08:06 <re_irc> <raphi ⚡️> Hello, I would like to use rust for an other platform (C-Based). I compile rust into a c-library and include this libray on the other platform. For i386 Architecture, everything works fine. But it should compile to arm7 too. The C-Based platform compiles with gcc 6.3.0 and i always has the error, that eabi version 5 isn't supportet. I tried every rust arm-target and also try to make an own arm-target. No look...

08:07 <re_irc> ... yet. Does everyone know, how to change the eabi-version on arm-targets?

08:08 <re_irc> < (@k900:0upti.me)> Uhh, what do you mean by a "C library"?

08:08 <re_irc> < (@k900:0upti.me)> Are you compiling Rust to C source or something?

08:09 <re_irc> <raphi ⚡️> crate_type = ["staticlib"] and using #[no_mangle extern "C" to define the function-interface

08:09 <re_irc> < (@grantm11235:matrix.org)> Or compiling to a static library and calling it from a C program?

08:10 <re_irc> < (@grantm11235:matrix.org)> ha, you beat me

08:12 <re_irc> < (@k900:0upti.me)> https://clang.llvm.org/docs/ClangCommandLineReference.html#cmdoption-clang-meabi

08:12 <re_irc> < (@k900:0upti.me)> Clang can produce EABI4

08:12 <re_irc> < (@k900:0upti.me)> So I'm assuming LLVM can too

08:12 <re_irc> < (@k900:0upti.me)> But you might need to dive pretty deep into rustc to make it do that

08:18 <re_irc> <raphi ⚡️> there is the "-Cllvm-args" rustflag (with arount 2000 options), but i didn't find a eabi setting. Or is there a way to direct define an llvm option somehow?

08:18 <re_irc> <raphi ⚡️> "-Cllvm-args=-meabi=4" won't work:

08:18 <re_irc> rustc -Cllvm-args="..." with: Unknown command line argument '-meabi=4'. Try: 'rustc -Cllvm-args="..." with --help'

08:28 <re_irc> < (@9names:matrix.org)> hmm, shouldn't it choose the correct ABI if you specify the target cpu correctly?

08:28 <re_irc> what does https://github.com/ryankurte/rust-gba/blob/master/arm7tdmi.json produce?

08:30 <re_irc> < (@grantm11235:matrix.org)> raphi ⚡️: It looks like rustc only allows you to pass certain options to llvm https://github.com/rust-lang/rust/issues/26338

08:32 <re_irc> < (@grantm11235:matrix.org)> You aren't the first person to have this problem https://github.com/rust-embedded/wg/issues/379

08:32 <re_irc> <raphi ⚡️> > hmm, shouldn't it choose the correct ABI if you specify the target cpu correctly?

08:32 <re_irc> > what does https://github.com/ryankurte/rust-gba/blob/master/arm7tdmi.json produce?

08:32 <re_irc> ELF Header:

08:32 <re_irc> ... long message truncated: https://psion.agg.io/_matrix/media/r0/download/psion.agg.io/6jiSqwUNz2zOiPK5j1VdfFKZaPFO174C (23 lines)

08:32 <re_irc> Still Version5 EABI => it's also compile to OS/ABI = UNIX (should be plain ARM)

08:33 <re_irc> < (@9names:matrix.org)> fascinating

08:43 <re_irc> <raphi ⚡️> > You aren't the first person to have this problem https://github.com/rust-embedded/wg/issues/379

08:43 <re_irc> Thank's. i left a comment there. Actually i programm same Code in Rust und C++ - just because this issue :(

08:43 <re_irc> But it's not seem to be an easy solution to fix that

08:44 <re_irc> <raphi ⚡️> > fascinating

08:44 <re_irc> i'm compiling with nightly (don't no if this make a difference)

08:54 <re_irc> < (@grantm11235:matrix.org)> The documentation for the "--meabi" parameter says that if it isn't set, it will use the default for the triple. So there must be some way to set it in a custom triple

09:03 <re_irc> < (@grantm11235:matrix.org)> Can you set the eabi to version 5 in gcc?

09:08 <re_irc> < (@grantm11235:matrix.org)> Later versions of gcc have an "-meabi" option, but I don't see it in the 6.3 manual

09:18 <re_irc> < (@osten_music:matrix.org)> Heya, I'm looking for some advice. I have circular DMA transmitting and updating a DAC, basically as fast as it possibly can, via SPI. I would like to limit the speed in which this operates, in the context of seconds, rather than ticks.

09:18 <re_irc> Basically I need to have a determined and accurate sample rate, but this method of circular DMA its hard for me to know exactly what that sample rate is, and if it deviates or fluctuates.

09:18 <re_irc> I've tried using a secondary hardware timer, updating periodically setting a flag, e.g. at 10kHz, but having these two interrupts causes a dramatic slowdown, so I would like to somehow limit it within the DMA context directly.

09:18 <re_irc> I can measure the time difference between transmit complete interrupts, getting a timestamp from the RTC, but from my attempts this is kind of unreliable, maybe the resolution of the RTC is too low.

09:19 <re_irc> ... long message truncated: https://psion.agg.io/_matrix/media/r0/download/psion.agg.io/iePHbiEVtbD9nK5yNuUz2bAgZ4xLo2Zw (7 lines)

09:23 <Darius> can you drive the DMA from a timer channel or similar?

09:23 <Darius> what MCU?

09:23 <re_irc> < (@osten_music:matrix.org)> stm32h743vit6

09:25 <re_irc> < (@osten_music:matrix.org)> I could use a hardware timer to generate the information for the DAC, pushing it onto a buffer thats then passed to the circular dma routine. Its possible, but I think its not the fastest way of doing it

09:25 <re_irc> < (@osten_music:matrix.org)> it causes a slowdown

09:25 <re_irc> <raphi ⚡️> > Later versions of gcc have an "-meabi" option, but I don't see it in the 6.3 manual

09:25 <re_irc> i can, but there are other precompiled libraries compile with version 0. So than, there will be a conflict :/

09:26 <re_irc> < (@osten_music:matrix.org)> maybe counting the number of bits being sent, and then calculating the transmitting frequency? I'm not sure exactly what to include in that equation though

09:27 <re_irc> <raphi ⚡️> > The documentation for the "--meabi" parameter says that if it isn't set, it will use the default for the triple. So there must be some way to set it in a custom triple

09:27 <re_irc> i thougth, the os oder 'none' of armv7a-none-eabi should set the correct eabi. But it wont' change anything

09:28 <Darius> osten: I would have thought you care about jitter and less so about raw performance

09:28 <re_irc> < (@osten_music:matrix.org)> yes thats true, it needs to be stable

09:29 <re_irc> < (@osten_music:matrix.org)> so then maybe the hardware timer + circular DMA is the best I can do then

09:29 <re_irc> < (@osten_music:matrix.org)> I wonder how the I2S drivers manage it

09:30 <re_irc> < (@osten_music:matrix.org)> When you configure it I2S to run at eg. 44100, will there be gitter?

09:30 <re_irc> < (@osten_music:matrix.org)> * jitter?

09:30 <Darius> I haven't used it but I assume you setup a clock and divisor and feed it samples and it does the pacing

09:31 <re_irc> <raphi ⚡️> https://github.com/llvm/llvm-project/issues/27212

09:31 <re_irc> look's like a bug of clang?

09:33 <Darius> https://deepbluembedded.com/stm32-dac-sine-wave-generation-stm32-dac-dma-timer-example is in C, but looks like what you want

11:29 <re_irc> < (@justacec:matrix.org)> Thought I would check here to see where I might hit up a chat channel for the rp2040 PIO. I am curious how to define more than 1 program in the asm file, compile it using the pio_proc! macros and then instruct each state machine to use specific programs. My reading of the documentation in the pio_proc and the pio_rs, and the rp2040_hal have not fruited me much on this. I have seen examples where the...

11:29 <re_irc> ... same program was used on each state machine in the sync examples, but that is not quite what I was looking for.

11:40 <re_irc> < (@dirbaio:matrix.org)> #rp-rs:matrix.org (https://matrix.to/#/#rp-rs:matrix.org)

12:44 <re_irc> < (@justacec:matrix.org)> :). Thanks

13:33 <re_irc> < (@chaosprint:matrix.org)> https://psion.agg.io/_matrix/media/r0/download/matrix.org/tEUqOuKceDXxqDIVTEyhDVMP/image.png

13:33 <re_irc> < (@chaosprint:matrix.org)> holy moly

13:36 <re_irc> < (@diondokter:matrix.org)> Does anybody know of crate that bridges "log" + Rust's assert to "defmt"? Seems like a lot of people roll one in their own project if using defmt is optional

14:14 <re_irc> < (@alex:matrix.sempto.net)> I'm unable to get defmt-rtt to work. Probe-run works and outputs the correct place in the program if I stop it with Ctrl + C, but I do not see any output from "defmt::info!("init");"

14:14 <re_irc> My target is pi-pico the debugger is pi-pico with dapper mime and I'm using RTIC if it matters.

14:15 <re_irc> < (@diondokter:matrix.org)> Try to see if "defmt::println!" works. If it does, then your setup works

14:15 <re_irc> < (@diondokter:matrix.org)> You may not be able to see the info because you haven't set the "DEFMT_LOG" environment variable

14:15 <re_irc> < (@diondokter:matrix.org)> :

14:18 <re_irc> < (@alex:matrix.sempto.net)> defmt::println! works

14:19 <re_irc> < (@alex:matrix.sempto.net)> And yes, turns out the problem was unset DEFMT_LOG

14:19 <re_irc> < (@alex:matrix.sempto.net)> Thank you!

14:20 <re_irc> < (@diondokter:matrix.org)> Np

20:16 <re_irc> < (@allexoll:matrix.org)> Hi everyone, is there some reason odd_number % 2 == 0? i'm running into some weird behaviour in my DMA implementation, when i'm checking if the buffers are aligned. if a u16 buffer is not aligned, it's giving me some weird results:

20:16 <re_irc> defmt::println!("src_ptr: 0x{:x}", src_ptr as usize);

20:16 <re_irc> defmt::println!("sizeof W: {:x}", core::mem::size_of::<W>());

20:16 <re_irc> defmt::println!("alignement src: {:b}", src_ptr as usize % core::mem::size_of::<W>());

20:16 <re_irc> ... long message truncated: https://psion.agg.io/_matrix/media/r0/download/psion.agg.io/jiuTuYQHQkpHdXVbzTAuMXPHQiSECgr3 (7 lines)

20:16 <re_irc> < (@allexoll:matrix.org)> to me it seems that ...19 % 2 would equal 1, not 0 ?!

20:16 <re_irc> < (@allexoll:matrix.org)> (where "W" is u16)

20:20 <re_irc> < (@allexoll:matrix.org)> and "src_ptr" is "*const W". is there a impossibility to misalign types in rust? I don't understand how an odd number % 2 would result in zero 😕

20:22 <re_irc> < (@grantm11235:matrix.org)> My guess is that there is some Undefined Behavior going on

20:30 <re_irc> < (@grantm11235:matrix.org)> If you post your code, I can take a look at it

20:32 <re_irc> < (@allexoll:matrix.org)> it seems to only happen in --release mode, so UB might be possible

20:33 <re_irc> < (@allexoll:matrix.org)> https://github.com/allexoll/lpc546xx-hal-dma-alignement

20:35 <re_irc> < (@allexoll:matrix.org)> https://github.com/allexoll/lpc546xx-hal-dma-alignement

20:35 <re_irc> < (@allexoll:matrix.org)> relevant part is the "copy_exact" function in src/dma.rs when called from "dma_copy_u16_1024_misaligned_src" in "self-tests/dma-m2m.rs"

20:39 <re_irc> < (@grantm11235:matrix.org)> This line is bad https://github.com/allexoll/lpc546xx-hal-dma-alignement/blob/dma/self-tests/tests/dma-m2m.rs#L796

20:39 <re_irc> < (@allexoll:matrix.org)> ah I think the issue is with the creation of the unaligned slice with "from_raw_parts_mut" :

20:40 <re_irc> "data must be non-null and aligned even for zero-length slices. One reason for this is that enum layout optimizations may rely on references (including slices of any length) being aligned and non-null to distinguish them from other data"

20:40 <re_irc> < (@allexoll:matrix.org)> yes

20:40 <re_irc> < (@allexoll:matrix.org)> new question then, anybody knows of a way to create an unaligned u16 slice? if that's not possible at all I'm fine with it too because that means that my test case is irrelevant as well!

20:41 <re_irc> < (@grantm11235:matrix.org)> A "&mut [u16]" is always aligned

20:41 <re_irc> < (@allexoll:matrix.org)> ahhhh thanks, that's amazing

20:42 <re_irc> < (@allexoll:matrix.org)> there goes my issue

20:43 <re_irc> < (@allexoll:matrix.org)> Is that specific to rust or specific to llvm? I'm pretty sure some architecture allow for unaligned access?

20:43 <re_irc> < (@grantm11235:matrix.org)> It is specific to (safe) rust

20:44 <re_irc> < (@grantm11235:matrix.org)> You can have unaligned raw pointers, but references are always aligned

20:45 <re_irc> < (@allexoll:matrix.org)> All right, thanks for the explanation ! if it needs to be unsafe then it's the user's risk

20:47 <re_irc> < (@grantm11235:matrix.org)> Technically you can create raw pointer safely, but using them requires "unsafe"

20:48 <re_irc> < (@grantm11235:matrix.org)> So if you have a function that takes a raw pointer and uses it for dma, make sure to mark the function as unsafe

20:48 <re_irc> < (@grantm11235:matrix.org)> Technically you can create raw pointers safely, but using them requires "unsafe"

20:49 <re_irc> < (@allexoll:matrix.org)> I'm unsure what would be the use case for that

20:50 <re_irc> < (@allexoll:matrix.org)> maybe i'll encounter something like that when doing dma from/to peripheral <=> memory, but for m2m I don't think that's needed

21:44 <re_irc> < (@alex:matrix.sempto.net)> Does anybody happen to have an example on how to secure the data in postcard with a CRC?

21:46 <re_irc> < (@alex:matrix.sempto.net)> If I understand it correctly it needs to be done in a flavor, but I thought I'd ask here before trying to implement it myself

22:48 <re_irc> < (@jamesmunns:beeper.com)> You could do it as a flavor, or use something like a heapless vec as the output buffer, calculate the CRC, and push that to the end of the buffer

22:49 <re_irc> < (@jamesmunns:beeper.com)> (author of postcard here, btw), happy to answer questions, don't have a CRC example offhand

22:50 <re_irc> < (@jamesmunns:beeper.com)> https://docs.rs/postcard/latest/postcard/ser_flavors/index.html#when-not-to-use-multiple-flavors talks about the pros and cons of flavors or not

22:53 <re_irc> < (@jamesmunns:beeper.com)> There's a unit test that does how to do a really terrible 8 bit checksum, but it's conceptually the same as CRC: https://github.com/jamesmunns/postcard/blob/main/src/de/flavors.rs#L166-L282

22:56 <re_irc> < (@alex:matrix.sempto.net)> : Then I can use the cobs crate to cobs encode, right?

22:57 <re_irc> < (@jamesmunns:beeper.com)> Yeah, you probably want:

22:57 <re_irc> Cobs(Crc32(Data)) if that makes sense

22:58 <re_irc> < (@jamesmunns:beeper.com)> So you can either do that with stacked flavors, or:

22:58 <re_irc> - Serialize

22:58 <re_irc> - Cobs encode

22:58 <re_irc> - Push Crc32

22:58 <re_irc> < (@jamesmunns:beeper.com)> (then backwards for deser)

22:59 <re_irc> < (@alex:matrix.sempto.net)> Yeah, I'd probably do that, sounds easier than implementing a flavor

22:59 <re_irc> < (@alex:matrix.sempto.net)> I'm a bit new to Rust and navigating the type system is still a bit of a struggle

23:00 <re_irc> < (@alex:matrix.sempto.net)> Well... It is the majority of the struggle to be honest.

23:00 <re_irc> < (@jamesmunns:beeper.com)> Yeah, either approach works :)

23:00 <re_irc> Flavors make some things easier, but also have some tradeoffs, but both approaches are valid!

23:01 <re_irc> < (@jamesmunns:beeper.com)> Feel free to me if you have any questions using postcard :)

23:02 <re_irc> < (@alex:matrix.sempto.net)> And thanks for making postcard by the way :) I needed something that was both reasonably efficient with the data on the line and well suited for microcontrollers. Postcard has worked really well for that

23:23 <re_irc> < (@alex:matrix.sempto.net)> : Wait I just realize a potential problem with that approach: I can't use CobsAccumulator if I do that, right?

23:23 <re_irc> < (@alex:matrix.sempto.net)> * realized

23:24 <re_irc> < (@alex:matrix.sempto.net)> Since it will not expect the crc at the end of the cobs decode

23:24 <re_irc> < (@jamesmunns:beeper.com)> Not directly, no, but cobs accumulator is also only like 100 lines of code

23:24 <re_irc> < (@jamesmunns:beeper.com)> So I would just copy and paste it into your project and add a CRC check and error condition

23:25 <re_irc> < (@alex:matrix.sempto.net)> Yep, that'll work. Thanks again!

23:36 <re_irc> < (@orclev:matrix.org)> So, I feel like I'm going crazy here. I'm trying to debug some code and getting an index out of bounds exception, but I'm not seeing how it's remotely possible. I've got a bit vector declared as follows:

23:36 <re_irc> touched: BitArr!(for 68, in u8, Lsb0),

23:36 <re_irc> for y in self.touched.iter_ones() {

23:36 <re_irc> then later on I have this:

23:36 <re_irc> what's driving me nuts, is when I hit the index out of bounds it's reporting that the value of y is 68, which shouldn't be possible, because the highest index for touched _should_ be 67 as the length of touched is 68.