Maybe soon. I am getting closer but not quite there yet. Send me a PM. Giving it a run on a network with multiple machines is the acid test and I don't have that.
alfaleader - I have it sorted I think.
For rgba data, the sender has to be specified as "NDIlib_FourCC_type_BGRA". Where I had it wrong was that the receiver can specify the format it "prefers". I took this to mean it might prefer that but if a sender is rgba it would get rgba data, when in fact it is converted to YUV from the rgba source. I am not sure how this works if different receivers prefer rgba or yuv. Anyway the type can be detected when the frame is received.
Then also the "progressive" flag comes into it. I am looking at all options for the type of video received. My findings are that the receiver has to prefer YUV for interlaced to be received properly. Prefer BGRA results in half the data, with each alternate line black.
As to the closed senders not being released from the sender list, I am finding that specifying zero timeout for NDIlib_find_get_sources always works OK and returns all the sources, even if one is closed, but specifying a timeout to detect any network change does not detect when a sender is closed, maybe the length of timeout is an issue here. I am profiling these functions to see what overhead there is in monitoring the network.
I get 16.5 msec for glGetTexImage at 1920x1080 like it is locked to vsync. With vsync off it is 10-11 msec independent of the frame rate specified.
Quick video to show it, i use vMix to capture... then NDI to spout -> get it in resolume
All is realtime with no delay what i can see.
Tested on WIN (sender) to WIN (receiver)
Tested also on OSX (sender) to WIN (receiver) not as good, windows the clear way to go
"Spout to NDI" is finished and seems pretty good according to tests by hive8. Now Nozomu Miura from Techlife has made a Syphon equivalent that is also ready for beta testing.
