00:01 lastofthe joined 00:38 lastofthe joined 01:54 FROGGS_ joined 02:18 itz_ joined 02:26 JimmyZ joined 02:36 lastofthe joined
lastofthe watches CFLAGS leak from his moarvm checkout to his rakudo checkout 02:38
I'm used to localizing compiler warnings to single files, not across repositories :) 02:39
04:26 itz joined 06:06 FROGGS[mobile] joined 07:23 woolfy left
FROGGS lastofthe: rakudo compiles ext_ops and need the MoarVM flags (it also include MoarVM header files) 08:02
08:59 brrt joined
brrt \o 09:10
jnthn o/ brrt 09:17
brrt \o jnthn
still in china?
jnthn brrt: Yes.
brrt also, do chinese have their day off on sunday, like we do?
jnthn Yeah, no work on Sat/Sun for me. 09:18
brrt that is.. interesting, in historical perspective :-)
how is it? 09:19
jnthn Work stuff is all fine. Finding plenty of tasty things to eat. It's really quite hot, however. 09:21
brrt oh, i see
i was complaining about 26 deg C here
when are you coming back? 09:22
jnthn I'm here the rest of September, then coming back. So, another few weeks here. 09:26
Today the forecast said it was 34 C, but with the humidity felt like 40. Having been outside, I'd tend to agree :)
brrt :-) i hope they do have airconditioning 09:29
jnthn Yes. :) 09:30
In the hotel. In malls. In the subway. At work. In taxis. :) 09:31
brrt that's something :-) 09:35
hmm, i'm wondering if you can advice me on the (posssible) JITting of the param_* ops 09:36
basically, i've argued that as-is is impossible 09:38
or 'not feasible given the differences between win64 and posix abi's'
jnthn Right, 'cus of the struct returning. 09:39
brrt i noticed that within each of them, there is significant dispatching to mediate between desired type and given type
i can't help but think that is suboptimal
jnthn Well, we do at least know what the desired type is... 09:40
So yeah, switching on that could be avoided if we're doing it
brrt ok. what about the required flag? not having it is a): cheaper to call b): unifies the behavior of optional and required parameters, i.e. the optional parameters jump and the required parameters throw 09:42
maybe i should just make a branch for it
jnthn I don't see that it has to be done the flag way it is at present. 09:43
Could lose that if there's a better way. 09:44
brrt ok :-) thanks 09:48
i'll go puzzle on it
have to be writing a proposal for the time being, though
jnthn :) 09:49
Hope all is well with the course so far?
brrt yeah, just very busy, lots of overlapping duties :-) 09:50
brrt bbiab 09:54
dalek arVM: 8bcc260 | jonathan++ | src/gc/orchestrate.c:
Remove an unused variable; timotimo++.
 10:11
arVM: 2eee415 | jonathan++ | src/ (7 files):
Add objects needing finalization to finalize list.

Not using it for anything yet; this just makes sure objects with types that have opted in to finalization reliably end up on the finalize list of their allocating thread.
arVM: 415e39c | jonathan++ | src/core/hll. (2 files):
Allow per-HLL finalize handler installation.
 10:34
11:20 brrt joined
brrt oh, lots of gc activity, nice 11:21
by the way, if i'm implementing something to deal with force_gc (a backend feature that was requested on #perl6), should i enter from MVM_gc_enter_from_allocator? 11:22
jnthn Yes
brrt will that deal correctly with the case when another thread has already started a collection (i.e. can it race?) 11:23
oh i see 11:24
it will
jnthn Yeah
It's just the one that is called if you're triggering a GC run, as opposted to being told you shoud join an already triggered one.
brrt ok. what happens if i don't add arguments to an op in the oplist? 11:25
(nothing, ok) 11:26
jnthn brrt: It's fine; there are a number of ops like that, such as noop and osrpoint 11:27
brrt i see :-)
jnthn dinner & 11:41
12:01 FROGGS_ joined 12:16 ggoebel1111113 joined 12:23 Ven joined
dalek arVM: c52ac9d | (Bart Wiegmans)++ | / (6 files):
Add force_gc instruction
 13:53
timotimo while falling asleep (or rather: trying to), i had a few random ideas go back and forth through my head 14:13
1) with the spesh framework we could build a coverage report tool and not only do line coverage, but also branch coverage and path coverage
2) if we had something like =SECTION in pod be taken into account by the code-gen and set something like $?SECTION (in addition to $?FILE), we could have errors in the core setting point to the file the code comes from 14:14
3) this idea has already decayed in my memory a fair bit, but something about finding out if we are in a git repository, if it's on github and then building a link to the github website for a given file and line number 14:15
JimmyZ src/gc/finalize.c:24:36: warning: assignment from incompatible pointer type [enabled by default] 14:16
tc->finalize[tc->num_finalize] = obj; 14:17
new warning under gcc
jnthn Hmm
I'll add a cast or some crap
hm, or maybe the collection type it's going into wants to chagne, in fact...
14:20 brrt joined
dalek arVM: 3ba1d01 | jonathan++ | src/core/threadcontext.h:
Hopefully fix a warning.

Reported by JimmyZ++.		 14:20
JimmyZ it fixed! 14:22
brrt \o/ 14:23
15:07 JimmyZ_ joined
brrt domestic happiness is.. when your SO finally removes the dust from his/her laptop, thereby silencing the fan 15:35
i'll be studying again :-) 16:01
brrt afk
16:01 brrt left
dalek arVM: b477a17 | jonathan++ | src/gc/ (3 files):
Start detecting objects needing finalize.

We don't do anything with them yet once we've detected they need to be finalized, but this seems to get the detection of that case working.		 16:01
jnthn Happy studying :)
16:04 lizmat joined 16:09 synopsebot joined 16:12 leedo joined 16:27 woolfy joined
dalek arVM: d94c91a | jonathan++ | src/ (4 files):
Collect to-finalize objects into finalizing list.

This will later be used in order to pass them off to the finalize handler in high level code.		 16:35
jnthn And I'll do that piece tomorrow. :) 16:38
timotimo are you going to bed now? 16:39
lizmat jnthn++
timotimo mh, will be afk for a bit
lizmat question: what signature should I use to get an unboxed string (native?) to call that candidate? 16:40
jnthn: use case: .chomp on a native str would not need to do an extra boxing/ unboxing 16:41
dinner& 16:42
jnthn lizmat: Calling a method *always* happens on a boxed object 16:55
lizmat: So it'd need to be a sub chomp(str $chompee) { ... } 16:56
16:56 colomon joined
jnthn Anyway, yes, sleep time her enow... :) 'night 17:04
timotimo gnite jnthn :)
17:22 zakharyas joined
japhb Dang it! Can someone please tell brrt to STAY ONLINE long enough for me to reply to him? He keeps PM'ing me with a mysterious request, and then going offline before I can respond to him, so I have no idea what he needs. :-/ 19:08
FROGGS_ japhb: he backlogs
japhb FROGGS: But he's PMing (presumably for good reason), so I can't respond in channel. 19:10
(But yes, hopefully he'll see the above remark. :-)
FROGGS japhb: I mean, he will read this here, so he'll know to stay up longer :o)
japhb: maybe you should chat via email or so 19:11
japhb The problem is, he tends to go off IRC immediately upon doing something else, so even in his prime hours I often can't reach him.
Yeah, I was thinking the same.
19:12 cognome joined
japhb brrt: I'm gjb at either my current or previous employer (google.com or sonic.net). 19:13
20:21 cognome joined 21:17 avar joined 21:21 betterworld joined 22:05 avuserow joined 22:17 betterworld joined 22:22 nine joined 23:12 dalek joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!