github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
00:37 Kaiepi left, Kaiepi joined 00:41 patrickz joined 00:45 patrickb left 01:06 patrickz left 01:10 Kaypie joined 01:11 Kaiepi left 01:27 Merfont joined, Kaypie left 01:32 Kaypie joined, Merfont left 01:57 squashable6 left 01:58 squashable6 joined 02:06 Kaypie left 02:07 Kaypie joined 02:35 Kaypie left, Merfont joined 02:48 TimToady left, TimToady joined 03:00 squashable6 left 03:04 squashable6 joined 03:30 Kaypie joined 03:32 Kaypie left 03:33 Merfont left 03:35 Kaiepi joined 03:47 Kaypie joined, Kaiepi left 04:51 Kaypie left 04:56 Kaiepi joined 05:30 AlexDaniel left 05:52 Kaiepi left 05:53 Kaiepi joined 05:54 Kaiepi left, Kaiepi joined 05:55 Kaiepi left 05:59 Kaiepi joined 06:26 Kaiepi left 06:28 Kaiepi joined 07:05 domidumont joined 07:06 domidumont left, domidumont joined 07:08 lizmat joined 07:16 Kaiepi left, Kaiepi joined
samcv AlexDaniel`, i should release after the movers deliver all my belongings. i finally get my stuff today :) 07:27
well. actually they're unpacking the stuff now
AlexDaniel` sounds great 07:48
07:58 lizmat left, Kaiepi left 07:59 Kaiepi joined 08:04 zakharyas joined
timotimo o/ from GPW 08:34
08:58 Kaiepi left 09:00 Kaiepi joined
Geth MoarVM: 16d03884ec | (Timo Paulssen)++ | 3 files
Put Inline-preventing Insturction In Spesh Log
09:25
timotimo i'm seeing AT-POS with a fixed callsite that still has checkarity and param_rp_o (and i) inside it for some reason 09:35
it's probably that the arg flag is ARG_OBJ for an param_rp_i, because it's got an intlexref 09:40
whoops laptop battery almost completely empty 09:41
anyway, the boxed primitive of native ref is "none", so there'd potentially have to be an extra case there which seems totally doable
09:44 travis-ci joined
travis-ci MoarVM build errored. Timo Paulssen 'Put Inline-preventing Insturction In Spesh Log' 09:44
travis-ci.org/MoarVM/MoarVM/builds/502471514 github.com/MoarVM/MoarVM/compare/a...d03884ec42
09:44 travis-ci left 10:29 Kaiepi left 10:32 Kaiepi joined
timotimo wow that typo 10:56
some errors from package downloads from apt repositories it looks like 11:00
11:05 domidumont left 11:07 travis-ci joined
travis-ci MoarVM build errored. Timo Paulssen 'Put Inline-preventing Insturction In Spesh Log' 11:07
travis-ci.org/MoarVM/MoarVM/builds/502471514 github.com/MoarVM/MoarVM/compare/a...d03884ec42
11:07 travis-ci left 11:33 Kaiepi left, Kaiepi joined 11:44 zakharyas left 11:49 lizmat joined 12:16 lizmat_ joined 12:17 lizmat left 12:20 lizmat_ is now known as lizmat 12:26 lizmat left 12:35 lucasb joined 12:40 domidumont joined 12:55 wictory[m] left 13:37 brrt joined 13:43 lizmat joined 13:45 lizmat left, lizmat_ joined 13:47 zakharyas joined 14:02 lizmat_ left
samcv run() seems to be leaking memory 14:52
i'm testing running this p.tyil.nl/ej0v and checking and every time it runs the memory usage increases 14:53
it's been reported to me it will eventualy exhaust the system's entire memory
japhb samcv: Is it a regression, or is it an old problem? 15:19
samcv old problem i think 15:21
i think they're running the latest rakudo star release. and i can reproduce on latest git
15:27 ugexe joined
brrt run() of what? 15:28
samcv p.tyil.nl/ej0v see this
ugexe is it unexpected that a tight loop opening handles and not closing them is leaking?
yoleaux 29 Jan 2019 13:27Z <lizmat> ugexe: stackoverflow.com/questions/544214...nk-library
20 Feb 2019 16:55Z <lizmat> ugexe: www.reddit.com/r/perl/comments/asc...s/egvtxmc/
samcv AlexDaniel`, starting to release now 15:31
ugexe the way to ignore input is to use :!out -- otherwise if a handle like :out is opened then it needs to be closed. 15:33
samcv timotimo, also you only need to capitalize the first letter of commit subjects, not every one
sorry if i wasn't clear before 15:34
ugexe s/input/output/
15:39 brrt left
samcv AlexDaniel`, about to release 16:08
16:22 brrt joined
Geth MoarVM: f64f2ee1d8 | (Samantha McVey)++ | docs/ChangeLog
Add another entry to the ChangeLog
16:25
MoarVM: 7b80f2c58a | (Samantha McVey)++ | VERSION
Bump version for release 2019.03
16:31 AlexDaniel joined
Geth moarvm.org: e155efe0a0 | (Samantha McVey)++ | 4 files
Release 2019.03
16:31
lucasb yay, thank you samcv & others! 16:37
since it was tagged as 2019.03, I assume the "regular" third-saturday release will *not* happen this month, correct? 16:39
samcv lucasb, i believe that is true
AlexDaniel said as much at least
lucasb ok, just to know :) 16:40
16:47 MasterDuke left
AlexDaniel yeah, and fyi next month I'm won't have time to make a release 16:48
and a May release is also unlikely from me
so find a new release manager, or wait till June :) 16:49
the spot was open for quite a bit and it's a bit worrying that nobody is volunteering
ok we didn't stress the urgency so maybe that's why, but now we probably should 16:50
brrt samcv++ 16:51
AlexDaniel samcv++ for sure
why is that code not running out of handles, btw? 16:52
samcv i'm going to try running valgrind on that memory leak on using run()
AlexDaniel, the run?
AlexDaniel yea
maybe check how many open handles it has 16:53
samcv how many handles are allowed? like 40k?
how do i check how many are allowed
AlexDaniel depends on the system
samcv err. open
but yeah it runs out of memory first 16:54
AlexDaniel ulimit -n
samcv 1024
AlexDaniel also use lsof on the process to see how many it has
samcv the server has 1024. same as me 16:55
dunno how that factors into it
AlexDaniel hmm I don't see the number of handles increasing 16:56
samcv yeah if that was it i'd think it'd run out of handles much faster than memory
AlexDaniel though I'm also not seeing any growth in memory usage… 16:57
AlexDaniel reaches for more precise measurement tools 16:58
samcv i'm gonna try it with just "echo" command. and taking the sleep away
yeah the memory usage is going up pretty fast 16:59
AlexDaniel not really
samcv well. not *that* fast but steadly increasing about 1mb every 5 seconds
AlexDaniel I don't think it's linear
samcv well. i will run it for some time 17:00
AlexDaniel graphs it
IIRC I already looked into this a few months ago, and the situation was that while the memory does seem to increase at first, it eventually settles 17:03
and it was somewhat better in the past, but the behavior changed after jnthn++'s changes
samcv AlexDaniel, i got 8GB 17:04
AlexDaniel holy shit
OK… sounds like an issue :)
samcv if i do: my $cmd = run « cat ubuntu.iso », :out; $cmd.out.close; in a loop
looped about 10 times 17:05
it's a 1.8GB file
so it isn't equal to, but seems maybe 1/2 the size of the file that's been outputted?
or maybe it is the full amount hmm 17:06
AlexDaniel here's the graph: imgur.com/a/nZGmcno 17:08
that's for the curl snippet 17:10
slurping large files, hmm… that should be investigated
samcv i have it sleeping. and it's at 8.2GB usage 17:13
well i had it do run() and then close filehandle then sleep 10 seconds. it does those 5 times. then it just sleeps forever. i wanted to see if it would ever free it
it is not
AlexDaniel speaking of sleeping, I need to get the other half of my sleep :) 17:16
samcv AlexDaniel, what's the best way to see what's allocating that memory? 17:17
AlexDaniel btw nqp and rakudo master branches are open for new contributions
samcv: oh, maybe timotimo knows
timotimo: ↑ ?
the profiler can probably tell that, right?
18:11 zakharyas left 18:12 brrt left 18:31 domidumont left
Geth MoarVM: ce2152df1e | (Samantha McVey)++ | docs/strings.asciidoc
Fix typo in strings documentation

  Thanks to MidCheck for spotting it.
19:01
19:18 travis-ci joined
travis-ci MoarVM build errored. Samantha McVey 'Fix typo in strings documentation 19:18
travis-ci.org/MoarVM/MoarVM/builds/502720637 github.com/MoarVM/MoarVM/compare/7...2152df1e68
19:18 travis-ci left 19:33 domidumont joined 19:54 zakharyas joined 19:56 domidumont left
timotimo d'oh samvc, roger that 20:38
yoleaux 18:07Z <Kaiepi> timotimo: you've done networking work right? can you take a look at github.com/rakudo/rakudo/issues/2733 and comment on it?
timotimo i think sometimes i am dropping nputs thoug? 20:41
Kaiepi grr, i can't build the js vm without bugging the maintainers of some of it's dependencies to make them build properly on openbsd 20:48
such as github.com/node-ffi-napi/node-ffi-.../issues/40
shit wc 20:49
timotimo doh 20:54
kaiepi, what do you need input on, exactly? i'm not sure if making the listen method immediately bind the socket would be a good idea; i'm not sure if i really saw it somewhere or not, but the tapping being the point where something actually starts is a common occurence, right? 20:59
Kaiepi yeah
i had the same thoughts, dropping connections before the server's tapped isn't a very good idea 21:00
timotimo m: my $s = Supply.interval(3); $s.tap(*.say); sleep 3; $s.tap(say *x2:); sleep 3 21:02
camelia ===SORRY!=== Error while compiling <tmp>
Unable to parse expression in argument list; couldn't find final ')' (corresponding starter was at line 1)
at <tmp>:1
------> 3); $s.tap(*.say); sleep 3; $s.tap(say *⏏x2:); sleep 3
e…
timotimo m: my $s = Supply.interval(3); $s.tap(*.say); sleep 3; $s.tap({ say $_ x 2 }); sleep 3 21:03
camelia 0
1
00
timotimo like here it starts over at 0
i'm not sure about other examples right now
but proc::async comes to mind, where setting up inputs and outputs before the process actually starts 21:04
21:08 zakharyas left 21:24 MasterDuke joined, MasterDuke left, MasterDuke joined
MasterDuke samcv: heaptrack is a good tool for profiling memory allocations 21:25
timotimo hey MD how are you? 21:27
MasterDuke now that the release has happened (++s all around), i'm going to merge github.com/MoarVM/MoarVM/pull/1059 and github.com/MoarVM/MoarVM/pull/1038 later today or tomorrow unless there are any objections 21:28
timotimo: going to go see into the spiderverse in a couple minutes, so hoping that's as good as i've been hearing
timotimo i liked it a whole lot
looking forward to iNtQP 21:29
MasterDuke samcv, et al.: did you see that unicode 12 was just announced? blog.unicode.org/2019/03/announcing...n-120.html 21:31
timotimo: i haven't figure out what that stands for yet... 21:32
timotimo just a mash-up o fint and nqp, no further thoughts from me yet 21:33
MasterDuke ha, i was trying to come up with movie titles it fit
timotimo oooh
because into the spiderverse
i will join playing fluxx now 21:34
MasterDuke timotimo: i haven't seen much of a performance difference when on my default_int branches. i don't know if that's because a lot of the hot spots were already explicitly intified? or if my changes aren't deep/comprehensive enough? 21:37
21:51 Kaiepi left, Kaiepi joined
MasterDuke timotimo: btw, just got a couple of these, don't know how important it is: `src/spesh/inline.c: In function ‘is_graph_inlineable’: src/spesh/inline.c:111:37: warning: assignment discards ‘const’ qualifier from pointer target type [-Wdiscarded-qualifiers] *no_inline_info = ins->info;` 22:06
22:35 patrickb joined
timotimo thats not important, but it's also easy to shut it up :) 22:58
i would just have to look up where exactly the const goes. well i guess trying both ways would get the compiler to tell mer which one is wrong
maybe the int-qp will make more code optimizeable 23:11
i will have to look into that a bit, i suppose
23:15 lucasb left 23:30 patrickb left