Wowki only works well with the Pi Pico... I'm not sure if it has ESP support.
#circuitpython-dev
midnight ember
willow totem
ah wicked thanks @slender iron
solar whale
Not that I object to learning a new tool. Hopefully a new or updated guide will follow soon.
midnight ember
Ohh that's good to know. good info @willow totem
The Wokwi simulator is excellent for helping people with issues in #help-with-circuitpython
willow totem
Yep wokwi supports ESP32's, Pico+PicoW, arduinos, stm32. They have components (led/LCD/etc) and screenshot taking with lcd peripheral.
My test was for heap size (actually done by gregorik from espressif but needed fixing up)
candid sun
ship it!
solar whale
π
willow totem
Yay, live release on stream
midnight ember
Sold to 9.0!
turbid radish
thorny jay
I just hope cardputer keyboard will be in a 9.0.x or 9.x
midnight ember
Releases closer to October get the Hacktoberfest sprint bug hunt treatment. Something to consider for a major release in the fall.
slender iron
I just hope cardputer keyboard will be in a 9.0.x or 9.x
we can do 9.1 pre-releases after 9.0.0 is branched too
tulip sleet
I just hope cardputer keyboard will be in a 9.0.x or 9.x
it will be in 9.0.x something, I assume
slender iron
thanks all!
candid sun
thanks for hosting @onyx hinge !
lone axle
Thanks for hosting Jeff! Have a great week everyone!
short tendon
Thanks!
gilded cradle
Thanks everyone
willow totem
@midnight ember the pi's with vscode-insiders pointed at them using remote explorer / remote SSH extension
midnight ember
The batch script at the end of this issue is a workaround/fix for "Github Desktop changes paths every version update". It works very well and can recommend.
Instead of pointing the path to the new Github Desktop .exe manually every version change just point it at the .bat and it will keep track of it every version change.
@tulip sleet Synergy was the virtual KVM I was thinking about. There's no mention of it being owned by Google at any time. I'm sure at one time there was a google branded fork of Synergy. https://en.wikipedia.org/wiki/Synergy_(software) It does still exist.
onyx hinge
fwiw the original synergy went under some other sort of closed source license. an open source fork of synergy is https://github.com/debauchee/barrier but I've never actually used it.
Here is the notes document for next Mondayβs 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 weβll read them off during the meeting. Hope to see you there! <@&356864093652516868>
https://docs.google.com/document/d/11hSXWvZY9pqH1RNT7wBU3gzwwDjjEJVO2yg_-E999D4/edit?usp=sharing
midnight ember
Thanks for hosting @onyx hinge!
short tendon
@lone axle I'm playing with ruff, and found one thing: https://docs.astral.sh/ruff/faq/#how-does-ruffs-linter-compare-to-pylint
Many of the rules we are disabling, ruff doesn't care about in the first place. Would we want to remove all the # pylint: disable=xxxx?
midnight ember
That's a good point. There are a lot of pylint comments embedded everywhere, in everything.
lone axle
Many of the rules we are disabling, ruff doesn't care about in the first place. ...
Nice, that is good to see that it's defaults are somewhat in line with our usage at least. I would be in favor of removing the pylint disable lines anywhere possible.
(which is perhaps everywhere)
short tendon
And as an FYI - added it to requests and only got one file changed
Also - it looks to match isort.
midnight ember
Are the spaces between imports really necessary? From my point of view I'll only ever import things that exist on the microcontroller. So I don't understand the need for sorting by 3 different types 2 of which mean nothing to me. Maybe it's more useful for Blinka users that have a Linux environment? If that's the case then I'm ok with the change otherwise I preferred the condensed chunk of imports. It helped make scripts smaller.
willow totem
Can I get the pythonistas role please, feel bad making others read my bits by always being text only
lone axle
Can I get the pythonistas role please, feel bad making others read my bits by al...
Done, you're now in circuitpythonistas too.
willow totem
Much obliged
fringe yarrow
I was wondering about debug builds.. I've done that but then I can't figure out how to see the debug output.. the serial console just puts out the normal stuff
lone axle
I was wondering about debug builds.. I've done that but then I can't figure out ...
On some boards there is a pin labeled DBG on the silkscreen that you need to hook up a USB to serial converter in order to read the output from it. (or using something else that can read from a serial line)
devout jolt
Hey @willow totem do you have tips on how you're using custom CircuitPython firmware on WokWi? I love using it for quick testing and teaching, but the CirPy UF2 images I've seen on Wokwi are pretty old
short tendon
Are the spaces between imports really necessary? From my point of view I'll onl...
whether ruff or isort is added, the spaces will be there. It's to show separation of where the import is from. Once your line-ending issues are resolved, it won't be a big deal
@lone axle and @onyx hinge did a quick ruff example. Took from micropython, our rules and some line-ending things:
https://github.com/adafruit/Adafruit_CircuitPython_Requests/pull/171
Most of the changes are because I also removed all the pylint comments. Since requests already has isort it didn't change any of those
lone axle
Would main be different from 9.0.0rc.1 right now? My assumption would have been that they were pretty much the same but I'm noticing some different behavior from a build of main vs. a downloaded rc.1 around USB Host stuff with the Feather USB Host device.
onyx hinge
@lone axle compared to 9.0.0-rc.1 there are two emrged PRS: #9020 and #9055
you can check this online: https://github.com/adafruit/circuitpython/compare/9.0.0-rc.1...main
lone axle
Thank you. I think something must be wrong with my local environment
my builds of main are succeeding but USB Host doesn't seem to work at all with them. When I download builds from S3 (downloaded abosolute newest as of a few min ago) it does work albeit with the unplug / replug. Somehow my local builds aren't complete or something
onyx hinge
check if your submodules are up to date?
lone axle
I had forgotten that at first but have updated them now
onyx hinge
that's the only guess I have at the tip of my tongue
willow totem
Hey <@375674153203662851> do you have tips on how you're using custom CircuitPy...
So this guy from espressif setup some heap size sketches in arduino micropython circuitpython +1 i forget. I just modded it, but the basic principle is it loves elf and bin files. I think a uf2 is out of the question except maybe for pico/w. This is wokwi-ci btw https://docs.wokwi.com/wokwi-ci/getting-started
But also vscode does similar, and I believe you can do the same online, but you need to click into the interface then press F1 (what a choice of key - if loads you forgot to click inside code not diagram), then a command pallete loads like vscode, and you search for upload
Here's the thread @devout jolt that mentions the original repo, and urish's changes to instead add workflow assets (I set reports to be uploads as release assets), and notification when the C3 is fixed https://github.com/wokwi/wokwi-features/issues/754
This bit is the bit of code you care about, it makes a disk image etc then merges all together after building https://github.com/tyeth/esp32-lang-lab/blob/main/tests/python/circuitpython/test-circuitpython.sh#L49-L139
devout jolt
Here's the thread <@352910176736772096> that mentions the original repo, and uri...
awesome thanks. I've only ever use it via the website and mucking about with "diagram.json" to specify different firmwares
willow totem
yeah similar, but had a play with CI a few times, and vscode a load for attempting to do cpy/mpy things, c++ works fine, build times online are too long, but I've got access to long builds. I really want CI to work for me, but there will eventually be a cost so still scoping what works best.
I secretly hope some philanthropicly minded org goes and buys out the project / Urish for a few years, only so it's fully open sourced and equally worth his time. 70% is in github anyway
devout jolt
I secretly hope some philanthropicly minded org goes and buys out the project / ...
ahha totally. how did you even learn about the wokwi command palette? I've never come across it
willow totem
pure luck, maybe the man told me when I asked how to do something, or maybe I spotted it on the wokwi discord or some random docs (he had vscode instructions in a google doc for a few months), but it was like a gotcha moment and remember feeling like they thought it was intuitive (but F1 π€¦ )
Useful server to be honest, great for silly / crazy questions, example projects, bugs although get some helpers fending off initial bug reports sometimes, but Urish is fantastic and loves the simulator to match reality (where poss)
short tendon
@midnight ember I would be curious if you checked out this branch: https://github.com/adafruit/Adafruit_CircuitPython_Requests/pull/171 and then ran pre-commit, if you would have any errors...
lone axle
ruff doesn't seems to care about strings using " instead of ' like black does. Is that something we care about? I would assume without this check we will eventually diverge from the consistency of using " unless necessary that black has built for us. Personally I don't really mind either way, but I do know that in my own code I kind of switch it around somewhat randomly based on how I'm feeling in the moment so without some tool assistance I don't use them consistently.
oh, actually I did not read enough of the documentation before running I've just now noticed that running as a linter is different from running as a formatter
okay, yeah nevermind it does make them consistent when you run the formatter. Interesting that the commands are seperated. I assume that ultimately we'd integrate this with pre-commit though so it ends up still being a single command to run?
short tendon
I did, my setup has both enabled
lone axle
I did, my setup has both enabled
can you share your pre-commit config?
proven garnet
I have a setup similar to what it might look like that mostly copies the CircuitPython pre-commit config (pylint, but also pydocstyle, isort, and pyupgrade).
lone axle
does it use https://github.com/astral-sh/ruff-pre-commit?
short tendon
Scroll up 3 messages. I have a whole PR
lone axle
ahh, thank you
I guess they renamed their github account
https://github.com/charliermarsh/ruff-pre-commit automatically redirects to https://github.com/astral-sh/ruff-pre-commit
short tendon
The important changes are in:
.gitattributes
.pre-commit-config.yaml
pyproject.toml
Should totally be able to update cookiecutter with these
orchid basinBOT
short tendon
It will make the same changes as isort (and maybe a few others)
midnight ember
Do you have the 1MB version? I just received the 65MB Daisy Seed. It's much smaller than I expected, KB2040 size. I'll look into spinning up a new dedicated linux pc this week, have a feeling cmake builds are in my near future.
marble hornet
Yay! π Congrats and a big thanks
tulip sleet
my builds of main are succeeding but USB Host doesn't seem to work at all with t...
tinyusb was recently updated
orchid basinBOT
lone axle
I think got updated with a fetch-all-submodules. I'll try a fresh clone later on.
slender iron
git status will tell you
lone axle
β― git status
On branch main
Your branch is up to date with 'origin/main'.
nothing to commit, working tree clean
spare jacinth
slender iron
that looks good to me
fringe yarrow
On some boards there is a pin labeled `DBG` on the silkscreen that you need to h...
I can't find any information about this. Just about using JTAG. Nothing jumps out at me as a "dbg" pin on the pinout of a bare esp32..
fringe yarrow
hrm.. yeah, i built this DEBUG=1 but i get no more info from UART0 than before, just the normal code.py output.
slender iron
on c3?
fringe yarrow
no, ESP-WROOM-32..
Sorry, I've changed gears from that i2c pulldown thing until I get another one to ensure it's not just hardware ghosts.. this is that Sunton cheap yellow display that I'm trying to get going π
slender iron
DEBUG=1 should turn on the CONSOLE uart from the IDF
fringe yarrow
is there a log level or something else I have to change? because it shows nothing besides normal code.py output on UART0
slender iron
make sure you clean and rebuild
fringe yarrow
that's probably it..
slender iron
idf settings changes (which DEBUG=1 does) aren't smart enough to trigger an IDF rebuild
fringe yarrow
too little flash to make a debug build, apparently π the fun never stops.
fringe yarrow
is there a list somewhere of mpconfigboard.mk parameters like that? I feel like I always have to hunt for things like that
and as i ask, i find the Documentation pdf
slender iron
generally it's CIRCUITPY_<modulename>. py/circuitpy_.mk has a full list
fringe yarrow
doesn't seem like disabling ulab or any other module changes the size of the final image.
ugh.. i wasn't making clean again.. disregard..
onyx hinge
https://github.com/jepler/wwvbpy/pull/72/files @lone axle here's one of my experiences switching to ruff
fringe yarrow
idf settings changes (which DEBUG=1 does) aren't smart enough to trigger an IDF ...
Is there a easy way to increase the debug log verbosity?
blissful pollen
Is there a easy way to increase the debug log verbosity?
You could add your own printfs if you know where in the code you are looking at
solar whale
π 9.0.0 loaded on several boards -- "no anomalies to report"
tulip sleet
π 9.0.0 loaded on several boards -- "no anomalies to report"
thank you for testing!
slender iron
Is there a easy way to increase the debug log verbosity?
The IDF's menuconfig has one more level I think
you can do make BOARD= DEBUG= menuconfig
vagrant wind
I found the problem. Once I run . esp-idf/export.sh in a terminal session the make html fails. I need to start another terminal sessione and then it works
empty salmon
Am I missing something with the synthio documentation? It has the following:
- attack_time: float - The time in seconds it takes to ramp from 0 volume to attack_volume
- decay_time: float - The time in seconds it takes to ramp from attack_volume to sustain_volume
- release_time: float - The time in seconds it takes to ramp from sustain_volume to release_volume.
I am guessing that attack_volume and sustain_volume should be attack_level to sustain_level. There is no release_volume.
tulip sleet
I found the problem. Once I run `. esp-idf/export.sh` in a terminal session the ...
Yes, export.sh activates a Python venv, so you have to reinstall from requirements-doc.txt in that venv
tulip sleet
@slender iron opinion about whether to add keypad_demux to 9.0.x or just leave it on main and publish a 9.1 beta shortly? I have a small number of other things for a beta that are not necessarily consequential (e.g. renaming nrf -> nordic)
I guess cardputer keyboard support on a stable release would be nice
proven garnet
https://blog.adafruit.com/2024/03/18/circuitpython-9-0-0-released/
9.0.0 @ 9:00 pm local time, feeling like an extra lucky release this time!
tulip sleet
9.0.0 @ 9:00 pm local time, feeling like an extra lucky release this time!
That was deliberate - I finished the post at 8:40 and decided to set it to publish at 9 π
onyx hinge
I guess we can release CircuitPython 13 at 1PM
empty salmon
So CP 10.0.0 on October 10 at 10 am ?
tulip sleet
only if it works out π
@lone axle et al: re library docs. Is it written down somewhere that we should document the constructor (__init__()) in the class docstring? I noticed some missing documentation in adafruit_hid, and it was because the info was in the __init__() docstring. Apparently the convention is not python-wide. There is a setting autoclass_content = 'both' in the sphinx config file, but we don't set that by default. https://blog.sandipb.net/2016/10/10/documenting-both-class-and-constructor-in-sphinx/
i wonder if there are other libraries with this issue. I worked on adafruit_hid years ago when this was all really early.
hmm, never mind, I think adafruit_hids conf.py is out of date
most libraries have autoclass_content = 'both'
lone axle
<@382939733107408897> et al: re library docs. Is it written down somewhere that ...
I don't believe it is written anywhere. The two places that I know of that it might make sense would be the RTD page of "Sharing a Library" guide: https://learn.adafruit.com/creating-and-sharing-a-circuitpython-library/sharing-our-docs-on-readthedocs, but currently this is focused on the actual RTD service integration, not the docstrings that the docs are built from. or maybe to a lesser extent this one about typing information: https://learn.adafruit.com/creating-and-sharing-a-circuitpython-library/typing-information which does include a section about type hints in __init__ but not specifically anything about it's interoperation with sphinx docs.
tulip sleet
thanks!
lone axle
We can try to add a check for that with the ruff addition automation. I wonder how many others are inconsistent with the majority, it'll be nice to get them on the "same page"
short tendon
We can require those via pre-commit with either pydocstyle or ruff...
onyx hinge
someday we also need to modernize what we do in the learn github repo for python scripts, right now there's pylint (old version) + no black
(and not via pre-commit either)
slender iron
<@252717193496756235> opinion about whether to add `keypad_demux` to 9.0.x or ju...
I'd just do 9.1 and do a pre-release
and leave 9.0.x for fixes only
It may be a good time to do the 1.22 MP merge and make upstream PRs
midnight ember
<@126221640559362048> I would be curious if you checked out this branch: https:/...
I don't have anymore sweeping changes to make so I wouldn't be a good person to test the new ruff stuff.
short tendon
You could checkout the branch and then re-update one of the newest and see what happens
You should get no lf errors
midnight ember
FoamyGuy brought up a question about the built-in binascii vs adafruit_binascii. Is binascii built-in for the majority of the boards or should I only use adafruit_binascii?
onyx hinge
@midnight ember for web-based stuff? All those boards are going to have builtin binascii, or at least I'd be rather surprised otherwise
midnight ember
makes sense
onyx hinge
someone needs to make "caniuse.com" but for micropython & circuitpython π
alternate useful idea: a way to query "has wifi but not binascii"
midnight ember
is that the programming equivalent to LMGTFY? π
midnight ember
alternate useful idea: a way to query "has wifi but not binascii"
That sounds more like something to run actions on to really know for sure. I wouldn't even know where to begin to look that up.
The list would probably be a mile long now. It would be a shorter list to know what board it's not included in.
lone axle
The support matrix makes finding it out fairly easy it turns out: https://docs.circuitpython.org/en/latest/shared-bindings/support_matrix.html?filter=binascii (I didn't know it had filtering, this is really convenient actually, I should use it way more often.)
It would be a shorter list to know what board it's not included in.
Got me thinking and it turns out this filter supports negation too! so you can search -binascii to find boards that do not support π€―
short tendon
@lone axle who can update https://learn.adafruit.com/welcome-to-circuitpython/pycharm-and-circuitpython? With 9.0.0 released we can add information about the board stubs
onyx hinge
is that the programming equivalent to LMGTFY? π
It is a website that shows support for a huge number of web browser features, by browser & version. https://caniuse.com/?search=canvas
lone axle
<@382939733107408897> who can update https://learn.adafruit.com/welcome-to-circu...
I am not certain of the entire list of who. I do have access to edit the page though so I could add a section for that.
onyx hinge
ooh and you can have a positive and negative filter both. "-binascii ssl" gives 0 matches. https://docs.circuitpython.org/en/latest/shared-bindings/support_matrix.html?filter=-binascii&filter=keypad is an example where there are matches: keypad but no binascii
lone axle
Nice!
onyx hinge
but of course there could be some Airlift-having device with no binascii
midnight ember
I highly doubt anyone with a circuit playground express is going to want to run base64 encoding. thank you. that'll do. going forward with using the built in.
hmm that could be a problem, i don't want to disclude airlifts.
midnight ember
It is a website that shows support for a huge number of web browser features, by...
The irony is I could have googled that. π
onyx hinge
I did also literally give you the website's name, you could have looked at it too π
midnight ember
ask Dan, I'm bad at looking at things unless someone shoves it in my face a couple times.
too many projects, too easy to get sidetracked with rabbitholes
lone axle
hmm that could be a problem, i don't want to disclude airlifts.
the examples are already using the built-in wifi for wifi.radio which excludes airlift AFAIK. TBH I would be pretty surprised if any of the SAMD21s would be able to use airlift and make requests successfully anyway. If binascii is on M4s I'd just roll with that since I think it's overwhelmingly likely airlift based project will be using something "bigger" than SAMD21 anyhow.
I didn't realize until now that circuitpython does not have base64 module like CPYthon does. Is that excluded intentionally to save space or would there be some interest in having it added as core module? It might be nice to make code a bit more portable between circuitpython and cpython.
midnight ember
think i'm just going to delete the twitter example. it's broken, their new api is broken, it's not worth the hassle. the documentation is horrible. i got Mastodon up and running in about 15 minutes. i shouldn't have to spend more than 24 hours of research trying to figure out which twitter api documentation is current and what works. i'd rather go back to playing with the crypto chip.
does anyone have any objections to deleting the broken adafruit_circuitpython_requests Twitter API example?
lone axle
removing it sounds fine to me.
midnight ember
even if someone needs binascii that doesn't have it on their board (including airlifts) they can just import adafruit_binascii and change the code a little to use that.
i just want to avoid support problems if the change will cause some.
midnight ember
I didn't realize until now that circuitpython does not have `base64` module like...
that is what lead me to finding the community library circuitpython_base64 to begin with.
tulip sleet
does anyone have any objections to deleting the broken adafruit_circuitpython_re...
sure
cedar moth
Good morning all. I was wondering if there was any ballpark estimate or rule of thumb on when the next merge from upstream MicroPython might occur?
slender iron
I was just thinking about it. Are you interested in MP 1.22?
cedar moth
I need to check the exact MP version but an update to the deque implementation was recently merged into MP (two nights ago) and it would be very helpful for a project I am working on. Not a show stopper by any means, I have been building my own version of CP with the proposed changes, but if it could be integrated into CP that would be helpful.
slender iron
Ah, it'll be in the next MP release then
cedar moth
OK, then no hurry! π
slender iron
They try to release every two months but I'm not sure when they will
You can make a PR to CP with the changes that MP accepted
cedar moth
Oh, OK. I will do that, thanks tannewt!
slender iron
np
midnight ember
OSN_CREDENTIALS_B = binascii.b2a_base64(b"" + str(OSN_CREDENTIALS) + "") # this works but appends \n
OSN_CREDENTIALS_B = binascii.b2a_base64(b"" + str(OSN_CREDENTIALS) + "", newline=False) # this doesn't
``` fails with `TypeError: function doesn't take keyword arguments`we must not fully implement it
tulip sleet
```py
OSN_CREDENTIALS_B = binascii.b2a_base64(b"" + str(OSN_CREDENTIALS) + "") ...
you can just do .strip() to remove the newline
midnight ember
foamyguy got it to work in a pr we're having the discussion in https://github.com/adafruit/Adafruit_CircuitPython_Requests/pull/167 ```py
import binascii
dir(binascii)
['class', 'name', 'dict', 'a2b_base64', 'b2a_base64', 'crc32', 'hexlify', 'unhexlify']
binascii.b2a_base64(b"hello")
b'aGVsbG8=\n'
binascii.b2a_base64(b"hello", newline=False)
b'aGVsbG8='
>>> import binascii
>>> binascii.b2a_base64(b'hello').strip()
b'aGVsbG8='
i was using slices BASE64_STRING[2:-3] which works about the same. just that foamyguy said there might be a way with the argument to not have it append the \n at all from the start.
tulip sleet
i think it comes from MicroPython, and the often leave off stuff like that when there's a workaround, to save space
midnight ember
that worked, thank you. think i'll use that.
fringe yarrow
In the ongoing saga of this Parallel wired ST7789 ESP32 all in one board.. I've removed the display definitions in board.c, pins,c, etc, and made a circuitpy image with CIRCUITPY_PARALLELDISPLAYBUS enabled to let me try to get the display working from within code.py rather than recompiling the image after modifying board.c over and over.. and after providing the right pin info to the paralleldisplaybus.ParallelBus constructor, the screen turns on, but shows the same lines and garbage that it did as compiled in.. so.. at this point I think the problem lies somewhere within the differences of how CircuitPython paralleldisplaybus is handled vs how the equavlent bus is constructed in the Arduino sketch.. I think this is the point where I set this aside.. π
slender iron
In the ongoing saga of this Parallel wired ST7789 ESP32 all in one board.. I've...
where do you get these boards?
fringe yarrow
where do you get these boards?
https://www.aliexpress.com/item/1005006284154750.html <- This is the one I've been working on for a few days.. The SD card, amplified GPIO speaker, and touch were all trivial to make work.. the display must need some configuration or kick in the pants that the arduino driver provides while the circuitpython one doesn't..
midnight ember
actually that didn't help as much as i'd hoped ```py
OSN_CREDENTIALS = str(osnusername) + ":" + str(osnpassword)
OSN_CREDENTIALS_B = b"" + str(OSN_CREDENTIALS) + ""
OSN_CREDENTIALS_B = binascii.b2a_base64(b"" + str(OSN_CREDENTIALS)).strip() # strip removes \n
BASE64_STRING = str(OSN_CREDENTIALS_B) # bytearray
TRUNCATED_BASE64_STRING = BASE64_STRING[2:-1] # slice bytearray head/tail
BASE64_STRING : b'VXNlcm5hbWU6MDEyMzQ1Njc4OQ=='
TRUNCATED_BASE64_STRING: VXNlcm5hbWU6MDEyMzQ1Njc4OQ==
A logic analyzer will really help with debugging this. You could get an rp2040 board to do it
midnight ember
@short tendon should i run black first or just try pre-commit?
fringe yarrow
That's the 28R which has a SPI screen.. I have one of those en-route, and if I'm honest, should be the one, if any, that get submitted for CircuitPy support.. it's by far more popular than the 22C, which I just happened to have the bad luck of getting
midnight ember
@short tendon ```
C:\Users\Devon\Documents\GitHub\Adafruit_CircuitPython_Requests\examples\wifi\expanded>pre-commit run --files requests_wifi_api_openskynetwork_private.py
[INFO] Initializing environment for https://github.com/charliermarsh/ruff-pre-commit.
[INFO] Installing environment for https://github.com/charliermarsh/ruff-pre-commit.
[INFO] Once installed this environment will be reused.
[INFO] This may take a few minutes...
check yaml...........................................(no files to check)Skipped
fix end of files.........................................................Passed
trim trailing whitespace.................................................Passed
ruff.....................................................................Passed
ruff-format..............................................................Passed
reuse....................................................................Passed
Thanks for checking
proven garnet
<@1152454697614069811> ```
C:\Users\Devon\Documents\GitHub\Adafruit_CircuitPytho...
@short tendon I think it might be helpful to swap the formatting and linting ruff commands so any linting errors get reported on new line numbers instead of old ones.
short tendon
Totally makes sense
short tendon
<@1152454697614069811> I think it might be helpful to swap the formatting and li...
Hmmm. The import fix happens in linting, not formatting. So they can both change lines...
proven garnet
I think it will still report unresolvable errors on the newer lines though. I had an issue using pydocstyle rules where my CI was reporting older lines for this reason, which is why it's on my mind.
Unless I'm misunderstanding what you mean by import fix - I assume you mean isort
short tendon
I do
proven garnet
Interesting it's not in formatting though, though I guess isort really is technically? linting???
short tendon
Yeah, it is strange
proven garnet
For what it's worth, I think you'll still get the correct line numbers even after the reordering.
short tendon
Sounds good. I also assume, those will change less often
@proven garnet do you think we should run PL? The rules are different and not sure how we would find out what ones to disable globally. Strange ones in that PR are: ["PLR2004", "UP030", "UP031"]
fringe yarrow
A logic analyzer will really help with debugging this. You could get an rp2040 b...
I suspect that whatever is required to make this screen work at this point is beyond a board definition and into modifying the common-hal/paralleldisplaybus code
manic glacierBOT
slender iron
I suspect that whatever is required to make this screen work at this point is be...
I'd be surprised if that were true. paralleldisplaybus is pretty standard and simple
proven garnet
<@135415127879974912> do you think we should run `PL`? The rules are different a...
I'm torn on PyLint. Ultimately yes, but it would mean changing all the # pylint: disable=thing to # noqa: newthing#### which is involved.
short tendon
Yeah, although there would be fewer of them
fringe yarrow
I'd be surprised if that were true. paralleldisplaybus is pretty standard and si...
There's a lot of assumptions in there, hardcoding dc_levels and the flag set without the opportunity to customize it for one board or another..
proven garnet
I don't think there are many relevant global disables - the .pylintrc files has them and is pretty consistent between libraries.
slender iron
ya, perhaps. a logic analyzer will tell you
midnight ember
C:\Users\Devon\Documents\GitHub\Adafruit_CircuitPython_Requests\examples\wifi\expanded>pylint requests_wifi_api_openskynetwork_private.py
************* Module requests_wifi_api_openskynetwork_private
requests_wifi_api_openskynetwork_private.py:12:0: C0411: standard import "import binascii" should be placed before "import adafruit_connection_manager" (wrong-import-order)
``` local pylint told me to put binascii before connection manager. i did that and isort failed during the job run.Also, do you know how to set UP to use a certain pyupgrade version?
fringe yarrow
Any idea why, with the latest pull from the git repo, my builds still say "CircuitPython 9.0.0-rc.1-7" leading to reminders from circup that there's a newer version..
slender iron
make sure to fetch tags too
onyx hinge
I don't think there are many relevant global disables - the `.pylintrc` files ha...
@midnight ember @proven garnet there are some examples-only modifications to the pylint settings made for examples only, fwiw. ```yaml
types: [python]
files: "^examples/"
args:
- --disable=missing-docstring,invalid-name,consider-using-f-string,duplicate-code
if using isort it will be necessary to disable pylint's wrong-import-order, since as you see they disagree.
midnight ember
consider-using-f-string was more like "you must use f-string or i'll bug you to death about it"
short tendon
I updated my ruff branch to use PL (pylint) and UP (pyupgrade), just to see the changes
I only needed to re-add 2 noqa statements
midnight ember
like this function that i've used for about 2 years now it complained about ```py
def _format_date(datetime):
return "{:02}/{:02}/{:02}".format(
datetime.tm_year,
datetime.tm_mon,
datetime.tm_mday,
)
so i rewrote it with only f-strings, probably an improvement that needed to be done anyway. now i have a nice f-string version added to my snippets.
```py
def _format_datetime(datetime):
""" F-String formatted struct time conversion"""
return (f"{datetime.tm_mon:02}/" +
f"{datetime.tm_mday:02}/" +
f"{datetime.tm_year:02} " +
f"{datetime.tm_hour:02}:" +
f"{datetime.tm_min:02}:" +
f"{datetime.tm_sec:02}")
We could also run ruff twice, once for code and the other for examples where we disable the isort rule
short tendon
The goal of linting and formatting is so code is safe (linting) and looks the same (formatting). It makes code reviews easier and takes peoples preferences to how it should look (there's like 5 ways to format a string) out of it. It would be up to the owners to determine what rules they want to enforce
stuck elbow
it makes code reviews harder, imo, because it removes important code smells
and most of the linting rules are completely useless
short tendon
Personally having different rules feels strange. Because then when you move example code into real code, it re-formats it.
proven garnet
Also @tulip sleet I can look at that fun pre-commit issue this evening, if you want to hand it off to me for a little bit.
tulip sleet
Also <@329766224093249548> I can look at that fun pre-commit issue this evening,...
that would be great -- I found the whole thing quite frustrating
midnight ember
honestly i've learned most of my formatting from learn guides and repos from circuit python. i think black and pylint had a lot to do with that.
midnight ember
Personally having different rules feels strange. Because then when you move exam...
right but the formatters are kind of like the glue that binds all of our code otherwise it would be the wild west for examples. there is a nice sense of consistency when reading learn guides because code is formatted before being shared.
proven garnet
right but the formatters are kind of like the glue that binds all of our code ot...
Relevant
midnight ember
It's not lost on me. I know. I really liked the formatting without the line spaces. :/
stuck elbow
foolish consistency is the hobgoblin of the little minds
short tendon
@midnight ember also, just so you know (and haven't checked specifically in CP), f-strings are faster and more memory efficient than any other string formatting. It is sad that we still need to concatenate long strings, so lose some benifit there...
midnight ember
oh i use f-strings for everything now. i'm not complaining. in fact it was very useful to kind of be forced to update that one to f-strings. i didn't even realize all this time that it needed to be updated too until pylint said something.
wraith crow
like this function that i've used for about 2 years now it complained about ```p...
Is there a formatting override to disable the f-strings check? In some cases I think the .format approach may be more readable.
midnight ember
probably something like # pylint: disable=consider-using-f-string
Now I have to go rename my entire nrf examples directory.
short tendon
@wraith crow yes, we can ignore f-strings if we want (globally, or selectively)
midnight ember
i like the recommendation as i do want to move everything to f-strings. i don't think it should be mandatory just yet though.
slender iron
@tulip sleet I'm running into reset issues with the usb host featherwing code and I'm tempted to remove all of the never_reset and bulk reset stuff
tulip sleet
<@329766224093249548> I'm running into reset issues with the usb host featherwin...
there's an issue for that π do you want to do it on main or 9.0.x?
slender iron
I was debating doing it on main
it may not save me from this problem though
because the finalizers aren't run in a particular order
so I could still deinit the spi before the max
tulip sleet
I was thinking of #8991, that's just finishing off pwmout
but we still need reset_forbidden, I assume (never EVER reset)
oh, it's #8960, that's everything
tulip sleet
@slender iron of interest re IDF iram, cache, etc.: https://github.com/micropython/micropython/issues/14005#issuecomment-1996507230
ok, it's only in ESP-iDF 5.0.x, so not an issue for us
good for us for not supporting older ESP-iDF
slender iron
ya, jim sent me a link to it
I need to write up the noxip thing for clang folks
and finish my impl
stuck elbow
Is there a formatting override to disable the f-strings check? In some cases I t...
especially since not all boards have f-strings enabled
wraith crow
Maybe I mis-read but I thought Dan said all boards do now have f-strings enabled....
stuck elbow
oh? I'm sure they didn't work on my samd21, but I didn't try with cp 9
wraith crow
π€· either way, there is still the option to disable the check I guess π
stuck elbow
there are also things you can't do with f-strings
short tendon
I'm pretty sure f strings are on any board. Didn't think you could turn that off. Many libraries use them...
tulip sleet
especially since not all boards have f-strings enabled
f strings have been on all boards since 7.3.0\
stuck elbow
f strings have been on all boards since 7.3.0\
I must have some old versions installed, sorry
short tendon
I think question number 1 for linting/formatting is do we want to sort imports. If so, it will probably be easier to add isort first and then switch to ruff (if that's the thing we're doing). I'm 100% happy to script it out and open all the PRs (I do know you guys mentioned direct commits to main)
tulip sleet
Do ruff and isort produce potentially different orders?
short tendon
No. I'm pretty sure ruff implemented isort
There should be no changes
That's why it feels like a good first step
I am also happy (if people can decide on rules) to do the whole thing.
tulip sleet
if we have to change the same places to do isort+pylint+black as opposed to just ruff+ruff, then we may as well move to ruff without an intermediate step. It could be done in batches of libraries since there would be cleanup
short tendon
Fair. Please let me know how I can help. I can probably script about 90% if not more. I can also create diff files so you could commit to main
tulip sleet
could be scripted, ifthere is manual cleanup needed could be done with pre-commit locally. We might be removing a number of # pylint: ... comments, if I understand correctly.
short tendon
Yes, I would remove all of those and replace with the new ones (if we are using rust:pylint)
If we open them as PRs, then anything not caught by scripts would be easy to find since it would fail
tulip sleet
we could generate PR's to switch over with a script. Many PR's would fail and then they could be cleaned up by hand. If there are easy cleanups those could be automated.
short tendon
This would be my flow:
- Pick 5-10 repos of different sizes (like ticks, minimqtt, portal base) and create 3 (maybe 4) PRS:
a. With sorting enabled
b. With pyupgrade enabled
c. With pylint enabled (different from the normal pylint)
d. There is also one for docstrings, not sure that would be wanted... - Create a Google doc that lists the PRs and allows for comments
- Either meet over the doc, or have someone review comments and make decisions
- Use said decisions to update all the repos, update cookiecutter and update docs to the rules and why (so someone doesn't come back in 6 months asking why we do x)
strange ravine
Probably better fit for #help-with-linux-sbcs but, i wanted to know will there support for blinka on the orange pi zero 3?
proven garnet
@tulip sleet It seemed like all the dependencies were up to date, so I'm not sure what was causing the error. However, I got the CI to pass by installing pre-commit via pipx as opposed to pip so it has an isolated environment and that worked.
I wonder if something is competing with it, but whatever it is, it'd take more digging. Using pipx is a pretty good permanent solution anyway though I think the contributing guide would need to be updated if that's the way it should be done locally too.
short tendon
I wonder if something is competing with it, but whatever it is, it'd take more d...
Anything I can help look at?
proven garnet
No, I think it's good now. If you're interested in figuring out how its using an older virtualenv, that'd at least solve the mystery. But it really only seems to be on the runner, since I know I'm running 3.12 on my system and have no issues.
I've tried checking versions with pip freeze, seeing if updating the system packages helped, and all nothing. My vote is that it's somehow using the system virtualenv which would be at 3.10 for the runner, but that's wild speciulation a guess.
short tendon
No, I think it's good now. If you're interested in figuring out how its using a...
What is it? I've missed a message somewhere
proven garnet
tulip sleet
I've tried checking versions with `pip freeze`, seeing if updating the system pa...
My vote is that it's somehow using the system virtualenv which would be at 3.10 for the runner, but that's a guess.
I was wondering if it was something like that too. I had hoped thatactions/setup-pythonwould somehow do the right thing and set up compatible versions of other things, but I guess not.
short tendon
I'm seeing this:
uses: actions/setup-python@v5
with:
python-version: "3.x"
Which should specify the python version if there's one you want...
short tendon
@tulip sleet and @proven garnet resolved. PR proof: https://github.com/adafruit/Adafruit_Blinka_bleio/pull/67 it's specifically a py3.12 issue. Doing some digging it's possibly an outdated version of setuptools
Feel free to update that other PR and close this one, or merge it in first and fix the other
proven garnet
<@329766224093249548> and <@135415127879974912> resolved. PR proof: https://gith...
The goal is to hopefully use 3.12 though π. It also doesn't look like setuptools is outdated if you do pip freeze on the runner.
short tendon
The goal is to hopefully use 3.12 though π. It also doesn't look like setuptoo...
So the actual issue is that pkgutil.ImpImporter that has been Deprecated since version 3.3 was removed in 3.12. I imagine it may take time for everything to catch up. Is 3.12 needed now? Or could we pin to 3.11?
proven garnet
<@329766224093249548> and <@135415127879974912> resolved. PR proof: https://gith...
Setuptools handles it correctly in 3.12 by putting that line inside of an if hasatrr() block, but that's not what's being used here. Here's setuptools 69.2.0 (which is what the runner claims to use in the system, but clearly not the pre-commit venv: https://github.com/pypa/setuptools/blob/6ee23bf0579c52e1cbe7c97fc20fd085ff2a25c7/pkg_resources/__init__.py#L2198-L2199)
It's currently pinned to 3.11 across all the repositories' CIs, including the libraries
tulip sleet
Right, I want to stop having to pin 3.11
Itβs mysterious to me why itβs using the wrong setuptools, but maybe itβs the default one for Ubuntu 22.04. Maybe we have to pip or pipe install the right one, but Iβm not sure where to put that requirement
short tendon
@tulip sleet and @proven garnet figured it out for real
it's the version of pylint we use: https://github.com/pylint-dev/pylint/blob/fc34a4b6abe56f3ac07ca15d846b1c1955545f85/pyproject.toml#L2 it needs setuptools~=62.6. It's removed in 3.0
Or you know we could switch to ruff.... π
π°π³οΈ success!
proven garnet
π°π³οΈ success!
Nice! We can run adabot to update the pre-commit hooks.
short tendon
Nice! We can run adabot to update the pre-commit hooks.
The only issue is that I needed to add a few # pylint: disable=too-many-arguments so updating may cause linting to fail in some projects. Also, I picked 3.0 in the PR, just because it was the first one that worked. There is a 3.1 (it looked like it had the same rule changes)
proven garnet
The only issue is that I needed to add a few `# pylint: disable=too-many-argumen...
I don't think that's an issue in the long run - most libraries should pass that check I think (I think the limit is 5 arguments which is plenty in most cases
I have a tool that can help locate and list any result CI failures
proven garnet
But yeah, this also competes with the ruff change so it may be worth figuring that out
short tendon
Yeah, does seem not the best use of time to fix linting to fix it again...
proven garnet
Haha yeah, but it makes sense now that it's within the venv. I vaguely remember an issue like this when we switched to 3.11 though I don't think it was setuptools but some other package that failed since something was moved or removed.
short tendon
I think it's funny how long that method had been deprecated. 3.3 was a long time ago...
@proven garnet did you see my thoughts above on how to figure out if the switch to ruff is good?
tulip sleet
@proven garnet @short tendon I was confused becaue I saw pylint 3.0.3 referenced here: https://github.com/adafruit/actions-ci-circuitpython-libs/blob/e9f0c3911c4e0cc6bf64f77355cca0ac143811b6/.pre-commit-config.yaml#L21 . I though this version was being used for lib CI actions, but now I see the separate specification in the .pre-commit-config.yaml for the blinka bleio library, etc.
short tendon
I can't even see how that library uses pre-commit...
I did look at that library yesterday as I was trying to see what was installed where and if something was installing a different setuptools
tulip sleet
I am pretty confused by that repo. I wasn't aware it existed
short tendon
I think it's mainly used here: https://github.com/adafruit/workflows-circuitpython-libs/blob/main/build/action.yml
other then the install.sh which is used to install what the CP library needs
You all have build an impressive build system, but it's a little like a rollercoaster...
proven garnet
<@135415127879974912> <@1152454697614069811> I was confused becaue I saw pylint ...
Ah, sorry, I should have said that. Yeah, currently all the Blinka stuff uses local versions of CI since dependencies might be different, but it may be worth unifying those.
proven garnet
I think it's mainly used here: https://github.com/adafruit/workflows-circuitpyth...
Yeah, it's wild. This is exclusively for libraries, which all pretty much have the same requirements for builds. The Blinka stuff all use their own local version of pre-commit config, pylintrc, etc.
proven garnet
<@135415127879974912> did you see my thoughts above on how to figure out if the ...
Yeah, I would say if we decide to switch to ruff then it's not worth linting to then fix to then lint. I haven't been to the meetings in a little bit, so not sure what the consensus is.
Tagging @lone axle who has been involved on that front.
This does remind me that this issue is why I wanted (and honestly should) create a repository for how these issues get solved - community libraries will break if we change the unified library action and downstream users (anyone who uses the cookiecutter) don't update their pre-commit config files.
short tendon
I would add to that doc, why the pre-commit linters,formaters,checkers,etc where chosen. Seems like a repo would be a good place to keep that history...
lone axle
This would be my flow:
1. Pick 5-10 repos of different sizes (like ticks, minimq...
Are these things you mention in like 'sorting', 'pyupgrade', 'pylint' like different modules within ruff? Or were the steps outlined here not talking about the switch to ruff?
And did you mean 3 or 4 PRs for like each of those letters a to c or d put into it's own individual PR?
short tendon
Are these things you mention in like 'sorting', 'pyupgrade', 'pylint' like diff...
Yes, in the pyproject.toml we would add:
[tool.ruff.lint]
select = ["I", "PL", "UP"]
Where each letter is the option we want to run.
And yes, 3-4 each. It would be so we can show what each one would change. To help see if we want to add them. I can script it all out, so it's not a lot of work
lone axle
Do you mean just for the first few to see what shakes out or for all of them even after the first few? I think I'd lean toward trying to collapse down to 1 PR per library if possible at least for when we do all of them. Even if the creation of those PRs can be scripted there are still manual steps for approve and then merge (unless there are more automation tools in adabot or somewhere that I don't know of). So at least for my process reviewing and merging 1 PR instead of 3 or 4 would result in substantially fewer clicks and keypresses through the process of all the libraries in the bundle.
Even in a single PR those discrete steps could be broken out for inspection by commiting them individually with the changes associated to each one. You can always view the branch at each of those commits within the same PR.
short tendon
oh yes. just for a few random ones to help make decisions
lone axle
What does pyupgrade do? I don't think I've heard of that
short tendon
And when doing the full thing, I would automate adding one at a time, running and committing
pyupgrade finds things that have changed and updates them. For example f-strings
lone axle
Is there one of these things select = ["I", "PL", "UP"] for formatting like the functionality that replaces Black? or is that set up some other way?
short tendon
no... it's strange. this:
[tool.ruff]
target-version = "py38"
[tool.ruff.lint]
select = ["I", "PL", "UP"]
tool.ruff == formatting
tool.ruff.lint.I == isort
tool.ruff.lint.PL == pylint
tool.ruff.lint.UP == pyupgrade
lone axle
Ahh, the pyupgrade one I have slight reservations on because micropython and thus circuitpython doesn't necessarily always support the brand newest of CPython syntax when it becomes available. Last I knew micropython was targeting 3.4 or 3.5 python syntax though with some newer features like f strings included at this point. I'm not sure if that has been moved forward perhaps though. We can see with testing on hardware when those inital PRs are in, but I worry that some day it's going to make an upgrade to some new syntax that isn't supported by circuitpython and then it will break the library on hardware but the linter / actions checks won't realize that.
short tendon
Totally don't need to add it. Just showing options
fringe yarrow
In one last ditch effort to determine why the lilygo tdisplay s3 works with the parallel display bus code while this cheap yellow display board does not, I determined that the only difference I could perceive in hardware, other than the resolution, was that the Lilygo device had a hardware reset pin while the CYD had tft_reset wired to the esp32 en pin. I went so far as to cut the trace to tft_rst and run it to a spare GPIO on the ESP32, tested continuity thoroughly to ensure the surgery was a success, change the code accordingly.. and it has made zero difference.. still lines and garbage.. I give up on this thing..
lone axle
I would say this plan sounds good to me. I do think 1 PR with separate commits would be fine instead of separate PRs, but for the smaller subset of libraries to check out I don't think it matters as much either way. I would add that part of the testing process on those first couple of libraries should include a trial on hardware both mpy'd and plaintext python. I recall some weird obscure issue in the past where mpy-crossed version of some syntax was causing trouble so I think it's worth doing that kind of testing especially on the pyupgrade changes.
And we can update the cookie-cutter repo with the specific configuration we land on once it's decided.
slender iron
@onyx hinge you know of a way to compute the length of an array in a type right? I feel like you added it somewhere. ESP removed a constant for how many endpoints chips have and hardcoded the array length
tulip sleet
<@426200547024961539> you know of a way to compute the length of an array in a t...
there is an MP_something for that
slender iron
this is for tinyusb
tulip sleet
#define MP_ARRAY_SIZE(a) (sizeof(a) / sizeof((a)[0]))
do you mean they literally made it 5 or whatever?? oy
short tendon
I would say this plan sounds good to me. I do think 1 PR with separate commits w...
I had only thought separate, since two could make the same/similar changes
tulip sleet
@slender iron I think I would open a complaining issue about that :forehead_slap:
slender iron
I can't figure out how they changed tinyusb to still work
onyx hinge
yeah MP_ARRAY_SIZE is the ticket @slender iron as long as it's really an array and not a pointer
slender iron
π
onyx hinge
unfortunately that construct results in 0 or 1 for a pointer, rather than an error
tulip sleet
I can't figure out how they changed tinyusb to still work
all the dcd stuff is gone
I'm mystified too
going afk
tulip sleet
slender iron
but it uses the old macro still
slender iron
π
proven garnet
Do you mean just for the first few to see what shakes out or for all of them eve...
The adabot patch is a good candidate for pushing to main some of these changes. Things like pyproject.toml, .pre-commit-config.yaml, and other such files are easy that way.
I then have a tool that will check and record which recent CI builds failed as a result
(also contained within adabot)
The nice thing about the adabot route is that it saves having to submit PRs, and just works by doing a git format-patch application to all the repos
I do have another tool that works similarly in that it pushes to main without a PR but is a little bit more programmatic (it basically takes a repo and can work with it in Python for things like updating TOML files or similar things that might be harder for a git operation (i.e., the files are very different across repos)
I don't think that will be necessary here but it is an option if a patch fails heavily. We can dry-run and see how many would need to be fixed otherwise.
lone axle
The adabot patch is a good candidate for pushing to main some of these changes. ...
Right, my understanding is that the plan is to do PRs only for the first few to get a closer look at the changes it's going to make. Then do the rest with a patch from adabot. I'll have to look into adabots CI build failure thing, I didn't know she had that ability.
proven garnet
I added it a couple years ago to help with looking for them for cases like this 
It utilizes the previous tool that iterates through the libraries via Python, I thought it was standalone but it isn't.
Here was an example use for converting setup.py to pyproject.toml: https://github.com/tekktrik/adafruit-bundle-toml-switcher
Here's a simpler use case for converting all the YAML files to use the unified composite actions, which I need to review so I remember how it works too π : https://github.com/tekktrik/CircuitPy-Composite-Action-Paster/tree/main
short tendon
@proven garnet can the patch also run commands? since it would also need to run pre-commit on the repo after the changes?
onyx hinge
fwiw I've just become aware that black has a CVE and recommends everyone upgrade to current. π’ this is via a non-working PR to adabot: https://github.com/adafruit/adabot/pull/370 https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2024-21503
proven garnet
<@135415127879974912> can the patch also run commands? since it would also need ...
The patch can't but the other tool I mentioned could
We could also do git submodule foreach on the bundle I think?
lone sandalBOT
tulip sleet
fwiw I've just become aware that black has a CVE and recommends everyone upgrad...
I guess the worst thing that would happen is malicious PR's that take a long time to run.
midnight ember
fwiw I've just become aware that black has a CVE and recommends everyone upgrad...
"or if you habitually put thousands of leading tab characters in your docstrings. " I feel like someone found that one accidentally. my recent experience with LF & CRLF spamming me with errors was quite rude looking... though that was pylint.
short tendon
So @lone axle, @proven garnet and @tulip sleet would me picking 5-10 repos, creating a branch, adding ruff: format, sort, lint, pyupgrade each as their own commit be a helpful step in making decisions?
lone axle
So <@382939733107408897>, <@135415127879974912> and <@329766224093249548> would ...
I think so π
short tendon
Any you would like in that list?
lone axle
pyportal, portalbase, pycamera, neopixel, requests, and any random I2C based driver make sense as a good test set in my mind.
lone axle
sounds good to me, though I've never done any testing with it personally. I think I purchased the hardware for it at one point but I'll have to dig around in the parts room a bit to find it.
proven garnet
pyportal, portalbase, pycamera, neopixel, requests, and any random I2C based dri...
I used the bme680 library typically for what it's worth but agree it's essentially arbitrary!
proven garnet
fwiw I've just become aware that black has a CVE and recommends everyone upgrad...
I can try to incorporate this into a patch
fringe yarrow
Because I need a new project to beat my head against.. I've ordered a Lilygo T5-S3 v2.3. It's an ESP32S3 with TF slot, I2C RTC, and 4.7" ed047tc1 EPD. Doesn't look like there's TOO much for support ouside of experimental drivers out there for that panel. We'll see how that goes..
manic glacierBOT
short tendon
I used the bme680 library typically for what it's worth but agree it's essential...
Can do this one. I looked:
You last purchased this item on September 25, 2020
So mines even stemma
midnight ember
CircuitPython.org has a new board rss feed in the top right corner but it hasn't been updated since May 2023. https://circuitpython.org/feed.rss
midnight ember
I'm trying to learn typing and the learn guide is nice but is there a list of all the types? The circuitpython_typing library example wasn't much help https://github.com/adafruit/Adafruit_CircuitPython_Typing/blob/main/examples/typing_simpletest.py
tulip sleet
CircuitPython.org has a new board rss feed in the top right corner but it hasn't...
The feed says <lastBuildDate>Thu, 21 Mar 2024 06:03:27 -0400</lastBuildDate>. I didn't hear that anyone fixed it -- not sure what changed since you looked.
lone axle
are RSS feeds meant to be chronological? Should the newest be always at the top or bottom? The order of boards in that feed seems all over the place.
maybe the client software is responsible for ordering them by pubDate
tulip sleet
are RSS feeds meant to be chronological? Should the newest be always at the top ...
the feed itself is in alphabetical order by board name. Each board has a PubDate. If the feed reader is sorting by that it will be confusing
lone axle
I didn't know how the RSS page was built so I went on a little exploration to figure out. It looks like ultimately the order would be determined by this loop in the feed template: https://github.com/adafruit/circuitpython-org/blob/ec8c7044857e010c500eac3c9fd280fb25f4aeab/feed.html#L11
Ahhh, I hadn't caught on it was alphabetical, that makes sense
tulip sleet
the sort order on the boards web page is by popularity (download count)
which makes for some surprises
lone axle
I added the feed to an rss reader on my phone and it does seem to handle the chronological sorting automatically
it would be cool to make this data available as JSON so it could be consumed by circuitpython. Then you could have a magtag, or matrix portal etc.. showing the latest new devices supported all the time.
proven garnet
I'm trying to learn typing and the learn guide is nice but is there a list of al...
I don't think there's an extensive list like that. It may help to add to that repository. If you want to make an issue to that repo I can assign myself and try to do that since I know I PR'd many of them.
I can also try to make a small example guide in the Playground page if you think that would be helpful.
midnight ember
I can also try to make a small example guide in the Playground page if you think...
Absolutely! That would help. I'll make an issue for the typing repo too.
Didn't notice RSS was alphabetical. Yes they're usually in chronological order since... like always. This is the first feed I've seen that isn't chronological. How you even managed to do that is neat. I thought it was a latest boards feed in chronological order. The first board in the list is from May 2023 so if you're expecting RSS in chronological order it will appear as if it hasn't been updated since May 2023.
Having 2 feeds, 1 in alphabetical order and 1 in chronological order would be appreciated. I love RSS feeds but there's no sense in subscribing to a feed that's in alphabetical order as my feed reader will only pull the top 25 and never update. RSS feeds are expected to be in chronological order. As FoamyGuy said last publish date is how RSS feeds notify you of updates.
That's how Adafruit's Blog RSS works.
midnight ember
I don't think there's an extensive list like that. It may help to add to that r...
Submitted as requested. https://github.com/adafruit/Adafruit_CircuitPython_Typing/issues/41
proven garnet
Thank you!
tulip sleet
Didn't notice RSS was alphabetical. Yes they're usually in chronological order ...
I think it isn't really a "feed" so much as just a regenerated summary. If it would help to sort it by publication date, you could open an issue for that
midnight ember
Where should I open it? Circuit Python main?
@proven garnet Tested and confirmed as requested https://github.com/adafruit/Adafruit_CircuitPython_datetime/issues/10
lone axle
Where should I open it? Circuit Python main?
Circuitpython-Org repo I think is the best place, that is where the change will need to be made to enable it.
midnight ember
Didn't even know that was a repo. Will do.
midnight ember
bookmarked, thank you.
midnight ember
Helped with closing an issue that was 967 days old. π€
onyx hinge
yeah but back when the issue was opened you probably couldn't have calculated that it would be open for 967 days
lone axle
Would it be worth adding a section at the top of the main displayio module docs section: https://docs.circuitpython.org/en/latest/shared-bindings/displayio/index.html#module-displayio that outlines the breaking changes from 8.x? That way anyone whose updated and found their project stopped working will have a chance to find what specific change is needed if the land on this base displayio docs page.
manic glacierBOT
there is an edit and it seems to have an option for changing branches but it doesn't show any branches in the list to choose from and typing 9.0.x in there and then enter doesn't seem to do anything.
tulip sleet
@lone axle do you see this?
lone axle
<@382939733107408897> do you see this?
I see the dropdown the but there are no actual choices in it only an empty search box
oh
tulip sleet
are you looking at your fork or adafruit/circuitpython?
lone axle
now it's populated
tulip sleet
!!
lone axle
okay, that saved but made way too many unrelated changes in the PR now. I'll start over on a new branch from 9.0.x
proven garnet
@lone axle I should have gone through all the circup instruction fixes
lone axle
<@382939733107408897> I should have gone through all the circup instruction fixe...
Thank you!
midnight ember
glad to know the RSS issue wasn't just my perspective. thank you Melissa.
midnight ember
How would you name the 2 feeds? Blinka RSS & Microcontroller RSS?
I like the idea of separate feeds. I would be far more interested in subscribing to a latest circuit python microcontroller latest boards feed since I don't use SBC's or micropython.
orchid basinBOT
digital shoreBOT
adafruit
orchid basinBOT
tulip sleet
@slender iron i was going to start looking at the deepsleep bug (9067) unless you have a clear idea of what this is.
slender iron
not really though I didn't read it closely
I am tempted to get sleep going on the Cx series at the same time
tulip sleet
it's waking up immediately on a pin alarm
i have one C3 board I could see about. But I think I would fix that first standalone
slender iron
ya, and the apis may have changed with 5.2
tidal kiln
is this related to something known?
Adafruit CircuitPython 9.0.0 on 2024-03-19; Adafruit-Qualia-S3-RGB666 with ESP32S3
>>> import cm_test
----------------------------------------
Fetching from http://wifitest.adafruit.com/testwifi/index.html
Traceback (most recent call last):
File "adafruit_connection_manager.py", line 271, in get_socket
OSError: [Errno 113] ECONNABORTED
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "cm_test.py", line 32, in <module>
File "adafruit_requests.py", line 597, in get
File "adafruit_requests.py", line 535, in request
File "adafruit_connection_manager.py", line 282, in get_socket
RuntimeError: Error connecting socket: [Errno 113] ECONNABORTED
>>>
where cm_test is this example from the library:
https://github.com/adafruit/Adafruit_CircuitPython_ConnectionManager/blob/main/examples/connectionmanager_helpers.py
tulip sleet
is this related to something known?
```py
Adafruit CircuitPython 9.0.0 on 2024-0...
nope, I haven't seen this before. Are all the libraries completely up to date?
tidal kiln
hmm. maybe. i just loaded the libs. but took them from a bundle zip from a guide example.
let me try a manual loading of the libs directly from the main bundle zip
tulip sleet
it could just be a bug in the cm test
tidal kiln
i originally hit it through the qualia support lib
but then recreated more directly as above
qualia.get_local_time() end up with same error
tulip sleet
is your settings.toml in good shape?
I'd try this with some other moregeneric ESP32-S3 board
tidal kiln
@danh library updates seems to have fixed it
Adafruit CircuitPython 9.0.0 on 2024-03-19; Adafruit-Qualia-S3-RGB666 with ESP32S3
>>> import cm_test
----------------------------------------
Fetching from http://wifitest.adafruit.com/testwifi/index.html
Text Response This is a test of Adafruit WiFi!
If you can read this, its working :)
----------------------------------------
>>>
could you download the bundle again and see if there's a library difference?
tidal kiln
still got the zip...looking...
got it from here fwiw:
https://learn.adafruit.com/qualia-s3-compass/code-the-compass
how do you diff .mpy's ?
$ diff lib1/adafruit_connection_manager.mpy lib2/adafruit_connection_manager.mpy
$ diff lib1/adafruit_requests.mpy lib2/adafruit_requests.mpy
no diff
tulip sleet
how do you diff .mpy's ?
do like strings foo.mpy and the version number should be one of the first strings
tidal kiln
still same. let me see if i can recreate the issue.
still same = versions agree using strings
1.0.1 and 3.0.1
@tulip sleet nevermind. can't reproduce. something must have just been weird.
tulip sleet
π€· that is odd, if it happens again ...
tidal kiln
@tulip sleet hmm...actually, still seeing some weirdness. qualia.get_local_time() not working again. and different error this time.
Adafruit CircuitPython 9.0.0 on 2024-03-19; Adafruit-Qualia-S3-RGB666 with ESP32S3
>>> from adafruit_qualia import Qualia
>>> from adafruit_qualia.graphics import Displays
>>> qualia = Qualia(Displays.ROUND28)
>>> qualia.get_local_time()
Connecting to AP peanuthamper
Getting time from IP address
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "adafruit_portalbase/__init__.py", line 492, in get_local_time
File "adafruit_portalbase/network.py", line 270, in get_local_time
File "adafruit_portalbase/network.py", line 237, in get_strftime
File "adafruit_requests.py", line 597, in get
File "adafruit_requests.py", line 535, in request
File "adafruit_connection_manager.py", line 236, in get_socket
gaierror: (-2, 'Name or service not known')
>>>
and after a soft reset:
Adafruit CircuitPython 9.0.0 on 2024-03-19; Adafruit-Qualia-S3-RGB666 with ESP32S3
>>> import cm_test
----------------------------------------
Fetching from http://wifitest.adafruit.com/testwifi/index.html
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "cm_test.py", line 32, in <module>
File "adafruit_requests.py", line 597, in get
File "adafruit_requests.py", line 535, in request
File "adafruit_connection_manager.py", line 236, in get_socket
gaierror: (-2, 'Name or service not known')
>>>
that looks like this code:
addr_info = self._socket_pool.getaddrinfo(
host, port, 0, self._socket_pool.SOCK_STREAM
)[0]
somethings getting into a weird state
after a power cycle:
Adafruit CircuitPython 9.0.0 on 2024-03-19; Adafruit-Qualia-S3-RGB666 with ESP32S3
>>> import cm_test
----------------------------------------
Fetching from http://wifitest.adafruit.com/testwifi/index.html
Text Response This is a test of Adafruit WiFi!
If you can read this, its working :)
----------------------------------------
>>>
i think it is just doing a DNS lookup
try some other board, and try older versions of 9.0.0. But I am wondering a little bit about your local network. Could try phone hotspot as an alternative
tidal kiln
any specific 9x?
tulip sleet
beta 0, alpha 4, but that's just a rough guess.
tidal kiln
Adafruit CircuitPython 9.0.0-beta.0 on 2024-01-27; Adafruit-Qualia-S3-RGB666 with ESP32S3
>>> from adafruit_qualia import Qualia
>>> from adafruit_qualia.graphics import Displays
>>> qualia = Qualia(Displays.ROUND28)
>>> qualia.get_local_time()
Connecting to AP peanuthamper
Getting time from IP address
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "adafruit_portalbase/__init__.py", line 492, in get_local_time
File "adafruit_portalbase/network.py", line 270, in get_local_time
File "adafruit_portalbase/network.py", line 237, in get_strftime
File "adafruit_requests.py", line 597, in get
File "adafruit_requests.py", line 535, in request
File "adafruit_connection_manager.py", line 236, in get_socket
gaierror: (-2, 'Name or service not known')
>>> import cm_test
----------------------------------------
Fetching from http://wifitest.adafruit.com/testwifi/index.html
Text Response This is a test of Adafruit WiFi!
If you can read this, its working :)
----------------------------------------
>>>
@gilded cradle hey, any ideas on above?
gilded cradle
Let me catch up...
tulip sleet
@tidal kiln I'll be afk for maybe an hour
gilded cradle
@tidal kiln I'm not familiar with connection manager. Name or service not known looks important, but it doesn't really give much info.
Also not familiar with gaierror
tidal kiln
am i being dumb in how im using the qualia lib?
gilded cradle
I'm going to look at the connection manager code and see if there's anything that stands out
@tidal kiln in adafruit_requests, line 535 is a comment on GitHub. Are you sure it's up to date?
slender iron
@tulip sleet I'm going to have to back out my i2c changes because the esp32-camera is using the old api still
tidal kiln
@gilded cradle hmm....well its from latest bundle. but latest bundle is:
gilded cradle
@tidal kiln yeah, there was a release of requests 3 days ago. Maybe bundles aren't building for some reason?
tidal kiln
yah. looks like. seriously doubt there have been no library updates since 3/7
gilded cradle
@tidal kiln maybe download some of the files manually from repos just to test your code and then we can fix the bundle releaser (not sure where it is or how it's done).
slender iron
I wonder why its broken
gilded cradle
Not sure, but I don't recall seeing a release in quite q while
I think this is the actions that releases the bundles, but not completely sure: https://github.com/adafruit/adabot/blob/main/.github/workflows/bundle_cron.yml
slender iron
lone axle
I wonder if there is a way to disable the idle based disabling.
gilded cradle
Ah ha, cool
lone axle
it's a bit odd that we have to make changes to Adabot every 60 days in order to keep the bundle building
gilded cradle
I think the bundle is on a cron, so will likely only run starting later tonight
slender iron
I don't see one in the settings
lone axle
maybe if the action were moved from Adabot over to the Bundle repo it wouldn't be able to hit 60 days because any time it releases I assume it counts as "activity" so it will be resetting it's own timeout anytime any library changes.
it could still clone and use adabot code from inside of the action I think even if it were moved to the Bundle repo.
tidal kiln
.github/workflows/cron_keepalive.yml?
gilded cradle
That makes sense to me @lone axle
proven garnet
Catching up but I think the cron settings live in circuitpython-org for certain things, that runs adabot
slender iron
proven garnet
I think the bundle is one of them
lone axle
Another option might be to commit the daily report into Adabot when it gets run from this action: https://github.com/adafruit/adabot/blob/main/.github/workflows/reports_cron.yml in addition to uploading to S3
if that committed the current file when it runs (replacing the old one if we don't want 2 complete archives) it would mean there is activity every day.
Ah interesting, that one has a direct API to re-enable it.
proven garnet
Is the issue that adabot needs a daily update?
gilded cradle
Maybe not daily, but probably at least semi-regularly.
lone axle
Is the issue that adabot needs a daily update?
at least once every 60 days it seems.
proven garnet
Why is that? I don't see anything in the workflow regarding that. I assume I'm missing something.
tidal kiln
it was in screen shot above
gilded cradle
Could we tie it to the bundle repo itself?
That's what @lone axle suggested.
proven garnet
Gotchya, sorry about that, thanks!
gilded cradle
Gotchya, sorry about that, thanks!
No worries
tidal kiln
@gilded cradle i tried to manually grab latest libs, but still getting same weirdness. but could still be related to something in the libs. will wait till bundle release is fixed and try again.
gilded cradle
ok, I suspect it will work tonight.
tidal kiln
yep. it'll be next week until i can try this again. but its just initial steps for something non-urgent. so can wait.
thanks all for looking into this!
short tendon
Do you all want to stick with the current line length? I personally like 120, but have no strong preference...
tulip sleet
Do you all want to stick with the current line length? I personally like 120, bu...
I think it would cause a lot less churn to stick with the current length, since otherwise parameter lists, etc, might be rewrapped onto one line.
midnight ember
I've run into many gaierrors in the past with requests projects that are online 24/7. I think it's wifi or dns lookup related. Used to get them all the time with 8.2.5, updated my main project to 8.2.7 and haven't had any issues. I think a lot of it also might have to do with the way the socket is continually retried inside a while loop.
Gai Error might be in part due to syntax <#help-with-circuitpython message>
response = requests.get(DATA_SOURCE).json()
response.close() # this has issues closing sockets and can lead to gaierror if using an explicit response.close()
vs
response = requests.get(DATA_SOURCE)
json_response = response.json()
response.close() # this is a little longer but plays much nicer with socket close
``` I haven't encountered a gaierror since about 8.2.5. It has more to do with syntax and placement of additional requests within a while true loop. it can allow it to iterate too many times if a failure occurs. instead of gracefully failing a gaierror is a crash. i think what i did to get around it just a general exception handler that will break or trigger microcontroller.reset. i've literally never had to touch it since. it's probably been up and running flawlessly for more than a month now.snippet to try ```py
while wifi.radio.ipv4_address:
try:
print("| | Attempting to GET Weather!")
if debug_OWM:
print("Full API GET URL: ", DATA_SOURCE)
print("\n===============================")
with requests.get(DATA_SOURCE) as owm_request:
# uncomment the 2 lines below to see full json response
# warning: returns ALL JSON data, could crash your board
# dump_object = json.dumps(owm_request)
# print("JSON Dump: ", dump_object)
try:
owm_response = owm_request.json()
if owm_response["message"]:
print(f"| | β OpenWeatherMap Error: {owm_response['message']}")
owm_request.close()
except (KeyError) as e:
splash_label.text = "Getting Weather..."
owm_response = owm_request.json()
print("| | Account within Request Limit", e)
print("| | β
Connected to OpenWeatherMap")
i will be updating it to 9.0 soon
midnight ember
Figured out what is responsible for changing all the LF's to CRLF's. It's Github Desktop. Here's a file in Notepad++ with line endings specifically setup to do LF and save as Unix. As soon as I import into Git it changes them and notifies me about it with a yellow warning.
empty salmon
Is it possible to set the "board description string" during init vs at compile time?
Alternately, is there an informational string which can be set during init?
stuck elbow
anything you print in boot.py goes to boot_out.txt
not sure where exactly you want to have access to that string?
empty salmon
Let me try to clarify: I am talking from the firmware for the board.
Is there any text which can be set in the board_init() which is accessible to a user from their code.py?
onyx hinge
Let me try to clarify: I am talking from the firmware for the board.
Is there an...
I don't think there is.
stuck elbow
I mean you have access to everything in memory, you could just set some string in any module
onyx hinge
most everything "in a module" is designed to be read-only, so that it goes in flash. Module dictionaries are read only, and of course strings are all immutable. So, it'll be more than just putting something in board init, unless there's some facility that already exists and I'm just not thinking of it.
onyx hinge
you could put a function in board module that returns the value in question, that might be easiest
board.submodel() or whatever your goal is
or you might be able to make the whole board module MP_DEFINE_MUTABLE_DICT but that also allows unlimited user modification (and I'm not sure that really works across soft resets)
empty salmon
I was not aware of board.submodule(). I will look for an example.
The issue is not a showstopper but it could be useful to the end user.
onyx hinge
oh, that is something I just made up
you can put about whatever you want in the board module, you could make your own function
onyx hinge
I'm back on an old idea: introducing a deinitialized TYPE. I believe that overall this can reduce code, but it entails a bunch of changes. This is only compile-tested on rp2040 macropad but it gives the idea: https://github.com/adafruit/circuitpython/compare/main...jepler:circuitpython:deinited-type2?expand=1
many "check for deinit" calls can be removed, because for instance once an I2C object has had its type changed to Deinitialized, you can't call the C function busio_i2c_scan from Python code with that object anymore.
This change decreases flash usage by 356 bytes(!) and for some reason even frees up 16 bytes of RAM. That's enough that I feel like I must have messed up my measurements. The bulk savings may come from properly marking raise_deinited_error() as NORETURN, not sure.
@slender iron @tulip sleet while I shouldn't go further down this tangent road right now your thoughts would be valuable.
onyx hinge
addendum: 392 bytes saved on rp2040 macropad just by marking NORETURN but 0 bytes on trinket_m0. I think this is because with LTO, the compiler is able to effectively determine that raise_deinited_error() does not return. rp2040 doesn't use lto.
onyx hinge
@gilded cradle do you know why 12 hours was being added to all the times? <pubDate>{{ board.date_added | date: '%s' | plus: twelve_hours | date_to_rfc822 }}</pubDate>
gilded cradle
@onyx hinge probably to mitigate an issue with tiemzones or something
empty salmon
@onyx hinge - is there documentation or an example of a board with a custom function (or property) exposed to the end user?
onyx hinge
@empty salmon not that I see with a full text search
tulip sleet
<@252717193496756235> <@329766224093249548> while I shouldn't go further down th...
This is an interesting idea. One question I have is how it would affect some singletons. Admittedly this could be pathological:
>>> import board
>>> i = board.I2C()
>>> i.deinited()
>>> i.unlock() # a harmless operation: works
>>> i.deinit()
>>> i.unlock()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ValueError: Object has been deinitialized and can no longer be used. Create a new object.
>>> i = board.I2C()
>>> i.unlock() # works again
>>>
I can imagine a use case where you deinit `board.I2C() to do something DigitalInOut'ish on its pins, like force an I2C bus reset. Then you create it again. Would your change prevent that?
@empty salmon what name it that you're computing in board.c that's going to vary? Can you explain the use case?
onyx hinge
@tulip sleet my proposal would make i.unlock() throw instead of being "harmless".
er the unlock-after-deinit already throws so that's fine, it would still throw
fringe yarrow
@slender iron In regards to that Parallel Display all in one board that's been keeping me up at night, at your suggestion, I've hooked D0-D7, WR,RD,DC,CS up to a logic analyzer and the results seem strange to me. I can clearly see the init sequence I have specified in the board.c, however, with commands sent without data parameters (ie 0x21, 0, i can see the 0x21 signal which lasts 10 milliseconds, however commands with data parameters only last for 7microseconds, with data parameters lasting 300 nanoseconds..
empty salmon
@tulip sleet - The use case is mostly theoretical at this point.
From my prior query, I have a board that will have a minor variance. Because of display availability, the board will ship with one of 2 displays. They are functionally so similar I would rather not end up with 2 firmware. Within board_init() I am able to detect which display it has and use the appropriate init code.
I was thinking it might be nice for the end user to be able to detect the variance from within their code.py.
It's not critical by any means.
tulip sleet
<@329766224093249548> - The use case is mostly theoretical at this point.
From...
here is an example of a function in board (defined in pins.c) that you would call and it would do something. It could query or use a value set in board.c.
empty salmon
Thanks @tulip sleet !
tulip sleet
i wonder in the long run in we might just as well merge board.c and pins.c, since pins.c implements board π
slender iron
<@252717193496756235> <@329766224093249548> while I shouldn't go further down th...
I'm definitely thinking about this area of the code in the back of my mind. I want to remove all bulk resets which puts more reliance on finalizers and deinit. A related piece is preserving some objects outside the heap. This gets trickier with display busses that use I2C and SPI. I don't really want to allocate them off the heap just in case.
slender iron
<@329766224093249548> - The use case is mostly theoretical at this point.
From...
I wouldn't tell the user. Just make board_init() setup board.DISPLAY and don't worry about it
@onyx hinge it crossed my mind to try and "move" management of memory from the VM to TLSF at the end of the VM's execution.
the memory would stay in the same place, it'd just change who manages it
empty salmon
@slender iron - that is my plan.
TBH: the value of having a board function is really for support in the event of an issue being reported.
slender iron
how are you detecting the difference? could you tell from a photo? that's one of the first things we usually ask for
onyx hinge
<@426200547024961539> it crossed my mind to try and "move" management of memory ...
I had a similar idea back in the day, but sort of the reverse: I was going to turn supervisor allocations into GC allocations whenever the GC was available.
tulip sleet
<@252717193496756235> - that is my plan.
TBH: the value of having a board funct...
is the board embedded in a case? Otherwise I'd suggest you mark the actual physical board to distinguish them
so you don't have to depend on software to distinguish them
from a support point of view (suppose one display turns out to fail more often) I think that might be better
empty salmon
I will add something to the silkscreen on the board to denote difference.
The other value to "being able to detect in code" is I typically develop a library of example code and if there was a reason to "do something different" then I could detect in code.
Again, "nice to have" vs "important to have".
tulip sleet
@slender iron ESP-IDF v5.1.3 broke pin wakeups that go high to wake up. I keep editing my post in the issue to add details. I have tracked down some IDF changes about this. I am going to see if this is a regression that was fixed later.
slender iron
<@252717193496756235> ESP-IDF v5.1.3 broke pin wakeups that go high to wake up. ...
you could test with my 5.2 branch
tulip sleet
if it's working sure
slender iron
ya, I had to roll back the i2c changes but it boots and runs
the camera code stops working once I insert an SD card so I need to debug that
tulip sleet
i cloned your repo but the main branch seems to be way out of date. I'm confused. I know you forked from micropython, not circuitpython
slender iron
I never update main
tulip sleet
ha!
slender iron
I only have feature branches
tulip sleet
and you never delete branches either π
slender iron
git switch -c feature_branch adafruit/main
I never delete anything if I can help it π
I just buy bigger hard drives
bundle was released overnight!
tulip sleet
how did it get kicked to run again?
slender iron
I clicked the "Enable workflow" button
tulip sleet
so it gets unenabled every 60 days if something-something doesn't happen? I missed part of that discussion
slender iron
so it gets unenabled every 60 days if something-something doesn't happen? I mis...
ya, if the repo has no commit activity
jim said 1.23 is a couple weeks out and has no mpy version changes
1.24 may not have mpy changes either
tulip sleet
sounds good; I could do the merge, and can also work on upstream PR's
slender iron
I think we did want to block the bulk reset removal on it
proven garnet
@tulip sleet I released the GPS library, thanks for the quick fix.
I assume that happened as a result of the release I did yesterday
tulip sleet
I decided it was too easy not to do immediately
tulip sleet
I think we did want to block the bulk reset removal on it
sure, I will await a good starting point
manic glacierBOT
ornate breach
is there a trick to running pre-commit on M1 Macs?
tulip sleet
is there a trick to running pre-commit on M1 Macs?
what's the issue, is it the Python version?
ornate breach
not sure the exact issue other than it appears to have issues with importing modules
it also happens on my intel mac so it doesn't appear to be related to the system architecture for me
[end of output]```likely an issue with my setup
orchid basinBOT
Today is March 22, 2024 so this looks to be updated with the latest boards. Now I have an RSS feed for CircuitPython.org latest added boards. This is great! Thank you.
mortal kernel
```AttributeError: module 'pkgutil' has no attribute 'ImpImporter'. Did you mean...
This sort of issue is a good reason to consider using a venv.
onyx hinge
@ornate breach it's a problem with very old pylint (still required in some of our repos). If you search the channel for "impimporter" you'll see other folks who've run into it. (including me) in github CI we've pinned python3.11 for running pre-commit for this reason. #circuitpython-dev message
old pylint + specific install method used by pre-commit + python 3.12 = broken
ornate breach
oh okay, that makes sense. somehow my intel mac is still on python 3.8 xD
my mac mini is on 3.12 though
onyx hinge
there's currently some interest in switching to ruff, which would replace pylint. which would be yay
manic glacierBOT
tulip sleet
@slender iron so 5.2 can't be in 9.1.0 because you need I2C updates for espcamera? I thought that both the old and new i2C APi was available, but maybe that was in 5.1.x.
slender iron
<@252717193496756235> so 5.2 can't be in 9.1.0 because you need I2C updates for ...
it can be but not the I2C API update
both are available in 5.2
short tendon
there's currently some interest in switching to ruff, which would replace pylint...
Interest * 1000
main furnace
I'm wondering what it would take to add displayio to the unix CircuitPython port. I would like to do unit tests for displayio. I've also uploaded the executable to ChatGPT and can have it write and run code for me.
fringe yarrow
Does anyone have a minute to walk me through the process of getting a valid CREATOR / CREATION ID to submit a new board? I believe I have something worth submitting in terms of popuarity, price point, and features. I think I get it, but want to make sure I do it right..
fringe yarrow
or i guess there's no reason not to use the UUID mapper, given the lack of branding of the hardware
slender iron
Does anyone have a minute to walk me through the process of getting a valid CREA...
Make a PR to here: https://github.com/creationid/creators
you can make a creator ID for the board designer and then a separate file for the board ids
fringe yarrow
I'll give it a go..
tulip sleet
I'm wondering what it would take to add displayio to the unix CircuitPython port...
Do you know about Blinka and https://github.com/adafruit/Adafruit_Blinka_displayio ? You can run CircuitPython code on CPython on Linux.
ornate breach
Make a PR to here: https://github.com/creationid/creators
This is handy, Iβm gonna get added π
alpine rapids
~~Hi! I'm working on adding a new board to CircuitPython and I'm almost done. The board is an ESP32-S3, and for some reason, the only way I can get it to properly install is by flashing the tinyuf2 combined.bin bootloader to the board with esptool.py, then install the firmware.uf2 that I've compiled. I've tried just flashing the firmware.bin (and tried the circuitpython-firmware.bin) that I've compiled, but those don't seem to work. I've also found, in the build directory/esp-idf subdirectory, this flash_args file:
0x0 bootloader/bootloader.bin
0x410000 circuitpython.bin
0x8000 partition_table/partition-table.bin
0xe000 ota_data_initial.bin
I'm able to locate the bootloader.bin and partition-table.bin files, but the other 2 files aren't anywhere in the build directory. Is there something I'm missing about how .bin files get built for Espressif boards? Just to reiterate, the .uf2 file works fine~~
edit: sorry, this can be ignored, I think I just needed to do a make clean
ornate breach
Make a PR to here: https://github.com/creationid/creators
is there a rhyme or reason to the creation ID we assign in the last 2 bytes?
ornate breach
Hey @lone axle I updated my community bundle PR https://github.com/adafruit/CircuitPython_Community_Bundle/pull/209
manic glacierBOT
Hi! This is my first attempt at adding a board to CircuitPython, I just received a Waveshare ESP32-S3 GEEK,
and noticed it wasn't yet supported in CircuitPython, so I did my best to follow the "How to Add a New Board to CircuitPython" guide on the Adafruit website. I've also tested the firmware built by my configuration files, and everythi...
tulip sleet
is there a rhyme or reason to the creation ID we assign in the last 2 bytes?
Something slightly mnemonic is often chosen, or if the mfr part number can be used, that's good too. But as long as it's unique, it doesn't really matter.
For creator ID's, if the mfr has a USB VID, that's a good choice.
manic glacierBOT
short tendon
Does CP support yield from?
tulip sleet
Does CP support `yield from`?
it should. I was confused by the mpy-cross error you got in that library. This works, for instance:
>>> def g():
... for i in range(3):
... yield i
>>> list(g())
[0, 1, 2]
>>> def h():
... yield from g()
>>> list (h())
[0, 1, 2]
but there are bugs: https://github.com/micropython/micropython/issues/9651
short tendon
I ask, because I was working through the ruff library tests and got that in pycamera. I'll make the change and see if I can compile it
Here are the ones I needed to ignore:
["PLR2004", "UP028", "UP030", "UP031", "UP032"]
from here: https://docs.astral.sh/ruff/rules/
Would be curious if anyone thinks we shouldn't ignore any of those...
tulip sleet
PLR2004 I can understand suppressing, because a lot of the "magic" numbers are definitely fixed and don't need naming. Plus names take space and are slower.
UP028 if it works is good (yield-in-for-loop)
UPO31 (don't use consecutive {0} {1} etc.) is worth appying
UPO32 (f-strings) I would also fix because it's almost always better and smaller
short tendon
okay, yield from compiled fine...
I worry on UPO32, since if could update something to break with these: https://docs.micropython.org/en/latest/genrst/core_language.html#f-strings-don-t-support-concatenation-with-adjacent-literals-if-the-adjacent-literals-contain-braces-or-are-f-strings
tulip sleet
I worry on `UPO32`, since if could update something to break with these: https:/...
I would just fix those particular cases as needed
short tendon
does mpy-cross fail if it's an invalid f-string style?
tulip sleet
ok, it's a mystery, mpy-cross ingesting the yield from in https://github.com/adafruit/Adafruit_Blinka_bleio is fine. I don't know why that failed during the CI build
tulip sleet
does `mpy-cross` fail if it's an invalid `f-string` style?
mpy-cross is just the interpreter, more or less, so you can just test the construct in the REPL or in a program
I don't think it fails hard, it gets the wrong result
short tendon
So I would worry someone writes code, run lints, changes it to an invalid f-string and it isn't tested and caught...
tulip sleet
So I would worry someone writes code, run lints, changes it to an invalid f-stri...
there are few or no examples of compile-time string concatenation in the libraries. I wouldn't worry about this as a blocker on converting to f-strings in general
it just means the PR's that change code need review by humans
which should happen anyway
idle owl
Is there a way to speed up refresh rate on eink in CircuitPython? The refresh rate on the default program that came with this thing was super zippy, and in CP, it's significantly slower. I know there are things built in to keep it from refreshing too often, but I don't understand why the speed is affected. The default program had a menu chooser, and flipping between things was super fast and responsive. It was likely MicroPython.
I expected that part to remain the same, but displaying this rectangle is super slow.
I guess I'm not sure what refreshing text or refreshing based on button presses looks like, I haven't tried yet.
tulip sleet
displayio is not that fast; If it was MicroPython it was probably using framebuffer, which is simple and mostly in C
but doesn't have groups, etc.
idle owl
Ah fair enough.
Bummer.
I'll have to seriously rethink my plans for this board then.
I thought of everything based on the speed of the program it shipped with. π€·π»ββοΈ
tulip sleet
which board?
idle owl
Pimoroni Badger2040W.
I was going to try to implement a game or two. Turn based, not live action or anything. But I'll have to keep the speed in mind now.
Make it a badge, have something interactive, and remote control the LEDs I added to it.
The latter has no cares about the display speed. So that's easy enough.
I think it'll be fine, it's mostly rearranging my expectations is all.
tulip sleet
you could have someone review your displayio code. There may be ways to speed it up. I ma not the greatest person for that though, I've never written a significant displayio program from scratch
idle owl
Tim is going to be helping me with the whole thing, I'm sure he can help there as well.
Thanks for the suggestion!
There's almost nothing to it at the moment. I haven't done displayio stuff either, really. So I'm struggling to even get a rectangle to display. π
The screen is black, but only a third of it should be, so obviously I don't know what I'm doing.
Β―_(γ)_/Β―
Fun little board, though. No RGB LEDs so I soldered a STEMMA QT cable to a NeoPixel stick, and tah-dah, RGB LEDs. I'll incorporate it into the 3D printed case we're designing.
Alright, lunch time here. Thanks for the info and suggestion, @tulip sleet! Appreciated as always.
short tendon
6 draft PRs for BME680, NeoPixel, PortalBase, PyCamera, PyPortal and Requests. All have commits in this order:
- Add ruff formatter
- Add ruff isort
- Add ruff lint
- Add ruff upgrade
- remove upgrade UP028
- remove upgrade UP031
- remove upgrade UP032
Happy to add to in-the-weeds if this would be a good use of time
tulip sleet
6 draft PRs for `BME680`, `NeoPixel`, `PortalBase`, `PyCamera`, `PyPortal` and `...
Thanks!! Yes, the details could be mentioned if there's something to decide on
manic glacierBOT
That's correct, the example is using SPI mode but the hardware has been implemented to support both that mode and the 4 Pin SDIO mode.
Notice the two columns of pin names at the bottom of the schematic section. The Left column are the SDIO pin names and the second column are the pins names for using SPI mode.
ornate breach
There's almost nothing to it at the moment. I haven't done displayio stuff eithe...
If you want to look at a simple game example, I have this here: https://github.com/skerr92/KeyBoy/blob/main/Firmware/Games/Loose_Cows/Loose_Cows.py
Itβs a cute little game to catch cows in a field
Using displayio (with st7789 TFT though)
idle owl
@ornate breach Yeah this is an eink, so it has to be something pretty static, and definitely turn based so there's an obvious reason to pause while it refreshes. π
This is neat though!
ornate breach
Yeah, makes sense. π I figured it would be a good reference either way. It took me a bit to figure it all out with displayio
idle owl
I get so mired in all of it. My brain starts to glaze over.
It's frustrating because I can write LED code with my eyes closed, but displayio is like a completely separate language.
But that is due to my history with coding, heh. I recognise that.
ornate breach
Itβs still a fun journey with displayio currently. I made one game and am working on others but it definitely takes a bit of mental fortitude to push through.
idle owl
I'm lucky to have Tim excited to help me out with it. π
ornate breach
Tim is really great at that kind of stuff, itβs definitely awesome of him to lend his talent π
idle owl
He wrote Snake for my last badge. This time we're coming up with some ideas, some more complicated than others. We'll see what happens. I'll definitely share the final result. π
midnight ember
It's frustrating because I can write LED code with my eyes closed, but displayio...
This is true. The syntax used in led animations and displayio might as well be two different languages. Think of it like this. If you had a wall of LED strips, perhaps even a wall of matrices, you'd want to find a way to output things to them in the most efficient way... and with layers. That's all displayio does. Pixels are pixels. π
idle owl
I had an LED wall for over a year. Never put one up in my office here.
I can't even call show() anymore in displayio. EVERYTHING CHANGED. :lolsob:
(I legit learned that today.)
midnight ember
Syndrome proved we can chain up to 25 matrix panels on a Matrix Portal S3. Matrix panels are a far more effective (and expensive) way of having a wall full of pixels.
stuck elbow
I can't even call `show()` anymore in displayio. EVERYTHING CHANGED. :lolsob:
I'm sorry
idle owl
Hah! Probably less work to mount as well.
midnight ember
aww. yes. there was transitional code for show from 8x to 9.x and would be deprecated in 9.x There are sooo many breaking changes in 9.x that it's kind of like learning circuit python all over again.
midnight ember
It should be worth it though. 9.x has a ton of great improvements.
midnight ember
finally figured out the CRLF issue for windows users. Go to C:\Users\Username\.gitconfig and modify ```py
[core]
autocrlf = input
This stackoverflow topic had the answer https://stackoverflow.com/questions/71995658/how-to-stop-github-desktop-from-changing-my-line-endings
autocrlf = false will fail. autocrlf = true will fail. autocrlf = input works.
midnight ember
Looks like I missed 1 example. Updated Youtube API example.
manic glacierBOT
manic glacierBOT
tiny peak
general question: Has anyone considered the potential for implementing a VNC server in CircuitPython that could present itself to a host computer over USB ethernet? I'm thinking of how:
- macOS Finder has a built in VNC client, and I'm guessing VNC is also easy to do on Windows and Linux
- Potentially VNC over USB-Ethernet would avoid the problems of DNS and TLS certificates that make it difficult to do HTTP-only web GUIs unless your server is running on localhost
Context for the VNC over USB-Ethernet idea:
- https://github.com/adafruit/circuitpython/pull/8831 (Add UVCFramebuffer to allow CircuitPython to send video to a host computer)
- https://forums.adafruit.com/viewtopic.php?p=1001421#p1001421 (#CircuitPython2024: IP over USB)
One of the main use cases for such a thing would be to let a small, cheap device present a big comfortable configuration GUI without having to add a large display to your BOM.
manic glacierBOT
midnight ember
Im not seeing slower wifi connect times with web workflow enabled either. Must have been my imagination. Sorry about that. :/
midnight ember
I believe the example being used is from the manual simpletest https://github.com/adafruit/Adafruit_CircuitPython_HTTPServer/blob/reuseaddr/examples/httpserver_simpletest_manual.py and it's throwing errors in 9.0 at least on the Pi Pico W anyway.
httpserver_simpletest_auto.py fails as well with the same error.
After trying a couple more examples from the repo couldn't get any of them to work on the Pico W.
main furnace
Trouble with UVC: When I enable UVC on my Raspberry Pi Pico the pico fails to boot. I can enter safe mode and remove boot.py and then it will boot.
# boot.py
import usb_video
usb_video.enable_framebuffer(32,32)
Adafruit CircuitPython 9.0.0-6-gac66da21fe on 2024-03-23; Raspberry Pi Pico with rp2040
I get the same result with Adafruit CircuitPython 9.0.0 on 2024-03-19.
onyx hinge
Trouble with UVC: When I enable UVC on my Raspberry Pi Pico the pico fails to bo...
I didn't immediately reproduce a problem on:
Board ID:adafruit_macropad_rp2040
UID:DF609C80673F1B26
boot.py output:
```with boot.py:```py
import usb_video
#usb_video.enable_framebuffer(128, 128)
usb_video.enable_framebuffer(32, 32)
```and no code.py. However with the framebuffer size set to 32x32 I don't get any video, either (I get an all-green frame, which may be just the behavior of qv4l2, the linux program I use to display usb video). At 128x128 it does work. I don't know why 32x32 would not work, though. I recommend filing an issue, especially since I didn't see quite the same behavior you did.no board crashes though
lone axle
I'm going to learn guides doing updates that match a recently submitted PR in the learn repo and noticed this:
It seems like this warning should be removed to me, at this point it may cause more confusion than help since it references old versions that are no longer listed by default.
Should I remove that while I've got this guide open or just leave it?
tulip sleet
I'm going to learn guides doing updates that match a recently submitted PR in th...
oh, sure, that's really old
lone axle
I think this may have been among the first of the displayio projects
tulip sleet
@onyx hinge We required SOREUSEADDR in 9.0.0, but it turns out that setsockopt in Pico W only implements one thing in common_hal_socketpool_socket_setsockopt(). Should it just call the lwip setsockopt routine instead? I was surprised.
onyx hinge
lwip_setsockopt isn't available, because it requires the fd-based layer and that cascades into a whole bunch of stuff
so we get to reimplement the bits we need π
tulip sleet
i will see what needs to be done to implement REUSEADDR, unless you think it's impossible. Should go into 9.0.x. I thought your refactoring of lwip would add this, but you're saying it's too much code?
also saw this:
// All socket options should be globally distinct,
// because we ignore option levels for efficiency.
but the common-hal does look at the option level
i don't know if that's you or something copied from MPy
onyx hinge
@tulip sleet in micropython it's lwip_socket_setsockopt in extmod/modlwip.c
tulip sleet
thanks, can just do what that does for SOREUSE
so that comment above is obsolete and will be more obsolete, I think
lone axle
Any Learn Editor ninja's in here? I'm trying to update these links to point to their new places in the docs:
So far the only one I changed is BusDisplay but I can't seem to get the bold to get re-applied
onyx hinge
thanks, can just do what that does for SOREUSE
take a look at https://github.com/adafruit/circuitpython/pull/9082 ?
lone axle
In the edit view the text does show as bold:
tulip sleet
were you already working on this or did i provoke you π
lone axle
well I mean the bold dropdown and and switch are enabled, within the preview the text does not visually appear bold like the other pre-existing links that I haven't touched yet.
tulip sleet
i think we started at the same time, I didn't see the email
@onyx hinge could you base it on 9.0.x?
onyx hinge
@tulip sleet oh sure!
tulip sleet
In the edit view the text does show as bold:
The actual HTML can be messed up. Check the <> option in ...
lone axle
Weird, if I delete the text entirely make new text and make it bold it appears bold at first but as soon as I add the link it visually reverts back to not bold.
Ooh thank you
I never knew there was a way to access the plain HTML π€―
tulip sleet
I have a lot of trouble with "contagious" formatting tags in Learn. Like, try to add a non-code-formatted period after a code-formatted word. Easiest thing to do is to hint enter to start a new paragraph, type the . and then delete the new paragraph newline
lone axle
Okay, I think I figured it out. Thanks again.
The front end places <strong> tags outside of <a> tag but visually that order doesn't seem to be respected by the rendering. <strong> needs to be inside of <a>.
tulip sleet
I think this is inherent in TinyMCE< which we use
onyx hinge
that was weird, something (probably a github browser extension) suggested an entirely wrong title for that PR
fringe yarrow
So I've just tried to set up a Parallel Display Bus ILI9341 on an entirely different ESP32 and entirely different screen, tested functioning with Arduino_GFX library in the arduino framework, but I get the same visual garbage that I was getting on the all in one unit I've been trying to make work for a few days.. Seems to me like, while the Parallel Display Bus functionality might work for the Lilygo TDisplay S3, it does not work for at least some other setups.
candid sun
Okay, I think I figured it out. Thanks again.
The front end places `<strong>` ...
anytime you see weird spacing check out the html view. another thing that can pop up is if you are copy/pasting text a bunch of <span> tags can sneak in
lone axle
Good to know, thank you!