can we rerun the action maybe ?
#circuitpython-dev
jaunty juniper
those missing assets are gonna make circup fail too
06/24/2023 16:45:08 WARNING: Unable to connect to https://github.com/adafruit/Adafruit_CircuitPython_Bundle/releases/download/20230624/adafruit-circuitpython-bundle-py-20230624.zip
(by the way I just realized that my circup.log is 1.47 GB)
do you think it could be a change in the "shogo82148/actions-upload-release-asset" action ? It doesn't look like there was one between yesterday and today. Want to try rerunning it ?
oh wait that was already rerun
so maybe we should delete the release, remake it or make a new release
lone axle
Yeah, it looked to me like the re-run failed for a different reason, because some of the files already exist. Although oddly it looked like one of the one it said already existed is the one from the XML file not found error if I understood them correctly.
I'm wondering if the root cause was just something transient and tomorrows will work fine (or whenever the next one builds. Maybe we should merge something today in order to ensure there is one tomorrow)
jaunty juniper
I think the it's the assets from the github release that "already exist", those ones ?
lone axle
Ohhhh, it was saying they exist in the release assets already, not in AWS.
jaunty juniper
I think we can delete the release and wait for tomorrow ? that should fix circup
though I'm not ready to delete a release/tag in github...
lone axle
I do think those are the ones. I don't know how it works but possible if those assets were deleted and then the action re-run maybe it could succeed.
Yeah, same. Not sure what other ramifications of deleting the release entirely, could cause more trouble than fix.
jaunty juniper
there's a little trashcan icon, so tempting 😅
slender iron
The boot drive won’t disappear if it doesn’t get all of the data. I’ve never seen an actual failure
manic glacierBOT
lone axle
I'm attempting to build a very old commit in order to track down a bug that I've found. But I'm getting this error about no make target for protomatter:
circuitpython/ports/atmel-samd (37e85ae) [!?] via C v11.3.0-gcc
❯ make BOARD=pyportal
Use make V=1, make V=2 or set BUILD_VERBOSE similarly in your environment to increase build verbosity.
make: *** No rule to make target 'lib/protomatter/core.c', needed by 'build-pyportal/genhdr/qstr.i.last'. Stop.
My best guess is that my submodules are not "sync'd" properly in order to make the build on this random old commit. I tried getting a fresh clone and doing make fetch-all-submodules, then checking out this commit and running the build and it outputs the above error.
Is there some other process I need to use in order for making builds from very old commits to get my submodules set to the proper revision in order to work?
blissful pollen
I'm attempting to build a very old commit in order to track down a bug that I've...
Assuming you have done make clean. It does seem like the makefile has reference to core.c but it doesn't exist.
When I've had issues I've removed the folder with the submodule and just re-synced/updated it
lone axle
I think that did end up being the fix for me (remove and re-sync) But it was a little more convoluted because it turned out the commit I was at was from before make fetch-submodules existed, so I had to track down the old way to do that (Hug report is in order Tyeth for help with that.)
it also turned out that commit was way older than where I needed to be looking to find the issue I was looking into though so I was able to "jump forward" pretty far in time and didn't have any more submodule weirdness.
pallid creek
It looks like circup is still failling because of the broken release 13hrs ago. Is there a way to tell circup to use the previous release instead?
jaunty juniper
we could still try removing the latest release, it can't make things worse, can it ?
pallid creek
I wouldn't think so.
lone sandalBOT
manic glacierBOT
tulip sleet
@lone axle @jaunty juniper I am trying to re-run the bundle build. I deleted today's release.
jaunty juniper
ah you did 👍
tulip sleet
i made a change yesterday in the assigned user for something else (updating circuitpython-org pages as adafruit-adabot instead of blank, which was registering as ladyada), but I didn't think it would affect this. The
build failed again in same way.
I will undo my previous change, jsut in case. It was only cosmetic (Limor sees way more commits by her than she would like)
hmm, it wasn't really blank: I can't set a blank one. I'll set it to myself and see what happens.
running yet again
this run is without deletion, so I will probably get a different error. Then I will delete the release and try agian.
... release is still gone, failed the same way, trying some other actions for bundle
pallid creek
whatever you did seems to have fixed circup. I can now list my local modules
tulip sleet
whatever you did seems to have fixed circup. I can now list my local modules
i removed today's bundle
@idle owl I was going to re-run the regular job. Should I re-run with your push?
some discussion above
i am going to be afk in less than an hour
idle owl
<@330227457296957440> I was going to re-run the regular job. Should I re-run wit...
I've been working on this for the last hour. I don't think we can manually trigger it with the push. I was going to add a workflow_dispatch which makes it possible to trigger a new job.
Oh you did do someting somewhere.
Regardless, that TOKEN doesn't exist anywhere, so I'm not sure how it ever worked without the change I made.
tulip sleet
all I did was set the circuitpython-org user name to adafruit-adabot; I don't really think it has to do with this.
I changed it to myself just in case that affected it, but it's not even used in this repo
idle owl
That sounds unrelated.
tulip sleet
if it doesh't work, you can just delete the bundle release and tag, and yesterday's bundle works.
I did that once already.
idle owl
Hold on trying to add a thing and not figuring out how
@tulip sleet Wondering if I have to give workflow_dispatch parameters if there is something else in the on: section. I can give it a specific branch easily which is fine because we want it run against main anyway. If that doesn't work, I'll toss it to you to get it to trigger another run that actually uses the updated file. tl;dr: If this build fails, I have one more thing to try, and if that fails, I'm returning the file to my original push.
@tulip sleet It worked! You can now manually trigger the Release Actions run from the GitHub UI. Triggered a new run. (Triggering a run on the previous job was not using the updated release.yml file.)
Well, new failure this time. Rerunning with debug enabled.
Might be the TOKEN change.
lone axle
I'm not sure it's actually related to the root cause anymore but I did find that those bracketed template things at the end of the URL printed in debug are for a "Hypermedia" url: https://docs.github.com/en/rest/overview/resources-in-the-rest-api?apiVersion=2022-11-28#hypermedia
idle owl
Hmm.
I don't really understand what all of that means.
(It failed sooner than before, so I'll swap the token back to what it was once this fails, if it seems like that's probably the issue now.)
This all seems like we're going after red herrings because the failure is obtuse.
This is the result of generating the upload_url with the updated token: ##[debug]..=> null ##[debug]=> null ##[debug]Result: null
Reverting that.
tulip sleet
i'd say switch it back; in the morning I'll see if the new bundle fails, and if so, I'll just delete it again, same Monday morning, and we'll get other eyes on it on Monday
idle owl
Though I wish I understood how it worked.
lone axle
It's a shorthand notation for URLs that can get specific fields populated with variables by some system. I think in our case they aren't problematic, the log printout doesn't seem to be able to handle it so well but the underlying system should be fine with it if I am understanding.
idle owl
It's a shorthand notation for URLs that can get specific fields populated with v...
Ok that makes sense.
idle owl
i'd say switch it back; in the morning I'll see if the new bundle fails, and if ...
Ok, this sounds like a better plan than continuing to go in circles.
tulip sleet
no vital bundle updates
idle owl
Ugh, I wanted to run it with debug. Can only do that on rerun apparently.
idle owl
no vital bundle updates
That was going to be my point in another reply to the failure emails, to the team.
@lone axle Thanks so much for looking into this with me. If nothing else, I now know a few more things about how this works.
@tulip sleet I saw you had triggered one of the runs, I should have messaged you. But I think you were idle, so I wrote it off as maybe you didn't trigger it and I didn't see it right. Parallel debugging happening here.
tulip sleet
i was in this channel with neradoc but then went away and saw it didn't work, but no problem, because older bundle was fine
once I deleted the partial new bundle
idle owl
I messaged Tekktrik about it. He's way more up on CI shenanigans than I am. He's not around right now though. If he can get eyes on it at some point, he might have some insight we didn't think of.
@tulip sleet Failed. On the new failure. Uh oh. I have no debug to know if it used the new file or not. Rerunning one more time with debug to check that.
Then I swear I'm done.
tulip sleet
then you'll revert?
idle owl
Already did!
But it now has the failure it did after I made the change.
So I'm not convinced it is using the reverted file.
That's why enabling debug logging.
It told me before when it wasn't using the new file.
tulip sleet
where did you make the change? I see no PR or revert PR in Bundle repo?
idle owl
where did you make the change? I see no PR or revert PR in Bundle repo?
I pushed the change to release.yml
Where I made it in the first place.
Directly. Did not PR.
@tulip sleet It is using the reverted file. So... at this point, the only change from before is including the workflow_dispatch which is completely unrelated to any of the rest of the Actions file. What I'm saying is, there are ostensibly no changes, and now it's not generating the upload_url, which it was doing earlier today.
(╯°□°)╯︵ ┻━┻
I'm done with this until Alec can take a look at it. Or until Monday. Whichever is soonest.
tulip sleet
let's just wait and see what happens tonight ... yeah
idle owl
I'm on board with this.
timber ivy
Howdy 👋 - I am trying to add support for the M5Stack Stick C Plus. The M5Stack Stick C (not Plus) firmware, when loaded on the Plus, gives a garbled display and I think might not be booting all the way. I spent a few hours running down cross-compiler stuff but I'm hung up on:
"/usr/bin/xtensa-esp32-elf-gcc" is not able to compile a simple test program.
once I use make BOARD=m5stack_stick_c_plus (referencing a new directory I made with a few tweaks to the stick c directory)
Is this channel the right place to seek advice on getting my toolchain working?
jaunty juniper
did you follow the guide to build CP ?
timber ivy
yes, this was after I got mpy-cross going
jaunty juniper
including that ? https://learn.adafruit.com/building-circuitpython/espressif-build
(what platform are you on ?)
timber ivy
I did not see that. I don't use Linux daily, but we have a gaming laptop dual-boots Manjaro, so that's what I'm using. Aside from having to compile xtensa-esp32-elf-bootstrap and xtensa-esp32-elf-gcc from AUR, most of that guide looks like what I fumbled through to get to where I'm at.
I didn't do the install.sh on the espressif ports dir, looking at that now, and any other hints on this page.
jaunty juniper
you shouldn't need to compile or do anything for gcc
esp-idf will download all you need to ~/.espressif includding all the binaries for gcc
plus its own python venv (in which you need to install the requirements-dev.txt)
also I gotta sleep, but more people will be there tomorrow in a few hours
(or monday for sure)
timber ivy
I think my problem was I started immediately trying to fight my way through errors after make BOARD (oh gosh, gotta install ninja, oh gosh, gotta install xtensa-esp32-elf-gcc) and never found this espressif-build guide. Thanks again!
Success! Cheers
I'm un-blocked. I'll try actually loading the firmware tomorrow. It is also late here.
lone axle
At first glance it looks like todays auto-release for the library bundle did succeed.
tulip sleet
At first glance it looks like todays auto-release for the library bundle did suc...
Yes, it looks ok: it has all the files I expect in the GitHub release assets, and when I use circup it downloads the new bundles just fine. tagging @idle owl
i have no idea what that hiccup was
lone sandalBOT
tulip sleet
<@&356864093652516868> Reminder of our weekly discord meeting, in about 1.5 hours, at 2pm US Eastern time. Add your notes to https://docs.google.com/document/d/1VFqasfzcuR1nmpfB-IxASRSf9isPpRWfG1O3ZAUnp8Y/edit?usp=sharing
proven garnet
@tulip sleet @idle owl sorry I was camping this weekend, I heard the CI went wonky, it seems like everything is fixed now?
tulip sleet
<@329766224093249548> <@330227457296957440> sorry I was camping this weekend, I ...
yes, weird failure, including on re-runs, but the next night's one was fine!
proven garnet
Sweet! Glad to hear!
slender iron
@onyx hinge yup, single buffer mode is basically what I was thinking but it'd DMA two different addresses instead of the same one
idle owl
cpnews@adafruit.com is best as there will be a guest editor in July
It me!
idle owl
Neither am I.
onyx hinge
[background: pipewire is the newer audio library/api/server for linux and many linux distros are in the process of replacing the old software, pulseaudio]
idle owl
@onyx hinge Wait what? USB A doesn't fit?
idle owl
I have a bare usb A to C adapter here, and a NeoKey BFF. Holding the entire A connector over the BFF and it fits. Obviously that doesn't take into account connecting the pads up...
Hmm.
idle owl
Should be.
This fits.
Which has pins coming out the back on both sides of the little PCB in there.
tulip sleet
is .6 the right spacing between the inner pin rows?
I measured with a ruler, yep 0.6:
idle owl
@onyx hinge Mostly wondering if there's another USB footprint part to use.
I will be out. I don't think moving it to Wednesday makes sense.
Also an option.
That's what I would suggest.
devout jolt
Thanks @tulip sleet !
gilded cradle
Thanks everyone
idle owl
@onyx hinge Can you also remove it from the hosting calendar?
lone axle
Thanks for hosting Dan! Have a good week everyone.
idle owl
Thanks everyone!
onyx hinge
@idle owl yes
slender iron
thanks!
onyx hinge
if I have permissions anyway
idle owl
<@330227457296957440> yes
Thank you! You have perms with your Adafruit account.
crimson ferry
@slender iron right before the meeting recording started, I wasn't fully tuned in, but you mentioned me and someone else... maybe some technical issue with my connection to the meeting? It looked fine and muted at my end.
slender iron
@crimson ferry just how discord hides your username when in streaming mode
onyx hinge
@timber mango you were scheduled to host the meeting on july 3, note that we canceled it so you get a pass.
idle owl
@onyx hinge Wrong person.
onyx hinge
oops sorry!
@candid sun you were scheduled to host the meeting on july 3, note that we canceled it so you get a pass.
candid sun
@onyx hinge ah okay, thanks!
tulip sleet
Here is the notes document for the next CircuitPython Weekly meeting, which will be on July 10, 2023. ***Note that we are skipping the week of July 3 due to a US holiday. *** The meeting will be 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/1ajdwXoubhG86rUVabqTcEzHvOKfUgK9TgPdDxqBDlSc/edit?usp=sharing
slender iron
for usb host on rp2040 I've run into the issue that making a non-default alarm pool (in order to run on core 1) requires malloc
I could 1) change pico-sdk to allow for a statically allocated pool or 2) remove the need for alarm pool in pio usb
tulip sleet
I could 1) change pico-sdk to allow for a statically allocated pool or 2) remove...
A change that does not require modifying pico-sdk (potentially forever) would be nice. They might take simple PR.
slender iron
I've also gotta see how many changes I have for placing stuff in RAM
I'm leaning 2 which changes the PIO USB library
skipping the alarm stuff simplifies all of that I think
orchid basinBOT
manic glacierBOT
slender iron
@lone axle need me to look at your esp32spi changes?
lone axle
<@382939733107408897> need me to look at your esp32spi changes?
That would be great! I do think it's ready for a look by someone else.
Thank you
slender iron
@lone axle what order should I look at them in?
lone axle
<@382939733107408897> what order should I look at them in?
This is the root one:
https://github.com/adafruit/Adafruit_CircuitPython_ESP32SPI/pull/167
These ones require the root one to be in place in order to test / evaluate:
https://github.com/adafruit/Adafruit_CircuitPython_Requests/pull/135
https://github.com/adafruit/Adafruit_CircuitPython_WSGI/pull/20
slender iron
👍
lone axle
This one deals with the same concept of the socket API: https://github.com/adafruit/Adafruit_CircuitPython_Wiznet5k/pull/125 but is not independent from the ESP32SPI once since it's for ethernet.
slender iron
@tulip sleet looks like 167 is your PR. Do you want to circle back to it or pass it off to foamyguy and I?
@lone axle is there a way we could change requests and wsgi first so that we don't need to do all three at once?
tulip sleet
@lone axle I was just going to drop the WSGI server from the ESP32SPI library. If you move any additional functionality to anothe repo, that's fine, but at this point maybe HTTPServer is sufficiently functional?
Are there any changes to be made in #167? I'm not sure I can reassign it to you, but you could push changes to it.
slender iron
foamyguy did already
lone axle
<@382939733107408897> I was just going to drop the WSGI server from the ESP32SPI...
Unfortunately HTTPServer cannot be used with ESP32SPI because the latter doesn't implement the bind() function on the socket, which HTTPServer uses. So WSGI library is the only way that I know of for any sort of http server functionality with ESP32SPI.
I have committed all changes that I was able to find that it needed to the PR branch.
tulip sleet
ok, now I remember. I kind of remember wondering whether we could a have dummy bind(), but it seemed tricky. It's fine to me to merge my PR when you are ready.
slender iron
Should we wait to do this switch until after the 4th of July to make sure we're around to fix things that come up?
tulip sleet
i am not going away
lone axle
<@382939733107408897> is there a way we could change requests and wsgi first so ...
I'm not 100% certain, but I think if requests is merged without ESP32SPI then ESP32SPI based devices will not be able to successfully use the new version of requests, they'd have to keep using the prior release.
Same for WSGI library I think. If that were merged I'm not sure if it would function with the older version of ESP32SPI.
tulip sleet
so do all of them in one bundle release, right, that's what your proposing?
lone axle
Yep, I think that would be ideal. That way when someone does get get the newer ones installed via default means they'll be getting ones that co-exist and function together.
It wouldn't be the end of the world if it's not done in the same release, but it would mean some folks may need to track down older releases of one or the other of the libraries in order to get things running.
tulip sleet
we have to do it some time. You've done a lot of testing, so I think we're ready. Then we also need to make Guide changes of some kind, right?
for doc purposes
lone axle
Possibly, it will depend what all is covered in the guides. If there is anything that covers the inner workings of ESP32SPI library, it would probably need some changes to reflect the new functions and behaviors that it has now.
But many / most of the guides are built on higher level libraries like adafruit_requests or adafruit_pyportal which abstract away the details anyhow, so project guides using these higher level libraries may not have anything to change.
I tested a small handful of the projects based on those higher level libraries like requests, and pyportal and all the ones I found were working as expected with the new versions. They didn't require any change in the actual project code. I didn't check all of the guide pages for the ones I tested though, it's possible there may have been explanations in there that could need tweaked, though I think it's likely many of them don't go down into the details enough to have info that would need changed.
jaunty juniper
which ESP32SPI boards have it frozen in, as well as the requests library (so compatible versions) ?
tulip sleet
matrixportal, I think
lone axle
Good thought. I'm pretty sure pyportal does have both frozen in as well.
tulip sleet
pyportal, pyportal titano, matrixportal have requests and ESP32SPI frozen. lots more too but they are Espressif
lone axle
Actually come to think of it, the fact that those are frozen in to many of the devices they're used on means that the bundle being all updated at once is perhaps a little less critical. The vast majority of people will just be using the frozen in versions rather than installing them anyhow.
jaunty juniper
did you want to wait for that update before 8.2.0 so we can update the frozen libraries ?
slender iron
I just approved all three PRs but didn't merge
tulip sleet
it's way too much churn to update the frozen libs in the rc
i was going to make rc.1 today with the few new changes since Friday. i could do 8.2.0 final Wednesday or Thursday morning, but could also wait until next Wednesday
@onyx hinge do you want the ulab update to be in 8.2.0? Does it fix significant bugs?
onyx hinge
@tulip sleet no, I don't think it's critical
if I'd thought of doing it a week ago I'd sing a different song
tulip sleet
the shape.set_boundary() fix, the no-screech rp2040, and a minor board fix are all interesting for 8.2.0 final and are merged. There isn't an unmerged PR left that is significant, then. The zlib doc fix does not need to be in 8.2.0 final.
I will make an rc.1 with the current set of merged PR's, and do 8.2.0 final tomw as a new name for rc.1, barring new issues.
I'll start on release notes, should be quick
slender iron
thanks!
onyx hinge
thank you for release-wrangling, Dan!
orchid basinBOT
idle owl
@onyx hinge Do you have a few minutes to go through a piece of code with me to help me explain it in the guide? It involves seesaw firmware if that matters. Basically I don't get button masking well enough to explain it. Or really at all.
onyx hinge
@idle owl I think I have to leave in about 10 minutes so unfortunately not
idle owl
Ok, no worries. Thanks, anyway.
idle owl
<@426200547024961539> Do you have a few minutes to go through a piece of code wi...
@tulip sleet Same question, since you appeared. 😄
Looking elsewhere also. No worries.
idle owl
@tidal kiln Hello! So... it has been requested that I actually explain what's going on in the code in the guide containing this demo, because Anne has run into multiple support issues with this type of code before because it's usually glossed over. I have a starter explanation that Anne wrote up, but she said verify it, and I don't understand any of the button parts well enough to verify anything. I do get the joystick code. But not the button code. https://github.com/adafruit/Adafruit_CircuitPython_seesaw/blob/main/examples/seesaw_gamepad_qt.py
idle owl
You set up variables that are set to the seesaw firmware pin they are attached to. Then.... button mask is where I get lost.
tidal kiln
first - that approach is only needed when using the "bulk" methods
tidal kiln
it could be done without a mask, but then would require multiple reads, each being I2C traffic, to read all the pins
idle owl
Hmm ok
tidal kiln
so it's not like it's a required all the time sort of things
idle owl
Ok got it
tidal kiln
it just helps speed things up, since it can get multiple pin states in a single I2C xfer
idle owl
Ah ok yeah
tidal kiln
if you did it without mask and separate calls, under the hood it'd be reading the same register each time, but then only returning the pin of interest - the masking stuff would be happening inside the firmware running on the seesaw chip
with the "bulk" access, the seesaw chip just returns the full register value and leaves it up to user code to deal with
idle owl
Ohhhh. So it does the work, and only speaks up when something important happens?
Wait no.
Oh! Ok, I think I get what's happening with bulk. The full value would be different with one or more buttons pressed, and the code is written to be aware of that?
tidal kiln
yah. easiest to think of just 8 pins as an example.
tidal kiln
digital pins are either high/low so a single pin state can be represented by a single bit 0 or 1
and a single byte register could then hold the pin state for 8 pins, with each bit representing 8 different pins
and for 32bit MCU's, it's same deal, just 4 bytes / 32 bits
so the non-bulk would
(1) read register PIN_STATE (which has the state of 8 pins)
(2) do some mask foo stuff to figure out if pin X is 0 or 1
(3) return true/false depending
so if you wanted pin Y, it'd read the same register again and do the same thing
and the I2C traffic could end up being a bottle neck
idle owl
Ahhhhh ok
tidal kiln
you'd use this call to do things without a mask:
https://github.com/adafruit/Adafruit_CircuitPython_seesaw/blob/d99b564ce5dc1784ec22e87626eea59235089916/adafruit_seesaw/seesaw.py#L214-L218
idle owl
Oh ok
tidal kiln
the bulk call is using this seesaw interface call:
https://learn.adafruit.com/adafruit-seesaw-atsamd09-breakout/gpio
idle owl
Fair enough
tidal kiln
it reads the 0x04 register - which, per the seesaw spec, is a 4 byte return
so after this call:
buttons = seesaw.digital_read_bulk(button_mask)
think of buttons as a string of 1's and 0's representing the pin state of 32 digital io pins
idle owl
Ok.....
tidal kiln
the masking is how to check for a given pin's 1 or 0
tidal kiln
yep. there's room, per the seesaw spec, for 32 pins.
idle owl
Ahhh ok
devout jolt
print(bin(buttons)) I find helpful after getting digital_read_builk()
tidal kiln
but what it actually is can be port specific - note the separate sections for SAMD and ATTINY with specific details
idle owl
but what it actually is can be port specific - note the separate sections for SA...
I did see that.
tidal kiln
and then the other thing to keep in mind is that for the ifs, 0 is False and anything else is True
tidal kiln
thats how you can just logical & with the mask and that's enough
idle owl
Because it will only be 0 or 1?
tidal kiln
nope. it'll be 0 or something other than 0
for example, let's say it's button 4 you're after
idle owl
OK
tidal kiln
bit 3 would be what you'd look at
idle owl
Ok
idle owl
`print(bin(buttons))` I find helpful after getting `digital_read_builk()`
This is really neat!
tidal kiln
0b00000000 # nothing pressed
0b00001000 # button 4 pressed
That's something other than 0.
tidal kiln
>>> mask = 0b00001001
>>> button = 0b00001000
>>> if button & mask:
... print("button pressed")
...
button pressed
>>> print(button & mask)
8
that would be a mask where buttons 4 AND 1 are of interest
idle owl
Ok I was confused there for a moment.
tidal kiln
but the button value itself is only showing button 4 is pressed
idle owl
OH! print(bin(buttons)) suddenly makes sense.
tidal kiln
but the logic was true, and the actual value it dealt with was 8
idle owl
The gamepad is 16 bits right? Because the Gamepad buttons are on pins between 0 and 16 inclusive. So, when no buttons are pressed, the mask looks like this becuase 0, 1, 2, 5, 6, 16 are of interest? 0b10000000001100111
And then the associated button is pressed, the 1 in those places become 0?
(at least that's what it's doing)
Sorry if I'm getting ahead of myself here. It clicked when you explained your example mask there.
I was confused before that moment.
about the value I posted above.
idle owl
Ahhhhhhhh
tidal kiln
the gamepad just isn't using any pins above 16
idle owl
Ok got it
tidal kiln
the simple example above would have also printed button pressed if button 1 was pressed instead of 4
so the example has:
if not buttons & (1 << BUTTON_X):
Oh hmm.
tidal kiln
to check for specific buttons
yep. and binary format print in python if you're at a repl is also helpful.
>>> BUTTON_X = const(6)
>>> BUTTON_Y = const(2)
>>> BUTTON_A = const(5)
>>> BUTTON_B = const(1)
>>> BUTTON_SELECT = const(0)
>>> BUTTON_START = const(16)
>>> button_mask = const(
... (1 << BUTTON_X)
... | (1 << BUTTON_Y)
... | (1 << BUTTON_A)
... | (1 << BUTTON_B)
... | (1 << BUTTON_SELECT)
... | (1 << BUTTON_START)
... )
>>> print("0b{:016b}".format(button_mask))
0b10000000001100111
>>>
can think of this as being another mask: (1 << BUTTON_X)
being created in place
using the << left shift operator
>>> print("0b{:016b}".format(1 << BUTTON_X))
0b0000000001000000
I printed (bin(buttons)) under the buttons = seesaw.digital_read_bulk(button_mask) call, which printed the same thing.
idle owl
```python
>>> print("0b{:016b}".format(1 << BUTTON_X))
0b0000000001000000
```
Because x is in the 5th bit spot?
tidal kiln
hey! that's new to me 🙂 knew about hex() figures theres also a bin()
idle owl
Or whatever the right word to use there.
tidal kiln
the_more_you_know.gif
idle owl
hey! that's new to me 🙂 knew about `hex()` figures theres also a `bin()`
That was Tod's suggestion 😉
tidal kiln
keen
idle owl
When printed there, it updates when you press the buttons. It's nice.
Once I understood the base value anyway.
tidal kiln
masks are general - don't always have to be for a single bit
idle owl
Ok
tidal kiln
but in the case of checking for a specific pin, they are. that is, all 0s except for the bit with the pin of interest
idle owl
Ok
tidal kiln
i guess it could be argued that the example could be written without using masks, to be more beginner friendly readable? or somehow show both, since there are those that would want to also see this approach. multiple examples?
idle owl
Eh...?
tidal kiln
for example:
if not digitalRead(BUTTON_X):
print("Button x pressed")
if not digitalRead(BUTTON_Y):
print("Button Y pressed")
if not digitalRead(BUTTON_A):
print("Button A pressed")
if not digitalRead(BUTTON_B):
print("Button B pressed")
if not digitalRead(BUTTON_SELECT):
print("Button Select pressed")
if not digitalRead(BUTTON_START):
print("Button Start pressed")
We use this in all of the guides for things using seesaw
I am unaware of us showing the other way.
tidal kiln
it's a tradeoff. the code above would just generate other feedback, since each digitalRead() call is I2C traffic.
idle owl
Agreed.
So I'd rather explain the existing code.
Anne's point is the seesaw firmware pins are magic numbers to people not In The Know, and then that leads to the rest of the issues with the code. In terms of support problems.
So is it ok to refer to a 16 bit value when referencing the button_mask creation? Or should I refer to 32 bit, and explain that this code isn't using anything over 16.
That the mask can be a subset.
tidal kiln
this is guide specific to gamepad?
manic glacierBOT
I do have one. The build runs Hello World. Photo of screen attached. It is not showing up as a circuitpy drive and I haven't gotten REPL to work via USB. I can list files over USB with adafruit-ampy, which I'm not too familiar with but found during my troubleshooting. Ampy get fails on module ubinascii. Additional guidance is welcome. I'm in the discord channel.
tidal kiln
hmm...maybe dont say one way or the other. just call it a mask.
idle owl
Ok
tidal kiln
an attiny seesaw has the potential to return more:
https://learn.adafruit.com/adafruit-seesaw-atsamd09-breakout/gpio#gpio-register-setup-on-attiny8x7-3111799
so don't want to give too much importance it being 16 bit
idle owl
Got it
Can I ping you after I try to write this up? So you can tell me where I didn't quite get it?
tidal kiln
sure
idle owl
Thanks. Will do.
tidal kiln
and hopefully explanation above helped?
idle owl
Hoping so. We'll find out!
idle owl
@tidal kiln DMing you the preview link. Hopefully this is good enough. Hoping to get it back to Anne today.
timber ivy
This is my open PR. Does anyone have an original M5Stack Stick C (not plus) to test the latest stable CircuitPython on? Specifically:
- Does REPL work over USB serial?
- Does it show up as a mass storage circuitpy drive?
jaunty juniper
This is my open PR. Does anyone have an original M5Stack Stick C (not plus) to ...
it's an ESP32 (no suffix), it cannot do mass storage.
if ampy works to list files, the REPL works, how did you try to connect to it ?
ampy needs to be changed to use binascii instead of ubinascii, I don't know if there's a version somewhere that does that, I know there was a PR to make it fallback to it but 🤷
thonny should work, our official solution is the web workflow
manic glacierBOT
timber ivy
if ampy works to list files, the REPL works, how did you try to connect to it ?
Screen on Manjaro Linux (I get some garbled output), cu on OpenBSD which is my actual primary workstation (it sees the USB serial port but it's not responsive). Both of these work great with my AdaFruit Circuit Playground Express. I need to try to figure out the web flow.
timber ivy
We got signal. Needed to manually set baud rate. And the web esp tool works.
empty stirrup
[just curious, I was trawling the github repo] and couldn't figure out how code that crossed the python/C boundaries was being tested (e.g. bitbangio.I2C). Is there some magic workflow that's testing these in some other way?
manic glacierBOT
desert tartan
What trips Halfkay mode in Circuitpython8.0 on Teensy 4.0
tulip sleet
[just curious, I was trawling the github repo] and couldn't figure out how code ...
it's tested by hand. Several years ago we were working on an automated hardware test bed, but there were many stumbling blocks. We would really like such a thing to catch regressions, but it's a lot of work.
empty stirrup
Makes sense. Btw, it looks like the circuitpython support in wokwi is new, but is something like it (https://github.com/wokwi/wokwi-ci-action/network/dependents) an option you had looked at before?
tulip sleet
Makes sense. Btw, it looks like the circuitpython support in wokwi is new, but ...
the bugs we encounter would typically not be diagnosed in a simulator. They often relate to the actual hardware implementations of various peripherals (I2C, etc.) or system-on-chip things (memory management, etc.)
empty stirrup
got it - thanks for the insights!
jaunty juniper
was it Tekktrik who looked into using wokwi for things like library tests ?
tulip sleet
maybe - I don't remember
slender iron
I think it was
devout jolt
I use wokwi to for basic "tests" on how CirPy differs from normal Python, and for that it's very handy
empty stirrup
I use wokwi to for basic "tests" on how CirPy differs from normal Python, and fo...
Do you run these tests "by hand" or via the CI action? (I'd love to see any examples of setting up the wokwi CI flow for CirPy in particular, just for sanity checks)
devout jolt
by hand, I use it to learn Python really
tulip sleet
@onyx hinge hi -- you had a comment about why I'm seeing merge conflicts for files like shared-bindings/canio/Message.h snd other files that have no counterpart in MicroPython. Something about them being copied?? Could you re-explain? I would like to have these not appear in the future.
onyx hinge
That seems familiar but I am not remembering right now
tulip sleet
not a big deal. I've dealt with them now, just merging them back to "ours". the other file is some random one.
onyx hinge
looks like there are ways to make merge not look for renames -- see man git-merge down at merge.renames and merge.directoryRenames.
tulip sleet
I'm not sure I want to suppress that in general...
I will keep chugging along for now.
these are spurious renames, and I was not sure how to understand the cause.
slender iron
@onyx hinge want to get my rp2040 usb host code going?
it is in my imx_usb_host branch
(just pushed it)
onyx hinge
I'll have a look -- that's in your fork?
slender iron
yup
onyx hinge
fetching .. looks like broadcom decided to update so it'll be a minute
slender iron
hrm, I didn't touch it at all. maybe you are fetching all of my branches 😬
onyx hinge
soemthing else to push?
$ git submodule sync; git submodule update
fatal: remote error: upload-pack: not our ref 1e960631fb4237b18436435ad43ed0f826993d94
Fetched in submodule path 'lib/tinyusb', but it did not contain 1e960631fb4237b18436435ad43ed0f826993d94. Direct fetching of that commit failed.
yeah I asked git to fetch all from you
slender iron
lemme look
slender iron
looks like I have some debugging stuff in it
onyx hinge
OK advise what I should try next
slender iron
@onyx hinge try fetching from my tinyusb again. I pushed to a branch
onyx hinge
OK checkout works
slender iron
pushed another commit to the cp branch to include the tinyusb commit
onyx hinge
OK, I have a binary. what should I expect to work / not-work?
I won't get out HW right now but just wanna know where it makes sense to start working..
slender iron
you should be able to import usb.core and usb_host
the regular feather and feather usb host have board_init code for it
slender iron
I was able to get serial input via HID reports
code is in supervisor/shared/usb/host_keyboard.c
onyx hinge
the core takes over the HID device entirely, as it stands?
slender iron
yes but the pyusb api has functions to disconnect to and from the kernel
(I haven't tested it.)
onyx hinge
hah OK
if it was about writing python code to interact with a usb device I was thinking about biting off USB MIDI but that may not be a good idea
slender iron
you could try it
I'm doing code for the intellikeys
(and totally found an issue with ctrl_transfer)
onyx hinge
MIDI's also pretty time sensitive..
slender iron
intellikeys is ctrl transfer. I think midi would be interrupt
intellikeys firmware loading at least
onyx hinge
once I learn more about USB I'll be able to put that statement into context
slender iron
onyx hinge
cool! you can hear all about my progress, or lack of, next week this time!
onyx hinge
just for a few minutes? I'm not on S&T tonight and I'm itching to go up and have dinner
slender iron
we don't need to if you feel like you are ready to poke at it
onyx hinge
I think either I'll be OK, or else just a few minutes video chat will not be enough to get me going. Leaning towards the former.
blissful pollen
Is the RP2040 USB host for CP still a WIP? had a thought on something to use it for whenever I get two minutes for projects again
onyx hinge
so I'll let you get prepped instead. make good use of your time away from all this!
slender iron
kk, we can just sync next week
onyx hinge
@blissful pollen I think if you tune in to S&T you'll get an update.
slender iron
@blissful pollen it is in progress but does seem to work now
see above for where to find it
onyx hinge
🤤 hungry, ttyl
blissful pollen
Thanks I think I have the stuff to wire up to a RP2040 device to try it out. Sadly haven't had much time / energy for projects the last couple weeks
slender iron
@onyx hinge one last cp push for you to snag. fixes the slowness
onyx hinge
yay!
lone axle
Is anyone seeing PRs or issues on the contributing page? https://circuitpython.org/contributing
I've got blank lists for PRs and issues.
still zephyr
Is anyone seeing PRs or issues on the contributing page? https://circuitpython.o...
Yep, me too
tulip sleet
Is anyone seeing PRs or issues on the contributing page? https://circuitpython.o...
I changed the user that is identified for these a few days ago to me, to test why all the commits were being marked as by ladyada. Somehow it is not working even though I should have the rights privs. I changed it again and am re-running to see what happens.
tulip sleet
Is anyone seeing PRs or issues on the contributing page? https://circuitpython.o...
actually, hmm, I think this may not be relevant to the missing issues. There is another issue we are working about github tokens.
@idle owl not urgent, but the PAT issue seems to be showing up, or maybe this should be fixed now ^^
idle owl
<@330227457296957440> not urgent, but the PAT issue seems to be showing up, or m...
Should be fixed. I'll look into it tomorrow if not.
idle owl
actually, hmm, I think this may not be relevant to the missing issues. There is ...
I'm not seeing Actions failures in the process. If that doesn't work, let me know and I'll dig tomorrow.
tulip sleet
me neither, maybe the missing content we are seeing here https://circuitpython.org/contributing is from Tuesday after the initial PAT problem and before the fix.
I'm doing a re-run, but I'm not sure it will work.
idle owl
Keep me posted. I'll check in tomorrow. Thanks.
manic glacierBOT
Just to be clear, the visible and IR LEDs on the M5Stick C Plus are active low:
https://m5stack.oss-cn-shenzhen.aliyuncs.com/resource/docs/schematic/Core/M5StickC_Plus/StickC_Plus_20200616.pdf
timber mango
So I added CircuitPython support for an ESP32-PICO-D4 based board and built successfully. How do I know where to upload my new firmware file to?
I mean, what offset should I use?
0x0? 0x1000?
The build documentation doesn't seem too helpful in this regard.
I uploaded a different firmware with the web uploader that was obviously not designed for my board there's a terminal at least. Mine is just dead I don't know if it's the wrong offset or another problem with my code.
Yeah, it seems 0x0 works for other firmware. Well mine is just borked then. 🙃
brazen hatch
I mean, what offset should I use?
for CircuitPython firmware.bin you always use 0x0
TinyUF2 is the complicated one not CP
tulip sleet
So I added CircuitPython support for an ESP32-PICO-D4 based board and built succ...
have you tried a generic ESP32 build that has 4MB flash support, which is what is in the module? e.g. adafruit_huzzah32_breakout
idle owl
@tulip sleet Looks resolved. PRs etc are listed now.
Is GitHub down for anyone else?
Status is green across the board.
But git pull is doing nothing, and visiting in browser says server not found.
tulip sleet
But `git pull` is doing nothing, and visiting in browser says server not found.
me too
tulip sleet
We're having a really bad day.
timber mango
have you tried a generic ESP32 build that has 4MB flash support, which is what i...
I uploaded the UM Pico firmware and got a terminal at least, but my own build is showing no signs of life.
I've checked over my code a few times, I can't figure it out.
tulip sleet
compare your board files with the other one, particularly the sdkconfig
timber mango
All I have in the sdkconfig file is the hostname. The other build has lots of lines about SPIRAM, but I don't have any so I didn't define it.
I'll take another look at my files.
tulip sleet
try the HUZZAH32 breakout build too, and make sure the serial lines you chose are correct
timber mango
I'm using an ESP32-PICO-D4, is that compatible with an ESP32 build?
idle owl
GitHub appears to be up again.
timber mango
I'll try building something that definitely works on my board and see if the build environment is the issue.
tulip sleet
I'm using an ESP32-PICO-D4, is that compatible with an ESP32 build?
yes
timber mango
Ah, thanks. I wasn't sure.
worn eagle
Any one know how to fix ESP_RESET in use error?
timber mango
Yep, looks like it's my build environment. A prebuilt binary that was previously working doesn't work after I build it.
Oh, hang on. I was uploading circuitpython-firmware.bin rather than firmware.bin. Well... now I just feel silly.
My code works perfectly fine.
timber mango
Is the web interface something you just get out of the board with CP 8.2.0 or do you need to enable something in the board config?
stuck elbow
Is the web interface something you just get out of the board with CP 8.2.0 or do...
you have to setup the wifi in settings.toml
timber mango
you have to setup the wifi in settings.toml
I was actually talking about how to enable it for a build.
stuck elbow
I think it's enabled by default on supported boards
timber mango
Doesn't seem to be working on my build. Maybe it's something I did wrong. Not sure.
stuck elbow
what's the board?
timber mango
SQFMI Watchy. An ESP32 board. Something I just added.
stuck elbow
esp32 definitely has it enabled by default, otherwise it's a pain to interact with it
but setting up the wifi is still a bit of a pain
timber mango
WiFi connection is working.
stuck elbow
from settings.toml?
timber mango
Yes. It's got an IP address.
stuck elbow
then just go visit circuitpython.local from your browser in the same network
timber mango
That's the problem. It's not working, it just times out. Just trying to see if I need to enable something.
I'll check my sdkconfig
stuck elbow
I'm looking on other esp32 boards and they don't seem to have anything enabled explicitly in their files
timber mango
Hmm, I rebuilt and reuploaded. Seems to work now. Not sure if it's what I changed or the board just needed a hard reset.
stuck elbow
ah, yes, after changing the settings.toml, you need a reset
but if you got an ip address...
in any case, the setting is CIRCUITPY_WEB_WORKFLOW, you can add it to the .mk file if you want to be explicit
timber mango
I did reset several times, but I think that was just a soft reset. Well I don't know. I also changed my mpboardconfig.
manic glacierBOT
tulip sleet
@onyx hinge @slender iron My plan is to wait until Wed July 5 for 8.2.0 final, given that Jeff was mostly off today and it would be good to have normal coverage for the stable release.
onyx hinge
<@426200547024961539> <@252717193496756235> My plan is to wait until Wed July 5 ...
Makes sense, no reason to release going into a holiday weekend
manic glacierBOT
onyx hinge
ah OK I don't know the difference between TinyUF2 and the others .. I guess some ports were made before the "tinyusb-ification" that made porting easier?
tulip sleet
ah OK I don't know the difference between TinyUF2 and the others .. I guess some...
yes, I think maybe the long-run plan is to merge the nRF code into TinyUF2
Adafruit_nRF52_Bootloader does use tinyusb
timber ivy
@tulip sleet you pointed out that the M5StickC-Plus LED outputs are active low, is there a standard way to deal with that so things like PulseIO work properly? I'm pretty new to this.
timber mango
Can I put my own board specific library in the firmware by default?
manic glacierBOT
stuck elbow
Can I put my own board specific library in the firmware by default?
yes, you can freeze it
manic glacierBOT
tulip sleet
<@329766224093249548> you pointed out that the M5StickC-Plus LED outputs are act...
We have had very few requests for inverted PulseOut. It hadn't occurred to us someone would wire a transmitter IR LED this way, but there it is, in the M5StickC. So the issue is open to implement it.
Most of the time for simple GPIO-driven LED's, you can just invert the logic in the program.
kind river
Any interest in a port for the ESP32-S3 on the new UNO R4 wifi? I'm currently running the devkit port on it and starting to work on a library for the RA4M1's UART bootloader.
tulip sleet
Any interest in a port for the ESP32-S3 on the new UNO R4 wifi? I'm currently ru...
I'm looking at the schematic for the Uno R4 WiFi, and see the quite limited connectivity of the ESP32-S3. What would you want to do with CircuitPython in that situation
You could try one of the existing builds for that module.
espressif_esp32s3_devkitm_1_n8
kind river
I'm looking at the schematic for the Uno R4 WiFi, and see the quite limited conn...
The intent is to do high level stuff in circuitpython on the ESP32 and realtime control/IO stuff in arduino on the RA4M1 because it seems a bit silly to run high level web stuff on the low power MCU.
tulip sleet
The intent is to do high level stuff in circuitpython on the ESP32 and realtime ...
I think you can start with the build above. You could modify it and remove nearly all the pins if you care 😆
kind river
I think you can start with the build above. You could modify it and remove nearl...
The devkit port is what I am using. I have an almost working bootloader library for the RA4M1
tulip sleet
are you interested in the board because it's 5v? Otherwise it's just kind of an engineering challenge/puzzle
kind river
Mainly so I can run realtime control and sensor fusion for IMU corrected steering in arduino. The 5v is also handy for Pololu's sumo bot
tulip sleet
have you chosen an existing protocol for communication between the two processors, or are you doing something bespoke for your purposes? (I mean higher level than SPI or whatever.)
kind river
I need to dig into the SCI pins and see if I can do SPI or just UART over the shared pins and look into existing RPC solutions once I get program loading working
my eventual goal is to write some wrappers that will let me basically just call exposed arduino functions from circuitpython but that's a ways off
tulip sleet
one more note: we have been trying to label pins that are (unusually for the application) the opposite of what one would expect. We add the suffix _INVERTED on the pin name. So the M5StickC-Plus LED pins would be LED_INVERTED, etc.
manic glacierBOT
timber mango
Put in a pull request for Adafruit_CircuitPython_NTP, it's broken for my timezone at moment.
https://github.com/adafruit/Adafruit_CircuitPython_NTP/pull/30
How long do these usually take to get addressed?
jaunty juniper
Put in a pull request for Adafruit_CircuitPython_NTP, it's broken for my timezon...
In the PR you write "This bug made it impossible to get the correct local time" but how ? It's just type hints, they don't have any effect in the real code
it's still good to fix the type hints, provided it works, though I suspect it actually needs to be converted to int after converting to seconds or there might be some precision loss later
timber mango
Oh, huh you're right. Well it must've been something else then? mpy-cross do anything with these hints does it?
timber mango
Yeah, it's half an hour off when I pass in the timezone. UTC is correct.
I'll take a look at my code...
jaunty juniper
I think you'll want line 59 to say
self._tz_offset = int(tz_offset * 60 * 60)
Yeah, that seems to fix the inaccuracy.
But hmm, it shouldn't have been that bad???
jaunty juniper
Circuitpython floats are 30 bits, which makes the math later lose precision
timber mango
Anyway I guess I need to push the commits to my branch and submit a new pull request? I'm not really an expert in this kind of Git.
jaunty juniper
as the epoch timestamps are generally larger than what fits a CP float (which is why time.time() is an int for example)
you can push on your branch, it will update the pull request
timber mango
Uh, done. Did that work?
midnight ember
tz_offset is supposed to be an int? didn't even know there were offsets that required a float.
offsets are usually unix offsets in seconds like minus or plus hours done in seconds.
timber mango
The parameter is passed to the API as an hour, so it's obviously going to need a float.
jaunty juniper
it's assumed to be an int in the code
timber mango
Anyway the docstring below says it's a float.
midnight ember
weird, maybe docstring is incorrect?
jaunty juniper
the issue is that math later, that becomes wrong because the sum is made a float, losing precision
self._monotonic_start = (
seconds
+ self._tz_offset
- NTP_TO_UNIX_EPOCH
- (destination // 1_000_000_000)
)
ah yes if using monotonic seconds sometimes aren't really truly seconds
timber mango
weird, maybe docstring is incorrect?
I would assume the docstring is correct, the API takes an argument in hours, then converts it into seconds. So you'd need a float for something like UTC+9.5h
I was talking about the NTP module init function.
Ah, sorry.
midnight ember
i've never used ntp library, nvm
jaunty juniper
CP floats be like that:
>>> 3897377815
3897377815
>>> f"{3897377815.:f}"
'3897376768.000000'
>>> 3897376768 - 3897377815
-1047
Maybe we should just... pass the argument in seconds in the first place?
jaunty juniper
that would be an API breaking change, I prefer fixing float values
timber mango
Yeah, true.
jaunty juniper
PR tested and approved, someone will come and merge and release (maybe not now, it's ID4 weekend for some people)
timber mango
It just feels kinda dumb that I converted the UTC seconds offset to an hour float with division only for this library to convert it back again, with all that accuracy loss. 😆
Anyway... I suppose I could just do a hack of setting the _tzoffset property directly. 🤔
jaunty juniper
eh the resulting seconds don't actually lose accuracy
timber mango
Any advice or merging a new board definition in? Defs are done and nearly done making a custom library for it.
jaunty juniper
um make sure to run pre-commit I guess ?
timber mango
Will do, well, thanks for floating a pull request for me.
I love it how whoever made this guide assumed there was absolutely no possiblility your board wouldn't have a status LED, probably several! 😆
onyx hinge
@timber mango thank you for the PR! I'm glad you & @jaunty juniper got to the root of the problem.
@midnight ember there are not only half-hour offset zones but also at least one quarter-hour offset zone (Nepal).
tulip sleet
<@&356864093652516868> Anti-reminder that there is no CircuitPython discord meeting today, due to the US holiday tomorrow. The meeting will resume next week, on Monday, July 10.
midnight ember
didn't know that, should still be able to do offset in seconds though even with a float.
my only experience with tz offset is with requests library stuff. using seconds for offset is very easy. converting fromfloat to int back to float with ntp module seems unintuitive in comparison?
jaunty juniper
yeah as long as we don't have millisecond time zones we should be set for now 😉
I'm thinking if a x * 3600 could result in a float ending in 999.99999 dropping the last second in conversion, maybe we should use round() instead of int() but it doesn't seem to be the case anyway
errant grail
Another option for scientific precision is to use jepler’s udecimal library https://github.com/jepler/Jepler_CircuitPython_udecimal
manic glacierBOT
onyx hinge
I'm thinking if a x * 3600 could result in a float ending in 999.99999 dropping ...
Half and quarter hours offsets will be exact in fp arithmetic, happily. Using round () wouldn't have hurt though
manic glacierBOT
manic glacierBOT
lone axle
is there a comprehensive listing anywhere all modules that are possible to enable/disable with these flags in the .mk file? https://learn.adafruit.com/building-circuitpython/customizing-included-modules Or anywhere that lists the size of those modules to help someone trying to make a build with certain module figure out what is best to cut based on their needs and the size of the modules?
brazen hatch
You can pretty much disable everythin. Some just need others to be disabled too.
A dependency chain if you will.
jaunty juniper
the list of what can be enabled/disabled in mpconfigboard.mk is pretty much the things in py/circuitpy_mpconfig.mk
then there's the .h
brazen hatch
I cannot, emotionally, physically or spiritually fix this.
Left is esp32, right is esp32c3. Same files.
C3 just decided to just ignore all .flush and not print out the line. The line only shows up after you start typing.
How does one man bugreport this even. It's not my stuff. it works on all other mcus. But I have no hope of singling it out.
Of course in repl the virtual console works fine. C3 is gonna be the end of me.
plucky tulip
Does anyone know whether ESP32-S2 have a limit on open sockets, or maybe CircuitPython limits them? I am working on Server-Sent Events on adafruit_httpserver and maximum number of connected client I can have is two, then the server does not accept connections, like it does not have available sockets.
EDIT 1: To be precise a am saving open sockets to a list. If there are more than two clients, server does not respond, if I disconnect one, responds normally. Considering that one socket is reserved for served the theoretical limit I am facing might be 3 not 2.
EDIT 2: If I create 3 separate servers, none of them respond to requests. 2 servers respond normally. Behaves similar to problem with SSE.
crimson ferry
@plucky tulip is web workflow enabled?
plucky tulip
yes
crimson ferry
that's eating up something, checking to see if I can give a better answer...
plucky tulip
I will test how it behaves with web workflow disabled
crimson ferry
I used this code to document Pico W socket limits https://github.com/adafruit/circuitpython/issues/7543#issuecomment-1509307970 maybe try on ESP32-S2?
I get stream=8, dgram=8, raw=8 (w/o web workflow)
CONFIG_LWIP_MAX_SOCKETS=8
plucky tulip
With web workflow enabled max 3 servers, but web workflow does not work.
Without web workflow same.
2 servers work with and without web workflow.
crimson ferry
web workflow and servers use different ports?
plucky tulip
yes, web workflow on 80, servers on 5000, 8000, 9000
no matter what 4th connection does not work
crimson ferry
I get 7 sockets with web workflow on (stream, dgram, or raw... same)
maybe there is some other limit
listen is one, each accept is (presumably?) one
(https requests use two each)
could run the socket counter after each accept to see how many are left
plucky tulip
each server closes socket after sending response, so they do not stack
crimson ferry
curiouser and curiouser
plucky tulip
keeps alive the listen one though
I wonder whether it is a hardware limit or a software lock, either way, 8 != 3
crimson ferry
if something was leaking sockets, I think we'd expect that after several iterations, no connections could be made
the stable 2 connections is... interesting
not even simultaneous
listen(1)?
plucky tulip
the stable 2 connections is... interesting
even three, two servers, and request that comes to one of them
crimson ferry
hmmmm:```
CONFIG_LWIP_MAX_ACTIVE_TCP=4
CONFIG_LWIP_MAX_LISTENING_TCP=4
well, at least it is good to know that is is not user error what I am experiencing, the 4 max sockets kind of limit the usage of SSE to only one client (each client needs an open socket), because one is for server and one for incomming requests, maybe the 4th one is used somewhere else or simply this limit is sockets<4
Thank you @crimson ferry for you help! 
empty stirrup
[low-priority] Submitted a minor PR just to get familiar with the contribution process; any suggestions on what would streamline the review for the dev team? (https://github.com/adafruit/Adafruit_CircuitPython_MAX9744/pull/17)
onyx hinge
@tulip sleet did you intend to close #8135 vs leaving it open as long-term?
tulip sleet
i did not intend to close it, thanks.
onyx hinge
@empty stirrup only thing I noticed was that the special keyword to link a pull request to an issue needs to be in the original message, not a reply .. so I put it there. https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue
empty stirrup
Ah! Thx for the tip.
onyx hinge
and thank you for the contribution.
empty stirrup
It's a fantastic project - thanks to all of you
onyx hinge
Is there a non-blocking stdin read in circuitpython? usb_cdc.console because the goal is to use any console-ish source of input, such as a USB HID keyboard with USB host..
tulip sleet
Is there a non-blocking stdin read in circuitpython? `usb_cdc.console` because t...
Nope, as far as I know. CPython doesn't have this -- there are OS-dependent tricks to figure this out, because in general you can't tell. So we had no model for this. And sys.stdin.read() doesn't have a timeout arg
onyx hinge
OK maybe something will need to be added..
tulip sleet
OK maybe something will need to be added..
https://stackoverflow.com/questions/3762881/how-do-i-check-if-stdin-has-some-data. Should this stream actually be sys.stdin? Maybe it should be something else that is smart about where it reads from.
@onyx hinge https://github.com/adafruit/circuitpython/issues/8121 (audio underflow) was partially addressed by https://github.com/adafruit/circuitpython/pull/8123 . Did you want to keep it open with a later milestone for a more complete fix?
onyx hinge
@tulip sleet I'd like to keep it open at the moment but move it to long term
tulip sleet
done
onyx hinge
thanks
@tulip sleet select.poll works with sys.stdin, which will read a char from USB CDC 0 or from the USB HID keyboard attached on USB host.
so a nonblocking or timeout read can be created based on poll
tulip sleet
Great! I didn’t know select was functional
tulip sleet
@onyx hinge @slender iron I could make 8.2.0 final today. There are a few new bugs noted, but nothing I see that is a showstopper. What think you?
onyx hinge
@tulip sleet it's fine by me, I don't have anything that I'm working on for 8.2.0.
onyx hinge
@idle owl do you know if all our circuitpython projects have "adafruit-travis" as the maintainer? seems weird. I was trying to figure out who/what to add as a co-maintainer for a pypi project I need to put under adafruit control
idle owl
<@330227457296957440> do you know if all our circuitpython projects have "adafru...
Uff, no. That isn't right. It's adabot.
Wait that's RTD
I accepted the invite.
onyx hinge
@idle owl are you able to accept the invitation "[PyPI] jepler has invited you to join the g722-1-mod project" that you probably got in your adafruit e-mail? the link doesn't work for me but I don't think I have access to log into pypi with that account.
thank you!
orchid basinBOT
tulip sleet
<@330227457296957440> do you know if all our circuitpython projects have "adafru...
according to my records, the username on pypi is indeed still adafruit-travis, for historical reasons
idle owl
Yeah it is.
I was thinking RTD.
I'm going to see if the PyPI folks can help with a username update, but it's a huge manual process, so the only way they could help is if they have some way to bulk add a maintainer to a huge list of projects. Based on the username-change issues on GitHub, I imagine they do not.
Worth asking anyway.
tulip sleet
I don't think it's that important, because it isn't tied to github.
idle owl
It bugs me. 😄
tulip sleet
i have the credentials, I assume you do too
idle owl
Yes.
onyx hinge
did we ever look into whether pypi organizations help us? wasn't that announced at pycon2023?
I mean there is probably still a terrible manual procedure to suffer through but
tulip sleet
I asked for it or voted for it or something. Yes, we could start to use that
idle owl
Yes they're in the settings.
tulip sleet
I am quite willing to work on transitioning to that, maybe using RTD and github as models. I am willing to do the tiresome click work.
idle owl
I want to ask about the username thing first, please.
Because the only way to do it is create a new account.
onyx hinge
sure, I don't want to step on any toes here
idle owl
I don't want to get this all set up and have to do it again.
slender iron
@tulip sleet thanks for making 8.2. It was fine with me. I was busy with daycare stuff.
still on Ari duty while he naps until Becca's parents get back
tulip sleet
yw - no surprises since last week
slender iron
👍
onyx hinge
@slender iron not sure how to make key repeat work with USB keyboard .. but I have great progress with a terminal-based editor running under CP https://github.com/jepler/editor
slender iron
<@252717193496756235> not sure how to make key repeat work with USB keyboard .. ...
That's awesome!
onyx hinge
I found some very minimal editor to build on, github has everything 🙂
slender iron
I was thinking you'd track the start of the keypress and the last time you emitted one.
Then when you get a "is there something?" or "give me the next char" you'd check the repeat
onyx hinge
hmmmm so move the repeat logic into the receive side instead?
slender iron
ya
onyx hinge
instead of that code running on the other core, hm
I'm not immediately in love with that idea, but I will keep it in mind..
slender iron
I don't think the HID report to serial code is running on core 1
it runs in tuh_task on core 0
onyx hinge
oh, I thought the usb host was on core 1...!
slender iron
core 1 is doing low level host stuff
onyx hinge
oh okay
slender iron
basically that a peripheral would do
every 1 ms send the usb packets and then queue stuff into tinyusb from the "interrupt"
slender iron
@onyx hinge how do you switch between running your code and editing it?
onyx hinge
@slender iron that's not implemented
onyx hinge
Thank you Dan! So excited for this new release, it is yet another great step for audio synthesis especially on rp2040.
tulip sleet
WARNING for nRF52 boards only: If your board has an nRF52 UF2 bootloader whose version is before 0.6.1, you will not be able to load CircuitPython 8.2.0 and later. The 8.2.0 firmware is larger than what the older bootloaders can handle. See these instructions (https://learn.adafruit.com/introducing-the-adafruit-nrf52840-feather/update-bootloader) for updating your bootloader.
cunning orchid
Out of curiosity, is https://github.com/adafruit/circuitpython/issues/4106 mainly waiting for someone to diff with MicroPython, and test? Or is the idea to wait until the threading API stabilizes more before even thinking about PRs?
I ask because I may look at poking at it once I grab a Pico W and an ESP32-S3 board, supposing the API stability isn't a blocker.
I've also been thinking about going a route similar to what was mentioned in the issue thread and treating the second CPU as a co-processor, rather than first-class CPU and providing a Python API as well as maybe a Rust/C API (I've become rather enamoured with Rust on embedded after seeing how easy it was for a POC on a SAMD21) to assign it async tasks.
onyx hinge
kicad to svg https://hacmanchester.github.io/LaserCutter.PCB.Making/KiCad/KiCad-ExportSvg/ @idle owl @slender iron
idle owl
kicad to svg https://hacmanchester.github.io/LaserCutter.PCB.Making/KiCad/KiCad-...
Duly noted. Thanks.
slender iron
Out of curiosity, is https://github.com/adafruit/circuitpython/issues/4106 mainl...
I don't think we'll ever implement it like micropython. We'll do the coprocessor approach.
while we haven't worked on the python api, we have recently started using the second core internally for picodvi and (soon) usb host
cunning orchid
I don't think we'll ever implement it like micropython. We'll do the coprocessor...
To be honest, that makes the most sense to me for the MCU use cases that come to mind.
slender iron
it doesn't really help solve the issue of code location and communication between processors
since those two examples use code compiled into the binary
the closest existing APIs for circuitpython is the ULP coprocessor API for ESP and the PIO API on rp2040
cunning orchid
Will take a look at those, thanks!
slender iron
here and the issue are the best places to discuss
(though the issue has some unproductive people involved)
jaunty juniper
but if you want specifically to make a uf2 with default files, on many ports you can save the flash into a UF2 that will then install everything when flashed on a board
so you prepare the content you want on a live board, save the UF2, and distribute that
the content of the flash won't survive a erase_filesystem, but the UF2 can be reapplied instead
timber mango
So are you talking about flashing them with a build before distributing or it that something I should include in the board definition?
jaunty juniper
yeah flash them with a build on manufacturing
timber mango
Ah, well no probs then.
onyx hinge
@slender iron earlier elsewhere I mentioned I was seeing spontaneous resets from the USB Host feather but now I wonder if it wasn't just me bumping the magnetic USB cable. I'll try again with a non-magnetic cable and see if it keeps happening
jaunty juniper
UnexpectMaker distributes his boards with default files, but he might just have a script that copies them over mass storage, I think his test jigg uses a Raspberry Pi
you can ask him on his discord
other that that you can do picotool save --all my_full_board.uf2 for RP2040
and esptool can get you a dump of the whole flash as a bin
timber mango
Yeah, I might do that. I was checking out his builds for clues but I don't see anything out of the ordinary so he's probably just putting the files on after the fact.
jaunty juniper
also kattni has some script that she uses to help flash and prepare boards that she takes to conventions and stuff for workshops
but that's probably not practical for production quantities
timber mango
(I mean, I just realised that I could put some sneaky bootstrap into the init code in board.c, it's just a matter of whether I should.)
empty stirrup
[ apologies if I missed an existing answer] - is there an existing CP library to run ML inference? If not, is there someone currently working on something like this?
tulip sleet
[ apologies if I missed an existing answer] - is there an existing CP library to...
not, that I know of. We have run ML models on our boards, but they are not CircuitPython, e.g. TensorFlowLite https://learn.adafruit.com/search?q=machine%2520learning
empty stirrup
Ack - thanks for the info!
slender iron
@onyx hinge I'm at my desk now if you want to sync up on usb host stuff
onyx hinge
@slender iron sure. I went down a non-circuitpython rabbit hole this afternoon but I can wrap that up in 10 and then get on video with you.
slender iron
👍
onyx hinge
@slender iron thanks for your patience, i'm at my desk now
idle owl
@onyx hinge @slender iron I have a question about the swirly grids. Product copy says "you can pretty much make anything on a 0.1 grid fit". The separate usage of "grid" there is confusing me a bit, and will ultimately confuse others. What is that sentence trying to say? Can you please assist with wording it better?
onyx hinge
@idle owl I'll try
idle owl
Thank you
onyx hinge
You know about placing things "on the grid" in fritzing, right?
onyx hinge
"on a 0.1 grid" means "if everything was placed on that kind of grid, like in an art or drafting program"
but I do not know a good way to explain that to humans who don't draw the connection
onyx hinge
Most boards with hole spacing that is a multiple of 0.1 inch will fit on the swirly grid
This includes Feathers, FeatherWings, and most Stemma QT boards
an examples of things that are NOT on an 0.1" grid would be arduino/metro
an example NOT on a 0.1" grid, because the horizontal distance is 1.15in ..
idle owl
> Most boards with hole spacing that is a multiple of 0.1 inch will fit on the s...
Ooooh
Thank you!
slender iron
@stuck elbow see these fonts? https://idleberg.github.io/playdate-arcade-fonts
manic glacierBOT
@dhalbert , it hangs here:
But with the modified CP as I told:
Other interesting thing, is that even with original binary CP 8.2.0, the board appears as N32R8, but is is a N8R2.
onyx hinge
of course devkitC3 has an S3 on it. 🤯 sorry for confusing that issue further.
idle owl
@onyx hinge @slender iron Another swirly grid question. Do either of you have any tips or tricks about them yet? A page of them is to be included in the guide, and I am wondering if there's anything to include, or simply make it a placeholder for now.
I figured out that, if done carefully, you can push the STEMMA QT cable connectors through the slots. Means it's easier to mount things on both sides, or create a cleaner look on the top. I'll add that at least.
onyx hinge
I .. haven't actually used the swirly grid yet for a "real project"
but that sounds like a good idea
slender iron
chatgpt makes my jaw drop when it works like I expect: https://chat.openai.com/share/f27a3302-1a9b-41e3-824b-0e6eb8addb9e
ornate breach
chatgpt makes my jaw drop when it works like I expect: https://chat.openai.com/s...
welp, almost. Netherlands/Dutch doesn't seem to have been caught properly
but pretty impressive regardless
one small hiccup that's easy to remedy
slender iron
I think it's fixed after the second prompt
stuck elbow
<@351136009922019331> see these fonts? https://idleberg.github.io/playdate-arcad...
seem a bit small for playdate's resolution
digital shoreBOT
adafruit
manic glacierBOT
This is the board, I bought on Aliexpress:
onyx hinge
onyx hinge
@slender iron was it only in a dream that you coded an in-place base64 decoder, that re-used the input area as the output?
slender iron
I did! in the web workflow file
onyx hinge
oh for password?
onyx hinge
doh, my greps weren't finding things because I was bisecting a problem and was in some early 2022 commit
_base64_in_place(_api_password, strlen(_api_password), sizeof(_api_password) - 1);
slender iron
@tulip sleet this might be the issue the person is having on the odroid go
(zero byte settings.toml)
onyx hinge
@slender iron uff .. what's happening with zero-byte settings.toml?
slender iron
<@252717193496756235> uff .. what's happening with zero-byte settings.toml?
the person in #8141 is having issues
and I suspect that might be the issue with odroid go too
may be worth testing it with an empty file
onyx hinge
was there a particular clue that leads you to this line of thought?
shared/filesystem.c: make_empty_file(&vfs_fat->fatfs, "/settings.toml"); many devices have empty ones
slender iron
ah, I didn't realize we made it
I was bringing it up based on the last comment of the issue
I don't think I've tested an empty one myself
onyx hinge
From e8c7b4a9f4d448dd5ecee5ba9ea7c46161e7c2e9 Mon Sep 17 00:00:00 2001
From: Jeff Epler <jepler@gmail.com>
Date: Fri, 7 Jul 2023 11:28:42 -0500
Subject: [PATCH] test a zero-byte environment file
---
tests/circuitpython/getenv.py | 2 ++
tests/circuitpython/getenv.py.exp | 1 +
2 files changed, 3 insertions(+)
diff --git a/tests/circuitpython/getenv.py b/tests/circuitpython/getenv.py
index 68dd328cfb..f01189a551 100644
--- a/tests/circuitpython/getenv.py
+++ b/tests/circuitpython/getenv.py
@@ -72,6 +72,8 @@ def run_test(key, content):
print(key, str(e))
+run_test("key", b"")
+
for i in range(13):
run_test(f"key{i}", content_good)
diff --git a/tests/circuitpython/getenv.py.exp b/tests/circuitpython/getenv.py.exp
index ba1b1b3f90..925dafb469 100644
--- a/tests/circuitpython/getenv.py.exp
+++ b/tests/circuitpython/getenv.py.exp
@@ -1,3 +1,4 @@
+key None
key0 'hello world'
key1 7
key2 '\n'
--
2.39.2
```I just wrote a test for an empty settings.toml and it didn't immediately turn up a problem👍
onyx hinge
urgh I have to stop worrying about making this the ultimate keyboard mapping and just consider that I can probably write something good enough... (like support for dead-keys and shift-altgr ... altgr support is needed for e.g., {} on french keebs though)
tulip sleet
<@329766224093249548> this might be the issue the person is having on the odroid...
hmm, did you try adding an empty settings.toml to your Odroid Go and retesting?
slender iron
no, I didn't
timber ivy
FYI, I was able to build 8.2.0 proper for my M5Stack Stick C Plus this week and test it out. Would you like me to start building out a board page for it in circuitpython-org?
tulip sleet
FYI, I was able to build 8.2.0 proper for my M5Stack Stick C Plus this week and ...
sure, thanks, that saves makermelissa time, and you have the board on hand. There is a guide: https://learn.adafruit.com/how-to-add-a-new-board-to-the-circuitpython-org-website
slender iron
manic glacierBOT
manic glacierBOT
manic glacierBOT
slender iron
I didn't know the host API was done!
Getting there. I’m a branch now and hopefully prd next week
thorny jay
Hi, is it possible that the toml parsing and os.getenv does not work on Trinket M0?
At least that is what my test comparing behavior on M0 and RP2040 reveal.
It would make a lot of sense if there is no space left for that on M0, but how am I supposed to know if it is available or not? Both seems to implement "os".
onyx hinge
@thorny jay yes, it's not supported there. no, the docs don't reflect it well.
if "well" means "at all"
due to technical limitations we don't have a way to put individual functions in the support matrix
thorny jay
<@681528999520108564> yes, it's not supported there. no, the docs don't reflect ...
Hopefully I have a good memory and this place exist.
I guess I am going to use the "secret" method importing a file, and protecting the import by a try.
Goal is to externalize the parameter, so that the code behaviour can be modified without touching code.py.
I decided to go M0 because I had a few trinked M0 in stock.
I had to use your TICKS function to have accurate timing and work around the absence of asyncio.
( https://docs.circuitpython.org/en/latest/shared-bindings/supervisor/index.html )
It has been a bit challenging, but I am getting close to my goal.
onyx hinge
whee now you can type on a french-layout keyboard (among hopefully many others) with a USB keyboard & CircuitPython in USB host mode! but it depends on Scott's ongoing not yet merged work too .. https://github.com/tannewt/circuitpython/pull/19
