#circuitpython-dev
tulip sleet
@slender iron do you have a few minutes for tinyusb q's?
slender iron
yup!
onyx hinge
@lone axle I can grab the blinka section if you want
turbid radish
Nearly 12,000 subscribers worldwide
Yes, I rely on everyone to find good content for the community
onyx hinge
would love to hear more about the plexiglass / acrylic hand techniques, I've done VERY bad jobs the few time I've tried it myself (cutting, not even bending)
errant grail
would love to hear more about the plexiglass / acrylic hand techniques, I've don...
I've been taking photos during the process in hopes to write a Playground Note on the technique. In this case KiCAD was used to incorporate one of LadaAda's Eagle PCB designs for the 3.5" TFT FeatherWing into a printed paper pattern that is then glued onto the plexi prior to machining. Getting an accurate pattern is key to the process.
onyx hinge
Thank you @lone axle for hosting the meeting. Happy to catch up with everyone!
errant grail
Thanks @lone axle !
uncut nexus
@tulip sleet Do you know what chip(s) Adafruit plans to use for Matter boards?
onyx hinge
@errant grail I'll look forward to probably seeing in the newsletter in that case!
lone axle
Thanks everyone, have a great week.
tulip sleet
<@329766224093249548> Do you know what chip(s) Adafruit plans to use for Matter ...
Espressif initially. I am testing on a Metro S3
uncut nexus
Thanks!
slender iron
(matter over wifi not thread)
lone axle
Thank you! I've got it on my list to over. I've got a few other items ahead of it, but be able to in the coming days.
manic glacierBOT
lone axle
Here is the notes document for next Tuesday’s CircuitPython Weekly Meeting. It is at the normal time of 11am Pacific / 2pm US Eastern here on Discord, but is on Tuesday instead of Monday. 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/1BTOyxonN09KjzL7GsxcN7HMlbOn4yzkcDvRoX4l-op0/edit?usp=sharing
lone axle
Here is the notes document for next Tuesday’s CircuitPython Weekly Meeting. It i...
I should have thought to ask about this in the weeds, but it slipped my mind.
Has anyone else received an email from a supply chain risk manager at NASA about the library bundle repo?
I received this the other day and then a second message following up on it from the same sender today.
Would there be a good person or place to direct this individual to for the information they're interested in about the bundle?
tulip sleet
I should have thought to ask about this in the weeds, but it slipped my mind.
...
send this on to management (pt)
stuck elbow
nasa is using circuitpython?
wraith crow
The ability for auditors to generate work always amazes me....
solar whale
nasa is using circuitpython?
I have used it to support NASA projects.
"Ground Support Equipment" -- not in flight
slender iron
I should have thought to ask about this in the weeds, but it slipped my mind.
...
I got one for the core from a different person. Seemed legit based on linked in. Didn't spend too much time on it
tiny peak
tannewt, I wrote up a playground guide walking through how to run the Zephyr shell with QT Py ESP32-S3 over either USB serial or BLE NUS with the Bluefruit Connect app. It takes a patch to get the BLE thing building, but it's an easy two-line fix.
jaunty juniper
(the frozen libraries versions have only been wrong for I assume 2 and a half year without anyone noticing, it's not an emergency 😄 )
tulip sleet
@lone axle did you see those emails from readthedocs about missing sphinx.configuration in the Bundle doc build?
lone axle
<@382939733107408897> did you see those emails from readthedocs about missing `s...
I don't see any messages about that.
tulip sleet
I don't see any messages about that.
I just forwarded one.
lone axle
I just forwarded one.
Thank you. I'll check into this and submit a fix. I wonder if / how many other libraries will have that issue
tulip sleet
that's the only one I got, but I'll forward any further ones.
lone axle
@tulip sleet I opened https://github.com/adafruit/Adafruit_CircuitPython_Bundle/pull/489 to add that configuration to the bundle. However I do think we'll need to patch the rest of the libraries as well.
I'm not sure why the others haven't generated similar warning messages to be sent, but we don't have that config in our yaml for libraries either. I found this failed build for the same reason https://app.readthedocs.org/projects/adafruit-circuitpython-mpu6050/builds/26830865/ on a library that had a release made very recently. Strangely it currently has "stable" that is newer than "latest" which I don't understand. But I think our banner tag in the readme files points to latest so it shows green even though the most recent stable build failed.
tulip sleet
<@329766224093249548> I opened https://github.com/adafruit/Adafruit_CircuitPytho...
thanks - replying elsewhere
tulip sleet
@lone axle I stumbled on this while looking for something else. The gpio part of the generated file listing screenshot looks wrong, and it also includes a .md: https://learn.adafruit.com/adafruit-pyportal/internet-connect
also the Project bundle is named examples.zip; I don't remember that being the case.
lone axle
also the Project bundle is named `examples.zip`; I don't remember that being the...
I think the examples.zip name is because it's pointing to an example script file in a library repo rather than a project file from the learn repo. The gpio part definitely seems wrong, I'll look into that, I'm not sure where it came from at first glance.
lone axle
<@382939733107408897> I stumbled on this while looking for something else. The `...
I've opened https://github.com/circuitpython/CircuitPython_Library_Screenshot_Maker/pull/29 that fixes the esp32spi_simpletest.py screenshot by adding filtering on the subdirectory files. We already filtered main directory files down to a list of shown filetypes, the subdirectories weren't filtered previously though so gpio and it's files got included in screenshots for anything inside of esp32spi/examples/
jaunty juniper
@lone axle hey I have a question about the use of displayio with blinka (on a Pi computer or MCP2221). Can you display an image with adafruit_imageload ? I'm bumping into errors with adafruit_imageload on a BMP (one I can PR, another I don't understand yet), is it supposed to work ? (There might be versions shenanigans ?)
stuck elbow
I think on the raspberry pi blinka is using PIL for the images?
jaunty juniper
yeah is there a way to make a TileGrid from PIL for use with Adafruit_Blinka_Displayio ?
lone axle
On Raspberry pi displayio is intended to work via https://github.com/adafruit/Adafruit_Blinka_Displayio. It should work with adafruit_imageload as far as I am aware.
The MCP2221 I am less sure about, I don't have specific experience with those expander breakouts. But if it's abstracted by Blinka I would expect it should work the same I think.
stuck elbow
but we are talking about displaying things on the monitor connected to the pi, not on an SPI display connected to the pins, right?
lone axle
I think PIL is or was used internally in Blinka_Displayio when it takes data from TileGrid / Bitmaps and sends it to the display.
I am talking about displaying things on a hardware display connected via SPI or I2C
stuck elbow
oh, I see
lone axle
to display on the HDMI monitor you need something like: https://github.com/FoamyGuy/Blinka_Displayio_PyGameDisplay but it's not updated for the newest Blinka_Displayio
jaunty juniper
ok I'll make a fix for one of the problems and open an issue for the other
lone axle
ok I'll make a fix for one of the problems and open an issue for the other
was this the other issue?
Traceback (most recent call last):
File "/home/timc/Adafruit_CircuitPython_ILI9341/examples/imgtest.py", line 38, in <module>
image, palette = adafruit_imageload.load(f"testimg.png")
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/timc/venv_blinka/lib/python3.11/site-packages/adafruit_imageload/__init__.py", line 92, in load
return png.load(file, bitmap=bitmap, palette=palette)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/timc/venv_blinka/lib/python3.11/site-packages/adafruit_imageload/png.py", line 177, in load
bmp[x, y] = pal.convert(
^^^^^^^^^^^^
File "/home/timc/Adafruit_Blinka_Displayio/displayio/_colorconverter.py", line 173, in convert
self._convert(self._output_colorspace, input_pixel, output_pixel)
File "/home/timc/Adafruit_Blinka_Displayio/displayio/_colorconverter.py", line 198, in _convert
rgb888_pixel.pixel = self._convert_pixel(
^^^^^^^^^^^^^^^^^^^^
File "/home/timc/Adafruit_Blinka_Displayio/displayio/_colorconverter.py", line 210, in _convert_pixel
pixel = clamp(pixel, 0, 0xFFFFFFFF)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/timc/Adafruit_Blinka_Displayio/displayio/_helpers.py", line 27, in clamp
return max(min(max_value, value), min_value)
^^^^^^^^^^^^^^^^^^^^^
TypeError: '<' not supported between instances of 'tuple' and 'int'
austere acorn
@lilac scaffold @strange ravine
https://github.com/adafruit/circuitpython/pull/9962
I created a PR with a fix for what I identified as the H7 problem -- with this, the latest CircuitPython runs on the open MV H7 camera.
Please let me know if it works with the builds for your boards as well.
strange ravine
yeeeeeeeeeeeeeeeeeeeaaaaaaaaahhhhhhhhh ill try it in a bit tysm
jaunty juniper
was this the other issue?
```
Traceback (most recent call last):
File "/home/...
yep, InputPixelStruct.pixel is declared as an int but is given a tuple. Changing _colorconverter.convert so that it passes an int gets me to show the image
I also have crashes with adafruit_display_text if a Label goes off the display area now
the original reason I looked into that is somebody in help_with a few days ago noticed that adafruit_imageload can't do jpeg with blinka, because there's not jpegio. Is there a known way to make Tilegrid from say a PIL image for use with displayio, or maybe should jpegio be added to blinka or something ?
(once we fix the current state of things)
lone axle
the original reason I looked into that is somebody in help_with a few days ago n...
I would consider the addition of jpegio to blinka displayio the proper long-term solution. Ultimately it strives to be in parity with core displayio as far as I understand.
That said I do think it would be possible and fairly straight-forward on RPi to use PIL to load a jpg and then get it's data into a Bitmap object for use with TileGrid and Blinka displayio, but I don't know of any sample code that illustrates it.
jaunty juniper
yeah my solution would be to convert it to BMP with PIL on the fly and load the BMP, that seems like a good enough generic solution
strange ravine
<@380535630939029504> <@606988156654321679>
https://github.com/adafruit/circuit...
huge
its working now, at least from what i can see
its getting late so ill do more tests tomorrow
orchid basinBOT
lone axle
The patch for adding sphinx config to rtd.yaml file has been run. There were a few that failed actions afterward that I'll check in to. And there are 8 that didn't get patched because their config was newer and just slightly different enough that the patch didn't apply. I'll submit PRs to those instead of worrying about trying to patch them seperately.
It does bring up the question of a potential future patch though. In cookiecutter and the newest repos we use ubuntu-lts-latest inside of RTD build system. But in the older libraries they use ubuntu-20.04. It looks like April 2025 is the EoL for 20.04. The good news is moving to lts-latest means we shouldn't have to patch again in the future.
strange ravine
@austere acorn so ive tried circuitpython on 2 of my custom H7 boards and im getting, interesting results
on the first one it will work, kinda, the circuitpy drive takes a bit to load but is rather unstable, maybe its my computer.
on the second board it straight up doesnt work, unsure why.
ill keep testing
tulip sleet
The patch for adding sphinx config to rtd.yaml file has been run. There were a f...
I didn't know there was lts-latest -- that's not too much churn but still pretty current.
strange ravine
no yeah its a problem with either circuitpython or my boards coz same happens when i try to use them on my main computer
lone axle
I didn't know there was lts-latest -- that's not too much churn but still pretty...
I think their available options inside RTD are slightly different from actions at least. We use ubuntu-latest in the github actions containers I believe. I'm not sure if those end up on the same release cadence or not, but it will be nice to have them both onto some form of "latest" so that they're moving forward automatically even though it may mean we have to fix things for the new version occasionally.
tulip sleet
I think their available options inside RTD are slightly different from actions a...
ah, thanks for clarifying, I thought that was a github designator. I think ubuntu-latest in github does mean lts
lone axle
It seems like Github may be experimenting with or rolling out a new "sub-issue" feature. Never seen this in the UI before that I can recall
tulip sleet
@mortal kernel re your PR, let's do a more interactive conversation about this.
shut copper
https://github.com/adafruit/adabot/pull/394 . One liner to fix an exception in adabot on empty results from Read The Docs @lone axle I also did that change request for circup, but I assume you got the notification on that.
lone axle
https://github.com/adafruit/adabot/pull/394 . One liner to fix an exception in a...
Thank you! got both on my list to look at.
strange ravine
manic glacierBOT
tulip sleet
@slender iron @onyx hinge Thach gave me a fix for the 9.2.2 bug of ESP32-S3 not showing up on USB ports. I tested it and it works. It is temporary in the sense that it should be fixed by ESP-IDF upstream. I will make a PR for now. I can also release a 9.2.3 today with the fix (+ other recent rather innocuous PR's). Think I should do a release?
onyx hinge
seems like it would be a nice regression to fix.
tulip sleet
Yeah - half a dozen people have reported it, and maybe more have not
slender iron
<@252717193496756235> <@426200547024961539> Thach gave me a fix for the 9.2.2 bu...
yup, agreed it would be good to release
slender iron
This doesn't cover the STM issue though right?
tulip sleet
This doesn't cover the STM issue though right?
No, and I'm not even sure the cause is the same. There is a separate thing about the order of checking for AHBIDL and a delay: https://github.com/hathach/tinyusb/issues/2943. I will try that on STM32.
I tried that fix on ESP32-S3 and it didn't fix it.
tulip sleet
https://blog.adafruit.com/2025/01/17/circuitpython-9-2-3-released/
Highlights:
- Fix ESP32-S3 regression: did not connect properly to USB without an interposed hub.
- New boards: Adafruit Sparkle Motion and Mini Sparkle Motion, M5Stack StampS3.
orchid basinBOT
spare jacinth
looking good, lets try and build it locally 🙂
i remember tinyuf2 worked funny on my board a year(ish) ago, though
small cypress
@slender iron Hey, while I didn't see all of the stream, I think I remember you saying something about how it might be useful if you could see logs from your HomePod mini when it comes to CircuitMatter work? Did you work out how to do that? While I can't work out how to get it to throw logs into the macOS Console app like other Apple devices do, Apple have logging profiles you can install from here and then you can get it to ouput logs and export them somewhere. Have a feeling you'd probably want the "HomeThread (HomePod)" one, plus maybe the general HomePod one. You can't quite watch them in real time, but it's probably better than nothing.
https://developer.apple.com/bug-reporting/profiles-and-logs/?platform=tvos&name=homepod
slender iron
@crude blaze MBEDTLS_BUILTIN is on anyway
slender iron
<@252717193496756235> Hey, while I didn't see all of the stream, I think I remem...
thanks for the tip! I haven't been circuitmatter recently and got over the hump with their test iOS app instead
crude blaze
<@375455651050160130> MBEDTLS_BUILTIN is on anyway
hm ok, weird that you don't get the include forwarded then
orchid basinBOT
Typing any phrase in search bar on "Blinka" page and then removing it does not correctly display all the cards.
The underlying issue is using .split() on undefined, which prevents the function from completing.
Spotted by @FoamyGuy during th Saturday stream: https://www.youtube.com/watch?v=8czmvgN3Q70&t=360s
thorny jay
Checking Circuit-Python.org with my filter to see the new board: https://circuitpython.org/downloads?sort-by=date-desc I was suprise to see some old board on top such as: https://circuitpython.org/board/gb_m4/
tulip sleet
Checking Circuit-Python.org with my filter to see the new board: https://circuit...
it's sorting by dataset.date, it looks like. There is a date_added field, but I don't know what dataset.date represents.
tulip sleet
@lone axle re an "all libraries update" thing like we just did. People are noticing and a little confused (e.g. https://forums.adafruit.com/viewtopic.php?p=1042524#p1042524) that circup is updating all their libraries.
In the case of an infrastructure change like this, but no code change for the user, perhaps we should not do a huge number of releases, but simply push the change to main on the libraries. If/when I library needs a release update, the change will be incorporated.
What do you think?
lone axle
<@382939733107408897> re an "all libraries update" thing like we just did. Peopl...
If we don't do a release then every library will get flagged in various Library has new commits since last release within _____ report sections on https://circuitpython.org/contributing/library-infrastructure-issues which makes it a bit more difficult for me to do a real release sweep each week to catch any libraries that had PRs merged without a release made because I typically load that page and then open in new tab the ones that appear in those lists to work from.
It would be a bit difficult to try to work out which libraries actually need a release and which ones just have a change that doesn't need one.
If we can come up with some solution that makes it easy at a glance to know which libraries "really need" a release and which ones don't then I'd be okay with it. But with the current way the reports work I would prefer to make the release even though the change was minor / irrelevant for usage on device.
tulip sleet
If we can come up with some solution that makes it easy at a glance to know whic...
I understand, thanks. I guess we could put "NO-RELEASE-NEEDED" or something like that in a commit message, but that may complicate the reporting mechanism a lot.
lone axle
I understand, thanks. I guess we could put "NO-RELEASE-NEEDED" or something like...
Maybe have circup ignore updates to the patch number, and then use the convention of always increasing the minor number for any change that is intended for devices. Then we can create releases for stuff like that but circup won't prompt people to update from them.
tulip sleet
Maybe have circup ignore updates to the patch number, and then use the conventio...
I think I would rather not do that. It's not what semver is meant to do, and it would probably raise as many support questions as the current scheme. I'm OK with all the releases. We can just announce it in some way next time.
shut copper
If there was some metadata with the library updates circup could present that to help guide the users.
manic glacierBOT
manic glacierBOT
jaunty juniper
So, I have looked more at making adafruit_httpserver work with adafruit_esp32spi, like I did a few versions ago with UniversalSocket. But by implementing the changes directly into the ESP32 library.
Added missing methods from Socket:
- 3 I haven't implemented, and "work" by being empty:
setsockopt,listen,setblocking - 2 are implemented:
binduses the ninafwstart_server()acceptgets a client socket from that server (if any)
Changes had to be made to the ESP32SPI library:
- re-add the
socknumargument to theSocket()constructor. It's not in the python socket API, but we need to be able to wrap a Socket object around a socket number returned by ninafw when getting a client connection to the server. ESP_SPIcontrol.socket_writereturns the number of bytes written, so that:Socket.sendnow returns the number of bytes written (as does sendto) to match the python socket() API.SOL_SOCKETandSO_REUSEADDRhave been added to SocketPool.
One change had to be made to the httpserver library:
- try/except the import of the
sslmodule in theservermodule.
I don't know if Nina supports opening ssl server sockets btw.
crimson ferry
So, I have looked more at making adafruit_httpserver work with adafruit_esp32spi...
nina does support ssl (edit: oh, server, not sure about that); fwiw the upstream nina just added an accept https://github.com/arduino/nina-fw/pull/79
it would be cool to use native ssl with esp32spi, if that’s possible
though most boards w/o wifi or eth don’t have ssl enabled
jaunty juniper
I think another thing I'll need to do is check adafruit_httpserver for memory use, it's gonna be really tight (especially as the library grew quite a bit)
jaunty juniper
are there rules for the default content of alarm.sleep_memory ? On an ESP I get zeros everywhere, on a Matrix Portal I get random data
fun fact: if adafruit_binascii is present, httserver loads it, which loads the entire library in memory because of how it's architectured (that could be optimized too), if absent it just prints a warning (maybe only in debug mode) that it's necessary for websockets
crimson ferry
reductions in memory footprint of the http server lib could help getting it to run on pico w too https://github.com/adafruit/circuitpython/issues/9003
manic glacierBOT
manic glacierBOT
manic glacierBOT
dawn rampart
Hello - I am setting up a new board with a built in LCD. The driver is ST7789. I have setup the driver and display in the firmware board.c - and it uses a fourwire bus. But the board also has a microSD card and the display and sd card share the clock and MOSI lines. Every time I try to setup the SPI to the SD card, I get the error that the CLK is already taken by the display bus. If I try to point the sdcardio to look at the display bus, I get the error that that bus is a fourwire bus, not SPI - which all makes sense. Thoughts on how I might be able to use both the LCD and the SD card slot? Board files: https://github.com/bwshockley/circuitpython/tree/master/ports/espressif/boards/waveshare_esp32_c6_lcd_1_47 shows how the display is initialized.
And these are the pins in use:
jaunty juniper
Hello - I am setting up a new board with a built in LCD. The driver is ST7789. ...
look at the Feather TFT boards, I believe the display shares the same SPI as board.SPI, so it's setup that way, which allows to share the bus
dawn rampart
look at the Feather TFT boards, I believe the display shares the same SPI as boa...
So I tried setting this up same as the feather esp32s2. Again it seems to work to access my build in LCD, but when I try using or setting up an SPI connection for the sdCard, I get an error that the CLK pin is already in use. OR I get the error, "TypeError: spi must be of type SPI, not function"
jaunty juniper
So I tried setting this up same as the feather esp32s2. Again it seems to work ...
what's your code ?
spi = board.SPI()
Well - now I'm stumped. I shouldn't work late at night I guess. I'm pretty sure I was trying to initiate spi via board.SPI() earlier and getting the error that clk was already in use, but now it seems to work fine.
stuck elbow
Well - now I'm stumped. I shouldn't work late at night I guess. I'm pretty sur...
note that unless you call release_displays, it will work on the first run, but not after reload
jaunty juniper
that's a bulid with built-in display
stuck elbow
ah, sorry then
jaunty juniper
tulip sleet
<@&356864093652516868> We'll have our weekly meeting (delayed one day due to a holiday) in this channel and the CircuitPython audio channel in less than 1.5 hours, at 2pm US Eastern time / 11am US Pacific time. Add your notes: https://docs.google.com/document/d/1BTOyxonN09KjzL7GsxcN7HMlbOn4yzkcDvRoX4l-op0/edit?usp=sharing
blissful pollen
Lurking no notes
manic glacierBOT
short tendon
@tulip sleet that's so exciting! If you are working on the Nina firmware, let me know how I can help!
lone axle
Thanks for hosting Dan! Have a great week everyone 👋
errant grail
Thanks!
random junco
thanks for hosting Dan, have a good week all
tulip sleet
thanks for attending!
mortal kernel
@danh Any insight into #9983 or would you like me to keep digging?
tulip sleet
@danh Any insight into #9983 or would you like me to keep digging?
sounds ike the DMA channel never got set up (16 would be used to indicate "no channel"). Maybe we should check for out of bounds and complains. Sure, go ahead, if you find it entertaining 🙂
mortal kernel
sounds ike the DMA channel never got set up (16 would be used to indicate "no ch...
It's one of those bugs that's both interesting and informative. I'll keep going.
modest hazel
is pulseio not available for the micro:bit v2? I tried modifying the makefile to "force" it and it's throwing an error undefined reference to 'pulseio_module'
slender iron
is pulseio not available for the micro:bit v2? I tried modifying the makefile t...
try cleaning and rebuilding
modest hazel
I did make clean at the port/nordic folder level, should I be doing that at the top level?
Also, I'm on 9.0.0
slender iron
I did `make clean` at the port/nordic folder level, should I be doing that at th...
provide the BOARD= too
modest hazel
okay, let me give that a whirl...
jaunty juniper
pulseio is listed on the board page https://circuitpython.org/board/microbit_v2/
is it not in there ?
modest hazel
Yeah, that's why I was confused when it didn't work
okay, that seemed to have worked, thank you!
jaunty juniper
the module is there in CP 9.2.3, what was your issue ?
modest hazel
I'm not sure why it wasn't working, but using make clean BOARD=microbit_v2 first seems to have fixed it. 🤷♂️
modest hazel
Or - maybe it didn't. The build went okay, and I moved this
CIRCUITPY_PULSEIO = 1 #force it
outside of all of the conditionals in the makefile, but when I use help(modules), pulseio isn't in the list. 😦
spare jacinth
instead of help() i would just try and import pulseio, in case the list getting printed gets messed up somehow
but im going to assume that fails as well 😬
modest hazel
Yup - module not found when I did that just now
spare jacinth
also, since makefile is more of a text language, maybe your comment is getting the value get "read" as 1 , i would also remove the comment (and maybe the spacing with equals sign) to make sure it is a verbatim 1 and not some combination of whitespace and 1
otherwise the checks on other files to see if it is enabled might fail, i think i saw at some point an issue/PR for similar issue
but most likely the issue is something else that i dont know about, this is more of a quick test than anything
modest hazel
okay, let me give that a spin.
okay, this looks promising...import pulseio now seems to be working - and that was just from moving the comment, so I think you're theory was right.
And, this seems to have solved the bigger problem I was fighting, which was getting an ultrasonic censor to work with my micro:bit, so thank you!
jaunty juniper
oh 9.0.0 why did you need 9.0.0 ?
modest hazel
I could update (and would if I needed to), but this is a custom build and 9.0.0 was where I last forked it. No other reason really.
jaunty juniper
oh in ESP32SPI we have this about settimeout:
If value is 0 socket reads will block until a message is available.
But this is what Circuitpython does (and C python)
0means non-blocking.Nonemeans block indefinitely.
also it uses time.monotonic() which is not great for measuring timeouts for projects that might want to stay running for days maybe
dawn rampart
(see https://discord.com/channels/327254708534116352/327298996332658690/13310362...
Okay - so I made the changes suggested and the firmware built on my machine just fine - loaded the firmware and everything appears to work as expected - I can access the DISPLAY as well as SPI - and the SD card together. I opened a pull request to update the board.c file in adafruit's repo and now I'm getting an error there.
boards/waveshare_esp32_c6_lcd_1_47/board.c: In function 'display_init': boards/waveshare_esp32_c6_lcd_1_47/board.c:43:28: error: implicit declaration of function 'common_hal_board_create_spi' [-Werror=implicit-function-declaration] 43 | busio_spi_obj_t *spi = common_hal_board_create_spi(0); | ^~~~~~~~~~~~~~~~~~~~~~~~~~~ boards/waveshare_esp32_c6_lcd_1_47/board.c:43:28: error: nested extern declaration of 'common_hal_board_create_spi' [-Werror=nested-externs] boards/waveshare_esp32_c6_lcd_1_47/board.c:43:28: error: initialization of 'busio_spi_obj_t *' from 'int' makes pointer from integer without a cast [-Werror=int-conversion] cc1: all warnings being treated as errors
Pull 9987
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/1JpgiZ0ps-KhwsvriKFn21Ss0hKLDFCD1_NPpAN1C2Vc/edit?usp=sharing
modest hazel
Hi - is there a way to get at least some of the functionality in the micro:bit v2 library working in circuitpython?
slender iron
Hi - is there a way to get at least some of the functionality in the micro:bit v...
You'll have to be more specific. I don't know what that library can do
modest hazel
You'll have to be more specific. I don't know what that library can do
Anything that can utilize the LED matrix, buttons, buzzer, accelerometer, etc. that is on the micro:bit
slender iron
led matrix is tricky I think because it is directly driven
buttons, buzzer and accel should be doable from CP
jaunty juniper
led matrix is tricky I think because it is directly driven
@modest hazel maybe we could adapt the _pew library to drive the led matrix on the microbit but I never got to actually look into it. It might be as simple as changing the pins, or as complicated as swapping port-specific calls if any...
this python code attemps to show a heart
modest hazel
I had thought of that specific library - my company modified it a bit to drive an RGB8x8 SPI matrix, but we were using the Pico and i took advantage of the second processor to just handle it in the background. I don't think I have that luxury on the micro:bit
It's not an urgent need by any means, just a nice-to-have at this point. Ideally I'd like to drive it in the C-level code and just expose an API.
jaunty juniper
it runs on the pewpew which is SAMD21, so I think it should work at least as well 🤷
modest hazel
Yeah, I'm not great with the concurrent task stuff but I could probably figure it out given enough time... 😉 At least I've got a starting point!
We are still in the proof-of-concept stage so it's just a lot of exploring and low-hanging fruit stuff right now.
lone sandal
mortal kernel
We're getting close to CircuitPython issue/pull number 10,000. Any celebrations or prizes planned?
stuck elbow
We're getting close to CircuitPython issue/pull number 10,000. Any celebrations ...
closed/wontfix
jaunty juniper
I'm sure there'll be a blog post
fleet hollow
@blissful pollen @onyx hinge Thank you for all the help on the #9776! I'm very happy to have that one buttoned up. 🚀
onyx hinge
🚀 🔉
blissful pollen
<@228322543507865601> <@426200547024961539> Thank you for all the help on the #9...
Thanks for all the work you put into it! I found the lofi with a sine wave sounds a lot like original Mario 🙂
onyx hinge
I guess NES is triangle-"ish" ? https://www.linusakesson.net/music/family-bass/index.php
correction, one channel is a triangle
blissful pollen
@fleet hollow Which of the PRs did you want to look at next?
fleet hollow
I've got a comment on #9804 for @onyx hinge to take a stab at. I'll be working on updating #9876 to be merge ready. Both are related to changes implemented in #9776.
It's not ready yet, but I've been working a lot on https://github.com/adafruit/circuitpython/pull/9909 since I've got good hardware to test it on. It works for basic loopbacks, but otherwise I'm running into output underflow issues and problems related to audiomixer. I'd love some insight with all the buffer dma schenanigans, but it's pretty complicated.
blissful pollen
Okay cool, feel free to tag me on any you want a review/test on. I'll take a look at 9909 as well to see what I see
fleet hollow
I've also been working a bit on arduino-pico lately for similar functionality. The implementation is much simpler than CircuitPython, but there might be something to glean from it. https://github.com/earlephilhower/arduino-pico/pull/2775
fleet hollow
@blissful pollen I've got #9876 merged back up to main. Feel free to review/test when you get the time. 🙂
jaunty juniper
the current frozen version of requests has a bug where some sockets are not closed when some kind of exceptions happen (fixed in https://github.com/adafruit/Adafruit_CircuitPython_Requests/pull/205)
can we update it, and maybe other libraries, and make a release ?
oh that could make all of them with the RTD updates, but requests does need the fix
viscid trench
Is there a way to enable compression functionality in uzlib in CP?
I'm looking at the CIRCUITPY_ZLIB option, but i think that only enables decompression..?
austere acorn
Is there a way to enable compression functionality in uzlib in CP?
I'm looking a...
not implemented (yet?) https://docs.circuitpython.org/en/latest/shared-bindings/zlib/
viscid trench
Oh yeah.. Just noticed the compressor in https://github.com/adafruit/circuitpython/tree/main/lib/uzlib but it would still need an implentation
blissful pollen
Oh yeah.. Just noticed the compressor in https://github.com/adafruit/circuitpyth...
I initially did some work on decompress but a couple years ago. If you do have any questions feel free to ask and I can see what I recall. I don't have any immediate plans to put compress in myself though.
tulip sleet
Is there a way to enable compression functionality in uzlib in CP?
I'm looking a...
are you compressing to a file or you need to send compressed data over a web request?
mortal kernel
@danh Here are some thought on RAM log I was hoping you could give me some feedback on:
I've been thinking about the "cooking" of line endings ('\n' -> '\r\n') that happens on some (but not all!) serial console output. I think it would be helpful to make whether or not line endings get "cooked" an attribute of the serial destination (UART, tud, BLE, etc.) instead of trying to get it right in the string to be printed. I'd like to see all serial output go through a single simplified path.
Related, it may clarify/simplify messages (including logging) to have three virtual destinations: STDOUT, STDERR, and STDLOG. Each destination would have one or more default physical destinations that could be adjusted at run time.
A feature of RAM logging is the ability to recover the log after a complete system crash. This works by walking the tlsf free blocks early in start up and reviving the ramlog_t object based on a magic number and its complement. For all ports except Espressif it's trivial to get a pointer to the tlsf heap. Espressif has its own modified version of tlsf that is buried deep in IDF that's problematic for this feature. Is it worth digging deeper into Espressif now or leave this until later?
jaunty juniper
@tulip sleet can we make a release with updated frozen libraries soon ? To get that fix in for requests (on portal boards typically) by updating 4.1.4 -> 4.1.9: https://github.com/adafruit/Adafruit_CircuitPython_Requests/pull/205
tulip sleet
<@329766224093249548> can we make a release with updated frozen libraries soon ?...
Yes, I am going to do a 9.2.4 soon with other necessary fixes. I’ll update frozen libraries too. Usually I don’t due to possible regressions, but this is important. Thanks for noticing this
jaunty juniper
thanks 👍
tulip sleet
A feature of RAM logging is the ability to recover the log after a complete syst...
I think this is a good idea. You could do a scan through all of RAM looking for the magic number boundary markers without walking the free list explicitly. Then you doon’t have to mess with tlsf and are also more robust against smashes that might make scanning the list unreliable
mortal kernel
I think this is a good idea. You could do a scan through all of RAM looking for ...
The reason that I start by scanning the tlsf free blocks is that I use a call to tlsf_malloc_addr() to revive the ramlog_t object. Otherwise, I have a problem of what to do with the ramlog_t once I've found it. If I call port_malloc to allocate space to copy it into, I run the risk (especially with a large ramlog_t in PSRAM) of having the newly malloc'ed area overlapping the ramlog_t I want to recover, possibly even corrupting it with block meta data from a split.
Also, the Esp hardware appears to sparsely populate RAM blocks in a larger address space, so without a cue from IDF/tlsf I won't have a clue of where to scan.
tulip sleet
The reason that I start by scanning the tlsf free blocks is that I use a call to...
So how about instead of using the regular heap, allocate a single ramlog object using port_malloc(), using a size specified in settings.toml ? If no size is specified, then ramlog is not available. That will get the object off the regular heap. If the allocation is done early, it will also be at the same place all the time. Or you could remember its location in a static. Or use a magic number to find it and its size.
you have to look at the existing object to see if it was in use before initializing it.
If the ramlog object was in use (not "closed"), or something like that, then when code.py or boot.py starts, it would be available to inspect, but writes into it would not be available (until explicitly reset). DOes that make any sense? What do you do currently with the ramlog on startup? Does it get copied to the filesystem?
mortal kernel
Thanks, Dan, for the great feedback! At this stage I've been using compile-time flags to select start-up behavior of ramlog (do nothing, attempt recovery, pre-allocate ramlog_t or start logging). Putting this into settings.toml makes sense. It will delay ramlog start until later in initialization, but I think leaving the compile-time flags in place will cover cases where logging is needed early in boot. CYW43 startup is a good example of a case where earlier ramlog startup is needed.
Here's the ramlog_t object to give you an idea of how it works at the lowest level:
typedef struct _ramlog_t {
uint32_t magic; // Magic id number
uint32_t c_magic; // ...and its complement
uint64_t last_log_time; // Last log message time in subticks (1/32768 second)
mp_uint_t log_buf_sz; // Size of the log buffer in bytes
mp_uint_t next_idx; // Index of next byte to write in the log buffer
bool logging_enabled;
bool wrap_enabled; // Wrap log buffer around when full
bool wrapped; // Log buffer has wrapped around since last reset
bool prev_nl; // Previous log byte was a '\n'
uint8_t log_buf[];
} ramlog_t;
On top of the low-level logging, there's a ramlog binding that encapsulates the ramlog_t object, a Python singleton object RamLog, and their associated control stuff. The RamLog object implements the stream binding, so it can be unloaded by Python code and can also be used as a target for Adafruit_Logging. The binding has methods to initialize/deinitialize, reset, enable/disable, enable/disable wrapping, and so on.
Here's an example unload():
import ramlog
import storage
ramlog_obj = ramlog.RamLog()
def unload(fname: str, reset=True, enable=True) -> None:
"""Unload the RamLog to a text file."""
# Test that filesystem is mounted rw
if storage.getmount("/").readonly:
print("Filesystem is mounted ro, cannot unload RamLog")
return
# Disable the RamLog to prepare it for unloading
ramlog_obj.log_enable = False
# Unload the RamLog
log_lines = log_bytes = 0
with open(fname, "w") as f:
while True:
line = ramlog_obj.readline()
if not line:
break
log_bytes += f.write(line)
log_lines += 1
print(f"Unloaded {log_lines} lines, {log_bytes} bytes to {fname}")
# Optionally reset/enable the RamLog
if reset:
ramlog_obj.reset()
if enable:
ramlog_obj.log_enable = True
broken falcon
i keep running in to really disappointing memory allocation errors with the pyportals. as soon as youve added some buttons, sounds, fonts and labels to make a cool interface, requests dont work anymore (30 kb free memory to fetch a 3kb json doesnt work, everybody says due to fragmentation). now, i have at least 30 pyportals. can they be upgraded with a esp32s3 or something with a lot of ram so these projects are possible. most importantly; is there anything else one can do to increase memory (like an extra chip or something)? or a quite deep dive into memory allocation function? or request itself?
My students are going to make an assignment were they present their AIO feeds (io.loop) , get historical feed data (http requests) and the current weather forecast. thats a nice versatile project, but it seems impossible at the moment. any tips?
stuck elbow
I suppose it would be possible to write a requests-like library that uses preallocated buffers instead of allocating them dynamically, just nobody did it yet
I think I also saw in here mentioned a json library that can parse it while it's streamed, without having to download it all at once
It was this, if I remember correctly: https://docs.circuitpython.org/projects/json_stream/en/latest/
slender iron
i keep running in to really disappointing memory allocation errors with the pypo...
freezing libraries you use can lower memory use too
(but increase flash use)
broken falcon
It was this, if I remember correctly: https://docs.circuitpython.org/projects/js...
well, i tried that rabbithole, but we never understood how to use it. the example isnt very good and nobody knows what the chunksize should be. maybe an example for fetching historical AIO data would be nice (to use in a sparkline, e.g.).
stuck elbow
I'm sure improving the documentation would be easier than soldering extra ram.
broken falcon
well, i tried that rabbithole, but we never understood how to use it. the exampl...
of this format: (im going to post this in the general chat as well.
{
"feed": {
"id": 1958829,
"name": "temp",
"key": "esp32s2.temp"
},
"parameters": {
"start_time": "2025-01-10T18:44:43Z",
"end_time": "2025-01-17T18:44:43Z",
"resolution": 60,
"hours": 168,
"field": "avg"
},
"columns": [
"date",
"value"
],
"storage": "agg",
"data": [
[
"2025-01-10T19:00:00Z",
"20.316816416666665"
],
[
"2025-01-10T20:00:00Z",
"20.3822235"
],
[
"2025-01-10T21:00:00Z",
"20.414748"
],
[
"2025-01-10T22:00:00Z",
"20.42719525"
],
[
"2025-01-10T23:00:00Z",
"20.414236416666668"
],
[
"2025-01-11T00:00:00Z",
"20.4240675"
],
[
"2025-01-11T01:00:00Z",
"20.42496075"
],
[
"2025-01-11T02:00:00Z",
"20.433898944444444"
],
[
"2025-01-11T03:00:00Z",
"20.461606541666665"
],...
Look here too: https://github.com/daggaz/json-stream
that's what the CP version is modelled after
manic glacierBOT
In the current implementation of the audiodelays.Echo.mix parameter, the source audio sample is first combined with the echo sample and then mixed again with the source audio sample using the value of mix parameter. This effectively only allows you to control the level of the effect audio and not the level of the source audio (see the following graph, red=source, blue=effect).
In this update, I...
manic glacierBOT
manic glacierBOT
manic glacierBOT
viscid trench
are you compressing to a file or you need to send compressed data over a web req...
Actually i was trying to optimize sending data over LoRa
onyx hinge
@slender iron do you have something queued up for #10000? otherwise, I have two minor PRs
onyx hinge
ding ding ding
also at some point we passed 5000 merged PRs. Amazing. https://github.com/adafruit/circuitpython/pulls?q=sort%3Aupdated-desc+is%3Apr+is%3Amerged [currently: 5143]
onyx hinge
huh there are only 9920 PRs+issues visible in https://github.com/adafruit/circuitpython/ -- there must be 80 that were hidden/deleted due to abuse. Or github burps and misses numbers sometimes.
jaunty juniper
why does that make me want to write a script that uses the github API find out which numbers are missing ?
manic glacierBOT
onyx hinge
🤯
midnight ember
@tulip sleet pinned google doc hasn't been updated from last week. i have a hug report to write. 🙂
onyx hinge
Here is the notes document for next Monday’s CircuitPython Weekly Meeting. It is...
@midnight ember [I also pinned it]
midnight ember
Thank you
tulip sleet
thanks for pinning it @onyx hinge
manic glacierBOT
tulip sleet
@onyx hinge This is an interesting, still in-process PR: https://github.com/raspberrypi/pico-sdk/pull/2209
mortal kernel
<@426200547024961539> This is an interesting, still in-process PR: <https://gith...
Thanks for mentioning this one. The CYW43439, IMO, is one of the least well understood parts of the Pico ecosystem.
onyx hinge
@tulip sleet I think we get this right in StateMachine _transfer -- we loop while dma_channel_is_busy. after the loop we clear the stall bit, then way below if requested we wait until the stall mask is set.
related: you can't yet access fdebug on rp1 via the kernel ioctl api: https://github.com/raspberrypi/utils/issues/114
mortal kernel
@onyx hinge Did you have an chance to look into potential PIO clock jitter?
onyx hinge
@mortal kernel you mean this old discussion? I didn't think about it since then.
mortal kernel
I'm interested in your thoughts on it. I did get a pair of Raspberry Pi Pico 2W boards. So far I've not had a failure with either a 2 or 3 PIO divider.
onyx hinge
This program is set pins, 0 / set pins, 1. mcu is RP2350B at 150MHz. Divisor is 2. The wave is square at 37.5MHz (150/4) as expected. It only moves around by +-1 scope capture cycle.
divisor 3 is also stable & square
divisor 3.5 is unstable, with pulses of different lengths.
2.5 is stable but not square.
(2+56/256) ~ 2.2539 is very unstable
import adafruit_pioasm
import board
import rp2pio
import microcontroller
DIVISORS = [2, 2.25, 2.5, 3, 3.5]
def cycle(seq):
while True:
yield from seq
program = adafruit_pioasm.Program("""
set pins, 0
set pins, 1
""")
for divisor in cycle(DIVISORS):
with rp2pio.StateMachine(program.assembled,
first_set_pin = board.A0,
set_pin_count = 1,
frequency = int(microcontroller.cpu.frequency / divisor)) as sm:
print(f"{divisor=} {sm.frequency=}")
input("Hit enter to cycle frequency")
```My test program@onyx hinge Thank you! I'm surprised that 3 is stable, but it is good news.
onyx hinge
divisor=2.25 sm.frequency=66551126
Due to rounding, I think this actually tested divisor (2+65/256)
real 2.25 has a very short repeating pattern of alternation
divisor=2.249 sm.frequency=66666666
afk for a grocery (coffee beans) run
mortal kernel
In combination with the RM2 (but not the older CYW43439) any instability in the clock around 22 MHz results in the gSPI interface oscillating between CPHA=0 and CPHA=1 on CYW43 to host responses. Ouch!
slender iron
<@252717193496756235> do you have something queued up for #10000? otherwise, I h...
Nope, all good
onyx hinge
@lone axle @slender iron will you be able to read core & libraries sections?
lone axle
Yes, I am.
onyx hinge
<@&356864093652516868> good morning! We're about 2 hours from our weekly meeting. Take a moment to add your notes to https://docs.google.com/document/d/1JpgiZ0ps-KhwsvriKFn21Ss0hKLDFCD1_NPpAN1C2Vc/edit?usp=sharing then if you can join us in the CircuitPython voice channel at 2PM ET/11AM PT. We love hearing what you're up to!
slender iron
<@382939733107408897> <@252717193496756235> will you be able to read core & libr...
Yup! I'm around
onyx hinge
@slender iron not sure if I'm overlooking something but on Metro RP2350 USB stops working when the program exits, meaning that a USB keyboard can't be used in the repl if it's set up by boot.py. This is presumably because the USB host power pin gets turned off. Is there something sensible to be done about this? One thing that crossed my mind is to add a "host power" pin to the usb_host.Port constructor, which can always be None if there's not a power control...
slender iron
<@252717193496756235> not sure if I'm overlooking something but on Metro RP2350 ...
constructor makes sense to me. The other thing you can do for a built in board is have board_init() create and manage it
onyx hinge
@slender iron should I do that for metro rp2350?
slender iron
I think the connector has I2C lines we could test if they are pulled up
onyx hinge
we'd still want to support the user changing geometry or color depth but I guess you can release_displays()....
@lone axle https://github.com/adafruit/Adafruit_Blinka_Raspberry_Pi5_Piomatter/
force display: add to /boot/firmware/cmdline.txt video=HDMI-A-1:640x480M@60D (other settings are valid but this is the one I tried)
lone axle
Thank you!
onyx hinge
please suggest modest improvements to piomatter, it is very barebones at the moment and not super friendly.
jaunty juniper
is there code anywhere to read a keyboard (let's say classic 6KRO for starters) with USB host, like there is for midi ? (like adafruit_usb_host_midi)
tulip sleet
@onyx hinge notice I copied the raw markdown from the report and did "Paste as Markdown" to get it as googld docs formatted text in the notes. It will go back to Markdown when you do File->Download as markdown
slender iron
is there code anywhere to read a keyboard (let's say classic 6KRO for starters) ...
circuitpython should automatically inject the keystrokes as serial input
jaunty juniper
circuitpython should automatically inject the keystrokes as serial input
that's not what I'm talking about, I just want to fully read the keyboard, including all modifier keys, combinations, etc. Like you would to make a USB-to-BLE converter or HID remapper thing.
The "kernel driver" delivers strings and also acts on actions like ctrl-C. It's an interesting feature, especially with set_user_keymap() (even though it doesn't do unicode characters), but it's no way to read the actual keyboard inputs.
(There's also a bug I yet have to open an issue on where reading with input() or sys.stdin gives a UnicodeError when pressing left-alt + another key.)
I started writing code to read a keyboard HID reports (on host USB Feather), and it mostly works but I still have more work to do on it, and i'm afraid of missing key changes when typing a little fast. I kind of wish there was a C keyboard driver that delivers a queue of events with timestamps (for each report change), keypad style.
slender iron
that's not what I'm talking about, I just want to fully read the keyboard, inclu...
You should be able to detach from the internal driver and then manually do it. Something like keymap makes sense too.
austere acorn
agreed! 😄
smt32H750 chip, with 1 MB ram but only 128kB flash
https://electro-smith.com/products/daisy-seed
All my work was getting CP to work out of the external QSPI flash
Thanks @slender iron !
I wanted to make sure if it's worth continuing, documenting and publishing my work for now.
Gotcha!
I will check it out!
People on the daisy seed discord keep asking about python support, but with the currently missing audio modules in CP, not sure how useful it would be anyway....
slender iron
I'd hope synthio would "just work" once enabled
mortal kernel
Thanks all. See ya later.
slender iron
errant grail
Thank you!
lone axle
Thanks for hosting Jeff, Have a great week everyone!
onyx hinge
👋
slender iron
People on the daisy seed discord keep asking about python support, but with the...
This is the DK I got for it because zephyr has a board def (that flashes to internal flash): https://docs.zephyrproject.org/latest/boards/st/stm32h750b_dk/doc/index.html
slender iron
I will check it out!
And this person seemed to be trying to get XIP working: https://github.com/zephyrproject-rtos/zephyr/discussions/68729
I'd expect SDRAM to "just work"
austere acorn
And this person seemed to be trying to get XIP working: <https://github.com/zeph...
I need to look into how Zephyr works ... does it have a file system that needs to be writeable? How to edit the python code?
XiP was one thing to get going, but that makes the QSPI flash read only, and I'm not sure it can even swap modes during execution.
slender iron
zephyr doesn't have a filesystem itself usually
circuitpython needs it though. it is tricky to do both because all of the flash code does need to be loaded into ram (or maybe we figure out how to do split builds.)
manic glacierBOT
manic glacierBOT
onyx hinge
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/1XVbpKF0WlpLap2RIqy4IgeEOFLTUaBdCRs_qWGoj1Qs/edit?usp=sharing
onyx hinge
something I tried that doesn't seem like it's going to work out: I noticed that circup takes a bit of time after downloading the bundles to unzip them. I thought maybe it would be faster if it just extracted files from the zips as needed. And Python has zipfile.Path which can work similar to pathlib.Path...
but iterating over files inside the zip file seems to be a lot slower than iterating over them in the filesystem, to the point that you might as well just unzip them.
though there could be something wrong with my WIP code too
manic glacierBOT
manic glacierBOT
jaunty juniper
something I tried that doesn't seem like it's going to work out: I noticed that ...
I've been wanting to make circup use the json file available with each bundle release, which would remove the need to iterate over files, both to get version numbers and dependencies
(and have the json be in the zip)
tulip sleet
@slender iron when you are available could we talk about C3 problems? Thanks. Maybe you have some ideas:
- C3 DEBUG=1 builds in 9.2.1 or later (at least, could also be earlier) don't work at all. They crash early with
Guru Meditation Error: Core 0 panic'ed (Instruction access fault). Exception was unhandled. - enabling CIRCUITPY_BLEIO with ESP-IDF 5.3.2 causes a boot loop. Does not crash with 5.3.1 -- I did not test the functionality. I've been looking at the
REGISTRATION_FUNCTIONSand any changes in the CMakefile.txt for the bt component -- nothing obvious yet.
1 is kind of hindering debugging 2 🙂
manic glacierBOT
mortal kernel
@slender iron I've been poking around in Zephyr source to see what BLE support might be there for RP2 and came up empty. Any pointers you can share?
jaunty juniper
so I have a freshly installed Feather Huzzah 32 where the Web workflow doesn't seem to run on 9.x (whereas a V2 for example doesn't have that issue), I don't think it is purposefully disabled, can someone who has the board check ?
slender iron
<@252717193496756235> when you are available could we talk about C3 problems? Th...
Yup! I'm at my desk now. Free except for a meeting in 45 minutes or so. (Also let ladyada know I'm free though too.)
slender iron
<@252717193496756235> I've been poking around in Zephyr source to see what BLE s...
I don't see any mention on the meta-issue https://github.com/zephyrproject-rtos/zephyr/issues/53810
tulip sleet
Yup! I'm at my desk now. Free except for a meeting in 45 minutes or so. (Also le...
np - lemme know here or elsewhere when you free up
slender iron
np - lemme know here or elsewhere when you free up
can chat now if you want
tulip sleet
can chat now if you want
if responds can interrupt
mortal kernel
I don't see any mention on the meta-issue <https://github.com/zephyrproject-rtos...
Looks like Zephyr has gone the AIROC WiFi route as opposed to RPi SDK, but no support for the BTstack (yet?).
slender iron
Looks like Zephyr has gone the AIROC WiFi route as opposed to RPi SDK, but no su...
I don't think it'd use BTStack itself. Zephyr has its own BLE host stack
orchid basinBOT
manic glacierBOT
manic glacierBOT
tulip sleet
CircuitPython 9.2.4 released!
Highlights of this release:
- Update frozen modules, including a fix for adafruit_requests
- audiofilters: add Distortion and implement LFO ticking
- Add canio support for i.MX boards
- Fix RP2 picodvi crash
- Fix STM USB crashes
https://blog.adafruit.com/2025/01/28/circuitpython-9-2-4-released/
manic glacierBOT
manic glacierBOT
mortal kernel
I don't think it'd use BTStack itself. Zephyr has its own BLE host stack
The naming is confusing. There's BlueKitchen's BTstack used by RaspberryPi's Pico SDK, but there is also Cypress/Infineon's BTSTACK that is used with their AIROC driver. Digging deeper, I can see that Infineon/Cypress has contributed an HCI extension driver that connects AIROC to Zephyr's HCI implementation making BTSTACK unnecessary.
manic glacierBOT
slender iron
The naming is confusing. There's BlueKitchen's BTstack used by RaspberryPi's Pic...
so AIROC does have BLE support in Zephyr?
mortal kernel
so AIROC does have BLE support in Zephyr?
As far as I can tell, it does. Take that with a grain of salt because I'm basing that on my reading of the code, not an actual working example. The BLE part of Zephyr on CircuitPython is coming into focus for me now. Zephyr has a generic HCI driver that interfaces by various means with a subset of the current zoo of BLE hardware implementations. CircuitPython will interface with this generic driver, obviating the need for port-specific BLE code. Am I seeing it?
orchid cliff
Hey guys - I have an "older" Metro M4 Express. I installed 9.2.4, and it seems to be missing vectorio I had 9.1.4 on it, which does have vectorio. The download page does correctly not show it as one of the modules now... No biggie I went back to 9.1.4
tulip sleet
Hey guys - I have an "older" Metro M4 Express. I installed 9.2.4, and it seems ...
we ran out of space, sorry
orchid basinBOT
onyx hinge
@tulip sleet so in https://github.com/adafruit/circuitpython/issues/10011 your conclusion is that one of the esp-idf internal watchdogs, which DOES enter safe mode, is firing? It is a bit confusing still.
(in the user's non-distilled application)
tulip sleet
<@329766224093249548> so in https://github.com/adafruit/circuitpython/issues/100...
that is the user's conclusion, yes, after extended wifi use, there is an internal watchdog
it is not my conclusion. But using micrcontroller.watchdog does not trigger safe mode
i agree it was confusing. Sometimes the user posted things that did not happen, which I misread at first
that is, the user said "sometimes I get this [below], but that did not happen under circumstance X"
slender iron
As far as I can tell, it does. Take that with a grain of salt because I'm basing...
Yup, that's my hope
orchid basinBOT
.. since this word is associated with the Feather ESP32 and is shown in large letters on the bottom of the V2 board.
slender iron
anyone know what shell is used by python on windows from asyncio.create_subprocess_shell()?
onyx hinge
@slender iron def _start(self, args, shell, stdin, stdout, stderr, bufsize, **kwargs): self._proc = windows_utils.Popen( args, shell=shell, stdin=stdin, stdout=stdout, stderr=stderr, bufsize=bufsize, **kwargs)I tracked through several layers, and I still can't tell you, but it's whatever windows_utils.Popen(shell=True) does.
windows_utils.Popen subclasses subprocess.Popen and passes shell= through transparently
slender iron
I think I'll jam a test script into our windows CI build to test it 🙂
onyx hinge
Changed in version 3.12: Changed Windows shell search order for shell=True. The current directory and %PATH% are replaced with %COMSPEC% and %SystemRoot%\System32\cmd.exe. As a result, dropping a malicious program named cmd.exe into a current directory no longer works.
This note in the docs makes me think it's cmd.exe. https://docs.python.org/3/library/subprocess.html#subprocess.call
tulip sleet
cmd.exe is pretty basic, not a unixy shell at all
slender iron
it looks like it does have > support though
I really need a zephyr on windows build CI job
blissful pollen
it looks like it does have > support though
My main dev machine is windows if you need to test / try anything out feel free to ping me
slender iron
thanks @blissful pollen. I'm getting a new windows-zephyr build going on CI
onyx hinge
slender iron
I can't remember how much of their display stuff we use
onyx hinge
not much I don't think. but maybe the content would be a clue how to use the qspi peripheral for displays in the first place. a trail of breadcrumbs to follow
orchid cliff
we ran out of space, sorry
No worries! Thank you 😀
blissful pollen
@fleet hollow just FYI hoping to look at those PRs soon / this weekend
fleet hollow
@blissful pollen It's mostly small stuff, so shouldn't be a big deal. Just getting that echo effect fine-tuned 😸
@blissful pollen By the way, I've been doing research into reverb algorithms lately. Seems like it's mostly a mix of comb and all-pass filters which require quite a bit of floating point calculations. I've successfully implemented freeverb (https://github.com/sinshu/freeverb), a public domain reverb effect from way back in 2000, on an RP2040 in arduino-pico. The most I could get out of it was mono output with a sample rate of 8000hz. Any more and I'd drop buffers. Even though it was a bit lo-fi, it still sounded really good. I think there's potential to fork that repo and convert as much of the calculations to 16/32-bit integers as possible. Granted, the RP2350 might handle it like a champ regardless.
blissful pollen
<@228322543507865601> It's mostly small stuff, so shouldn't be a big deal. Just ...
Great. I was looking at my chorus effect and realized I have to look at all the fixes since I started to pull them across as well.
blissful pollen
<@228322543507865601> By the way, I've been doing research into reverb algorithm...
I looked at freeverb too. The effect doesn’t seem too hard to put in place. Mostly just take ram and cpu. May need to warn people about sample rate / stereo if it really takes a lot. The teensy audio library has it built in as well.
slender iron
would it be weird to have files checked into the REPO that list the modules enabled on a board? that way the stub stuff doesn't need to generate it
slender iron
@tulip sleet have time to talk CP config this afternoon?
tulip sleet
<@329766224093249548> have time to talk CP config this afternoon?
I missed your ping somehow, but I'm still available.
jaunty juniper
would it be weird to have files checked into the REPO that list the modules enab...
is generating it a big hassle ? It is done to update cp.org too, which is a separate workflow I believe.
it would be nice to have a static source of that information other than a file in the cp.org repo
(or the RTD docs matrix)
tulip sleet
would it be weird to have files checked into the REPO that list the modules enab...
I think it needs to be generated automatically at some point, not by hand (too easy to get out of sync).
slender iron
the challenge is that it is based on the final zephyr board device tree
jaunty juniper
yeah that would require pre-commit check and something similar to make translate
slender iron
I'm thinking that I can autogenerate it, put it in boards/ and then when the CI builds the board it'll make sure it is the same
but for other uses it can just be read directly
(I'm figuring out how to fix the docs build)
without installing zephyr and doing a build
tulip sleet
is the device tree generated from CIRCUITPY_WHATEVER = ... settings?
slender iron
the device tree is generating CIRCUITPY_ setting equivalents
tulip sleet
the device tree is generating CIRCUITPY_ setting equivalents
I would hope it's the other way round, so we aren't depending on the device tree as ground truth
slender iron
device tree is what zephyr uses for config though
tulip sleet
but can we generate the device tree from some set of our own settings, and then after that let zephyr run its own config?
we don't have to use gnu make variables, we can use something else (though I would hate for it to be kconfig)
tulip sleet
e.g. a .toml or .json file generates a .mk file, and/or it generates the device tree
slender iron
I'm working to get the build generating a toml file with per-module info
I can even add comments for humans about why a setting is what it is
manic glacierBOT
orchid basinBOT
manic glacierBOT
manic glacierBOT
manic glacierBOT
jaunty juniper
Is the only way to change the partition table to change a board's sdkconfig ? To build a QT PY ESP32S3 with BLE I had to change mpconfigport.mk, which I expected, but I can then set CIRCUITPY_BLEIO=1 as an argument to make without having to change the board files. Having the same flexibility for partition tables would be nice
tulip sleet
Is the only way to change the partition table to change a board's `sdkconfig` ? ...
You can set FLASH_SIZE_SDK_CONFIG. See ports/espressif/Makefile for a lot of logic involving that variable. In most cases is it set with ?= so you can override it.
jaunty juniper
ah thanks, let me try that then
jaunty juniper
You can set `FLASH_SIZE_SDK_CONFIG`. See `ports/espressif/Makefile` for a lot of...
yeah FLASH_SIZE_SDKCONFIG=esp-idf-config/sdkconfig-flash-4MB-no-ota-no-uf2.defaults works thanks
combining CIRCUITPY_LEGACY_4MB_FLASH_LAYOUT=0 and UF2_BOOTLOADER=0 also works to select the 4MB-no-ota-no-uf2 variant
tulip sleet
yeah `FLASH_SIZE_SDKCONFIG=esp-idf-config/sdkconfig-flash-4MB-no-ota-no-uf2.defa...
The LEGACY stuff is mostly for C3, so it includes the no-uf2 (since C3 can't have UF2 anyway). It is tested in some nested ifs. All the S3 boards have legacy layout since they were defined before that was a choice
you could use no-uf2 but I think there's no real advantage to that
4MB-no-ota is fine