|
00:02
colomon joined
02:40
colomon joined
02:47
ilbot3 joined
03:24
camelia joined
05:45
vendethiel joined
06:47
japhb joined
07:17
japhb joined
09:58
rurban__ joined
10:33
tgt joined
|
|||
| dalek | arVM/native-ref: 86bf121 | jnthn++ | src/6model/reprs/P6opaque.c: Liberalize mixin prefix test in P6opaque. Up to the point we hit the first attribute-contributing class, we can skip the checking. This makes the Perl 6 MOP re-parenting not break the ability to mix in to meta-objects. |
11:26 | |
|
12:03
mj41 joined
13:04
colomon joined
13:21
rurban__ joined
13:55
colomon joined
15:22
zakharyas joined
|
|||
| FROGGS | perl6-m -MNativeCall -e 'sub PolygonShape_Set_ZN14b2PolygonShape3SetEPK6b2Vec2i(OpaquePointer, int32) is native("libBox2D") { * }; _ZN14b2PolygonShape3SetEPK6b2Vec2i(OpaquePointer, 0)' | 16:11 | |
| moar: /build/buildd/box2d-2.3.0+ds/Box2D/Box2D/Collision/Shapes/b2PolygonShape.cpp:122: void b2PolygonShape::Set(const b2Vec2*, int32): Assertion `3 <= count && count <= 8' failed. | |||
| I find that quite encouraging | |||
| ohh, I pasted the subname wrong, it is called like in the actual call | 16:12 | ||
| I guess I have to use the thiscall calling convention here? | 16:17 | ||
| hmm no, seems not | 16:20 | ||
| "All mangled symbols begin with _Z (note that an underscore followed by a capital is a reserved identifier in C, so conflict with user identifiers is avoided); for nested names (including both namespaces and classes), this is followed by N, then a series of <length, id> pairs (the length being the length of the next identifier), and finally E." | 17:08 | ||
| that seems to agree with this example: _ZN14b2PolygonShape3SetEPK6b2Vec2i | 17:09 | ||
| sort of | |||
|
17:10
mj41 joined
|
|||
| FROGGS | _ZN 14b2PolygonShape 3Set E PK 6b2Vec2 i | 17:10 | |
| m: sub foo::bar() { 42 }; say foo::bar | 17:14 | ||
| camelia | rakudo-moar d9f58f: OUTPUT«Could not find symbol '&bar' in method <anon> at src/gen/m-CORE.setting:14609 in any find_method_fallback at src/gen/m-Metamodel.nqp:2810 in any find_method at src/gen/m-Metamodel.nqp:992 in block <unit> at /tmp/sKXHLm8vbH:1» | ||
| FROGGS | a very minimal case generates the symbol correctly: gist.github.com/FROGGS/7c8180851f3b89821631 | 17:39 | |
|
17:42
FROGGS[mobile] joined
|
|||
| timotimo | is that the gcc specific mangling or something? | 18:28 | |
|
18:40
FROGGS[mobile] joined
|
|||
| FROGGS | timotimo: en.wikipedia.org/wiki/Name_mangling..._functions | 18:47 | |
| must be g++ on an ubuntu machine I think | |||
| dalek | arVM/native-ref: 71fd576 | jnthn++ | src/ (3 files): Implement iscont_[ins]. |
18:48 | |
|
18:51
rurban__ joined
|
|||
| FROGGS | I guess in b2PolygonShape::Set(b2Vec2 const*, int), which is `thiscall`, I need to pass the "THIS" as the first arg? | 19:46 | |
| jnthn has really no idea, sorry :( | 19:48 | ||
| Also, wtf: "Spesh inline: unhandled arg instruction" | |||
| FROGGS | O.o | 19:49 | |
| jnthn | Seems it might be a constant named arg that triggers it. | 19:56 | |
| Ah, no | 20:09 | ||
| FROGGS | 0000000000000000 T _ZN7b2WorldC1ERK6b2Vec2 | 20:11 | |
| 0000000000000110 T _ZN7b2WorldD1Ev | |||
| the first is a constructor, the latter a destructor btw | |||
| ("C" vs "D") | |||
| jnthn | eek | 20:14 | |
| Such mangle... | |||
| FROGGS | I enjoy it :o) | ||
| dalek | arVM/native-ref: 679d06f | jnthn++ | src/spesh/inline.c: Add missing inline arg count check. We only collect information on up to a certain number of args. Without this, we try to resolve arg instructions beyond the end of the buffer, which innevitably works out rather badly. |
20:17 | |
| arVM/native-ref: b23351a | jnthn++ | src/spesh/inline.c: Improve error by including unhandled op. |
|||
| arVM/native-ref: da8fbed | jnthn++ | src/spesh/inline.c: Handle argconst_ instructions in inline. Fix order of args passed to lexical ref function. |
|||
| arVM/native-ref: d5ef005 | jnthn++ | src/6model/reprs/NativeRef.c: Implement by-index native lexical ref support. |
|||
|
21:19
Peter_R joined
|
|||
| FROGGS | ahh: | 21:20 | |
| my $gravity = OpaquePointer<140737326674832> | |||
| Program received signal SIGSEGV, Segmentation fault. | |||
| b2World::b2World (this=0x4d36840, gravity=...) at /build/buildd/box2d-2.3.0+ds/Box2D/Box2D/Dynamics/b2World.cpp:55 | |||
| even for a constructor I have to pass THIS | |||
| jnthn | ah :) | 21:22 | |
| FROGGS | [Inferior 1 (process 5976) exited normally] \o/ | ||
| THIS as in malloc'd to the right size (... I don't know how to calculate) | 21:23 | ||
| jnthn | Sounds like good progress :) | ||
| FROGGS | I mean, look at class World, I dunno how vtable and all these things are layed out in mem: box2d.googlecode.com/svn/tags/v2.3..../b2World.h | 21:24 | |
| jnthn | FROGGS: Can't help but wonder if the easier way would be to write a small wrapper that exposes a C-friendly API... | 21:29 | |
| Though, that would need compiling. | 21:30 | ||
| FROGGS | jnthn: that'd mean that every user of a C++ would have to do that | ||
| which... is not trivial | |||
| jnthn | Well, if they're using C++ they're already used to pain, at least... :P | ||
| FROGGS | :P | ||
| I see how far I get, if it proves to be too hard, I'll take another route | 21:31 | ||
| but atm, it is a nice project | |||
| jnthn is curious what other languages have a C++ FFI approach that doesn't need some kind of C wrapper | 21:34 | ||
| FROGGS | jnthn: python | 21:36 | |
| and Java probably | |||
| you'd probably know about C# | 21:37 | ||
| jnthn | FROGGS: search.cpan.org/~plicease/FFI-Platy...ang/CPP.pm may be of interest, fwiw | ||
| FROGGS | jnthn: very nice! | 21:39 | |
| "You need to alloate the memory needed for the object before you call the constructor[...]" aye :o) | 21:40 | ||
| ohh damn, it, I can call a method on my b2World instance! | 21:46 | ||
| timotimo | <3 | 21:47 | |
| FROGGS | this could be end up very sweet... like have native methods in a CStruct... | 21:49 | |
| so I am doing this: | 21:52 | ||
| # alloc $world | |||
| new_World($world, b2Vec2.new(x => 42e0, y => 42e0)); | |||
| World_Dump($world); | |||
| and I get in my shell: | |||
| b2Vec2 g(4.200000000000000e+01f, 4.200000000000000e+01f); | |||
| m_world->SetGravity(g); | |||
| b2Body** bodies = (b2Body**)b2Alloc(0 * sizeof(b2Body*)); | |||
| [...] | |||
| the vector I pass to the world's constructor is the gravity | 21:53 | ||
| to allocate the world I just fill a buf8 with a hundred elements :o) | |||
| for C++ I guess we'd want to have 'is native' and 'is symbol' on class definitions too, as well as on methods | 21:55 | ||
|
22:00
FROGGS[mobile] joined
|
|||
| jnthn | Not going to make it all the way to "is rw" on native params today, but got many (most, I hope) of the groundwork in place by now :) | 22:04 | |
|
22:04
FROGGS_ joined
|
|||
| TimToady | \o/ anyway :) | 22:05 | |
| timotimo | i'm liking it, too | ||
| TimToady likes to see the C++ progress too | |||
| it was really strange to hear that XS doesn't support C++, since the original XS was written specifically to call into C++ code | 22:06 | ||
| jnthn went through 3 different designs for integrating the native ref stuff with QAST::Var before finding one he liked. :) | |||
| TimToady | and part of why P5 supported OO was to ease the mismatch with C++ | ||
| timotimo | jnthn: i'd rather have you try multiple different things than pull through the first thing you try even if it's not 100% to you rliking | ||
| FROGGS_ | ohh, we already allow 'is native' on methods and we probably already pass self/this along | 22:10 | |
| jnthn | FROGGS_: I...uh...wow, that's a happy accident! o.O | 22:11 | |
| timotimo | methods are routines ... dunno if it know sabout the invocant? | 22:12 | |
| jnthn | (You're right, I think, it just wasn't actually a design consideration. :)) | ||
| Well, not a concious one | |||
| I wonder if we could have a is repr('CPPClass') or so | |||
| FROGGS_ | we probably need to | 22:13 | |
| timotimo | can the given mangling convention of a library be figured out somehow? perhaps just listing the available symbols in a library or so? | ||
| TimToady | that was part of why we invented repr poly in the first place... | ||
| FROGGS_ | star-m: use NativeCall; class Foo is repr<CStruct> { method malloc(int32) is native is symbol<malloc> { * } }; Foo.new.malloc(32) | 22:15 | |
| camelia | star-m 2015.01: OUTPUT«Unknown type Mu used in native call.If you want to pass a struct, be sure to use the CStruct representation.If you want to pass an array, be sure to use the CArray type. in sub type_code_for at /home/camelia/rakudo-star-2015.01/install/languages/perl…» | ||
| FROGGS_ | I could have bet that it failed differently :/ | ||
| like, showing that we pass self along | |||
| jnthn is off for the night | 22:20 | ||
| o/ | |||
| timotimo | good night jnthn! | ||
| FROGGS_ | gnight jnthn | 22:21 | |
| japhb | FROGGS_++ # Working on C++ support. | 22:30 | |
| Really, I can't even tell you how much that would improve things for me at $day-job, where lots of the deep infrastructure is written in C++, and SWIG is all over the place. | 22:32 | ||
| FROGGS_ | ohh, that sounds motivating :o) | 22:33 | |
| m: class Foo { method bar { } }; Foo.^find_method('bar').signature.params».type.say | |||
| camelia | rakudo-moar d9f58f: OUTPUT«(Foo) (Mu)» | ||
| FROGGS_ | what's that Mu? the implicit *%? | ||
| m: class Foo { method bar { } }; Foo.^find_method('bar').signature.params».name.say # :o) | 22:34 | ||
| camelia | rakudo-moar d9f58f: OUTPUT«Nil %_» | ||
| FROGGS_ | introspection ftw | ||
| japhb | *chuckle* Well, that's one way to find out. :-) | ||
| japhb <3 deep Perl 6 introspection | 22:35 | ||
| FROGGS_ | aye | ||
| :o) | |||
| timotimo | areed | 22:44 | |