02:03 <M9names[m]> <thejpster[m]> "Has anyone tried ELD as a linker..." <- I gave it a go.... (full message at <https://catircservices.org/_irc/v1/media/download/AdxYpcO9vPQsvoKaU2vJ8LPOf93iD4_ooO1Sdd2IKP1eiq_IPLlcB6EwBCBwc54PcT7Q3eteRxniprkhsHGxYZK_8AAAAAAAAGNhdGlyY3NlcnZpY2VzLm9yZy9HdE5uVWRYelFuaXZOdGNiVXFKVEVCWnk>)

02:08 <JamesMunns[m]> I audibly laughed at it not liking what defmt does.

02:08 <JamesMunns[m]> I would love to see the bug report for that and what they think of those shenanigans

02:10 <i509vcb[m]> theres a part of me wondering if the linker script is generally turing complete

02:10 <i509vcb[m]> Like generating an entire program in just the linker

02:35 <M9names[m]> i take it back, it's only one line of the defmt linker script it doesn't like.... (full message at <https://catircservices.org/_irc/v1/media/download/Acp0cYn8o51dqn9StdxiaqYmFR21vGqsY0qXPEFwwcQKGdcr1vJPoqYZnCaKfIJxnr0vbJQdYmjL-2KqmpR8z9O_8AAAAAAAAGNhdGlyY3NlcnZpY2VzLm9yZy9sQ2hweXRUUE1nY3B6Y1ZERkZNcXpiWGg>)

02:36 <i509vcb[m]> Do we need to RiiR the linker lol

02:36 <M9names[m]> you mean like this? https://github.com/davidlattimore/wild

07:48 <thejpster[m]> I recall pietro is writing a linker too

09:46 <thejpster[m]> I just had the most terrible idea.... (full message at <https://catircservices.org/_irc/v1/media/download/AaPFtvgjF0QBHfQe0Cr1jBroNflvaIDxipB6IIVCWS2NdaJXy_dVGV86DsEodTrrknoGYsTcwoSCSFwswktZXqi_8AAAAAAAAGNhdGlyY3NlcnZpY2VzLm9yZy9kaWJGd0xIZVp5UGVWdGtCRmhITGh1aWk>)

09:47 <diondokter[m]> This is indeed terrible, but I like it!

09:49 <thejpster[m]> To get two binaries to share one RTT buffer just needs a fork of defmt-rtt with a multi-core spinlock and the ability to pass in the RTT buffer address rather than statically allocating it. But that’s ok - I need to fork it anyway so I can specify that the buffer lives in non-cached memory (according to the MPU).

09:51 <thejpster[m]> Or I guess a defmt-ipc that throws data from R to M over a completely separate queue, which M then copies to the RTT buffer. At least then M can still do logging even if R crashes. And actually it should log if R crashes. That’s kind of the point of the M core.

09:51 <thejpster[m]> Anyway, now I need to see if I can run code on the M core and if RTT works.

11:15 <JamesMunns[m]> > defmt-ipc that throws data from R to M over a completely separate queue

11:15 <JamesMunns[m]> If you want, I can hack you up an IPC safe version of bbqueue, as long as atomics are properly coherent between the R and M cores

11:18 <JamesMunns[m]> The one thing that I'd need you to impl is a "notification" trait impl, so that the R core can signal the M core "hey you have pending data", probably using some kind of mailbox peripheral or swi?

11:19 <JamesMunns[m]> https://docs.rs/bbq2/latest/bbq2/traits/notifier/trait.Notifier.html, and optionally https://docs.rs/bbq2/latest/bbq2/traits/notifier/trait.AsyncNotifier.html if you want to be able to async await things.

11:24 <JamesMunns[m]> (mnemos has a version of bbqueue that is/was intended to be ABI safe, and has manual "construct a Sender/Receiver from a base ptr address" methods, https://github.com/tosc-rs/mnemos/tree/main/source/abi/src/bbqueue_ipc, I've meant to forward port this to bbq2, but haven't needed it yet, just waiting for an excuse 😃 )

11:26 <JamesMunns[m]> @thejpster are both cores 32 bit, or is the R 64-bit?

11:27 <RobinMueller[m]> What is the easiest way to test my defmt::Format impl on a host PC? for example, write the defmt formatted structure to a string again for example printouts?

11:55 <thejpster[m]> It’s a four R5s and one M4 and I don’t know if they are atomically coherent

11:56 <thejpster[m]> <RobinMueller[m]> "What is the easiest way to..." <- The repo has tests. I’d do what that does.

11:57 <thejpster[m]> The AM243x-Launchpad is worth buying at $90, if you can live with openocd.

12:49 <JamesMunns[m]> I just ordered a stack of other hardware, having async on the cortex-m would actually be interesting, because you could `select` on four async bbqs to feed the m's rtt buffer

12:49 <JamesMunns[m]> I just ordered a stack of other hardware, so I'm probably going to pass on the launchpad.

12:49 <JamesMunns[m]> Having async on the cortex-m would actually be interesting, because you could `select` on four async bbqs to feed the m's rtt buffer, if you have to service four other cores

21:39 <thejpster[m]> You could do that without async too? Just WFI then check each queue.

23:40 <JamesMunns[m]> You could certainly do anything async or not! It's useful if your m core has other things to do too.