02:06 colomon joined 02:48 ilbot3 joined 02:51 colomon joined 06:04 harrow joined
FROGGS happily deletes a lot of jnthn's code 09:10
09:12 sivoais joined 09:22 vendethiel joined 09:38 woolfy joined, woolfy left
FROGGS jnthn: the segfault from yesterday can be turned off by doing: 10:06
src/spesh/log.c:61 @@ void MVM_spesh_log_add_logging(MVMThreadContext *tc, MVMSpeshGraph *g, MVMint32
- case MVM_OP_getlexstatic_o:
+ //~ case MVM_OP_getlexstatic_o:
jnthn: does that make any sense to you?
m: use Test; class Foo { method bar { } }; my $foo = Foo.new; my @a = <a b c d>; sub baz(|) { ok $foo.bar, "bar" }; for @a { for @a { baz @a } } 10:07
camelia rakudo-moar 7f2085: OUTPUT«(signal SEGV)not ok 1 - bar␤␤# Failed test 'bar'␤# at /tmp/9vhi17JrwM line 1␤not ok 2 - bar␤␤# Failed test 'bar'␤# at /tmp/9vhi17JrwM line 1␤not ok 3 - bar␤␤# Failed test 'bar'␤# at /tmp/9vhi17JrwM line 1␤not ok 4 - bar␤␤# Faile…»
FROGGS this one
the funny thing is that the logging itself seems to trigger it 10:11
commenting out the case MVM_OP_getlexstatic_o in optimize.c does not change anything 10:12
I dunno how to continue from here :o( 10:27
jnthn FROGGS: Doesn't make much sense to me that logging would do it, no... :S 11:08
FROGGS: It's more likely something using the logged info is to blame 11:09
FROGGS hmmm
jnthn in facts.c there's a log_facts that incorporates the facts from the logs. Commenting the call to that will help test that hypothesis. 11:10
FROGGS do you mean I shall comment out this one? 11:12
//~ if (po != MVM_OP_getlexstatic_o && po != MVM_OP_getlexperinvtype_o)
//~ log_facts(tc, g, bb, ins);
jnthn Yes 11:13
FROGGS that does not change anything
jnthn Hmmm
FROGGS I commented out every mention of MVM_OP_getlexstatic_o, and only the one in log.c makes the error go away 11:14
jnthn o.O 11:16
That's really odd.
FROGGS aye
that's my conclusion also :o9
:o)
jnthn I want to continue on natives stuff at the moment 11:17
Is there an issue for this bug?
Looks like now
*not
jnthn added one and linked the irc log from above 11:18
FROGGS jnthn++ 11:19
11:32 tgt joined 11:59 rurban__ joined
dalek arVM/native-ref: eecd6ee | jnthn++ | src/ (3 files):
Implement native attribute reference taking ops.

This is an initial, basic, implementation; we'll be able to make it a bit more performant in a few ways later (without any outside API changes, which is the key thing).		 12:00
12:08 colomon joined
dalek arVM/native-ref: 77b0480 | jnthn++ | src/6model/reprs/NativeRef. (2 files):
Stub native reference read functions.
 12:37
arVM/native-ref: 8d142cd | jnthn++ | src/6model/containers.c:
Flesh out native ref container fetch functions.
arVM/native-ref: 59ef11f | jnthn++ | src/6model/reprs/NativeRef.c:
Implement native ref reading.
arVM/native-ref: ec64685 | jnthn++ | src/6model/reprs/NativeRef. (2 files):
Stub native reference write functions.
 12:54
arVM/native-ref: b759791 | jnthn++ | src/6model/containers.c:
Implement native ref container store functions.
arVM/native-ref: 407a85f | jnthn++ | src/6model/reprs/NativeRef.c:
Implement native ref attribute writing.
 13:01
14:09 tgt joined 14:45 rurban__ joined 14:46 lue joined 14:50 [Coke] joined
dalek arVM/native-ref: cebc244 | jnthn++ | src/ (3 files):
Implement ops for taking positional references.
 15:17
arVM/native-ref: 732696c | jnthn++ | src/6model/reprs/NativeRef.c:
Implement positional ref read/write.
15:33 colomon joined 16:05 harrow joined 16:49 colomon joined
jnthn lexrefs, y u no work... 17:13
Ah, they do, but the Rakudo optimizer doesn't know not to optimize lexicals out when I use the ref-taking by name op. Fair enough. 17:23
timotimo good catch. that means we will have $x++ ob native types work? 17:24
dalek arVM/native-ref: eb2e520 | jnthn++ | src/6model/reprs/NativeRef.c:
Native lexical ref read/write functions.
arVM/native-ref: ca82f8a | jnthn++ | src/ (3 files):
Stub native lexical ref taking ops.
arVM/native-ref: ba84cf3 | jnthn++ | src/6model/reprs/NativeRef.c:
Implement native lexical reference by name ops.
jnthn timotimo: Well, I've several more steps to get us there.
timotimo: But yeah, hopefully by tomorrow we have that bit.
Already today got us rw accessors for native attrs and fixed the Buf assignability problem somewhat :) 17:25
timotimo \o/ 17:39
i like to hear that
much closer to Struct-like classes now 17:40
18:30 FROGGS[mobile] joined 19:14 colomon joined 19:33 zakharyas joined 19:35 colomon joined 20:12 FROGGS[mobile]2 joined
FROGGS jnthn++ # :o) 20:13
20:21 V_S_C joined
timotimo jnthn: would you like to tell me about what stuff could be optimized with the native-ref branch in spesh? or perhaps tell about the "generate an accessor/... for me please, dear VM" thing 20:40
or perhaps you have a clever idea for how to compile BUILDPLAN and BUILDALLPLAN to actual code
is that all we were thinking of? buildplan, buildallplan and accessor methods? 20:42
jnthn Yeah. I was thinking of providing some compiler services API that can be used for this. 20:48
And it's only available if you're currently at compile time
timotimo who gets to decide? :D 20:49
jnthn Basically it'd mostly use code available in Perl6::Actions and Perl6::World
"The compiler" :)
I think we can do it with a contextual *if* we're careful.
timotimo how about composing stuff for classes with the MOP? 20:50
that's not at compile time …
jnthn Then we fall back to the existing approach 20:51
If you do your generation at BEGIN time, you get the fast stuff
timotimo OK :)
well, if my ADT module generates the classes when it gets use'd by something else, it's at run time during compile time 20:54
so it should be fine, i guess?
jnthn That's runtime of the module 20:57
Not sure if we can get away with that
timotimo hehe.
i wouldn't get BEGIN time inside the module multiple times if there's multiple different use lines, so that's not so awesome 20:59
21:18 sivoais joined 22:04 FROGGS_ joined, ChanServ joined
FROGGS dalek: ? 22:46
I lust committed: updating dyncall (still 0.9) and libuv (0.11.29 to 1.0.0) 22:47
review: github.com/MoarVM/MoarVM/commit/79...34c787f254
@all: please test... works perfectly on my linux box 22:48
22:48 d4l3k_ joined
timotimo cool. any interesting features? 22:49
FROGGS and yes, libuv made their 1.0.0 release, so it is important for us to test it, since this will be packaged and we perhaps have to live with that for a while
timotimo or important bugfixes?
FROGGS read the top 139 lines here: github.com/joyent/libuv/blob/master/ChangeLog 22:50
mostly (minor?) fixes and docs
for dyncall it is just build system stuff I think: github.com/MoarVM/dyncall/compare/......903ef9c 22:51
22:51 ingy joined
timotimo OK 22:51
well, it's still good everything is good
22:52 harrow joined
FROGGS aye 22:52
23:02 Util joined 23:09 zakharyas 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!