So it's not as much of a rush as I was initially thinking.
#circuitpython-dev
idle owl
tulip sleet
product is out of stock right now
idle owl
Right but that won't affect the guide. Or the folks that already ordered one.
tulip sleet
ah ok, it did ship some, ok. I will get ready for an rc.2 with new release notes (which are minimal and I can add to)
idle owl
Thank you!
@tulip sleet How are we looking for 8 stable? There's a thing I wanted included, but haven't had the time to do, and am not sure I ultimately will.
tulip sleet
<@329766224093249548> How are we looking for 8 stable? There's a thing I wanted ...
not sure what you mean by "How"? You mean what will determine we are stable? I think in the guide you could just refer to circuitpython.org and say download 8.0.0 or later.
idle owl
not sure what you mean by "How"? You mean what will determine we are stable? I t...
This isn't related to the guide I'm working on now. By "how" I meant do we have some idea when it's going to happen, or is it still a bit out.
tulip sleet
we really don't have more to add to 8. We are picking up things like new boards because they are standalone and are in the rc period. I think we are waiting for any showstopper issues to show up, and none have yet.
idle owl
we really don't have more to add to 8. We are picking up things like new boards ...
Good to know. Thanks!
still zephyr
@proven garnet I continued digging the problem of the missing __version__ information in the files. I made a list using a basic GREP search. here. Think we have more files missing the info... https://gist.github.com/jposada202020/3d20534385c434577b3e50cf7bfc28c1. warning, some __init__ files could be wrongly reported, as some libraries have empty __init__ let me know if there is something that we would need to work on. ☺️ thanks
jaunty juniper
oh there's an issue with mission __version__ ?
still zephyr
I did one this morning for displayio_layout I did not know all the files missing it at that time
And not sure if it matters or worth the effort to correct them 🙂
jaunty juniper
yeah a library only needs one file containing the version
still zephyr
I found trying to understand why circup was saying that displayio_layout needed a major version ergo last night question..
still zephyr
I only found 3 or 4 libraries that circup could not find the version and I think one of them was displayio_layout,
still zephyr
That is good, I think we will need to put a note somewhere here for folks to know https://github.com/adafruit/circuitpython/blob/main/docs/design_guide.rst
jaunty juniper
I'll PR an update to circup
still zephyr
Good Ill will make the PR for the design_guide and the three libraries that could be a quick fix. Thanks Neradoc!
jaunty juniper
before
❯ circup freeze
Found device at /Volumes/CIRCUITPY, running CircuitPython 8.0.0-beta.6-62-g7522522a0-dirty.
adafruit_displayio_layout==None
after
❯ circup freeze
Found device at /Volumes/CIRCUITPY, running CircuitPython 8.0.0-beta.6-62-g7522522a0-dirty.
adafruit_displayio_layout==1.19.10
yay 
still zephyr
sorry did not want to get you into a rabbit hole.. 😦
jaunty juniper
nah that's good
also note that putting the version and repo strings in more than one file uses more disk space than necessary
maybe we could have a pre-commit check installed by cookiecutter that looks if there's a valid __version__ ? Is there something like that ?
manic glacierBOT
blissful pollen
@lone axle I tried the IS31FL3741 on 8.0 rc.1 and like you it does not work. Not sure offhand why that would be
manic glacierBOT
onyx hinge
@tulip sleet I wonder if we shouldn't update ulab before final release of 8
tulip sleet
<@329766224093249548> I wonder if we shouldn't update ulab before final release ...
That is fine with me. Go ahead with a PR
onyx hinge
(I think 7531 is fixed)
tulip sleet
I too thought that was fixed
tulip sleet
@onyx hinge it looks like something changed out from under pre-commit last night. I did an innocuous PR merge and got this (ran twice to make sure): https://github.com/adafruit/circuitpython/actions/runs/4060983872
I have not had a chance to look at it in detail yet. Looks like black
never mind, MicroDev found it; I hadn't finished reading my email
onyx hinge
@tulip sleet is it related to this? The black version has changed:
- black==22.12.0
- black==23.1.0
ah because we pip-install black and don't pin it
tulip sleet
that is a lot of churn
onyx hinge
it is.
tulip sleet
lots of new white space. I like the """ on its own line; other than that, not sure why they did it
onyx hinge
https://ichard26.github.io/blog/2022/12/black-23.1a1/
If you’re surprised we’re doing this, our stability policy allows us to update the stable style in the new year. This minimises disruption while keeping the door for improvements open.
happy january
proven garnet
maybe we could have a pre-commit check installed by cookiecutter that looks if t...
Yeah that sounds like a good plan
Some kind of build CI or pre-commit check at least so we can catch it before release.
onyx hinge
@tulip sleet I'll hold off pushing this ulab thing until 7532 is merged
tulip sleet
i am mostly ok with the change, only silliness, it seems to add an extra blank blank line before methods/functions that have a # comment before the def line.
we might consider excluding the upstream test .py files, if they are not already excluded... will look at that
ok, never mind, it's not methods, it's that top-level functions are separated by two blank lines, not one, and it has nothing to do with comments
I am impressed it moved this """ onto the next line and preserved the comment style:
//| :param int rotation: The rotation of the display in degrees clockwise. Must be in 90 degree increments (0, 90, 180, 270)
//| """
Processing the inline comments via black is my doing and I feel like it ended up working pretty well.
tulip sleet
it did; I thought it was black only that fixed up those files, but it's the whole pre-commit process, yes?
onyx hinge
the tools/codeformat script does a number of sequential steps
one of them involves running a separate script, tools/black_bindings.py. It changes the file from being "C with Python inside //| comments" to being "Python with C inside ##| comments", blacks that Python file, then does the reverse comment-swapping to create its final output
I should also contribute the script that does the swap but doesn't run black; this lets you use normal Python indenting and syntax highlighting while working on the in-line docs
tulip sleet
you could put that in tools/
onyx hinge
I'm not sure how many folks would discover to use them, but it doesn't mean I shouldn't put it there
timid bolt
Hi all, I wrote another doc about async, this time with a different focus. As always, I would love to hear your thoughts about it.
https://docs.google.com/document/d/1xlzTEdP8ovTtz5Sx8voP8R9M8fY88LuyiWBYu0muiRc/edit?usp=sharing
tulip sleet
@onyx hinge your ulab PR just needs a make translate
slender iron
@timid bolt we have a "light sleep" until interrupt function here: https://github.com/adafruit/circuitpython/blob/135a2ba681d5cf1ef5330ec7380189a9ebc9c111/supervisor/port.h#L89-L91
onyx hinge
thanks for running that @tulip sleet ! I just got back
tulip sleet
i realized today you would have gone out; np
idle owl
@jaunty juniper @still zephyr I mentioned your issue to one of our Learn devs. A suggestion for now is to try enabling "Reduce motion" in your Learn Account Settings. This should stop the guide preview GIFs from animating. That said, it also stops all of the other non-critical GIFs from animating as well. There are a set that still run because they are considered "instructional". If nothing else, enabling this setting could tell you whether that is your issue or not. Log into your Adafruit account, go to "Account" next to your name in the top right, click "My Account" from the dropdown, then click on the "Account Settings" tab (in the black bar above "My Adafruit"), choose "Accessibility" from the left menu list, and then check the box.
still zephyr
<@199811382793863168> <@607375730610798634> I mentioned your issue to one of our...
Thanks I will try it. Thank you very much 🙂 I have already changed the settings,
idle owl
Thanks I will try it. Thank you very much 🙂 I have already changed the settings...
Excellent! Let me know how it goes.
timid bolt
When native code crashes (e.g., hits an assert) what is meant to happen in the REPL? Should it just hang or can it print the assert 🤞🥺
tulip sleet
When native code crashes (e.g., hits an assert) what is meant to happen in the R...
it resets and goes into safe mode. But asserts are not normally enabled in non-DEBUG builds
there are various reasons shown for safe mode
some crashes cause a hard reset, some do not. It also depends on the underlying layer: e.g., ESP-IDF may hard-crash due to watchdog timeouts
timid bolt
Looks like I'll need a jtag adapter. Do you have any recommendations on which one to get?
tulip sleet
j-link
verygood sw support
it comes with a cable, but get this breakout too: https://www.adafruit.com/product/2743 -- very convenient
timid bolt
any thoughts on this one https://www.adafruit.com/product/3839
tulip sleet
any thoughts on this one https://www.adafruit.com/product/3839
it doesn't have as complete coverage of the chips we use
manic glacierBOT
timid bolt
it doesn't have as complete coverage of the chips we use
there are more chips listed on black-magic.org than on the adafruit product page 🤷♂️, but I'll try to get the first
tulip sleet
they have added support for more chips, but the J-Link folks are extremely good at covering a wide range of chips.
J-Link is kind of the "standard". Expensive for commercial use, but the edu models are not crippled
they also respond well to bug reports
stuck elbow
on the other hand, a dap programmer is like $10, and you can flash a black magic probe on a $2 blue pill board
ornate breach
J-Link is kind of the "standard". Expensive for commercial use, but the edu mode...
The edu models also recently came back in stock but at a price increase ($19.95 —> $38)
From Segger directly that is
I recently picked one up
tulip sleet
STlink's and clones don't work properly on non-ST boards.
stuck elbow
stlinks are ancient
tulip sleet
at segger it's now $60, which is the price of the old full-sized Edu
stuck elbow
we have things like dapcat and daplink now
tulip sleet
stlinks are ancient
some people have been tempted to try to debug SAMD's with them, no go
ornate breach
Wow just noticed they’re $60 now
tulip sleet
still $38 at our store, so buy 'em while they're hot
stuck elbow
some people have been tempted to try to debug SAMD's with them, no go
I programmed my samd21 ones with them successfully
ornate breach
Guess I lucked out getting it for $38
stuck elbow
one weird thing with them is the vcc pin is output, not input, so you don't want a separate power for the device, like with jlink
or simply don't connect the vcc
ornate breach
It is interesting how the price of things is just jumping seemingly month to month
stuck elbow
It is interesting how the price of things is just jumping seemingly month to mon...
it's the invisible hand of the market
ornate breach
“The market”
stuck elbow
it's still a market, even if it's a monopoly
ornate breach
I know, just smoke and hand waving to make it seem more complicated to keep prices low
On one hand, I at least have tools now to debug properly
timid bolt
are these debuggers expensive because of ARM licensing?
ornate breach
are these debuggers expensive because of ARM licensing?
I’d guess that and increases supply chain costs
rugged spindle
so would a daplink work on all the boards? I've seen CMDIS-DAPLINK that do JTAG/SWD (and using openocd) from AliExpress for < $10
ornate breach
You could probably do most boards I’d imagine
tulip sleet
@timid bolt one thing to note is that we haven't succeeded in getting gdb to work all that well via J-Link or openocd on the Espressif chips. We tend to just use print statements to UART, via the ESP_LOG macros. There is a very limited number of hw breakpoints on these chips, and there seem to be other problems as well. gdb via J-Link works quite well on SAMD and nRF. Not sure about i.MX.
tulip sleet
are these debuggers expensive because of ARM licensing?
i am not sure that is it -- it's a very software-intensive support thing, and they support thousands of chips. Just look at the J-Link release notes.
tulip sleet
I don’t see why i.MX would have any issues with gdb
not inherently, just how well it works or not
timid bolt
does j-link work for rp2?
ornate breach
It seems NXP likes things well supported so hopefully that ends up being the case 🙂
ornate breach
does j-link work for rp2?
Yup
Generic M0+ last time I tried
They may have added it with the most recent update though
tulip sleet
you can also use one RP2040 to debug another -- they set that up, i think via openocd
ornate breach
Yeah, that’s true
You don’t even need another RP2040 to debug, you can use the other core to run the debugger
tulip sleet
if you don't want to buy anything now, you could just try the RP2040 debug stuff, which the creators use themselves, and for testing aysncio stuff, that should probably be fine
timid bolt
wow, is there a link to this rp2040 debug stuff?
ornate breach
Nice thing about the RP2040 is being able to run your debugger on the same chip without really affecting resources too much
tulip sleet
websearch for "picoprobe", and I htink skerr is talking about https://github.com/majbthrd/pico-debug ... but the repo is archived ?!
ornate breach
Straight from the Adafruit blog 🙂
Oh wait
Darn I thought it was different
Maybe it still works 🤔
blissful pollen
I love finding comments that give you hints that the code you are looking at was copied from elsewhere (which is exactly why I noticed it). OnDiskBitmap has the line // load the wave
ornate breach
Lol
timid bolt
picoprobe sounds great! I guess I'll try that before ordering a debugger. Thanks!
wraith crow
still $38 at our store, so buy 'em while they're hot
Well that window closed fast, looks to me like Adafruit just adjusted their price to match 🙂
ornate breach
Wow, that fast
stuck elbow
almost as if someone was listening...
wraith crow
I don't know but I image that Segger would notify their resellers of the price increase and perhaps require them to match it
tulip sleet
I think I saw it at $38 when I first looked on the segger site, and then it went up!
timid bolt
yikes, it was still sitting in my cart. I guess I shouldn't discuss these things here. 😅
the "pico-debug" project is not super convenient because it seems it consumes the usb port, so then I can't also run the python repl
tulip sleet
yikes, it was still sitting in my cart. I guess I shouldn't discuss these things...
we didn't rat on anyone 🙂
i think it's a coincidence
timid bolt
I guess I could tell CP to use the hardware uart instead of usb for "stdout", but I think I should just bite the bullet and get a hardware solution. Then I'll be set up to work on multicore CP 😉
ornate breach
The change happened around 5pm, likely when price changes/new stock would regularly be pushed
timid bolt
btw, just curious, what hardware do you use for debugging risc-v?
tulip sleet
print statements
ornate breach
OpenOCD has limited RISC-V support
Nothing Espressif though
Mostly just major RISC-V platforms
timid bolt
does the j-link adapter work with OpenOCD/RISC-V?
tulip sleet
the "driver" stuff mentioend there is Windows-specific. I don't think there'd be any issue on Linux
timid bolt
tbh, I'm usually fine with printf debugging, but in this particular case I'm getting random crashes, (probably in the gc I guess,) so a debugger would be handy
tulip sleet
which board /chip is it?
timid bolt
adafruit_feather_rp2040
tulip sleet
i have not actually had a bad enough bug to have to use gdb on the rp2040 yet, so I can't say what the best choice is. But I think the openocd with another RP2040 will work fine
slender iron
@timid bolt I'm a fan of pyocd. i think it does have rp2040 support already
blissful pollen
Got a GIF loading! 🎉 now to get it to display
still zephyr
Got a GIF loading! 🎉 now to get it to display
YAY
orchid basinBOT
stuck elbow
Got a GIF loading! 🎉 now to get it to display
with the imageload library?
blissful pollen
with the imageload library?
Trying in the core for animated GIFs in displayio for speed.
stuck elbow
Trying in the core for animated GIFs in displayio for speed.
with bitbank's C library?
blissful pollen
Yup
timid bolt
I ended up ordering a pico w for 'picoprobe' debugging. Funnily enough the swd cables and headers cost more than the board!
Adafruit doesn't seem to have their own pico w board. Is one in development?
crimson ferry
there may be contractual limitations on the driver
jaunty juniper
what do you mean by their own pico W ? An RP2040 board with the CYW43439 module ?
crimson ferry
although, this seems OK https://github.com/georgerobotics/cyw43-driver/blob/main/LICENSE.RP
jaunty juniper
I don't know if the module is available for 3rd parties
stuck elbow
and the price is probably not as good as for the rpi foundation
crimson ferry
you can buy some dev board forms of the module, but yeah not sure about the module itself
timid bolt
oh I see, the pico w is just a regular pico with a CYW43439 bolted on. I was wondering what all this #ifdef CYW43 in the code was about. 😀
jaunty juniper
yeah that's the metal box that is not on the regular pico 😉
ornate breach
yeah that's the metal box that is not on the regular pico 😉
There’s just magic smoke under it is all
crimson ferry
CYW43439KUBGT modules on digikey $7.73 quantity 1, $4.20 quantity 1000 ...that's double or more of espressif modules
ornate breach
Yuuuup
RPi pays like $1.80 or $2 for each of the CYW43 modules because they bought like half a million
crimson ferry
Adafruit would be hard-pressed to do an under-$10 board, but maybe a $20-$25 Feather
ornate breach
I barely eeked out a $7 RP2040 board
crimson ferry
but dang, a BFF with u.FL would be pretty sweet :-p
ornate breach
Back when the RP2040 was $0.50 a piece on a 500 piece reel
Goood times
glares at $500 reels from Digi-Key
At least Adafruit gives us a price break on 100+ qty
storm minnow
So did the RPi foundation pay to have every RP2040 ever made, then sell to distributors, or do distributors order them from the fabs themselves?
stuck elbow
So did the RPi foundation pay to have every RP2040 ever made, then sell to distr...
the former
storm minnow
Ahh… imagine standing in a warehouse with every RP2040 ever made, before they scatter across the world…
stuck elbow
there were multiple batches
storm minnow
Sadly their blog appears to lack search…
timid bolt
In the REPL, if I start typing in multiline code, then it automatically indents lines, which is nice. However if I paste "pre-formatted" code into the REPL, the auto-indenting messes up the proper indenting. Is there a way around this?
jaunty juniper
In the REPL, if I start typing in multiline code, then it automatically indents ...
paste mode with ctrl-E
slender iron
So did the RPi foundation pay to have every RP2040 ever made, then sell to distr...
note, it isn't the foundation that does the rp2040 and raspberry pi SBCs. There is a commercial entity, raspberry pi trading for commercial stuff
storm minnow
Ohhhh…. I didn’t know this
jaunty juniper
also raspberrypi.org vs raspberrypi.com
storm minnow
I didn’t realize there was a split
slender iron
they've been more formal about it recently
storm minnow
Makes sense
orchid basinBOT
lone axle
Is there any relatively easy way to see a diff between the state of two different builds linked on S3? I'm attempting to find the differences between these two
jaunty juniper
state ? what do you mean ? you can do a diff on github
jaunty juniper
in git in general
tulip sleet
@crimson ferry thank you for that backtrace!!
jaunty juniper
lone axle
by state, I'm hoping for a list of commits and/or a list of which files have differences and what the differences are.
I know git can probably make this comparison but I don't really know how to do it beyond a new branch like for a PR compared with the current main.
jaunty juniper
(starting at https://github.com/adafruit/circuitpython/compare)
lone axle
Yes! thank you.
do you know if it's linked in the front end anywhere? I can bookmark it if not, but If I know where to find it I can probably remember the path to it
jaunty juniper
even the github docs says to "append /compare to your repository's path" so... 🤷
blissful pollen
Anyone know have thoughts on memory structure alignment in a library Vs main code base?
Short example is in the gif library is does (summarized) ‘memset(gif, 0, sizeof(GIFIMAGE))’ where gif is a passed pointer to the gifimage variable in the python object. But that line is overrunning it’s memory and wiping my file object pointer.
If I run the exact same line of code in my shared-module .c it works properly.
Or offhand does someone know what memory alignment is used?
(If this doesn’t make much since I’ll try to post a gist later)
timid bolt
Anyone know have thoughts on memory structure alignment in a library Vs main cod...
I will look if you post the gist.
crimson ferry
Did we lose a socket?
import time
import wifi
import socketpool
from secrets import secrets
time.sleep(2) # wait for serial
wifi.radio.connect(secrets['ssid'], secrets['password'])
time.sleep(0.1)
print(f"{wifi.radio.ipv4_address}")
pool = socketpool.SocketPool(wifi.radio)
socks = []
while True:
socks.append(pool.socket(pool.AF_INET, pool.SOCK_DGRAM))
print(f"{socks}")
[tio 15:56:54] Connected
192.168.6.205
[<Socket>]
[<Socket>, <Socket>]
[<Socket>, <Socket>, <Socket>]
Traceback (most recent call last):
File "code.py", line 15, in <module>
RuntimeError: Out of sockets
web workflow disabled: espressif correctly shows 8, I thought we had 4 in raspberrypi
how do we have enough for web workflow:
https://github.com/adafruit/circuitpython/issues/7333#issuecomment-1349491317
still zephyr
Not sure if this is something worth to write an issue or not, but this code..
Adafruit CircuitPython 8.0.0-rc.1 on 2023-01-30; Adafruit Feather RP2040 with rp2040
>>> import displayio
>>> bitmap = displayio.Bitmap(-40,40,1)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
MemoryError: memory allocation failed, allocating 10485600 bytes
>>>
🤔
proven garnet
https://blog.adafruit.com/2023/02/02/circuitpython-8-0-0-release-candidate-2-rel...
I already have the poster and am ready to hang it 
🎊
tame creek
Not sure if this is something worth to write an issue or not, but this code..
``...
is it documented to accept a negative number there?
still zephyr
is it documented to accept a negative number there?
I guess not, :), I stumbled upon this after trying to set a tilegrid.
width (int) – The number of values wide
arguably, it could be a doc bug
tulip sleet
Could you open an issue? It should validate the args to be >0. Thanks.
still zephyr
Yes. Will do 🙂
manic glacierBOT
manic glacierBOT
proven garnet
Not sure this really affects us because I think we typically have used API tokens as secrets where applicable, but just a heads up the $GITHUB_TOKEN is moving to read-only default: https://github.blog/changelog/2023-02-02-github-actions-updating-the-default-github_token-permissions-to-read-only/
proven garnet
I don't think we really use it, the exception being adabot, but adabot uses tokens in secrets instead.
The only exception is the new replacement action in the core, but it gets write access for issues explicitly
Also doesn't affect existing repos
onyx hinge
Not sure if this is something worth to write an issue or not, but this code..
``...
it wouldn't hurt to report it as an issue. Probably carefully choosing the incorrect numbers could cause a hardfault as well, similar to https://github.com/v923z/micropython-ulab/issues/576
thorny jay
Hi, I am watching John Gallagher video on how to setup a PicoW for his class ( https://www.youtube.com/watch?v=ND5UhTQlfWw&t=308s ). And he goes on listing all the library he will use and how to copy them from the bundle.
So I was wondering if there was a way for him to provide a mini-bundle like the learn guide are doing, to have a single zip file that contain everything he need for his student.
If not possible as such, maybe he could have a learn guide (or user page on learn) and a demo code that use all those library for the "system" to generate that zip file for him.
Also, if you know his Discord handle, maybe you could ping him about your response?
tulip sleet
Hi, I am watching John Gallagher video on how to setup a PicoW for his class ( ...
He could always just make up a .zip by hand. He is @JohnG here and is also active on Twitter as @Gallaugher. Note the "U" in his name.
thorny jay
He could always just make up a .zip by hand. He is `@JohnG` here and is also act...
The "problem" is to have obsolete library version if doing that. Not sure how the "all project files" work behind the scene in the learn guide, but that use case of user provided list of file was discuss at the time.
tulip sleet
The .zip file is generated at download time (maybe it's cached a little bit), but yes, it would be updated as versions change. How it works is pretty specific to the Learn Guides. One could imagine a web-based "zip creator" that had URL's specified for it that were fetched individually and then zipped up.
jaunty juniper
I don't think that the script used in the learn guides is available anywhere right now ? But it would really be easy to write, even using circup (which can be run on local directory instead of a drive, as long as there's a valid boot_out.txt in it)
circup --path some_directory install --auto
I also made a web "bundler" where you can drop a file and get a zip with it and all dependencies
https://neradoc.me/bundler/
tulip sleet
I don't think that the script used in the learn guides is available anywhere rig...
Learn uses ruby behind the scenes
thorny jay
Thanks, I'll stop solving a problem that @broken hedge did not even express, but I think there are great solution above.
The video is still useful as is, because going to the bundle and downloading what you need is good knowledge.
proven garnet
Hi, I am watching John Gallagher video on how to setup a PicoW for his class ( ...
Using the build-mpy action could be applicable if you also wanted to compile .mpy files from some while you're at it. Making zip files for projects automatically is its goal!
You would just have a repo with the files. If you didn't want to compile any .mpy files at all you should still be able to use it with:
on:
release:
types: [published]
jobs:
upload-mpy-zips:
runs-on: ubuntu-latest
steps:
- name: Checkout the current repo
uses: actions/checkout@v3
with:
submodules: true
- name: Run MPY Action
uses: adafruit/build-mpy@v1
with:
github-token: ${{ secrets.GITHUB_TOKEN }}
circuitpy-tag: "7.3.3"
mpy-manifest-type: "exclude"
that doesn't retrieve the dependencies though ?
proven garnet
No, you're right, it doesn't. I made the actions assuming projects would add dependencies within a lib/ folder in the repo. But no, no auto-fetching deps.
That way dependencies were static (it wouldn't accidentally pick up an API-breaking change)
Slash that seems like where circup already is the best solution 😄
broken hedge
Thanks, I'll stop solving a problem that <@217065986384330763> did not even expr...
HI David. Glad you found the video useful! Yes, I did want my students to be comfortable exploring the original "lib" files, finding what they need, and copying these over. I encourage students to work with components that we don't cover in class, so my hope is that this helps them understand how to work with libraries they need so that if they use something we haven't covered, they'll be able to dive into the libraries & get what they need in the future. This might also help them decide which libraries they can delete if they don't if they run short on space. There are also some cool tools out there like circup, which I haven't used yet, but that seem really great at keeping "lib" directories up to date. Others might have more insight on circup. Cheers! https://learn.adafruit.com/keep-your-circuitpython-libraries-on-devices-up-to-date-with-circup
thorny jay
HI David. Glad you found the video useful! Yes, I did want my students to be com...
Don't worry, I know CircUp and Discotool. I am not new here, but I still discover things. It is just that some times ago, before the zip files from learn guide even existed, I described that need to get just the files you need rather than to describe getting the full zip, copying over the mpy files and folder. I was testing many guide to see if they were still working with a new CP version, and picking the library was the slowest part of the process.
I have only watched one or two videos, but there is really a lot of content. That's great.
onyx hinge
well that's weird. (local build but all should be as normal)
Adafruit CircuitPython 8.0.0-rc.1-12-g80d961292-dirty on 2023-02-03; Adafruit Feather ESP32-S2 TFT with ESP32S2
>>> import wifi
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
espidf.IDFError: Invalid argument
I have some out of date submodules, maybe that's it
tulip sleet
@onyx hinge interesting PR -- I will defer it past 8.0.0 if that is OK with you
lone axle
@slender iron how do you load the bangle.js device once you've got a build from your PR branch? Does their 'app loader' page allow you to select the firmware file, or load via some other means?
slender iron
I used nrf connect on android to load it
thorny jay
FOSDEM 2023... It is so close (in less than 12H) and I am so unprepared (trying to select what I want to see). Anybody from the community going there?
manic glacierBOT
still zephyr
Does anybody has problems bulding the packages for a library release?
lone sandalBOT
proven garnet
Does anybody has problems bulding the packages for a library release?
For PyPI?
Everything should be working within the release workflow, what issue are you experiencing?
Ah, I looked at your repo you submitted to the Community Bundle and see the error. It looks like a permissions error - the Adafruit org might have a setting on by default that you may not. The solution is probably to explicitly allow permssion to upload a release asset.
It does look like (I think) you have a GitHub token with the correct secret name in the repository. Does it have the repo scope, with write permissions?
I think for fine-grained tokens its "Contents", and for classic tokens it's "repo" (or the finer "public_repo" since your's is public)
still zephyr
Ah, I looked at your repo you submitted to the Community Bundle and see the erro...
Ok going to look into that. I guess is in the library setting, right?
proven garnet
Tokens can be found in your settings under "Developer Settings", but the secrets can be found in the repository settings. You can't see a secret value after adding it, so you're best bet is to generate one with the correct permissions and then add that. Alternatively, you should be able to modify the workflow to use $GITHUB_TOKEN and add permissions to that in the YAML file, but if you want to keep it consistent with the Adafruit-style, use the other way.
@still zephyr 
still zephyr
<@607375730610798634> <a:this:407945287534116864>
Thank you very much I will take a look at that 🙂
still zephyr
Tokens can be found in your settings under "Developer Settings", but the secrets...
Thank you very much for your help, it worked 🙂 I really appreciate it !!!!!!
proven garnet
No problem! Glad it worked!
I'm trying to keep a record of common issues like this so thanks for making me aware of it!
manic glacierBOT
lone axle
I am probably missing something fundamental about types in C. This this code / behavior doesn't really make any sense to me. Does anyone know why this evaluates to false, or what could be an appropriate change to make in order to have it evaluate to true like it should numerically?
uint32_t negative_val = -1;
if (negative_val < 1){
// This does not print, condition evaluates to false seemingly
mp_printf(&mp_plat_print, "negative_val is < 1:\n");
}
u means unsigned
lone axle
I am assuming that uint32_t means unsigned int 32 bits. And it does make sense that -1 would not be a valid value for an unsigned int. But it seems to allow the value to be stored in it none the less. If you print it's value with %d string format it will print out the -1 showing that it is indeed storing that value. But then when you go to make a logical expression with it things get very weird.
lone axle
u means unsigned
why does this line suceed then? shouldn't it be some kind of error for type / value mismatch?
uint32_t negative_val = -1;
yeah that's C, it converts it to a positive value
blissful pollen
C is always trying to be "helpful"
jaunty juniper
and %d is signed, while %u isn't
blissful pollen
The short answer would be the only valid value you can check on a uint32_t would be 0. The concept of -1, -2, etc. does not even exist for a uint
jaunty juniper
I believe the value actually stored will be the 2-complement
blissful pollen
Anyone offhand know of an example of a native class creating another native class in code?
jaunty juniper
or something like that
lone axle
It doesn't seem like it's flipping to positive though.
uint32_t negative_val = -1;
mp_printf(&mp_plat_print, "negative_val is %d:\n", negative_val);
This prints
negative_val is -1:
The %d auto converted it back I believe
jaunty juniper
It doesn't seem like it's flipping to positive though.
```
uint32_t negative_val...
use %u
lone axle
The %d auto converted it back I believe
auto converted back to what though? if -1, -2 etc don't exist as concepts I don't understand how it could have done that?
lone axle
use %u
Thank you! this one printed negative_val is 4294967295:
so %d flips that back around then I guess
blissful pollen
auto converted back to what though? if -1, -2 etc don't exist as concepts I don'...
C really just treats everything as bytes in memory. Your code tells it how to interpret that byte in memory. There is no type safety involved.
So if memory address 0x1000 contains 0x20
%c will return a space
%d will return 32
%x will display 0x20
There is no meaning to it. So when you say uint32_t tmp = -1 the compiler auto converted -1 to uint32_t because that is what you said it was. Hope that makes some sense
lone axle
Okay I think I'm following what is happening with the variables now at least. Thank you both!
The next, more practical question though. These areguments: https://github.com/adafruit/circuitpython/blob/5f43a63a701447595e5647621769e1ca9289ac85/shared-bindings/displayio/Bitmap.c#L65-L67 are all uint_32.
I'm trying to add validation that will raise an error if the user passes in a negative value. Since the logic can't be trusted when the values are negative, my next thought is to make these arguments into signed int's so that they can exist as negative and behave as expected. But that feels kind of awkward since my ultimate goal is specifically disallowing negatives.
despite feeling awkward does that seem like the correct / only way for us to actually validate that user passed a positive value successfully?
blissful pollen
So in that case the get_int returns a signed value. if you look under the hood. Those variables are converting it to unsigned
jaunty juniper
you can take them as ints and then convert back to (uint32_t) when you pass them later ?
actually, is there a test for positive integers ?
tame creek
for two's complement, just test the most significant bit
blissful pollen
I thought Dan made some common checks to save on translation space that does it all at once. Not sure if this was one of them
jaunty juniper
I think that would work by the way
uint32_t width = mp_obj_get_int(all_args[0]);
if( (int_32_t) width < 0) ...
analog bridge
@lone axle use mp_arg_validate... functions defined in py/argcheck.c
blissful pollen
Hmm my example on the issue in the IS31 code is being blocked when I try to post it. Not sure why
jaunty juniper
blocked here ? on github ?
blissful pollen
No discord blocking message... But just noticed it exists in that code as well
lone axle
No discord blocking message... But just noticed it exists in that code as well
it got filtered out due to a string that it interpreted as a bad word. I can see it in the "recycle bin"
jaunty juniper
yeah that bot goes a little too far
blissful pollen
I think I missed the ticks to mark it as code. Not a big deal. Just was wondering
lone axle
I think the tick marks may not hav ehelped in this case. It found a substring in the code that it didn't like. I think it would filter it whether or not it was in code. But not certain.
jaunty juniper
nope, didn't like it with the ticks
analog bridge
How to return a str object from a QSTR?
blissful pollen
How to return a `str` object from a `QSTR`?
qstr_str(qst)
analog bridge
qstr_str(qst)
I found this:
mp_obj_t mp_obj_new_str(const char *data, size_t len) {
qstr q = qstr_find_strn(data, len);
if (q != MP_QSTRnull) {
// qstr with this data already exists
return MP_OBJ_NEW_QSTR(q);
} else {
// no existing qstr, don't make one
return mp_obj_new_str_copy(&mp_type_str, (const byte *)data, len);
}
}
so I think MP_OBJ_NEW_QSTR(q) can do the job
it works!
timid bolt
The only real difference between signed and unsigned ints is the meaning of comparison operators (< and >). All other operations are the identical.
lone sandalBOT
manic glacierBOT
manic glacierBOT
lone axle
Does anyone know what hardware goes with the MCP2515 library? https://github.com/adafruit/Adafruit_CircuitPython_MCP2515 The description says it's a CAN bus controller. So my best guess was maybe it could be used with the CAN M4 devices like https://www.adafruit.com/product/4759. But I didn't find any reference to that driver chip or usages of the library in the product pages or learn guides for CAN stuff. It looks like there is a core module for interacting with CAN on those device. Perhaps this is a library that is/was for a piece of hardware that didn't end up coming to be, or it was superseded by the core implementation?
stuck elbow
maybe something like https://www.sparkfun.com/products/13262 ?
lone axle
Thank you.
onyx hinge
It doesn't make sense that a delay would be needed here, because the wifi is up. communication with the dhcp server and dns server apparently succeed. even opening (but not reading from) the tcp connection apparently succeeds. I wish instead of me you had a networking expert.
crimson ferry
yes, wifi seems to be fine, and some level of socket operations also seems ok. but at the level of TCP, something is not quite right without delay.
(I am not a networking expert either)
onyx hinge
I, yeah. I appreciate that this is a vexing issue but I need to not get too deep into it right now due to other priorities 😕
crimson ferry
there is a workaround 🙂
onyx hinge
but I am still confused about the script for reproducing the problem, too. how do I see the results, if it runs 0 seconds after boot? Or does re-running code.py exhibit the same problem each time? (so it's not just after boot)?
crimson ferry
yes, once it doesn't start up right, it stays broken until: working (delayed) code is used and board is reset
onyx hinge
btw I do appreciate that you distilled it down into a version that doesn't use adafruit_requests, it eliminates a lot of stuff
proven garnet
@tulip sleet Thanks for clarifying that sound levels were possible on the CPX - my understanding previously was that it was a hard "can't".
Specifically due to the buffer
tulip sleet
it might be that it fits now; we had a lot of trouble with frozen libs being too big. I am not sure from the dialogue in the issue exactly what was wrong.
but even if it fits now it might not fit in the future
proven garnet
@lone axle what do you think about allowing a configurable Python version for the CI workflows? Default would still be latest but it would give community contributors the ability to use CI with the intended version (e.g., they may not want to support 3.11 on Blinka platforms)
lone axle
<@382939733107408897> what do you think about allowing a configurable Python ver...
I'd be in favor of that I think. It may depend on how difficult it is to configure, but assuming it can be made fairly straight forward, that seems like a nice option to give folks. I could see posibilities for an issue that is present under whatever is brand new and used by the CI. But isn't present in their local environment and could be hard to fix if they can't easily try whatever version is needed locally.
timid bolt
I, yeah. I appreciate that this is a vexing issue but I need to not get too deep...
I could look at this in a few days. (I have a Pico W in the mail.)
onyx hinge
I could look at this in a few days. (I have a Pico W in the mail.)
If you're motivated to that'd be great.
jaunty juniper
@crimson ferry you seem to use the term "web workflow" for "auto connect". To clarify, I seem to reproduce the issue without the web workflow (CIRCUITPY_WIFI_SSID and CIRCUITPY_WIFI_PASSWORD defined but no CIRCUITPY_WEB_API_PASSWORD)
or at least it should be disabled in that case
proven garnet
I'd be in favor of that I think. It may depend on how difficult it is to configu...
Yeah, it's simple to make optional arguments with defaults. I'll go and handle that this week (honestly probably today)
Thanks for the input!
lone axle
For sure. Thank you for always working to improve the infrastructure!
crimson ferry
<@360465312518504448> you seem to use the term "web workflow" for "auto connect"...
you're right, I conflate the two in my head
(auto-connect necessary for web workflow, I think, but could be used on its own, though I don't)
jaunty juniper
you're right, I conflate the two in my head
well to be fair it looks like the web workflow is enabled anyway, which I'll verify and open an issue about, because that's a problem
crimson ferry
I mean, theoretically, you could connect to wifi later, and then use web workflow, not sure if the code supports that... there are quite a few things set up for web workflow
jaunty juniper
and serves empty pages instead of asking for password
onyx hinge
<@&356864093652516868> hi! we'll be having our weekly meeting in a bit over an hour. Now's a great time to add your notes to the document. https://docs.google.com/document/d/1BAM7-G2YJIgL8ptZG8rY5rKig1ptj1nfbDeuzX6dQzA/edit?usp=sharing
crimson ferry
@jaunty juniper https://docs.circuitpython.org/en/latest/docs/workflows.html#web seems to have the two concepts a bit co-dependent (1st paragraph)
web workflow = auto-connect + web server + mDNS + websocket +...?
jaunty juniper
yeah maybe I'm wrong and we don't intend to support auto connect on its own
crimson ferry
🤷♂️
slender iron
I think that auto-connect will start mdns and the webserver
jaunty juniper
currently it does yeah
slender iron
we could add another key to cause it to start
or one that would prevent it from starting
jaunty juniper
and urls like http://cpy-007331.local/fs/ return a blank page
slender iron
ya, when the password isn't set
jaunty juniper
so I would just say no password = no workflow, which was what I assumed all along
slender iron
that's reasonable but not how I implemented it 🙂
Could add some plain text to the response too. or a link
thorny jay
Sorry, skipping the meeting I am wasted by my FOSDEM week-end (maybe I am too old for this). I had a long "report" in a notepad, but not on this phone. 😉
There where two MicroPython talk, 1H00 + 30 minutes. And the usual MicroPython stand (or half table).
idle owl
Sorry, skipping the meeting I am wasted by my FOSDEM week-end (maybe I am too ol...
No worries! Take care of yourself. You can do your long report anytime. We'll be back next Monday. 😉
(Or Tuesday. No idea if it's a holiday.)
orchid basinBOT
onyx hinge
We're still wrapping our internal meeting so we'll probably start a few minutes late today.
slender iron
brb
idle owl
BRB.
slender iron
I'll be in the same boat next week 🙂
idle owl
On it.
slender iron
@stuck elbow mute please
stuck elbow
sorry, I didn't notice
midnight ember
wow major undertaking, nice project.
midnight ember
It uses a for loop inside a with.
From CPython to circuit python back to C. Abstraction inception.
turbid radish
I try all sources, it can be hit & miss, please email it in for sure
slender iron
yup, happy to
midnight ember
Only 2 issues for 8.0, getting close.
tulip sleet
Only 2 issues for 8.0, getting close.
i just pushed those forward
midnight ember
With Scorpio out the NeoPxl8 library seeing some action makes a lot of sense.
jaunty juniper
NOTE: Please be advised that this is a change to the document only the product has not been changed.
slender iron
are the buttons just on/off? Maybe a switch scanning matrix would be cheaper
like a mechanical keyboard
midnight ember
yes, i don't know enough to know how to do that yet 😦
oblique iron
16 switches would need 8 gpio. 4x4 matrix. You would just pay attention to directionality to prevent ghosting with diodes.
slender iron
blissful pollen
yes, i don't know enough to know how to do that yet 😦
It's not too bad once you wrap your head around it. I set a small one up via breadboard to learn and practice.
oblique iron
Hi, I design keyboards and I just so happened to click over when that was suggested. If you can't figure it out, let me know.
midnight ember
Ahh I didn't think about a 4x4 diode matrix in that way. That's neat. Will definitely look into that thank you.
jaunty juniper
I am still convinced it's only a marginal issue, that you might simply never encounter for a keyboard design
errant grail
We just revived a couple of air plants after finding a great source for care information. They are happy again.
idle owl
We just revived a couple of air plants after finding a great source for care inf...
What is this source?
midnight ember
There's no word on expected behavior if you do use GPA7 as an input. Would it just crash or be glitchy? Who knows.
idle owl
I haven't even had these for very long, I was wondering if they weren't getting enough water at the place we bought them from.
jaunty juniper
but to be fair, Microchip hasn't given any clear info
oblique iron
I am still convinced it's only a marginal issue, that you might simply never enc...
1N4148 diodes are less than a penny each. 16 cents to prevent any chance of issue is worthwhile. For computer keyboards, it does happen because of the speed of the scanning.
midnight ember
Air plants are quite fragile and tempermental, don't feel bad if you kill them.
idle owl
Air plants are quite fragile and tempermental, don't feel bad if you kill them.
That's fair. I seriously hope I don't kill them, a. because I only got them recently, and b. they're sooooooo fascinating. And the new ones I have coming, some of them have bonkers flowers. I hope to eventually see them.
storm minnow
Air plants?
idle owl
Air plants?
They are plants that do not need soil.
storm minnow
Ahh… they do sound fascinating
errant grail
What is this source?
idle owl
Thank you!!
errant grail
The last page of the .pdf has some other useful links.
idle owl
Perfect.
storm minnow
The closest I’ve gotten to air plants is doing aeroponics
midnight ember
Zoom zoom logic analyzer, neat project.
errant grail
Air plants were supposed to be the no-maintenance felines of the plant kingdom.
midnight ember
I've helped more than a few people in the circuit python discord using Wokwi instead of breaking out a physical device. Can highly recommend it.
Thanks for hosting Jeff!
slender iron
thanks all!
gilded cradle
Thanks
errant grail
Thanks!
tulip sleet
lol
still zephyr
Thanks
midnight ember
Let's go 8.0 
ember iris
Thanks all!
onyx hinge
ugh my audio today sounds like it was recorded inside one of those weird huge food-service tomato sauce cans. oh well, it's the audio we have.
lone axle
ugh my audio today sounds like it was recorded inside one of those weird huge fo...
I have a backup recording if you want to see if it's any different?
onyx hinge
you're welcome to check but I'm assuming that's just how my audio sounded.
lone axle
Skipped around a bit, not listened to the whole thing. But sounds normal in my recording as far as I can tell. https://www.dropbox.com/s/fbharug8ksnw390/circuitpython_weekly_meeting_2_06_23.mkv?dl=0 should be available to download from here if you want to use it.
maybe on the loud side, if anything, I should drop the input volumn a tad in OBS next time.
onyx hinge
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/1qKQK-hmTK52jADRMLvGza0wRwAt7R3cGG8Cr0u6xrxY/edit?usp=sharing
onyx hinge
Here is the notes document for next Monday’s CircuitPython Weekly meeting. It is...
@lone axle yeah it sounds pretty similar to what I was hearing .. a bit metallic-echoey
thanks for running the backup recording, it's good to have a safety net for any number of reasons
idle owl
@onyx hinge At least you had audio for everyone. My favorite is the one I posted with only my audio. 🙄
onyx hinge
has your newer mac stopped sabotaging your audio quite so much?
idle owl
Entirely.
midnight ember
As soon as he said "I haven't used OBS in years" my first thought was someone should do a backup.
I almost started one but was like nah someone else will hear that and do one. 😛
orchid basinBOT
proven garnet
digital shoreBOT
thorny jay
I found my laptop and notepad and my excuse message for not having notes, and some report about FOSDEM:
I have not done much except going to Fosdem.
Maybe I am too old for this... maybe I am not ready for the in person stuff.
But that was crowded and almost no one wearing mask.
So I only went to day one and followed day two in streaming.
(also in between I had an unrelated party until 2AM and maybe that is why I am wasted today)
For @turbid radish I have a picture of the MicroPython stand: https://hostux.social/@NOeVoting/109806087336088426
The stuff presented were very PyBrick oriented with LEGO bricks everywhere (globe, train, robot).
But I don't know, I did not even talk to them, I don't know if there was something new or if they were selling stuff.
What I can say is that usually there is that lady, maybe from Swizerland, holding the stand, and this year they were two.
(Pine64 was next and maybe had a watch that can run MicroPython, but I did not go there either)
I am sure I missed opportunity for meeting interesting people, and I am sorry if anybody from the community was present.
I saw a few old friends, even a co-worker, (and then the party), and that was maybe enough social interaction for me.
There were two speak about MicroPython, the first one very active with demo, remotely it was not always possible to see the LEDs:
https://fosdem.org/2023/schedule/event/python_micropython_intro/
https://fosdem.org/2023/schedule/event/python_hacking_esp32/
The FOSDEM team is still working on checking the video before publishing, so that could be advertised next week in the CPnews.
brazen hatch
limpid flame
Hey, I am trying to build a circuit python firmware.uf2 for a custom board. But I keep getting some errors when building the firmware.uf2 file. Even with some already established boards.
I am not to familiar with this type of programming, so pls be lenient.
Errormessage: Makefile:403: *** unknown IDF_TARGET . Stop.
Could maybe someone more experienced have a look at it?
slender iron
https://github.com/adafruit/circuitpython/releases/tag/8.1.0-alpha.0
Incorrect t...
we tend to do this when we branch off
so that S3 builds are labeled with 8.1
@limpid flame can you post more of the error?
limpid flame
<@349289583340683265> can you post more of the error?
There is not much more
circuitpython/ports/espressif$ make -j8 BOARD=sensebox_MCU_2_1 TRANSLATION=es
Use make V=1, make V=2 or set BUILD_VERBOSE similarly in your environment to increase build verbosity.
Makefile:403: *** unknown IDF_TARGET . Stop.
slender iron
did you install and export from the esp-idf directory?
limpid flame
at least i think so
limpid flame
WSL Ubuntu 22 on Windows 11
/tmp/circuitpython/ports/espressif$ git status
On branch main
Your branch is up to date with 'origin/main'.
Untracked files:
(use "git add <file>..." to include in what will be committed)
../broadcom/
../cxd56/
bindings/espcamera/
bindings/espulp/Architecture.c
bindings/espulp/Architecture.c:Zone.Identifier
bindings/espulp/Architecture.h
bindings/espulp/Architecture.h:Zone.Identifier
boards/brainboardz_neuron/
boards/espressif_esp32s3_devkitc_1_n32r8/
common-hal/espcamera/
esp-idf-config/partitions-32MB.csv
esp-idf-config/partitions-32MB.csv:Zone.Identifier
esp-idf-config/sdkconfig-32MB.defaults
esp-idf-config/sdkconfig-32MB.defaults:Zone.Identifier
../litex/
../mimxrt10xx/
../nrf/
../raspberrypi/
../stm/
nothing added to commit but untracked files present (use "git add" to track)
slender iron
hrm. I don't see anything suspicious about esp-idf
it does seem like you checked out a newer version and it left some files behind
limpid flame
it does seem like you checked out a newer version and it left some files behind
Do you mean a newer version of the repo or some dependences?
In the biginning i forked the repo and added the custom board to the /boards.
limpid flame
what does that mean for me? Can i just delete the "messed up" repo and glone it again?
tulip sleet
https://github.com/adafruit/circuitpython/releases/tag/8.1.0-alpha.0
Incorrect t...
no, tag 8.0.0, branch 8.0.x, and tag 8.1.0-alpha.0 are all at the same commit
slender iron
its probably not the problem
but I'm not sure how it got that way
the next step I'd suggest is adding V=2 to your make command to see what its failing on
limpid flame
like this?
make -j8 V=2 BOARD=sensebox_MCU_2_1 TRANSLATION=es
slender iron
that should work
tulip sleet
did you do source esp-idf/export.sh before?
slender iron
try it without -j 8 too
tulip sleet
you cannot just run the shell script, you have to source it.
limpid flame
try it without -j 8 too
all the same error code
limpid flame
you cannot just run the shell script, you have to source it.
I do not know what you mean tbh
slender iron
do you have IDF_TARGET in your board file?
looks like its in mpconfigboard.mk. did you typo it?
limpid flame
do you have IDF_TARGET in your board file?
https://github.com/A-Spork/circuitpython/tree/main/ports/espressif/boards/sensebox_MCU_2_1
That´s the added boards code, if you mean that
slender iron
I see IDF_TARGET in other boards: https://github.com/A-Spork/circuitpython/blob/main/ports/espressif/boards/microdev_micro_s2/mpconfigboard.mk#L6
manic glacierBOT
limpid flame
Okay, then this should be the problem. Let me add and try it again.
To create the new board i just compared a old board with the same pins etc and one board with the same chip.
Not to sure if it even possible to work that way
slender iron
ya, that should work
jaunty juniper
CHIP_VARIANT and CHIP_FAMILY are not used for ESP builds, you might have copied from a non-ESP board ?
likewise pin_PA04 etc are atmel and stm pin names
limpid flame
I coppied form the lolin s2 mini as it should use the same chip and and older version with atmel https://github.com/adafruit/circuitpython-org/pull/669
I just saw that you also helped a collegue of mine too
timid bolt
I could fix this issue. Does that sound like a good idea?
https://github.com/adafruit/circuitpython/issues/3913
limpid flame
Okay, getting a different error message with the updated version
circuitpython/ports/espressif$ make BOARD=sensebox_MCU_2_1 v=2
Use make V=1, make V=2 or set BUILD_VERBOSE similarly in your environment to increase build verbosity.
mkdir -p build-sensebox_MCU_2_1/genhdr
GEN build-sensebox_MCU_2_1/genhdr/moduledefs.h
make: *** No rule to make target 'boards/sensebox_MCU_2_1/sdkconfig', needed by 'build-sensebox_MCU_2_1/esp-idf/config/sdkconfig.h'. Stop.
slender iron
I think you'll want to start from another ESP board
slender iron
I could fix this issue. Does that sound like a good idea?
https://github.com/ada...
sure! looks weird
jaunty juniper
limpid flame
yep, start with <https://github.com/A-Spork/circuitpython/tree/main/ports/espres...
Okay, but even if i try to build this board it does not work at the end😅
jaunty juniper
what's the error then ?
limpid flame
jaunty juniper
you need to do those after enabling the ESP-IDF environement:
https://learn.adafruit.com/building-circuitpython/build-circuitpython#install-required-python-packages-3086211
that should install the missing modules
limpid flame
I did, but it did not work.
ERROR: Could not open requirements file: [Errno 2] No such file or directory: 'requirements-dev.txt'
ERROR: Could not open requirements file: [Errno 2] No such file or directory: 'requirements-doc.txt'
jaunty juniper
they are at the root of the circuitpython repo
limpid flame
I think i messed that up. Thx
jaunty juniper
so like pip3 install --upgrade -r ../../requirements-dev.txt
limpid flame
Okay, i went threw th tutorial again and everything finished without error, but when i try to build the example board, i get the error messages again
make BOARD=circuitplayground_express
Use make V=1, make V=2 or set BUILD_VERBOSE similarly in your environment to increase build verbosity.
MKMANIFEST ../../frozen/Adafruit_CircuitPython_CircuitPlayground/frozen_cpx ../../frozen/Adafruit_CircuitPython_HID ../../frozen/Adafruit_CircuitPython_LIS3DH ../../frozen/Adafruit_CircuitPython_NeoPixel ../../frozen/Adafruit_CircuitPython_Thermistor
(cd build-circuitplayground_express/frozen_mpy && find * -name *.py -exec printf 'freeze_as_mpy("frozen_mpy", "%s")\n' {} ; )> build-circuitplayground_express/manifest.py.tmp && mv -f build-circuitplayground_express/manifest.py.tmp build-circuitplayground_express/manifest.py
find: ‘*’: No such file or directory
make: *** [../../py/circuitpy_mpconfig.mk:589: build-circuitplayground_express/manifest.py] Error 1
tulip sleet
did you do make fetch-submodules at the top level? If you did esp-idf/export.py in this terminal window, it will have an incorrect environment, set up for espressif builds only. Start a build in a fresh terminal window (you don't have to repeat the setup steps). Also do a make BOARD=circuitplayground_express clean first
@limpid flame ^
limpid flame
did you do `make fetch-submodules` at the top level? If you did `esp-idf/export....
I ran make fetch-submodules and it ran without any errors
I will try it in a fresh terminal
In the new terminal i get the error above
manic glacierBOT
tulip sleet
I ran make fetch-submodules and it ran without any errors
I will try it in a fr...
did you do the make clean?
all make clean does is delete the build-<BOARDname> directory for that board
limpid flame
Okay. No I did not do that in the fist try, but even if ido that in a new terminal and try to build then, I get the same error
jaunty juniper
what is your version of make ?
limpid flame
$ make --v
GNU Make 4.3
Built for x86_64-pc-linux-gnu
Copyright (C) 1988-2020 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later http://gnu.org/licenses/gpl.html
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
jaunty juniper
ok that should be good
limpid flame
all `make clean` does is delete the `build-<BOARDname>` directory for that board
I hope that´s the right way to do it
jaunty juniper
well pip install semver then
I would think it would be in the requirements or as a dependency, but maybe not ?
limpid flame
do not think so. it was not installed by now at least
limpid flame
well `pip install semver` then
it worked
make BOARD=circuitplayground_express
Use make V=1, make V=2 or set BUILD_VERBOSE similarly in your environment to increase build verbosity.
247980 bytes used, 5460 bytes free in flash firmware space out of 253440 bytes (247.5kB).
12060 bytes used, 20708 bytes free in ram for stack and heap out of 32768 bytes (32.0kB).
Converted to uf2, output size: 496128, start address: 0x2000
Wrote 496128 bytes to build-circuitplayground_express/firmware.uf2
Had to install a few packages but now i get this, which is how it should be, I guess
jaunty juniper
@tulip sleet looks like we need to add semver to the requirements ?
jaunty juniper
make BOARD=circuitplayground_express
Use make V=1, make V=2 or set BUILD_VERBOSE...
yeah that looks good
jaunty juniper
those are in the requirements
limpid flame
then i messed sth up
jaunty juniper
I tested making a venv and installing requirements-*.txt and importing semver in python, it was not found
tulip sleet
That is odd, I thought it was there, or implied by something else.
jaunty juniper
yeah, maybe some module removed their dependency to it
limpid flame
Okay, so the example runs now, and the errors i get should be from the custom board. Any recommendations how to compare or create the code for the new board?
I am really sorry to bother you that much, but i really do not know how to do sth like this, but need the new board for my Bachelors Thesis
jaunty juniper
you can build the lolin S2 ?
limpid flame
sadly no
jaunty juniper
did you forget:
source esp-idf/export.sh
sure i did forget it👍
jaunty juniper
might have to reinstall semver in that environement
orchid basinBOT
limpid flame
onyx hinge
oddly, that's not the content of bitmap_font.py for me. Is it possible you ran some git commands that changed (updated) that submodule beyond the version that we've commited in git? the version we have is quite old [2019] at this point, from before we added type annotations.
limpid flame
Now as you ask, i did ran apt-get update in a different folder.
but no git commands
could that have caused this problem?
tulip sleet
Now as you ask, i did ran apt-get update in a different folder.
that should not matter at all
did you run esp-idf/install.sh before source esp-idf/export.sh
limpid flame
did you run `esp-idf/install.sh` before `source esp-idf/export.sh`
ran both
tulip sleet
Now as you ask, i did ran apt-get update in a different folder.
However, running pip3 install ... after source esp-idf/export.sh does matter, because you are using a different python environment. After you do source esp-idf/export.sh, you may need to do pip3 install semver. It would not hurt to do that.
limpid flame
so in a new terminal? Not sure if i understand that right
sudo apt install python3-pip
pip3 install --upgrade -r requirements-dev.txt
pip3 install --upgrade -r requirements-doc.txt
esp-idf/install.sh
source esp-idf/export.sh
pip3 install semver
limpid flame
not yet
jaunty juniper
those after export for espressif builds though (but once)
pip3 install --upgrade -r ../../requirements-dev.txt
pip3 install --upgrade -r ../../requirements-doc.txt
The "fatal error" at the top is not good
btw, not so great to use /tmp for this: it can be cleared. Better to use your WSL homedir or something underneath it
also /tmp might not be big enough
limpid flame
btw, not so great to use /tmp for this: it can be cleared. Better to use your WS...
So i should set it up fresh in homedir? Was a trip to get it running.
tulip sleet
well, eventually.
limpid flame
Okay. Gonna do that tomorrow. Big Thanks for helping me this time. Hope its okay if i may ask again. Good night
devout jolt
Hi, I am making a library but am having problems with making the docs. In the Learn Guide https://learn.adafruit.com/creating-and-sharing-a-circuitpython-library/sharing-our-docs-on-readthedocs, it says to do this:
cd docs
sphinx-build -E -W -b html . _build/html
but it appears that cookiecutter does not create the "docs" directory and all the files needed for that.
What is the recommended way to generate the "docs" directory?
tulip sleet
Hi, I am making a library but am having problems with making the docs. In the Le...
there should be a docs dir, maybe underneath where you are looking?? https://github.com/adafruit/cookiecutter-adafruit-circuitpython/tree/main/{{ cookiecutter.__dirname }}/docs
still zephyr
there is no Makefile
I used cookiecutter yesterday and it created the doc directory
stuck elbow
cookie cutter asks whether to create docs, the default is n iirc
devout jolt
hmm wonder why it didn't work for me
I just used it
I will try again
and copy the files over I guess to my repo
still zephyr
just checking if you have installed Sphinx in your .venv/.env
devout jolt
perhaps I didn't have it installed at the time of running cookiecutter? (because it didn't mention it until the rtd section)
oh I see. There was a "sphinx docs [no]" option question that I just next'd through because it was after a bunch of adafruit-specific questions
okay, nevermind I am good now. thanks!
still zephyr
oh that would explained it 🙂
slender iron
this pip error makes no sense to me: The conflict is caused by: The user requested pc-ble-driver-py nrfutil 6.1.7 depends on pc-ble-driver-py>=0.16.4 The user requested pc-ble-driver-py nrfutil 6.1.6 depends on pc-ble-driver-py>=0.16.4 ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts The user requested pc-ble-driver-py nrfutil 6.1.5 depends on pc-ble-driver-py>=0.16.4 The user requested pc-ble-driver-py nrfutil 6.1.4 depends on pc-ble-driver-py>=0.16.4
tulip sleet
@slender iron further up it did this:
Collecting pc-ble-driver-py
Downloading pc_ble_driver_py-0.11.4.tar.gz (8.0 MB)
this is windows, so maybe there is no pc_ble_driver_py for Windows (in particular) newer than 0.11.4
unfortunately
slender iron
pip says there is. could be a cache thing
I'll look tomorrow. gotta get garbage out now
jaunty juniper
ok I don't know what happened in the circup CI:
An unexpected error has occurred: CalledProcessError: command: ('/home/runner/.cache/pre-commit/repoaqv54v2u/py_env-python3.7/bin/python', '-mpip', 'install', '.')
return code: 1
stdout:
Processing /home/runner/.cache/pre-commit/repoaqv54v2u
Preparing metadata (setup.py): started
Preparing metadata (setup.py): finished with status 'error'
stderr:
error: subprocess-exited-with-error
× python setup.py egg_info did not run successfully.
│ exit code: 1
╰─> [1 lines of output]
error in pylint setup command: 'python_requires' must be a string containing valid version specifiers; Invalid specifier: '>=3.5.*'
[end of output]
note: This error originates from a subprocess, and is likely not a problem with pip.
error: metadata-generation-failed
× Encountered error while generating package metadata.
╰─> See above for output.
pylint 2.6.0 does have python_requires=">=3.5.*" which is weird (using >= with a * wildcard), is it like no longer supported by setuptools or something ?
onyx hinge
pylint 2.6.0 does have `python_requires=">=3.5.*"` which is weird (using >= with...
yes they changed something in the upper levels of python packaging, let me find you a reference
jaunty juniper
it looks like we use pylint 2.15.5 in other repos, maybe I can do a PR to update pylint in circup and the pylintrc accordingly
onyx hinge
I'm not finding my reference. But yeah I think the solution is going to be to bite the bullet and update to newer pylint. how widely is this going to affect stuff? (i.e., if it's learn that's bad news)
jaunty juniper
learn does pip install --force-reinstall pylint==2.7.1 in the CI it seems, which doesn't use the * version (but "~=3.6")
onyx hinge
@jaunty juniper https://github.com/pypa/setuptools/issues/3772 is what I ran into, the message is different but I think it's the same "removal of behavior people relied on a decade but hey we told you so" that underlies both..
I mean maybe not the same exact code change but the same philosophical impulse.
jaunty juniper
Your code has been rated at 9.50/10
Is that a passing grade ?
❯ pylint circup/__init__.py | wc
105 1401 15294
oh but mostly pylintrc hmmm
why is "Consider using {} instead of dict()" a thing ? Consider using symbols instead of english words ? ugh
there's a few "good" ones:
Missing timeout argument for method 'requests.get' can cause your program to hang indefinitely
Iterated list '_requested_libraries' is being modified inside for loop body, consider iterating through a copy
crimson ferry
Where's the best place to document differences between ports? Thinking at the moment about wifi diffs between espressif and raspberrypi. We already know of some. Just playing around with mDNS again... looks like raspberrypi has a limit of one advertised service, and if web workflow is enabled, that's it. espressif can do multiple regardless of web workflow, haven't gone beyond four yet.
jaunty juniper
Where's the best place to document differences between ports? Thinking at the mo...
maybe we can change that in mpconfigport.h ?
https://github.com/adafruit/lwip/blob/239918ccc173cb2c2a62f41a40fd893f57faf1d6/src/include/lwip/apps/mdns_opts.h#L58-L61
crimson ferry
maybe, not sure about side effects
I thought at first it was the limited sockets, but appears independent of that (at least to a point)
lol, on espressif, I can advertise more mdns services than there are available sockets... I assume it would fail when trying to connect to too many
tulip sleet
Where's the best place to document differences between ports? Thinking at the mo...
I have added "Limitations:" sections to various places in the Python API doc. These document what's different about various ports.
crimson ferry
ooh, I will look and maybe suggest some additions (is that PR-able?)
tulip sleet
oh sure; they are in the in-line doc in the C code for each module, etc.
mental nexus
@stuck elbow Any suggestions for this user on how to debug flickering of an ESP32-S3 driving a dot clock display with the ESP-IDF?
stuck elbow
<@!351136009922019331> Any suggestions for this user on how to debug flickering ...
never used any of those, so no idea
timid bolt
I published a branch of my prototype for changes to async and PIO. https://github.com/gneverov/circuitpython/compare/main...gneverov:circuitpython:async
There is a lot of code there, but if you only read one file, look at this one (https://github.com/gneverov/circuitpython/blob/async/ports/raspberrypi/lib/test_sm.py). It is an example program that uses async/await to send/receive data from a RP2 PIO state machine.
I appreciate your feedback. Especially about what is PR-able. I think that some sort of "design review" of these changes would be helpful. If you have a use case for async or PIO that is not currently feasible on CP then we should talk.
slender iron
@timid bolt please make a draft PR so we can comment inline
tulip sleet
@timid bolt How would you feel about having an audio or video chat at some point soon? I would find that a lot more efficient to get the background on your thoughts on this, and to tell you where we are coming from.
timid bolt
<@1065334137243836517> How would you feel about having an audio or video chat at...
Totally agree. 😁
tulip sleet
We are in US ET (-5 GMT, me), CT (-6, @onyx hinge), and PT (-8 @slender iron) timezones. Since this is our day job, we are pretty flexible as to timing. Even today would be fine for me at least, unless it's getting late for you
slender iron
@timid bolt is in PT too. We chatted yesterday
timid bolt
Cool. How about 10 or 11 PT today?
tulip sleet
that's fine for me too, 2pm ET. @onyx hinge join if you'd like. @slender iron did you text chat or audio/
onyx hinge
@timid bolt @tulip sleet I can join at 2PM ET, sure.
onyx hinge
@slender iron this PR? https://github.com/adafruit/circuitpython/pull/7396 I can review it if you like
I need a break from writing guide for a bit anyway
onyx hinge
@slender iron so that PR switches espressif, nrf, and rp2040 to stackless and turns off pystack? I think it's fine (I haven't studied the details of 'stackless' in detail but not having an arbitrary limit seems nice). It feels a little weird to have it in the style of conditional-undef, that's not like we do most things. But maybe it's necessary, since we are overriding things in circpy_mpconfig.h which comes 'before' mpconfigport.h? If you're content with that style of changing the setting, and we have plenty of time to 'let it bake' on 8.1 to shake out problems, is fine with me.
slender iron
I haven't looked in detail to see how its done. generally I think now is time to start getting stuff like that in
onyx hinge
my recollection of you turning ON pystack in the first place is that it was to benefit the small samd21 boards and this leaves that in place which also seems good.
jaunty juniper
I would like to know more about the effects, it seems to me that it means that it will be harder to optimize for allocations
maybe get more understanding on what it does from the MP people
timid bolt
<@1065334137243836517> <@329766224093249548> I can join at 2PM ET, sure.
Great, see you all at 2 pm ET.
onyx hinge
Afk for a walk and lunch until then
timid bolt
I'm not sure what "stackless" is. A cursory look at the code seems like it allocates python stack frames on the heap at every function call. I'm not sure what the motivation for that is; it would seem to introduce a big performance hit. For one it will be putting a lot more pressure on the GC.
If the default stack limit is fine for 90% of people, then it might be the wrong tradeoff to make 90% of people's code slower for the benefit of 10%. One can always refactor their code to not use recursion and avoid deep stacks, so the 10% have a workaround anyway. And deeply recursive code is generally not a good idea in the first place.
timid bolt
But wouldn't the solution then to be to increase the default stack size and not turn on stackless?
slender iron
ya, that could work too but then you reducing the ram available for heap
jaunty juniper
the impact on memory fragmentation is also a concern to me, especially when using displayio libraries on pico for example
slender iron
k, please comment on the pr so they proposer can see the concerns
timid bolt
Would you consider making the pystack size boot-time configurable to be an acceptable solution?
It looks like it may be possible to dynamically allocate _pystack in start_mp.
slender iron
ya, that could work. we did allow changing stack size when we shared the c stack.
timid bolt
oh, what was wrong with sharing the c stack?
slender iron
large call frame end up on the heap (iirc)
slender iron
async may end up being stackless anyway
I doubt the pystack is used for multiple concurrent threads
timid bolt
yes, coroutines are effectively stackless, but there are still a lot of function calls that happen between "awaits"
You would need a pystack for every thread. FWIW, it don't think CP should ever have preemptable threads; there would only be hardware threads, so that's not a large number of stacks. But if it were up to me, I would not even share a heap between hw threads, in which case there would not be a memory saving of multi-thread "stackless" anyway.
onyx hinge
I'm slightly delayed, will join in a few minutes
timid bolt
where do you join?
tulip sleet
do you want to do private or not?
timid bolt
I'm not sure what that means
tulip sleet
in a public channel or in DM's
timid bolt
public is fine
tulip sleet
can go to Amelia Earhart voice channel
brazen hatch
I too am in the chat (semi afk) if you guys want to talk about stackless more.
brazen hatch
This would kind of be the best of both worlds
something like sys.stacksize
0 for stackless
signed int for number of frames to be allocated.
realloc on set
wraith crow
Before you started testing stackless where you able to get the your code running well with larger stacksize settings? I may have been using the wrong parameters but when I started troubleshooting the pystack errors I didn't have much luck adjusting the stack size. I did find the stackless option before working too long on the stacksize parameters though.
brazen hatch
Before you started testing stackless where you able to get the your code running...
Yea, it all worked a-ok.
crimson ferry
we have supervisor.runtime.next_stack_limit for C stack
crimson ferry
not the pystack
brazen hatch
brb, imma go read the docs
crimson ferry
it's a bit open loop though: "The size of the stack for the next vm run. If its too large, the default will be used." https://github.com/adafruit/circuitpython/issues/4520
crimson ferry
and we have CIRCUITPY_RESERVED_PSRAM= in setings.toml (espressif)
memory knobs are quite handy
brazen hatch
we have `supervisor.runtime.next_stack_limit` for C stack
Uhh, I have been studying parallel search algos for exam, imma just slap it onto my fat 8mb board and see what it does.
Docs pain me.
brazen hatch
we have `supervisor.runtime.next_stack_limit` for C stack
tldr, no effect on static pystack. (8.0.0 stock)
I have no idea what it does.
set it to 100k and pystack exausted.
tried reload and reset
slender iron
we used to have a combined c and python stack
thats when next_stack_limit was added
it didn't get switched to the pystack when we enabled it
iirc the reason to do pystack was to avoid large allocations of stack frame onto the heap
because they'd fail when memory was fragmented
brazen hatch
Maybe if it gets reworked it would be the simpliest solution for dynamic stack size.
no fancy settings.toml, but good old next_ keys
Not a fan, but it's something I can pr myself. probably. hopefully..
Is it meant to work on reload or reset?
brazen hatch
I mean, I crawled through the code..
shared-bindings/Runtime.c->supervisor/stack.c
So apparently it will always boot with default stack size and then after the fact reallocate on next reload with the new stack size.
Or so it should..
Perhaps it is not then being realloc'ed on reload?
Crawling through shared-bindings/supervisor/__init.c__
Traced back to supervisor/shared/reload.c
Failed to trace port_wake_main_task. Cannot find relevant code.
Cannot find anywhere where pystack is actually realloc'ed.
It may be as simple as adding it during reload.
Hmm port_wake_main_task actually is port specific and xTaskNotifyGive feels like it slowly leads me to freertos code, so bye I have had enough. Imma yell at puter more tomorrow
Opening issue.
brazen hatch
Oh I was totally off with my horrible reversing.
I should have started from main, yea this is going very well 🥲
Yea it totally just gets alloc'ed by main.c from CIRCUITPY_PYSTACK_SIZE.
All that code was old unused code.
So that ain't going to happen without nvram of sorts, which I do not know how to use.
Perhaps settings.toml is simplier then next_stack_size
Though I do not know if that code would work so early?
If I can get the value from somewhere during boot and read it into main, it should be easy enough to override the rest.
Thanks for coming to my ted talk.
Goodnight.
slender iron
ya, I think settings.toml is simpler than next_stack_size
I think it should work early
brazen hatch
It should also be easier for code to verify as to if it has been set or not.
But whileI am at it, I will make a getter
proven garnet
Does CircuitPython discard module docstrings?
idle owl
Does CircuitPython discard module docstrings?
Discard how?
Like are they part of code size when compiled or whatever?
I'm fairly sure they're not included. Are you talking libraries or core?
proven garnet
Both
idle owl
OK, I'm not entirely certain. I feel like yes, they're not included. But I might be wrong.
proven garnet
That's my thought as well, but I don't have hardware on hand to test, but also wanted to confirm.
idle owl
@tulip sleet would probably know.
tulip sleet
yes, they are discarded
proven garnet
Context is questioning whether adding more information to a module docstring as an adverse effect (or any at all) on anything.
Sweet thanks!
tulip sleet
none at all
idle owl
That's what I thought.
proven garnet
Thanks y'all!
timid bolt
I'm working on a CircuitPython library. So I'm editing code in a git repo on my desktop PC. Then in order to test the changes I have to copy them to a device. Is there a convenient tool for syncing library code between your PC and device? Just wondering what other people do.
onyx hinge
if I don't just edit directly on the device, I use some variation on rsync -@2 to efficiently copy just changed files within a package. I'll just write out in a terminal and hit cursor-up enter, or if I'm really organized set it as my 'makeprg' in vim
and if there's just a single source file nothing is much better than anything else
(it helps to tell your editor not to write temporary files to CIRCUITPY if you edit files there; I know how to ask vim not to do that but I don't know about others)
timid bolt
Thanks. Is there a way to reload modules without resetting the board (like importlib.reload from CPython)?
jaunty juniper
you have to restart the code
ctrl-C ctrl-D or supervisor.reload() (or autoreload)
onyx hinge
timid bolt
Sounds good. Thanks.
proven garnet
I'm working on a CircuitPython library. So I'm editing code in a git repo on my ...
I have my own tool that does this (write PC code to board) for Linux. Just uses a process to detect changed code files from the PC and write them to the board.
chilly jay
Heya, so I have some questions again regarding CircuitPython. My university tasked me with implementing a non blocking writer for writing to an SDCard file.
Now, this could be done either purely in Python land or also in a C module, so I'd like to know (before I start implementing lots of boilerplate code and such) whether it is possible and also a good idea to malloc a buffer in which I could store strings before writing them to the card once a specific size is reached or flushing function is called.
stuck elbow
Heya, so I have some questions again regarding CircuitPython. My university task...
what if the board is restarted in the mean time?
chilly jay
as far as I know that should be neglectable. The reason why they want me to implement something like that is because they write a lot of sensor test data and writing lots of small data takes too long apparently.
stuck elbow
I think that should be handled at a higher level, where you know what the data is and how to best buffer it
chilly jay
I guess they could also just append their text to a list and then write that list to the disk every now and then
tulip sleet
Heya, so I have some questions again regarding CircuitPython. My university task...
Do they want the SD card writes to actually be non-blocking always (so the latency is known), or do they just want most "writes" to go fast (e.g. so you would mostly just buffer)
chilly jay
They want most writes to go fast.
tulip sleet
you can make a buffer in Python with a bytearray. Are you being asked to extend CircuitPython to do this? sdcardio or sdio? Main idea could be to add non-blocking writes to shared-module/sdcardio/SDCard.c. Find out what the blocking operations there are and either do them. This is a lot of low-level work
chilly jay
Though non-blocking would also work and it would even be fine if some data gets lost because of a ringbuffer or something.
tulip sleet
and requires knowledge of how SD cards work, their latencies, etc.
chilly jay
you can make a buffer in Python with a bytearray. Are you being asked to extend...
I am not exactly tasked with extending CircuitPython, just that the writes should either be faster or not block the code at all.
They would be fine with using a wrapper too I guess
tulip sleet
"most writes should be faster", it sounds like
chilly jay
yep
tulip sleet
so just make a buffer as big as you can, append to it, and then when it gets to full, write the whole thing out, clear it, and repeat
chilly jay
So just a simple python bytearray as a buffer would suffice? No need to extend or create a custom c module?
tulip sleet
no reason to
the actual write is still going to be blocking, but there will be fewer write operations. You may want to use some multiple of the filesystem block size (512 bytes or larger, depending on the size of the filesystem). Is this text or binary data?
at what rate are you writing data? 100 bytes/second? 100 bytes/minute??, etc.
chilly jay
It would be text data. 100 bytes a second sounds about right.
tulip sleet
don't concatenate strings into a new buffer each time you add stuff, that can be costly in terms of storage allocation. Allocate the buffer once and keep track of the next free byte, etc.
what board is this?
tulip sleet
it has a large amount of PSRAM, so you can allocate a big buffer.
chilly jay
Very nice. I'll get to work then. Thanks for the quick and detailed response!
tulip sleet
Very nice. I'll get to work then. Thanks for the quick and detailed response!
use sdcardio, not adafruit_sdcard. sdcardio is native C internally
chilly jay
I think they also used adafruit_sdcard previously so that would also explain things
But now with sdcardio and the buffer, it should be blazing fast for most writes 😄
tulip sleet
If you can use sdio, that would be faster
but you need the right kind of card breakout
not sure if S2 has sdio, I'll look
chilly jay
from what I've seen the board doesn't support sdio natively at least.
thorny jay
Need to take a feather with build in SDcard, those are wired for sdio.
chilly jay
Need to take a feather with build in SDcard, those are wired for sdio.
for now they do not want to change from FeatherS2's and FeatherS3 since they would have to order new ones with another complicated batch of paper work and funding.
tulip sleet
it probably will be fine
brazen hatch
I can't do the configurable pystack.
It's currently a file definition
moving it elsewhere really doesn't work
and all the funcs that work with it are written in minecraft echantment language
I give up.
I have understood how the _pystack array is used and allocated.
This didn't help.
Pretty sure it would take a boat load of changes to make it configurable.
crimson ferry
there's a decoder for the Standard Galactic alphabet 😉
slender iron
@tidal kiln do you do much logic analyzer stuff?
tidal kiln
<@337989520962093057> do you do much logic analyzer stuff?
wouldn't say "much", but have done some. for troubleshooting i2c, spi, etc. need something scoped?
slender iron
I've been working on making the scorpio useful as a logic analyzer
wondering if you want to try it. Do you usually use a saleae?
tidal kiln
yep. got an older model saleae. 8 chan. just fast enough for spi. works fine for my needs.
slender iron
how fast do you usually run it?
tidal kiln
only as fast as needed. may also have a host PC issue, as fastest setting often leads to "couldn't keep up" errors.
24MS/s is fastest possible. but i can't achieve that on my setup. 16MS/s is more typical for SPI, slower for I2C.
slender iron
I think the appeal for me is that anyone with an rp2040 will have a (slowish) logic analyzer
jaunty juniper
I would love that 😉
slender iron
@jaunty juniper want to try it?
tidal kiln
for I2C i'm more at 2MS/s
jaunty juniper
I don't have specifically a scorpio though
slender iron
any rp2040 will work
tidal kiln
i also haven't snagged a scorpio yet
slender iron
scorpio just has the logic level translator to protect against 5v
tidal kiln
yep
jaunty juniper
yeah I have a pico to spare
tidal kiln
i have several picos
slender iron
kk, let me push the latest
right now it is cli capture and then you can load into pulseview to see the trace
so you can install pulseview now if you want
tidal kiln
found an appimage i downloaded circa july 2021. think i was giving the sigrok stuff a try back then. appears to still be latest ver 0.4.2