github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
01:13 Altai-man_ joined 01:15 sena_kun left 01:36 guifa2 left 01:37 guifa2 joined 02:36 lucasb left 03:36 quotable6 left, nativecallable6 left, coverable6 left, tellable6 left, shareable6 left, linkable6 left, bisectable6 left, releasable6 left, greppable6 left, benchable6 left, unicodable6 left, evalable6 left, bloatable6 left, statisfiable6 left, reportable6 left, committable6 left, sourceable6 left, notable6 left, squashable6 left, evalable6 joined, benchable6 joined, unicodable6 joined 03:37 bisectable6 joined, statisfiable6 joined, greppable6 joined, squashable6 joined, sourceable6 joined, reportable6 joined 03:38 bloatable6 joined, quotable6 joined, committable6 joined, nativecallable6 joined, coverable6 joined, linkable6 joined, tellable6 joined, notable6 joined, releasable6 joined 03:39 shareable6 joined 04:14 sena_kun joined 04:16 Altai-man_ left 04:33 Kaeipi left, Kaeipi joined 06:09 Kaeipi left 06:10 Kaiepi joined 06:14 Altai-man_ joined 06:16 sena_kun left 07:07 leont joined 07:23 patrickb joined 07:33 zakharyas joined 08:00 zakharyas left, zakharyas joined
nwc10 good *, #moarvm 08:36
[Coke]: you pasted some test failres in gist.github.com/coke/cc744eade897a...14b93495ec
I can't replicate those. What platform is that, and does master suffer the same failures?
Geth_ MoarVM/YYY-A-Better-Hash: 53 commits pushed by (Nicholas Clark)++
review: github.com/MoarVM/MoarVM/compare/6...241f5c9764
08:40
jnthn good moarning o/ 08:41
tellable6 hey jnthn, you have a message: gist.github.com/cae0675546cb02a855...9ae1984d81
nwc10 hi.
(Almost) ships passing in the night
in that, we're heading out to a far away place with poor mobile coverage at lunchtime 08:42
[no, not Germany :-)]
so, you were about to get a pull request (For the purpose of tracking the problems)
the sodbag works. And it's faster. But/until
I was paranoid, and we've exposed an existing data race (described somewhere in scrollback)
and this morning I tried just eliminating MVM_CF_HAS_OBJECT_ID 08:43
and it's about a 10% slowdown
*some* of that can be won back with the last known not-yet-done optimisation for the hash implementation. But not 10%
so, bother, seems to be "Atomic ops" needed. At least for write barrierers and object IDs 08:44
or am I cofnused and you're still on holiday, but temporarily in some "non holiday" Monad
lizmat nwc10: it's my understanding that jnthn is back from holiday, but also at $work now 08:59
nwc10 I hope that there aren't too many coffee slicks to clear up there. 09:00
Geth_ MoarVM: nwc10++ created pull request #1334:
YYY a better hash
09:01
nwc10 nine: could you check the last paragraph of that draft pull request?
I need to go AFK and mow the lawn. Else we will return to a jungle. 09:02
09:14 sena_kun joined 09:16 Altai-man_ left
jnthn Yes, I'm back, and catching up on all the things :) 09:58
What does MVM_CF_HAS_OBJECT_ID race with, ooc? 09:59
oh, right, I should find the scrollback :)
nwc10 scrollback link in that PR 10:23
and of anyone here, I think nine understands it best
and I'm mostly not here
and soon (until Monday) very not here
jnthn Long weekend vacation? :) 10:25
nwc10 roughly this. Although I'm not sure how relaxing it will be 11:04
11:14 Altai-man_ joined 11:16 sena_kun left 11:30 zakharyas left
[Coke] nwc10: those results were from my mac. 12:48
... and I see there was just another push, so I can kick off another mac run.
updating to latest rakudo/nqp/moarvm-your-branch and rebuilding. 12:49
12:50 zakharyas joined
[Coke] btw, do we can about const char warnings? 12:50
I remember back in the day, I think it was Andy Lester purging them from parrot 12:51
nwc10: same errors. Let me try against a non-nwc build and see if it's you or not. 13:20
s/same/similar, at least/
13:42 MasterDuke joined
[Coke] nwc10: unfortunately, the failures occur only with your branch. 13:47
nwc10: gist.github.com/coke/353adbf0ab607...b4847e312f - though I see the mainline dev is using a detached head on MoarVM. let me know if I need to force that to be updated. 13:51
14:00 lucasb joined 14:05 Kaiepi left, Kaeipi joined 14:10 squashable6 left 14:11 Kaiepi joined, squashable6 joined 14:12 Merfont joined 14:13 Kaiepi left 14:14 sena_kun joined 14:15 Merfont left, Kaeipi left 14:17 Altai-man_ left
MasterDuke nwc10: everything builds and tests ok on YYY-A-Better-Hash 14:45
14:53 Kaiepi joined
[Coke] ... wait, the branch name changed? 14:54
14:55 Kaiepi left
[Coke] retesting on yyy 15:01
15:13 Kaiepi joined 15:29 domidumont joined
timotimo it's easy to see how removing one line from X gives you Y, but when you remove another, what do you get? I? |? V? 16:39
[Coke] / 17:06
you just need the right font. 17:07
ugh. --force-rebuild (build everything anyway as is) is still (correctly) overridden by specifying a branch on the config invocation, so i've tested XXX branch 3 times now instead of the YYY branch. trying one last time... 17:08
17:14 Altai-man_ joined 17:16 sena_kun left 17:32 zakharyas left
[Coke] ok. finally can say "same issues on the YYY branch as on the XXX branch." 17:35
17:45 domidumont left 19:09 zakharyas joined 19:14 sena_kun joined 19:16 Altai-man_ left 20:50 zakharyas left 21:14 Altai-man_ joined 21:16 sena_kun left 21:24 AlexDaniel joined, AlexDaniel left, AlexDaniel joined 23:04 patrickb left 23:15 sena_kun joined 23:17 Altai-man_ left 23:31 vrurg left 23:35 vrurg joined