nwc10 Good *, #moarvm 10:08
diakopter good 10:09
jnthn o/ nwc10
FROGGS morning
jnthn_ diakopter: What are the code ID and callsite ID? Just unique IDs for each one of them? 10:47
diakopter yeah 10:51
to be traced back to static frames and callsites within static frames, respectively
er, the lines numbers and sub names and filenames associated with those things, I mean 10:52
diakopter www.youtube.com/watch?v=WCqm4H3m3Ew 10:54
FROGGS EYOUPOSTEDITTWICE 10:59
:P
FROGGS .oO( Dear Diamond Diamond Kopter... )
timotimo i think i'll go ahead and do some little big int work now. 18:21
i can't stand the low amount of moarvm commits recently :P
TimToady timotimo++ 18:27
timotimo diakopter: should i put the malloc for the mp_int * into the allocate 'method' of the P6bigint repr? 18:52
that seems like the most sensible place to put it 18:53
diakopter usually in the init
the allocate is meant just be interaction with the gc/managed allocator
wow, I've dropping lots lately
timotimo oh, hm. 18:54
diakopter what about init?
timotimo what about it? 18:55
it has the malloc in it now
diakopter oh
timotimo and in gc_free i put the free
diakopter good
the MVM_checked_free_null macro is convenient there
timotimo i'll have a look 18:56
i'm running into a coredump, thoughd 18:58
diakopter eww
TimToady diakopter: now that moar is getting along pretty well, you see any big blockers for starting in on P5 interop sometime soon? 19:00
timotimo er, a segfault
MVM_repr_alloc_init seems to return a null pointer in the "i" slot
does that mean using MVM_repr_alloc_init is not the right function to use?
diakopter well in somes cases, init takes parameters, iirc, so yeah you'd want to split it up into calling the alloc and init separately 19:01
timotimo i'm not using malloc wrong, right? body->i = malloc(sizeof(mp_int))?
diakopter as long as ->i is pointer-sized, seems okay 19:02
how are you testing it for being null?
timotimo gdb
diakopter hrm
timotimo yeah, the malloc returns 0 19:06
japhb__ Is FOSDEM still going on, or is that over now?
lizmat FOSDEM is over
timotimo er, nope.
lizmat was only Sat/Sun
FROGGS o/ 19:07
japhb__ Ah, OK. Any rough ETA on video?
lizmat Wendy and I just got home and unloaded the van
check the FOSDEM website
it was all done by the organization, not by any of the devroom people
japhb__ video.fosdem.org/2014/ # Gah, I was going to say "Yay, lots of videos!" and then I went "Boo, all named badly!" and "Wow, small files ... hmmm" finally ending with "Ah shucks, that top filename says it all." :-( 19:09
jnthn: Slide decks, pretty please with your choice of garnish on top? ;-) 19:11
Any other 6'ers/Moarvians give a talk?
diakopter: Did you manage to create the slide deck for your talk before you got sick? 19:12
timotimo i'm confused. it segfaults in a pretty benign looking mp_copy
huh. 19:14
in copy_to, would i need to check if the target object already has a mp_int allocated for it?
well, copy_to isn't executed anyway. 19:15
(before the segfault happens)
whhhhyyyyyyy >:( 19:31
jnthn ffs, the netsplits have made a real mess of my irssi windows... 19:35
japhb__: I linked slides earlir on #perl6 19:36
japhb__ jnthn: Ah, sorry, I'm about a day backlogged there.
I'll go look for them on my next mental break
.oO( Hopefully not a psychotic one )
jnthn japhb__: jnthn.net/papers/2014-fosdem-perl6-today.pdf 19:37
japhb__ Oh, thanks! 19:38
jnthn ok, about time to get off the airport train...
Probably back tomorrow, in the moments I can steal while teaching :)
timotimo oh, hm 19:39
i've been printing the result of (MVMP6bigint *) casting a P6Opaque that was merely containing a P6bigint 19:40
jnthn &
timotimo (ooc, why do we do that?)
at least that tells me why i've been finding null pointers to be in the "i" slot 19:41
okay, now i'm convinced i need help. 19:49
it seems like get_boxed_ref is misbehaving quite radically 19:50
diakopter: would you be so kind and check out the little_big_int branch? (you may need to have it overwritten when pulling or something) 19:52
get_boxed_ref returns something that's not the mp_int i've been trying to build 19:53
timotimo diakopter: can you tell me why the objects that get created are a P6opaque rather than "just" a P6bigint? is that because they also have class-stuff? 21:22
timotimo well, i suppose that's because it's a p6bigint boxed into a p6 class 21:43
FROGGS froggs@TL02L-R8RXTCW-linux:~/dev/MoarVM$ perl6-m -e 'my $fh = "/usr/bin/wc".IO.open(:p, :w); $fh.print("hello world\n"); $fh.close;' 21:43
1 2 12
froggs@TL02L-R8RXTCW-linux:~/dev/MoarVM$ perl6-p -e 'my $fh = "/usr/bin/wc".IO.open(:p, :w); $fh.print("hello world\n"); $fh.close;'
1 2 12
\o/
no segfault, just memleak!
here is the openpipe diff for everybody who is curious: gist.github.com/FROGGS/998bacca7b3d3abab644 21:50
next: read from a child's stdout and then get rid of mem leaks 21:51
timotimo \o/ 22:30
i'm running into a situation where OBJECT_BODY(obj) inside get_bigint seems to return crap 22:31
and i don't know enough about the stuff to figure out why it's going wrong
(gdb) print *(MVMObject *)data 22:32
$32 = {header = {owner = 0, flags = 0, size = 0, forwarder = 0x2f03d20, sc = 0x30000000000000},
st = 0x0}
jnthn timotimo: I rather doubt data, if it's a thing passed into a REPR func, is an MVMObject 22:50
It already *is* the object body 22:51
timotimo yes it should be the body 23:16
why did i think it would be that?
japhb__ jnthn: Got a chance to read your slides. I really liked this new example, and how you ran with it. The code was a little dense (pseudo-golfing idioms), but fun. 23:31
And FWIW I've used the "factor out a special iterator into a private method that is used kinda like for" idiom and really like it. 23:32
timotimo jnthn: can you help me with my boxing trouble? :\ 23:55
timotimo in the new little_big_int branch, you'll see that get_bigint is looking at a wrong offset or something? 23:57

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!