00:02 Skarsnik joined 00:21 [Tux] joined 00:28 skids joined 03:00 japhb joined 03:02 lizmat_ joined 03:26 japhb joined 03:32 geekosaur joined 03:56 geekosaur joined 06:44 geekosaur joined 07:30 RabidGravy joined
[Tux] test 21.761 08:10
test-t 12.156
csv-parser 25.390
12.210 on a second run
08:27 lizmat joined
RabidGravy :-\ 09:48
wasn't it below 10 at some point earlier in the year?
timotimo tux.nl/Talks/CSV6/speed4.html
it hasn't dipped below 11.9 or something 09:49
RabidGravy yeah, false memory
[Tux] $ grep 'test-t 1[01]\.' *log 09:50
2016-01-22 16:09:12 test-t 11.937
2016-02-09 11:28:24 test-t 11.750
2016-02-10 08:35:50 test-t 11.839
2016-02-11 08:38:52 test-t 11.884
2016-02-15 18:20:21 test-t 11.776
2016-02-19 11:55:51 test-t 11.672
2016-02-21 11:07:21 test-t 11.956
2016-02-23 08:34:45 test-t 11.934
2016-02-25 08:23:23 test-t 11.828
2016-02-26 09:15:45 test-t 11.889
timotimo not bad!
we'll get back there soon.
[Tux] hopes that that info is not perceived as throttle 09:51
RabidGravy it would be an interesting if somewhat tedious endeavour for someone with a lot of time on their hands to bisect through the timings
[Tux] unlikely I will be posting new timeings this week 09:52
want me to automate the graphs for the upcoming week?
RabidGravy :)
timotimo i thought they were already automated, huh.
[Tux] nope. all manual 09:53
the graphs *are* automated from the manual commands though
timotimo okay!
[Tux] tests automated build … 09:56
RabidGravy [Tux]++ 09:59
lizmat will be doing a 20 minute ad-hoc Q&A session at the NLPW this afternoon at 16:00 10:01
jnthn lizmat: Cool :) 10:07
Hope NLPW is being a nice time so far 10:08
RabidGravy nice weather for it 10:09
jnthn hah, not here :P
lizmat being at the registration desk, I haven't seen a single presentation yet
jnthn lizmat: aww
lizmat but people seem happy and having fun...
RabidGravy sunny and quite mild in London
lizmat the Hallway++ track is also quite busy with some ad-hoc hacking going on (not perl 6 though :-( )
RabidGravy: same here, but tomorrow and Sunday are going to be the first 20C+ days here this year, so we're good :-) 10:10
[Tux] cron job wil now refresh timings and graphs every day at 09:09 Europe/Amsterdam 10:11
RabidGravy ooh, almost outside drinking weather
jnthn We had really nice sunny weather yesterday. Today is overcast and damp. 10:12
So, I have most of the rest of the day to work on Perl 6 things... :)
And like months worth of things folks want me to do with it :P 10:13
RabidGravy cool
timotimo github.com/search?q=This+should+ne...=%E2%9C%93 <3
lizmat github.com/search?utf8=✓&q=%22...rchresults 10:14
gives significantly fewer results, but still a lot :-)
timotimo ah
RabidGravy while there are some semi-awake people here, any feelings why a .wrap() might contribute to an SCRef getting leaked (and compilation failing as a result?)
lizmat RabidGravy: feels like an RT to me with the proper golfing :-) 10:15
timotimo shouldn't an SCRef actually be serializable? i mean, shouldn't it just add a proper dependency and things work out fine?
[Tux] refresh speed4.html: the timing log is now available to all 10:16
timotimo maybe we're missing a repossession barrier in some operation that's related to .wrap
RabidGravy the circumstances are quite convoluted, there's a "class trait" which adds a role that over-rides the compose to wrap the methods 10:19
it only fails for "explicit" methods and not attribute accessors if that helps 10:20
dalek p: 8bea1eb | jnthn++ | src/vm/jvm/ (3 files):
Implement nqp::setdebugtypename on JVM.
p: 5cb3a0b | jnthn++ | src/vm/moar/QAST/QASTOperationsMAST.nqp:
Move setdebugtypename op to be with object ops.
p: 24789b4 | jnthn++ | src/vm/jvm/stage0/ (10 files):
Update JVM stage0.
p: e400d72 | jnthn++ | src/vm/moar/stage0/ (11 files):
Update MoarVM stage0 bootstrap.

This means we'll be able to use the nqp::setdebugtypename op in NQP's MOP.
p: 77005fc | jnthn++ | src/how/NQP (5 files):
Scatter various nqp::setdebugtypename.
p: 2632192 | jnthn++ | src/ (16 files):
Merge branch 'use-setdebugtypename'
kudo/nom: 34942c8 | jnthn++ | tools/build/NQP_REVISION:
Bump NQP_REVISION for nqp::setdebugtypename.
kudo/nom: ae43baa | jnthn++ | src/Perl6/Metamodel/Naming.nqp:
Remove `#?if moar` that we no longer need.
jnthn sortiz: In rt.perl.org/Ticket/Display.html?id=127813 I assume you missed the $s off the parameter vars? :)
sortiz jnthn, indeed. I can reproduce it trivially if more info is needed. 11:35
jnthn sortiz: I suspect the is repr golfs away too? And it probably even golfs to a sub call? 11:45
11:50 lizmat joined
sortiz Umm, dunno, Need to do more testing. 11:50
timotimo jnthn: do you know something about it being unsafe to remove Want nodes if we find them in sink context in the optimizer? 11:57
sortiz jnthn, if I change that parameter to positional, works with and without the constraint, so is named related. 11:58
jnthn timotimo: Well yeah, anything can carry a flatten/named, including a Want node 11:59
timotimo oh!
jnthn timotimo: So can be that.
timotimo that sounds like a good hint
jnthn sortiz: Interesting...
sortiz: It's probably a code-gen bug
sortiz jnthn, Another fact: Only fail if the constraint is a native int type, Int works. 12:02
timotimo that'd make a big difference in the underlying code
jnthn sortiz: And only if it's unsigned, by any chance? 12:03
sortiz The type and operand change, but fails: "operand type 32 does not match register type 24" 12:05
But with [u]int64 works! 12:06
timotimo wait. if a Want node has nameds or flattening, it wouldn't ever be in $!void_context, would it? 12:11
i was only going to build a "if the void context is known, replace the want node with its third child"
jnthn timotimo: Ah, true 12:13
timotimo i might start work on a star release
(out of nowhere)
sortiz jnthn, not method related, same problem calling as a sub.
jnthn OK, so that's another golf too
sortiz m: use NativeCall; sub Foo(uint32 :$bar) { say $bar }; Foo(:bar(4)); 12:21
camelia rakudo-moar ae43ba: OUTPUT«Bytecode validation error at offset 40, instruction 6:␤operand type 160 does not match register type 152␤ in block <unit> at /tmp/MJZ3yGoxGX line 1␤␤»
dalek p: 52fb397 | jnthn++ | t/nqp/19-file-ops.t:
Fix broken skip count.
jnthn heh, spelling switch wrong is a good way to confuse a C compiler, and yourself until you spot it... :) 12:23
sortiz m: sub Foo(uint32 :$bar) { say $bar }; Foo(:bar(4)); # btw don't need NativeCall.
camelia rakudo-moar ae43ba: OUTPUT«Bytecode validation error at offset 40, instruction 6:␤operand type 160 does not match register type 152␤ in block <unit> at /tmp/5nZ_TgBdbq line 1␤␤»
sortiz jnthn, minimal golf ^^^ 12:24
jnthn sortiz: Thanks; please drop it into the RT :) 12:29
sortiz jnthn, done. Thanks. 12:31
[Coke] (april fool's day) No. No we are not. 12:52
someone in -toolchain wants the public key used to sign the tarballs. 12:55
Where should we put such things? 12:56
timotimo um, is it wise to hand around a single key, rather than having a chain of trust instead?
[Coke] timotimo: It's my key. I'm signing the releases.
timotimo we probably also sign rakudo-star releases?
how long are you going to be around today? :) 12:57
[Coke] I assume if someone else ever does a release and they have a key, they'll sign it on their own.
I just got into work.
timotimo Multiple exceptions were thrown by LEAVE/POST phasers 12:58
Attempt to read past end of string heap when locating string
well, this is pretty fantastic %)
jnthn timotimo: Corrupt bytecode file of some kind? 13:00
timotimo i wonder how
jnthn Dunno
timotimo happened close to the end of install-modules 13:01
could be because i've asked it to install Shell::Command again, after it was already in there from panda installing it 13:03
13:05 skids joined
timotimo i ran make modules-install a second time and it worked 13:08
13:21 lizmat joined
[Coke] RT: 1301; PERF: 13 13:30
mentioned in perlchain, but repeating here: rakudo.org is very broken. 13:43
we need SSL working so we can host the keys we use to sign the tarballs. 13:44
14:04 lizmat joined
dalek p: 1e4a73a | (Pawel Murias)++ | src/vm/js/ (3 files):
[js] Make nqp-js-on-js handle native call stuff.
p: 3598599 | (Pawel Murias)++ | src/vm/js/nqp-runtime/reprs.js:
[js] Support using P6num and P6str as REPRs in nqp-js-on-js.
p: 66da2a2 | (Pawel Murias)++ | docs/qast.markdown:
Document typevar.
p: 9959496 | (Pawel Murias)++ | README.pod:
[js] Document how to build nqp-js-on-js.
16:39 geekosaur joined
dalek p: 512c9a1 | jnthn++ | tools/build/MOAR_REVISION:
Bump MOAR_REVISION for fixes, improvements.
kudo/nom: ecb7f70 | jnthn++ | src/vm/moar/ops/perl6_ops.c:
use_capture can allocate; add MVMROOT.

This fixes a crash/memory corruption that was most likely to show up in any hot-path code that did a lot of deferral (nextsame et al).
kudo/nom: 507acf1 | jnthn++ | src/vm/moar/ops/perl6_ops.c:
Avoid an allocation on an error path.

Doesn't fix anything, just a bit neater.
kudo/nom: a87fb43 | jnthn++ | tools/build/NQP_REVISION:
Bump to get lastest MoarVM, NQP fixes.
ast: be1b7fe | jnthn++ | S32-str/utf8-c8.t:
Test for RT #126756.
ast: 8516c62 | jnthn++ | integration/weird-errors.t:
Test for RT #127748.
jnthn There we go, two SEGVs down. 17:36
[Coke] \o/ 17:47
RabidGravy Harr! 20:49
21:42 AlexDaniel joined 21:52 perlpilot joined 22:03 lizmat joined
dalek p: 1168dac | (Stéphane Payrard)++ | README.pod:
supress section about nqp-rx. dead since 6 years

Maybe the same should be done for Parrot? Not sure.
p: ed2d728 | lizmat++ | README.pod:
Merge pull request #278 from cognominal/patch-1

supress section about nqp-rx. dead since 6 years