Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2017-10-13

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:11 Geth_ joined #moarvm
00:13 Geth_ joined #moarvm
00:13 Geth_ joined #moarvm
00:14 Geth_ joined #moarvm
00:16 Geth_ joined #moarvm
01:56 ilbot3 joined #moarvm
01:56 Topic for #moarvm is now https://github.com/moarvm/moarvm | IRC logs at  http://irclog.perlgeek.de/moarvm/today
04:31 ugexe should calls to getenv() be using wgetenv() on windows in instances when getenv() is expected to contain a path?
04:34 ugexe if so note that there is a uv_os_getenv that can do this
06:00 lizmat joined #moarvm
06:02 domidumont joined #moarvm
06:18 rba_ joined #moarvm
06:31 domidumont joined #moarvm
06:40 lizmat joined #moarvm
06:58 Ven joined #moarvm
07:40 rba joined #moarvm
08:12 brrt joined #moarvm
08:12 zakharyas joined #moarvm
08:24 lizmat joined #moarvm
08:33 KDr2 joined #moarvm
08:54 lizmat joined #moarvm
08:56 jnthn ugexe: I dunno the ins and outs of that, but I think environment handling on Windows was patched a while back to make it do the right thing with regard to encodings and so on (think that's related to why we declare a wmain instead of main function on Windows also).
09:34 Ven joined #moarvm
09:44 lizmat joined #moarvm
09:50 ZofBot joined #moarvm
10:14 lizmat joined #moarvm
10:31 lizmat joined #moarvm
11:15 lizmat joined #moarvm
11:16 MasterDuke `MVMuint64 arg_idx   = arg_match & MVM_MULTICACHE_ARG_IDX_FILTER;` seems like it might be the problem when i try changing MVM_INTERN_ARITY_LIMIT to 16
11:16 MasterDuke MVM_MULTICACHE_ARG_IDX_FILTER is `(2 * MVM_INTERN_ARITY_LIMIT - 1)`
11:17 lizmat_ joined #moarvm
11:17 MasterDuke and when i get segfaults, arg_idx = 16
11:18 MasterDuke the next line is ` MVMRegister  arg       = args[arg_idx];`, and then it dies on then next line `MVMSTable   *st        = STABLE(arg.o);`
11:19 lizmat__ joined #moarvm
11:24 lizmat joined #moarvm
11:46 Ven joined #moarvm
12:00 ZofBot joined #moarvm
12:06 samcv MasterDuke, can you check the current PR for join?
12:06 buggable joined #moarvm
12:06 samcv i updated it yesterday or so
12:07 MasterDuke samcv: will do
12:09 buggable joined #moarvm
12:10 buggable joined #moarvm
12:13 [equa] joined #moarvm
12:25 cog_ joined #moarvm
12:29 MasterDuke samcv: i find it easier to follow now
12:30 zakharyas joined #moarvm
12:31 [equa] left #moarvm
12:33 lizmat joined #moarvm
13:27 Ven joined #moarvm
13:28 Ven_ joined #moarvm
13:32 samcv nice :)
13:39 brrt joined #moarvm
13:51 MasterDuke what's the website for those moarvm daily benchmarks?
13:57 Ven joined #moarvm
13:57 MasterDuke .ask jnthn think https://github.com/MoarVM/MoarVM/pull/689 is good now?
13:57 yoleaux MasterDuke: I'll pass your message to jnthn.
14:33 Ven_ joined #moarvm
14:39 Ven_ joined #moarvm
14:50 Geth_ ¦ moarvm.org: MasterDuke17++ created pull request #8: Remove period
14:50 Geth_ ¦ moarvm.org: review: https://github.com/MoarVM/moarvm.org/pull/8
16:10 ZofBot joined #moarvm
16:19 ZofBot joined #moarvm
16:47 camelia joined #moarvm
17:20 MasterDuke samcv: have you thought about implementing nqp::eqat(ic|im|icim) for the jvm?
17:20 samcv well jvm doesn't support NFG
17:21 MasterDuke hm. what about js?
17:21 samcv java does support normalization though. apparently
17:23 samcv not sure if it supports Graphpme_Cluster_Break though
17:23 MasterDuke in INTERPOLATE there is an `#?if moar <do eqat* stuff> #?endif #!if !moar <do ordbaseat stuff> #?endif`
17:23 samcv icu does
17:23 MasterDuke but ordbaseat is NYI for jvm, so neither works for it
17:25 MasterDuke might make sense to at least add NYI eqat* ops for jvm?
17:26 samcv MasterDuke, yeah
17:26 MasterDuke oh, and ordbaseat is also NYI for js
17:27 MasterDuke guess i could just change that `$!if !moar` branch to die, since it would just throw NYI exceptions anyway
17:28 samcv i would leave it until it's implemented
17:29 MasterDuke well, couldn't equat* for jvm/js be implemented in terms of ordbaseat once it's implemented?
17:30 domidumont joined #moarvm
17:31 samcv MasterDuke, tell me which file you're looking at so i can go to it and look
17:33 MasterDuke https://github.com/rakudo/rakudo/blob/nom/src/core/Match.pm#L337
17:34 samcv ok yeah leave that code
17:36 MasterDuke why not just create eqat* for js/jvm that throw NYI like the ordbaseat do now? then i can remove those if/else and the behavior will be exactly the same as it is now
17:38 MasterDuke the #!if !moar code that exists doesn't work for jvm/js anyway, it's just so the compilers don't complain about unknown ops
17:38 samcv i don't want to have to go back and search for that code. let alone remember
17:39 MasterDuke ?
17:39 samcv and if i make nfg a thing on jvm i'll add ordbaseat before adding those ops
17:40 samcv if you delete it. and i implement ordbaseat on jvm. then i'll have to remember to go into the history and find it. which i may not remember, or someone else may not know
17:40 samcv unless you want to comment it out and add a die in there
17:42 samcv but i'd prefer it just be left alone. maybe with a comment above it to say they don't have ordbaseat yet. i will add a comment to it
17:42 MasterDuke i'll add it, i'malready working on a PR for that code
17:48 samcv ok cool
18:10 AlexDaniel- joined #moarvm
18:10 MasterDuke hm. or implement eqat* in terms of ordbaseat for jvm/js right now. then both will work when ordbaseat is actually implemented and we can remove the if/else
22:32 MasterDuke timotimo: did you see my comments about what happened when i increased MVM_INTERN_ARITY_LIMIT? any suggestions?
22:33 timotimo wasn't able to look into it
22:35 MasterDuke i wasn't able to fix it, so i'm going to PR my INTERPOLATE change as is. if people think my workaround isn't good then i'll look into it more
22:36 timotimo i wonder if i should (just?) merge the FSA valgrind thing
22:37 MasterDuke do it
22:45 jnthn Yowser... https://travis-ci.org/croservices/cro-http/jobs/287751428#L1423
22:45 yoleaux 13:57Z <MasterDuke> jnthn: think https://github.com/MoarVM/MoarVM/pull/689 is good now?
22:45 yoleaux 17:22Z <jdv79> jnthn: I just RT'd the one from a week ago - RT#132287 - not sure how to debug that.  never had to debug something threaded like that before.
22:45 yoleaux 19:14Z <jdv79> jnthn: - looks like the scheduler might have enough worker threads... - https://gist.github.com/anonymous/49f84d38087bf12d70eeeda95696a29b
22:45 synopsebot RT#132287 [open]: https://rt.perl.org/Ticket/Display.html?id=132287 [REGRESSION][CONC] stall/block in async heavy code
22:46 timotimo whoopsie-daysies
22:47 jnthn oh wait, it's actually on the 2017.08 Rakudo
22:47 jnthn Not on HEAD
22:47 jnthn So false-ish alarm
22:48 timotimo oh phew
22:48 timotimo latest is just starting
22:55 jnthn I wonder a tad if somehow there's insufficient affinity workers leading to a deadlock in the one above
22:56 timotimo jdv's? it's possible
22:56 timotimo telemeh outputs which thread acquires which low-level lock; that could maybe shed some light, unless it's using something higher-level locking thing
22:56 jnthn I did consider having the supervisor steal from blocked affinity queues and shove the work into the general queue to try and break any such deadlocks
22:56 MasterDuke doubt it's related, but i just had to hard-reset my desktop because a spectest hung it
22:57 weabot joined #moarvm
22:57 timotimo that doesn't sond so bad
22:57 timotimo MasterDuke: perhaps ram exhaustion?
22:57 jnthn I couldn't actually create a situation that needed it, though :)
22:57 jnthn But maybe now we have one
22:57 MasterDuke timotimo: doubt it, only had one or two other things open
22:57 jnthn Or it could be something else entirely.
23:00 jnthn Palackého
23:00 timotimo annoyingly it's not easy to figure out afterwards
23:00 jnthn d'oh
23:00 timotimo oops, root password leaked
23:00 jnthn haha
23:01 jnthn Just a street name
23:01 jnthn That isn't actually where I expected it to be
23:01 timotimo that cro job came up green on latest
23:01 timotimo oh, already 4 minutes since it finished
23:01 jnthn ah, nice
23:07 timotimo i wonder what kind of tooling you could get to cross the gap between a cro app and whatever js is running on the other side
23:07 timotimo i really need to look for a set of snippets or something for react/redux %)
23:07 ggoebel joined #moarvm
23:12 weabot joined #moarvm
23:15 rba joined #moarvm
23:16 timotimo yes, action type and action creator snippets, that's really useful
23:31 nine joined #moarvm
23:32 TimToady joined #moarvm
23:53 timotimo joined #moarvm
23:56 jnthn sleep &
23:57 timotimo gnite jnthn

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary