Zoffix . 00:00
oh right it told me in private 00:02
japhb Zoffix: I think it is hard for anyone to live up to the example Audrey left for us (even she had trouble being everything she wanted to be, and I'm fairly certain she'd qualify as a straight-up genius.) 00:16
BenGoldberg This conversation reminds me Azer Koçulu, who unpublished 250 of his javascript modules on NPM, some of which was in very active use (most notably a string "leftpad" function). 00:17
japhb But she left an enduring legacy in the community to be better not just as programmers but as people and as friends. It's worth keeping that legacy alive, and even when we start to lose control, reminding ourselves to be the people we wish we could form a community around.
Zoffix Yeah. I also read this post of hers: pugs.blogs.com/audrey/2009/08/my-ho...gging.html 00:22
I recall my time as an op in #css. Reading a log right now of a conversation and 2007's Zoffix is just linking newbies W3C's specs instead of giving any useful answers and being a giant dick to everyone... 00:27
samcv good * 00:28
i'm back from seeing the total eclipse. traffic was so bad coming home
Zoffix And I'd say a lot of that could've been curbed if someone told me to STFU instead of letting me do whatever I want 'cause I was a channel op and answers to 99.99% questions.
\o 00:29
I only seen a slightly darkened sky and blinded myself a bit by trying to look at the sun :)
geekosaur we had about 80% here, I used a pinhole projector to watch like the last time we had an eclipse 00:30
samcv i had some binoculars. was at full eclipse for 2 minutes or so. 2 minutes is shorter than you might think. though it's exactly as long as you might think 00:31
Zoffix :)
geekosaur I see "8 second rule" got redefined for the Internet 00:33
used to be "people remember a total eclipse lasting 8 seconds"
samcv heh 00:34
oh yeah i was working on changing synthetic's moarvm struct to have a different format. Instead of having MVMGrapheme32 base; MVMGrapheme32 *combs; MVMint32 num_combs; i'm instead going to have it just be a single array 00:36
instead of storing base separately. and then i will store `base_index` which will point to which in the array is the base character. since currently ignoremark and other ops don't work properly for graphemes whose first codepoint is NOT the base character 00:37
m: say "\c[arabic number sign]9" ~~ /:m 9 /
camelia Nil?
samcv m: say "\c[arabic number sign]9" ~~ /:m "\c[arabic number sign]" /
camelia ??9??
samcv but making a change like this means i have to make changes tons of other places :\ and getting segfaults 00:38
[Coke] makes a double batch of homemade onion dip 01:11
news.perlfoundation.org/2017/08/uni...rship.html 01:39
japhb Didn't even realize that was a thing we needed, but good, I think? 01:50
samcv oh nice i didn't know that 02:39
so that means i can ask them random questions?
BenGoldberg Probably. 02:45
unicode.org/consortium/levels.html#associate 02:46
"As an Associate member, your organization has full access to email lists and technical committee meetings allowing you to contribute to the ongoing development of the Unicode Standard."
I expect that the Unicode Consortium would be happy if some member of The Perl Foundation (or someone appointed by them) were to join one or more of their mailing lists. 02:51
samcv i thought everyone could get email lists?
well i am on their main one but there's so much volume
BenGoldberg doesn't know. *Shrug*
samcv unless there's any secret lists for contributors or members or whatever
hmm
geekosaur committee lists, likely
BenGoldberg Being able to send someone to the next technical committee meeting sounds more interesting. 02:52
(Assuming it's physical, not virtual ;))
BenGoldberg wonders what TUC thinks of rakudo's synthetic codepoints. 02:53
Something ranging from "Ooh clever!" to "Madness!" to "yuck!" 02:55
geekosaur ...and then the next Unicode standard includes an NFG appendix... 02:57
samcv LOL 02:58
geekosaur (which will inevitably be *different*)
MasterDuke .ask jnthn in the atomicint example in the docs `my atomicint $total = 0; start { for ^20000 { $total?++ } } xx 4; say $total;`, shouldn't it be `await start`? with await i always get 80000 printed, without it's always some random number less than that 04:52
yoleaux MasterDuke: I'll pass your message to jnthn.
nine t/spec/S17-lowlevel/atomic.t seems to be failing here more often than not. At least when I run spectest with TEST_JOBS=22 which is kinda ironic considering the functionality it tests ;) 05:57
6 hours 25 minutes to Zurich :) 06:55
lizmat_ commute to SPW& 07:10
samcv sends best wishes 07:11
brrt yes, have fun 07:14
lizmat is being commuted 08:35
nine Oh great... track is blocked near Rosenheim 09:07
already one hour late 09:42
lizmat_ :-( 09:47
so far only minor delays on account of roadworks
Zoffix .tell jnthn I added the missing `awaits` in atomic docs 10:03
yoleaux 10:01Z <gfldex> Zoffix: you are not the reason why I stepped back from Perl 6. Your behaviour was just a casus belli. I wanted to actually use Perl 6 instead of making it more useable for others for a change. Sadly every time I try, I end up sending complains toward jnthn reguarding crashs and moar eating all my rams.
Zoffix: I'll pass your message to jnthn.
Zoffix MasterDuke: yeah, there should've been awaits.
nine Finally moving again. 2 hours 5 minutes late 10:35
Geth nqp/jit_nativecall: 068419589f | (Stefan Seifert)++ | src/vm/moar/QAST/QASTOperationsMAST.nqp
Map the new nativecallinvokejit OP

nativecallbuild now has a return value indicating whether we were able to JIT compile code for the call site
13:55
rakudo/jit_nativecall: 503bcdb367 | (Stefan Seifert)++ | lib/NativeCall.pm6
Use the new nativecallinvokejit for calling JITed NC code directly
Zoffix .tell tbrowder Small world. Just noticed that half a decade ago you reviewed one of my Perl modules :) cpanratings.perl.org/dist/Spreadshe...ataToExcel 15:04
yoleaux Zoffix: I'll pass your message to tbrowder.
dogbert17 todays c language question: both coverity and gcc complains about this line, github.com/MoarVM/MoarVM/blob/mast...lize.c#L16 15:17
the complaint is 'warning: statement with no effect [-Wunused-value]'
the question is why does it complain, this macro is used in other places and the compiler does not complain about those 15:19
coverity states 'CID 141567 (#1 of 1): Misused comma operator (NO_EFFECT)extra_comma: Part tc of statement tc , obj has no effect due to the comma.' 15:20
pmurias looks into it 15:22
dogbert17 ++pmurias
Zoffix Because the macro is defined as empty and so the code after preprocessing ends up as (tc, obj); 15:25
tbrowder Zoffix: ref Spreadsheet::DataToExcel, I'm still using it, but could use a Perl 6 version! 15:27
yoleaux 15:04Z <Zoffix> tbrowder: Small world. Just noticed that half a decade ago you reviewed one of my Perl modules :) cpanratings.perl.org/dist/Spreadshe...ataToExcel
Zoffix And the rest of the uses use something more complex as arguments, which perhaps make it think there is some effect, so it doesn't warn? 15:28
tbrowder: me too, though Rakudo version is a bit too hard for me :) I don't know how to make Excel without using a module 15:29
pmurias so I assume the macro should be fixed?
Zoffix pmurias: it's defined as something else when debug mode is used. I think the warning is benign. 15:30
dogbert17 I noticed that in all other uses the line is enclosed by a '#if MVM_GC_DEBUG' but not in this particular case
Zoffix Ohhhh
dogbert17++
pmurias: ^ that would explain it. The line should be put inside that stuff.
/src/gc/debug.h defines that macro to do things when #if MVM_GC_DEBUG and for it to be empty when it's turned off, so the line where the macro is used should go undere the same thing 15:32
dogbert17 Zoffix: I get the impression that you've managed to work through your C language book :) 15:33
pmurias Zoffix: I made a PR to fix that
dogbert17 pmurias++
Zoffix Or, and I'm not an expert here, but when the case when #if MVM_GC_DEBUG is off, define macro as MVM_ASSERT_NOT_FROMSPACE(a, b) instead of just MVM_ASSERT_NOT_FROMSPACE, that way the variables will be replaced with empty too and we won't need to stick #if MVM_GC_DEBUG around the macro 15:34
dogbert17: nah, I'm still only on page 265 out of 831, but I'm gettin' there! :)
dogbert17 perhaps we'll expose some nasty GC bug when that PR is applied
Zoffix: and how's your vacation going
Zoffix Oh, I forgot I was on one :) 15:35
I guess good. :)
dogbert17 :)
just avoid reddit and twitter :)
pmurias Zoffix: the later is what the PR does
Zoffix pmurias++ sweet 15:36
dogbert17: nah, I'm trying something new and reddit and twitter are good now.
.oO( It's working! )
dogbert17 relocates & 15:38
Geth rakudo: Skarsnik++ created pull request #1137:
Add a source-address parameter to IO::Socket::INet.pm
16:29
Skarsnik gah this pr sucks. It added some dumb commit >< 16:30
Zoffix git pull --rebase will avoid merge commits and will rebase instead 16:36
Skarsnik I think I will wait for the previous pr to be pulled before redoing this x) 16:46
Zoffix Observation: last release was 200th release of Rakudo, based on the available git release tags: github.com/rakudo/rakudo/releases 17:39
Skarsnik Yes !
samcv i'm trying to decide on the maximum length of a needle allowed for knuth morris pratt algorithm 19:32
so basically, we need to allocate an array the same size as the size of the string, so a 32bit integer array for the length of the needle. i haven't merged the KMP algorithm in yet but 19:33
atm i had it set at 100. but i could set it much higher. the thing is, it must process the needle before starting the index operation, so it does take some time
though most of the time needles are a lot smaller than 100 long 19:34
idk maybe i should set it to 200? anybody have any ideas? 19:35
moritz doesn't the algorithm become more efficient with larger needle length? 19:36
I'd go with something like 16k as the upper boundary or so
Skarsnik running out of space on / is not fun x) 19:39
samcv 16k? 19:40
i mean it has to process the whole size of the needle though
moritz is that slow? 19:41
samcv it's not slow but it does take the time you need to access all the elements in the needle
perlpilot samcv: what would happen when they specify a needle bigger than the maximum? Does it use a different algorithm? 19:42
samcv it falls back to how it did before which is brute force
like a needle that's 1000004 long it ends up being slower 19:43
perlpilot samcv: maybe 4K?
samcv a needle 100004 it's about even
yeah 4k i could do
also it has to allocate double the size of the needle in memory fyi 19:44
to do the algorithm
4k seems fine though 19:45
not sure if i should allocate onto the stack or use malloc. could put 4k graphemes on the stack if i wanted to 19:47
(i think)
not sure of stack sizes on non-linux os's 19:48
that's only like 16kb
should be fine and avoid a call to malloc 19:49
perlpilot you could make it auto-tune to the OS in use :)
samcv well since i already know the max size of 4000 graphemes that's 16kb which is small enough to put on the stack. though if i didn't have an upper bound i'd want it to use malloc 19:50
also if the needle is made up of strands it flattens it for the operation as well. so we could end up using even more space (though the flattened needle is malloced not put onto the stack) 19:51
but in my tests that is faster since accesses to the needle are just straight array lookups 19:52
Geth rakudo/nom: 593fa5f87f | (Samantha McVey)++ | tools/build/NQP_REVISION
Bump NQP/Moar to get Knuth-Morris-Pratt string search

Plus other fixes and improvements. Full MoarVM changes brought in: 371b579a Set the needle limit to 4096 for KMP algorithm 01a35b27 Shorten text of 'failed to resolve' exception 62f66cbf Ensure strings created with MVM_string_repeat are in NFG form ... (12 more lines)
23:20
¦ rakudo/nom: version bump brought these changes: github.com/perl6/nqp/compare/2017....g99f01d2cd
samcv bam
MasterDuke i don't see the moarvm bump in nqp? 23:25
samcv MasterDuke, ok fixed :) sorry 23:43
Geth nqp: 99f01d2cd8 | (Samantha McVey)++ | tools/build/MOAR_REVISION
Bump MoarVM to get Knuth-Morris-Pratt string search

Plus other fixes and improvements. Full changes brought in: 371b579a Set the needle limit to 4096 for KMP algorithm 01a35b27 Shorten text of 'failed to resolve' exception 62f66cbf Ensure strings created with MVM_string_repeat are in NFG form ... (12 more lines)
¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...-g151a2563