Camelia, the Perl 6 bug

IRC log for #moarvm, 2013-10-14

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

All times shown according to UTC.

Time Nick Message
00:15 diakopter eh; better attend to more pressing todos
00:44 diakopter lue: I get the same segfault
00:45 diakopter 'cept it was built using asan, but .. it's not spitting out the debug asan output
00:45 diakopter oddly
00:46 lue I'll try with GCC (like I've done before) and see if things improve.
00:48 lee_ i'm back to getting: '65' is not a valid number at line 2132, near ",\n    66,\n"
00:48 lee_ building nqp
00:49 lue nope. Still dies.
00:55 diakopter with gcc??
00:55 diakopter it's probably not actually gcc; it's probably clang pretending to be gcc
00:58 BenGoldberg joined #moarvm
01:09 woosley joined #moarvm
01:17 * [Coke] tries on os x.
01:22 FROGGS_ joined #moarvm
01:37 lue diakopter: compiling MoarVM with GCC gives me warnings that are not filled with pretty colors. I'm sure it's gcc :)
02:46 diakopter ok
02:47 diakopter TimToady: akive?
02:47 diakopter alive
03:20 diakopter lue: but clang gives warnings in colors..
03:21 diakopter magenta "warning:" text and green pointer arrow line
03:21 lue I know. I'm saying, since my gcc run didn't have pretty colors, it was probably actually gcc :) [first time round was the superior colored clang, in fact]
03:22 diakopter oh.
03:22 diakopter oh, I missed the "not" there; sorry :)
03:22 diakopter even though I "read" it like 3-4 times :D
03:22 lue Kind of a critical word to miss in any sentence :D
03:23 lue I'm glad though it's not compiler-specific (considering my LLVM is only ever the latest pull from the repos, it's a possibility).
03:24 diakopter mine isn't; mine's clang version 3.3 (tags/RELEASE_33/final)
03:24 diakopter Target: x86_64-apple-darwin12.5.0
03:24 diakopter hmm, I'll try clang on linux
03:25 lue (It seems the problem comes from MoarVM, and not the compiler used itself, is what I was trying to get across with the previous statement)
03:26 lue My clang says its 3.4, although it would have a revision attached to it had I enabled such in cmake.
04:45 JimmyZ zhuomingliang
04:45 JimmyZ oh
05:07 benabik joined #moarvm
06:10 FROGGS joined #moarvm
06:25 FROGGS next things we miss: p6trialbind and p6bool
06:29 JimmyZ it needs ext-ops
06:34 FROGGS and that means?
06:34 FROGGS what is 'ext-ops' ?
06:36 JimmyZ FROGGS: https://github.com/MoarVM/MoarV​M/blob/ext/docs/extops.markdown
06:38 FROGGS thank you
09:36 diakopter not_gerd has been working on it
09:36 diakopter in a branch
09:36 JimmyZ great
09:43 diakopter .tell jnthn that clang build problem on mac is identical on linux, so you can reproduce it on host07 if you like
09:43 yoleaux diakopter: I'll pass your message to jnthn.
09:44 jnthn diakopter: Which clang build problem? :)
09:44 yoleaux 09:43Z <diakopter> jnthn: that clang build problem on mac is identical on linux, so you can reproduce it on host07 if you like
09:47 diakopter the one lue reported in the backlog
09:47 diakopter er
09:47 jnthn k
09:47 diakopter actually
09:48 diakopter no
09:48 diakopter I hadn't updated nqp
09:49 diakopter stupid airport wifi gave me an ip address, but no access to dns
09:49 * diakopter goes to override to 8.8.8.8
09:51 diakopter jnthn: actually yeah it's doing that can't parse Ops thing
09:52 jnthn k
09:53 diakopter jnthn: (but that's not the error on mac
09:53 diakopter )
09:55 diakopter error on the mac can be seen in lue's nopaste
09:57 diakopter erm
09:58 diakopter ok never mind; I was still on gcorch branch :(
09:58 jnthn Oh...so it's more orchward to reproduce it in master?
09:59 diakopter o wait.. same in master.
09:59 diakopter Ops parse thing
09:59 diakopter not the error lue pasted
09:59 diakopter lue must've been on the gcorch branch..... o_O
10:00 diakopter nqp builds absurdly quickly on mac/clang though
10:00 diakopter (except for Ops.moarvm!)
10:00 diakopter (which builds impossibly quickly) ;)
10:29 FROGGS so, do I understand right that we would create a rakudo/src/vm/moar/ops/perl6_ops.c, add things like perl6settypes, compile it to a dll, and load/access it via upcoming ext-ops feature?
10:29 FROGGS jnthn: ^^
10:30 jnthn FROGGS: Soemthingng like
10:32 FROGGS cool
10:33 * FROGGS .oO( "Do we always have to call it "thing" Jim?" )
10:37 * jnthn finally gets his admin tasks done and starts pondering the two talks he needs to write...
10:49 grondilu joined #moarvm
11:38 colomon joined #moarvm
13:10 colomon_ joined #moarvm
13:10 arnsholt_ joined #moarvm
13:13 odc joined #moarvm
13:42 jnap joined #moarvm
14:43 benabik joined #moarvm
14:53 gshank joined #moarvm
14:54 diakopter not_gerd: u around?
15:20 camelia joined #moarvm
15:57 cognominal joined #moarvm
16:11 benabik joined #moarvm
16:23 FROGGS joined #moarvm
16:29 FROGGS so, what do we do today?
16:30 jnthn Write Git slides!
16:30 moritz take over the world!
16:30 FROGGS moritz's answer sounds better
16:31 jnthn Mine keeps me out of trouble :P
16:31 FROGGS gah, coward! :P
17:11 benabik joined #moarvm
17:12 TimToady well, each of you should take over half the world, then compare notes
17:12 diakopter ..
17:12 diakopter empty diff
17:13 TimToady you weren't supposed to take over the *same* half the world, you danglesockets!
17:15 * diakopter *just now* found the Alt-number and Alt-qwert shortcuts for switching to windows
17:15 diakopter in irssi
17:19 timotimo oh my
17:21 jnthn really?
17:24 timotimo you have not known about that? for how long? o_O
17:24 timotimo there's also a "cycle to next window with activity" shortcut
17:28 diakopter I was using Esc-number
17:28 diakopter so, slightly slower
17:29 * TimToady also has his page-left and page-right keys bound to "tab" left and right
17:29 TimToady (in firefox too)
17:30 timotimo that is always equivalent to alt-number, diakopter
17:31 timotimo it's not something irssi does, it's what terminals do
17:40 diakopter had no idea
17:41 timotimo :)
17:49 woosley joined #moarvm
18:08 lizmat_ joined #moarvm
18:22 jnap joined #moarvm
18:49 BenGoldberg joined #moarvm
18:51 FROGGS hi BenGoldberg
19:25 BenGoldberg nqp-m: say( NQPMatch )
19:25 camelia nqp-moarvm: OUTPUT«(signal SEGV)»
19:32 wsri joined #moarvm
19:35 lue diakopter: I assure you I've never gone beyond master branch. (Also, it's not clang-specific. It happens with gcc too.)
19:59 FROGGS arghh, this print-to-stderr segfault is driving me crazy
20:10 diakopter BenGoldberg: I think that's just crash on null?
20:10 diakopter FROGGS: it's possible we're using libuv wrong there.. maybe you're not supposed to open stderr more than once
20:10 diakopter (so we should cache those)
20:11 jnthn Is it not that it gets sad if the writes didn't finish?
20:13 FROGGS diakopter: I thought about that one
20:13 FROGGS diakopter: but we don't open stderr by default as it seems, so that should be safe
20:14 FROGGS jnthn: I see the printed text.. how should it not have finished?
20:16 FROGGS I added this for example: https://gist.github.com/FR​OGGS/38abae848760ad27216a
20:16 FROGGS but nothing I do helps
20:16 FROGGS trying on windows now
20:21 FROGGS nqp-m: my $err := nqp::getstderr(); nqp::printfh($err, 'test')
20:21 camelia nqp-moarvm: OUTPUT«test»
20:22 FROGGS why doesn't it explode here or on my windows box?
20:27 lue jnthn, diakopter: any possible clues as to what may be causing my instant segfault with nqp-m ? Should I prefix executions of moar with gdb or valgrind to try and find something? (And if so, I should probably recompile moar with debug symbols, no?)
20:27 FROGGS lue: what is your problem?
20:28 lue FROGGS: https://gist.github.com/lue/6968277
20:28 lue ... waitaminute. I think I'm a moron.
20:29 lue I should probably switch out of nqp/moarboot and into nqp/master sometime, huh? :)
20:29 FROGGS :o)
20:29 FROGGS yes, that sounds wise
20:29 FROGGS moar-support on rakudo though
20:29 FROGGS the others are on master
20:30 lue I'm just concerned with compiling NQP-moar again at the moment (because I've done it before.) :)
20:31 lue Whaddya know, it's inching along. :) Sorry for raising false alarms, diakopter.
20:35 lue I assume this is a known error? Routine declaration requires a signature at line 4112, near "(MAST::Nod"
20:35 FROGGS hmmm, you should not get this
20:36 FROGGS it sort of *was* an error and is meant to be fixed
20:36 jnthn I don't think it is, though :(
20:36 lue .oO(At least this time it's not a PEBKAC error, then.)
20:38 cognominal joined #moarvm
20:43 benabik joined #moarvm
21:12 benabik joined #moarvm
21:32 BenGoldberg joined #moarvm
21:51 retupmoca joined #moarvm
22:28 benabik joined #moarvm
22:37 diakopter jnthn: boo
22:41 jnthn diakopter: AARRGHH
22:44 diakopter jnthn: hoo
23:02 diakopter jnthn: ping
23:04 diakopter jnthn: I've got most of a HLL line-hitcounter (profiler) locally
23:06 jnthn ooh
23:06 jnthn Just raw hit counts?
23:07 diakopter yeah
23:07 jnthn k
23:07 diakopter I can do time-in-routine too
23:07 diakopter and I know how to do it to not lose too much precision
23:07 diakopter (average after batches of 50 or so per routine)
23:08 diakopter and not take GB of disk like VS' profiler
23:08 diakopter for 10 seconds of running
23:08 diakopter 30 GB
23:08 diakopter stupid.
23:09 diakopter trouble is recursion... a routine can total more than 100%
23:09 diakopter so the average time-in-routine won't be relatable to total time-in-routine and number of invocations
23:10 diakopter if it's recursive
23:10 diakopter (count recursive calls separately)
23:11 diakopter could even have it only track real named subs/methods instead of every block
23:12 diakopter jnthn: I can do it without resorting to hash lookups, too
23:20 jnthn nice :)
23:20 jnthn sleep &
23:20 jnthn oh, and diakopter++ for hacking on this
23:21 jnthn btw, the tuit supply is looking ropey for tomorrow :(
23:21 diakopter ropey?
23:21 diakopter as opposed to..?
23:21 jnthn Well, I'd hoped I'd get more stuff done today on my talks for Wed :)
23:22 jnthn And have tomorrow mostly for Perl 6 things
23:22 diakopter I didn't get the rope reference
23:22 jnthn Oh...
23:22 jnthn ropey, ropy [ˈrəʊpɪ]
23:22 jnthn 1. Brit informal
23:22 jnthn a.  inferior or inadequate
23:22 jnthn :)
23:23 jnthn That explains it :)
23:23 diakopter k
23:23 jnthn Gosh, I'm such an informal British chap sometimes...
23:23 jnthn ok, really sleep
23:23 jnthn o/
23:23 diakopter o/
23:24 diakopter chap? like chapped hide? chapped lips? chaps as slacks?
23:28 benabik joined #moarvm
23:34 [Coke] '65' is not a valid number at line 2132, near ",\n    66,\n"
23:34 [Coke] (finally tried a build on os x)
23:35 diakopter [Coke]: yeah 'tis busted with gcc on mac and clang on mac and linux
23:35 diakopter but gcc on linux seems ok I think
23:36 diakopter (also msvc)
23:36 diakopter I'm sure it's some number of tiny gc things because of the fragility of the syste
23:36 diakopter system
23:41 diakopter probably more expedient would be a full manual code review

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