brrt hi #moarvm, just wanted to report that i finally handed in my proposal 09:13
JimmyZ Hello, I can't see it 09:14
moritz brrt++
brrt: please join #soc-help on irc.perl.org 09:15
JimmyZ hopes there is a gist version
brrt will post a gist version sure :-) 09:16
ok will do moritz
maybe after i restart X though, brb
brrt gist.github.com/bdw/9682548 -> html version in a gist 09:22
should've used markdown? yeah, probably
JimmyZ sounds good 09:33
brrt thanks :-) 09:40
perl foundation wishes me to actually fix a bug
so i'm off to the issues tracker
is this fixed by now? github.com/MoarVM/MoarVM/issues/66 - because i see Configure.pl emitting 'updating submodules' when i run it 09:43
moritz brrt: no, it's not fixed yet 09:44
the submodules are only about updating the source code 09:45
brrt: the problem here is that once the stuff in 3rdparty/ is built, you have to type 'make realclean' to rebuild it
brrt: because the MoarVM makefile has no deps between the 3rdparty sources and the 3rdparty object files 09:46
brrt ok, seems doable
brrt takes a look
github.com/MoarVM/MoarVM/issues/62 this one seems fun, too 09:47
JimmyZ \o/, fix all of them! 09:49
yeah, the gc one seems more fun 09:50
brrt i know 09:51
jnthn .tell brrt On your proposal, the bytecode graph building part will be done by the time you get there, I think. Possibly some of the type stuff too. 10:42
hmm, no message bot?
FROGGS gone :/
jnthn .tell brrt I think your intermediate set will want some things that can't exist in regular MoarVM bytecode too - unsafe things that we prove the correctness off.
ok, I guess somebody can point him at backlog if he doesn't see it himself :) 10:43
FROGGS I'll try to :o)
nwc10 jnthn: unsafe things that we *can't* prove the correctness of? 10:45
jnthn uh
nwc10 well, a static verifier can't prove the correctness of, so mustn't be in loaded bytecode, which can't be trusted completely
jnthn unsafe things we can prove the correctness of.
nwc10 oh, I see 10:46
jnthn Well, there's multiple levels of this
The verifier proves a bunch of static properties
nwc10 potentially unsafe things that we can prove the correctness of at generation time, but can't do at load time
jnthn But if you're JITting something with guards, then you can rely on more stuff in protected code-paths.
Like, you can dig straight into an object and drag out an attribute, once you know its type.
brrt: "Notably in the django world were everything is always broken" => harsh... ;) and s/were/where/ :) 10:47
brrt: By the way, will you be at Dutch Perl Workshop? 10:49
brrt jnthn - seen backlog, seems ok 11:19
when is the dutch perl workshop? 11:23
24th of april 11:27
... would be nice, but i'm not sure what kind of study obligations i'll have
nwc10 oh, a Thursday
you could attend the beer^Wfluxx and whisky track
brrt 25th of april, sorry :-) 11:30
yes, maybe in the afternoon
nwc10 oh, a Friday. That makes more sense. :-)
you are free on Saturday? This is probably important 11:31
brrt :-) yes, i'm free on saturday
nwc10 I have not been to the Dutch Perl Workshop, but I understand that everyone stays up very late.
or at least, quite a few people choose to
brrt updated the proposal :-) 11:35
brrt jnthn, are you there? 13:56
i was challenged by pmurias on thinking up a list of opcodes that could be JIT-ed 13:57
and i'm thinking - only a pretty small set, really
tadzik Internal error: Unwound entire stack and missed handler -- does that sound like a moarvm thing? 15:38
hm, on jvm it's a different exception 15:39
heh, most curious :) 15:40
it happens when C calls Perl 6 which calls a method with not enough arguments given 15:41
FROGGS tadzik: MoarVM/src/core/frame.c:552: MVM_panic(1, "Internal error: Unwound entire stack and missed handler"); 15:57