05:07 <mtrbot-ml_> [mattermost] <sb10q> thanks

08:36 <mtrbot-ml_> [mattermost] <sb10q> @astro what is the purpose of OCM3 in link.x?

08:41 <mtrbot-ml_> [mattermost] <sb10q> mh ok sometimes zynq splits the upper 64KB of memory and puts it a different address, because it's zynq

08:41 <mtrbot-ml_> [mattermost] <sb10q> mh ok sometimes zynq splits the upper 64KB of memory and puts it at a different address, because it's zynq

12:55 <mtrbot-ml_> [mattermost] <sb10q> @astro have you considered splitting the async TcpStream into tx and rx halves? that would be consistent with sync_channel, and also make the moninj implementation simpler and cleaner

12:56 <mtrbot-ml_> [mattermost] <sb10q> @astro but that's just brainstorming; if that's difficult or creating any mess, don't do it

16:01 <mtrbot-ml_> [mattermost] <astro> @sb10q the OCM splits are configurable but only once our code is loaded and running

16:01 <mtrbot-ml_> [mattermost] <astro> @sb10q I could give the split a try, there's already ref counting for sockets in smoltcp. yet in my (tokio) experience passing tx/rx off into separate program parts turned out to be cumbersome because you have to share data between them over another sync_channel. alternating between send() and recv() in an async fn that owns the socket is much more convenient.

19:29 <mtrbot-ml_> [mattermost] <astro> libdyld loads elf on arm \o/

19:30 <mtrbot-ml_> [mattermost] <astro> now to the refactoring!

23:02 <mtrbot-ml_> [mattermost] <sb10q> I tried reconfiguring them with openocd, I systematically get "data abort" errors afterwards

23:02 <mtrbot-ml_> [mattermost] <sb10q> Bah. It's probably faster that I write a sdram loader than mess around with those OCM settings

23:03 <mtrbot-ml_> [mattermost] <sb10q> Cool!

23:04 <mtrbot-ml_> [mattermost] <sb10q> What I wanted to do was use send and recv on the same socket in separate tasks. Including sending while another task is blocked on recv

23:59 <mtrbot-ml_> [mattermost] <astro> do you prefer a tx/rx split, or would a `impl Clone for TcpStream` with ref counting work even better for us?

23:59 <mtrbot-ml_> [mattermost] <astro> smoltcp tcp sockets already have ref counter we can use