Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2014-04-13

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

All times shown according to UTC.

Time Nick Message
00:09 retupmoca should a parameterized role be repossessed like it's generic?
00:10 retupmoca except that would mean it doesn't exist anywhere after serialization...
00:10 retupmoca does each SC need it's own copy of a parameterized role?
00:21 retupmoca unless it's just the generic role itself having issues...
05:18 ashleydev joined #moarvm
05:18 Util joined #moarvm
05:18 tadzik joined #moarvm
05:18 JimmyZ joined #moarvm
05:18 odc joined #moarvm
05:18 lue joined #moarvm
05:18 camelia joined #moarvm
05:18 lizmat joined #moarvm
05:18 brother joined #moarvm
05:18 ggoebel11119 joined #moarvm
05:18 rurban_ joined #moarvm
05:18 daxim_ joined #moarvm
05:18 flussence joined #moarvm
05:18 TimToady joined #moarvm
05:18 eternaleye joined #moarvm
05:18 woolfy joined #moarvm
05:18 FROGGS joined #moarvm
05:18 cognominal joined #moarvm
05:18 cxreg joined #moarvm
05:18 dagurval_ joined #moarvm
05:18 avar joined #moarvm
05:18 ingy joined #moarvm
05:18 moritz joined #moarvm
05:18 BinGOs joined #moarvm
05:19 bcode joined #moarvm
05:19 timotimo joined #moarvm
05:19 _sri joined #moarvm
05:19 jnthn joined #moarvm
05:19 sergot joined #moarvm
05:19 yawnt joined #moarvm
05:19 dalek joined #moarvm
05:19 lee_ joined #moarvm
05:19 tokuhirom joined #moarvm
05:30 synopsebot joined #moarvm
05:31 xiaomiao joined #moarvm
05:38 xiaomiao joined #moarvm
05:50 Ven_ joined #moarvm
05:55 harrow joined #moarvm
07:18 ChanServ joined #moarvm
07:37 xiaomiao joined #moarvm
09:19 ChanServ joined #moarvm
09:19 synopsebot joined #moarvm
10:50 FROGGS[mobile] joined #moarvm
11:49 colomon joined #moarvm
12:36 dalek MoarVM: 4f73c4b | (Steve Mynott)++ | src/platform/threads.h:
12:36 dalek MoarVM: add sched_yield for NetBSD
12:36 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/4f73c4b228
12:36 dalek MoarVM: 2055a67 | jonathan++ | src/platform/threads.h:
12:36 dalek MoarVM: Merge pull request #91 from stmuk/master
12:36 dalek MoarVM:
12:36 dalek MoarVM: add sched_yield for NetBSD
12:36 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/2055a6724e
12:41 zakharyas joined #moarvm
13:38 dalek MoarVM: 161a98b | jnthn++ | src/core/frame.c:
13:38 dalek MoarVM: Fix callsite size calculation in tweak_cs.
13:38 dalek MoarVM:
13:38 dalek MoarVM: Avoids trying to access memory beyond its allocated boundary. Found by
13:38 dalek MoarVM: nwc10++ using ASAN/Valgrind.
13:38 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/161a98bcf1
15:00 hoelzro joined #moarvm
15:06 FROGGS joined #moarvm
15:28 colomon joined #moarvm
15:32 tadzik mm, beer
15:32 jnthn tadzik: IPA again? :)
15:33 tadzik today it's Dark Lager
15:33 tadzik I'm curious what it is
15:33 tadzik it's from the creators of "The Empire Strikes", and it's named "Black Hole"
15:33 jnthn ooh :)
15:34 tadzik ooh, it is good :}
15:34 tadzik next time we meet I'll bring a backpackfull of those
15:35 jnthn :D
15:35 jnthn I'm on the way to Stockholm and found a place last time with a very, very nice Swedish imperial IPA on tap
15:37 jnthn it's 9% or so :) But I might take one tonight :)
15:38 tadzik this one is 11.5
15:38 tadzik I didn't notice until I checked
15:38 tadzik ah, no
15:38 tadzik actually 4.2
15:38 tadzik the weight vs volume trick
15:38 jnthn ...small difference :P
15:39 tadzik oh, the bottle even recommends a particular flavor of glass
15:39 tadzik and recommended food
15:39 tadzik how fancy :)
15:41 jnthn tadzik: If you can hunt it down, http://www.ratebeer.com/beer/great-divide-espresso-oak-aged-yeti-imperial-stout/85174/ comes highly recomended.
15:42 tadzik oooh, the Warsaw Beer Track has some new points on it
15:43 tadzik jnthn: I'll look around for it :)
16:40 colomon joined #moarvm
16:48 retupmoca joined #moarvm
17:06 nwc10 jnthn: works on "my" machine
17:06 jnthn nwc10: Gets rid of ASAN fail?
17:06 nwc10 that one, yes
17:06 nwc10 need this for the second: http://paste.scsys.co.uk/348393
17:07 nwc10 to fix the naughtiness of
17:07 nwc10 m: "\x1F63E".encode("UTF-16")
17:07 camelia rakudo-moar 3e1f32: ( no output )
17:08 japhb joined #moarvm
17:25 dalek MoarVM: 551786d | nicholas++ | src/strings/utf16.c:
17:25 dalek MoarVM: Fix buffer overflow in UTF-16 encoding.
17:25 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/551786d4be
17:50 colomon joined #moarvm
17:53 nwc10 jnthn: the final thing that ASAN is unhappy about, also spotted by valgrind: http://paste.scsys.co.uk/348420
17:54 btyler joined #moarvm
18:32 nwc10 jnthn: valgrind also complains (plus fix): http://paste.scsys.co.uk/348444
18:32 nwc10 jnthn: however, there are other places that assign has_flattening but don't (yet) assign is_interned
18:34 jnthn ah, yes...that really needs to be zero
18:35 jnthn And may explain some of the intermittent bugs folks hvae been seeing with spesh enabled
18:35 nwc10 yes, I did wonder that
18:35 nwc10 as in, if that's garbage but true, it enters the block incorrectly
18:36 jnthn right
18:39 jnthn nwc10: Just made a pass through all the places where has_flattening is set, and is_interned is not, and reviewed them.
18:39 nwc10 I found the right by trial and error
18:39 nwc10 tried two other places, no change from valgrind
18:40 nwc10 and I don't know anything near enough to make the right choice
18:45 dalek MoarVM: fd4cb89 | jonathan++ | src/core/ (4 files):
18:45 dalek MoarVM: Make sure is_interned is zeroed as needed.
18:45 dalek MoarVM:
18:45 dalek MoarVM: Patch partially thanks to nwc10++ (the args.c case), who also found
18:45 dalek MoarVM: the general issue. Other 3 changes are other places vulnerable to the
18:45 dalek MoarVM: same kind of bug.
18:45 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/fd4cb89c69
18:48 japhb At this point, are the known r-m segfaults and ASAN/valgrind complaints all dealt with?
18:50 nwc10 not the crazy threading one in ttp://paste.scsys.co.uk/348420
18:53 colomon joined #moarvm
19:27 zakharyas joined #moarvm
19:52 nwc10 jnthn: yes, works on "my" machine apart from that threading test
20:33 timotimo jnthn: should i count the nameds against the intern arity limit, too?
20:47 jnthn timotimo: um...hmm
20:47 jnthn timotimo: probably
20:49 timotimo now i get a strange error. probably because i have some parts implemented, but not most of them :)
20:52 jnthn hah, yes :)
20:52 jnthn nwc10: OK. That one's on my todo list then :)
20:53 timotimo merge conflicts :\
21:00 dalek Heuristic branch merge: pushed 18 commits to MoarVM/named_to_positional by timo
21:15 * jnthn writed some feedback on https://github.com/MoarVM/MoarVM/pull/89
21:25 timotimo jnthn: what's the direct consequence of me starting to intern callsites that have nameds?
21:25 jnthn timotimo: the is_interned check in frame.c that guards trying spesh will be true
21:26 jnthn timotimo: And then you'll hit args.c, which assumes it's never getting a callsite with nameds.
21:26 jnthn And then drunk unicorns will fly out of the screen.
21:27 brrt joined #moarvm
21:28 jnthn o/ brrt
21:31 timotimo ah, yes
21:32 brrt hi jnthn, timotimo
21:32 brrt good afternoon :-)
21:32 timotimo hi :)
21:34 timotimo This type does not support positional operations
21:34 timotimo hmm
21:34 jnthn tbh, if I was a drunk unicorn I'd probably say that too... :P
21:35 jnthn It probably means you've a P6opaque somewhere a VMArray is expected.
21:35 timotimo in MVM_interp_run, can i easily tell whether i'm in specialized bytecode or not?
21:35 jnthn yes.
21:35 brrt how so?
21:35 jnthn Look at tc->cur_frame->effective_bytecode
21:35 jnthn And see if it's the same as tc->cur_frame->static_info->body.bytecode
21:36 jnthn If they're different you're in specialized bytecode.
21:36 timotimo tanks
21:36 jnthn bombs
21:37 * jnthn ponders if his plan for nqp::timer will work out on JVM
21:37 timotimo er, huh
21:37 timotimo (gdb) print tc->cur_frame->static_info->body.bytecode
21:37 timotimo $4 = (MVMuint8 *) 0x7ff96c2ad6ee <Address 0x7ff96c2ad6ee out of bounds>
21:37 timotimo but they are the same pointer
21:38 jnthn huh indeed
21:39 jnthn if they're the same then you're running the original bytecode.
21:39 timotimo maybe i'm breaking inside a frame that got called from spesh'd bytecode
21:39 timotimo oh well. i'll have a look at spesh/args.c and see what i can do
21:39 jnthn That's possible.
21:39 jnthn Well, you can look at caller
21:40 jnthn Also you can set MVM_SPESH_DISABLE=1 and see if your problem goes away
21:40 jnthn If it does, then I'd probably get a working Moar, get the spesh log, then get your changed Moar, get its spesh log, and diff.
21:40 jnthn Because in theory nothing shoulda changed.
21:42 timotimo right
21:43 timotimo it still segfaults even with DISABELE=1
21:44 jnthn did you spell DISABLE right in trying it?
21:44 brrt (bloody legacy x86 registers)
21:44 timotimo MVM_SPESH_DISABLE=1
21:44 jnthn right
21:44 jnthn So the breakage is nothing to do with spesh.
21:44 timotimo the error happens in MVM_string_equal called from MVM_args_get_named_obj
21:45 timotimo it's trying to find the string "inheritable" in the args
21:49 brrt (good luck folks, i'm going to sleep again :-))
21:49 brrt left #moarvm
22:11 timotimo well, when i don't intern callsites that don't have nameds, it still works. so ...
22:11 timotimo i'll continue thinking about this later. tomorrow or so.
22:11 jnthn ok :)
22:11 jnthn No hurry :)
22:12 * jnthn is currently tied up pondering timers :)
22:37 colomon joined #moarvm
22:59 colomon joined #moarvm

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