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)
> => #
> irb(main):003:0> v2 = v1.class.new(4,5,6)
> => #
> 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:
>
> type='{_vec3_t="x"f"y"f"z"f}'/>
>
>
>
>
>
>
>
>
>
>
> 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
> [email protected]
> http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel
smime.p7s
Description: S/MIME cryptographic signature
___
MacRuby-devel mailing list
[email protected]
http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel
