nucleo-l432kc i2c appears to be nonfunctional

[rayozzie]

I have been trying to use this nucleo board for I2c, to no avail.

The nucleo-l432kc pins are wired as SCL PB6 (D5) and SDA PB7 (D4).

I have several devices on the i2c bus.

After repeatedly failing to be able to speak with the devices using machine.Tx(), I then loaded https://github.com/toyo/I2Cscan and tried scanning the bus for any devices. None could be seen.

Without any rewiring, I wrote an Arduino sketch to do the same, and it worked perfectly.

I then removed the wires and hooked them instead to a feather-nrf52840, and used that target for tinygo. It worked perfectly.

As such, my conclusion is that the nucleo-l432kc I2C implementation is not functioning properly.

Thank you for your consideration of the issue.

[deadprogram]

@rayozzie can you try the dev branch of TinyGo with your board to see if it addressed your specific issue? Appears to work as expected on STM32F4 board on TinyHCI.

[kenbell]

@deadprogram - there was a problem with i2c on the stm32f407 some time back? I think we 'fixed' it by putting a delay in the app. I wonder if something similar here - perhaps some part of the initialization code isn't waiting for something to complete? In my test I was sending i2c traffic in a loop - don't know if perhaps the first transmission is missing?

[rayozzie]

@deadprogram i'm following instructions to build dev branch so i can try it. MacOS big sur. Currently blocked by:

$ git clone --recursive https://github.com/tinygo-org/tinygo.git Cloning into 'tinygo'... ... Cloning into '/Users/rozzie/dev/tinygo/lib/musl'... fatal: repository 'https://git.musl-libc.org/musl/' not found fatal: clone of 'git://git.musl-libc.org/musl' into submodule path '/Users/rozzie/dev/tinygo/lib/musl' failed Failed to clone 'lib/musl' a second time, aborting

$ cd tinygo

tinygo $ git checkout dev M lib/CMSIS M lib/avr M lib/binaryen M lib/cmsis-svd M lib/compiler-rt M lib/mingw-w64 M lib/nrfx M lib/picolibc M lib/stm32-svd M lib/wasi-libc Branch 'dev' set up to track remote branch 'dev' from 'origin'. Switched to a new branch 'dev'

tinygo $ git submodule update --init Cloning into '/Users/rozzie/dev/tinygo/lib/musl'... fatal: repository 'https://git.musl-libc.org/musl/' not found fatal: clone of 'git://git.musl-libc.org/musl' into submodule path '/Users/rozzie/dev/tinygo/lib/musl' failed Failed to clone 'lib/musl'. Retry scheduled Cloning into '/Users/rozzie/dev/tinygo/lib/musl'... fatal: repository 'https://git.musl-libc.org/musl/' not found fatal: clone of 'git://git.musl-libc.org/musl' into submodule path '/Users/rozzie/dev/tinygo/lib/musl' failed Failed to clone 'lib/musl' a second time, aborting

tinygo $

[aykevl]

Looks like musl is having trouble. Luckily you probably don't need musl, at least not as long as you're not building Linux binaries. I recommend to clone tinygo without the --recursive flag and then running git submodule update --init.

[rayozzie]

thx. continuing to fight the good fight:

tinygo $ git submodule update --init Submodule 'lib/CMSIS' (https://github.com/ARM-software/CMSIS.git) registered for path 'lib/CMSIS' Submodule 'lib/avr' (https://github.com/avr-rust/avr-mcu.git) registered for path 'lib/avr' Submodule 'lib/binaryen' (https://github.com/WebAssembly/binaryen.git) registered for path 'lib/binaryen' Submodule 'lib/cmsis-svd' (https://github.com/tinygo-org/cmsis-svd) registered for path 'lib/cmsis-svd' Submodule 'lib/compiler-rt' (https://github.com/llvm-mirror/compiler-rt.git) registered for path 'lib/compiler-rt' Submodule 'lib/mingw-w64' (https://github.com/mingw-w64/mingw-w64.git) registered for path 'lib/mingw-w64' Submodule 'lib/musl' (git://git.musl-libc.org/musl) registered for path 'lib/musl' Submodule 'lib/nrfx' (https://github.com/NordicSemiconductor/nrfx.git) registered for path 'lib/nrfx' Submodule 'lib/picolibc' (https://github.com/keith-packard/picolibc.git) registered for path 'lib/picolibc' Submodule 'lib/stm32-svd' (https://github.com/tinygo-org/stm32-svd) registered for path 'lib/stm32-svd' Submodule 'lib/wasi-libc' (https://github.com/CraneStation/wasi-libc) registered for path 'lib/wasi-libc' Cloning into '/Users/rozzie/dev/tinygo/lib/CMSIS'... Cloning into '/Users/rozzie/dev/tinygo/lib/avr'... Cloning into '/Users/rozzie/dev/tinygo/lib/binaryen'... Cloning into '/Users/rozzie/dev/tinygo/lib/cmsis-svd'... Cloning into '/Users/rozzie/dev/tinygo/lib/compiler-rt'... Cloning into '/Users/rozzie/dev/tinygo/lib/mingw-w64'... Cloning into '/Users/rozzie/dev/tinygo/lib/musl'... fatal: repository 'https://git.musl-libc.org/musl/' not found fatal: clone of 'git://git.musl-libc.org/musl' into submodule path '/Users/rozzie/dev/tinygo/lib/musl' failed Failed to clone 'lib/musl'. Retry scheduled Cloning into '/Users/rozzie/dev/tinygo/lib/nrfx'... Cloning into '/Users/rozzie/dev/tinygo/lib/picolibc'... Cloning into '/Users/rozzie/dev/tinygo/lib/stm32-svd'... Cloning into '/Users/rozzie/dev/tinygo/lib/wasi-libc'... Cloning into '/Users/rozzie/dev/tinygo/lib/musl'... fatal: repository 'https://git.musl-libc.org/musl/' not found fatal: clone of 'git://git.musl-libc.org/musl' into submodule path '/Users/rozzie/dev/tinygo/lib/musl' failed Failed to clone 'lib/musl' a second time, aborting

[kenbell]

For me, this works: git clone git://git.musl-libc.org/musl/ this fails: git clone https://git.musl-libc.org/musl/

The submodule is configured as:

submodule.lib/musl.path=lib/musl
submodule.lib/musl.url=git://git.musl-libc.org/musl

(from git config --file .gitmodules -l)

Any chance you have a git config that's changing 'git://' into 'https://' (using the insteadof config option?)

[rayozzie]

Can’t do more today because thanksgiving, but that is almost certainly the problem. Thx for the debug assist.

update: yep, it was this: [url "https://"] insteadOf = git://

[rayozzie]

Ok, so I got the dev version build and cleanly installed [thanks].

$ tinygo No command-line arguments supplied. TinyGo is a Go compiler for small places. version: 0.22.0-dev

However, now when I try to build the simple i2c scanner, I get the following: ../../../Library/Caches/tinygo/goroot-go1.16.5-1a5e73fa17769497a6324b7152c1ac20bacc16ab6b2aadb7ed99e128c5cddbe4-syscall/src/runtime/runtime_stm32_timers.go:16:2: package device/stm32 is not in GOROOT (/Users/rozzie/Library/Caches/tinygo/goroot-go1.16.5-1a5e73fa17769497a6324b7152c1ac20bacc16ab6b2aadb7ed99e128c5cddbe4-syscall/src/device/stm32)

(yes, i tried doing an rm -rf of the tinygo library cache)

$ tinygo info LLVM triple: x86_64-unknown-macosx10.12.0 GOOS: darwin GOARCH: amd64 GOARM: 6 build tags: darwin amd64 tinygo math_big_pure_go gc.conservative scheduler.tasks serial.none garbage collector: conservative scheduler: tasks cached GOROOT: /Users/rozzie/Library/Caches/tinygo/goroot-go1.16.5-1a5e73fa17769497a6324b7152c1ac20bacc16ab6b2aadb7ed99e128c5cddbe4-syscall

[deadprogram]

You probably need to run make gen-device to rebuild the device wrappers for the various processors.

[rayozzie]

That was it. Thx.

So, net-net, with kids climbing all over, I was able to build and run the l432kc i2c scanner under dev, and it still fails.

[kenbell]

I've opened a bug (above) for scanning-type behaviour - it's definitely broken and probably affects a wide set of stm32 MCUs. I see lots of dropped messages.

[kenbell]

I found a timeout issue - the PR above should help make i2c scanning work.