00:54 benchable6 joined 01:55 ilbot3 joined, committable6 joined 02:03 committable6 joined 02:19 bisectable6 joined 03:19 bisectable6 joined, committable6 joined 05:57 brrt joined
brrt good hi 05:57
good find timotimo
06:08 domidumont joined 06:12 domidumont joined 06:27 domidumont joined 06:32 patrickz joined 06:47 brrt joined, leont joined 07:42 leont joined 07:54 rba joined
jnthn morning o/ 08:14
brrt good monring jnthn
jnthn o/ brrt
samcv o/ jnthn brrt 08:15
brrt \o samcv 08:16
samcv grant was approved :) 08:17
08:19 robertle joined
brrt \o/ 08:20
good work samcv++
jnthn samcv++ \o/
08:28 releasable6 joined
nwc10 ASAN excitement in t/02-rakudo/repl.t paste.scsys.co.uk/565286 08:41
09:24 rba_ joined 09:49 robertle_ joined
brrt clickbaits www.youtube.com/watch?v=N5_drt7TEqE 10:46
my sentence construction
is not entirely awesome 10:47
stmuk brrt: your talks are excellent and very clear in my experience 11:05
brrt stmuk: thanks :-). i'm trying to be critical mostly though 11:07
i want to learn from it and do better
11:12 rba joined
dogbert17 jnthn: wrt t/spec/S17-promise/stress.t, irclog.perlgeek.de/moarvm/2017-10-06#i_15266626 11:22
stmuk 11:38
11:52 zakharyas joined
timotimo testing 12:10
hmpf something's going wrong with glowing-bear in my windows vm
i can see new messages show up but i can't press enter or click send when i write my own
MasterDuke glowing-bear? 12:12
timotimo a weechat relay frontend in the browser 12:13
MasterDuke ah
timotimo wait a minute
i'm writing messages from inside my vm
what did i change?
MasterDuke with dragon?
timotimo Testing testing 123
As soon as Dragon is turned on I can no longer use enter to send messages but apparently you can double click on the send message icon 12:14
12:14 ggoebel joined
timotimo But fortunately I can just say press control enter instead of press enter and it seems to work 12:15
MasterDuke most software is not very accessible. one of my coworkers is blind, and the trouble he has with just about everything...
timotimo No it turns out I first have to say press shift enter and then backspace and then press control enter and then it works
And actually I can leave out the backspace in the middle too 12:16
MasterDuke timotimo: can rr recordings be shared? if i put the one i have on the whateverable server would you be able to replay it? 12:18
timotimo i can replay it by ssh-ing onto the server at least; i'm not sure if they have been able to make it portable yet
MasterDuke ah. i made it on my machine, instead i should make one on the server? 12:19
timotimo that might make it easier, yeah
MasterDuke are you actually interested in taking a look (won't bother otherwise)? 12:20
and feel free to say no 12:21
timotimo right now trying to do as little as i can so my wrists will shut up about their pain
MasterDuke you're typing all this? 12:22
timotimo Only sometimes
I assume you can tell by proper capitalization or the lack thereof
MasterDuke heh, thought that might be a tell 12:23
timotimo brrt: we should call it reg_Ɵ instead of reg_sz 12:26
brrt lol 12:33
timotimo for people who can't type Ɵ we can also accept Åæz or Åæs 12:34
nwc10 :-) 12:36
incoming rocket: www.youtube.com/watch?v=SB4N4xF2B2...e=youtu.be 12:37
12:37 rba joined
nwc10 (OK, outgoing *currently*) 12:37
timotimo was about to say
MasterDuke brrt: in the jit to documentation there are several things listed, and i know one feature of the new jit is that it's supposed to be easier for non-you to work on, it might be nice to give a rough difficulty level to the different next steps 12:38
nwc10 meanwhile, the less time critical thing I was typing was "there's 'Ɵ' in my (correctly written) adress, hence that "suggestion" would be useful in my current change-of-address correspondance"
(...straƟe)
MasterDuke timotimo: huh, rakudo built and installed just fine on the server 12:39
brrt: *todo documentation 12:40
12:43 rba_ joined
brrt oh, that thing 12:43
that's all things that are, i apologise for the ehm, presumption, my-level kind of things
they're fairly deep, is what i mean 12:44
not that i wouldn't want anyone else to take them off my hands
MasterDuke heh, they did seem that way
lizmat stage 1 landed ok :-) 12:45
brrt sorry about that šŸ˜¬
nwc10 boring++
lizmat nwc10: I'll take it
nwc10 me too
"the HD full colour live stream keeps cutting out just before the rocket lands. Every time. Life sucks" -- # nice problem to have 12:47
timotimo that's because they can't fake HD full color yet
all of these rocket things are just attempts to make the population believe the earth is anything but flat
clearly
lizmat next thing you'll hear is that they complain about the noise :-)
brrt ā€¦ SpaceX is a fairly amazing organization 12:48
timotimo good talk, brrt 12:55
brrt thanks 12:57
13:00 domidumont joined
lizmat brrt: any chance of a blog post explaining people how to work with the new JIT, to coincide with mentioning the video of your presentation in tonight's P6W ? 13:02
brrt: or a link to the documentation you think would be a good intro? 13:03
timotimo brrt: should i look into refreshing the "parse graph.c" script? :) 13:04
brrt ehm, lizmat, maybe i can give youā€¦. 13:11
github.com/MoarVM/MoarVM/blob/mast...erview.org 13:12
timotimo: that would be awesome i think 13:13
lizmat: i will try and write a blog post in the near future, but i usually take way to long to finish one 13:14
lizmat thanks for considering it :-)
MoarVM oops: Spesh inline: fix_coderef NYI 13:19
timotimo: ^^^ ring a bell ?
say + ^100000 .grep( *.indices("7") == 2 ) # gives error 13:20
evalable6 7290
lizmat say + ^100000 .grep( { .indices("7") == 2 } ) # creates profile ok
evalable6 7290
13:23 MasterDuke_ joined 13:32 evalable6 joined
timotimo lizmat: could need to recompile rakudo so its C parts get all the new header definitions 13:37
hm. actually ... don't know :\
lizmat you mean we need a bump ?
timotimo nah, what i means occurs when a newer version of moar has been built but the C portions of rakudo have been built with an older version 13:38
but it's probably not what's going on here
lizmat I would doubt that
indeed
timotimo how did you trigger this?
we probably want to turn that oops into a "well, don't inline this, then" thing for the time being
lizmat $ 6p 'say + ^100000 .grep( *.indices("7") == 2 )' 13:39
MoarVM oops: Spesh inline: fix_coderef NYI
timotimo hm, versions? it works fine locally
oh
i missed the +
doesn't oops though 13:40
lizmat so maybe it's already fixed and we need a bump ?
timotimo wow my local moar is way out of date 13:41
no, i was just holding it wrong 13:42
jnthn Can't reproduce it here either, fwiw 13:44
timotimo the fix_coderef NYI error message has been there since 2015-05 and i haven't seen it tripped yet
lizmat I guess we have some OS dependencies in there then, because it's 100% repeatable for me 13:52
more data points: 13:54
say + ^99 .grep( *.indices("7") == 2 ) # works
evalable6 1
lizmat say + ^100 .grep( *.indices("7") == 2 ) # produces the correct result, but then says: MoarVM oops: Spesh inline: fix_coderef NY
evalable6 1
13:55 robertle joined
lizmat say + (77..175).grep( *.indices("7") == 2 ) # works 13:55
evalable6 1
lizmat say + (77..176).grep( *.indices("7") == 2 ) # produces result, then says oops 13:56
evalable6 1
jnthn I can't even get it with: MVM_SPESH_NODELAY=1 MVM_SPESH_BLOCKING=1 ./perl6-m -e 'say + ^100000 .grep( *.indices("7") == 2 )'
lizmat so it's the 101st iteration that makes it go
jnthn Built HEAD of everything just to check
lizmat jnthn: you need the --profile !!!
jnthn Oh
ah, then I get it
Ah, fine 13:57
File a MoarVM issue
lizmat will do :-)
jnthn If it only happens under --profile I'm a lot less worried :)
timotimo oh, so "6p" is what you use to --profile
lizmat sorry, should have been more explicit: the "6p" is very clear to me, but maybe not for anybody else sorry
timotimo i totally didn't see the "creates profile ok" bit
jnthn I was wondering what the p meant :P
lizmat timotimo: yeah, saves some typing 13:58
6 basic "time perl6 -e"
6l "perl6 --ll-exception -e'
6p "perl6 --profile -e'
jnthn Nice :)
lizmat alias 6='time perl6 -e' 13:59
alias 6l='perl6 --ll-exception -e'
MasterDuke_ i have some aliases for valgrind, gdb, perf, heaptrack, etc.
lizmat alias 6p='perl6 --profile -e' 14:00
^^ for the lazy people :-)
fwiw, I also have "alias 5='perl -E'" :-)
timotimo Oh I need to steal some of these 14:03
also I just figured out that the problem with Dragon and press enter it just happens that in chrome but not in Firefox 14:04
on the negative side it looks like I can only dictate new words into the input field but not select previous words or change anything
lizmat jnthn: github.com/MoarVM/MoarVM/issues/720 14:06
MasterDuke_ param_on_o is what's causing the BAIL of indices 14:13
brrt well, the question is why that is still a optional named parameter 14:29
14:36 tadzik1 joined, tadzik1 left
dogbert17 timotimo: have you seen this before? gist.github.com/dogbert17/cbc78851...01701f571d 15:04
15:11 brrt joined 15:28 rba joined
dogbert17 jnthn: stumbled upon a gc related error during spectest. Should I report it or is it old hat? gist.github.com/dogbert17/18d11a6e...3932f80423 15:33
jnthn dogbert17: looks new 15:34
dogbert17 ok, will create an issue then 15:36
MasterDuke_ jnthn: do you happen to have any suggestions on debugging what i'm seeing here? gist.github.com/MasterDuke17/5ca4c...c515c8ab76 16:09
this is my 'use the FSA for string storage' branch. it's trying to fsa_free 0 bytes of a null pointer 16:10
jnthn Umm...no, that looks odd 16:12
16:13 AlexDaniel_ joined
MasterDuke_ if i used rr correctly, when i ran it backward from the segv, it was showing the address of the body.storage.any getting zeroed here github.com/MoarVM/MoarVM/blob/mast...ect.c#L168 16:14
jnthn That'd imply it got collected 16:15
So...missing mark?
But..that'd be odd
MasterDuke_ i kind of feel like i'm running into something that i didn't cause with my fsa changes, but somehow uncovered 16:17
but i wouldn't bet a whole lot on that
jnthn Yeah but it's a bit odd it'd not blow up without your changes.
Though I guess MVM_free will ignore a NULL pointer 16:18
ilmari free(NULL) is required to be a no-op by POSIX
jnthn I guess you've tried using MVM_GC_DEBUG set to some value, or tweaking the nursery size in collect.h to GC stress things a bit?
If not, try those...
MasterDuke_ haven't yet, will give it a shot 16:19
jnthn OK, those will maybe provide some further hints of what's up
MasterDuke_ should MVM_GC_DEBUG=1|2 cause stuff to be printed? 16:21
jnthn No, it'll cause an MVM_panic if it detects anything 16:22
MasterDuke_ still segvs with that set to 1 or 2 16:23
jnthn Hm, OK 16:24
And with nursery shrunk?
I normally set it to something around 16KB
MasterDuke_ trying that now 16:26
still segv 16:27
16:39 dogbert2 joined
dogbert2 jnthn: the problem mentioned earlier can now be found in github.com/MoarVM/MoarVM/issues/721 16:40
16:43 robertle joined
timotimo jnthn: how do you feel about the idea of consolidating FSA-alloced pointers at some point? compaction, basically? 16:44
i've yet to measure, but i fear that the memory usage i'm experiencing from the heap analyzer could be in big part to internal fragmentation of the fsa pages 16:45
Geth MoarVM: zoffixznet++ created pull request #722:
Fix typo in comment		 17:06
17:09 domidumont joined 17:13 rba joined 17:20 rba joined
MasterDuke .tell brrt nice talk. now which templates are the most needed? 17:30
yoleaux MasterDuke: I'll pass your message to brrt.
17:38 lizmat joined
stmuk 17:39
17:40 leont joined
Zoffix "The hFile parameter must refer to a file stored on a seeking device; for example, a disk volume. Calling the SetFilePointer function with a handle to a non-seeking device such as a pipe or a communications device is not supported, even though the SetFilePointer function may not return an error. The behavior of the SetFilePointer function in this case is undefined." 17:46
From msdn.microsoft.com/en-us/library/w...s.85).aspx looks to be why $*OUT.tell always gives 0 on Windows
MasterDuke jnthn, timotimo: no change with MVM_GC_DEBUG=1|2 and/or setting MVM_NURSERY_SIZE to 16k or 4k, still a segv. anything else to try? 17:53
timotimo does the location of the segv change at all?
MasterDuke nope
timotimo hrumpf :(
MasterDuke i know
timotimo the community/ page on the perl6.org site still links to Perl 6 Testers even though zef hasn't been sending reports that way for a couple of months 17:54
hm, ww, but also: 17:55
there's a "getting involved" part to that page that could mention the squashathon
Zoffix huh... I fixed the bug 18:15
ZofBot: and I even know what I'm doing this time! 18:17
ugexe timotimo: i think something is wrong with the /report endpoint 18:19
timotimo i have only vague recollection of looking into it last time
dogbert2 found a SEGV 18:23
in something called p6decontrv 18:24
Zoffix sounds familiar 18:26
Though I think it was fixed
(awhile back)
dogbert2 here's the usual gist gist.github.com/dogbert17/48792636...00548ebded 18:27
Zoffix: OT but I picked up Doom yesterday for my new comp 18:29
Geth MoarVM: 669b344296 | (Zoffix Znet)++ (committed using GitHub Web editor) | src/io/syncfile.c
Fix typo in comment
 18:30
MoarVM: fbef62210b | (Jonathan Worthington)++ (committed using GitHub Web editor) | src/io/syncfile.c
Merge pull request #722 from zoffixznet/patch-2

Fix typo in comment
jnthn timotimo: We can't move FSA-allocated memory, we just don't keep track of it like that 18:31
timotimo: We could throw out totally empty pages, or we could introduce a limit on the number of pages we allow and fall back to malloc 18:32
I am just a tad worried that something I introduced primarily as a way to get good re-use of MVMFrame, ->work, ->env, and to serve as a safepoint mechanism for concurrent data structures (which also get high re-use rates due to thier typical tree nature) is now becoming used more widely. Yes, it's faster, but it's fast because it can make assumptions. One of its base assumptions from day one has been that it doesn't have to care about fragmentat 18:35
So I'd say if we've problems with fragmentation then we might just be using the FSA inappropriately. 18:36
timotimo ah, mhm 18:49
in the mean time i came up with a way that would probably give me much better memory use anyway; write large amounts of data into single arrays and shove them off to the other thread for analysis 18:50
much like the spesh log mechanism, actually
18:51 lizmat_ joined
timotimo also, maybe i'll use one array per size class so i won't waste as much memory 18:51
Geth MoarVM: zoffixznet++ created pull request #723:
[WIN32] Fix ->seakable detection on TTY handles		 18:52
timotimo also, my idea of FSA fragmentation relied on VMArray using the FSA already. it doesn't though 18:53
so i must search my problem somewhere else :)
19:13 lizmat joined
lizmat nullprogram.com/blog/2017/10/06/ # interesting, a branchless UTF-8 encoder 19:24
19:36 travis-ci joined
travis-ci MoarVM build failed. Jonathan Worthington 'Merge pull request #722 from zoffixznet/patch-2 19:36
travis-ci.org/MoarVM/MoarVM/builds/285725606 github.com/MoarVM/MoarVM/compare/6...ef62210b78
19:36 travis-ci left
ugexe that was a load bearing comment apparently 19:38
Zoffix Nah, it was missed plan bump in nqp test. Fixed in github.com/perl6/nqp/commit/ce60c2bf38 19:40
MasterDuke is this safe? github.com/MoarVM/MoarVM/blob/mast...iter.h#L33 20:05
memcpy isn't needed? 20:06
timotimo strings are immutable already
MasterDuke but if s is GCed or something like that?
timotimo we don't expose grapheme iterators to user code, do we? 20:07
i'd assume whereever we have a grapheme iterator on some string we'll also be rooting the string (and keeping it alive that way) 20:08
MasterDuke check out gdb.log in gist.github.com/MasterDuke17/5ca4c...c515c8ab76
does result also need an MVMROOT here? github.com/MoarVM/MoarVM/blob/mast...ops.c#L552 20:09
timotimo only if there's something that allocates an object in there somewhere 20:10
MasterDuke `result = (MVMString *)MVM_repr_alloc_init(tc, tc->instance->VMString);` ? 20:13
timotimo the GC happens before the function returns, and repr_alloc_init makes sure the return value is already the right one 20:14
MasterDuke hm. then nothing stands out in the backtrace? 20:20
timotimo unless it's reliably crashing under substring ... don't think so :\ 20:21
MasterDuke the backtrace is always the same
except that the ones in rr and valgrind don't have the substring stuff, that might be the nursery size change i just made 20:22
timotimo it could just be the program spends most of its time substringing stuff 20:23
MasterDuke could be. it's one of the final steps in rakudo's `make install` 20:24
timotimo anything interesting in dump_backtrace?
MasterDuke gdb.log in gist updated 20:26
afk for a bit 20:29
timotimo that doesn't look far into execution 20:34
lizmat and another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2017/10/09/...r-empathy/ 20:48
timotimo cool 20:55
(i had to wait a few minutes because i forgot to re-enable my zram swap stuff after rebooting and was daring enough to start both browser and mail client)
wait, out of cpan? 20:57
oh, that kind of "getting out,
"
21:07 rba joined
Zoffix What version of libuv are we using? 21:26
timotimo 1.14.1 is what we have in master
there was a pull request that included upgrading to a newer one
Zoffix This one github.com/MoarVM/MoarVM/pull/719 21:27
Looking at the windows proc issue. Found one closed libuv issue that had similar symptoms that said to use 1.x branch for fix; so I guess that's not the issue we have 21:28
timotimo yes that's the one
timotimo restarts a few of the failed travis builds for that PR 21:29
Zoffix Issue being `run $*EXECUTABLE, '-e', ļ½¢q|'|ļ½£;` fails with "The syntax of the command is incorrect" and `run $*EXECUTABLE, '-e', ļ½¢q|"|ļ½£;` with '"' is not recognized as an internal or external command,; and `run $*EXECUTABLE, '-e', ļ½¢(q|"|)ļ½£;` with `')"' is not recognized as an internal or external command,`
timotimo *sigh* what's with travis and osx jobs not starting 21:30
MasterDuke timotimo: fyi, this gist.github.com/MasterDuke17/5ca4c...tfile1-txt is the actual command i was running 21:32
timotimo i'm already failing to go easy on my wrists >_< 21:37
MasterDuke use that tame dragon you've got! 21:38
timotimo it's hard to use for gdb :D
MasterDuke train your cat? 21:39
timotimo good idea
Zoffix If I mod libuv and dump out the args it's trying to run with, it uses `C:\rakudo\perl6.bat -e "say q|foo\"bar|"` which fails on Windows. If I escape the quote by using double double quote instead of backslashing it, then it works. 22:23
Zoffix makes a trip to #libuv 22:26
22:30 ggoebel joined
Zoffix Didn't get anything concrete but got some suggestions to try logs.libuv.org/libuv/latest#22:29:44.993 22:45
Gonna take a crack at it tomorrow, when I'm less sleepy
Geth MoarVM: cfb0bffc09 | (Zoffix Znet)++ | src/platform/win32/io.c
[WIN32] Fix ->seakable detection on TTY handles

Fixes RT#132254: rt.perl.org/Ticket/Display.html?id=132254 Fixes all t/02-rakudo/repl.t failures
The SetFilePointer() routine silently fails[^1] when the handle is not a seekable handle, so check the handle is a disk handle[^2] before ... (7 more lines)		 22:47
synopsebot RT#132254 [new]: rt.perl.org/Ticket/Display.html?id=132254 [WIN32] $*OUT.tell always gives 0
MoarVM: 631b3bf0f0 | (Zoffix Znet)++ (committed using GitHub Web editor) | src/platform/win32/io.c
Merge pull request #723 from zoffixznet/fix-Win32-IO-tell

  [WIN32] Fix ->seakable detection on TTY handles
Zoffix Oh wait, I did get concrete suggestions :D 22:53
Geth MoarVM: ugexe++ created pull request #725:
Fix cwd error message		 23:00
MoarVM: b2c5ce0ba7 | (Nick Logan)++ (committed using GitHub Web editor) | src/io/dirops.c
Fix cwd error message
 23:03
MoarVM: bd8e6687a7 | lizmat++ (committed using GitHub Web editor) | src/io/dirops.c
Merge pull request #725 from ugexe/patch-2

Fix cwd error message
23:19 travis-ci joined
travis-ci MoarVM build passed. Zoffix Znet 'Merge pull request #723 from zoffixznet/fix-Win32-IO-tell 23:19
travis-ci.org/MoarVM/MoarVM/builds/285820777 github.com/MoarVM/MoarVM/compare/f...1b3bf0f02c
23:19 travis-ci left 23:56 travis-ci joined
travis-ci MoarVM build passed. lizmat 'Merge pull request #725 from ugexe/patch-2 23:56
travis-ci.org/MoarVM/MoarVM/builds/285825133 github.com/MoarVM/MoarVM/compare/6...8e6687a721
23:56 travis-ci left

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!