15:07 <_whitenotifier-3> [nmigen] Success. 85.87% (+5.11%) compared to 4310254 - https://codecov.io/gh/m-labs/nmigen/compare/43102541032f0bb51775bc1756fd125b984d6de8...1fedc1d140cfb600a39a13871217b1f6a844f6c1

15:07 <_whitenotifier-3> [nmigen] Success. Coverage not affected when comparing 4310254...1fedc1d - https://codecov.io/gh/m-labs/nmigen/compare/43102541032f0bb51775bc1756fd125b984d6de8...1fedc1d140cfb600a39a13871217b1f6a844f6c1

16:22 <whitequark> rjo: so, artiq has the io traits.

16:24 <whitequark> in libio. and the implementations, awkwardly, in sched.rs

16:24 <whitequark> but i think now that rust has native async support, that's what should be used instead

16:24 <whitequark> the whole artiq scheduler is a hack working around lack of nice async

16:27 <whitequark> rjo: regarding peek. i don't feel comfortable providing you with two slices. that binds me hard to one specific internal representation.

16:27 <whitequark> but i think i can give you a slice of slices.

16:28 <whitequark> it's not currently possible to allocate a variable length slice in stable rust, but it will be, soon

16:37 <whitequark> rjo: regarding DMA. i don't think that's possible.

16:41 <whitequark> poll() takes a &mut SocketSet. if your code is running when the handler fires (which is very likely) you are probably borrowing a socket

16:42 <whitequark> the ownership on that is kinda sketchy anyhow, the main problem being that access to any socket excludes all others

16:42 <whitequark> so it might be possible to enable the DMA interrupt iff you have released the socket set

16:43 <whitequark> that'd be easy. i can make an example for it, i think. but the ergonomics of using smoltcp without something like artiq's sched layer are unfortunate.