Launching RetroArch with --host delays controller detection

I’m finding when RetroArch is launched with --host from the command line, the Controller detection is delayed for 60 seconds.

I’m using the latest nightly build (july 7, 2017) for Windows x64, running on Windows 10. I have an XBOX360 wireless controller with an adapter.

When I launch RetroArch without the --host option, the controller is detected immediately and is responsive.

When I launch RetroArch with the --host option, the controller is not responsive, and after 60 seconds, the controller is detected and the yellow “100% xinput controller (User 1) configured on port #0” is displayed.

Is launching with --host still a valid/recommended thing to do?

that’s very strange. When you say “60 seconds,” do you mean actual 60 seconds on the nose? Or “a long time that feels like 60 seconds”?

if it’s exactly 60 seconds, it should be easy to find whatever’s doing it, lol. Just search the source tree for “3600” i.e., 60 seconds at 60 fps.

Using a stopwatch as a quick test, it clocks in at about 58 seconds. I don’t think I can say precisely 60 seconds. I’ll check out the source and see if ‘3600’ comes up anywhere.

Edit: 3600 does come up in net_natt.c as the leaseDuration in UPNP_AddAnyPortMapping() and UPNP_AddPortMapping(). Since it may not be exactly 60 seconds (and I also dont know how these functions behave anyway), i’d be hesitant to point at these. It also comes up in civetweb.c.

I think the best I can say for now is if you have Windows, give it a shot. Plug a USB control in and launch with --host, --libretro, a core and a rom.