Laurent Sansonetti
lsans****@apple*****
Mon Jun 11 01:45:28 JST 2007
Hi Kimura-san, On Jun 10, 2007, at 8:03 AM, kimura wataru wrote: > Great! > > I have a few questions. > - It seems the method cache mechanism is disable at framework/src/ > objc/mdl_objwrapper.m. > Is the mechanism disable by default under the latest revision? > I changed the value of OCM_AUTO_REGISTER to "1", then got a good > result > from Laurent's bench.rb. Oops, I forgot to set it to "1" :-) I will commit it as enabled by default. (The #define was just for benchmarking.) > - I'm not clear about ffi. I wonder the > method cache mechanism ignores > overriding methods by ObjC categories. ObjC runtime flushes its > cache against > some occasions, such as loading a bundle. This mechanism need not > flush > like ObjC runtime? Good question, and normally everything should be fine as before. The fact is that methods can be added via categories (when loading a bundle), but not removed. What this new change does is just to define the Ruby method, but it still calls ocm_send (from a FFI closure) which will at runtime retrieve the ObjC implementation (IMP) and call it. So even if the ObjC method implementation changes, it will call the new one. Laurent > On Fri, 8 Jun 2007 00:24:11 -0700, Laurent Sansonetti wrote: >> I committed a working experiment as r1821! >> >> It automatically defines the method directly in C by generating >> Libffi >> closures. >> > > -- > kimura wataru > > _______________________________________________ > Rubycocoa-devel mailing list > Rubyc****@lists***** > http://lists.sourceforge.jp/mailman/listinfo/rubycocoa-devel
