Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2013-12-21

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

All times shown according to UTC.

Time Nick Message
00:24 TimToady reproduced the mbc failure this time, sigh
00:26 TimToady trying some git submodule syncs this time
00:27 TimToady and a reconfig
00:27 TimToady (after a realclean
00:27 TimToady )
00:27 diakopter )
00:28 TimToady on my supercomputer, m-spectest got up to /S02-literals/types.t before hanging permanently
00:37 TimToady still mbc-busted, trying realclean, reconfig in MoarVM
00:42 diakopter du4F)
00:53 * lue wonders if that "uv__loop_delete" assertion will be Moar's Null PMC ...
00:58 [Coke] TimToady: you can load Test.pm?
01:01 TimToady on one machine, but not on the other
01:01 TimToady in fact, on the other, I haven't even quite got back to the state of compiling Test.pm (mbc, m-BOOTSTRAP)
01:01 TimToady probably about time to blow the whole thing away and try again from git
01:02 TimToady funny thing is, these are both Linux Mint 15 machines, 64-bit
01:08 diakopter lue: no, it's just a platform quirk
01:08 diakopter combined with a misthink by me|others
01:09 diakopter TimToady: yeah I found a missing dependency or two from scratch
01:09 lue diakopter: I meant more in that "this problem is deeper than just Rakudo" feeling the error gives. Time will tell if the assertion error lasts as long though :)
01:09 diakopter (as written above)
01:10 diakopter lue: no, I'm saying I know for a fact it's a tiny subsystem and an error condition that can only be reached at the very end of execution
01:10 diakopter as opposed to the null pointer exception or nullpmc that will likely be just as prevalent on moar as it is currently on jvm and parr
01:11 lue OK, wasn't sure how literally would you were transposing the idea of Null PMC.
01:11 diakopter well that error can occur in a huge number of places
01:11 diakopter but the libuv loop delete assertion is in only one place and is affected by code in only a ocuple places
01:15 lue OK, the error just happens to crop up more often than others because of the specific places it gets triggered, not because it's a generic signal for bad code. Got it.
01:35 jnap joined #moarvm
02:22 TimToady by adding a --stagestats (filling in a missing parameter?) I got a Test.pm compile
02:22 TimToady now spectesting on my laptop
02:24 colomon joined #moarvm
02:26 TimToady t/spec/S02-magicals/vm.t ...................................... ok
02:27 TimToady that took at least two cpu minutes to get ok of a 3/3
02:30 TimToady hmm, spectest does not seem to be accessing lib/Test.moarvm nor blib/Test.moarvm
02:31 [Coke] do you get an error with ./perl6-m -Ilib -MTest -e1; ?
02:34 TimToady oddly enough, not updating the access time on Test.pm either
02:35 TimToady no error, but it also doesn't update the last access time
02:35 * TimToady scratches head
02:36 TimToady is there a way to open a file that doesn't update last access?
02:36 TimToady or would not closing it have that effect?
02:36 TimToady are these files mmapped in?
02:37 TimToady I suppose I could do an strace...
02:44 TimToady Hah!
02:44 TimToady it's *not* accessing either Test.pm or Test.moarvm
02:44 TimToady stat("/home/larry/nom/lib/test.pm", 0x7fff9fccace0) = -1 ENOENT (No such file or directory)
02:44 TimToady stat("/home/larry/nom/lib/test.pm6", 0x7fff9fccace0) = -1 ENOENT (No such file or directory)
02:44 TimToady stat("/home/larry/nom/lib/test.moarvm", 0x7fff9fccace0) = -1 ENOENT (No such file or directory)
02:44 TimToady note the lowercase
02:44 TimToady this is Windows branedamage
02:45 [Coke] ARGLEBARGLE
02:45 [Coke] TimToady++
02:46 TimToady diakopter: ^^^
02:47 TimToady dinner &
02:47 diakopter gah
02:47 diakopter <- don't know where to fix that
02:48 TimToady .oO(Redmond?)
02:48 TimToady rafk &
02:49 [Coke] If no one figures it out, I can check it out after blogventing.
02:51 colomon joined #moarvm
03:31 colomon joined #moarvm
04:39 TimToady a lot of the hanging tests seem to do with enums
04:43 TimToady but not all of them
04:52 TimToady well, there's a lot of others too, with no apparent pattern
04:52 diakopter hrm
04:53 diakopter hanging burning cpu or not?
05:01 TimToady yes, there're about 20 burners
05:02 TimToady probably just need to cut one down to a minimal test, assuming it's not GC craziness
05:04 TimToady well, chances are most of 'em are the faked continuations
05:11 TimToady well, that's also a mere guess...
05:21 diakopter it's almost certainly a loop as opposed to polynomial or exponential or factorial or...
05:25 TimToady failed 179 test files
05:25 TimToady which ain't too shabby at this point
05:26 TimToady otoh I noticed roast's standards for e and pi are quite low
05:28 diakopter "Windows branedamage" is damage caused by using Windows, or damage to Windows caused by some kind of brains?
05:33 jnap joined #moarvm
05:36 [Coke] TimToady: when running the speculative spec tests, I use a local ./perl6 that is a ulimited perl6-x
05:36 [Coke] that might at least help you get past the hangers.
05:37 TimToady first minimal looping test: class B is C { }
05:37 lue ooc, to those who know MoarVM, can you specify multiple --libpath paths? Either through use of multiple --libpaths or a [delimiter]-separated list to one --libpath ?
05:42 TimToady oh cool, Internal error: invalid thread ID in GC work pass
05:43 diakopter heh.
05:43 diakopter there shouldn't be more than 1 thread :)
05:44 diakopter TimToady: can perl6-m compile that input to a .moarvm? and can moar --dump that .moarvm?
05:49 TimToady here's another looper: class Bar {}; role B { has $.attr is rw = 42 }; my Bar $b .= new(); $b but B;
05:50 TimToady dunno, was on the way to getting that snippet, which loops
05:50 TimToady lemme see if I can minimize the GC carpage instead
05:54 TimToady doesn't really minimize, chopping random bits out makes it loop instead
05:56 diakopter heh,
05:57 TimToady moar --dump segfaults on it
05:58 TimToady but then, it segfaults on lib/Test.moarvm too :)
06:00 TimToady anyway, t/spec/S14-roles/basic.rakudo.moar if you want to fiddle with it yourself
06:00 TimToady about half the time I can get a segving compile to pass by adding --stagestats
06:01 TimToady and sometimes it even succeeds before it segvs, so probably segv is from exiting somehow
06:02 * TimToady yawns hugely
06:38 * diakopter huge yawnsly
07:35 jnap joined #moarvm
08:00 [Coke] ... I really wish I had been trying to build stuff in between watching that stargate movie. whoops.
08:03 [Coke] add in an nqp:say to get some diagnostic output, introduce a segfault.
08:18 [Coke] diakopter: it'd be nice to have whatever host06's "trace" command is available.
08:19 [Coke] I was trying to fix the issue TimToady was seeing with search for wrong-cased names, but I don't see it in the code, want to see if it's just his install
08:34 cognominal joined #moarvm
08:35 jnap joined #moarvm
09:36 jnap joined #moarvm
10:35 woolfy joined #moarvm
10:35 lizmat_ joined #moarvm
10:37 jnap joined #moarvm
11:56 FROGGS TimToady / diakopter: --dump is known to be broken
12:33 tgt joined #moarvm
12:38 jnap joined #moarvm
13:33 colomon joined #moarvm
13:39 jnap joined #moarvm
14:08 cognominal joined #moarvm
14:11 jnap joined #moarvm
15:15 wsri joined #moarvm
15:33 TimToady [Coke]: sorry, realized this morning  I mighta typed -Mtest instead of -MTest
15:38 * TimToady apologizes to all those with Windows branedamage :)
15:44 TimToady if I'd stuck the 'use Test' into the -e I'd've noticed; I'm not a habitual user of -M, so the initial capness of the M ate the initial capness of the T in my brane
15:45 FROGGS still good to know :o)
15:46 TimToady just regret the wild-goose chase I sent [Coke]++ on
15:49 cognominal joined #moarvm
15:53 FROGGS sure, but it is good that this was a level 8 error instead of something else
16:18 cognominal joined #moarvm
17:00 FROGGS joined #moarvm
17:30 colomon joined #moarvm
18:18 ssutch joined #moarvm
20:49 diakopter nwc10: the solution is to have the thread mark the mutex with its thread id after locking it, so in case it's locked, and it's marked with its own thread id, it .. increments a depth counter (like you said, faking up a recursive one) .. and then needs to be sure to decrement it when unlocking, and if it decremnts it to zero, unset the owner threadid and then unlock the mutex
20:51 diakopter nwc10: I think that's the only one to worry about..
20:51 diakopter nwc10: nice find. :) nwc10++
20:58 tgt joined #moarvm
21:00 [Coke] TimToady: eh, I understand a little bit moar now.
21:00 [Coke] I am bummed that I cannot load Test.pm though. :|
21:01 [Coke] I get a segfault trying to moar --dump lib/Test.moarvm
21:02 * TimToady did: strace install/bin/moar --libpath=/home/larry/nom/install/languages/nqp/lib perl6.moarvm -Ilib -e 'use Test; say "hi"' >/tmp/st 2>&1
21:02 TimToady that seems to work
21:02 TimToady someone said the --dump is a known deficiency
21:04 [Coke] if I just run moar lib/Test.moarvm, I get:
21:04 [Coke] Unhandled exception: Bytecode validation error at offset 2552, instruction 368:
21:04 [Coke] extension op 'p6box_i' not registered
21:16 timotimo huh? weird.
21:16 timotimo did you try a full clean rebuild?
21:16 timotimo we couldn't have made it through the setting if we didn't have p6box_i, could we?
21:17 diakopter yeah; the dependency of Test.moarvm on perl6.moarvm is apparently not being added
22:25 jnap joined #moarvm
23:25 jnap joined #moarvm
23:45 nwc10 joined #moarvm

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