#circuitpython-dev
lone sandal
tulip sleet
@slender iron Re Pants: Pants V1 says it supports C/C++ (and many others). But Pants V2 does not say it supports C/C++ (yet)
slender iron
@tulip sleet I think someone started cc support in v2
It looks pretty extensible from python too which is nice
tulip sleet
i will take a closer look 👍
slender iron
The environments feature looks handy for setting the toolchain
And they have a way of setting defaults for subfolders which is like our port level defaults
tulip sleet
bundles are a bit messed up because there is not 9.x-mpy bundle because there was no reason to build one. Also community bundle failed for other reasons (fixed that). Maybe I will force a bundle change somehow with a gratuitous library change or adding the Test library
meanwhile making circup more robust. it is upset at no bundles
tulip sleet
so does Pants substitute for all the makefiles, or for the config stuff, or what? you were thinking of ninja a while ago
or it's a substitute for our CI actions and python scripts
slender iron
There is likely a library that needs a release
All the makefile stuff
Both the building part and the config part
Ninja would only do building. We’d need python on top of it
Maybe something like toml is a better way to do config. But pants has BUILD files that are python but can’t import anything. Not sure if they can do all of Python or not
tulip sleet
hmm, you mean they are fake Python or they are just not allowed to import
slender iron
I think it’s real python but I’m not sure
Definitely says it disallows import
I like that they call it build metadata
tulip sleet
why disallow import, for security reasons or to prevent circumventing dependency analysis?
sorry to keep you here -- do you know how to list which libraries need releases? there is a script, but where??
tulip sleet
i'm looking there, but not seeing it. Eva had scripts
slender iron
I did reviews recently for foamyguy
tulip sleet
ok, I found display_text needed a release
tulip sleet
agh, python 3.12 is preventing the release.
manic glacierBOT
manic glacierBOT
CircuitPython version
9.0.0 Beta. NA?
Code/REPL
# NA
Behavior
https://circuitpython.org/board/espressif_esp32c6_devkitc_1_n8/ HEADING says it's for ESP32-S3-DevKitC-1-N8.
It shows up in search results(https://circuitpython.org/downloads?q=esp32-s3-devkitc-1-n8) too:
thorny dove
I just flashed 'Adafruit CircuitPython 9.0.0-alpha.2 on 2023-10-27' onto an Unexpected Maker ProS3. The only file that appeared in the CPY drive was: 'boot_out.txt'. Is this correct? I'm used to see also 'lib' and settings.toml for example. See image: https://imgur.com/9LmJfWU
stuck elbow
if there was a filesystem already, the other files won't be created
they are only created when a new filesystem is made
you can force it by doing import storage; storage.erase_filesystem()
orchid basin
thorny dove
they are only created when a new filesystem is made
I erased the flash, then flashed combined.bin and after a reset I copied the .uf2 to the drive
mortal kernel
I erased the flash, then flashed combined.bin and after a reset I copied the .uf...
If you're running MacOS Sonoma, you may experiencing https://github.com/adafruit/circuitpython/issues/8449
manic glacierBOT
hybrid dragon
Is the goal with the circuitpython asyncio implementation to remain as close as possible to micropython's and so PRs to support features like wait and Queue wouldn't be merged?
Was about to spend time cleaning up a Queue implementation PR and a separate refactoring so wait could also be included -- but looking at the various files it feels like that might be the case?
I see reading through some other PRs there's also a want to keep the mpy footprint as small as possible
tulip sleet
Is the goal with the circuitpython asyncio implementation to remain as close as ...
not necessarily. Internally we already have __await__(), and MicroPython doesn't have that.
See https://github.com/adafruit/circuitpython/pull/7711
https://github.com/adafruit/Adafruit_CircuitPython_asyncio/pull/40 for some initial work to make things more like CPython. Sounds like you might have already seen them.
We are trying to keep the footprint small, but not necessarily as small as MicroPython.
hybrid dragon
Gotcha. Yeah, I did see that PR. Not a hundo percent sure why it was closed or if there was more about it. I'll open a few PRs and we'll see how they feel. :)
plucky tulip
Was there recently any change to CI build system for libs for CircuitPython?
I see that it fails importing a module:
AttributeError: module 'pkgutil' has no attribute 'ImpImporter'. Did you mean: 'zipimporter'?
which suggests there was an update to Python 3.12.
It is not limited to CircuitPython: https://github.com/pypa/pip/issues/12179#issuecomment-1649148523
tulip sleet
Was there recently any change to CI build system for libs for CircuitPython?
I ...
There are a lot of problems breaking many builds because the 3.x Python on GitHub CI is now 3.12.
We are fixing them one after another.
plucky tulip
There are a lot of problems breaking many builds because the 3.x Python on GitHu...
I assume it is not possible to select previous Python version? That is not a very well thought move from GitHub side.
tulip sleet
it is possible, but some of our yaml files had "3.x" instead of a specific version.
we are pinning them to 3.11 for now.
plucky tulip
Ok, I see. Thank you 
tulip sleet
Also circup is broken, and I have a fix for that so it will not give up so quickly.
hybrid dragon
With the community bundle?
atomic summit
Can anyone officially confirm if native modules (aka Micropython native modules) works in CP?
Also, are ESP32 native emitters (aka Viper) supported? I can't seem to find anything official on googley about either. Thanks in advance!
tulip sleet
Can anyone `officially` confirm if native modules (aka Micropython native module...
viper worked in the past, but we haven't tested it for a long time. theacodes used it. I think there may be some stumbling blocks to native modules -- also not tested.
atomic summit
viper worked in the past, but we haven't tested it for a long time. theacodes us...
Ok, that's better news rthat I expected - atleats it's not specifically being excluded. Gives me some hope I might be able to get something going for my BlizzardS3 "3D" project. Thanks Dan!
blissful pollen
I did get some native code (AHRS calculation) running I think for the SAMD51. I eventually got it going (with some help from Jeff) but it isn't well document even for micropython (that I found). I can probably find my old examples if they weren't accidently removed if you ever want to see.
atomic summit
Ok, it seems that @micropython.native is supported, but when I try to use viper I get...
SyntaxError: invalid micropython decorator```Michael is about to try to get native modules going - I'll report back on this success/failures later.
atomic summit
Ok, looks like native code in CP is not supported
ValueError: native code in .mpy unsupported
actually @micropython.native is not enabled in CP 9.x - but is in 8.2.x
SyntaxError: invalid micropython decorator in CP built from master.
I'm not really sure what's involved with enabled it all again, but if I find some spare time, I'll take a look.
viscid pine
// compiled, it just passes through the function unmodified.```
onyx hinge
Happy Monday and Group hug all! I'm still out for a couple of weeks.
manic glacierBOT
slender iron
Ok, looks like native code in CP is not supported
```ValueError: native code in ...
My opinion is that we shouldn't support native code in mpy because it'll be a support burden because it increases the API surface of CP including caring about cpu architecture.
The native emitter would be ok to turn on. It is off by default but a couple of boards do enable it.
candid sun
<@&356864093652516868> We'll have our weekly meeting in about 90 minutes from now in this text channel and in the circuitpython voice channel. Please take the time to add your notes in advance to the document: https://docs.google.com/document/d/1Pz4-FdEH1Du5bWghjJ7dUheGoq61ijuFpowd5qFlIFI/edit -- I look forward to everyone's updates!
slender iron
issue 8533 is an easy core PR if folks want to make one
pallid creek
I'm trying to build CP on my new mac for the first time. When building a port I get an error about arm-none-eabi-gcc not found. I went make to the Mac instructions. They reference going to the Linux instructions and looking for the right mac build. Thinking there had to be a better way, I found someone had made a MacOS Homebrew cask for gcc-arm-embedded. Does anyone know if this method would work?
slender iron
I'm trying to build CP on my new mac for the first time. When building a port I ...
I think I use that on my mac
pallid creek
It seems to get further with this gcc installed, but then failed on a type error in the shared-module/warnings/__init__.h file. line 29.
29 | void warnings_warn(const mp_obj_type_t *category, const compressed_string_t *message, ...);
I can't tell if the issue is from my gcc or just a bug in the port and I should try to compile an older version from a branch (like the latest 8.x)
pallid creek
just now? ... yep, that fixed it! thanks.
slender iron
pallid creek
Since this works, do you think we could update the MacOS instructions page to say to use homebrew if you are using the standard arm boards?
slender iron
@tulip sleet probably has an opinion about it
generally the arm sourced gcc is what we "officially" build with
tulip sleet
That is good, if you can choose the version
thorny jay
<@&356864093652516868> We'll have our weekly meeting in about 90 minutes from no...
I will not participate as I have no notes, and had no CP activities since... weeks.
I just want to highlight that Europe just had their Daylight Saving switch, so all Adafruit streaming and meeting will be one hour ealier for a few weeks... until USA does it's own time change.
Maybe I will be able to join Show and Tell. 😉
pallid creek
brew info gcc-arm-embedded shows all of the files that are part of the toolchain. Currently 12.3.rel1. It looks like there aren't previous versions because this particular formula migrated from a different filename.
tulip sleet
`brew info gcc-arm-embedded` shows all of the files that are part of the toolcha...
i can revise the build guide so for gcc12 and up, it says to use brew
pallid creek
Is that doc built out of the github repo? If so I could make a PR for it.
tulip sleet
Is that doc built out of the github repo? If so I could make a PR for it.
no, it's a CMS system. You could use the "Feedback" link in the left sidebar to make a suggestion
pallid creek
ah. okay. thx.
candid sun
we'll get started in about 2-3 minutes
tulip sleet
@inland tusk I think you put yourself in the wrong channel
turbid radish
Hi @idle owl !!!
slender iron
how are we fixing the 3.12 issues in libraries? re: https://github.com/adafruit/Adafruit_CircuitPython_PCF8575/pull/3
turbid radish
Would someone be able to summarize for me the CircuitPython Hacktoberfest results?
gilded cradle
how are we fixing the 3.12 issues in libraries? re: <https://github.com/adafruit...
We pinned it to 3.11, so it should work if it's re-run.
lone axle
😼
slender iron
classic off by one error
inland tusk
Not hearing anything
lone axle
I can hear deshipu now. You might need to try reconnecting Charles
inland tusk
still Not hearing anything
slender iron
I can hear eightycc
crimson ferry
NINA 1.2.2?
tulip sleet
NINA 1.2.2?
yes
tulip sleet
1.2.3 (if we have it), does not work
at least in the Arduino version of NINA-FW. I'm looking at what changed to cause it to stop working
errant grail
@idle owl - I completely dog-eared the fifth edition of this reference: https://www.amazon.com/Master-Handbook-Acoustics-Seventh-Everest-dp-1260473597/dp/1260473597/ref=dp_ob_title_bk
turbid radish
Doesn't sound like a lot (sound pun)
idle owl
😊 👋🏻
lone axle
NINA 1.2.2?
I think there may potentially a few variations that all report version 1.2.2 as well. So some can work, but others raise the error.
idle owl
<@330227457296957440> - I completely dog-eared the fifth edition of this referen...
Ooooh, nice! Thank you!
idle owl
<@330227457296957440> - I completely dog-eared the fifth edition of this referen...
The description makes it seem like a lot of it is how to build things right in the first place. How much does it cover dealing with existing structures etc?
errant grail
The description makes it seem like a lot of it is how to build things right in t...
I used it to characterize studio spaces to retrofit diffusers and absorbers based on the analysis
crimson ferry
I think there may potentially a few variations that all report version 1.2.2 as ...
yeah, the versioning was fuzzy before 1.4.0. NINA_W102_Feb_17_2018.bin == 1.2.2 I believe (and I think devices still ship with that). I don't seem to have any of the files between 1.2.2 and 1.4.0 on devices or in my filesystem, though there may be some in the wayback on Discord (...1.3.0 was WPA Enterprise; some very early devices I think shipped with 1.2.1, but it had an SSL issue that was quickly fixed)
idle owl
I used it to characterize studio spaces to retrofit diffusers and absorbers base...
Got it. Thanks!
errant grail
The description makes it seem like a lot of it is how to build things right in t...
Understanding the resonance of the space is critical when selecting the retrofit devices. I ended up primarily using diffusers instead of absorbers except when a couple of bass traps were needed.
idle owl
Understanding the resonance of the space is critical when selecting the retrofit...
Fascinating. I put up those high density felt acoustic panels, and two types of weather stripping. The two types is because one is better than the other, but the better one is too thick for part of my very-apparently-not-straight door. The change is significant. I'm going to panel my housemate's office door as well, but not weather strip it because it's typically open, and we have cats. That said, my wife's office is directly below me and we still haven't resolved that. This reference could help!
errant grail
Fascinating. I put up those high density felt acoustic panels, and two types of ...
Yes, weatherstripping and sealing any seemingly small holes will really help. It's surprising how much sound can come through.
idle owl
Yes, weatherstripping and sealing any seemingly small holes will really help. It...
I walked the line between fully sealed versus being able to close without a significant amount of effort. I'm happy with the results.
slender iron
idle owl
Thanks for hosting, Liz! Thanks everyone!
gilded cradle
Thanks
errant grail
Thanks!
lone axle
Thanks for hosting Liz, Have a great week everyone!
candid sun
thanks folks!
idle owl
@errant grail Thanks again for the recommendation!
errant grail
<@339922377746022402> Thanks again for the recommendation!
You're welcome!
candid sun
<@&356864093652516868> Here is the notes document for next Monday’s CircuitPython Weekly meeting. It is at the normal time of 11am Pacific / 2pm Eastern here on Discord. Everyone is encouraged to attend! Please add your hug reports and status updates even if you’ll be attending the meeting - it’s super helpful! If you are unable to attend but would still like to include updates, feel free to include them in the notes and we’ll read them off during the meeting. Hope to see you there! <@&356864093652516868>
https://docs.google.com/document/d/1DPmDRY-Tt5awLeP4GrmR6ofnTgVacjJYlBTbvBeHYoM/edit
tulip sleet
@lone axle I am trying make a release on the CircuitPython_Org_Bundle. I have a PR to update its use of Python. But the PR runs build.yml to test, and it was complaining about a gawk error: \; is not a regexp escape. I fixed that, but build.yml is virtually the same as the one in the community bundle repo, which is not currently failing. After I fixed that but now it thinks some library in the bundle has no tag: see https://github.com/dhalbert/CircuitPython_Org_Bundle/actions/runs/6698769554/job/18201550163.
I am not sure if both of these bundles' .yml's are messed up or this is barking up the wrong tree. Their yml's are pretty different than the ones for the main adafruit bundle.
I have to go buy some Halloween candy. Will be back in a while.
slender iron
arm released gcc 13 today!
(and I'm fighting code size issues on CI but not locally)
tulip sleet
well ,we could try ti
lone axle
<@382939733107408897> I am trying make a release on the CircuitPython_Org_Bundle...
I think that warning does get printed when building the community bundle too: https://github.com/adafruit/CircuitPython_Community_Bundle/issues/187 it's come up a few times as a "red herring" when looking for other problems.
I was just doing some builds of the community bundle locally as well and I do see that warning output. But my builds suceed inspite of it.
I believe it originates in one of these lines: https://github.com/adafruit/CircuitPython_Community_Bundle/blob/bbb29f432cd1383f387101fe11e4da06de48ccc7/build.sh#L31-L32 but I do not understand them at all.
tulip sleet
if you change \: to : it goes awy, so that's a minor fix.
yeah, exactly.
so maybe there is just something wrong with the commit chosen for displayio_annotations now, though I thought it would update. Maybe this is irrelevant and we should just merge my PR anyway, and then make a new release, which will update the libraries? There are are couple of libraries that actually will update -- a new bundle was never built after they were updated.
When we went to 8.x, I kicked the can down the road hacked this bundle by just copying the 7.x bundle to 8.x, but I can't do that anymore.
lone axle
I'll review and merge that PR.
I'll let you do the release if you'd like. I see there is a new tag created. I usually create the tag on the release page, I'm not sure I know the right way to do it if the tag is created ahead of time and don't want to do it wrong and throw further wrenches into it.
crimson ferry
If a fork branch and/or entire fork is deleted out from under a closed but unmerged PR, will the code changes still be available in the PR?
(once it's merged, they stick and GitHub suggests deleting the branch)
slender iron
If a fork branch and/or entire fork is deleted out from under a closed but unmer...
I don't think so but aren't sure
crimson ferry
I'll ask in #help-with-git , maybe someone has experience with it
tulip sleet
I'll ask in <#763904151155245066> , maybe someone has experience with it
@lone axle the release did not build the bundles. I'll see if the nightly job does anything.
crimson ferry
<https://stackoverflow.com/questions/21309262/can-i-delete-a-forked-github-repos...
thanks, sounds like it would be a new PR for changes, and that's fine
lone axle
<@382939733107408897> the release did not build the bundles. I'll see if the nig...
Looks like no luck on the automated build either. I briefly looked into the dispalyio_annotations repo. I think the cause of build failures there is a very old actions configuration attempting to use container versions or dependencies that aren't available any more.
Later on I can update the actions yml files in the circuitpython org repos to the latest version from other libraries and see if that gets it building or dig further if needed.
tulip sleet
Looks like no luck on the automated build either. I briefly looked into the disp...
Thanks!
@slender iron https://forums.adafruit.com/viewtopic.php?p=991181 User wanted to use the Adafruit_USB_Host_MIDI library. Is it working yet? It hasn't been released and isn't in the bundle. Thanks.
thorny jay
Hi, I am trying to upgrade a little project from 8.2.7 to 9-a-2 but I am running into discotool/circup and mpy compatibility issue. I was wondering if the "infrastructure" problem have been solved, like generating the bundle or that kind of things?
I can totally move from mpy to .py file and fix that manually, well, I hope I still can do that.
thorny jay
Hi, I am trying to upgrade a little project from 8.2.7 to 9-a-2 but I am running...
I fixed the blocker by moving all my library to .py (circup install --py module after circup uninstall module is the trick... to be done manually one by one).
I fixed the loss of display.show(x) by changing it to display.root_group=x (thanks to hint on the REPL).
I gave up on preparing for the future and hided the FutureWarning by doing import warnings and warnings.simplefilter('ignore') thanks to some hint I saw in show and tell.
Now I am fine and my code work... but I really think some learn guide on migrating to CP 9.0 will be much needed, to deal with depreciation, and explaining the new things.
Well, new release are usually announced by a blogpost, but maybe it deserve more, documentation and guidance? Now that the cat (9.x) is out of the box, there might be a surge in support need, well I guess I should go to #help-with-circuitpython 🙂
tulip sleet
I fixed the blocker by moving all my library to .py (`circup install --py module...
which library are you using that is not installable via circup? Or is it just that the library is not updated yet?
thorny jay
All of them.
tulip sleet
We have not made a pass through the libraries to get rid of .show(), etc. Or maybe we did, come to think of tit.
thorny jay
The .show() was in my code.
tulip sleet
Did you update circup to 1.4.0? pip3 install --upgrade circup
thorny jay
Yes I am in 1.4.0, but it fail to detect my device, so I have to use --path D:\ ... or use discotool that does that for me.
tulip sleet
I can update the release notes to include updating circupt
thorny jay
It seems to fail to download the bundle at all.
And it is not in not in https://circuitpython.org/libraries either.
tulip sleet
could you post the output of a circup failure?
I have to update the circuitpython.org libraries page.
we were interrupted in updating to the 9.x bundle by a cascade of failures due to GitHub updating the default Python to 3.12, which broke some things for us
9.0.0-alpha.2 is very, very alpha. We appreciate your testing -- there's a lot to fix in CircuitPython itself and in the infrastructure
thorny jay
C:\Project>circup update
Could not find a connected CircuitPython device.
C:\Project>circup --path D:\ update
Found device at D:\, running CircuitPython 9.0.0-alpha.2.
Downloading latest bundles for circuitpython/CircuitPython_Org_Bundle (20231029).
py:
There was a problem downloading that platform bundle. Skipping and using existing download if available.
None of the modules found on the device need an update.
C:\Project>discotool update
- Running circup on QT2040 Trinkey
> circup --path D:\ update
Found device at D:\, running CircuitPython 9.0.0-alpha.2.
Downloading latest bundles for circuitpython/CircuitPython_Org_Bundle (20231029).
py:
There was a problem downloading that platform bundle. Skipping and using existing download if available.
None of the modules found on the device need an update.
So Circup fail to detect, but discotool find my device. 🙂
And Circup cannot download the bundle.
tulip sleet
those errors are only due to the CircuitPython_Org_Bundle, which is not used very much, though maybe you are using. libraries from there.
thorny jay
I am fine, I solved most of my problem, I just feel it need reporting.
tulip sleet
the regular bundle already got downloaded, and was not reported.
I think it was looking for d:\CIRCUITPY. Is your drive visible as that?
thorny jay
Oh, I was wondering why there where 3 bundles, but I did not dig more than that.
thorny jay
I think it was looking for d:\CIRCUITPY. Is your drive visible as that?
It is visible, it is D:\
Nice try: C:\Project>circup bundle-remove circuitpython/CircuitPython_Org_Bundle Cannot remove built-in module: circuitpython/CircuitPython_Org_Bundle
tulip sleet
Adafruit_CircuitPython_bundle is all the adafruit-supported libraries, CircuitPython_Community_Bundle is contributed libraries supported by individuals, and CircuitPython_Org_Bundle is other contributed libraries that are supported by the CircuitPython community. The last one is little used, and we're not sure we should keep it.
I think it's failing because it didn't downlaod properly
I have never tried the bundle-remove command
Could you file an issue in the circup repo about the remove?
thorny jay
I tried today for the first time. 🙂
tulip sleet
thanks!
tulip sleet
It is visible, it is D:\
Do you see this in an Explorer Window?
I just installed circup 1.4.0 on a Windows 10 box, and it found the board as D:\
thorny jay
Wait, it does not say CIRCUITPY ... I need to figure why:
tulip sleet
if you do storage.erase_filesystem(), (backup first!) I think you may get CIRCUITPY back
you may be able to rename the volume from Windows, but I'm not sure that will work
thorny jay
you may be able to rename the volume from Windows, but I'm not sure that will wo...
I erased and that fixed the thing, D:\ is not visible with label CIRCUITPY and is detected by circup.
I am not totally sure if I manually changed the label in the past, or if it is a result of my upgrade to 9.0.0-a-2.
slender iron
<@252717193496756235> <https://forums.adafruit.com/viewtopic.php?p=991181> User ...
I think it should work. I got lazy finishing the code though
tulip sleet
I think it should work. I got lazy finishing the code though
so release as 1.0.0 or 0.9.0 and add to the bundle? Or does it need testing?
I asked the forum user if they tested it.
slender iron
Does it pass ci? I think it worked for me but not Jeff
I’m at the vet now and haven’t gotten my laptop out
tulip sleet
I’m at the vet now and haven’t gotten my laptop out
It does pass CI. No hurry at all. The forum user can test it.
slender iron
@tulip sleet at my desk now
manic glacierBOT
ashen cedar
Eyo, I was trying to see yalls history with circuitpython and waveshare 7.5 inch.
I got the screen working on my pi 4 and python scripts with pillow library and such and it was great. But i wanted to see if i could do it with a pico and so super low battery usage
lone axle
It looks like this repo never had it's default branch changed to main instead of master https://github.com/adafruit/actions-ci-circuitpython-libs
lone axle
I think part of the issue with the CircuitPython_Org bundle is Adabot was never updated to do the automated check for updates on submodules and make releases. So no bundles are not generated when updates are made to the libraries within it.
Adding the functionality for that will require some tweaking inside adabot because there are spots where it's got hardcoded URLs of https://github.com/adafruit/
I'm trying to work out how to manually update submodules and commit to the bundle to see if I can get a good build that way
Success! 🎉 https://github.com/circuitpython/CircuitPython_Org_Bundle/releases/tag/20231031.2
It appears that's got circup back to successfully downloading this bundle as well.
tulip sleet
Success! 🎉 https://github.com/circuitpython/CircuitPython_Org_Bundle/releases/...
Fantastic - thanks!
manic glacierBOT
brazen hatch
It's porting time!
I saw the devkits making it in, so it will be a good chance for me to test c6
Let's hope it ain't possesed like c3
tulip sleet
@lone axle what do you want to do with https://github.com/adafruit/actions-ci-circuitpython-libs/pull/21 ? Should we leave it as is, as Melissa suggested?
lone axle
<@382939733107408897> what do you want to do with <https://github.com/adafruit/a...
I am on board with leaving it as-is. I definitely was mistaken about it being related to my issue, I was fooled by an unrelated mention of setuptools in the error that I had.
I'll close it.
mortal kernel
It's porting time!
Any thoughts re: Zigbee support?
pallid creek
I was wondering if the PDF support matrix was every turned back on?
slender iron
Let's hope it ain't possesed like c3
I think they are pretty similar so don't get your hopes up.
slender iron
I was wondering if the PDF support matrix was every turned back on?
We don't plan on turning it back on. You can build it locally. IMO the html version is much more usable
we could have it upload to s3 I think
brazen hatch
Any thoughts re: Zigbee support?
Oh, I'm not a home automation guy, but it's a cool protocol.
Once I get a few more devices with it, I will try using it.
pallid creek
We don't plan on turning it back on. You can build it locally. IMO the html vers...
Why did it exist originally? It is useful for people making books and PDF courses?
slender iron
originally it was all one pdf but the displayio split made the pdf too big to render. to fix it, I split it out into a separate pdf. readthedocs errors when two pdfs are generated
so we disabled the support matrix pdf
brazen hatch
Ah has anybody other than me tested C6?
Cuz like.. the wifi is broken for mine..
It scans but all connections fail with Error 15.
It could be the bug where the network securities are mixed, but I still haven't gathered the time to propely document what is going on on that
slender iron
I tested it on the espressif board when I added it
brazen hatch
Hmm, it could then be the network securities being messed up.
I have 3 open, 1 enterprise and a bunch of hidden ap's around me.
So that bug is greatly exadurated here.
It may be a great opportunity for me to try to fix it here.
keen eagle
Ah has anybody other than me tested C6?
I just flashed the an ESP32-C6-DevKitC-1-N8 with CircuitPython 9.0.0-alpha.2 and I can confirm, attempting to connect to wifi gets an error:
import wifi wifi.radio.connect("NOPE", "ALSO NOPE") Traceback (most recent call last): File "<stdin>", line 1, in <module> ConnectionError: Unknown failure 15
crimson ferry
where are the docs for https://docs.circuitpython.org/en/latest/README.html#differences-from-micropython? should this also mention web workflow:
Supports native USB on most boards and BLE otherwise, allowing file editing without special tools.
(at the top of that page too, where USB and BLE are discussed)
tulip sleet
where are the docs for <https://docs.circuitpython.org/en/latest/README.html#dif...
It is in the toplevel README.rst
slender iron
@tulip sleet want to chat memory allocation debugging in an hour?
tulip sleet
ok sure
slender iron
gonna eat lunch and take a short walk
tulip sleet
i should take a walk too, then
slender iron
I debugged https://github.com/adafruit/circuitpython/issues/2885 and it is a pathological case
tulip sleet
a bunch of those kevin walters nrf52 bugs went back up on my radar a bit after going through all the isuses
we can talk about it
slender iron
ya, that's fine
👍
I used it as an excuse to learn more about the split heap stuff
slender iron
@tulip sleet I'm back from my walk
tulip sleet
<@329766224093249548> I'm back from my walk
now I"m back too
tulip sleet
@slender iron are you occupied?
slender iron
<@252717193496756235> are you occupied?
back now. just chatted with my sister
tulip sleet
any time or we can wait till later if you want
brazen hatch
I did test it throughtly, it's a totally unrelated bug to anything else.
I am home now, where everything is standard across the boards (pun intended).
slender iron
I am home now, where everything is standard across the boards (pun intended).
C6 stuff is a low priority for me now so don't expect any adafruit 👀 on it soon
brazen hatch
C6 stuff is a low priority for me now so don't expect any adafruit 👀 on it soon
It's to be expected, ESP32-CX is pretty annoying to work with.
I will try to figure it out, since I have an example to work from.
This is an example of ESP-IDF console component.
Type 'help' to get the list of commands.
Use UP/DOWN arrows to navigate through command history.
Press TAB when typing command name to auto-complete.
Press Enter or Ctrl+C will terminate the console environment.
esp32c6> join --timeout=5000 KeyFalse [Redacted]
I (11507) connect: Connecting to 'KeyFalse'
I (11517) pp: pp rom version: 5b8dcfa
I (11517) net80211: net80211 rom version: 5b8dcfa
I (11527) wifi_init: rx ba win: 6
I (11527) wifi_init: tcpip mbox: 32
I (11527) wifi_init: udp mbox: 6
I (11537) wifi_init: tcp mbox: 6
I (11537) wifi_init: tcp tx win: 5744
I (11537) wifi_init: tcp rx win: 5744
I (11547) wifi_init: tcp mss: 1440
I (11547) wifi_init: WiFi IRAM OP enabled
I (11547) wifi_init: WiFi RX IRAM OP enabled
I (11557) phy_init: phy_version 202,b4b3263,May 17 2023,20:14:14
W (11577) wifi:(bf)761:0x600a7cac:0x01b4b4b0
W (11577) wifi:(agc)0x600a7128:0xd21c0800, min.avgNF:0xce->0xd2(dB), RCalCount:0x1c0, min.RRssi:0x800(-128.00)
W (11587) wifi:(TB)WDEV_PWR_TB_MCS0:19
W (11587) wifi:(TB)WDEV_PWR_TB_MCS1:19
W (11587) wifi:(TB)WDEV_PWR_TB_MCS2:19
W (11587) wifi:(TB)WDEV_PWR_TB_MCS3:19
W (11597) wifi:(TB)WDEV_PWR_TB_MCS4:19
W (11597) wifi:(TB)WDEV_PWR_TB_MCS5:19
W (11597) wifi:(TB)WDEV_PWR_TB_MCS6:18
W (11607) wifi:(TB)WDEV_PWR_TB_MCS7:18
W (11617) wifi:(TB)WDEV_PWR_TB_MCS8:17
W (11617) wifi:(TB)WDEV_PWR_TB_MCS9:15
W (11627) wifi:(TB)WDEV_PWR_TB_MCS10:15
W (11627) wifi:(TB)WDEV_PWR_TB_MCS11:15
I (13967) esp_netif_handlers: sta ip: 10.42.0.228, mask: 255.255.255.0, gw: 10.42.0.1
I (13967) connect: Connected
In the github actions for libraries, I'm guessing there's no way to get the circuitpython async/await behavior working, right?
(Context: Trying to add tests for Adafruit_CircuitPython_asyncio and python 3 async/await is not havin' any of my shenanigans.)
tulip sleet
@wraith crow Note that I opened issues for all the bundle libraries that have .show()
wraith crow
In the library bundle? I guess I should take a look before I go much further 😄
tulip sleet
it was a lot more than I thought when I started 🙂
no, in each repo there's a new issue, so you can link that issue.
wraith crow
Ah okey thanks!
tulip sleet
I found them by ag-ing in the bundle
wraith crow
I'm working of a grep of the bundle
tulip sleet
same difference 😆 I just thought I'd mark them all as a kind of to-do list
They all have the same title
wraith crow
They're actually pretty quick, of course I almost immediately stumbled on the LEDs still using .show and realized I needed to slow down a bit 😁
tulip sleet
yes, that was confusing
and there are some framebuffer non-displayio ones with .show() also, but usally self.show()
thorny jay
Do you really want an issue for this:
CircuitPython core code crashed hard. Whoops!
Hard fault: memory access or instruction error.
Please file an issue with your program at github.com/adafruit/circuitpython/issues.
Press reset to exit safe mode.
Press any key to enter the REPL. Use CTRL-D to reload.
]0;🐍REPL | 9.0.0-alpha.2\
Adafruit CircuitPython 9.0.0-alpha.2 on 2023-10-27; Adafruit Feather RP2040 USB Host with rp2040```
I am just more or less running the learn guide code for read QR code with tiny code reader: https://learn.adafruit.com/reading-qr-codes-with-the-tiny-code-reader/code
I cannot go to 8.x as I need some 9.x feature.
My plan is to roll forward to the latest version 9 I can find.@lone axle re .show() I have not been making major or minor version releases, just bug fixes (x.y.z+1), since the user API did not change. watcha think?
I see you bumpd layout to 2.0.0
i could re-release what i released so far
random junco
https://github.com/search?q=org%3Aadafruit+is%3Aissue+in%3Atitle+convert+.show&t...
Do you want help with any of the PRs to update to use root_group?
tulip sleet
Do you want help with any of the PRs to update to use `root_group`?
Thanks! Feel free to do some. The issues are meant as a to-do list. It will be a while before this is critical.
random junco
Sounds good, think I can handle changing one line. 🙂
tulip sleet
check for root group being set to None when it should be CIRCUITPYTHON_TERMINAL (if you want to show the terminal).
lone axle
<@382939733107408897> re `.show()` I have not been making major or minor version...
Sorry, it seems I never fired up discord this morning. I'm good with either way. In the moment I chose it I went with 2.0.0 since there are contexts where the new code won't run (albeit quite old versions of circuitpython at this point.)
But a person who does have one of those older versions installed would update the library and then have it "stop working"
tulip sleet
they'd have to be running 7.x.x or earlier, yes
so maybe I'll do a minor version bump on future releases. Major version usually implies API incompatibilities
lone axle
I'll do minor number for any others that I release. It'll be a while before I am back to any of them though. I just took a quick stroll through github notifications before getting started with other work.
tulip sleet
But a person who does have one of those older versions installed would update th...
in most of the libraries, fortunately there is not actually a .show() in the library -- it's only in the examples
lone axle
True
random junco
The first PR I did to update with root_group is failing the CI at https://github.com/adafruit/Adafruit_CircuitPython_SPD1656/actions/runs/6734034840/job/18304005944?pr=7 with error /home/runner/work/Adafruit_CircuitPython_SPD1656/Adafruit_CircuitPython_SPD1656/adafruit_spd1656.py:docstring of adafruit_spd1656.SPD1656:15:'any' reference target not found: displayio.FourWire - is this related to the displayio splitting Scott has been doing? I didn't touch that file when editing, so not sure
mortal kernel
@danh Continuing my deep dive into oofatfs I've found that FAT32 and EXFAT support is turned on just for the Broadcom ports. This means that any other boards with an SDCard slot will have a hard size limit of 4G (largest FAT16 filesystem supported by oofatfs). Also, an SDCard formatted on a PC and inserted in a board running CP is likely to either not mount or if it does it may become corrupted.
lone axle
The first PR I did to update with `root_group` is failing the CI at <https://git...
I believe it is related to the displayio split, Fourwire is now in it's own module I think https://docs.circuitpython.org/en/latest/shared-bindings/fourwire/index.html
displayio.Fourwire -> fourwire.Fourwire might fix it, but I haven't actually tried in any code so I'm not positive.
mortal kernel
I've looked at https://github.com/adafruit/circuitpython/issues/5706 but it may not be this problem.
random junco
I'll give it a try, thanks!
tulip sleet
@danh Continuing my deep dive into `oofatfs` I've found that FAT32 and EXFAT sup...
I am not sure that 4GB is the limit, because I've used 16GB cards and larger on PyPortal
without reformatting
mortal kernel
I am not sure that 4GB is the limit, because I've used 16GB cards and larger on ...
The limit is in the oofatfs formatting code.
tulip sleet
The first PR I did to update with `root_group` is failing the CI at <https://git...
intersphinx in the doc build is using < https://docs.circuitpython.org/en/latest/>, so maybe yes
tulip sleet
The limit is in the `oofatfs` formatting code.
but referencing parts of a preformatted card outside the 4GB limit is going to fail?
mortal kernel
Not sure, but it could result in some unexpected weirdness.
tulip sleet
the FAT32 file size limit is 4GB, but I don't think the volume size limit is 4GB?
FAT32 can support 32GB volumes
mortal kernel
Right, FAT32 can get up into the TB range on volume size.
tulip sleet
I'm not even sure the formatting code is available to use?
the intention was only to turn it off for small builds
the formatting code, that is
mortal kernel
Yeah, just gonna go there. As it is FAT32 and EXFAT are off for all ports except Broadcom.
Pertinent flags are MICROPY_FF_MKFS_FAT32 and MICROPY_FATFS_EXFAT
tulip sleet
so i see the flag is MICROPY_FATFS_MKFS_FAT32, which is on for most:
py/circuitpy_mpconfig.h
267:#ifndef MICROPY_FATFS_MKFS_FAT32
268:#define MICROPY_FATFS_MKFS_FAT32 (CIRCUITPY_FULL_BUILD)
ffconf.h:76:#define FF_MKFS_FAT32 MICROPY_FATFS_MKFS_FAT32
mortal kernel
Ouch! Missed that.
tulip sleet
np - that's a relief
@gilded cradle Do you have an opinion about how we should bump the version numbers of libraries where we remove show()? In many cases the changes are only in the examples. Should it be x+1.0.0, x.y+1.0, or x.y.z+1? I.e. is it just a bug fix, or a minor or major bump? See my discussion with foamyguy not quite two hours ago above
gilded cradle
<@478616919721771019> Do you have an opinion about how we should bump the versio...
I would just do a minor bump since the interface of the lib is likely not changed, but it's bigger than just a bug fix.
tulip sleet
@lone axle ☝️
wraith crow
@danh So what do you think remove the root_group=None's or leave em?
@tulip sleet My tag didn't take 😆
tulip sleet
I have started to hit tab after typing an @ tag
go ahead and leave them in. The code will still work. We are just trying to fix code that will break
thanks for thinking about it
approved
random junco
Thanks for the linking tip Dan
brazen hatch
Oh.. Oh no...
Even the scan example is broken..
I think idf5.1 is at fault
Guess it's upstream bug time.
brazen hatch
Here we go I guess.
https://github.com/espressif/esp-idf/issues/12517
stone rock
hey im trying to build circuitpython but im stuck on ‘makefile fetch-all-submodules’
makefile is not recognizing
ornate breach
hey im trying to build circuitpython but im stuck on ‘makefile fetch-all-submodu...
Change to make fetch-all-submodules
stone rock
sorry i meant to write ‘make’ not ‘makefile’
wraith crow
I sometimes forget to change to the circuitpython folder before trying to run the fetch-all-submodules
You can also reduce the time significantly by changing to the ports/whicheverport folder and running make fetch-port-submodules
hybrid dragon
In the github actions for libraries, is there any way to get the circuitpython async/await behavior working?
(Context: Trying to add tests for Adafruit_CircuitPython_asyncio and python 3 async/await is not havin' any of my shenanigans.)
My assumption is no - I asked a few days ago, thought I'd ping it again juuust in case it catches someone that knows a way around it.
wraith crow
I'm not real familar with async/await but you may have better luck in #help-with-circuitpython the folks hanging out in this channel are generally focused more on circuitpython development
hybrid dragon
Ah, okay - I was thinking since it was a library to do with circuitpython & the build system designed by the same folks this'd be the right place. I'll ask there, too.
onyx hinge
In the github actions for libraries, is there any way to get the circuitpython a...
We can and do test this during the main circuitpython build. Look in tests, there are existing asyncio tests. But the test is done with whatever is the submodule ref not latest or what's in a PR.
tulip sleet
@wraith crow @random junco thanks for all the .show() fixes!
hybrid dragon
That's really helpful, thank you!
So if any new features make it into the asyncio library these tests would be the next place to update.
In particular if it's something new, since it looks like that'd fail the uasyncio compat test.
I'll dig into how that's used as part of CI. Maybe I can bring a similar binary into the other repos for running tests in an actual environment instead of cpython..
tulip sleet
@lone axle @slender iron @gilded cradle @random junco @wraith crow One thing about all these updates to fourwire.FourWire, etc., for the displayio examples in the library repos is that the displayio refactor is only in 9.0.0 builds. Normally we would defer these changes until 9.0.0 final was released. But because the library doc builds look at http://circuitpython.readthedocs.io/en/latest/, they require this change. I went ahead and merged in those changes, but it will be a support burden to to tell people "no, for 8.x.x, you have to use displayio.fourwire if you copy the examples".
I'm not sure this is a big deal, because it's only for examples in libraries for certain display boards. We could not depend on latest, but instead the current released doc for the cross-references. (And is that already factored out into a separate place, or is it in each library repo?) On the other hand, then the library docs can't refer to new features.
river quest
@crude blaze hihi
hybrid dragon
Is github having an issue or was this repo moved? https://github.com/adafruit/Adafruit_CircuitPython_asyncio 
oh wait, discord sees it? I see a 404. Weird..
Oh, github status says it's havin a bad time
Nevermind! :) I'll wait for github to resolve their problems.
slender iron
@tulip sleet I'm at my desk for a little while. need anything from me?
lone axle
Should we remove this warning about CircuitPython version? It's quite outdated at this point
slender iron
sure!
midnight ember
I've seen more than a few learn guides with warnings with CP 6 or CP7. I've used that guide by Jepler a lot when constructing the 12-panel matrix and didn't find anything wrong with the code, examples, information. It's an excellent excellent guide that got me up and running much faster than I could have done on my own that's 💯
tulip sleet
I've seen more than a few learn guides with warnings with CP 6 or CP7. I've use...
if you see these CP6 or CP7 warnings that are kind of obsolete, feel free to mention them here or use the "Feedback" link in the left sidebar of the guide to note it
main furnace
What toolchain do I neet for RT1011?
tulip sleet
What toolchain do I neet for RT1011?
just the ARM gcc, note that we just upgraded to gcc 13 on main.
main furnace
Cool, thanks for the tip.
tulip sleet
i will add some new info to the Building guide
hybrid dragon
I'll be trying to figure this out myself, but if someone knows the answer to these things it saves me some time:
- Is the unix port of circuitpython well maintained?
- Is there a reason beyond time constraints as to why it isn't being used in CI for testing libraries? I'm in particular looking at
Adafruit_CircuitPython_asyncio. - Is there a build of it uploaded anywhere that CI could access?
manic glacierBOT
LILYGO T-Deck
Specifications
CPU: ESP32-S3
Flash: 16MB
PSRAM: 8MB
Display: 2.8" ST7789 (320x240)
Radio: 2.4Ghz WiFi, Bluetooth 5, and LoRa (optional)
tulip sleet
I'll be trying to figure this out myself, but if someone knows the answer to the...
It is maintained only well enough to run the tests
many libraries require external hardware to function. There are asyncio tests already, inherited from MicroPython.
We don't publish a build. We just run the "coverage" build: make -C ports/unix VARIANT=coverage
to run the tests
it is somewhat more important in the MicroPython world
hybrid dragon
It is maintained only well enough to run the tests
Got it, thank you!
many libraries require external hardware to function.
That's true.. and making them testable by breaking them into smaller pieces kind of run into the pystack limitations.
Yeah, there's asyncio tests that are part of the circuitpython build but it means changes to asyncio the library is slightly more roundabout, hence me digging into it a little.
tulip sleet
> It is maintained only well enough to run the tests
Got it, thank you!
> many...
There are tests that test the asyncio library and tests that test _asyncio. asyncio the Python library is in the MicroPython repo. We factored it out, and try to keep it roughly up to date by updating it from the repo. Maybe you know this already
hybrid dragon
I know some of it. It's somewhat different at this point from what I can tell vs uasyncio from the micropython library. I was looking into adding tests to verify additional aspects of it that weren't part of the micropython pieces. As long as the intent is to keep it in line with micropython's featureset instead of getting it closer to CPython's then that path seems reasonable.
My thinking was that the tests in the circuitpython repo were to ensure that circuitpython behaved as expected - especially with changes to the async/await behacior - towards asyncio. However, pull requests towards the adafruit/Adafruit_CircuitPython_asyncio repo don't get validated that they don't incur regressions or otherwise have glaring problems.
I was poking at maybe being able to support some tests via CPython & pytest but the async/await behavior from circuitpython seems a bit different from CPython's - at least, under the hood. And most of the asyncio features rely on those differences.
For any other libraries that's really not that much of a concern - as all those internal differences are encapsulated, so pytest and cpython should technically work? 🤷 Maybe a bad return on investment going down this path, since like you said it was factored out from micropython and heavily reliant on internal features. Not many other libraries would need this to run tests.
tulip sleet
I know some of it. It's somewhat different at this point from what I can tell v...
So what I am saying is that https://github.com/adafruit/Adafruit_CircuitPython_asyncio is derived almost exactly from https://github.com/micropython/micropython/tree/master/extmod/asyncio. During our recent merges from MicroPython we tried to update our separate library with the upstream changes from extmod/asyncio. _asyncio is optional C-language helper functions, but the library can work without it.
uasyncio and _uasyncio dropped the u in the latest version of MicroPython.
Our async/await impl is slightly different from MPy's. We have __await__() implemented separately, and we distinguish between coroutines and generators; MPy does not. We would like to discuss with the MPy folks PR'ing that change back.
Did I already point you at geneverov's experimental work redoing some things in the C implementation and also trying to take the CPython asyncio library and strip it down? The resulting library is still about 4x larger than the MicroPython version.
MPy/CPy also don't have Exception groups, and that I don't think is on the short term list for MPy
hybrid dragon
I did see the changes to the C implementation
tulip sleet
I agree we should add testing to the asyncio CPy library to exercise the tests that are in the circuitpython repo. Right now that only happens when we update the submodule in circuitpython
tulip sleet
the person who did await/async in CircuitPython did it separately from the MPy implementation. That was way before we added the asyncio library: https://github.com/adafruit/circuitpython/pulls?q=is%3Apr++async+author%3Akvc0+
we originally thought asyncio was too big, too complicated, and too confusing. This was before asyncio task groups, and before a lot of simplification of asyncio had happened. I had read about curio and trio .
manic glacierBOT
CircuitPython 9.x is now running on the device with display output.
The rest of the hardware needs to be tested to ensure the assigned pins under board are accurate.
manic glacierBOT
manic glacierBOT
manic glacierBOT
mortal kernel
Is it OK to submit just a pull request or should it have a corresponding issue?
stuck elbow
it's fine
mortal kernel
Besides some of the SAMD-based boards, does anyone know of others with small flash drives?
winter mortar
Besides some of the SAMD-based boards, does anyone know of others with small fla...
You mean separate flash chip? The RP2040 Feather & possibly others with the RP.
mortal kernel
You mean separate flash chip? The RP2040 Feather & possibly others with the RP.
Flash in the SoC or separate. I'm looking for very small, like the M0 Trinket that exposes 64KB of its flash as a filesystem.
Context: This is for an optimization to filesystem overhead on very small devices.
tulip sleet
Besides some of the SAMD-based boards, does anyone know of others with small fla...
Feather M0 Basic, Gemma M0, Trinket M0, QT Py (SAMD21), many of the Trinkey boards, Arduino Zerodon't have an external flash chip. Anything in ports/atmel-samd/boards with INTERNAL_FLASH_FILESYSTEM = 1 in mpconfigboard.mk. You can install Feather M0 basic on Feather M0 Express
sorry, missed the "besides", there are also nRF boards with internal-only CIRCUITPY
some are fairly small
mortal kernel
some are fairly small
Thanks! Any thoughts on whether scaling down the FAT12/FAT16 root directory size from 512 entries to something more reasonable for extremely small filesystems should be a #define based on board or an inline test based on filesystem size? I've tested using the latter, and it frees up 25% of the flash filesystem on the Trinket M0.
tulip sleet
I didn't know you could make it smaller. I'd do it based on filesystem size. what size improves things by 25% on 64kB?
mortal kernel
I didn't know you could make it smaller. I'd do it based on filesystem size. wha...
Going to 32 entries from 512. After overhead for reserved sector and FAT it leaves about (working from my memory here) 60 512-byte clusters free for the data area.
tulip sleet
32 is probably a little small, 64 or 128 might be better for libraries and data files
mortal kernel
32 is probably a little small, 64 or 128 might be better for libraries and data ...
Keep in mind it's the root directory. Libraries and other sub-directories allocate out of the data area.
tulip sleet
ah ok
would it be flexible enough to read current format, and only drop size on a reformat?
do all OS's handle this?
mortal kernel
would it be flexible enough to read current format, and only drop size on a refo...
Yes, works with existing filesystems, so nothing changes until a reformat. I've tested Linux, Windows, and MacOS.
slender iron
I know some of it. It's somewhat different at this point from what I can tell v...
Our goal with CP is to be more like CPython than MicroPython. So if you have behavior differences, then it'd be good to make CP match CPython.
slender iron
ember iris
ahh that explains why my discord is loading slowly
turbid radish
Hi @idle owl !!
idle owl
Congrats!
slender iron
(I'm happy to read too if you need a break)
hybrid dragon
That's good to know! The biggest issue with that today that I've seen has to do with a balance of features to package weight.
errant grail
Sorry to hear about your mother, Scott.
slender iron
That's good to know! The biggest issue with that today that I've seen has to do...
Ya, we definitely target being a subset of CPython
we want code to be portable from CP to CPython but not necessarily the reverse
turbid radish
Thanks all!
idle owl
Thanks, all!
errant grail
Thanks!
ember iris
Thanks all!
lone axle
Thanks everyone! Hope everyone has a good week!
candid sun
thanks for hosting Tim!
gilded cradle
Discord finally just let me online. Guess the meeting is over now.
It would be nice if we had a link to the doc outside of discord in case we can't get on discord, we can at least add notes.
lone axle
I see them listed if I go to drive.google.com and click "shared with me" on the left. I'm not sure if that's the same for everyone or not though.
gilded cradle
I do not see them there. :/
Nevermind, I had to switch user accounts. I see them now.
Thanks @lone axle
lone axle
Here is the notes document for next Monday’s CircuitPython Weekly meeting. It is at the normal time of 11am Pacific / 2pm Eastern here on Discord. Everyone is encouraged to attend! Please add your hug reports and status updates even if you’ll be attending the meeting - it’s super helpful! If you are unable to attend but would still like to include updates, feel free to include them in the notes and we’ll read them off during the meeting. Hope to see you there! <@&356864093652516868> https://docs.google.com/document/d/1GChyaFlm9dC-TNY4DgkKVXk51ZxJCHfd42bJOstZmIg/edit?usp=sharing
cold oar
Hi there - can anyone point me a library for detecting a variety of hardware button press patterns - short hold, long hold, double click, triple click, etc?
wraith crow
You might try asking in #help-with-circuitpython but it looks like the debounce library might do some of what you're looking for https://docs.circuitpython.org/projects/debouncer/en/latest/
manic glacierBOT
mortal kernel
@tulip sleet I'm curious re: reduced flash usage on 9.0.0, how was that achieved?
tulip sleet
we swapped our set of changes for reducing the size of type objects for MicroPython's. That was probably a wash. Going to gcc13 made the biggest difference.
Also we removed long-lived storage allocation, thought that probably was not a lot of code.
gcc13 gained >1200 bytes on small builds: https://github.com/adafruit/circuitpython/pull/8546#issuecomment-1787472714
mortal kernel
gcc13 gained >1200 bytes on small builds: <https://github.com/adafruit/circuitpy...
Fascinating. I'll want to update my local chains from 12.3 to 13. BTW, I did a deep dive into oofatfs differences between upstream, MicroPython, and CircuitPython. I've got notes, where would be a good place to put them?
tulip sleet
Fascinating. I'll want to update my local chains from 12.3 to 13. BTW, I did a d...
You can just make an issue. The object of it would be "update oofatfs" or something like that.
meager fog
@crude blaze hihi
mortal kernel
@tulip sleet I'm having the same failure for explorerkit_xg24_brd2703a for en_US during CI of PR 8567 on main that I was hitting on 9.2.x. From the log, it looks like it may be related to the GitHub Python update.
tulip sleet
i'll take a look
tulip sleet
<@329766224093249548> I'm having the same failure for explorerkit_xg24_brd2703a ...
https://stackoverflow.com/questions/73742774/gcc-arm-none-eabi-11-3-is-not-implemented-and-will-always-fail but I don't know why it's only that board
and why it just started showing up
mortal kernel
and why it just started showing up
Mysterious. Anything I can do to help?
I was seeing the same warnings building locally with gcc 12.3.
tulip sleet
but only on that board, is that right?
mortal kernel
Can't say, it was the only one I tried.
tulip sleet
the differences in the port/silabs/boards files don't seem consequential
mortal kernel
I looked at those too, agree, nothing consequential. I wonder if SiLabs changed something in their SDK that's board specific? Are we picking a specific checkin or taking top-of-tree?
tulip sleet
mortal kernel
The reason I thought is was Python related was because of these messages:
if (not build_all) and (language is LANGUAGE_FIRST) and (exit_status is 0):
/home/runner/work/circuitpython/circuitpython/tools/../docs/shared_bindings_matrix.py:194: SyntaxWarning: invalid escape sequence '\d'
for m in re.findall('-D([A-Z][A-Z0-9_]*)=(\d+)', line):
/home/runner/work/circuitpython/circuitpython/tools/../docs/shared_bindings_matrix.py:218: SyntaxWarning: invalid escape sequence '\s'
"\s+:target:\s+(http\S+(docs.circuitpython|readthedocs)\S+)\s*",
Note: Not building languages {'cs', 'hi', 'el', 'ko'}
building boards with parallelism 4
That is a problem, but not a big one. non-raw strings now complain about undefined escape sequences. I can do a PR for that along with some miscellaneous things. I don't think it is related.
I am re-running the one failed job to see if it's intermittent (maybe build order). I think will ping the silabs person about this, assuming it still fails, to see if they have some idea. I am not too averse to merging 9.0.0PR's if only one board is not working
mortal kernel
No problem for me to wait. Just concerned this will hit any subsequent pulls.
Watching that step run again. That's some amazing CI foo you've got there.
tulip sleet
Watching that step run again. That's some amazing CI foo you've got there.
ok, it worked the second time 🤷
there is some reason why it happens, but I have no idea
slender iron
@tulip sleet need anything from me? at my desk for a little while now
tulip sleet
there are a few PR reviews, not a big deal
tulip sleet
don't have to do them now
thorny jay
Hi, when I download a project bundle from a learn guide, there is a README.txt that start with
This bundle contains two folders, one each for CircuitPython 7.x and CircuitPython 8.x. You must use the files that match the version of CircuitPython you're using, e.g. if you are using CircuitPython 7, you will copy the files from the CircuitPython 7.x folder to your CIRCUITPY drive.
However, since a few days, the folder are CircuitPython 8.x and CircuitPython 9.x.
That file is likely dynamically generated as it say what is the download URL and download date... not sure where in Adafruit this is generated.
Maybe you should change the text, or find a version agnostic wording?
tulip sleet
Hi, when I download a project bundle from a learn guide, there is a README.txt t...
Thanks, I will check with the Learn Guide folks about this.
slender iron
@tulip sleet split heap PR is ready for review too when you have a chance
I expect it'll need tuning before the stable release
tulip sleet
<@329766224093249548> split heap PR is ready for review too when you have a chan...
what do you think of the "large stuff doesn't load" issues?
we can adjust the split size?
I wasn't sure if you wanted to tune before merging?
slender iron
ya, we may need to tune it
I think it'd be better to get multiple examples and then look at it
maybe start an issue to collect examples, especially learn guide code that fails
one way to fix might be making more static allocations dynamic
slender iron
I made the starting size a toml setting
and allocations will double after that I think
but we may need to tune how it works when running out of outer heap
kk, I merged your PR and approved another that is waiting on CI. feel free to ping me when you need to be unblocked
I can hop on my laptop as-needed
now time for daycare pickup
tulip sleet
will do -- I'll review and merge, and I can release an alpha with big caveats
midnight ember
Is it possible to add larger PackSize to the lc709203f library or is the chip limited to a maximum size of 3000mah?
If it is limited can a code comment be added that 3000 is the absolute maximum size? I'm using it with a 10000mah battery and while it does work with a PackSize of 3000 I'm a little concerned it won't have an optimal lifespan using the wrong PackSize.
I'm using it with an older Adafruit ESP32-S3 Feather that has the lc709203f battery monitor chip.
https://www.adafruit.com/product/5035 there are also larger 4400 & 6600 mah batteries too
blissful pollen
Is it possible to add larger PackSize to the lc709203f library or is the chip li...
Looking at the datasheet (https://www.mouser.com/datasheet/2/308/ENA2237-D-321387.pdf) there doesn't seem to a limit. The values in the library are written to a register that says "Compensate for total impedance between the Battery and Fuel Gauge". So I think you could add more values but you would have to know what value to set. Or measure the impedance.
The only limit seems to be that is measures 1 cell batteries.
midnight ember
Looking at the datasheet (https://www.mouser.com/datasheet/2/308/ENA2237-D-32138...
hmm I'm going to guess the double and triple 18650 packs have a BMS chip built in to act as a single cell. maybe it would work if I modified the library? will look into it and maybe do a PR if it works.
manic glacierBOT
manic glacierBOT
manic glacierBOT
hybrid dragon
Would adding exception() / result() to _asyncio's Task class count as a breaking feature? I'd assume not. Just curious before I dig in.
tulip sleet
Would adding `exception()` / `result()` to `_asyncio`'s `Task` class count as a ...
no, it's upward compatible
hybrid dragon
Wicked.
tulip sleet
you could consider proposing it for MicroPython and submit it upstream as well
we'd put it in 9.x in any case
hybrid dragon
I'll take a peek
Can't be too much harder to add them there. Not like it'll actually do anything yet, either. Needs support via the asyncio lib for that.
wraith crow
I'm poking around Scott's split heap PR trying to figure out (without much luck) where the mechanism is that doubles the heap allocation size. Have you looked at it close enough yet to get me pointed in the right area?
tulip sleet
I'm poking around Scott's split heap PR trying to figure out (without much luck)...
this behavior is inherited from MicroPython. See gc_try_add_heap() in py/gc.c. Scott has switched us over to using this.
wraith crow
Thanks that's close to where I was, just wanted to make sure I wasn't wasting my time trying to figure out exactly what's going on in there 😄
tired nacelle
That was me
Moving here the chat we had in the general chat
I use CP 9/ESP32-S3 and I'm getting NotImplementedError: main_clock when I try to use the I2SOut class. Is full mclock support in the roadmap? Can I pair with someone to help implement that? I attempt to use ESP32-S3 with SGTL5000.
tulip sleet
Moving here the chat we had in the general chat
> I use CP 9/ESP32-S3 and I'm ...
well, that looked easy! did you test your PR and does it work?
tired nacelle
well, that looked easy! did you test your PR and does it work?
Yes, it looks kind of straightforward since the functionality is already there in Espressif API.
did you test your PR
Not really. I will have to setup the toolchain (...) Do you have a label to flag the PR "on hold" for test?
tulip sleet
Yes, it looks kind of straightforward since the functionality is already there i...
when it finishes building you can download a build artifact and try it. Artifacts will show up below the "graph thing" on this page when it has finished
tired nacelle
I'm also thinking about making a clone of AudioControlSGTL5000 from Teensy as a device. It has some interesting utilities in there.
tulip sleet
just add a comment about your testing results
tired nacelle
Quick question - what's the easiest way to build in an AWS machine e.g. Ubuntu? Is there any kind-of oneliner script?
tulip sleet
Quick question - what's the easiest way to build in an AWS machine e.g. Ubuntu? ...
yep, you're on the right track, and if the artifact works for you, we can merge, and then there will be a build after that finishes in https://adafruit-circuit-python.s3.amazonaws.com/index.html?prefix=bin
tired nacelle
lol - esp-idf prerequisites is like everything... golang, java... I hope I'm wrong and it's just yum doing housekeeping
tulip sleet
there are lots but not those 🙂
hybrid dragon
you could consider proposing it for MicroPython and submit it upstream as well
Looking a bit at the issues for micropython I'm less inclined to start with their repo for trying to bring asyncio closer in line with CPython's implementation of task.
It looks like it has been discouraged a number of times for the (not so incorrect) belief that the way CPython does things for asyncio is more complex and doesn't offer value over what exists and works. I'll stay on the optimistic side that these were comments relating to prioritization efforts - and if they were offered a pull request they'd be thankful.
As far as code goes, the modasyncio implementations are near-identical, so the same changes could probably be applied to both pretty easily.
But with all that in mind I'll code it first against CircuitPython & then port it usptream to micropython. 🤷 Just to optimize my efforts.
But onto a bit of a question -- if I'm implementing some python methods in C and need it to throw an error - in this case CancelledError - do I need that error available at runtime or compile time? In other words, do I need to pull that out of the asyncio external module and implement an error in _asyncio?
I think during compile time so I'd have to define it as part of the modasyncio.c but I'm not the most familiar with circuitpython internals.
tired nacelle
when it finishes building you can download a build artifact and try it. Artifact...
When you said "finish building", I thought on my own build system. So I spend an hour to make it build in a linux VM from scratch. The instructions were quite up-to-date. With some tweaking I got it to start building. Then I got some errors like .espressif/espidf.constraints.v5.1.txt doesn't exist and later /../esp-idf/components/newlib/reent_init.c:29:6: error: 'struct _reent' has no member named '__sdidinit'.
At that point I realized that likely you didn't mean that. So I found the artifact on the build system and set it up and after some debugging of my test code, it worked pretty great.
tired nacelle
I'm also thinking about making a clone of `AudioControlSGTL5000` from [Teensy](h...
Regarding that - micropython-sgtl5000 looks pretty ok.
tulip sleet
When you said "finish building", I thought on my own build system. So I spend an...
I'm assuming you did esp-idf/install.sh, and then source esp-idf/export.sh before you did the build. Is that right? Also you mentioned yum so you were using redhat or Centos or Fedora? We've never tried building on those platforms, so we're not sure how they would work out.
tulip sleet
Regarding that - [micropython-sgtl5000](https://github.com/rdagger/micropython-s...
yes, it looks easy to modify
mortal kernel
@tulip sleet I've been successfully debugging with PicoProbe and OpenOCD and have some hints that I'd like to share. It looks like the place for them to go is BUILDING.md. Are there any special steps I need to take for this kind of doc pull request?
tulip sleet
<@329766224093249548> I've been successfully debugging with PicoProbe and OpenOC...
We tend to write up that sort of thing in Learn Guides instead of in the repo. For instance, https://learn.adafruit.com/building-circuitpython is more authoritative, complete, and up to date than what is in the repo.
Consider using the new Adafruit Playground system: https://adafruit-playground.com/
i could add a link to your playground writeup from the Building guide
mortal kernel
We tend to write up that sort of thing in Learn Guides instead of in the repo. F...
Got it. Playground looks, well, fun. Another quick Q: I'm close to a proof-of-concept pull for your "bad blocks" idea to get to FAT16. Is there anything special I need to do to make a "pull for discussion but don't merge"?
tulip sleet
Got it. Playground looks, well, fun. Another quick Q: I'm close to a proof-of-co...
you can open it as a draft PR
re README-type files: I know there is this tendency to say "everything you need to know is in the repo", but different communication modalities can work better for different purposes.
the problem is always keeping things up to date
mortal kernel
re README-type files: I know there is this tendency to say "everything you need ...
Agreed.
tulip sleet
If you run into Playground bugs (and there are definitely bugs), report them here and we'll pass them on or send to support@adafruit.com.
mortal kernel
If you run into Playground bugs (and there are definitely bugs), report them her...
Thanks for the tips!
atomic summit
Anyone tried macOS 14.2 beta 1 and 2? Any glimmer of hope for a usb resolution?
tulip sleet
Anyone tried macOS 14.2 beta 1 and 2? Any glimmer of hope for a usb resolution?
i could update to these and try. I only use macOS for testing, so no problem to upgrade.
atomic summit
i could update to these and try. I only use macOS for testing, so no problem to ...
Thanks Dan.
tired nacelle
I'm assuming you did `esp-idf/install.sh`, and then `source esp-idf/export.sh` b...
Yes, it was even more wild. It was the amazon operating system for AWS EC2. I did run those commands, yes. There's also a shortcut get_idf they suggest on their guide since you don't know the exact installation path/directories in every case. https://docs.espressif.com/projects/esp-idf/en/latest/esp32s3/get-started/linux-macos-setup.html#step-4-set-up-the-environment-variables . What OS do you use with your github build workflow? I think I see ubuntu-22.04. Ok, I will also try that just to make sure I have a nice build environment around.
tulip sleet
Yes, it was even more wild. It was the amazon operating system for AWS EC2. I di...
I use Ubuntu 22.04, and most others do. Scott uses Arch, which gets new compilers faster. Some people use WSL with Ubuntu under Windows, but it has pretty slow disk I/O, or at least it used to.
mortal kernel
Ubuntu 22.04 here, too. I also use the latest gcc 13.2 toolchain from Arm: https://developer.arm.com/downloads/-/arm-gnu-toolchain-downloads
stuck elbow
hmm, looks like asyncio is broken in 9, is that expected? it seems to stay in the main loop and never runs any tasks
tulip sleet
hmm, looks like asyncio is broken in 9, is that expected? it seems to stay in th...
we didn't know that it's broken, but it's expected there are bugs
it passes the tests.
did you update to the latest asyncio library version?
tulip sleet
Thanks Dan.
I updated to the latest Developer Beta (I am in the developer program, for testing reasons). No luck, seeing delayed writes, and sometimes I/O errors
random junco
Bummer. Thanks for testing, Dan
hybrid dragon
hmm, looks like asyncio is broken in 9, is that expected? it seems to stay in th...
Broken how?
I've been looking at it a lot lately, would be happy to help with debugging / work through problems.
I was able to run it (though not as mpy) without too much issue today with a simple example.
manic glacierBOT
atomic summit
I updated to the latest Developer Beta (I am in the developer program, for testi...
Ahh, bummer 😦 Thanks for trying though.
tulip sleet
@slender iron I am working on a 8.2.x -> main merge, and I will update the frozen modules. Maybe after that is a good time to do an alpha.3
slender iron
@tulip sleet sounds good. Looks like some qualia code doesn’t run with split heap
A couple reports of it on the forums
I asked them to file an issue because I’m not sure when I’ll have time to get to it
tulip sleet
I asked them to file an issue because I’m not sure when I’ll have time to get to...
i have a qualia board and can take a look when the issue comes up
tulip sleet
@atomic summit I am merging 8.2.x into main and I have to redo some aspects of board defs. You have two new boards, the bling and the tinywatch. a few q's and I'll have more maybe as I proceed?
Does either board have PSRAM? If so, what size, mode (dio/qio, etc.) and frequency?
CONFIG_LWIP_LOCAL_HOSTNAME="UMTinyS3" on the bling, is that right?
atomic summit
<@195765176035835904> I am merging 8.2.x into main and I have to redo some aspec...
All have PSRAM - Bling & TinyWATCH have 2MB, QIO.
That should be CONFIG_LWIP_LOCAL_HOSTNAME="UMTinyWATCHS3"
Sorry 😦
tulip sleet
That should be `CONFIG_LWIP_LOCAL_HOSTNAME="UMTinyWATCHS3"`
and UMBling for the BLing!?
manic glacierBOT
manic glacierBOT
manic glacierBOT
mortal kernel
@tulip sleet Do you think there will ever be a use case for an internal flash drive in excess of 32GB? I'm trying to decide whether to support ExFat for the internal filesystem in some virtual MBR improvements that are part of FAT16 proof of concept I'm working on.
I think this is a future-proofing issue? Supporting ExFat is certainly do-able, so now might be the time to take it on. I can fence it with FF_FS_EXFAT so it doesn't bloat tiny ports.
stuck elbow
to be honest, I don't see you ever wanting to juggle that much data with a poor little tiny microcontroller
mortal kernel
to be honest, I don't see you ever wanting to juggle that much data with a poor ...
Yeah, it's a lot of data. The trouble is that CP kinda supports ExFat for its internal filesystem already, albeit with a misreported virtual MBR.
stuck elbow
would it save space to remove that support? ;3
mortal kernel
Now that I think about it, you could hit the case where CP makes an internal ExFat filesystem on a Broadcom port with a large SDCard.
mortal kernel
would it save space to remove that support? ;3
Only at the expense of losing ExFat for SDCards.
stuck elbow
Only at the expense of losing ExFat for SDCards.
boo
tulip sleet
Now that I think about it, you could hit the case where CP makes an internal ExF...
Yes, I was going to mention that. Note that overall control for CircuitPython is via MICROPY_FATFS_EXFAT
mortal kernel
Yes, I was going to mention that. Note that overall control for CircuitPython is...
OK, ExFat it is. As it stands, the "fake", i.e. virtual MBR always reports a FAT12 partition for the internal flash filesystem. For bigger flash, like the 16M device on the RP2040 Metro, the virtual MBR returned by USB MSC is wrong and might cause problems with the host OS. I can split my new virtual MBR code into a separate pull if you like.
tulip sleet
OK, ExFat it is. As it stands, the "fake", i.e. virtual MBR always reports a FAT...
I think the only reason we have a virtual MBR is to save flash on the tiniest systems. Is it worth conditionalizing that so you can use real MBR when there's room?
are we saving just 512 bytes, or how much, with the virtual?
given your space-saving improvements on tiny filesystems, maybe the virtual is more trouble than it's worth. I don't remember if that's our change or it comes from MicroPython.
mortal kernel
Interesting thought. I'd wondered why we were making a virtual MBR to begin with. Getting rid of it would cost one 512B sector, but would save some code space plus simplify flash.c. oofatfs has code that constructs a perfectly usable MBR which we currently throw into the bit bucket.
tulip sleet
Interesting thought. I'd wondered why we were making a virtual MBR to begin with...
Yes, with the big improvement on 64kB flash systems, I'd just as soon get rid of that code. It might be something that is provoking issues with various third-party disk utilities, especially on Windows (though maybe not, who knows). And gaining back more code space on tiny systems is very good.
We really appreciate your work on all this!
mortal kernel
The down side is the transition. Let me think about how that would work.
tulip sleet
ah, yes, you mean the old filesystems are not compatible. hmmm.
mortal kernel
Any history of forcing a filesystem reformat with a new major release?
tulip sleet
Any history of forcing a filesystem reformat with a new major release?
we have changed the sizes of filesystems on certain boards, yes, with warning. In https://github.com/adafruit/circuitpython/releases/tag/8.0.0
Warning
The flash partitioning for 2MB and 4MB flash ESP32 and ESP32-C3 boards has changed since 8.0.0-beta.2.
When you load 8.0.0-beta.3 or later over beta.2 (or vice versa), on these boards, CIRCUITPY will be erased and reformatted.Warning
The flash partitioning for the Raspberry Pi Pico W has changed since 8.0.0-beta.1.
When you load 8.0.0-beta.2 or later over beta.1 (or vice versa), on the Pico W, CIRCUITPY will be erased and reformatted.
this was from beta to final for new boards. I'm not sure we have done it for existing boards.
maybe there could be some heuristic about not seeing what looks like an MBR in the zeroth cluster and using a fake one instead. On storage.erase_filesystem(), a real MBR would be added. But switching versions back and forth over the boundary would be painful.
This is something to discuss with @slender iron, and we'll discuss it internally as well.
mortal kernel
This is something to discuss with <@252717193496756235>, and we'll discuss it in...
Good deal. In the meantime, I'll think about your heuristic idea.
crimson ferry
learned about CIRCUITPY_HEAP_START_SIZE in a PR, maybe there are more new ones... are new "standard" environment variables going to be documented at https://docs.circuitpython.org/en/latest/docs/environment.html? (settings.toml file)
mortal kernel
@tulip sleet Here's a weird, possibly bad idea. The data structures in a "standard" MBR don't overlap anything but the boot code area of a VBR. That includes an ExFat VBR. So, how about putting an MBR into the VBR of an existing filesystem? The "overlaying" MBR's first partition would point to block 0, so a host OS would find the existing VBR and the existing filesystem. Maybe. A reformat would wipe this kludge and put the filesystem's VBR at block 1. Interestingly, that the MBR and VBR boot signatures are the same and that they don't overlap leads me to think this might have been designed in, although I can't find any doc that says this explicitly.
tulip sleet
<@329766224093249548> Here's a weird, possibly bad idea. The data structures in ...
I don't know enough about this to understand if it would work. Are our filesystems partitioned? What I read says that an unpartitioned filesystem (e.g. a floppy) has just a VBR. Do we already have an MBR followed by a VBR?
mortal kernel
<https://fileadmin.cs.lth.se/cs/Education/EDA385/HT09/student_doc/FinalReports/F...
Yes, we present a fake MBR for all internal flash filesystems for all ports (including the Broadcom port).
tulip sleet
and that MBR lists one partitiion, and points to a VBR?
I should just look at the code...
mortal kernel
Any external filesystem gets a real MBR courtesy of oofatfs. Yes, the fake MBR points to the VBR.
tulip sleet
supervisor/shared/flash.c flash_read_blocks(). hmm, the MBR always says the first partition is FAT12. Why does this even work for larger CIRCUITPY's >32MB?
mortal kernel
`supervisor/shared/flash.c` `flash_read_blocks()`. hmm, the MBR _always_ says th...
Good question. By the way, that's the root cause of failure to mount on Android.
tulip sleet
just because Android doesn't support FAT12 at all?
mortal kernel
just because Android doesn't support FAT12 at all?
Weirdly it supports FAT12. It just doesn't like the FAT12 partition type.
tulip sleet
Weirdly it supports FAT12. It just doesn't like the FAT12 partition type.
so, from https://en.wikipedia.org/wiki/Partition_type, it does support part type 04 and up?
mortal kernel
I did try other partition types, and yes it does. The rub is that if I put a FAT12 in another partition type it works on MacOS and Windows. My Ubuntu 22.04 blows chunks though, to the extent that a reboot is necessary to get automount working again.
OK, it now occurs to me that the answer could be to simply elide the MBR on ports so small they can only support FAT12.
Here's my current code for determining the correct partition type:
// Determine filesystem's partition type
// Note: Expects only FAT12, FAT16, or FAT32. No EXFAT support.
uint8_t fs_fat_bits = (fs_clusters < 4085)? 12 : (fs_clusters < 65525)? 16 : 32;
if (fs_heads_per_cylinder > 255 || fs_sectors_per_track > 63 || fs_sectors >= 65536) {
if (fs_fat_bits != 32) {
part_type = 0x0E; // FAT16 (LBA, >=32MB)
} else {
part_type = 0x0C; // FAT32 (LBA)
}
} else if (fs_fat_bits == 12 && fs_sectors < 65536) {
part_type = 0x01; // FAT12 (CHS)
} else if (fs_fat_bits == 16 && fs_sectors < 65536) {
part_type = 0x04; // FAT16 (CHS, <32MB)
} else if (fs_fat_bits != 32 && fs_sectors < fs_sectors_per_track * fs_heads_per_cylinder * 1024) {
part_type = 0x06; // Big FAT16 or FAT12 (CHS)
} else if (fs_fat_bits != 32) {
part_type = 0x0E; // Big FAT16 (LBA, >=32MB)
} else {
part_type = 0x0C; // FAT32 (LBA)
}
i have to go out, sorry. Do you mean elide the MBR in favor of the VBR? Is this forward-compatible -- I'm just considering if you flip back and forth between 8.x.x and 9.x.x, with or without reformatting in 9.x.x with the new format.
mortal kernel
i have to go out, sorry. Do you mean elide the MBR in favor of the VBR? Is this ...
No problem, get back to me when you can. If we elide the MBR is would be both forward and backward compatible. Pre-9 would continue to fake the MBR, 9 would elide it without altering the underlying filesystem. For filesystems larger than FAT12, we'd continue to present a fake MBR using logic similar to above for 9 and above.
tulip sleet
that would be really interesting to test on all these platforms
so the fake MBR would not use partition type 01h, but something appropraite, is that right?
eventually we could drop the fake MBR, but break backward compatibility
mortal kernel
I'm tooled up here to test the various host OSs. Elision is any easy change, let me try that first and get back to you.
tulip sleet
we may be willing to have a break at 9.0.0 in any case. there are plenty of other things that are changing. The cleanest thing to do is to present a proper real MBR, I guess, though 01h still breaks Android, so making small filesystems work on both Android and Linux does not seem possible, unless we always do the fake FAT16
maybe fake vs real is not such a big deal, it's more making it more accurate. It does reduce the code size a bit, but only a bit
mortal kernel
I can't help thinking (hoping) that there's a magic combination of MBR partition type or MBR elision and FAT12 that will work for everything, call it the "unified field theory of partition IDs".
tulip sleet
unfortunately programming is more like law than physics
mortal kernel
They don't call some of us "language lawyers" for nothing.
mortal kernel
@tulip sleet Eliding the MBR didn't work out, but I found a partition type that does! When I initially tested alternative partition types, I didn't try 0x06 (Big FAT12/16 CHS). Gave it a try, and it is the magic number that works on Android, MacOS, Linux, and Windows.
tulip sleet
<@329766224093249548> Eliding the MBR didn't work out, but I found a partition t...
Great! If you can test on iOS that would be great, else I can.
mortal kernel
I lack a USB adapter for my iPhone, so please do.
I'm going to put the virtual FAT16 work aside and go back to my RP2 bluetooth work. Funny how sometimes a tiny fix can resolve a big issue.
tulip sleet
I lack a USB adapter for my iPhone, so please do.
tested on an iPad running OS 17. Seems fine, works same before and after. On a Samsung Android 11 tablet, both a Trinket M0 (64kB) and a CIrcuit Playground Bluefruit work fine with the fix. They identify as "USB Storage 1", not CIRCUITPY, in the Files app.
mortal kernel
tested on an iPad running OS 17. Seems fine, works same before and after. On a S...
Shows up as CIRCUITPY on Android 10. I don't think it's an issue.
tulip sleet
may be some peculiarity of the Files app in Samsung Android
I wonder if this improves the Sonoma situation, or it's still an "8MB and under" issue
mortal kernel
Tried it, no joy. Sonoma that is.
That's the only reason I can see to keep on with the fake FAT16. But even if that did work, it's not a complete fix. All the loaders would need to be reworked, too.
tulip sleet
It's really an Apple bug that needs fixing, we shouldn't have to work around. ... Did you ever get anyone's attention there?
mortal kernel
Tried through an old contact in Dev support, but no reply. All my old droogs in CoreOS have left.
tulip sleet
thanks for trying, in any case. They did eventually fix the Ventura UF2 problem. RPi was banging the drum on that case, so it was more public. Not sure how much this affects their general Pi Pico user population. The RPI-RP2 fake drive appears as a much larger drive, so it's not incredibly affected, but it might still do some write delays on that. I haven't looked carefully.
mortal kernel
I've been checking up on the feedback issue I opened. Still ignored, and still showing no similar reports despite others cross referencing it.
How about direct action? I could load up the car with some 5 gallon plastic pails and some past-their-pull-date hippies to drum them and stage a protest outside their new HQ.
stuck elbow
How about direct action? I could load up the car with some 5 gallon plastic pail...
tell them to fix MTP once you are at it, that would solve all filesystem problems once and for all
and we could use something a bit more modern than fat then
keen eagle
The Sonoma bug also affects more than just CircuitPython devices; I'd hope that other communities might also try to push back on them.