interned, and that was because of the MVM_INTERN_ARITY_LIMIT 23:10
timotimo understood
MasterDuke heh. took me quite a bit longer to figure out, only a couple sentences to explain 23:11
timotimo i know that feeling :)
jnthn :) 23:12
Sleep time; 'night 23:13
timotimo gnite jnthn :)
MasterDuke gains on my branch are significantly less for <$s> compared to just $s, and even less for @s (with 4 elements) 23:14
Zoffix Ah, I see
MasterDuke later, thanks for the help!
Zoffix: see something related to what i'm doing? 23:17
Zoffix MasterDuke: nah, it was re "Zoffix: This is about callsites"
MasterDuke ah, i see ;) 23:18
hm. changing MVM_INTERN_ARITY_LIMIT to 16 and then trying to rebuild nqp also results in immediate segfaults 23:19
backtrace if anyone is interested, looks the same as the one i got when just re-running my perl6 code after rebuilding moar yesterday: gist.github.com/MasterDuke17/60679...587e405610 23:22
samcv AlexDaniel-, hah. well i knew it wouldn't work :P 23:43
Zoffix Well, until we figure out how to fix Geth, I'm gonna add more hooks to send stuff to Geth in the meantime 23:52
done moarvm, rakudo, roast, and docs. The rest ain't that important 23:55

Powered by Cro and the Raku® Programming Language.

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

Thank you!