Hi Fjölnir, As I think I mentioned in IRC, my best guess is that this is a bridge support bug. Have you filed a bug with Apple at all? Something else you might try (if you have the time) is to access your struct using the pyobjc bridge, which also makes use of bridge support, to see if this is MacRuby specific. At any rate, please also file a ticket for the MacRuby project so we don't loose track of this. If nothing else we'll be able to mark it as a third-party issue. Cheers, Josh On Tuesday, March 13, 2012 at 8:41 PM, Fjölnir Ásgeirsson wrote:
One thing I forgot to mention: If I switch the struct to use doubles, then everything works fine (which is probably why this didn't get caught sooner, since the appkit types all use doubles)
On 2012/03/13, at 22:06, Fjölnir Ásgeirsson wrote:
I've been struggling with accessing C structs from ruby. The problem is:
I expose a struct from my framework (definition below). However when accessing/allocating it from ruby it seems the offsets are off by 4 bytes (as if macruby think it's supposed to be a double)
Definition:
#ifndef __SCRIPTINGBRIDGE__ union _vec3_t { float f[3]; struct { float x; float y; float z; }; struct { float r; float g; float b; }; struct { vec2_t xy; float andY; }; }; typedef union _vec3_t vec3_t; #else typedef struct _vec3_t { float x; float y; float z; } vec3_t; // Bridgesupport freaks out when given the above union so I show it a struct with the same layout #endif
Example output:
irb(main):001:0> framework "./GLMath.framework" => true irb(main):002:0> v1 = vec3_create(1,2,3) => #<vec3_t x=1.0 y=3.0 z=NaN> irb(main):003:0> v2 = v1.class.new(4,5,6) => #<vec3_t x=4.0 y=5.0 z=6.0> irb(main):004:0> printVec3(v1) 0x676f1310 - 0x676f1314 - 0x676f1318 Vec3: [1.00, 0.00, 3.00] => nil irb(main):005:0> printVec3(v2) 0x676f1310 - 0x676f1314 - 0x676f1318 Vec3: [4.00, 0.00, 5.00] => nil
in the above example, vec3_create & printVec3 are C functions, shared using the scripting bridge. So when the struct is allocated from the c side, ruby only sees the first & last items, and the last item is on the offset of the second. when I then pass it to printVec3, it gets the first and last items only (last item taken from the place of the second item, => offsets are double what they should be). then when I allocate it using FFI in ruby, it's correcly read from the ruby side, but when I pass it to a c function, it gets the first item in the correct place, then the second item in the place of the last.
The type definitions in my bridge support file are:
<struct name='vec3_t' type='{_vec3_t="x"f"y"f"z"f}'/> <function name='vec3_create' inline='true'> <arg type='f'/> <arg type='f'/> <arg type='f'/> <retval type='{_vec3_t=fff}'/> </function> <function name='printVec3'> <arg type='{_vec3_t=fff}'/> </function>
Right now I'm working around this by wrapping everything in an object on the objective-c side, but that is very slow and bloated.
Has anyone ever gotten structs to work properly or know what is causing this?
Note: if I return a pointer to the struct, and then assign to it's address using a boxed struct created from ruby, that works fine it seems. It's only if I return a struct directly. (However, myPointer[0].x = foo; would not work. I'd have to do a=myPointer[0];a.x=foo;myPointer[0]=a)
– Fjölnir
_______________________________________________ MacRuby-devel mailing list MacRuby-devel@lists.macosforge.org (mailto:MacRuby-devel@lists.macosforge.org) http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel
_______________________________________________ MacRuby-devel mailing list MacRuby-devel@lists.macosforge.org (mailto:MacRuby-devel@lists.macosforge.org) http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel
Attachments: - smime.p7s