02:49 FROGGS__ joined 03:02 colomon joined 04:21 colomon joined 05:00 colomon joined 05:17 colomon joined 06:56 colomon joined
[Tux] test 21.004 09:24
test-t 13.355
csv-parser 48.868
09:44 RabidGravy joined
[Tux] cpupower frequency-set -g performance 09:46
test 20.612
test-t 13.008
csv-parser 49.729
[Tux] realized I rebooted after a power failure recently 09:47
masak why does QAST::VM have the three pieces [jvm], [moar], and [loadlibs] 11:04
one of those is not like the others
:)
lizmat good *, #perl6! 12:16
did I miss a memo about the release being delayed?
dalek ast: b7507b8 | lizmat++ | S09-typed-arrays/native-int.t:
Add tests for int @a.append|prepend
12:49
ast/6.c-errata: 266882d | lizmat++ | S09-typed-arrays/native-int.t:
Make two more tests for failure less specific

This is currently a WIP across releases, the semantics of the check doesn't change: it's an error, and it stays an error :)
12:57
ast: 5541d6e | lizmat++ | S09-typed-arrays/native-num.t:
Add num @a.append|prepend tests
13:16
lizmat Newbie code segfaults: RT #127748 13:45
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=127748
dalek ast/6.c-errata: ea25fa5 | lizmat++ | S06-signature/introspection.t:
Sigs with "where { ... }" in them can never eqv

Previously, .perl output was taken for eqv, which was both wrong and very CPU intensive for different signatures. This was since then fixed in 2016.03, so tests asking for equivalence are wrong here.
14:07
ast/6.c-errata: 960fde0 | lizmat++ | integration/advent2009-day20.t:
Untodo passing test
14:12
kudo/nom: 1ad9175 | coke++ | docs/announce/2016.03.md:
Add first pass at release announcement
14:49
[Coke] anyone have any last minute updates to the ChangeLog? 15:02
16:03 Begi joined 16:04 stmuk joined, awwaiid joined 16:09 Begi left
psch just to have this here, in case someone wants to chime in later: 16:09
r-j currently doesn't build, because of commit 93d597f40
the problem comes during serialization of the BOOTStrArray type object
nqp-m: say(nqp::elems(nqp::what(nqp::list_s()))) 16:10
camelia nqp-moarvm: OUTPUT«-4294967287␤»
psch $ ./nqp-j -e'say(nqp::elems(nqp::what(nqp::list_s)))'
org.perl6.nqp.sixmodel.TypeObject$AggregateRTE: Cannot do aggregate operation on a type object
the exception type is just a debugging aid 16:11
16:11 colomon joined
psch i kind of want to say that moar is not taking this serious enough, but i do recognize that putting concreteness checks or similar there makes it (probably noticably) slower 16:11
afaict right now it's just following a pointer 16:12
16:19 skids joined
psch so, the obvious solution would be having a sixmodel.TypeObject return a long instead of dieInternal 16:20
while the more involved solution involved adapting the serialization code for VMArray to take TypeObjects into account
...while nqp-m keeps cheating :S
skids m: [Coke]: WRT Changelog ".reverse on ranges returns Seq now, not List" 16:23
camelia rakudo-moar 1ad917: OUTPUT«===SORRY!=== Error while compiling /tmp/TFTyXP60fS␤Confused␤at /tmp/TFTyXP60fS:1␤------> [Coke]:⏏ WRT Changelog ".reverse on ranges retur␤ expecting any of:␤ colon pair␤»
skids ergh reflex m:
dalek kudo/nom: 6ef16ca | coke++ | src/core/List.pm:
Revert "Don't bother creating native str array if possible"

This reverts commit 93d597f4075cfc13aa925a662bc3ee6e4986bda0.
12:29 < liztormato> [Coke]: please revert the commit that is causing the jvm buid failure
16:35
kudo/nom: 468d431 | coke++ | docs/announce/2016.03.md:
Pull in some changelog
16:38
kudo/nom: f6c45b8 | coke++ | tools/build/NQP_REVISION:
[release] bump NQP revision
kudo/nom: 5eaa15d | coke++ | VERSION:
[release] bump VERSION
p: d2e5dad | coke++ | tools/build/MOAR_REVISION:
bump moar revision
16:40
p: d6fadbd | coke++ | VERSION:
bump version to 2016.03
[Coke] make test failure on nom. 16:45
Test Summary Report
-------------------
t/04-nativecall/13-union.t (Wstat: 65280 Tests: 26 Failed: 0)
it's actually the use v6 causing the failure here. 16:47
[6~[6~
Doesn't appear to be impacting any of the other files, though. Just that one. 16:48
My plan will be to remove the use there. I have to run out to the store; if someone has a better idea before I get back, great. :) 16:49
16:58 AlexDaniel joined 17:47 Ben_Goldberg joined
[Coke] oh. no, it just fails *sometimes*. :P 18:05
yah, I can get it to fail every few times running make test. 18:12
18:22 geekosaur joined
lizmat is back 20:23
timotimo with spesh_disable set i get 10 out of 10 non-crashes 20:34
lizmat just double checked it: still fails with it: 20:36
$ MVM_SPESH_DISABLE=foo perl6 2
Segmentation fault: 11
timotimo 6b85e65d08ce8b5d59027dc2ed799f68f1b25d38 segv.pl6
^- sha1sum
to be sure
lizmat $ perl6 --optimize=2 2 20:37
360360
fails with --optimize=0 and --optimize=1
anything above 2 it works
what's the default again ?
timotimo i think it's 3 for the core setting and 2 otherwise? 20:38
aye, 2 is the default 20:39
anyway. i know where to put the check for a null value so we can get a panic instead of a segfault
that's not a good improvement
i'll be AFK for a bit to prepare dinner and such, after that i'll have a look at the generated bytecode and why that MVMContext's body ends up containing a null pointer 20:40
lizmat argh: I had removed the () around True
timotimo hehe
lizmat so, with that put back, it segfaults on 3 also
and on 2, but not on 0 and 1 20:41
timotimo how about spesh disable? 20:42
lizmat still segfaults 20:43
This is my SHA1: $ 6 'use nqp; say nqp::sha1("2".IO.slurp)'
38C3A8A7984CAD0DE5868D91B5716255EA8013FA
fwiw, I can't get t/04-nativecall/13-union.t to fail with "prove -v -e 'perl6 --ll-exception' t/04-nativecall/13-union.t" 21:02
it fails 3/4 times using make 21:03
dalek kudo/nom: 1bbef98 | lizmat++ | t/harness:
We don't need "lib" and "." anymore

As a side effect, this also seems to fix the spurious errors as discussed in RT #127750. Although technically, this is a workaround, having always a "use lib lib, ." for each test was a fossil from older days when we still had @*INC (aka pre-CURLI)
21:23
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=127750
ast: c8ee946 | lizmat++ | S11-modules/require.t:
Fix only test file that broke because of 1bbef98
21:24
lizmat so I think we can safely say there are still some dragons in the precomp department :-( 21:25
timotimo ;( 21:27
lizmat hmmm... looks like my last change had a significant effect on spectest runtime 21:38
Files=1106, Tests=51998, 265 wallclock secs (13.73 usr 4.18 sys + 1653.57 cusr 146.87 csys = 1818.35 CPU)
timotimo is that good? 21:39
lizmat from a few days ago, with almost 300 fewer tests: 21:40
Files=1106, Tests=51726, 280 wallclock secs (14.32 usr 4.32 sys + 1654.18 cusr 158.33 csys = 1831.15 CPU)
timotimo *nice* 21:41
22:03 BenGoldberg_ joined
[Coke] ok. I have to now 'make install' before make test or lots of failures. 23:54
but even if I do that, that test from before still fails. 23:55
timotimo huh, it does? 23:59
but you've already Configure.pl'd?