#circuitpython-dev
manic glacierBOT
onyx hinge
OK I assigned those.
I don't know in particular what i2sin would "do" -- act like a sample type so it could go through the new audio filters? or what?
honestly I don't know how the new stuff all fits together into a cohesive whole.
fleet hollow
Definitely agree here. PIO does the job in a lot of cases, but once you try to force input data into the AudioSample API (via RawSample), it gets messy. I started working on an I2SIn class within the core on a local fork, but I haven't wrapped my head around the adjustments needed within the audio dma routines to get it working.
slender iron
I don't know in particular what i2sin would "do" -- act like a sample type so it...
I kinda think of it as an infinite sample. coordinating the buffers could be tricky
fleet hollow
I kinda think of it as an infinite sample. coordinating the buffers could be tri...
It could open up the door to diy effects pedals built on CircuitPython. I have this draft PR for a distortion effect which would be the first thing that comes to mind. https://github.com/adafruit/circuitpython/pull/9776 Would be really cool to control these kind of effects with a variety of sensors which aren't necessarily easy to integrate with within other environments.
native sandal
Can I fix tinyusb to get rid of the #warning "This header file has been moved, thus#include <xtensa/xtensa_api.h>is deprecated. Please use#include <xtensa_api.h> instead"?
Or does that come from somewhere else?
manic glacierBOT
mortal kernel
onyx hinge
@danh I've been seeing some odd issues starting up an RP2350 board with PSRAM co...
I have a feather rp2350 that I soldered a psram to (that one part adafruit sells). I haven't done much with it, but I noticed that it fails to enter the bootloader other than at power-on, which is weird.
mortal kernel
I have a feather rp2350 that I soldered a psram to (that one part adafruit sells...
I've seen odd behavior on both a hand-soldered feather rp2350 and a pimoroni pico plus 2w. I'm not seeing any weirdness entering the bootloader, how's it failing for you?
onyx hinge
@mortal kernel neither the bootloader or circuitpython seem to start when holding boot and pressing reset. Holding boot while plugging in works
tulip sleet
@onyx hinge @mortal kernel It would be good to open an issue about this. 9.2.1 has a PSRAM fix to supposedly handle cache flushing properly. I'm not sure if this would cause failure to start up properly though.
onyx hinge
@tulip sleet it's hard to see how my issue can be circuitpython code, since whatever condition this is occurs when exiting reset, no circuitpython application code could run. however, the psram module could be in a different state that messes things up, I guess..
eightycc's issue yes, that could be the cache flushing thing.
tulip sleet
there has been a lot of discussion of PSRAM stuff in the RPi forums and in pico-sdk. I haven't looked in pico-feedback yet, but I think I am subscribed to the repo. It could be some start-up race condition, but I don't know why the bootloader would try to use PSRAM at all: I"d assume it isn't. Maybe a bus issue??
onyx hinge
is there a RP2350 reference design with psram? One totally un-verified thing I wondered about was whether the psram's CE/ line is not being driven high, so it might start interfering on the shared bus. https://datasheets.raspberrypi.com/rp2350/hardware-design-with-rp2350.pdf 3.2 on page 12 shows such a pull up resistor (R13). It's labeled DNF but then the text says "To be able to use this device, it will obviously have to be populated , as well as R11 (0Ω), and R13 (10KΩ)."
feather rp2350 has a resistor pull up so that's not it
R15 is populated on this board
mortal kernel
<@426200547024961539> <@849815081251766292> It would be good to open an issue ab...
Agreed. I'll open an issue and give reworking #9783 with the new hardware_xip_cache in SDK 2.1.0 a try.
@onyx hinge I'm not seeing your issue on the PSRAM-equipped boards I've tested.
onyx hinge
Maybe the place for this board is the recycling bin
mortal kernel
Soldering that PSRAM chip isn't easy. I vacillated between using hot air or an iron, but went with my trusty Hakko set at 675 with a hoof tip.
I keep a candy jar for dead parts. Someday if someone really gets under my skin I'm going to offer them the contents.😜
onyx hinge
I'm reasonably sure the solder job is OK (no two pins shorted) but I'm no pro and I could well have messed up something else while I was there
I originally ordered 2 feather rp2350s but the 2nd one has not turned up so I haven't been able to test a second one
mortal kernel
Anyone know where I can find the source for flashnuke? It doesn't seem to fully erase a newer 16MB flash.
random junco
Good evening
I'd recommend asking for coding help in #help-with-circuitpython - this channel is for the development of CircuitPython itself.
fathom sand
Adafruit CircuitPython 9.2.1 on 2024-11-20; Waveshare ESP32S3 LCD 1.28 with ESP32S3
I have problems with this board / this CP version.
There is only a CH343 UART but no CIRCUITPY drive or any other USB-Device!
I can access the REPL over the COM-Port and use Thonny to see the file system on the board.
The REPL output shows on the TFT fine!
Also connecting to Wifi works good.
I can not use WEBREPL because adafruit_wgsi is not locally available.
Copying this library into the Lib dir worked.
But: I can not find adafruit_webrepl anywhere to install locally....
Can someone take a look at the build parameters for this board?
Also, a UF2 file for this build would be desirable instead of the .BIN file which needs to be uploaded by webrepl.
BTW: I bought 2 of these board and both behave identically.
Here the modules on the board:
`` help("modules") __future__ canio math storage __main__ codeop max3421e struct _asyncio collections mdns supervisor _bleio countio memorymap synthio _eve digitalio microcontroller sys _pixelmap displayio micropython terminalio adafruit_bus_device dualbank msgpack time adafruit_bus_device.i2c_device epaperdisplay neopixel_write touchio adafruit_bus_device.spi_device errno nvm traceback adafruit_pixelbuf espidf onewireio ulab aesio espnow os ulab.numpy alarm espulp paralleldisplay ulab.numpy.fft analogbufio fontio paralleldisplaybus ulab.numpy.linalg analogio fourwire ps2io ulab.scipy array framebufferio pulseio ulab.scipy.linalg atexit frequencyio pwmio ulab.scipy.optimize audiobusio gc rainbowio ulab.scipy.signal audiocore getpass random ulab.scipy.special audiomixer gifio re ulab.utils audiomp3 hashlib rgbmatrix usb binascii i2cdisplaybus rotaryio usb.core bitbangio io rtc vectorio bitmapfilter ipaddress sdcardio warnings bitmaptools jpegio sdioio watchdog board json select wifi builtins keypad sharpdisplay zlib busdisplay keypad_demux socketpool busio locale ssl
jaunty juniper
Adafruit CircuitPython 9.2.1 on 2024-11-20; Waveshare ESP32S3 LCD 1.28 with ESP3...
What do you mean by webrepl ? That's a Micropython thing.
The web workflow is documented here:
https://docs.circuitpython.org/en/latest/docs/workflows.html#web
I quickly looked at the board page, and it seems like the native USB is not broken out to a USB port, only the CH343 ? So there would be no way to have CIRCUITPY or the tinyUF2 bootloader.
https://www.waveshare.com/esp32-s3-lcd-1.28.htm
From the What's On Board section:
5. CH343P
USB to UART chip
11. USB Type-C connector
USB to UART, for programming and log printing
hollow ether
i'm trying to build circuitpython (main) for the first time and getting a gcc version error when I try to build for my board. is this the right place to ask for help? [i'm on Mac M1, installed arm-gnu-toolchain-13.2.rel1-darwin-arm64-arm-none-eabi.pkg, successfully did gmake -C mpy-cross but gmake BOARD=nullbits_bit_c_pro in ports/raspberrypi gives me lots of "Minimum GCC version 13" errors]
stuck elbow
i'm trying to build circuitpython (main) for the first time and getting a gcc ve...
make sure to set your PATH to use the toolchain you installed
fathom sand
What do you mean by `webrepl` ? That's a Micropython thing.
The web workflow is ...
Thank You for the explanation.
I got "web workflow" working and understand why the drive is not "visible".
Question: Is there an API using the IP connection that makes the drive accessible like "web workflow" is offering?
This could speed up data transfer to and from the drive significantly.
jaunty juniper
Thank You for the explanation.
I got "web workflow" working and understand why t...
I don't think there's anything that uses the web workflow API like that. I hope someone picks up the idea of making it webdav compatible, which would make it possible to treat it somewhat like a network drive: https://github.com/adafruit/circuitpython/issues/6644
hollow ether
make sure to set your PATH to use the toolchain you installed
perfect, thank you! something like this got me going: echo /Applications/ArmGNUToolchain/13.2.Rel1/arm-none-eabi/bin | sudo tee -a /etc/paths [turns out i also had a very old install of v8 of the toolchain that was found first, and I needed a new shell to pick up the path]
fathom sand
I don't think there's anything that uses the web workflow API like that. I hope ...
For development purposes, a plugin for Thonny would be a first big step.
https://github.com/thonny/thonny/wiki/Plugins
A general network drive implementation would be a "catch all" for most needs but would require either big modification of the code running in the board or a driver for each OS.
I saw your: https://github.com/Neradoc/discotool-circuitpython-manager
may this be a starting point?
jaunty juniper
The idea of implementing WebDAV is that it is already supported by many many applications and operating systems, but it still needs a bunch of addition to the http API, and maybe some deeper changes to keep data in sync.
As for the app, I need to get back to it, one of my goals was to feature a "sync" mode that would allow editing the files of the board locally with any software you want and auto sync any changes (or with the press of a button), but I have been away from Circuitpython stuff for a while and am just coming back
fathom sand
You: > "but it still needs a bunch of addition to the http API, and maybe some deeper changes to keep data in sync."
I: >but would require either big modification of the code running in the board
That's what I meant...
jaunty juniper
yeah
fathom sand
I will investigate what it takes to create a Thonny plugin that can read and write files from and to the internal file system using the existing API.
If an app needs to publish and consume outside files, this task can be handled easily inside the app with some sample code for the app devs.
A light weight -importable- FTP-server module comes to mind here...
orchid basinBOT
manic glacierBOT
onyx hinge
@tulip sleet re https://github.com/adafruit/circuitpython/issues/9847 do you recall if there's a reason cdc & msc come before hid? if a hid boot device has to be first.... it's inconvenient!
manic glacierBOT
lone sandalBOT
devout jolt
I have a question on this PR. pre-commit is failing in the CI but I have pre-commit installed and it ran. What's going on?
Weird. if I run pre-commit run --show-diff-on-failure --color=always --all-files like in the CI, it fixes the problems. I have already run pre-commit install but it wasn't running. nevermind, I guess I have a workable solution
tulip sleet
<@329766224093249548> re https://github.com/adafruit/circuitpython/issues/9847 d...
I think you are already proceeding on this, but I think I tested it with the current setup.
tulip sleet
<@329766224093249548> re https://github.com/adafruit/circuitpython/issues/9847 d...
Here's a comment I wrote years ago in `supervisor/shared/usb/usb_desc.c:
// CDC should be first, for compatibility with Adafruit Windows 7 drivers.
// In the past, the order has been CDC, MSC, MIDI, HID, so preserve that order.
if I remember right, some older BIOS's don't even support a composite configuration descriptor for boot devices.
onyx hinge
I missed that comment, thanks.
tulip sleet
@onyx hinge are you continuing on this at the moment? I have Windows available and am trying things out.
onyx hinge
<@426200547024961539> are you continuing on this at the moment? I have Windows a...
@tulip sleet the USB thing? No, I didn't have anything else I knew of to check on Linux.
mortal kernel
Is there a way to exclude a common-hal module for a given board?
I'd like to exclude _bleio entirely for non-W Pico boards.
tulip sleet
I'd like to exclude _bleio entirely for non-W Pico boards.
CIRCUITPY_BLEIO = 0 in mpconfigport.mk or mpconfigboard.mk`
sorry typo
mortal kernel
The problem I'm having is that there appears to be a mechanism in the build that's compiling and linking everything in ports/raspberrypi/common-hal. I'm looking for a way to control that.
@danh Maybe this is the first port where this is desirable?
mortal kernel
Spotted why all the common-hal code is getting compiled. In ports/raspberrypi/Makefile starting at line 699 there's this:
SRC_COMMON_HAL_EXPANDED = $(addprefix shared-bindings/, $(SRC_COMMON_HAL)) \
$(addprefix shared-bindings/, $(SRC_BINDINGS_ENUMS)) \
$(addprefix common-hal/, $(SRC_COMMON_HAL))
SRC_SHARED_MODULE_EXPANDED = $(addprefix shared-bindings/, $(SRC_SHARED_MODULE)) \
$(addprefix shared-module/, $(SRC_SHARED_MODULE)) \
$(addprefix shared-module/, $(SRC_SHARED_MODULE_INTERNAL))
# There may be duplicates between SRC_COMMON_HAL_EXPANDED and SRC_SHARED_MODULE_EXPANDED,
# because a few modules have files both in common-hal/ and shared-module/.
# Doing a $(sort ...) removes duplicates as part of sorting.
SRC_COMMON_HAL_SHARED_MODULE_EXPANDED = $(sort $(SRC_COMMON_HAL_EXPANDED) $(SRC_SHARED_MODULE_EXPANDED))
...(elided lines)
OBJ += $(addprefix $(BUILD)/, $(SRC_COMMON_HAL_SHARED_MODULE_EXPANDED:.c=.o))
somber gazelle
hi all! why BLE HID doesnt work on CP 9 on a nice!nano? no matter what i try, it always ended with "adafruit_ble" library line 445, in _init_devices TypeError: 'int' object is not iterable
onyx hinge
@somber gazelle unfortunately, we recently introduced a bug in the BLE library. It looks like a fixed version hasn't been released yet. https://github.com/adafruit/Adafruit_CircuitPython_BLE/issues/208 for more details. You can try the fixed version by grabbing the file marked "bundles" from https://github.com/adafruit/Adafruit_CircuitPython_BLE/actions/runs/12090444392?pr=209 and manually putting it on your device (note: you have to be logged into github to see/use that "bundles" link)
or you can also grab the previous version https://github.com/adafruit/Adafruit_CircuitPython_BLE/releases/tag/10.0.9
candid sun
<@&356864093652516868> We'll have our weekly meeting in about 60 minutes from now in this text channel and in the circuitpython voice channel. Please take the time to add your notes in advance to the document: https://docs.google.com/document/d/1Z86RigQ9nB_2Gdk6oS2w0tTHTqy6FGG3T22MQpQI08c/edit?tab=t.0 -- I look forward to everyone's updates!
tulip sleet
hi all! why BLE HID doesnt work on CP 9 on a nice!nano? no matter what i try, it...
I just made a release of ta proposed fix. Please try https://github.com/adafruit/Adafruit_CircuitPython_BLE/releases/tag/10.0.11
If it doesn't work, let us know.
blissful pollen
Just lurking today, no notes or anything
onyx hinge
@honest nova very old pylint doesn't work with newer python. in the adafruit libraries we have bumped the pylint version, I think. in the long term we are moving to ruff for linting. Looks like we were using v2.17.4 before switching to ruff. https://github.com/adafruit/cookiecutter-adafruit-circuitpython/commit/a3622a12f11cb4f2c86f0cdfea801ce27f9cf872#diff-5fc0fdfcce62208c1e92b0458d4586847ec66365ca0bd44595053abd7e83e4acR21
turbid radish
Sorry for the abbreviated newsletter, I was sick for a few days
onyx hinge
and the full latest pre-commit file, which you can hopefully just copy into a lib (rather than using cookie cutter again) https://github.com/adafruit/cookiecutter-adafruit-circuitpython/blob/main/{{ cookiecutter.__dirname }}/.pre-commit-config.yaml
@turbid radish we want you to take care of your self & get better. we can wait for the hot python news if necessary.
still zephyr
Thank you for the info @onyx hinge
slender iron
me too 🙂
zephyr would also help us manage background tasks on other cores (think synthio on a second core and displayio too)
👍
mortal kernel
@onyx hinge Cool use of memorymap.
onyx hinge
@still zephyr which thing did you change to version 5.0.0? I didn't see any version 5 in https://github.com/jposada202020/CircuitPython_BMP581/blob/master/.pre-commit-config.yaml
oh that file's a year old
still zephyr
Hello I changed in this library https://github.com/jposada202020/CircuitPython_scales/blob/main/.pre-commit-config.yaml
onyx hinge
- repo: https://github.com/pre-commit/pre-commit-hooks
rev: v5.0.0```I think the enable/disable needs to be in a critical section to guarantee it's going to be done in time.
still zephyr
I have just found out this morning
tulip sleet
there was some discussion of this in the philhower core issue on E9
onyx hinge
[I think @still zephyr is not listening to us live but I hope you'll listen later]
still zephyr
yes I will 🙂
onyx hinge
I like the idea in the abstract 🙂
so the parameter type for e.g., DigitalInOut's first parameter would be Pin | PinConfiguration ?
turbid radish
Thanks Liz!
errant grail
Thanks everyone and to Liz for hosting!
lone axle
Thanks for hosing Liz. Have a great week everyone!
manic glacierBOT
candid sun
Here is the notes document for next Monday’s CircuitPython Weekly meeting. It is at the normal time of 11am Pacific / 2pm Eastern here on Discord. Everyone is encouraged to attend! Please add your hug reports and status updates even if you’ll be attending the meeting - it’s super helpful! If you are unable to attend but would still like to include updates, feel free to include them in the notes and we’ll read them off during the meeting. Hope to see you there! <@&356864093652516868>
https://docs.google.com/document/d/1aZ43h3GXE9me3MBCjRob9WCtkmxUE2rZCtBT9z7Z4p8/edit?tab=t.0
lone axle
When an external display is initialized with displayio it stays alive beyond the end of code.py. Is it possible to re-use that same display instance in subsequent runs of code.py? Currently all of the scripts that do this call release_displays() and then re-initialize it. I'm curious if it's possible to just keep using the instance that is already initialized.
onyx hinge
@lone axle unless it's the board.DISPLAY, no.
devout jolt
More generally, is there a way to stash variables in a similar place to where board.DISPLAY lives, that persist across Ctrl-Cs?
manic glacierBOT
lone axle
More generally, is there a way to stash variables in a similar place to where bo...
I don't know which all devices support it but alarm.sleep_memory provides a way to store raw bytes that persists across ctrl-c / ctrl-d, but is lost if power goes away.
lone axle
I don't think it's quite the same as how board.DISPLAY works, but provides similar functionality
lone axle
blissful pollen
I’m more thinking of what you’re just asking. I was intrigued that while using M...
Do you have a link to any sample code for this offhand (I can google it if not don't go searching). Curious if MP can do it if there is something that could be used.
jaunty juniper
resetting the VM after boot.py and on reloads is a design choice for CP though, it's not MP doing anything special
still zephyr
Have any of the Community Library Authors have tried the workflow to publish on Pypi lately. I have <<Username/Password authentication is no longer supported>> error. Tried to change the Github action yaml file, with some examples that I found, but with any success. If any can point me in the right direction it will save me some time. Thanks 🙂
https://github.com/jposada202020/CircuitPython_BMI160/actions/runs/12132932832/job/33827619137
mortal kernel
@tulip sleet This is what I asking about yesterday regarding _bleio HCI vs. native selection. I wrote it up mostly to clarify my thinking. This isn't an urgent issue, so please have a look when you have time.
tulip sleet
<@329766224093249548> This is what I asking about yesterday regarding _bleio HCI...
It would be absolutely fine to have CIRCUITPY_BLEIO_NATIVE and CIRCUITPY_BLEIO_HCI and not CIRCUITPY_BLEIO. There should be check that both are not set.
A similar thing is CIRCUITPY_TOUCHIO and CIRCUITPY_TOUCHIO_NATIVE. Should probably be CIRCUITPY_TOUCHIO_GENERIC and CIRCUITPY_TOUCHIO_NATIVE.
when these flags were first invented, they were disjoint across ports, and there was less conflict. But not so true anymore.
mortal kernel
"The vpath mechanism will always select the native source modules if both are selected" should be "..if both are present."
tulip sleet
If you feel like making a PR, that would be welcome.
mortal kernel
If you feel like making a PR, that would be welcome.
I can do that. I'd like to find a more durable solution than continuing to use vpath for the second problem.
tulip sleet
I can do that. I'd like to find a more durable solution than continuing to use v...
I think this could be done by using the source list concatenation mechanism used when enabling other modules. In other words, if CIRCUITPY_BLEIO_NATIVE and CIRCUITPY_BLEIO_HCI were treated as completely separate modules, would it regularize things? Then the vpath special case could be removed.
mortal kernel
I think this could be done by using the source list concatenation mechanism used...
Good idea. I'm not gonna say vpath is the root of all evil, but it's up there on my list.
mortal kernel
@danh Is there an Adafruit breakout that works with devices/ble_hci/common-hal/_bleio?
tulip sleet
@danh Is there an Adafruit breakout that works with `devices/ble_hci/common-hal/...
Anything that says AirLift. Also PyPortal and anything else with an ESP32 coprocessor.
If you already have a plain ESP32 board, you could wire it up as an airlift.
lone axle
Have any of the Community Library Authors have tried the workflow to publish on ...
I do think that you need to use an API token now instead of the password. The environment variable names are the same I believe so I don't think the actions script itself needs to change iirc. To get your token log in to pypi.org go to account settings and look for API token section.
onyx hinge
Have any of the Community Library Authors have tried the workflow to publish on ...
I recommend checking out https://packaging.python.org/en/latest/guides/publishing-package-distribution-releases-using-github-actions-ci-cd-workflows/ as the current best practice. However, cookecutter doesn't set things up for this method. Instead, it uses an older method where you generate a token: In github the secret pypi_username is set to __token__ and the secret pypi_password is set to the token value pypi-..... https://pypi.org/help/#apitoken
still zephyr
Hello thanks Jepler and Foamyguy, Yes my libraries used the method included with the cookiecutter and it worked very welll indeed, however it stop working saying that the username/password method. Not sure way because the method works currently for the Adafruit Libraries. Anyway I setup the new method after a lot of trial of error using this article
https://medium.com/@blackary/publishing-a-python-package-from-github-to-pypi-in-2024-a6fb8635d45d
Combined with the GH actions included here
https://github.com/adafruit/workflows-circuitpython-libs/blob/main/release-pypi/action.yml
The result was this
https://github.com/jposada202020/CircuitPython_BMI160/blob/master/.github/workflows/release_pypi.yml
Just for future reference if any other persone find the same problem as I did
Thank you for all the inoput and help ❤️ ❤️
slender iron
When an external display is initialized with displayio it stays alive beyond the...
I intend on redoing this. Lemme find the issue
a GC stack size setting
tulip sleet
this is interesting https://github.com/micropython/micropython/pull/15620#pullre...
yeah, I've been following this
crimson ferry
is it just me or is anyone else having trouble commenting on issues in GitHub?
(#9855: there seems to be a small delay between setting the AP IP and it becoming available)
crimson ferry
n/m, it's working now
orchid basinBOT
lone sandalBOT
manic glacierBOT
onyx hinge
@blissful pollen @fleet hollow Heya! After writing my review comments on #9861 I noticed this nonstandard setter-with-kwargs style in audiodelays & audiofilters, too. Likely it was all propagated from an example in audiomixer.
Removing the kwarg handling will probably reduce code size a bit (it did by 88 bytes for the case in audiomixer), so it might be worth doing. I found the following likely candidates: ```c
shared-bindings/audiodelays/Echo.c:static mp_obj_t audiodelays_echo_obj_set_delay_ms
shared-bindings/audiodelays/Echo.c:static mp_obj_t audiodelays_echo_obj_set_decay
shared-bindings/audiodelays/Echo.c:static mp_obj_t audiodelays_echo_obj_set_mix
shared-bindings/audiodelays/Echo.c:static mp_obj_t audiodelays_echo_obj_set_freq_shift
shared-bindings/audiofilters/Filter.c:static mp_obj_t audiofilters_filter_obj_set_filter
shared-bindings/audiofilters/Filter.c:static mp_obj_t audiofilters_filter_obj_set_mix
Would you like me to file issue(s) about this?Yes please. I think you’re right about where I came from. I’ll try to get to it in the next week or so.
fleet hollow
Yes please. I think you’re right about where I came from. I’ll try to get to it ...
I copied over a lot of code from audiodelays, so I'll defer to @blissful pollen on this.
fleet hollow
Yes please. I think you’re right about where I came from. I’ll try to get to it ...
However, I can update audiofilters separately if needed, but it'd be better to make all of these updates within a single PR.
onyx hinge
OK, issue filed. Thank you in advance for addressing it.
@tulip sleet the other day you said something about critical sections with respect to the E9 workaround. What situation is it trying to prevent? Is that something that can be implemented in pure CircuitPython?
fleet hollow
@onyx hinge Just read over your issue. Do I need to be a part of the adafruit organization on GH to be assigned to an issue? I don't want to step on any toes, but I don't have any issue with an elevated role so that I can better contribute to CircuitPython.
onyx hinge
@fleet hollow yeah I think it's probably a github feature that I can't assign "random people on the github" an issue within an organization. So your involvement will have to be on the honor system 🙂
fleet hollow
<@527668354568159232> yeah I think it's probably a github feature that I can't a...
Good with me! I had some free time, so I've already updated the setters and created the PR (see above). I didn't quite get your 88 bytes per setter result, but I think that may have been due to filesystem block segments in your previous PR if you're using UF2 output size as the metric. With FLASH_FIRMWARE, it's more 35 bytes per setter.
onyx hinge
Good with me! I had some free time, so I've already updated the setters and crea...
my example also removed a probably unique error string, mp_raise_ValueError(MP_ERROR_TEXT("level must be between 0 and 1"));
fleet hollow
my example also removed a probably unique error string, `mp_raise_ValueError(MP_...
Yeah, that must have been it. Regardless, less is more!
onyx hinge
ugh it's distressing that we're so close to full on m4 express boards now 😦 https://github.com/adafruit/circuitpython/pull/9861 is failing on itsybitsym4 ja translation
tulip sleet
<@329766224093249548> the other day you said something about critical sections w...
From the erratum:
To read the state of a pad pulled-down
GPIO from software, enable the input buffer by setting GPIO0.IE immediately before reading, and then redisable immediately afterwards.
So if an interrupt occurred in these "immediately" gaps, the reading could come out wrong. Earl Philhower and I discussed the difficulty of doing this here: https://github.com/earlephilhower/arduino-pico/issues/2380#issuecomment-2334997756 and for several posts afterward.
I don't think it is worth coming up with a software fix for normal use. It's easier to make sure the hardware is set up so the condition doesn't happen.
onyx hinge
yeah I don't think I'll polish my work any further
I dunno how long it is between the Python change to the IE bit -> digitalio read but surely it's most of a millisecond. and yet my code seemed to work.
tulip sleet
I'm sure it would work nearly all of the time. The problem is the small probability it would not work, due to a gc, an extended interrupt, etc.
The ease of using memorymap is admirable -- it's just in this case we really need an atomic section to guarantee correctness.
onyx hinge
If I understand this graph right, the excess leakage current exists during a particular voltage range. If IE is off long enough for the voltage to drop BELOW that, then it will stay low. (screen grab of the current official documentation for E9)
We won't be providing a software workaround in the SDK as it isn't clear what you'd do. The user can use an external resistor[.]
TFW the datasheet says there's a workaround and the sdk devs say "it isn't clear what you'd do"
tulip sleet
I'm not sure they're saying "we don't know what would work". It's true you have to specify the use case via the API, viz. https://github.com/adafruit/circuitpython/issues/9541#issuecomment-2341792598.
I feel like they've avoided describing the fault in a way that would make it clearer when it "latches", when it doesn't, etc. "Here's the chip schematic; here's why it occurs." That explanation is missing.
onyx hinge
I've never seen that level of detail in an erratum. But I do miss the part of datasheets that show "output equivalent circuit", "input equivalent circuit", etc. This went out of style somewhere along the line.
once we weren't thinking about how many "TTL loads" we had on a gate output anymore I guess
mortal kernel
The ease of using memorymap is admirable -- it's just in this case we really nee...
The example I gave in the memorymap doc shows how to use xor and bit flip to make the GPIO control register accesses atomic. This allows use of a register shared by multiple GPIOs that in turn might be shared by user and IRQ-level code. It's atomic, but might not be suitable for the purposes you're discussing.
mortal kernel
@danh @onyx hinge I'm factoring this code out of individual port Makefiles into circuitpy_defns.mk. It looks like it's generic for all ports. Is there any reason not to?
SRC_COMMON_HAL_EXPANDED = $(addprefix shared-bindings/, $(SRC_COMMON_HAL)) \
$(addprefix shared-bindings/, $(SRC_BINDINGS_ENUMS)) \
$(addprefix common-hal/, $(SRC_COMMON_HAL))
SRC_SHARED_MODULE_EXPANDED = $(addprefix shared-bindings/, $(SRC_SHARED_MODULE)) \
$(addprefix shared-module/, $(SRC_SHARED_MODULE)) \
$(addprefix shared-module/, $(SRC_SHARED_MODULE_INTERNAL))
# There may be duplicates between SRC_COMMON_HAL_EXPANDED and SRC_SHARED_MODULE_EXPANDED,
# because a few modules have files both in common-hal/ and shared-module/.
# Doing a $(sort ...) removes duplicates as part of sorting.
SRC_COMMON_HAL_SHARED_MODULE_EXPANDED = $(sort $(SRC_COMMON_HAL_EXPANDED) $(SRC_SHARED_MODULE_EXPANDED))
@mortal kernel superficially I don't know a reason not to.
tulip sleet
@danh <@426200547024961539> I'm factoring this code out of individual port Makef...
looks good to me too, assuming it's the same across all
@onyx hinge does nlr_push_jump_callback() call the routine and then still raise the exception?
onyx hinge
@tulip sleet It's a newer thing in micropython. According to the documentation, the callback gets called when nlr is unwinding the stack past the location of the node; otherwise, if no exception is thrown, it gets forgotten and is never run.
"do this in case of exception, let exception propagate"
mortal kernel
Thanks to you both. That's what I was hoping.
tulip sleet
I have seen the old way of basically doing try-catch in C. This is nice. Thanks.
onyx hinge
yeah it might be worth reviewing if there are other places to use it. however, micropython didn't use it much yet either.
lone axle
@short tendon Do you think you may have some time to look into https://github.com/adafruit/Adafruit_CircuitPython_WSGI/issues/21 ? This library worked prior to the esp32spi switch to connection_manager but wasn't converted for the new way. I had a look into it with an eye towards making the necessary changes but I'm feeling a bit lost. If you do have some time and want to try to tackle converting this one or even could try to point me in the right direction it would be most appreciated.
slender iron
@mortal kernel don't spend too much time improving the makefiles. I'm hoping to replace them with my zephyr migration
mortal kernel
<@849815081251766292> don't spend too much time improving the makefiles. I'm hop...
I’ll restrain myself. I’m doing just enough to get serial HCI or native BLE working for raspberry pi. The factoring makes it easier.
onyx hinge
@tulip sleet @slender iron itsybitsy m4 being full is blocking a number of otherwise good PRs. I'm not sure what to disable. floppyio?
tulip sleet
<@329766224093249548> <@252717193496756235> itsybitsy m4 being full is blocking ...
I think that makes sense
onyx hinge
I don't know how on the edge itsybitsy m4 is, whether we're in for a sea of such problems coming right up. I could go look but I really don't want to know.
floppyio was already off 😧
msgpack? zlib?
rgbmatrix?
onyx hinge
ok, we didn't disable terminal fonts for ja/ko/ru translations on itsybitsy m4 yet. we had on metro m4 previously.
lone axle
Does anyone with knowledge of seesaw have a moment to take a look at the trio of https://github.com/adafruit/Adafruit_CircuitPython_seesaw/pull/137, https://github.com/adafruit/Adafruit_Seesaw/pull/105/, and https://github.com/adafruit/Adafruit_seesawPeripheral/pull/28. They are all related to adding support for the ATtiny416. The changes seem in-line with existing code in the respective repos to my eye, but I know basically nothing about the seesaw internals. Or anyone know who might be a good person for this?
mortal kernel
@danh @onyx hinge In a few cases in py/circuitpy_defns.mk SRC_COMMON_HAL is augmented after it is dead. This affects handling of CIRCUITPY_TOUCHIO_USE_NATIVE, CIRCUITPY_SSL_MBEDTLS, CIRCUITPY_AUDIOMP3_USE_PORT_ALLOCATOR, and CIRCUITPY_HASHLIB_MBEDTLS. The fix is easy, but the bug's been there a long time which makes me hesitate to try.
Might be related to issue 9624.
onyx hinge
@danh <@426200547024961539> In a few cases in `py/circuitpy_defns.mk` `SRC_COMMO...
@mortal kernel I'm not sure what you mean by "dead". An incorrect (missing some files) list of input files will usually cause a link time error.
onyx hinge
I am GUESSING here that you might have an incomplete idea of how make calculates values (because everyone, including me, does)
default:
@echo ALL_FILES=$(ALL_FILES)
@echo ALL_FILES_SORTED=$(ALL_FILES_SORTED)
@echo ALL_FILES_SORTED_CAPTURED=$(ALL_FILES_SORTED_CAPTURED)
WHATEVER = baz.c
ALL_FILES = $(WHATEVER) $(STUFF)
ALL_FILES_SORTED = $(sort $(ALL_FILES))
ALL_FILES_SORTED_CAPTURED := $(sort $(ALL_FILES))
WHATEVER += foo.c bar.c
STUFF := quux.c baz.c
make -f surprising.mk
ALL_FILES=baz.c foo.c bar.c quux.c baz.c
ALL_FILES_SORTED=bar.c baz.c foo.c quux.c
ALL_FILES_SORTED_CAPTURED=baz.c
A variable is expanded "when it is used", so surprisingly ALL_FILES_SORTED contains foo.c and others when it is expanded during the execution of the default rule, even though the default rule appears before the definition of all the values. A := assignment expands the value right away, so it only includes baz.c. The GNU Make manual refers to these as "deferred" and "immediate" assignments. (as a total aside a recent update to the POSIX standard standardizes ::= and :::= to match two slightly different existing implementations of :=. not joking)
This is weird and contrary to the expectations we have from a lot of languages.
tulip sleet
@danh <@426200547024961539> In a few cases in `py/circuitpy_defns.mk` `SRC_COMMO...
Just to be clear, did you mean after it is defined ?
mortal kernel
I am _GUESSING_ here that you might have an incomplete idea of how make calculat...
That's amazing and as you say, weird. Thank you for the explanation, I now understand.
onyx hinge
That doesn't mean we shouldn't re-order things in the Makefiles "if it makes sense to do so". that would include for clarity or consistency.
OTOH tannewt has said he hopes to jettison the makefiles at some point in the future (replacing the build system has been something he's itched to do for years!) so doing real work is probably better than block moving lines of variable assignments for no net change in functionality.
mortal kernel
Just to be clear, did you mean `after it is defined` ?
By "dead" I meant "assigned after last read". After reading Jepler's explanation I see that my assumption of assignments happening in textual order was bad.
mortal kernel
OTOH tannewt has said he hopes to jettison the makefiles at some point in the fu...
Completely agree. Thought I'd found something, but never mind...
onyx hinge
OK, just be careful lest you become a makefile expert in the process.
mortal kernel
OK, just be careful lest you become a makefile expert in the process.
Hah! Like becoming the shop JCL expert once upon a time.
onyx hinge
is that the thing that has to appear at the top of a print job?
no that must be something else
oh that's PJL. I knew PJL once, never JCL.
slender iron
Yes, in my Zephyr work I'm using async python to build the circuitpython bits instead of make.
(and really happy with it so far)
mortal kernel
is that the thing that has to appear at the top of a print job?
IBM mainframe. Job Control Language. Used to this day by z/OS.
onyx hinge
@lone axle sorry not sorry for stepping on your toes on that turtle issue. what can I say, it piqued my interest. [re https://github.com/adafruit/Adafruit_CircuitPython_turtle/pull/45]
and thanks for the review of course
lone axle
<@382939733107408897> sorry not sorry for stepping on your toes on that turtle i...
😄 I appreciate it honestly. I imagine I could have made it to the solution but it would have definitely taken me much longer and adding many prints to get there.
manic glacierBOT
wraith crow
Using absolute newest for the ESP32-P4-Function-EV-Board I now see the CIRCUITPY drive 🎉 and tio can connect over the USB-OTG port...
I noticed there is now a UF2 file being built, is there a UF2 bootloader available somewhere? I took a quick look at tinyusb and didn't see anything for the P4
slender iron
Using absolute newest for the ESP32-P4-Function-EV-Board I now see the CIRCUITPY...
no bootloader yet
but someone had requested a uf2 id
silent torrent
I’m using a pi and blinka to draw and redraw this solar charge panel. I note that every time it draws a new object, it slows down the pi. Is there a way to speed it up by just “filling” objects already drawn with a diff colour?
tulip sleet
I’m using a pi and blinka to draw and redraw this solar charge panel. I note tha...
This channel is more for core circuitpython development. If you don't get an answer here, ask in #help-with-circuitpython
onyx hinge
Is there any way to make usb_cdc.data.write ignore "connected" and write anyway? or is that nonsense? (like pyserial(dsrdtr=False))
The host is considered to be connected if it is asserting DTR (Data Terminal Ready). Most terminal programs and pyserial assert DTR when opening a serial connection. However, the C# SerialPort API does not. You must set SerialPort.DtrEnable.
lone sandalBOT
onyx hinge
huh I wonder if hidden = True used to not get reset across soft-reboots
common_hal_busdisplay_busdisplay_set_auto_refresh(self, true);
circuitpython_splash.x = 0; // reset position in case someone moved it.
circuitpython_splash.y = 0;
supervisor_start_terminal(self->core.width, self->core.height);
if (!circuitpython_splash.in_group) {
common_hal_busdisplay_busdisplay_set_root_group(self, &circuitpython_splash);
}
}
```seems we used to reset the offset [which is code that could be deleted if we leave the entire splash group read-only] but not the hide flag.I think so, that's why it would work to hide the root_group if you don't want the REPL to ever show up outside of a hard reset. Which, if I understand correctly, there's no way to do in the current version anymore ?
tulip sleet
<@&356864093652516868> Our weekly meeting will be starting in about 1-1/4 hours, at 2pm US ET. Add your notes to https://docs.google.com/document/d/1aZ43h3GXE9me3MBCjRob9WCtkmxUE2rZCtBT9z7Z4p8/edit?tab=t.0
tulip sleet
slender iron
yup!
onyx hinge
yup!
errant grail
errant grail
Thanks all and to Dan for hosting!
random junco
Thanks all - have a good week
onyx hinge
thanks! see y'all around
lone axle
Thanks for hosting Dan, have a great week everyone.
tulip sleet
bye thanks everyone!
onyx hinge
@slender iron (use your mental model of RP2 because that's what I'm doing 🤣 ) if only one state machine is running I could just OUT all the pins but only the ones with PIO set in their function would actually be reaching the GPIO pins, right? I won't shed too many tears if nobody else can PIO while running a HUB75 display. Because, yes, the pins are very much non-sequential.
still haven't found anything that says that the PIO pin numbering differs from the GPIO pin numbering but who knows. There are a lot of layers of software between here and there and I haven't looked at them all
slender iron
<@252717193496756235> (use your mental model of RP2 because that's what I'm doin...
ya, you could even disable output for the intermediate pins from pio
tulip sleet
Here is the notes document for next Monday’s CircuitPython Weekly Meeting. It is at the normal time of 11am Pacific / 2pm US Eastern here on Discord. Add your hug reports and status updates to the document before the meeting. If you are unable to attend but would still like to contribute, feel free to add your notes and we’ll read them off during the meeting. Hope to see you there! <@&356864093652516868> https://docs.google.com/document/d/1aGwPWHpPvQm2wmEwwj1MR9NBBY96ah5IbBDRB9N8zag/edit?usp=sharing
willow totem
Is it known on 9.2.1 that the hostname access for circuitpython.local (but device name version) doesn't work for /code/
e.g. http://cpy-s3_reverse_tft-70041dce0bc4.local/code/ gives the wifi/ble/usb choice then choosing wifi gives
using the IP address/code/ works.
Raised https://github.com/circuitpython/web-editor/issues/273
lone axle
Currently when you try to import a module that requires longint support the stacktrace doesn't seem to be able to give you any more information beyond the import line. i.e.
code.py output:
Traceback (most recent call last):
File "code.py", line 1, in <module>
OverflowError: No long integer support
raised from
from adafruit_hx711.hx711 import HX711
But nothing about which line inside of hx711 were problematic.
Is there any better way than just looking around inside of the library to find the parts that utilize the longint?
Also, do we have / should we have some mechanism that knows which libraries require longint support? With such a mechanism circup could warn if you attempt to install a library that relies on it onto a device without it.
onyx hinge
@lone axle ```py
# Convert to 32-bit signed integer
if value & 0x800000:
value |= 0xFF000000
I'm not sure that line is *correct*, fwiw. I think it should be `if value & 0x80_00_00: value -= 0x1_00_00_00` which should also happen to avoid the long integer.
The line written there might make sense in C/C++/Arduino when dealing with a signed 32-bit type, and was copied without testing.
```py
for i in (0x7f_ff_fe, 0x7f_ff_ff, 0x80_00_00, 0x80_00_01):
if i & 0x80_00_00:
i -= 0x1_00_00_00
print(i)
```prints
8388606
8388607
-8388608
-8388607
as for detecting it, we don't have a pre-made tool for it. HOWEVER, mpy-tool (which helps with freezing in modules) in the circuitpython source tree has a line about it: ``` elif is_int_type(obj):
if mp_small_int_fits(obj):
# Encode directly as a small integer object.
return "MP_ROM_INT(%d)" % obj
elif config.MICROPY_LONGINT_IMPL == config.MICROPY_LONGINT_IMPL_NONE:
raise FreezeError(self, "target does not support long int")
My gut says it's not worth spending the time, especially since you'd need a different mpy-tool for each circuitpython version in the bundle, but this program could help with the checking you want to perform..Ah, thank you! I found that line is indeed the one that causes it in that library. It got me curious about the more general question while looking into it.
also thank you for the insight about it's functionality. Bit manipulation particulars are something I'm still not quite fully in tune with. I did not / would not have been able to spot the other issue with that line.
onyx hinge
As written, it would give this result which is not right...
8388606
8388607
4294967295
4294967295
@lone axle as a proof of concept, mpy-cross could gain a new flag "-mno-long" which causes it to error when building an mpy file. https://gist.github.com/jepler/181ab2c5ccd4175d3635d459ee739e2c
Hypothetically, circuitpython-build-tools would try building a module with this flag. If it errors, it would store a new piece of metadata (where?) noting that the package needs long integers, and then mpy-cross it without the flag to get a working file.
$ echo 'def f(): return 0x7 << 31' | ./build/mpy-cross -mno-long -o /dev/null -
OverflowError:
$ echo 'def f(): return 0x7 << 31' | ./build/mpy-cross -o /dev/null -
$
basically not call pio_gpio_init on that pin right? so the pinmux is still set to gpio, presumably
plucky flint
Good morning folks... Not sure where else to post this, but it feels relevant here: https://madewith.mu/mu/users/2024/12/10/retirement-plans.html
TL;DR - we're sunsetting Mu in 2025, and the link is to our retirement plan so we honour and respect our users with minimum disruption.
More context: Mu has had a good run, there are now plenty of alternatives and (most importantly for me personally), we're all volunteers and life moves onto other fun stuff, so sunsetting with a final release is the right thing to do.
Thank you, and season's greetings to all Adafruit friends! 🤗 🐮 ❤️🐍
tulip sleet
@plucky flint I just saw this announcement this morning in the repo, and am sorry to see this. Thank you all very much for Mu, which served us so well as the "easy editor" for beginners. Given the long time since the last release, we have already been thinking about how to move on. I started a discussion thread in #help-with-circuitpython about the alternatives that people see. I would also be interested in what you see as the alternatives, in case we have missed some.
plucky flint
<@354682505271443468> I just saw this announcement this morning in the repo, and...
We discussed if we should mention alternatives and decided against simply because we don't want to appear to endorse one thing over another (which may not be known to us). The community will work it out and, to be honest, this is very much a case of something must die for something else to be born. A Mu-ish hole will appear and folks will fill it with something wonderful... I'm excited to see what that might be... and hope for something creative, compassionate and playful.
random junco
Good morning folks... Not sure where else to post this, but it feels relevant he...
Thank you to you and the maintainers for Mu. It's really nice to see such careful thought and planning going into sunsetting Mu, it's appreciated.
tulip sleet
We discussed if we should mention alternatives and decided against simply becaus...
I appreciate the neutraility. For CircuitPython, we have several known alternatives already, including our own code.circuitpython.org (browser-based). For CPython users, that is a different audience, and it will be interesting to see how that goes.
plucky flint
Quite. One of the most important aspects of Mu was that we wanted folks to graduate away from it. We did our job if folks could take their first steps with Mu and gain the confidence to move elsewhere for a "proper" editor experience - as their needs and understanding developed.
manic glacierBOT
onyx hinge
huh this project gets totally different styling than the normal readthedocs. I wonder what I need to fix. https://adafruit-blinka-rp1pio.readthedocs.io/en/latest/_generate/adafruit_rp1pio.StateMachine.html#adafruit_rp1pio.StateMachine
fleet hollow
huh this project gets totally different styling than the normal readthedocs. I w...
Looks like you selected the alabaster theme in conf.py, https://github.com/jepler/Adafruit_Blinka_rp1pio/blob/53991ceebaec95e931a9d81942696406db76a00b/docs/conf.py#L114. Seems like that configuration is the default sphinx one. Did you use cookiecutter to create the repo?
onyx hinge
Looks like you selected the alabaster theme in conf.py, https://github.com/jeple...
It's from a pybind11 project template, not adafruit circuitpython cookiecutter...
fleet hollow
It's from a pybind11 project template, not adafruit circuitpython cookiecutter.....
Oh sorry, I am not too familiar with pybind11 (or Blinka, tbh). I think all that is needed is this:
import sphinx_rtd_theme
html_theme = "sphinx_rtd_theme"
I'll try it
fleet hollow
I'll try it
I am definitely not certain since there may be other requirements to get that working on RTD.
lone sandalBOT
lone axle
does LONGINT support work differently on raspberrypi ports than on others say atmel? i.e. if I made a build for an rp2040 device with LONGINT_IMPL = NONE in mpconfigboard.mk would it be expected to raise the same OverflowError: No long integer support that an atmel device would if too large of a value appears in the code?
tulip sleet
does LONGINT support work differently on raspberrypi ports than on others say at...
It works the same
fleet hollow
@onyx hinge I've been thinking about your comment on audiofilters.Distortion regarding BlockInput ticking here: https://github.com/adafruit/circuitpython/pull/9776#discussion_r1872006686. Would you like me to go ahead and add a blocks property to the class and handle the ticking of those objects just as synthio does? Or are we still debating the best way to approach this problem?
blissful pollen
@fleet hollow @onyx hinge Just a FYI I've seen the tags on github on a few audio things, just have not had time / brain space to keep up. If there is something you really need my eyes on just ping me here.
onyx hinge
@fleet hollow I don't know that they need a blocks property. But they need to call shared_bindings_synthio_lfo_tick once per 256 samples(*) to set the global data that causes synthio_block_slot_get to actually advance the LFOs that are used.
blocks is used as a way to tick blocks that aren't actually used by anything. or which aren't necessarily used by anything. So for instance if you wanted to have an LFO that is polled from CircuitPython code in order to set an LED color. Or, if you have an LFO that you want to keep advancing its phase regardless of whether it's associated with a playing note.
For the first case, having some object that is not a synthesizer/audio source at all might be the right solution.
For the second case, I don't know if that is applicable to the e.g., an audio echo effect.
(*) shared_bindings_synthio_lfo_tick could have a "number of samples" parameter added to the C function, so that you could calculate a different number of samples each time. Of course, since the arithmetic is all discrete you could get slightly different results from advancing LFOs by 256-sample@44.1kHz units vs advancing them at 256-sample@8kHz or 512-sample@24kHz rates...
lone axle
It works the same
It would appear there is something strange going on with the build process perhaps. I am making builds for adafruit_feather_rp2040 and there doesn't appear to be a difference in the resulting build whether or not I have LONGINT_IMPL = NONE in the mpconfigboard.mk file. I noticed these outputs during the build show the exact same numbers whether or not I include that line in the .mk file
FLASH_FIRMWARE: 865904 B 1020 KB 82.90%
On both builds with and without that line I can open the repl and run this:
>>> 0xFF000000
4278190080
and get the same results. On a different device i.e. atmel with longint disabled putting this into the REPL results in the OverflowError.
I checked the hash of the firmware.uf2 with a make clean in between builds with this command and it outputs the same hash for both builds.
echo -n build-adafruit_feather_rp2040/firmware.uf2 | sha256sum
7df4a376f4e0398b6add5f8b8b6f17cb4af6410b1a05656a7fc7c70f3d0b106d -
I'll look a more into it tomorrow. Maybe I've somehow messed up my environment and it's not picking up changes to the mk file or something.
tulip sleet
It would appear there is something strange going on with the build process perha...
I think it is being overriden by the first line in ports/raspberrypi/mpconfigport.mk, which is
LONGINT_IMPL = MPZ
so just change that and rebuild
after make clean
manic glacierBOT
Is it really necessary to limit the width? Having it expand to actual width seems way more logical
Line in question: https://github.com/adafruit/circuitpython/blob/27456d525e645acc68126a3d4ad831aa01aa2281/supervisor/shared/web_workflow/static/style.css#L2
manic glacierBOT
lone axle
I think it is being overriden by the first line in `ports/raspberrypi/mpconfigpo...
I tried changing MPZ to NONE in ports/raspberrypi/mpconfigport.mk. With that change the build doesn't seem to succeed fully. There are some errors about mp_obj_int_get_uint_checked that don't get printed during a build without this change. The flash size table does get printed, and does show different sizes so it does seem that at least some parts of the build suceed.
arm-gnu-toolchain-13.2.Rel1-x86_64-arm-none-eabi/bin/../lib/gcc/arm-none-eabi/13.2.1/../../../../arm-none-eabi/bin/ld: build-adafruit_feather_rp2040/shared-bindings/memorymap/AddressRange.o: in function `memorymap_addressrange_make_new':
AddressRange.c:(.text.memorymap_addressrange_make_new+0x32): undefined reference to `mp_obj_int_get_uint_checked'
arm-gnu-toolchain-13.2.Rel1-x86_64-arm-none-eabi/bin/../lib/gcc/arm-none-eabi/13.2.1/../../../../arm-none-eabi/bin/ld: AddressRange.c:(.text.memorymap_addressrange_make_new+0x66): undefined reference to `mp_obj_int_get_uint_checked'
Memory region Used Size Region Size %age Used
FLASH_FIRMWARE: 853412 B 1020 KB 81.71%
RAM: 61348 B 256 KB 23.40%
SCRATCH_Y: 0 GB 4 KB 0.00%
SCRATCH_X: 2 KB 4 KB 50.00%
collect2: error: ld returned 1 exit status
make: *** [Makefile:752: build-adafruit_feather_rp2040/firmware.elf] Error 1
However there is no resulting firmware.uf2 file that gets written.
tulip sleet
looks like memorymap is assuming longints (which it will certainly need), so you'll need to turn off memorymap too.
lone axle
I don't necessarily need to get a raspberrypi build without longints. It just happened to be one of the devices I grabbed to test something and I was perplexed by the results. The value in the port mk file taking precedence makes sense and solves what was the most mysterious part in my mind.
tulip sleet
ideally memorymap being on should be guarded to require longints
lone axle
ahhh, I see shared-bindings/memorymap... in there now. I'll try one with that turned off.
Yep, that did the trick and I do now have a successful build without longint that does behave the same as the others. Thank you!
slender iron
@tulip sleet I wrote the tlv stuff. I'd say strip what needs to be stripped to work on CP. We always have the original version
manic glacierBOT
slender iron
@tulip sleet now I remember the fancy typing stuff was a pr by someone
tulip sleet
hah, I had to just throw it all out. Getting closer. Now on AES fixups.
I am thinking that maybe there might be a separate Adafruit_CircuitPython_MiniCircuitMatter library that stays pruned, so the other one can be fancier or have better type checking.
manic glacierBOT
manic glacierBOT
generally super long lines are considered bad for readability
Well, the messages are long anyway, except its scroll able right now.
It doesn't wrap lines/break the lines.
So removing the max width has no downside, other than having it be able to display longer lines if your screen width is wider..
Your point would only make sense if the console had wrap lines enabled, which it doesn't.
tulip sleet
@slender iron There are a couple of places in session.py which use a fixed nonce for encrypting with AESCCM, e.g.
s2k_cipher = AESCCM(
s2k,
tag_length=crypto.AEAD_MIC_LENGTH_BYTES,
)
sigma2.encrypted2 = s2k_cipher.encrypt(b"NCASE_Sigma2N", bytes(tbedata.encode()), b"")
Do you remember this? The doc says never ever reuse a nonce, so was this meant to be just temporary?
keen jay
Hi everyone. I'm not sure if this is the right channel to use, but I've started working on French translations and I had a question about how to proceed if I noticed there were potential inconsistencies in tone or nomenclature with certain previously translated strings. In such cases, should I just add comments on the relevant string/glossary page or should I just go ahead and replace?
tulip sleet
Hi everyone. I'm not sure if this is the right channel to use, but I've started ...
can you give a few examples? Do you disagree with the nomenclature, for instance?
keen jay
Of course. One good example is regarding "safe mode". It was previouly translated as mode sûr, however this doesn't reflect the way safe mode is translated currently. Windows has been using the phrase mode sans échec, which better describes what that mode entails. Another example would be the currently registered glossary term for "in use". Currently, the selected translation for that word is "utilisé". Without context, this connotes past tense. "En utilisation" would be more apt
(granted, I understand that there are several ways to approach this)
tulip sleet
"le truc est en utilisation" et pas "le truc est utilise'" ?
are there academie francaise guidelines for CS techinical terms, or are they useless or ignored?
keen jay
"le truc est en utilisation" et pas "le truc est utilise'" ?
Very good question. I can definitely look that up. I apologize if I came off as rude about this
tulip sleet
The previous translators were @ajs256 and @lisapple.
keen jay
Also, I really must say here that I am not trying to criticize or undermine anyone
tulip sleet
no, not at all. I am an low-ish intermediate French speaker, but I know that there have been tussles about what terms to use when: e-mail, courrier, etc.
One issue has been that French is one of the most verbose translations, just due to the nature of French, so there may have been some effort to make the translations shorter.
You don't need to apologize at all. I am just trying to figure out if there's any guidance at all that we could point to.
keen jay
One issue has been that French is one of the most verbose translations, just due...
Ah! 10-4. That makes sense. I can focus on concision. French does tend to go all verbose
tulip sleet
If a single translation causes a build to be too big, we shrink the build (often by turning off features) for everyone, so we don't favor one language over another.
keen jay
You don't need to apologize at all. I am just trying to figure out if there's an...
I know that l'Office de la langue française (the Québec entity that oversees the language) does have an official glossary for technical terms. I can refer to those and compare with the French (from France) entity, to ensure that the translations don't have too much of a regional bias
tulip sleet
if you see inconsistencies, it's fine to rectify those. Alsowhen words are reused, it makes the translation smaller. The two previous translators I know of: one is in the San Francisco Bay Area, but could be a native speaker, and the other is in France. And you are in Quebec?
keen jay
if you see inconsistencies, it's fine to rectify those. Alsowhen words are reuse...
Yes, I live in Québec. I do however tend to prefer International French terms instead of Québec-specific terminology (like how we say "courriel" for email in Québec whereas in France they use e-mail or even mail)
tulip sleet
we coudl have a a quebecois translation, but that's more work. We recently took a vacation to QC and I remember some podcasts that dicussed a few computer terms that differed from Metropolitan French.
keen jay
we coudl have a a quebecois translation, but that's more work. We recently took ...
That is one big can of worms to let loose though! But yes, if there's interest in having a French Canadian translation, I am willing to help
tulip sleet
@jaunty juniper here is in France, and you might bounce things off him if that would be helpful.
There is even an "en_gb" translation, which is almost the same as the en_us, but has a few spelling differences.
We are very grateful for you working on the translation! I don't want to hold you back at all. It would be nice for it to be more complete.
keen jay
Excellent. I'll send them a message!
tulip sleet
I would say if you find that the QC nomenclature is noticeably different than Metropolitan, we can start a new translation. But start out with the current one and see how it goes.
keen jay
Sounds good!
slender iron
<@252717193496756235> There are a couple of places in session.py which use a fix...
Those are in the spec I think. You don’t want to change them
tulip sleet
Those are in the spec I think. You don’t want to change them
that's surprising, but yes, it looked like some given name.
I will move on in any case.
keen timber
Hi all! Trying to drive a coin hopper and quite new to this. Just wanted to see if this wiring seems correct. I’m hoping to use the Adafruit Funhouse (https://learn.adafruit.com/assets/102128) for driving it.
Pin 1 Supply - Using external power supply (positive)
Pin 2 Ground (0v) - Using external power supply (negative)
Pin 3 Motor on +ve - Use A0 - VCC
Pin 4 Motor on -ve - Use A0 - A0 pin (use CircuitPython to set pin direction to INPUT and then Pull.UP
Some additional info from the hopper’s datasheet is here, thank you so much:
Provide Pin 1 with 24vDC and Pin 2 with 0v constantly.
Motor On+ should be pulled high to a reference voltage of between 5vDC and 12vDC for interfacing ‘directly’ with host machine control circuitry. If connection of Motor On+ to a higher voltage is desired, then use of a pull up resistor is recommended to limit the current to between 8mA and 23mA. To drive the hopper the host machine must pull Motor On- down to below 1.0v DC. This can be achieved by use of a switch, relay or open
collector transistor.
austere acorn
Hi all! Trying to drive a coin hopper and quite new to this. Just wanted to see ...
This is the channel for discussions about CircuitPython core development. Try https://discord.com/channels/327254708534116352/330406777009209346 and it would be helpful if you provided a link to the device you're trying to connect to the Funhouse.
keen timber
This is the channel for discussions about CircuitPython core development. Try ht...
Oh sorry! Will move it thank you! Have a good day!
digital shoreBOT
adafruit
manic glacierBOT
manic glacierBOT
shut copper
In the https://circuitpython.org/contributing/library-infrastructure-issues page there is an initial section titled "A programmatic error occurred"
All of those CI builds appear to be passing. Setting up the adabot locally as it looks like this output may come from there, although at current gtihub API rate limiting this may take days to run. Am I looking in the right area? Or where can I see what those programmatic errors might be?
onyx hinge
In the https://circuitpython.org/contributing/library-infrastructure-issues page...
err_msg = [
f"Error: retrieving latest release information failed on '{repo['name']}'.",
f"Information Received: {repo_release_json['message']}",
]
self.output_file_data.append("".join(err_msg))
return [ERROR_OUTPUT_HANDLER]
```it looks like this represents adabot having trouble retrieving info from github. I don't think it's related to CI actions failures.ERROR_OUTPUT_HANDLER = "A programmatic error occurred"
those items seem to have readthedocs configuration that doesn't match what is expected. For instance https://app.readthedocs.org/projects/circuitpython-pixelmap/ while the script is looking for /api/v3/projects/adafruit-circuitpython-pixelbuf/builds
I don't think there's anything you can do about it as a contributor. @lone axle might be able to put it on a list to check.
pallid creek
What is the single source of truth for CircuitPython boards? I’d like to make a tiny search page for boards so you could look for , say, all boards with an SD card and get code snippets to use them. Similar to the main boards page. https://circuitpython.org/downloads
Is there a JSON file or similar for each board in a repo somewhere ?
jaunty juniper
What is the single source of truth for CircuitPython boards? I’d like to make a ...
there is data in the repo, it's not json but the board files have a structure, with a feature list in the top section. https://github.com/adafruit/circuitpython-org/tree/main/_board
there's also the data generated when building Circuitpython that lists available modules in the files.json file: https://github.com/adafruit/circuitpython-org/tree/main/_data
it's not a very detailed list of features
like SD card is not part of the listed features
lone axle
In the https://circuitpython.org/contributing/library-infrastructure-issues page...
Some of the ones listed on that page can be tricky to troubleshoot due to the tooling being set up to run on everything at once. For any individual issue on that page the easiest way to learn about what it is indicating is to grep for the message on the page inside of the adabot repo. i.e. git grep -n "A programmatic error occurred" will find the area of code inside adabot that generates that portion of the page, from there you can trace it backwards to find the failing conditions that led to it's inclusion.
For the API rate limit thing, the limit is much lower for anonymous requests with no token associated with them. If get a github API token for your account and make an environment variable named ADABOT_GITHUB_ACCESS_TOKEN with your token as the value, then run adabot in that environment it will use your token for the requests which raises the limit high enough that it can finish running locally in a more reasonable time.
thorny jay
we coudl have a a quebecois translation, but that's more work. We recently took ...
I am French speaking Belgian... my feeling is that in Belgium we use more English words for IT stuff than what French would use.
Also I learned about computer reading the English manual of my ZX81 first computer... and when I started to use the internet, the internet was only in English while French people were still on the "Minitel" (A kind of state sponsored BBS on dump terminal).
I am not sure a Belgian-French (like my AZERTY keyboard) would make a lot of sense... but I would sure enjoy reading a diff between French-French and Quebec-French. 😉
onyx hinge
<@&356864093652516868> Howdy friends! We're about 35 minutes away from our last meeting of the year. Join us in the CircuitPython voice chat, and add your notes to https://docs.google.com/document/d/1aGwPWHpPvQm2wmEwwj1MR9NBBY96ah5IbBDRB9N8zag/edit?usp=sharing
onyx hinge
sure can
turbid radish
slender iron
thanks everyone!
onyx hinge
thanks @lone axle !
slender iron
happy holidays
lone axle
Thanks everyone, have a good week!
wraith crow
☝️
lone axle
Here is the notes doc for the next meeting https://docs.google.com/document/d/1f2Em2jVmoOt-aoOyxfzj5u1yUQXcxz6sGIjCWcWIAHY/edit?usp=sharing It will be held on January 6, 2025 at the usual time of 2pm US Eastern / 11am US Pacific. Enjoy the next couple of weeks off, and we hope to see you for the next meeting on January 6th! <@&356864093652516868>
tulip sleet
@slender iron my TLV issues may be do to tag ordering. Mine may be different due to lack of __mro__ and vars() in CPy. Still looking at that.
In the Matter TLV spec, they specify a canonical ordering:
A.2.4. Canonical Ordering of Tags
Where a distinguished ordering of tags is required (e.g. for the purposes of generating a hash or
cryptographic signature of elements within a structure), the following ordering rules SHALL be
used:
• Anonymous tags SHALL be ordered before all other tags.
• Context-specific tags SHALL be ordered before profile-specific tags.
... [ etc.]
Are you enforcing this in some way? I don't see that, but might have missed something.
@slender iron also, I'm getting different results when doing .encode() vs .encode_into(). e.g. in test_chunked_message.py, I was doing .encode() to just check the encodings, and saw that in test_only_one() .encode() produced:
15-00-36-00-04-00-04-01-04-02-04-03-18-28-02-24-ff-0b-18
but .encode_into() produced
15-36-00-04-00-04-01-04-02-04-03-18-28-02-24-ff-0b-18.
The second byte 00 is not in the latter. The way you go through the pieces is different. Not sure if this is a bug or just a difference. I'll look further.
shut copper
I went on a journey to understand the example naming rules based on the https://circuitpython.org/contributing/library-infrastructure-issues page and the Good First Issue to begin contributing. And that all led to these repos in the list of missing sensor/library name:
https://github.com/adafruit/Adafruit_CircuitPython_Requests
the examples directory consists of more directories, however each directory contains correctly named files, i.e. requests_xxx.py
https://github.com/adafruit/Adafruit_CircuitPython_ESP32SPI
here there is a mix of directories and files, although again all the example .py files are correctly named.
https://github.com/adafruit/Adafruit_CircuitPython_DisplayIO_Layout
again flagged although all naming appears correct.
That led me to adabot, and a dive into the ruleset enshrined in the code. So I can see why these are getting flagged and it is a pretty simple fix to make this work accordingly. However after testing these three now passed but these new ones now failed:
https://github.com/adafruit/Adafruit_CircuitPython_TestRepo
This one is interesting because the repo says it exists to test adabot, and the naming used is example directory named testrepo_xxx but it seems the content of the directory does not conform to the naming rule of testrepo_xxx.py.
https://github.com/adafruit/Adafruit_CircuitPython_Radial_Controller
This one too has directories named radial_controller_xxx containing anything goes .py files.
So which ruleset is correct or should both be correct?
lone sandalBOT
still zephyr
Hello I am in the process of archiving some old repos that I am not longer maintaining. What would be the process to remove them from the community bundle library? Thanks
tulip sleet
Hello I am in the process of archiving some old repos that I am not longer maint...
adding @lone axle to this thread. You can submit a pull request to remove them. Which ones are they? Someone else may want to take them over.
slender iron
<@252717193496756235> my TLV issues may be do to tag ordering. Mine may be diffe...
generally the tag order is done by the declaration order which is also the tag number.
tulip sleet
generally the tag order is done by the declaration order which is also the tag n...
I think the order may be coming out different due to the __dict__ or something.
slender iron
<@252717193496756235> also, I'm getting different results when doing `.encode()`...
I think the 00 is a bug because 15 is an anonymous struct iirc. The 00 may be an incorrect tag number
slender iron
I think the order may be coming out different due to the `__dict__` or something...
you could sort it by id first
__mro__ handles inheritance I think too
tulip sleet
`__mro__` handles inheritance I think too
I wrote an __mro__ like thing to return the class and all the superclasses.
still zephyr
adding <@382939733107408897> to this thread. You can submit a pull request to re...
Mostly the sensor libraries. I can do the PR just wondering how to deal with the adabot part of it.
tulip sleet
Mostly the sensor libraries. I can do the PR just wondering how to deal with th...
adabot gets the list of libraries dynamically
lone axle
Indeed, I think if you submit a PR to the bundle removing the submodule of the library and it's entry in the list file the rest should get taken care of automatically after it's merged.
still zephyr
Yes the problem is I not sure how to remove the submodule☺️. Ill google it or copilot it 😇
lone axle
I went on a journey to understand the example naming rules based on the https://...
I'm not entirely sure what the specific rules around directory name vs. filename and their usage of library name prefix. The .py files should generally be prefixed with the library name because there are some tools that assume that will be the case, though I'm not sure what all of them are specifically. Directory names, and names of files inside the directory I don't know about for certain.
the rotary controller one, I think your instinct about why it's flagged is correct. these files https://github.com/adafruit/Adafruit_CircuitPython_Radial_Controller/tree/main/examples/radial_controller_rotary_trinkey are named boot.py and code.py as a sort of "full set" of files that can be copied to the device to make it do it's thing.
I don't think there are a whole lot of libraries where we'd have that "full set" checked in to the repo, but in some cases like this one it does make sense and I'd be inclined to say the tooling should allow it. However I don't know of / what other ramifications it could cause. As long as care is taken to check on the relavent guide pages that could embed this code and make sure they continue to work and show the appropriate code I think it would be okay to change it so that this issue won't get flagged in the report.
slender iron
@tulip sleet you setup o test usb host? My guess is that the memory protection is causing an issue on the second core
tulip sleet
<@329766224093249548> you setup o test usb host? My guess is that the memory pro...
I can do some kind of testing. I have the Pimoroni RP2350B breakout and could wire it up with a USB-B breakout.
I can also test it on an RP2350A
slender iron
Either should be fine. I doubt it is pin specific
tulip sleet
maybe what you suggest or is it using high-bank pin numbers?
I have to look...
uses GPIO 32, 33
onyx hinge
I think PIO support for high GPIOs in CircuitPython was a "?". I never tested it.
slender iron
Ah that could be the problem too. Worth verifying it works on lower pins first
lone axle
If anyone has a free moment sometime and could review: https://github.com/circuitpython/CircuitPython_Library_Screenshot_Maker/pull/26 it's appreciated. This repo doesn't seem to allow requesting review from circuitpython librarians.
tulip sleet
If anyone has a free moment sometime and could review: https://github.com/circui...
since it's in the circuitpython org, it doesn't have the adafruit team
lone axle
It's got a big list of indivuduals and adafruit-adabot. These are probably also a member of circuitpython org. I wonder if it's possible to make the team a member of both orgs
tulip sleet
I think the scope of a team is an org, but we could make an identical team
but has to be maintained by hand
@lone axle I'm confused:
# check if it's a wildcard import e.g.
# from module.submodule import *
if cur_import.name.endswith(".*"):
but it's looking for .*, not just *
or <whitespace>* (use a regexp for that)
lone axle
<@382939733107408897> I'm confused:
```py
# check if it's a wildcard import e.g....
replied in github.
stuck elbow
@lone axle you know that split takes an optional argument that tells it how many times to split, and also there is rsplit that goes from the right side?
so you can do things like if "foo.py".rsplit('.', 1)[-1] == 'py':...
tulip sleet
I was confused because I thought it was literally trying to match the string from module.submodule import *.
lone axle
rsplit I think maybe I knew about, but don't often think to use it in the moment. optional args for split I did not know, that is handy. Thank you.
tulip sleet
so the comment made me more confused 🙂
lone axle
Would it be good to add more details to the comment about how findimports transforms the import string?
tulip sleet
You could just say "findimports may return an import like foo.bar.*
slender iron
@tulip sleet did you try usb host yet?
tulip sleet
<@329766224093249548> did you try usb host yet?
i'm just making sure it's working on an RP2040 first. It's being flaky.
Sometimes it works, sometimes it doesn't.
@slender iron a USB Host Feather RP2040 is working, but when I try a jumpered USB socket on different pins on the same board, I get RuntimeError: usb_host in use even after a power cycle
aha, ok, it's hardwired in board.c to construct the usb_host object in advance
I'll use a regular RP2040 Feather to test that
ok, got that working.
@slender iron ok got test program working on D10 and D11 on a Feather RP2350
Metro will arrive this afternoon, and I'll test
slender iron
Ok nice! Metro issue must be the high PIN numbers
mortal kernel
@danh Are you testing with SDK 2.1? There are a number of changes there to high numbered pins on the RP2350.
tulip sleet
@danh Are you testing with SDK 2.1? There are a number of changes there to high ...
I saw all that and was worried about it.
but we haven't moved forward yet
mortal kernel
but we haven't moved forward yet
I was thinking that without SDK 2.1 you might have problems with high numbered pins. PR 9831 advances to 2.1. Maybe you're testing with the older SDK in error?
tulip sleet
I was thinking that without SDK 2.1 you might have problems with high numbered p...
maybe -- I did have to revert a PR, but I don't remember whether I reverted it on the tip of main or not. I'll look
I reverted 9659 on the tip of main and gave that to Limor to test. But I just got the Metro board myself and will do some more testing.
we may still be assuming something incorrect in pio about the pin numbers
I was going to test host usb, and then test simple PIO stuff on higher-numbered pins (neopixel would be an easy test)
mortal kernel
maybe -- I did have to revert a PR, but I don't remember whether I reverted it o...
The Pimoroni Plus 2 exposes some high numbered pins on the SP/CE connector. Testing with that might help indicate if it's SDK or Metro.
tulip sleet
i have that board, but now I have the 2350B metro too, and it's a lot easier to get to the n>31 pins. I don't have a breakout cable for the SP/CE connector, just the connector to the wifi addon board.
also I bought the 2350B PGA board, and then realized just today it has no USB connector 😦
too much trouble to wire it up and UPS was delivering the metro today anyway
mortal kernel
Let me know if there's any way I can help.
mortal kernel
@danh Would you like me to pick up #9897?
tulip sleet
@danh Would you like me to pick up #9897?
I was going to work on this myself or as @onyx hinge about it, but your help or doing it would be welcome, if you think it's not a big job. Do you feel you already have experience on this, or have studied it? It's needed to get USB Host to work on the upcoming Metro RP2350B. It's more than just making things uint64_t.
I looked at https://github.com/micropython/micropython/pull/15619.
RP2350 adds GPIOBASE to PIO to specify whether to start pin numbering at 0 or 16, so you can use pins 0-31 or 16-47. In MicroPython, this setting is exposed directly. I think we could instead determine the setting by examining the specified pins in the constructor (and complaining if they're out of range).
There are also some include paths with /rp2040/ in the current impl, and I haven't looked at what should be done with those.
mortal kernel
I was going to work on this myself or as <@426200547024961539> about it, but you...
Let me take a closer look. I'll put what I find in the issue and let you know if it fits my skill set.
onyx hinge
@tulip sleet I would prefer to avoid having users manually track whether they're using the "high" or "low" pins. The sdk structures will track which banks in the pinhi field of pio_sm_config. we can check for configurations that span banks with code similar to the parameter checking there ... ```c
#if PICO_PIO_USE_GPIO_BASE
uint used = (~config->pinhi >> 4) & PINHI_ALL_PIN_LSBS;
// configs that use pins 0-15
uint gpio_under_16 = (~config->pinhi) & (~config->pinhi >> 1) & used;
// configs that use pins 32-47
uint gpio_over_32 = (config->pinhi >> 1) & used;
uint gpio_base = pio_get_gpio_base(pio);
invalid_params_if_and_return(PIO, gpio_under_16 && gpio_base, PICO_ERROR_BAD_ALIGNMENT);
invalid_params_if_and_return(PIO, gpio_over_32 && !gpio_base, PICO_ERROR_BAD_ALIGNMENT);
I will also consult with management about the timing of the impl.
onyx hinge
in fact I guess we can just check for return of PICO_ERROR_BAD_ALIGNMENT and say "you can't mix pins GPIO0..15 with pins GPIO33..47"
tulip sleet
@onyx hinge how do you feel about working on this? Maybe it is a quick job for you, but also you may be taking time off?
onyx hinge
@tulip sleet I'm working regular-ish hours through the end of this week anyway. but I'm not sure where the priority lies between this & my pi5 rgbmatrix work.
tulip sleet
let's ask "management"
onyx hinge
I could sure spend an hour on it once my metro arrives later today though
tulip sleet
i have little experience with PIO or its impl, except to change neopixel timings
tulip sleet
@onyx hinge I had a misunderstanding above. I'll delete that. Don't bother to reply.
pin base is not GPIOBASE
onyx hinge
I may be confused too, we'll see
tulip sleet
@onyx hinge we could add an arg to specify GPIOBASE, defaulting to 0. So that would specify the low wraparound (or not) on the RP2350B too, if I understand correctly. Then it's just something to document. As to whether the PIO is set to 0 or 16, that's a matter of bookkeeping, and makes finding a free SM a little harder, but still possible.
Maybe on RP2350A, specifying GPIOBASE=16 should just be an error? I don't see that there's a use for that.
is the register avialable on RP2350A?
tulip sleet
maybe the arg is a bool: allow_high_gpios or use_high_gpios or some name like that.
onyx hinge
gpiobase: Literal[None, 0, 16] = None (meaning infer from arguments)
and only 0/None are accepted if you're not on 2350B
tulip sleet
what is the max number of consecutive pins that can be used?
onyx hinge
[and I think that None should result in 16 for the "does it wrap or not" case]
32
this will be constructor argument #41 😭
tulip sleet
so if the first pin is 8, what is the last pin if you ask for 24 pins on RP2350B. GPIOBASE has to be zero.
on RP2040 the last pin is 0
onyx hinge
from what I understand it would be 0 on RP2350B as well, (8+24)%32 + GPIOBASE
I wonder if that's accepted or rejected now. I honestly don't think that's something anyone DOES on purpose. unless you stuck yourself with a real bad pinout and that was your only way out
especially since what, 29 through 31 don't actually have IO pads on 2040? but they do on 2350B
and on pico W you'd be stomping on the wifi chip pins
tulip sleet
i don't see that said explicitly in the datasheet, hmm. I looked for all instances of GPIOBASE. I think maybe it could be inferred. Not saying you're wrong, just trying to find something clear.
onyx hinge
I think this wraparound case is "here's something the hardware happens to do", not "this is a honking great idea"
tulip sleet
it's a hack. They should have made the bit registers 64 or 128 bits wide in RP2040
(and not implement the upper parts)
onyx hinge
Relocate GPIO 0 (from PIO’s point of view) in the system GPIO numbering
I'm saying that based on this sentence (though I thought there was a figure)
tulip sleet
> Relocate GPIO 0 (from PIO’s point of view) in the system GPIO numbering
I'm sa...
is clearer, I think
onyx hinge
the PIO instance either addresses pins 0-31 or 16-47 based on \ref pio_set_gpio_base
onyx hinge
widening everything to 48 or 64 bits must have taken too much area or made them miss timings or something
tulip sleet
so "infer" above means if the base pin is >=16, then set GPIOBASE? Is the number of pins in the arguments or only in the program?
onyx hinge
"pio_claim_free_sm_and_add_program_for_gpio_range" oh goodie
tulip sleet
I am worried about inferring.
being ambiguous
does the PIO program itself know the base pin, or are its pin numbers all relative to the base pin set up in the SM registers?
i.e. do you have to write a different program if GPIOBASE is 16 vs 0?
onyx hinge
there are 3 PIO blocks each with 4 State Machines. GPIOBASE is per PIO. There are also the "{in,out,set, sideset}" bases & counts that are per state machine. When your program says "nop side 0", the affected pin(s) depend on the PIO GPIOBASE, the SM sideset base, and the SM sideset count.
same for an out instruction, it depends on the PIO GPIOBASE, SM out base, and SM out count.
wait gpio depends on PIO GPIOBASE but not any SM base register
tulip sleet
so the neopixel program is the same in all cases, you just need to set the base to be the neopixel pin and the count is 1
onyx hinge
yeah, and because there's just 1 pin there's no worry about the wrapping
say the actual pin is 24, you could put it in a PIO with GPIOBASE=0 and set the sideset base to 24, or GPIOBASE=1, sideset base 8
tulip sleet
so when you say "infer", how would it determine the GPIOBASE? Clearly if the pin# is >31, it's 16. But if the base pin is >=16 but the base_pin + count is >31, then it's ambiguous.
onyx hinge
Suppose my imagined gpiobase constructor parameter for StateMachine is the default, None, and you have a base < 32 and base+count >=32, then it would choose gpiobase 16, because I don't mind making the person who wants wraparound behavior ask for it
as a not entirely unconnected thought, how do you write in CircuitPython "if the pin's number is >= 16"? I don't think you can?
tulip sleet
I am thinking just make the default be 0, and allow 16 on RP2350B, and complain starting pin is out of range.
but don't provide None
onyx hinge
there's the connection, you found it
you're writing generic library code that a user will pass some pins into, how do you know what value for gpiobase?
tulip sleet
bleh, I see
onyx hinge
there are around 32 uses of StateMachine constructor in the adafruit bundle
gpiobase=None seems like the best shot at compatibility before/after adding support for these high gpios without adding complications to every StateMachine user. So maybe: don't add gpiobase= now, implement only what gpiobase=None would give you and add gpiobase=0/16 later if needed. along with "here's how to get a pin number, officially"
tulip sleet
re the huge number of args, many are defaulted, right? Could have a "Config" data class that doesn't take zillions of args, and gets passed to constructor, or has a factory function. Not sure that huge number of args is any better than setting a bunch of attributes.
or maybe we need a Python wrapper that makes the setup easier instead of putting it all in C
onyx hinge
and so this hypothetical person who had a bad design based on rp2040 that needed wraparound, re-spun it with an rp2350b, but didn't change the pinout and didn't otherwise use the new GPIO30 & GPIO31 (since wraparound is stomping these pins) can contribute support for gpiobase=0 to force their compatible wrapping for their very dubious design
Yes most of the arguments are defaulted
and many more are automatically determined by pioasm and passed via **pio_kwargs
tulip sleet
`gpiobase=None` seems like the best shot at compatibility before/after adding su...
this sounds ok, since the wraparound case is already rare, I think. I've never seen anyone do such a design.
onyx hinge
self._sm = rp2pio.StateMachine(
program.assembled,
frequency=125_000_000,
first_out_pin=first_pin,
out_pin_count=14,
auto_pull = True,
pull_threshold=14,
**program.pio_kwargs)
a typical constructor call uses only a few of the parameters
(and "more modernly", out_pin_count and following parameters can ALSO come from pio_kwargs; so it's not that bad from the coder's point of view)
tulip sleet
the main thing right now is just to support the usb host pins. another use might be sdio on the Sd card pins, which also spill into the upper pins
onyx hinge
though we don't have sdio support in circuitpython at all yet, so 🙂
our sdk does have pio_claim_free_sm_and_add_program_for_gpio_range so yay
tulip sleet
so you don't really even need to add the arg right now, you can infer from the base pin
and that would be documented and that's that for now
maybe I would say base pin must be >=32 to cause GPIOBASE=16
add the case like 16+10 would just default to the wrap behavior for now. Eventually someone will be using something like the PGA2350B and complain.
onyx hinge
turns out we already always rejected wraparound. ```#define NUM_BANK0_GPIOS _u(30) // pico-sdk rp2040
// our code
static uint64_t _check_pins_free(const mcu_pin_obj_t *first_pin, uint8_t pin_count, bool exclusive_pin_use) {
uint64_t pins_we_use = 0;
if (first_pin != NULL) {
for (size_t i = 0; i < pin_count; i++) {
uint8_t pin_number = first_pin->number + i;
if (pin_number >= NUM_BANK0_GPIOS) { mp_raise_ValueError(MP_ERROR_TEXT("Pin count too large"));
}
So base=8, count=24 would error when it got to checking pin 30so no worries about preserving that behavior yay!
tulip sleet
turns out we already always rejected wraparound. ```#define NUM_BANK0_GPIOS _u(3...
ha! and no one said "but why not?". home free
onyx hinge
"free" still is feeling like at least an afternoon's work but it's all good 🙂 not being tied by this past weird behavior I only dreamed up is good!
fleet hollow
I'm working on creating a new audio input class which works with the audiosample API. When debugging an instance of the class (with CIRCUITPY_AUDIOCORE_DEBUG enabled), I can run audiocore.get_structure(audio_in) and audiocore.get_buffer(audio_in) without any issue, but audiocore.reset_buffer(audio_in) throws the following error: "TypeError: '' object does not support 'protocol_audiosample'". They both call mp_proto_get_or_throw(MP_QSTR_protocol_audiosample, sample_obj), but I don't know why reset_buffer specifically wouldn't work here. I'm thinking the empty object qstr might be a clue, but the object name works in other cases (ie: print(audio_in)). Any leads I could follow?
blissful pollen
I'm working on creating a new audio input class which works with the audiosample...
Is it getting a null object (hence not showing anything)? That would cause the exception
fleet hollow
Is it getting a null object (hence not showing anything)? That would cause the e...
It seems to be good on the python side, and the other functions (audiosample_get_buffer_structure and audiosample_get_buffer) follow a similar path. I currently don't have a way to debug CircuitPython firmware line by line, but that may be what's needed.
blissful pollen
If the code is somewhere I could take a look (likely later today). My most common method is just using printf debugging too, if you know how to do that in CP?
fleet hollow
fleet hollow
If the code is somewhere I could take a look (likely later today). My most commo...
I was able to figure out how to use mp_printf(&mp_plat_print, ...). I'm surprised I hadn't used it before, tbh.
blissful pollen
I was able to figure out how to use `mp_printf(&mp_plat_print, ...)`. I'm surpri...
Years from now when I've forgotten everything else, that line will be burned into my brain.
fleet hollow
Years from now when I've forgotten everything else, that line will be burned int...
I've narrowed it down! I should be able to take it from here. Thanks!
blissful pollen
I can't spot anything with a quick glance now, but my guess is for some reason it is coming in NULL.
And great!
manic glacierBOT
shut copper
I'm not entirely sure what the specific rules around directory name vs. filename...
Here is a proposed change to adabot to support this naming update https://github.com/adafruit/adabot/pull/391
lone axle
Here is a proposed change to adabot to support this naming update https://github...
Thank you! I've got a few other things to tackle first, but I've added this to my list to take a closer look at and test out locally.
onyx hinge
@fleet hollow huh did you figure out a cause? it works here (CircuitPython 9.2.1-19-gb5c38c2600-dirty on 2024-12-04) with a RawSample. I do notice that the tests don't seem to ever call audiocore.reset_buffer though. ```>>> a = audiocore.RawSample(b'123')
audiocore.get_structure(a)
(1, 0, 3, 1)
audiocore.get_buffer(a)
(0, <memoryview>)
audiocore.reset_buffer(a)
@onyx hinge The problem was with the specific class I was working on. I have already fixed it (but stumbled onto other issues, of course). Ended up being a bad pointer when setting up audio dma. 🤷
onyx hinge
<@426200547024961539> The problem was with the specific class I was working on. ...
OK, I won't worry about it then!
fleet hollow
OK, I won't worry about it then!
By the way, I know you're pretty busy, but if you ever want to go over the audioeffects updates that are waiting for review, I'd be happy to dive in.
onyx hinge
By the way, I know you're pretty busy, but if you ever want to go over the audio...
I know - thanks for being understanding of that.
blissful pollen
By the way, I know you're pretty busy, but if you ever want to go over the audio...
I'm on vacation (just at home) after Friday so feel free to poke me if you need reviews. I should have more time.
fleet hollow
I know - thanks for being understanding of that.
Of course! And great work on the RP1 PIO updates. That's huge. 😮
olive vessel
slowly but surely I am making progress on this port for an STM32L433-based board. At @tulip sleet's prompting I connected an STLINK and managed to step debug my way towards solving a few issues with the flash setup and pin mapping. I have reached the end of main.c and it looks like CircuitPython is running with no faults, but I am not seeing a CIRCUITPY drive on my machine. I've enabled CIRCUITPY_USB_MSC so I am not sure what's up, esp since I'm not seeing a fault and the device isn't in safe mode. My port is here: https://github.com/blues/circuitpython/tree/blues_cygnet/ports/stm/boards/cygnet. Is there anything I've missed in mpconfigboard.mk or any other reason this would be the case?
manic glacierBOT
onyx hinge
hmmm gcc doesn't seem to warn about this. There's a diagnostic -Wnarrowing but that's only in C++ for some reason. int32_t f(void) { uint64_t result = ~0ull; return result; } Ah, it's -Wconversion but this throws tons of errors so I can't enable it
tulip sleet
hmmm gcc doesn't seem to warn about this. There's a diagnostic `-Wnarrowing` but...
hunh, that's too bad :/
jaunty juniper
I was thinking how about switching the RTD theme to a dark mode one ?
onyx hinge
I was thinking how about switching the RTD theme to a dark mode one ?
@jaunty juniper can it be made to obey the browser preference?
seems that no, not for sphinx_rtd_theme. https://github.com/readthedocs/sphinx_rtd_theme/issues/224
blink and that issue will be 10
onyx hinge
hmmm gcc doesn't seem to warn about this. There's a diagnostic `-Wnarrowing` but...
it's making the change bigger but I'm introducing structures to hold the masks, which forces me to touch every location (since conversion between scalar and structure is always an error)
I think I'm getting close. it builds again and "low neopixel" still works.
e.g.,
for (size_t pin_number = 0; pin_number < NUM_BANK0_GPIOS; pin_number++) {
- if ((pins_we_use & (1 << pin_number)) == 0) {
+ if (PIO_PINMASK_IS_SET(pins_we_use, pin_number)) {
continue;
}
Thank you for working on this
jaunty juniper
seems that no, not for sphinx_rtd_theme. https://github.com/readthedocs/sphinx_r...
yeah there's an extension to add a toggable dark mode, but that does not default based on the browser setting apparently... I guess browser extensions for forcing dark mode on websites exist 🤷
slender iron
slowly but surely I am making progress on this port for an STM32L433-based board...
do you have printf logging? You can enable it within tinyusb. you can also use a usb sniffer to see how far it gets
onyx hinge
@jaunty juniper I'm not an expert but there are CSS selectors that access the browser preference. css @media (prefers-color-scheme: dark) { /* Dark theme styles go here */ }
@media (prefers-color-scheme: light) { /* Light theme styles go here */ }
olive vessel
Thanks @slender iron, I'll give both a shot.
lone axle
I have a new small change for the screenshot utility if anyone gets a moment and can review: https://github.com/circuitpython/CircuitPython_Library_Screenshot_Maker/pull/27 This adds adafruit_connection_manager to the list of libraries that will trigger settings.toml to be included in the screenshot.
tulip sleet
I have a new small change for the screenshot utility if anyone gets a moment and...
approved
lone axle
Thank you
onyx hinge
@errant grail I'll forget it before next meeting but please accept this hug report for nau7802 library!
errant grail
<@339922377746022402> I'll forget it before next meeting but please accept this ...
You’re welcome! Was very happy to see it become part of the Adafruit library family.
manic glacierBOT
Here is a screenshot illustrating the phase error for an RM2 response to a read of the gSPI test register:
manic glacierBOT
onyx hinge
@tulip sleet USB D+ is green right?
ah yes https://www.adafruit.com/product/4449 says so
tulip sleet
<@329766224093249548> USB D+ is green right?
green-colored wire? I don't know! and the cables might not be consistent. I was using
it's not working?
onyx hinge
no, no errors, but not working.
mortal kernel
White D+, green D- is standard coding.
tulip sleet
maybe the production description is wrong
onyx hinge
most likely the code is wrong
tulip sleet
I was using this, which has embossed labels https://www.adafruit.com/product/3629
onyx hinge
oh! oh ! ```>>> pwr = digitalio.DigitalInOut(board.USB_HOST_5V_POWER)
pwr.switch_to_output(True)
list(usb.core.find(find_all=True))
[<Device>]
🎉
tulip sleet
is green D+ on that then?
onyx hinge
@tulip sleet yes
slender iron
Kicking off #CircuitPython2025! I'm excited to hear what everyone is looking forward to. https://blog.adafruit.com/2024/12/20/circuitpython-in-2025-circuitpython2025-python-circuitpython/
mortal kernel
I'm celebrating getting an RM2 to work with an RP2040. At this stage it's a hack pending resolution of an upstream issue spread across the Pico SDK and cyw43-driver. Having this working allows capturing gSPI activity, which is a small step forward on a couple of issues. Hugs to everyone here, I hope to "see" more of you in the coming year.
onyx hinge
@slender iron 🎉 to you as well re: zephyr
slender iron
🙂 Seems promising
mortal kernel
@slender iron ❤️ what you're doing with Zephyr.
onyx hinge
huh this is weird. my keyboard doesn't appear in the device list until I press a key on it... ```Adafruit CircuitPython 9.2.1-43-gaf0cdc4065-dirty on 2024-12-20; Adafruit Metro RP2350 with rp2350b
paste mode; Ctrl-C to cancel, Ctrl-D to finish
=== import board, digitalio, time, usb, usb_host
=== pwr = digitalio.DigitalInOut(board.USB_HOST_5V_POWER)
=== port = usb_host.Port(board.USB_HOST_DATA_PLUS, board.USB_HOST_DATA_MINUS)
=== pwr.switch_to_output(True)
=== print("Here are the devices (but believe in yourself):", list(usb.core.find(find_all=True)))
=== _=input("but believe in yourself and hit enter on USB-attached keyboard")
=== print("Here are the devices now:", list(usb.core.find(find_all=True)))
=== for dev in usb.core.find(find_all=True):
=== print(f"{dev.idVendor:04x}:{dev.idProduct:04x}: {dev.product} {dev.manufacturer}")
Here are the devices (but believe in yourself): []
but believe in yourself and hit enter on USB-attached keyboard
Here are the devices now: [<Device>]
046d:c313: USB Multimedia Keyboard LITEON Technology
I've tried adding a sleep as long as 8s but it didn't make a difference.
@slender iron does that ring any bell for you ^^ ? I'll test with a USB Host Feather (the RP2040 one) just in case I introduced this by mistake
slender iron
Keyboards are handled by tinyusb for us
Not sure how that interacts with the device list
@mortal kernel working to migrate us to a zephyr baseline so we can share lots of the common things
Make new platforms easier too
onyx hinge
@slender iron it "should" appear in the device list, as it has to be in that list so you can detach the "kernel driver"...
onyx hinge
OK, it's something that happens when you busy-wait but not when you time.sleep. I'll file an issue.
tulip sleet
@onyx hinge thanks from me on working on the high PIO stuff and figuring out so many angles on it.
nimble sparrow
Huh, when did heapq disappear
stuck elbow
Huh, when did `heapq` disappear
nimble sparrow
Yep, and it's various default builds - I have code going back a bit that used it but haven't had to for a while
stuck elbow
you will need to copy it into lib I suppose
nimble sparrow
It's definitely been enabled on rp2 builds in the past, so I wonder if I just missed a discussion about it being unstable on rp2350
It's no trouble, I can work around it for now
stuck elbow
I think it might have been moved in micropython, and then the change came to circuitpython with one of the updates
nimble sparrow
That'd make sense
fleet hollow
I'm on vacation (just at home) after Friday so feel free to poke me if you need ...
Hey @blissful pollen, I hope your vacation is going well so far! Here's the main PR that needs review if you have the time. It adds another effect to audiofilters, but the main update is that it addresses LFO ticking outside of synthio.Synthesizer on all audioeffects classes which implement BlockInput. I have a few other PRs that will likely need to be updated if this ends up merging. So, I think this is the one to get moving on first. https://github.com/adafruit/circuitpython/pull/9776 Thanks! ❤️
blissful pollen
Cool, I'll take a look at it the next day or two. Hopefully push me to make a couple changes I need to as well. I have to catch up on some of the discussion you guys have had.
blissful pollen
Hey <@228322543507865601>, I hope your vacation is going well so far! Here's the...
Caught up on the discussions. Have to wrap my head around the Block stuff. I added it on a suggestion (and like the idea) but haven't looked into it much.
I do get the pain with stereo/mono it drove me crazy when I was looking at it. (same with 16/8bit). I have thought for some effects is it worth making it only say 16 bit or only stereo. Is there enough use of the other cases to justify the complexity.
Also this may have been mentioned but when I was working initially a lot was based on AudioMixer which I realize is fairly old at this point. There could be a point where we take a step back at the whole audio system, but I don't want to bite off something too huge / potentially breaking.
fleet hollow
Caught up on the discussions. Have to wrap my head around the Block stuff. I add...
There's discussion of removing the single channel output feature since it's only used on a small handful of boards (which I think could split the interleaved data at the end of the chain). I think that would simplify the stereo situation greatly. This PR fixes a few of the issues with stereo within audiodelays.Echo, https://github.com/adafruit/circuitpython/pull/9876, but I'd rather focus on get the distortion one out first in order to get LFOs squared away.
