Hardy Griech

I stumbled into the same thing, see #2701, gcc compilations. For clang (19.1.5 and 21.1.1) this works, especially if you take my branch from MR #2700.

Hmmm... not getting the point. Do you think the problem is in the NCM driver or in the glue code? Perhaps my own glue code could be of help?: https://github.com/rgrr/yapicoprobe/blob/master/src/net/net_glue.c

ok... understood. Good point. Loop above seems to be too complicated for my eyes. Is something like this working: ```C void tud_network_recv_renew(void) { TU_LOG_DRV("tud_network_recv_renew()\n"); static bool should_process = false; //...

Hmmm... you are right (perhaps I need some time to rethink it ;-)) Perhaps a simple semaphore could do?: ```C void tud_network_recv_renew(void) { static bool sema = false; if (...

you are right (again)... perhaps it could be solved by letting recv_transfer_datagram_to_glue_logic() return an indication that some work has been done (or that there is some work still waiting)?

got that, but to be honest I'm not happy with two variables for this logic.

Could you please check PR #2713

perhaps @konkers knows best?