Perl 6 language and compiler development | Logs at colabti.org/irclogger/irclogger_log/perl6-dev | For toolchain/installation stuff see #perl6-toolchain | For MoarVM see #moarvm
Set by Zoffix on 27 July 2018.
AlexDaniel timotimo: hmm, Geth? 09:57
and hack?
timotimo sorry, i somehow managed to not "get it" that "geth is down" means "hack might need restarted" 10:29
AlexDaniel timotimo: can we figure out why it happens? 10:38
timotimo probably, with a bit of work 10:40
the first step might be to ensure we have a working shell on the vm even when the disk has disappeared, for example by leaving a root shell open (sounds like a great idea!) logged into a statically compiled busybox that's been loaded into memory completely 11:16
that way we'd be able to poke around in /sys and /proc and such
nine Where does hack send its logs to? 11:40
timotimo i don't know 11:50
and i'm not entirely sure how to find out :)
hm. an rsyslog daemon is running, but /var/spool/rsyslog is empty, so it's probably not going there 11:54
oh, maybe the logs actually go to www.p6c.org 11:58
at least there's an entry in /etc/rsyslog.d/remote_www.p6c.org.conf on hack
yeah, they live in /var/log/hack.p6c.org/ 12:00
timotimo it doesn't seem like the logging actually reaches www, though 12:02
timotimo i wonder if rsyslogd on hack dies from lack of disk and doesn't continue with the next output, so before it can send logs it waits forever for some disk access to finish 12:03
nine Sounds reasonable. Is it possible to switch the outputs? 12:05
timotimo surely 12:08
hm 12:11
the include directive is in front of the rules that split the log messages into different files
but i don't know if it's "ordered" 12:12
nine Wait, does it log anything at all to www? 12:16
timotimo yes
nine Ok, so we're really just missing the interesting log entries
timotimo i think so
or perhaps those aren't even reaching rsyslog
no clue, tbh
nine Is rsyslogd the only loggin daemon or is journald involved, too? 12:18
timotimo i believe journald also runs, yeah
it gets kernel messages
ah, and a bunch of other stuff, too 12:19
looks like everything
nine Maybe using systemd-journal-remote is more reliable?
timotimo perhaps rsyslogd gets its data from journalctl?
could be. do you want to set it up? :)
nine Nice. The journal supports both push and pull 12:25
But hack is a debian system and I don't have sudo rights and this apt stuff looks much more complicated than zypper 12:28
timotimo heh. 12:31
i'm lots distracted ATM 12:32
nine Btw. looks like I actually don't need $*MAST_FRAME at all. I can put it into an attribute in QASTCompilerMAST just as easily.
timotimo oh! 12:33
that should give another percent or two
nine timotimo: 1-2 % was a really good guess 14:25
timotimo damn 14:27
nine I wonder if it'd be faster to use integer constants instead of string literals in compile_var 15:00
timotimo nqp isn't very good at constants, but you should try it anyway 15:05
nine Maybe. But a closer look at fresh_register yielded about the same speedup as all my other efforts today ;) 15:31
timotimo you mean you just doubled your improvements for the day? 15:33
nine Sounds much better :) 15:34
lucasb m: dd ([slip,slip],[Slip,Slip]) 15:39
camelia ([], [Slip, Slip])
lucasb Shouldn't a Slip type object be as slippery as a concrete slip value?
timotimo no, type objects are always itemy 15:41
lucasb ah, right. thanks 15:42
nine timotimo: btw. your hint to set $!index as box_target in MAST::Local was just great!
Saves so much coercion and boxing 15:43
timotimo <3 15:44
glad to hear it
hm, actually, they might sometimes not be itemy, probably some way to override that behaviour 15:45
lucasb 'Nil.new =:= Nil' is always true. Do you think makes sense to make argument-less slip 'Slip.new =:= Empty' be true? 15:57
in Slip.pm: multi method new(--> Empty) {} # ? :) 15:59
timotimo that way you could "does" Empty and get the same changes every time you write Slip.new 16:03
lucasb hm, not sure I understood :) 16:07
m: use MONKEY-TYPING; Empty does role { method foo { 42 } }; augment class Slip { multi method new(-->Empty) {} }; dd slip.foo
camelia 42
lucasb ^^ related idea?
timotimo yeah 16:08
Geth nqp: 06b0001d84 | (Aleks-Daniel Jakimenko-Aleksejev)++ | tools/build/MOAR_REVISION
[MoarVM Bump] 4454cfd99 fix jit perf map with […]

MoarVM bump brought: github.com/MoarVM/MoarVM/compare/2...g4454cfd99
17:45
¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...g4454cfd99
rakudo: 0d5a813240 | (Aleks-Daniel Jakimenko-Aleksejev)++ | tools/build/NQP_REVISION
[NQP Bump] 06b0001d8 [MoarVM Bump] 4454cfd99 f […]

NQP bump brought: github.com/perl6/nqp/compare/2018....g06b0001d8
17:46
rakudo: version bump brought these changes: github.com/perl6/nqp/compare/2018....g06b0001d8
f1fa5adfff | (Timo Paulssen)++ | 2 files

so that assigning a Seq into a native array won't make it grow more and more each time
lizmat Files=1255, Tests=76369, 344 wallclock secs (15.70 usr 5.35 sys + 2429.93 cusr 225.53 csys = 2676.51 CPU) 17:59
timotimo writing a test for this nao 19:11
i "git reset" the wrong repo %) 19:17
Geth roast: 1725e7e076 | (Timo Paulssen)++ | S09-typed-arrays/native.t
test repeated STORE Seq on native arrays

an unfiled bug in rakudo caused those to result in just all values being concatenated.
19:44
lizmat timotimo: good catch! 20:08
timotimo tanks 20:27
may want to go through all method STORE across the codebase and check for this issue 20:28
and write tests perhaps
lizmat :-) will do 20:35
Geth rakudo: 2e5be5b8e4 | (Elizabeth Mattijsen)++ | src/core/Hyper.pm6
Handle PredictiveIterators a bit better

  - one check less per iteration at the expense of a little initial overhead
  - throw exception immediately if lengths don't match
   - rather than iterating until one of them is exhausted
lizmat m: my @a[10] = ^6; @a = 5; dd @a # one problem found 20:55
camelia Array element = Array.new(:shape(10,), [5, 1, 2, 3, 4, 5, Any, Any, Any, Any])
timotimo ah, interesting 20:57
i wonder if that's perhaps specced?
lizmat m: my @a[10] = ^6; @a = ^3; dd @a # another problem found 20:58
camelia Array element = Array.new(:shape(10,), [0, 1, 2, 3, 4, 5, Any, Any, Any, Any])
lizmat timotimo: not sure
gonna fix it and see if any spectests break
timotimo OK 20:59
Geth rakudo: 7262b4e840 | (Elizabeth Mattijsen)++ | src/core/Shaped1Array.pm6
Fix re-initialization of 1dim shaped arrays

Nudged by timotimo++
22:59
lizmat timotimo: no noise from spectest 23:00
will continue looking for others tomorrow
afk&
timotimo frames per second: 23:15
4963.6734 4871.4332 4782.5587 4660.3378 718.3257
actually it's lines per second, but it redraws the screen after every individual line, so ... :) 23:16
m: say "full screen redraws were between { 4963 / (786 / 2) } and { 718 (786 / 2) } per second" 23:28
camelia 5===SORRY!5=== Error while compiling <tmp>
Two terms in a row
at <tmp>:1
------> 3e between { 4963 / (786 / 2) } and { 7187⏏5 (786 / 2) } per second"
expecting any of:
infix
infix stopper
statement e…
timotimo m: say "full screen redraws were between { 4963 / (786 / 2) } and { 718 / (786 / 2) } per second"
camelia full screen redraws were between 12.628499 and 1.826972 per second