japhb .ask lizmat re: rakudo 5db5b1db , couldn't you test for being concrete or being Any? That would only leave the much smaller number of type objects that are Mu but not Any. Not perfect, but an improvement I think. 00:43
yoleaux japhb: I'll pass your message to lizmat.
Geth rakudo: snarkyboojum++ created pull request #1136:
Adding name for snarkyboojum
06:43
lizmat Files=1223, Tests=67668, 275 wallclock secs (13.98 usr 5.24 sys + 1861.57 cusr 194.07 csys = 2074.86 CPU) 08:35
.tell japhb I'd much rather see a HLL way to test for NQPMu, as we need that in some other places as well
yoleaux lizmat: I'll pass your message to japhb.
Geth rakudo/nom: 8944721366 | (Adrian White)++ | CREDITS
Adding name for snarkyboojum
08:41
rakudo/nom: 636a3c1264 | lizmat++ (committed using GitHub Web editor) | CREDITS
Merge pull request #1136 from snarkyboojum/nom

Adding name for snarkyboojum
ugexe gist.github.com/ugexe/31e22b592431...0fca34e3a0 # why does this code act different? it looks the same, but it fails spectests (and additionally fixes Grammar::HTTP which was broke from uncurse branch merge in march-ish) 14:41
i replace self.to with the code inside method self to { }, and replace nqp::subst_i with an actual `-` 14:42
sub_i rather 14:43
t/spec/S05-transliteration/79778.t .. This type cannot unbox to a native integer: P6opaque, Num 14:44
the only test failures are transliteration
TimToady: might be of interest ^ 14:56
[TuxCM] This is Rakudo version 2017.07-149-g93a5eff7c built on MoarVM version 2017.07-382-g348b4262 15:03
csv-ip5xs 1.348 - 1.367
test 12.322
test-t 4.385 - 4.415
csv-parser 14.124
[Coke] any pointers on how to do a build on windows? I have a win 10 eval VM that came with a version of VS. 15:08
ugexe use `x86 Native Tools Command Prompt for VS 20xx` not cmd.exe 15:11
as that will have the compiler toolchain stuff in PATH 15:12
you can also use rakudobrew in windows so you dont have to remember which "make" to use! 15:15
timotimo [Coke]: i think you just get strawberry perl so you can use Configure.pl 15:18
and then use nmake?
ugexe install chocolately and follow github.com/rakudo/rakudo/blob/nom/...or.yml#L48
but nmake isn't usually PATH'd in the normal command prompt
lizmat [TuxCM]: disappointing :-( 15:19
timotimo oh no :(
stmuk [Coke]: github.com/rakudo/star/blob/master...ws-msi.pod might help (ignoring the WiX toolset/MSI stage) 15:46
BTW I tried a Reactos (open source Windows) build yesterday but gcc segfaulted (probably the fault of the OS) 15:48
nine It shouldn't be that hard finding the test-t regression, should it? 15:53
japhb . 15:58
yoleaux 08:35Z <lizmat> japhb: I'd much rather see a HLL way to test for NQPMu, as we need that in some other places as well
japhb .tell lizmat Fair enough. 15:59
yoleaux japhb: I'll pass your message to lizmat.
Skarsnik btw it that just a TODO that the family (ipv4 or ipv6) is never passed to NQP/Moar in github.com/rakudo/rakudo/blob/nom/...et/INET.pm 16:04
Sound dumb but localhost for me resolve to ipv6 and ipv4
ugexe i think it used to get passed but since it didnt do anything yet it got removed 16:06
Skarsnik I am adding a source-address param to bind the socket (on connect) to the ip you want, but like I try localhost (ip to reachĆ  from iplocal (ip to bind) it does not work since localhost is ipv6 16:08
Not sure how to handle this aside throwing an error x) 16:12
timotimo m: say 0x7ffffffff000 - 0x7ffffffde000 16:16
camelia 135168?
timotimo oh, d'oh 16:17
ugexe isnt that an OS level detail?
(what localhost binds to) 16:18
Skarsnik hm, let me check, I think it's just because a host can resolve to a ipv4 and ipv6 address 16:19
geekosaur you need to filter the list for correct protocol like I mentioned yesterday (granted it was a bit offhand and perhaps not clear)
timotimo it's whatever you write in /etc/host
geekosaur but it's also complicated by hosts that do ipv4 emulated
timotimo you mean you get connections from ipv4 endpoints via the ipv4-inside-ipv6 address range? 16:20
geekosaur it works both ways, your server program sees an ipv6 socket that is also an ipv4 socket 16:21
in which case the ipv6 localhost is correct 16:22
("thanks" linux)
Skarsnik Sure. You have an A record (for IPv4) and an AAAA record (for IPv6). Typically the AAAA record gets resolved first, then the A record.
(copy pasted from somewhere)
geekosaur (not all linux dustributions do this, but enough do that you need to check for your ipv4 socket being an ipv6 socket)
Skarsnik so yeah an hostname can be ipv4 or v6 16:23
geekosaur but in general, if you are binding to an address, you should bind to the *address* not a hostname
you might be able to provide a symbolic constant meaning "the loopback address for this address family" but don't use a hostname 16:24
Skarsnik I also feel like I kinda want to give the user the ip back x)
[Coke] tries out a win10 eval and puts together instructions on how to get a build using their eval version. 16:31
Skarsnik what do you think of 16:48
skarsnik@devel:~/perl6/socket/rakudo$ ./install/bin/perl6 -e 'my $conn = IO::Socket::INET.new(:host<localhost>, :port(80), :source-address("pikonyo")); $conn.print("GET index.html\r\n"); say $conn.recv()'
The socket won't be binded. Destination and Source address family does not match. Destination is IPv6. Source is IPv4
geekosaur this is system dependent, network stack dependent, network configuration dependent, and people have been fighting with it for years. perl 6 is not going to "fix" it 16:50
Skarsnik Yeah that why I tried to give an 'useful' error 16:52
*try
also I get a Cannot unbox a type object (Str) to a str. 16:57
when I don't provide a source-address but I defined the attribute as Str $.source-location = ""; and in the new that do the connect I put it to "" too.
timotimo default values only work if there's no custom BUILD
so perhaps there is one?
Skarsnik The file is github.com/rakudo/rakudo/blob/nom/...et/INET.pm 16:58
I added Str :$source-address is copy = "", in the new method that do connect 17:02
forget to do make && make install in rakudo x) 17:04
lizmat Skarsnik: there's also a --make-install param in the configure step
yoleaux 15:59Z <japhb> lizmat: Fair enough.
Skarsnik what does that do? 17:05
timotimo it runs "make install" for you when the configure step is done
Skarsnik I don't need that, it's just I always forget to make && make install when changing a pm file in rakudo 17:06
lizmat just "make install" is shorter :-)
it will do a "make" for you
timotimo but you'd have to " && make install" 17:07
Skarsnik could be nice to not have to rebuild the core.setting and stuff every time ^^ 17:09
at least it give you time to do some chores 17:10
I am not sure how to make it work : gist.github.com/Skarsnik/24778366a...et-pm6-L86 I tried having a default value to "" 17:14
Zoffix [Coke]: the doc build is still busted. It isn't updating 17:33
Which is annoying, 'cause I keep linking people to Texas versions of atomic ops, but they ain't there -_- because the site didn't update
Geth nqp/ugexe-patch-1: a5b2f39709 | (Nick Logan)++ (committed using GitHub Web editor) | src/QRegex/Cursor.nqp
Fix 'Substr length cannot be negative' from Match

I don't understand why this should change anything, but it does.
Fixes Grammar::HTTP - see:
  github.com/ugexe/Perl6-Grammar--HTTP/issues/6
  irclog.perlgeek.de/perl6-dev/2017-...i_14521237
17:34
nqp: ugexe++ created pull request #367:
Fix 'Substr length cannot be negative' from Match
[Coke] Zoffix: it is only some commits that seem to cause the (&#$ trouble.
lizmat afk&
[Coke] Zoffix: kicked off another build using "force-update" - those seem to go through, will keep an eye on it. thanks for the heads up 17:35
(I don't understand why that works: it's still using all the same infrastructure. 17:37
Zoffix Weird. Thanks. 17:40
Man, the atomic ops sure kicked up a "discussion" :) 99 comments in total so far www.reddit.com/r/perl6/duplicates/...nd_perl_6/ 17:41
Mostly people from 1960s saying they don't got a ? button on their keyboard :P 17:42
Skarsnik for something probably 2 users will really have a use for it? xD
tadzik yeah, I read that too
I don't think they're that wrong though
Skarsnik I don't even know how to do unicode char, so I can understand ^^
Zoffix hehe, yeah, and for which we have Texas versions, so the whole discussion is moot IMO
tadzik I put Ā«Ā» on my keyboard layout, ā€œā€ too, but I don't think it's a sane way to go to add everyshit possible to it too
Zoffix tadzik: so use Texas? 17:43
That's the whole point of it.
tadzik yeah, I can and will
Zoffix I don't use the ā€œā€ quoters. I use q||
qq|| even
[Coke] I also install App::Uni and use "uni newline" if I need to get something I can cut and paste.
tadzik thing is, Ā«Ā» have nice and short ascii alternatives, ? does not
Zoffix tadzik: because it's not an oft-used thing. 17:44
[Coke] (I use that to get the uni "NL" when updating doc OUTPUTs)
tadzik Zoffix: is that why the blogpost advertises it as useful even for non-threaded code? ;)
then again, I'm a hater of emojis and stupid shit in unicode, so I'm perhaps a bit biased 17:45
and last I checked I'm not a 5yo but I still don't know how to make a smiley face that's not rotated like ":)" 17:46
not on my pc, and not on my phone either
Zoffix tadzik: well, before the update, you had to write like three strategically-placed lines of code and no one complained. Now it's just a single routine and people are saying it's too long. We also added an even shorter Unicode alternative and people are complaining they don't have a ? button on their keyboard. To me, this all sounds like people are desperate to complain about anything. 17:47
[Coke] hurm. using strawberry perl and gmake, I get gcc as my compiler on windows. I assume ideally we'd want to be getting Visual Studio, yes?
Skarsnik probably? 17:48
Zoffix .ask lizmat Shouldn't ?+= and ?-= be called atomic-add-assign and atomic-sub-assign instead of atomic-add-fetch and atomic-sub-fetch, since ?= is atomic-assign, not atomic-fetch?
yoleaux Zoffix: I'll pass your message to lizmat.
tadzik Zoffix: well, I feel like people are mostly hostile to emoji-culture creeping its way into the programming world. As you said, there'd be probably no reddit drama if it was just atomic() with no operator to it 17:49
ugexe atomic operations are named after the order in which they do their operation
and what you get back 17:50
Zoffix .ask jnthn Docs for atomicint say overflow will wrap around silently. But in C, overflow on signed integers is undefined and I recall asking whether the same were true for our native ints and you said yeah because that's what hardware does, but doesn't the same apply to atomicint too?
yoleaux Zoffix: I'll pass your message to jnthn.
Zoffix ugexe: then why atomic-assign and not atomic-assign-fetch?
ugexe: also atomic-inc-fetch doesn't do any assignment, while atomic-add-fetch does, yet that fact isn't indicated in the name. 17:51
Anyway
Zoffix re-disappears
(well, yeah, it does do assignment, I guess; so never mind that point) 17:54
ugexe atomic-assign does not return anything, the -fetch stuff does 17:59
jnthn The position of the fetch before the add or inc is to indicate whether you get the value before or after the operation 18:01
yoleaux 17:50Z <Zoffix> jnthn: Docs for atomicint say overflow will wrap around silently. But in C, overflow on signed integers is undefined and I recall asking whether the same were true for our native ints and you said yeah because that's what hardware does, but doesn't the same apply to atomicint too?
jnthn Zoffix: I suspect everywhere we run does do wrap-around semantics, but yeah, strictly we should perhaps just say undefined behavior.
geekosaur yes, strictly you should. but also yes, de facto all platforms we care about wrap 18:02
and in fact so do most platforms we *might* care about; other behaviors quietly vanished into history a couple decades ago
Zoffix Ah, ok. thanks. 18:03
m: my atomicint $x = 42; say atomic-assign($x, 100)
camelia 100?
Zoffix m: my atomicint $x = 42; say $x ?= 100
camelia 100?
jnthn Note there's no fetch in that operation at all, you just get back exactly what you put in
Zoffix Damn. Just my lack. The doc build worked, but my table isn't rendering right :P 18:04
Oh, the operator symbols them selves have parts of the table lines 18:05
OK. Thanks. Now I get the Texas atomic naming :) 18:06
jnthn Also, inc vs add - they're very close under the hood
Both are "assignments" in that they update the content of a variable 18:07
Zoffix .tell lizmat nevermind. ugexe++ explained my mistake
yoleaux Zoffix: I'll pass your message to lizmat.
timotimo so, uh, anybody know anything about the pre-workshop meetup? 18:11
how do i indicate i'll be there and it'd be nice if a seat existed for me to sit on
jnthn timotimo: Mail the organiers
*organizers
timotimo let's see 18:13
done 18:15
ugexe m: grammar Foo { token TOP { <to> }; token to { .+ }; }; say Foo.parse("xyz"); # the source of Grammar::HTTP failures finally golfed 18:19
camelia ??? to => ???
timotimo ah 18:22
that's from when Grammar and Cursor became the same thing
YAML used to have a token "list" which clashed with the .list method on the match object
jnthn 'twas Cursor and Match became the same 18:23
timotimo oops 18:26
[Coke] manages to kick off a MS VS build... 18:54
optimize.c(666) warning == : operator has no effect, did you intend = ? 18:55
timotimo 666, eh? 18:56
[Coke] AHA 19:04
/* General is container check, so answer is yes. */
known_result == 1;
timotimo hah, good catch
[Coke] that's a buggo!
still on line 666 in master
should I resend to #moarvm?
will 19:05
OK. I have a set of build instructions that got me a working MS VC instance, 'nmake test' *mostly* passes. 19:07
(all failures are related to cpp and nativecall)
many failures in spectest. 19:10
Skarsnik hm interesting 19:18
I thought NC has msvc mangling stuff for c++
geekosaur platform C++ specs are encouraged to have standard name mangling, but nothing enforces this. msvc's C++ name mangling doesn't abide by the x86/x86_64 platform spec 19:20
and I don't know anyone actually implemented it for msvc, or if the necessary stuff (this includes runtime hooks for object manipulation etc.) is even documented without an NDA 19:21
[Coke] I have N cores allocated to this VM, and have test_jobs=5... but prove is just showing a single test being run at a time in 'spectest'. :(
AlexDaniel: I'll get a report out of here about the failures when it finally finishes. 19:22
ugexe TEST_JOBS or test_jobs? case matters on windows (although I dunno if the script itself normalizes this) 19:23
AlexDaniel cool 19:24
[Coke] it's got 5 slots showing in prove. only one is being used.
1/47 0? 0? 0? 0? 19:25
Looks like we have a hanging test on windows. 19:35
Skarsnik hm. Can I easily commit stuff on a repo, then reuse this commit to put in on a fork (like putting my change on MoarVM in my MoarVM fork) 19:36
[Coke] I don't understand the question; you want to update a fork with commits from the original repo that occurred post-fork? 19:37
perlpilot Skarsnik: cherry-pick? 19:38
Skarsnik: I mean, the "can I?" question most likely has a "yes" answer. But I don't quite get your question either. 19:39
Skarsnik well I changed stuff in Moar but in a repo Config.pl of rakudo pulled
now I want to put these change on my fork to make a PR 19:40
AlexDaniel Skarsnik: did you ā€œgit commitā€ it already?
Skarsnik No. I did nothing yet ^^
I could probably do a git diff and use it thou 19:41
AlexDaniel well if you did, then you'd be able to just ā€œgit pull /path/to/the/moar/repoā€ and it would pull the commit
(from your fork)
but yeah, git diff > stuff.patch 19:42
or just copy the files :)
[Coke] I can't cut and paste out of the VM. :| 19:47
Skarsnik What do you use? 19:48
ugexe its often an option you have to enable 20:00
geekosaur not to mention that if either host or guest is *nix, you get to guess whether it used PRIMARY or CLIPBOARD 20:01
ugexe also if you are using the command line then cut and paste works different
[Coke] ugexe: Yes, I went through the process of enabling clipboard and drag and drop, both ways, restarted the guest OS... nothing.
lizmat . 20:02
yoleaux 17:48Z <Zoffix> lizmat: Shouldn't ?+= and ?-= be called atomic-add-assign and atomic-sub-assign instead of atomic-add-fetch and atomic-sub-fetch, since ?= is atomic-assign, not atomic-fetch?
18:07Z <Zoffix> lizmat: nevermind. ugexe++ explained my mistake
ugexe using VMware fusion on OSX with windows guest I highlight text in the command prompt, press command+c which copies, then command+v pastes 20:03
pastes inside of and outside of the guest 20:04
[Coke] AlexDaniel: win64 failures: gist.github.com/coke/3d1f76b63d325...684f87e6d8 20:30
lizmat wow, that's quite noisy
AlexDaniel hm, IO stuff I can understand 20:31
but S02-literals\quoting.t ? :/
[Coke] one sec.
updated. 20:34
looks like q/x stuff.
I am happy to run this stuff every so often now that I have it in place. I'll post the instructions as well in docs/windows.md or something. 20:35
ugexe if the spectest can run in under 50 minutes on appveyor hardware... 20:36
I'm getting the same failures fwiw 20:37
[Coke] it took MUCH less than 50 minutes here. 20:38
trying to run with the installed perl6, I get Could not find Test::Util for one of them...
Zoffix Many of the Windows failures existed for ages. There are two open roast Issues: github.com/perl6/roast/issues/197 github.com/perl6/roast/issues/197 20:56
Second one: github.com/perl6/roast/issues/232
And I suspect a good portion of these aren't issues with rakudo, but just faulty tests.
AlexDaniel: qx// is shell quoting. 20:57
Zoffix &
AlexDaniel yea, I see
stmuk_ I see roast failures on 2017.08 FreeBSD (when it passed on 2017.07) 21:01
failures with t/spec/S17-supply/interval.t t/spec/integration/eval-and-threads.t and the suite then hangs on t/spec/S17-procasync/basic.rakudo.moar 21:02
lizmat just realized we have a "summary" function on irclog.perlgeek.de ? 21:59
timotimo yup! 22:00
there's a mode where you can edit which lines should go into the summary and one that shows only those
lizmat yeah, I just realized that 22:01
does that still work? Haven't been able to find one with a summary, though
good night, #perl6-dev! 22:05
long drive ahead tomorrow!
Zoffix Do you think I'm toxic or causing damage to the community? 23:07
Zoffix is watching node community implode and is self-reflecting a bit... 23:08
geekosaur you are sometimes a bit rough, tbh. but we don't really have the kind of toxic people I encounter in many other communities 23:19
(I am sometimes a bit rough too, not helped by having difficulty spotting it in myself until afterward :/ )
Zoffix reads medium.com/@audrey.tang/lessons-i-...f5d8107e34 23:23
AlexDaniel Zoffix: I think there were some toxic moments in the past (for example I miss gfldex), but lately you're super sweet 23:37
Zoffix There were someone else who left and never came back, when I tried to argue about what is and isn't a module... 23:38
geekosaur there was someone who left and demanded everything they'd ever done be removed 23:39
*that* is a form of toxicity, and I'm still waiting for some "very special" person to use that to completely destroy an open source project
"I have provacy law on my side, your CLA won't stand up in court" = "I reserve the right to completely destroy anything I ever contribute to, and as such should not be permitted to contribute to any project that anyone else actually cares about" 23:41
I also fully expect them to realize that they need to use the same privacy law bullying to get the record of that removed from public IRC logs, so nobody will be able to prove that a toxic bully is about to slit their project's throat 23:46
AlexDaniel eh, people go crazy sometimes, I don't think we should make such a big deal about it 23:48
23:51
Zoffix . 23:58
yoleaux:
zzzzzzzzzzz .tell Zoffix test 23:59
no bot?
yoleaux zzzzzzzzzzz: I'll pass your message to Zoffix.