Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2013-12-11

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

All times shown according to UTC.

Time Nick Message
00:21 * [Coke] thinks now is a great time to fix moar!
00:21 timotimo isn't it always? :)
00:24 [Coke] I have 2 hours to kill and a laptop and wifi
00:24 * [Coke] tries to see if he can track down this nqp-m segfault
00:25 [Coke] looks like moar builds by default with -g - is there a way to get more debug symbols?
00:28 [Coke] https://gist.github.com/coke/7903006
00:33 [Coke] minor bugnit, can't use a relative path when configuring nqp-m --prefix
00:39 [Coke] https://gist.github.com/coke/7903139 - actual line numbers and stuff. whee.
00:39 TimToady yes, I discovered this a week or two ago
00:40 TimToady it's worse when configuring from the rakudo level
00:40 [Coke] is there an open bug? my bad.
00:40 [Coke] oh, the relative path thing
00:40 [Coke] the bug is reporting from             OP(isinvokable): {
00:42 [Coke] from inside that block, but that might be macro'd too much for me just now.
00:42 TimToady is this while compiling the rakudo setting?
00:43 [Coke] updated https://gist.github.com/coke/7903139 with the func args.
00:43 [Coke] TimToady: this is a segfault building nqp-m
00:43 [Coke] (current stopper on the daily roast machine)
00:43 TimToady are you in master?
00:44 TimToady and are you using MoarVM master?
00:45 * TimToady configures his rakudo with: perl Configure.pl --backends='parrot,jvm,moar' --prefix='/home/larry/nom/install' --gen-parrot --gen-nqp=master --gen-moar=master
00:45 TimToady undetected version skew can cause problems in nqp-m
00:46 [Coke] I'm just trying to build nqp, not rakudo.
00:46 TimToady on what arch?
00:46 [Coke] I don't specify, so I would expect to be getting master on moar.
00:46 [Coke] whatever host06 is... checking...
00:46 [Coke] Linux host06 3.5.0-17-generic #28-Ubuntu SMP Tue Oct 9 19:31:23 UTC 2012 x86_64 x86_64 x86_64 GNU/Linux
00:46 TimToady specifying master explicitly has seemed to matter sometimes to me
00:47 TimToady dunno why
00:47 [Coke] what branch of rakudo are you on?
00:48 cognominal joined #moarvm
00:49 * [Coke] wipes ./nqp and tries again with TimToady++'s config (minus parrot,jvm)
00:50 [Coke] TimToady++ # cargo culting done right.
00:51 [Coke] I wonder if we need to bump the required version in both places.
00:53 TimToady nom
00:53 [Coke] ah. that's probably my problem. I was on moar-support on rakudo.
00:53 TimToady oh, wait, so am I, duh
00:53 [Coke] bet that is old and has an old nqp ref.
00:54 * [Coke] ah, whew, I had one thing right.
00:54 * [Coke] tries to build rakudo-m
00:54 TimToady but install/ gets confused when you switch branches, so you generally have to blow it away
00:54 TimToady the various configures and the various makes are not interacting well with git
00:55 [Coke] I start fresh every day, so I have a slight advantage.
00:55 TimToady on mine I get up to where rakudo wants to compile m-BOOTSTRAP, and get a heap corrpution
00:56 TimToady on Windows® people seem to get as far as compiling the setting
00:56 [Coke] Heap corruption detected: pointer 0x2aba5552acf8 to past fromspace
00:56 [Coke] so now I'm caught up to now.
00:57 TimToady yeah, please fix that one :D
00:57 TimToady which both jnthn++ and diakopter++ seem to run away from screaming...
00:58 TimToady (they get karma for being entertaining, y'see)
01:08 diakopter :) it's not that I run from that screaming, it's that I run toward everything else screaming
01:09 TimToady berserkers...
01:18 japhb_ joined #moarvm
01:27 [Coke] diakopter: can we get valgrind on host06?
01:31 TimToady okay, I'm not in the sudoers list, and this incident will be reported.  :)
01:37 diakopter REPORTED
01:39 diakopter [Coke]: done
01:59 [Coke] diakopter++
02:02 [Coke] I am going to wish I captured the output of that command...
02:06 [Coke] valgrind sez: https://gist.github.com/coke/7904039
02:08 * [Coke] reruns with -v and --leak-check=full, but is unsure if valgrind noticed the heapcorruption
02:28 benabik joined #moarvm
02:32 [Coke] here's a more verbose run: http://feather.perl6.nl/~coke/barf.out
02:37 [Coke] I wonder how many of those losses are due to the crash from the heap corruption
02:37 TimToady fix the crash and we'll know :P
02:40 TimToady but I'd guess your guess is right
02:48 [Coke] so, I can't gdb it (no backtrace), and valgrind isn't helpful. I think this is the point where I run away screaming.
02:48 diakopter [Coke]: well if the global destruct phase doesn't run, nothing get freed
02:49 [Coke] I wonder if I need to do a gdb run and just step through the whole thing until it splodes.
02:49 diakopter it won't help track down the problem
02:49 TimToady trouble is that something is leaving a timebomb
02:50 TimToady incorrectly rooted or some such
02:51 TimToady or somebody is assuming something doesn't move when it does
02:56 [Coke] diakopter: pointers on how to solve it most welcome. I was just hoping to find the place where it exploded.
02:58 * [Coke] packs up for the commute home.
07:07 FROGGS joined #moarvm
07:11 FROGGS [Coke]: valgrind is not helpful because all happens in our allocated area, so there is no suspicious invalid read or so...
07:11 FROGGS [Coke]: but this was interesting: https://gist.github.com/FROGGS/8efa32fce9e288e9f7f0#file-past_fromspace-bash-L25
07:11 FROGGS [Coke]: the exploding thing is an MSMStaticFrame
09:16 odc joined #moarvm
10:14 nwc10 is it worth creating a way to make the GC very agressive, and (at least) do a nursery sweep at every point it might do one
10:14 nwc10 to rapidly show up unrooted thingies
10:14 nwc10 yes, I realise that this will be slow
10:15 nwc10 but it's probably faster than how a developer currently finds porblems
10:15 nwc10 or even just valgrind
10:15 FROGGS it might help insofar you would hit the wrongish thing earlier
10:15 FROGGS at least that is what I'd hope
10:28 diakopter nwc10: sure, just decrease the nursery size pathologically
10:42 tgt joined #moarvm
12:33 tgt joined #moarvm
15:47 eternaleye joined #moarvm
16:01 jnthn evening o/
16:01 benabik joined #moarvm
16:01 nwc10 good UGT, jnthn
16:03 jnthn Mostly, hunting the GC bug from my side blocks on a chunk of undistracted time I can dedicate to it. :)
16:09 * jnthn managed to attend talks by the creators of Erlang and Eiffel this week :)
16:16 timotimo cool :)
16:17 timotimo something good to steal for perl6 or moarvm perhaps?
16:18 jnthn Well, the Erlang one had some good stuff on fault tolerance.
16:20 timotimo that's always cool
16:20 timotimo perl6 could learn a bit of marketing from Outlaw Techno Psychobitch.
16:29 FROGGS joined #moarvm
16:42 jnap joined #moarvm
18:22 jnap joined #moarvm
18:47 jnap1 joined #moarvm
19:15 ssutch joined #moarvm
19:16 ssutch joined #moarvm
19:30 ssutch joined #moarvm
20:45 benabik joined #moarvm
21:00 jnap joined #moarvm
21:27 jnap joined #moarvm
22:32 [Coke] lizmat: I was going to say you shouldn't enable me, but my boss decalred he's taking the team out for drinks and dinner this evening, which makes a date switch much more attractive.
22:32 [Coke] ww
22:53 jnap joined #moarvm
22:54 ssutch joined #moarvm
23:35 ssutch joined #moarvm
23:47 jnap joined #moarvm
23:55 benabik joined #moarvm

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