07:10 <re_irc> <korken89> On the topic of "uninitialized", one this I should add to heapless (and maybe fix for James Munns bbqueue) is to allow to construct "const methods" directly on "MaybeUninit<Container>", like a "new_in_place(&mut MaybeUninit<Container>) -> &mut Containter". Turns out the compiler is not good at optimizing away the copy of a "MaybeUninit::write" if you have non-0 starting values.

07:10 <re_irc> <korken89> * thing

07:11 <re_irc> <korken89> It will happily place the container on the stack, and then copy it in - instead of constructing it in place

07:11 <re_irc> <korken89> Neeeeooooowwww says your stack space

11:53 <re_irc> help: consider further restricting this bound

11:53 <re_irc> |

11:53 <re_irc> <barafael> I'm updating a crate from embedded time 0.10 to 0.12. Compiler tells me at some point:

11:53 <re_irc> 31 | Clock: embedded_time::clock::Clock + embedded_time::clock::Clock,

11:53 <re_irc> | +++++++++++++++++++++++++++++

11:54 <re_irc> <barafael> that makes no sense :D if I do this, compiler asks for one more :D

11:54 <re_irc> <K900> You probably have a mix of versions

11:54 <re_irc> <K900> And those traits are from different versions

11:54 <re_irc> <barafael> what I want to do is "Delay::new(self.clock).delay(Milliseconds(200u32));" where "clock" is of type "Clock"

11:56 <re_irc> <chrysn (@chrysn:matrix.org)> When debugging a mix of versions, "cargo tree -d" is often helpful.

15:30 <re_irc> <dirbaio> korken89: "uninit.as_mut_ptr().write(Container::new_from_whatever())" usually does optimize well

15:31 <re_irc> <korken89> Nice, the next time I'm fighting with that I'll give it a try!

15:32 <re_irc> <dirbaio> constructing the thing then writing it to its final place with a raw ptr write usually optimizes well

15:32 <re_irc> <James Munns> korken89: I need to read back context, but bbqueue statics live totally in bss, at the cost of initialize-on-split

15:33 <re_irc> <dirbaio> only if there are no conditional branches or side effects between both :S

15:33 <re_irc> <James Munns> I also have a hack to do this for heapless mpmc queue in byte slab

15:34 <re_irc> <dirbaio> because the compiler can't let other code see the final memory location in a "half-written" state :(

15:34 <re_irc> <dirbaio> and this includes drops and panics

15:35 <re_irc> <James Munns> https://github.com/anachro-rs/powerbus/blob/main/byte-slab/src/byte_slab.rs#L142-L156

15:37 <re_irc> <James Munns> (And https://github.com/jamesmunns/bbqueue/blob/main/core/src/bbbuffer.rs#L245-L282)

15:40 <re_irc> <James Munns> I start the struct in a weird-but-valid-state where all tracking vars are zero, in order to guarantee the whole struct is zero or uninit. Otherwise that single u32 “infects” the whole struct, which means a 64KiB struct would live in .data for ONE WORD of initialized data.

15:42 <re_irc> <James Munns> This does require atomic ops or a CS to handle safely though.

15:42 <re_irc> <James Munns> (bbqueue uses atomic polyfill for this, byteslab should but I havent switched it over yet.

15:44 <re_irc> <dirbaio> oh yeah I hate the .data bloat

15:46 <re_irc> <James Munns> Thats why I had to work around mpmc queues init

15:47 <re_irc> <James Munns> My slab allocator was ending up in .data because I use mpmc queue as a freelist

15:47 <re_irc> <James Munns> Which was bad, for a large allocator pool (slow to flash, waste of flash space)

17:18 <re_irc> <bradleyharden> Maybe someone here will know. A few years ago, I read a post about how to judge/rank APIs. The scale was essentially "impossible to use incorrectly" to "impossible to use correctly". I tried to dig up that post, but I can't find it at all. Does anyone know what I'm talking about?

17:18 <re_irc> If I remember correctly, it was a very bare bones website, and I think the API examples were in C.

17:49 <re_irc> <yruama_lairba> bradleyharden: there wasn't advice to make a good API ?

17:50 <re_irc> <bradleyharden> I don't remember. Maybe? I think it was mostly focused on showing examples from several different APIs

17:50 <re_irc> <bradleyharden> Both good and bad examples

17:53 <re_irc> <yruama_lairba> examples are already good indications

17:55 <re_irc> <yruama_lairba> but i guess Rust help greatly to prevent API missuse

17:57 <re_irc> <bradleyharden> I'm just trying to dig up that article. I've searched everything I can think of, but I can't find it. I was hoping someone here might know what I'm talking about

18:58 <re_irc> <thejpster> Can we do anything about the github teams spam?

19:09 <re_irc> <adamgreig> any ideas? it all gets spam filtered for me

20:21 <re_irc> <9names (@9names:matrix.org)> bradleyharden: Sounds like Rusty Russell's API ranking.

20:21 <re_irc> https://ozlabs.org/~rusty/index.cgi/2008/03/30#2008-03-30

20:23 <re_irc> <9names (@9names:matrix.org)> There's also the complete list with negative values:

20:23 <re_irc> http://sweng.the-davies.net/Home/rustys-api-design-manifesto

20:24 <re_irc> https://ozlabs.org/~rusty/index.cgi/2008/03/30#2008-03-30

20:24 <re_irc> <9names (@9names:matrix.org)> Sounds like Rusty Russell's API ranking.

20:24 <re_irc> https://ozlabs.org/~rusty/index.cgi/tech/2008-04-01.html

22:03 <re_irc> <justacec> This is a bit off topic, but I think that a few of you guys might know... So, I am pretty confused as to what the LoRa-E5 is. It says that it has a STM32WLE5JC MCU as well as a SX126X? But the STM32WLE5JC has a built-in radio right? Or is it that ST just merged a SX126X into their chip and are calling that the radio... Or did SeedStudio actually combine these two components. So confused... Does anybody have any...

22:03 <re_irc> ... experience with this?

22:04 <re_irc> <justacec> Here is a link to the item: https://www.seeedstudio.com/LoRa-E5-Wireless-Module-p-4745.html

22:16 <re_irc> <justacec> Ok, after sending my message a few minutes ago, I stumbled across this Gem: https://forum.seeedstudio.com/t/lora-e5-does-not-contain-a-semtech-sx1262/262389/15. So that kind of answers the question. The SX126X is imbedded in the STM32WLE5JC.

22:20 <Lumpio-> Wonder what kind of "legal action" they were going to take

22:31 <re_irc> <firefrommoonlight> > Or is it that ST just merged a SX126X into their chip and are calling that the radio.

22:31 <re_irc> > Yep

22:31 <re_irc> <firefrommoonlight> > Or is it that ST just merged a SX126X into their chip and are calling that the radio.

22:31 <re_irc> > Yep

22:31 <re_irc> <firefrommoonlight> ->

23:43 <re_irc> <yruama_lairba> hi, hi have a small refactoring issue with rtic. i want to put inside a lib functions taking a shared object, put unfortunatly, type of the shared object is known outside the app module.

23:43 <re_irc> <yruama_lairba> how to workaround that ?

23:59 <re_irc> <henrik_alser> yruama_lairba: https://github.com/rtic-rs/cortex-m-rtic/blob/master/examples/extern_binds.rs