@silver tapir enabled π
#circuitpython-dev
slender iron
silver tapir
thanks.
hybrid scarab
The CookieCutter is about as fun as gargling glass though
slender iron
I like evolving APIs to keep folks used to things evolving
lapis hemlock
@onyx hinge Jeff, don't forget to pull this in https://github.com/adafruit/circuitpython/issues/3265, before 6.0 goes out.
hybrid scarab
Keep people on their toes, just randomise all the method names, GO!
thorny jay
If there is one example library that only care about one chip, and then other library that rely on that for implementing board specific... then I am interested in analysing how that work and how it is done.
slender iron
I don't think anyone would like that very much
thorny jay
I don't really want to invent something, but I can copycat.
slender iron
sounds like JS obfuscation
thorny jay
Pimoroni ( or @hybrid scarab ) made some interesting variation in very similar library, doing the same thing mostly but with slight difference in the available function.
Like on library had brightness, the other not.
hybrid scarab
Oh yeah that's just classic me making a mess of things
thorny jay
And I was trying to port from one to the other trying to figure out why it was different.
hybrid scarab
I'll let you in on a secret- I have no idea what I'm doing
thorny jay
Well some hardward do not have brightness...
Big warning... I learned a lot about python by reading code from Gadgetoid!
hybrid scarab
(okay CookieCutter isn't that bad)
thorny jay
Ok, now I should figure out what resistor I need for SDA/SCL and if it's Ok to have not exactly the right value because of what I have in stock... and maybe why this is needed on the FeatherCap and what is the difference between a Pi and a Feather.
onyx hinge
Well this is weird. circuitpython's ports/unix crashes with "import ulab.linalg"
so does "import ulab.fft", so it's not special to linalg...
lapis hemlock
This works on micropython import ulab from ulab import linalg
But I konw that on circuitpython you import differently.
hybrid scarab
You should get away with a pretty wide range of SDA/SCL resistors in my experience, as long as you're not doing anything challenging with i2c
ivory yew
at low speed it doesn't matter much
at high speed it does
what's this about pull-up sensing?
hybrid scarab
It throws a little message if you don't have pull-ups on your i2c lines, it's cute
hybrid scarab
But also makes me think I've broken something
onyx hinge
yeah in micropython import ulab.linalg is not accepted, it says no module named 'ulab'
ivory yew
that is ... weird, @onyx hinge. That should be valid.
lapis hemlock
I am trying to figure out what has changed since 11a7ecf, which is the version you have.
onyx hinge
@ivory yew in micropython it's a (I think) deliberate choice; they don't facilitate having hard-coded submodules. We added patches to make it work better
ivory yew
ah
onyx hinge
import ulab.linalg works on circuitpython, phew
lapis hemlock
Oh, so was it a false alarm, then?
onyx hinge
I mean, on hardware
@lapis hemlock for what it's worth, ulab.linalg.cholesky exists for me without your change in ulab
'0.54.0'
>>> ulab.linalg.cholesky
<function>
Oh my APDS9500 I'd almost forgotten about that
lapis hemlock
I haven't checked it, I worked from the issue that was posted on circuitpython, and then I realised that it didn't work on micropython.
@onyx hinge So, what do we do, then? Just close the issue, and walk away?
onyx hinge
Figure out why it didn't work for the user?
lapis hemlock
I would also like to know, why it worked for you, and why it didn't work on your unix port.
Something is definitely fishy here.
onyx hinge
yeah but almost certainly not ulab's fault
lapis hemlock
As I said, there was a problem in 0.54.0, I just don't understand, why it didn't affect circuitpython, and why the user claimed that it did.
onyx hinge
the import-crash is not ulab's fault
thorny jay
One day, I will have a deep look at ulab. But it seems to be math... some of it I did understand in the past and succeeded at exam. Even if it was fun at the time, I manadge to forget almost everything. Now I start to see a lot of interesting use of math, maybe I should re-learn. My kids are not far enough that I cannot help them with homework. Let's hope one of them get deep into math and I have to help them!
lapis hemlock
@onyx hinge OK, but then I think either you, or @tulip sleet should respond to that specific issue.
It wouldn't be proper, if I mingled in circuitpython matters.
onyx hinge
let me check whether circuitpython 5.3 behaves or not ...
hybrid scarab
@half sedge PM25 library works nicely, I wonder if we can get Enviro+ added into the simpletest π
idle owl
@hybrid scarab That depends, do you want to submit an update to the simpletest? π
hybrid scarab
Got me there π
manic glacierBOT
hybrid scarab
hybrid scarab
Why is the PM25 header... b"BM" π
Also is the any reason why- in CircuitPython- you would do
if (self._buffer[0] != 0x42) or (self._buffer[1] != 0x4D):
instead of
if not self._buffer[0:2] == b'BM'
the first uses less memory but I'm not sure it's worth the worse readability
hybrid scarab
Oh lord what eldritch horror does indexing into a bytearray involve?!
slender iron
the slice generates a new bytes object
hybrid scarab
So raw bytes + overhead * 2
Ah yes I'm aware of the magic of numbers < 32 bits
This is clearly written with a very keen eye for memory usage
slender iron
yup, and bytes objects are actually two allocations
hybrid scarab
Similarly I would do self._buffer[2:] = data
slender iron
one for the object and one for the array contents
I would only worry about this if you are doing it very often
hybrid scarab
But the O.G. code has a for-loop, copying each byte from the bytearray returned by serial read into the bytearray of the buffer
hybrid scarab
Now IIRC some serial libraries like to return surprising things
slender iron
we've had a lot of libraries written by folks coming from C so they may lack python-isms like slicing
hybrid scarab
Aye you did mention
slender iron
including libraries ported from arduino
hybrid scarab
This strikes me as some tread-very-carefully code though
slender iron
which is why I'm excited to get you on board π another person coming from python land
hybrid scarab
But I will base a PR upon David's work to package-ize this, to hopefully bring some idiomatic python (or idiotic python if I trounce some careful optimisations) into the mix
I come from javascript land, I'm just passing through Python while the jobs are exciting π
slender iron
I would only do the optimization if its in frequently called code
hybrid scarab
Yeah this is... infrequent to say the least
slender iron
then readable is better
hybrid scarab
Usually every second, but the PM sensor will actually speed up if it detects a change
hybrid scarab
Knowing my luck this will catch fire on the Pi
hybrid scarab
@thorny jay I don't know if this will fix anything, but it's more idiomatic and takes the complete SoF sequence into account- https://github.com/dglaude/Adafruit_CircuitPython_PM25/compare/library-to-package...pimoroni:library-to-package
Actually it might break SoF sync
Since if it's never read a-byte-at-a-time it's possible to get one byte out of sync, and break indefinitely
I think flushing the input buffer is probably the right way to fix that on any system that has one, though, and not my weird bytewise read
thorny jay
Yeah, checking one byte only might not be the best... and checking the two from the signature might be best. I believe the UART code is doing minimal check and attempt to sync. Then code in init.py is (or could) do more test.
That split and reuse of code between I2C and UART is confusing to me. I guess I2C has no issue of that kind.
hybrid scarab
I can't seem to beak it on my m4
for _ in range(5): # try a few times, it can be sluggish π€
thorny jay
I was poorly inserting the connector and moving the cable to try to produce bad reading.
hybrid scarab
I'm guessing i2c.readinto(self._buffer) just performs a straight read of 32 registers right from the device into the buffer
But, yes, generally i2c doesn't have the same kind of problem space that fuzzy horrible UART does
To be fair you can still get erroneous bytes on i2c, but it'll normally bail the whole protocol and fail with an IOError before you manage to do anything with them
thorny jay
I believe that what I did in PM25 is the best I can do. But feel free to improve. At least it changed from an unreliable library to something that work for me.
hybrid scarab
Yeah I'd be interested to know if my change breaks reliability for you
I can omit the change to reading the SoF packet, the other changes just clean up some non-idiomatic Python
- for i in range(31):
- self._buffer[i + 1] = remain[i]
- # print([hex(i) for i in self._buffer])
+ self._buffer[2:] = data
for example
thorny jay
So now I need to understand how to deal with a PR on a fork that is a branch that is already a PR from me. I guess this is standard practice, but I usually only do one level. This is going to confuse me... I guess I need to git pull from your repo, take that branch and test on the board and wiggle the cable.
hybrid scarab
Checking just one SOF byte means it's possible for the sensor to read total garbage, but since there's a checksum test anyway it'll throw a RuntimeError("Invalid PM2.5 checksum")
Haha, I haven't raised as a PR against your feature yet
I just linked the GitHub diff URL so you could see the changes
I aim to raise them as a PR once yours is merged
hybrid scarab
Keeps the breakage nice and separate π
thorny jay
And it will not be my take.
hybrid scarab
The comments can probably be deleted from the __init__.py too, they are... perhaps a little obvious π
Not that they hurt
thorny jay
Not sure how much is mine and how much was there first. From memory, I just made a few more check in the UART version, and then the splitting.
hybrid scarab
They've been there since initial commit anyway- https://github.com/adafruit/Adafruit_CircuitPython_PM25/blame/master/adafruit_pm25.py#L88
thorny jay
If you go into the detail, you will see that my contribution is minor or mechanical.
hybrid scarab
Yeah you poll the start byte on a loop, versus just checking it once
thorny jay
and that also maybe empty the buffer if needed.
hybrid scarab
Near as I can tell, there's either a very small buffer or no buffer to speak of
thorny jay
What I don't get is how things works... I mean, the hardware send 32 bytes every seconds? The driver does not read every second, it is only because the user code read every second. So what if I don't read for 10 seconds? Do I get 10 seconds old data? What if I try to read every 0.5s? How does that "sync"?
Does the hardware know that we want to read?
I have been VERY confused with this hardware.
hybrid scarab
Two quick successive reads will fail if the serial timeout is 0, or block until some bytes are read
thorny jay
I lost all my basic understanding of UART, buffer and things like that.
So reading too fast will just fail or wait.
But reading too slow???
hybrid scarab
So the PM25 hardware just squirts 32 bytes over UART every 1sec, and if it sees a particulate change big enough it will speed up to something like 50x/sec (I can't remember the exact numbers)
thorny jay
Also, there is that thing where the value only change every 3 seconds or so. Was is with PM?
But in some test, I could see always the value repeating 3 times.
hybrid scarab
Reading too slow will, depending on your client:
- Slowly fill up your UART receive buffer with garbage/unread data
- If it's a FIFO buffer you will get the data in order, as long as you keep up with reads- so "old" data
- If your buffer overruns... AAAHHH
- If there's no buffer at all, you will get data as you request it
Could be the hardware itself doesn't update even as fast as the value is propagated out to UART I'm not sure
But the PM25 doesn't know you're asking for data, it's just diligently throwing it out there into the abyss
thorny jay
I read some documentation, and it was like: "Yeah, we send data every seconds, but in reality it only change every 3 seconds".
hybrid scarab
And that's half the problem with people running the sensor for six weeks and slowly filling the receive buffer on the Pi with unread data until you get a short read and it blows up
thorny jay
There is a reset pin... but not in use on the EnviroPlus. But I have the breakout now, so I could see if that part is helpfull.
hybrid scarab
I have PMS_RESET documented as being on D11 in the schematic
(though it might be an old schematic)
thorny jay
OMG, maybe now I start to prefer the I2C version? Maybe that one overwrite the memory with the lattest value and when you read, you always read the lattest.
hybrid scarab
Yeah with i2c you're explicitly asking for the contents of the registers, and clocking them out one-by-one as and when you ask for them
thorny jay
Wait... need to check if your code is using it. But in Adafruit simpletest, I remember it was not put to use.
hybrid scarab
So it makes not a lick of difference if you're too fast or too slow
thorny jay
I believe the issue is that it is asymetrical UART. You are not asking for the value, you are flooded with values.
hybrid scarab
Yep!
Like sticking your head under a waterfall for a drink of water, versus turning on a tap
This is broadly why:
- I hate UART
- People have to contrive horrible packet formats to make it work, including checksums and SoF bytes
Oh and
3. Inevitably some design-for-UART format ends up implemented in i2c and makes me cry
thorny jay
That is to make migration from UART to I2C more easy... they keep the same format.
hybrid scarab
Though I will yield that having a checksum is probably still useful in very extreme cases over i2c- ie: if you want to be 1000% sure the data you read is perfect and hasn't been bitflipped
But IMHO 99.99% of all sensor implementations wont care about a bitflip because it will be too incidental for any sensible algorithm to care
hybrid scarab
If you're reading temperature from a sensor at 1000Hz and you need to know in 1/1000th of a second if the temperature has just gone from 3C to 131C and isn't a result of a bitflip then... yes.. you're gonna care about checksums π
thorny jay
Here for PM... I guess filtering creazy data, taking some kind of average, ... that would be the way to go.
hybrid scarab
Yeah the accuracy on the data is questionable anyway
hybrid scarab
Though with UART you have to have a SOF and a checksum because you could be reading total random garbage
It's the only way to tell whether you've got a PM connected, or some random GPS module
And I really need to go to bed π
thorny jay
"RuntimeError: SDA or SCL needs a pull up" ... Maybe I should have put 4.7K rather than 1K ???
hybrid scarab
1K should work, in theory
As long as your resistor value isn't so low that the chip can't pull the pin low
hybrid scarab
I think the rule of thumb might be "use the highest value you can get away with before your signal turns into mush"
But that probably only applies if you're worried about burning tiny amounts of power or something
Like- if you put a 10M resistor on there then it might take so long for the line to go to 3.3V when it's released that your device has tried to pull it low again before it finishes
thorny jay
Yeah, I have read 2 or 3 page that google gave me to read... and yeah, gave 1K a try... now I am going to try 4.7K. If that also fail... then... I will start to instigate further.
Not even sure if that is a Pull Up resistor or some magic adaptation between what the Pi does and what the Feather does.
hybrid scarab
I've done some slightly crazy things with i2c (running it over long cables for Flotilla) but that's about as far as my "engineering" creds will take me
thorny jay
I NEVER had to worry about that because every board I use do the right thing.
hybrid scarab
Yeah Feather not having onboard pullups is weird to me, but I guess it makes those pins multi-purpose
Also no pre-populated resistor will ever be the right resistor for some specific edgecase, I guess
slender iron
@onyx hinge what do the _reconstruct calls for rgbmatrix and sharp do?
ooooh, folks will be happy. terminalio is 6k
onyx hinge
@slender iron when the framebuffer displays are initially created they allocate heap memory. "reconstruct" lets them recreate them as supervisor allocations on soft-reset.
slender iron
ok, I think I fixed a bug then where the move memory would exit early if there was no terminal
onyx hinge
(if they were ever allocated as a "built in" display it would probably not be necessary as they would already be supervisor allocations)
oh cool, thanks for fixing my bugs
slender iron
np, I think it's very rare that it'd happen
onyx hinge
it's in that "turn off terminal" PR?
slender iron
yup
onyx hinge
This would break any demo that used terminalio.FONT for displaying ASCII text and numbers, they would have to use external fonts instead, but only on ja and ko builds?
onyx hinge
OK, making sure I understood the impact of this change.
idle wharf
I'm on adafruit/main, built and flashed espressif_saola_1_wrover.
Should I see a network module in help('modules')?
crimson ferry
@idle wharf I think it was an artifact from MicroPython https://github.com/adafruit/circuitpython/pull/3164
idle wharf
@crimson ferry Sorry, I worded that poorly. I don't see a network module. I assumed with the native_wifi merge for an esp32s2 port I would see one now. I wonder if only the IDF\C part is implemented so far and not the CP Library...
Actually, looks that is exactly what that PR says... "coming soon" π
Thanks @crimson ferry !
crimson ferry
I'm looking at Scott's fork now to see how it manifests
anticipation is high for playing with the new networking stack π
idle wharf
I'm borderline giddy
crimson ferry
same. I like the finer granularity in the modules and eventually the libraries
idle wharf
@crimson ferry I think I missed the information on that. How will the modules be changing?
crimson ferry
@idle wharf Looking through the native-wifi branch, looks like there are new modules, see the shared-bindings and common-hal in that branch, e.g., changes committed here https://github.com/tannewt/circuitpython/commit/a97ed1af0e940bc507ba0352efaf7f91227b79d8
The finer-grained library re-factoring is just my impression from some commentary in the Deep Dives.
idle wharf
Gotcha... As a C dev, you're able to see deeper in the dives than I can. π
crimson ferry
uh, it's actually quite alien to me (not a C dev; not a Python dev really either)... it's taken me over a year working with CP to be able to navigate those directories and grok any meaning from it at all
slender iron
@idle wharf the native_wifi branch adds wifi and socketpool modules
idle wharf
@slender iron AH, so I need to build the native_wifi branch in your fork, correct?
slender iron
yup!
will hopefully get a PR out tomorrow for the basics
it'll require my version of requests I think too
idle wharf
Got it, thanks! I'll try out a built tonight on the Wrover and the FeatherS2
crimson ferry
@slender iron Seems the library APIs will change? Will the break be 5.x vs. 6.x (different libs in the bundles)?
slender iron
they will a bit but they'll apply backwards
all of the existing native networking is disabled
we have a python wiznet driver anyway
crimson ferry
we never used that anyway did we? not since 3.x or 4.x with esp32spi or wiznet
(libraries)
slender iron
not really no. anything we've used heavily is all python side
manic glacierBOT
forest musk
@tulip sleet I've just filed the bug posted by the bot (previous post)... not sure whether I'm compiling the firmware with incorrect parameters or what... but it seems that BLE cannot bond because of an issue with NVM.
tulip sleet
thanks, I will take a look tomorrow
forest musk
thansk!
manic glacierBOT
orchid basinBOT
hybrid scarab
My kingdom for a CircuitPython library manager π
Though tbf I can search the "lib" folder in Windows
solar whale
@hybrid scarab this may be of interest https://learn.adafruit.com/keep-your-circuitpython-libraries-on-devices-up-to-date-with-circup
lone axle
+1 for Circup, I've started using it more recently. If you are familiar with using pip or npm for other projects to install and keep updated the requirements then you'll probably be able to pick up Circup quickly and use it for installing and updating Circuit Python libraries on your devices.
hybrid scarab
Aha thank you @solar whale
thorny jay
I guess circup is better... but I have a small python script for that. I download the new bundle, unzip, create a few folder, then run my script. It check the file/folder I have in D:\ (CircuitPython drive) and copy those from the bundle. Then I have my folder full of all the file I should copy over. So it assume that my board already has the set of usefull library, and get me the latest. Not fool proof, but I used it a lot recently: https://gist.github.com/dglaude/f5f3423288861bc44fe758212dbaac61#file-update-py
Initially the Gist was about PyPortal, and I created a list of file/folder I wanted. But then it became a way to get the latest version of all thing.
ionic elk
The bdmicro_vina_m0 atmel-samd board is failing CI for me with just 76 bytes over. Probably had to do with me adding a new translation. Seems to me like no board should probably be that close without a little margin for error, but should I take something out or modify the translation somehow?
tulip sleet
@ionic elk I just fixed that yesterday, and gave it plenty more room; is this new?
tulip sleet
idle owl
@tidal kiln If you have the cycles, can you check this out? I'm assuming they don't have the right version of something, but I could be wrong. Either way, I don't have the cycles. https://forums.adafruit.com/viewtopic.php?f=60&t=168434
lone axle
I think it's most likely that they need to update to the newest version of Adafruit_CircuitPython_HID
tidal kiln
@idle owl @lone axle yep. agree. most likely. i responded and will cover it.
idle owl
@tidal kiln Much appreciated!
crimson ferry
sys.getsizeof would be really handy
idle wharf
Hey just a quick FYI on trying the native wifi branch from @slender iron's repo.
I successfully built and flashed the FeatherS2 and I could see the wifi module. Didn't see a Mac address nor did wifi.radio.start_scanning (not the right name) work. But I'm not sure I was doing it correctly and didn't spend much time looking at it.
On the Espressif WROVER, I couldn't get it to flash. I had a parameter error.
Anyway, nothing specific for anyone to act on, but I wanted to share the info in case it was a pattern someone else is seeing. I'll pull the branch and give it all another go tonight, be more methodical and take notes.
slender iron
thanks @idle wharf. I'll be poking it after lunch. splitting pulseio now
manic glacierBOT
crimson ferry
I had some now-buried questions in the #help-with- channel about memory usage of imageload vs. OnDiskBitmap: https://discordapp.com/channels/327254708534116352/537365702651150357/745374089954197644 Do I have a wrong intuition about this, or what I am I missing?
slender iron
looks
@crimson ferry this is a better place to ask implementation questions because the help-with channel is for more basic qs
OnDiskBitmap should use relatively little because it never loads the image into memory
it always fetches pixels from the disk to put them in the final bitmap portion that is output to the display
crimson ferry
does it continually fetch from disk on every refresh? if not, how can it be displayed without taking up enough memory for all the pixels and color depth?
slender iron
ya it does fetch every time. that's why it's slow
crimson ferry
wow, OK, but still I saw a big memory drop after exiting the with... I couldn't account for that
slender iron
ya, that seems weird did you gc.collect before doing gc.mem_free?
crimson ferry
no, did that a couple lines later after some clean-up and checked again and free mem was back up
slender iron
k, may be some intermediate memory then
crimson ferry
thanks, that helped
slender iron
np, Bitmap/OnDiskBitmap are really meant to be the two extremes
small in memory and fast vs large on disk and slow
slender iron
@silver tapir usb doesn't work for me either... I get a dmesg error saying cannot set config
solar whale
@slender iron just finished soldering up a v3 soala feather adapter -- The QWIIC connector works!.. I had a real struggle getting the microusb connector soldered w/o shorts..... Thanks for creating the board -- it will make things a lot easier to test.
solar whale
I used flux, but not enough I guess -- and definitely first next time!!
slender iron
π practice helps too and you are well on your way
solar whale
Yes -- this was great practice. And a good test of patience and resilience
solar whale
what is the purpose of the R/C parts you added?
solar whale
ah -- good.
manic glacierBOT
thorny jay
Hi @onyx hinge I have now the FeatherWing for M0 and M4 (actually, I am a bit confused because there is a different bonnet for nRF52840... but that's fine).
@thorny jay I haven't used the 64x64 matrix. The code should support the 5 address lines A through E, but the existing featherwing for feather m4 express would need to be modified to actually bring the signal to the header.
@onyx hinge Would love to get your feedback if you try it
I have identified where the E should be. But I need to figure out what pin from the Feather are in use for the [RGB][12] and other [ABCD] and what pin am I left with to solder E to it. It seems easy, I just need to make an informed choice.
onyx hinge
@thorny jay I think that you are free to choose from any pin for the "address pins". Before committing, I would test that the RGBMatrix constructor will accept your chosen pin.
manic glacierBOT
onyx hinge
It's the RGB and LAT (I think?) pins that are highly restricted.
onyx hinge
This is the one part of the Arduino code where some knowledge of the underlying hardware is required. rgbPins[] and clockPin must all be on the same GPIO PORT peripheral (e.g. all PORTA, all PORTB, etc.). The other pins have no such restrictions
OK, so it's RGB and clock ...
that bit is from https://learn.adafruit.com/adafruit-protomatter-rgb-matrix-library/arduino-library so it's a bit hidden. "beware the leopard"
prime cove
Is there a reference for which circuitpython features work on which hardware platforms e.g. the SAMD51(grand central) doesn't currently support RS485
thorny jay
@onyx hinge I added board.A1 and it did not complain. It did when height was still 32: "ValueError: 5 address pins and 6 rgb pins indicate a height of 64, not 32
" so there is a fair among of testing of the parameter. I guess now I need to solder that:
'''
matrix = rgbmatrix.RGBMatrix(
width=64, height=64, bit_depth=1,
rgb_pins=[board.D6, board.D5, board.D9, board.D11, board.D10, board.D12],
addr_pins=[board.A5, board.A4, board.A3, board.A2, board.A1],
clock_pin=board.D13, latch_pin=board.D0, output_enable_pin=board.D1)
'''
lone axle
If anyone finds themselves with a moment to spare today, I would be very grateful to get a quick review on this PR https://github.com/adafruit/Adafruit_CircuitPython_Display_Text/pull/88. This issue came from a recent change that was initially released yesterday, so it's only been in the bundle for 1 day I believe. But I do think it's likely going to be a relatively widespread issue so the sooner we can get the fix merged in the fewer people who will end up with the affected version of the library.
ivory yew
darn, I was gonna look but Jeff beat me to it. π
thorny jay
Hi @slender iron the recent discussion about rgbmatrix make me think about ESP32S2. If ESP32S2 is going to make a Feather, then choosing the pin so that it can respect the rgbmatrix constrain (to have various GPIO in the same register) would be great. It could even be better if the FeatherWing for M0+M4 or the FeatherWing for nRF52840 could work directly for the ESP32S2, that way there would be no need for yet another FeatherWing. I am clueless about the constain and if they would be the same for ESP32S2. But it would be sad not to verify if it will be possible before commiting to pin and pin usage? Cc: @onyx hinge
onyx hinge
It's above my pay grade @thorny jay
solar whale
@prime cove there is this https://circuitpython.readthedocs.io/en/5.3.x/shared-bindings/support_matrix.html but I'm not familiar with RS485 support -- is it supported on any boards?
onyx hinge
By which I mean, I'm not the one who makes such decisions. It's an idiom, intended to be humorous, and perhaps missed the mark
lone axle
Thank you Jeff and stargirl!
prime cove
@solar whale thank you, that was what I was looking for. RS485 settings are listed under the busio>uart documentation, but running them gave an unsupported error
onyx hinge
It looks like maybe esp32s2 and mimxrt10xx have at least some rs485 support, while samd, cxd56, nrf, stm do not. That's just based on quick searching within the source code and could be inaccurate.
prime cove
Thanks! Any idea why the samd does not? I believe it's supported in hardware, though even without it you just need to be able to toggle 1 or 2 GPIO pins and have a uart to get it to work with most interface chips.
solar whale
@prime cove not sure, but this looks ominous for SAMD https://github.com/adafruit/circuitpython/blob/main/ports/atmel-samd/common-hal/busio/UART.c#L72
and a bit more information here https://github.com/adafruit/circuitpython/issues/2354
prime cove
I suppose the best way for me to approach it would just be a wrapper class around the UART to handle DE and RE GPIO. I don't think being limited to 8 data bits is an issue; most 485 devices I've run into have configurable parameters or just use 8N1 or 8N2. Thank you for the help @solar whale and @onyx hinge
solar whale
good luck!
slender iron
@prime cove if you are comfortable in C you could expand the SAMD UART implementation
RS485 is a bit weird because its a specific form of a broader protocol (UART)
usually we have functionality enabled or disabled at the module level but this case is different
@thorny jay that's a good point! I'll chat with limor and philb about it
@onyx hinge want to chat CAN?
onyx hinge
@slender iron I'm around, if you want to go to voice or video you'll have to give me 5 though
slender iron
@ionic elk can you pick up https://forums.adafruit.com/viewtopic.php?f=60&t=168430
its a 405 feather
@onyx hinge that'll be easier and no rush
onyx hinge
OK I'll ping you back in a few
slender iron
k np
ionic elk
@slender iron I will check it out, thanks for the ping
slender iron
np, thank you
onyx hinge
@slender iron I'm back at my desk, do you want to DM-video, try one of the public voice chats (that didn't work well for us last time for some reason), or what?
slender iron
we can try amelia first
idle wharf
@slender iron @thorny jay I saw that thread about S2 Feather Pins...
Have you also mentioned this to @atomic summit ? I have his pre-release FeatherS2 and I know he's changed two pins in his production design.
https://unexpectedmaker.com/shop/feathers2-esp32-s2
onyx hinge
https://gist.github.com/jepler/e1fc3705242794e91d7b7c28c7422844 @slender iron here's the draft pyi of what canbus might look like
from pyb import CAN
while True:
print("I am the receiver")
can = CAN(1, CAN.NORMAL, prescaler=2, sjw=1, bs1=14, bs2=6, auto_restart=True)
can.setfilter(0, CAN.LIST16, 0, (11, 12, 13, 14))
while True:
while can.any(0):
_id, rtr, fmi, data = can.recv(0, timeout=10000)
print(str(_id), '1' if rtr else '0', str(fmi), str(data), sep=',')
#f.flush()
data = int.from_bytes(data, 'little')
print(_id, rtr, fmi, data)
In a package, I have an abstract class (the parent class) that should not be instanciated without overriding a function. Should I use @abstractmethod decorator and/or raise NotImplementedError() to indicate that? I may also want to have abstract @properties so that you cannot derive without defining (in my case width and height). I am really rusty on OO concept and their name... and clueless about the Python way of doing thing, and what exist in CircuitPython. Any hint is welcome. Or better, existing CP code that does that so that so that I can copycat.
For the record, I learned OO with Ada9X ... in a year where the value of X was still unknow. π
Ok, I am lost in slackoverflow... I will do a few test, and it is always possible to make comment on the PR. PM25 does use "raise NotImplementedError()".
slender iron
@thorny jay I think raising not implemented is the easiest way to do it
@onyx hinge can can messages be non-byte lengths?
onyx hinge
@slender iron not in micropython's implementation at any rate
slender iron
@onyx hinge here is where I'm at: https://gist.github.com/tannewt/89ee082de9f5a98675df901cdcfbe951
onyx hinge
@slender iron I think the "id" is some number that the sender and receiver agree on, I guess it could be said to identify the sender's address..
slender iron
is that what the filter matches on?
onyx hinge
yup
slender iron
ah
onyx hinge
@slender iron the Data length code is a 4 bit number, and only values from 0 to 8 may be placed in it. so yes, the messages are always whole bytes. Not just as a matter of implementation.
idle wharf
Hey @slender iron I've been doing the testing I mentioned I'd do last night. I can't get my builds of Adafruit/main or your native_wifi branch to flash onto the Saola and boot CP.
At first they would only flash if i left off --no-stub from the esptool command. I then updated ESP-IDF and now that isn't an issue. I can use make BOARD=espressif_saola_1_wrover flash and the board successfully flashes.
BUT, it still does not boot to CP, mount the drive or show up in /dev/tty.* the output on the external USB (the one on the wrover board) is here: https://gist.github.com/askpatrickw/36be63d1b96fd57b043c0d0065c87af3
the last line is I (627) cpu_start: Starting scheduler on PRO CPU.
Worst part is nightly builds from S3 work. AND my builds for the FeatherS2 work. I'm stumped... Any suggestions?
slender iron
@idle wharf I'm not sure why CP isn't starting
but USB is definitely broken atm
gonna look into that right now
idle wharf
OK.. well that you are aware of a related issue makes me feel better. I'll wait and try later. tx !
slender iron
π keep an eye on the commits
slender iron
do we have a way to check that we don't create a usb descriptor with too many endpoints?
onyx hinge
@slender iron no, I don't think so
When I did cleanup of some of the #ifs and makefile settings around USB configuration I think you mentioned it was a "would be nice"
is it what bit you this time?
slender iron
yup, I enabled MIDI
onyx hinge
ah
slender iron
and now it's disabled again π
onyx hinge
how would one check the number of supported endpoints? does tinyusb provide it as a macro or anything?
slender iron
no, I think we'd need a port setting or something
Β―_(γ)_/Β―
I'm not going to worry about it now
thorny jay
@idle owl For the PM25, maybe it would be more readable if I split the simple test into two copy, one purely I2C and one purely UART. Each will be more simple and not make unecessary include.
idle owl
@thorny jay I believe we'd like a self-contained simpletest specifically for the simpletest file. If you'd like to do something different as an I2C or UART example, that would be something we could consider.
I'm heading out for the night. I will reply to any further discussion tomorrow. Have a good rest of your day!
thorny jay
I have been overwelm by the number of comment and the interface of GitHub to resolve that. I hope I did not mess too much, I did everything "online".
Ok, I think I fixed everything.
idle owl
@thorny jay You didn't mess anything up, it's simply little tweaky things the doc rendering needs.
thorny jay
Doc, I need to check your Sphynx thing. Have a good night.
slender iron
@idle wharf here is my requests version: https://github.com/tannewt/Adafruit_CircuitPython_Requests/tree/http11
just pushed a new native_wifi with fixed usb
idle wharf
@slender iron Awesome! Iβll check it out tonight. First up. Go Flyers! (Hockey sport ball)
manic glacierBOT
turbid radish
?serverinfo
digital shoreBOT
adafruit
thorny jay
@idle owl maybe this issue will be interesting for you. I wanted to run the Pedometer learn guide https://learn.adafruit.com/clue-step-counter-st-lsm6ds33 on the CLUE. But the step count was missing from the library. Now the library https://github.com/adafruit/Adafruit_CircuitPython_LSM6DS has been fixed and splitted from library to package. So I wanted to manually do things... but now I discover that the CLUE library also used LSM6DS and so will fail as soon as it is release: https://github.com/adafruit/Adafruit_CircuitPython_Clue . This is where I give up (for the moment?)! There is some library dependancy and release management that will be required. Possibly, a lot of CLUE related learn guide will stop working when any of those two library change. Maybe there is a way to smooth the transition, like having two different library, the old one, and the package. And not remove the old one until a lot of things have been migrated.
thorny jay
I did give a "try" and catching import error, but I don't find a way with "import" "from" "as" to make the package look like the library or the library look like the package. It might be possible, but I failled. If there is a solution, then I am interested for PM25 and IS31FL3731, as the same kind of thing take place there.
onyx hinge
hm I have to remember how to connect to the debug interface on the SAM E54..
slender iron
I'm on a roll
slender iron
anyone want to help test led matrix improvements here: https://github.com/adafruit/Adafruit_CircuitPython_IS31FL3731/pull/32 ?
tulip sleet
@slender iron are you declaring all functions in .h files as extern now, for gcc10? I thought you were doing that for globals, but not necessarily functions. I think extern is the default for non-static functions.
tulip sleet
Use of extern with C functions:- By default, the declaration and definition of a C function have βexternβ prepended with them. It means even though we donβt use extern with the declaration/definition of C functions, it is present there. For example, when we write.
slender iron
variables and types I think
tulip sleet
I see stuff like that a bunch of places. ... OK, I just did a merge and a bunch of extern's were added
slender iron
to functions?
tulip sleet
yes, in _bleio
slender iron
definitely have been adding them. not sure where
but I've been chasing compiler complaints so they must be needed
tulip sleet
so the point is it's not necessary for functions, but is necessary for globals. On the other hand, it's a lot easier to be consistent.
i have NOT tried to build with gcc10 yet. I'll take another look. I just wanted to know if you had made a deliberate decision. tnx
slender iron
π
thorny jay
anyone want to help test led matrix improvements here: https://github.com/adafruit/Adafruit_CircuitPython_IS31FL3731/pull/32 ?
Thanks @slender iron Actually, I tested all the driver with my hardware. It give strange result as they don't use the same height and width, but the strange result are consistant (I should video record the original output and the new output, but mostly that part is unmodified). My last fight was with the documentation and how the documentation of package should be done. Some library only list one file: https://github.com/adafruit/Adafruit_CircuitPython_LSM6DS/blob/master/docs/api.rst where other list multiple: https://github.com/adafruit/Adafruit_CircuitPython_CircuitPlayground/blob/main/docs/api.rst . I don't really know how to see the output of Sphynx to see the result, so I made blind change.
slender iron
ah, ok great!
doc build instructions are here: https://learn.adafruit.com/creating-and-sharing-a-circuitpython-library/sharing-our-docs-on-readthedocs
solar whale
I keep running into this -- is there something "different" about the protomater submodule? or is it this PR -- I don't see this with other PRs ```[Jerry-desktop-mini:/Volumes/CircuitPythonBuild/circuitpython] jerryneedell% git fetch origin pull/3237/head:pr_3237_test
remote: Enumerating objects: 49, done.
remote: Counting objects: 100% (49/49), done.
remote: Total 62 (delta 49), reused 49 (delta 49), pack-reused 13
Unpacking objects: 100% (62/62), done.
From https://github.com/adafruit/circuitpython
- [new ref] refs/pull/3237/head -> pr_3237_test
Fetching submodule lib/protomatter
fatal: remote error: upload-pack: not our ref 2408e9c033f5ec050967b1592b29a950a831d6c2
fatal: the remote end hung up unexpectedly
If I repeat the command -- it does not complain and the branch exists ....
idle owl
@thorny jay We are supposed to keep up with changes like those. The appropriate thing to do is file an issue on the library with the dependency failure (so in this case CLUE), and mention it needs to be updated. This sort of thing will happen, and exactly like we have to update guides associated with particular libraries when we make breaking changes to a lib, we should be updating libs using other libs when we change the lib. Adding a way to have different versions of a library is not a good solution, it will lead to confusion and support issues.
onyx hinge
@solar whale I wish I knew
that PR has unintentionally modified the ref of lib/protomatter multiple times. As far as I can tell, the ref 2408e... was the ref of lib/protomatter during one of the intermediate commits but not at the tip of 3237's branch.
solar whale
I think I'll just wait for the PR to be merged ...
onyx hinge
if you want to do testing you can pick up uf2s from the PR's actions .. do you know how to do that already?
.. we love it when you do testing
solar whale
yes -- good idea.
onyx hinge
I wish I could figure out what types things into the REPL of every circuitpython board I plug in. I've removed modem manager. I've failed to identify anything else udev or systemd would be "running on" the device. But I always get some characters repeated back (slightly munged) from the boot-up messages into the REPL. "D Ado-reload is o"
"oh linux, you try so hard"
The spew actually occurs when I open the terminal device, whether or not it's a traditional terminal program
e.g., true < /dev/ttyACM4
onyx hinge
does this happen to anyone else?
mental nexus
I get βKeyboard Interruptβ randomly and gibberish on the REPL input after I leave a board alone for a while and work on the computer (older iMac). Some random USB signals that check in on devices on the bus? But if doesnβt sound like what youβre getting.
onyx hinge
@slender iron what do you think of "Listener" vs "Queue" vs "Receiver"?
I kinda like CAN.listen() -> Listener
slender iron
@onyx hinge fine with me
onyx hinge
What if .next() raises StopIteration in the case of timeout? That would let you process messages until there's a gap of 'timeout' between messages.
onyx hinge
oh, that permanently makes t he object not iterable anymore?
if Listener.__iter__() returns self, and Listener.next() raises StopIteration for timeout, couldn't you repeatedly loop on the same listener?
slender iron
ya, I suppose
onyx hinge
This is the structure I'm imagining, but maybe it is not good structure:while True: for message in listener: process message do other stuff when gap in messages
NORMAL = auto()
LOOPBACK = auto()
SILENT = auto()
LOOPBACK_SILENT = auto()``` should this be one enum, or 2 bools? "silent" and "loopback" as bool properties(and "silent" is the same as not giving a TX pin, ... I think ?)
slender iron
the alternative is:
for message in listener:
if not message:
print("timeout")
continue
# process message
two bools seems better to me
easier to document too
onyx hinge
if message is None:
do other stuff when gap in messages
continue
process message
(yeah I just said the same thing as you)
slender iron
I don't expect the "do other stuff when a gap" is useful
because a time out is the end of a gap potentially
onyx hinge
then what would you write inside 'if message is None:' -- just continue? break? sleep(1); continue?
slender iron
true, if you have nothing to do then you don't need to set a time out
"do other stuff" would probably do before the if
onyx hinge
oh if you set timeout=0 or timeout=None (whichever is standard circuitpython?) then you'd never get a None return so you don't have to deal with it?
slender iron
right, instead of using stopiteration on timeout
onyx hinge
I'll keep thinking about this, but now it's closing in on pizza making time
I do appreciate having your slightly revised doc to work from
idle wharf
My guess is "because memory"... but I'm curious why CP doesn't implement .__dict__ on objects. In Python its one of my favorite hacky ways to figure out what I'm doing.
onyx hinge
could you fake it? Something like (untested / incomplete) ```class DictProxy:
def init(self, obj):
self.obj = obj
def __getitem__(self, idx):
return getattr(self.obj, ix)
def __setitem__(self, idx, value):
return setattr(self.obj, idx, value)
# etc
BTW WIFI is working on my Espressif ESP32-S2 Saolo as advertised !
(sysname='esp32s2', nodename='esp32s2', release='5.0.0', version='5.0.0-beta.0-2632-g80b15f6b3-dirty on 2020-08-19', machine='Saola 1 w/Wrover with ESP32S2')
Start Scan
SSID: Jenga
SSID: henther
SSID: xfinitywifi
SSID: xfinitywifi
SSID: PS4-FDBC7527F6EA
SSID: FBI_Van
SSID: MOTO4ACA
SSID: ARRIS-E5AD
SSID: WaveG Public WiFi
SSID: TP-LINK_3D37
SSID: INTRANET
SSID: xfinitywifi
SSID: xfinitywifi
SSID: The_GreenDragon
SSID: Pawnee Commons Wifi
SSID: KeepYourWifiInLine
Scan Completed
WIFI Connected: True
IP: 192.168.1.27
GoogleDNS Ping Response: 0.008
@onyx hinge I code more procedurally (linerally?) whatever the word is, so I would tend to try and make a function. Are you saying, I could define a class and that class would take an object and return the same object with an added property of .__dict__ ?
idle wharf
WiFi also working on the FeatherS2 as advertised...
(sysname='esp32s2', nodename='esp32s2', release='5.0.0', version='5.0.0-beta.0-2570-gd615cc9fd on 2020-08-19', machine='FeatherS2 with ESP32S2')
Start Scan
SSID: henther
SSID: FBI_Van
SSID: Turn On, Tune In, Drop Out
SSID: WaveG Public WiFi
SSID: TP-LINK_3D37
SSID: INTRANET
SSID: xfinitywifi
SSID: comcast-gw-sea-guest
SSID: comcast-gw-sea-2g
SSID: Crouton
SSID: The_GreenDragon
SSID: Pawnee Commons Wifi
SSID: Ko_Olina
SSID: JP
SSID: Medialink_1E7250
SSID: xkcd910
SSID: Smartypants
Scan Completed
WIFI Connected: True
IP: 192.168.1.28
GoogleDNS Ping Response: 0.013
@onyx hinge I guess I should add... I know I could make it happen like you mention. I was curious as to the thought behind not including it.
onyx hinge
I think you'd have to ask around at Micropython, it seems to be a decision made pre-circuitpython. Probably "it's not needed, it constraints implementation choices, leave it out". Except there also seems to be an option for a read-only or copy __dict__ property in the source of circuitpython. curious.
https://github.com/micropython/micropython/pull/1757 doesn't really seek to make it clear to outsiders what the trade-offs are
idle wharf
There is a nice little example though in the linked issue which is still open. I tried this on my FeatherS2.
print(dir(wifi.radio))
for p in dir(wifi.radio):
val = getattr(wifi.radio, p)
print(p, val, type(val))
Output
['__class__', 'connect', 'enabled', 'ipv4_address', 'mac_address', 'ping', 'start_scanning_networks', 'stop_scanning_networks']
__class__ <class 'Radio'> <class 'type'>
connect <bound_method> <class 'bound_method'>
enabled True <class 'bool'>
ipv4_address 192.168.1.28 <class 'Address'>
mac_address None <class 'NoneType'>
ping <bound_method> <class 'bound_method'>
start_scanning_networks <bound_method> <class 'bound_method'>
stop_scanning_networks <bound_method> <class 'bound_method'>
Anyway, fun stuff to research. Thanks @onyx hinge !
thorny jay
Is there an easy way to check all the requirements.txt from all the repo part of the library bundle or community bundle and see if they make a refrence to "adafruit-circuitpython-lsm6ds"? Then check in https://github.com/adafruit/Adafruit_Learning_System_Guides if guide would be affected by the release of 4.1.0 of LSM6DS?
lone axle
I just made a script that searched learn guide repo for projects using a specified library last week. I think it could help with that
thorny jay
With library restructuration into package, the import and/or usage is different. I will trigger that problem in PM25 and in IS31FL3731 ... so I am pretty much interested into detecting cascade effect of such change.
lone axle
My script is in this issue: https://github.com/adafruit/Adafruit_Learning_System_Guides/issues/1209 when run inside the learn guide repo it will search for all projects that contain python files and then search all of those python files for adafruit_button in my case, but could be modified to any other library, Or could also search for constructor use or something beyond the import statement if desired.
frosty prairie
moved the question to #help-with-circuitpython .
lone axle
@frosty prairie the #help-with-circuitpython channel is probably the best place for this since it's not related to development of the project. It is fine to ask about questions micropython, but I will say there is likely a lot more experience with Circuit Python specifically than with Micropython amongst the community. So it may take a bit longer and be harder to get micropython specific things answered.
thorny jay
Hi @gilded cradle I have split into separate files the library IS31FL3731. So now I have copyright statement for each separate board. So I would like to know the board you added to the library, and those already present and done by Tony DiCola. I have the feeling you worked on the Bonnet because those imply Blinka work to implement on Raspberry Pi. I'll try to check the GitHub history, but with major rework like black or pylint, it is hard to use blame and know what lines are coming from you. My PR is here: https://github.com/adafruit/Adafruit_CircuitPython_IS31FL3731/pull/32
slender iron
anyone want to do a release today?
gilded cradle
@thorny jay I think I just worked on improving overall functionality, so I'm not particular how you split it up.
thorny jay
It is split by board. I keep both of your name in every file, except the two Pimoroni board I added to the library.
gilded cradle
Sounds good
thorny jay
@stoic parrotMelissa If I want to write a library to support Pimoroni 16x16 Unicorn Hat HD, should I use framebufferio like you do for matrix LED? I am a bit confuse between the framebuf, the pixelbuf and maybe other.
gilded cradle
I'm not sure how that's setup hardware-wise. I believe that's for the Pi and there currently is no PyPixelBuf for the Blinka. I would think it would be neopixels though.
thorny jay
I think I need a big buffer because what Pimoroni does is to use numpy and basically it is SPI transfer of the buffer.
gilded cradle
ok, maybe use the neopixel spi library then
thorny jay
No it is not neopixel. And yes, it's for the Pi but the goal is to use that from a Feather.
thorny jay
I have already used the neopixel library for Unicorn pHAT (4x8) and Unicorn HAT (16x16). But I made no driver no nothing as the code from Adafruit is sufficient... except if I could use the fact that the geometry is know to better handle the matrix.
But the Unicorn HAT HD does not use NeoPixel nor DotStar, they use normal RGB LED (non addressable) and they have their own electronic to drive that. So they implement an SPI protocol that I need to implement in Circuit Python.
gilded cradle
Sounds like it's similar to an RGB panel which we don't have support for yet
thorny jay
So it pretty much look like RGB matrix for which there is support in CircuitPython: "RGB LED Matrices with CircuitPython". When you say "we don't have support for yet" you mean in Blinka? I am not really interested in the Pi or SBC. I am interested in driving Pi hardware from an MCU.
They even have Ubercorn and you can daisy chain those one behind the other to make larger board.
gilded cradle
Yes, I meant Blinka. I thought that's what you were referring to since you said it's a HAT.
thorny jay
Maybe you have not seen the stuff I have done so far. From an MCU (Clue or Feather) I controled: Blink! + Scroll pHAT HD + LED Shim + Unicorn pHAT + Unicorn HAT + the RGB LED and segment LED of the Rainbow HAT.
I am pretty sure all of those would also work on a Pi... I mean on a Pi with Blinka and CircuitPython.
But that is not really very interesting as Pimoroni already provide Python Library.
gilded cradle
Ok, I hadn't and I'm understanding now. Maybe look at https://learn.adafruit.com/rgb-led-matrices-matrix-panels-with-circuitpython then.
thorny jay
Ok, I'll try to understand the framebufferio that seems to be used. The goal is to be able to use DisplayIO and all the other stuff on top. So I need to provide the right API so that it is transparent.
I hope to have it all in Python as I don't want to go into the "firmware".
gilded cradle
Yeah, I understand.
slender iron
someone want to help get this into pypi: https://github.com/adafruit/Adafruit_CircuitPython_TC74 ?
idle owl
@slender iron Can you file an issue and assign it to me on the TC74 repo? The docs aren't working, Actions isn't setup, there's a lot of other things doing on. PyPI should happen automatically once all of that is working.
idle owl
@slender iron PyPI creds do not need to be added. They're global now. It's everything else that needs to happen.
slender iron
haha, ok. that's why I'm not doing it. I'm behind the times
idle owl
@slender iron For the community bundle, do we recommend running cookiecutter and so on before submitting?
The few I checked have our CoC in them, so I assume it was run on those. But I didn't check all of them.
slender iron
ya, I think it's a good idea
not a requirement though
I think my sharing guide recommends it
idle owl
Alright, that's exactly what I was about to say to someone. But second guessed it. Thanks.
slender iron
np, thank you
simple pulsar
Are there any known issues with using PWMOut after AnalogIn on a pin (with deinit()) on an nRF52840 based board like the CLUE? It doesn't seem to work for me.
manic glacierBOT
idle owl
@slender iron Can you review this right quick? You have to make changes for Actions to fire for the first time. I found a couple of things to fix. https://github.com/adafruit/Adafruit_CircuitPython_TC74/pull/2
slender iron
yup
idle owl
Thanks.
@slender iron Ok, RTD is set up, Actions is running properly, and it's uploaded to PyPI.
slender iron
thank you!
silver tapir
I'm as usual checking to see if there are strings-to-translate to spanish in Weblate. I found an error message that might have some extra spaces. Not sure if this is a bug or not.
https://github.com/adafruit/circuitpython/blob/main/ports/atmel-samd/common-hal/pulseio/PulseOut.c#L104
The string itself is a line of text, the \n, and then 37 spaces, and then the second line. Not sure if the spaces belong there, can someone please confirm?
spice crypt
I'm facing a very difficult bug (I think), with I2CPeripheral. Most packets I intend to send are fine, but there are some combinations that will refuse to send with r.write()
This fails to send: 0xa8, 0x2f
Saleae logic output:
Setup Write to [0x0B] + ACK
0x14 + ACK
Setup Read to [0x0B] + ACK
<blank>
This sends fine: 0x67, 0x30
Saleae logic output:
Setup Write to [0x0B] + ACK
0x14 + ACK
Setup Read to [0x0B] + ACK
0x67 + ACK
0x30 + NAK
spice crypt
To expand on my i2cperipheral problem above, the ranges of ints that send are separated by a span of 128.
0-127 -> Can send
128-254 -> Cannot send
255-511 -> Can send
etc.
Perhaps just a quick question. What function can I use in the C source files to be able to log to the REPL for troubleshooting purposes?
BTW, I've tried v4.3.1, v5.3.1, and latest v6, as well as on M0, and M4
spice crypt
Reproducible code:
import board
from i2cslave import I2CSlaveregs = [0] * 16
index = 0
reg_check = 0while True:
with I2CSlave(board.SCL, board.SDA, (0x0b, 0x41)) as slave:
while True:
r = slave.request()
if not r:
# Maybe do some housekeeping
continue
with r: # Closes the transfer if necessary by sending a NACK or feeding the master dummy bytes
if r.address == 0x0b:
print("address 0x0b")
if not r.is_read: # Master write which is Slave read
b = r.read(1)
# if not b or b[0] > 15:
# break
index = b[0]
b = r.read(1)
if b:
regs[index] = b[0]
elif r.is_restart: # Combined transfer: This is the Master read message
print("Restart")
n = r.write(int.to_bytes(reg_check,2,'little')) #FAILS TO WRITE PACKETS AFTER 127
reg_check +=1
#else:
# A read transfer is not supported in this example
# If the Master tries, it will get 0xff byte(s) by the ctx manager (r.close())
spice crypt
I see it's getting stuck on the following in: ports/atmel-samd/common-hal/i2cslave/I2CSlave.c
if (!self->sercom->I2CS.INTFLAG.bit.DRDY) { return 4; //changed this to return 4 for debugging }
slender iron
@spice crypt to print to serial from C do mp_printf(&mp_plat_print, <regular printf args>);
silver tapir
I only have a Saola wrover, and not a wroom. Are they physically the similar?
I want to know if I can use the image that illustrates the wrover in the downloads page, as the image for the wrover.
mental nexus
My PR is failing for "Test Documentation Build"
/opt/hostedtoolcache/Python/3.8.5/x64/lib/python3.8/site-packages/autoapi/extension.py:133: RemovedInAutoAPI2Warning: autoapi_template_dir will be expected to be relative to conf.py instead of relative to where sphinx-build is run
344
345
warnings.warn(
346
/opt/hostedtoolcache/Python/3.8.5/x64/lib/python3.8/site-packages/recommonmark/parser.py:75: UserWarning: Container node skipped: type=document
347
warn("Container node skipped: type={0}".format(mdnode.t))
348
349
Warning, treated as error:
350
/home/runner/work/circuitpython/circuitpython/shared-bindings/_typing/index.rst:59:duplicate object description of _typing.ReadableBuffer, other instance in shared-bindings/_typing/index, use :noindex: for one of them
351
##[error]Process completed with exit code 2.
I suspect it's something to do with the documentation but I'm struggling to figure out where is the issue in my code.
Any hints on how to decode the error message?
Oh, no! Just realized I have 41 files changed.. Ugh!
thorny jay
You have to locally install Sphynx and see the result there.
I only have a Saola wrover, and not a wroom. Are they physically the similar?
I want to know if I can use the image that illustrates the wrover in the downloads page, as the image for the wrover.
@silver tapir I want to know which one to order to closely follow the streaming... I guess I need the wroom. Just one need a physical antenna added to the tiny coaxial connector.
silver tapir
Correct.
raven canopy
@mental nexus with the new stubs and AutoAPI, its a little esoteric to find some problems. I think I've found the issue; will comment on your PR if it pans out.
mental nexus
Thanks @raven canopy I appreciate any help.
crimson ferry
@thorny jay The Saola Wrover has the 2MB PSRAM, same price.
@silver tapir I think the only difference is inside the module, and the wording on the module.
(wroom left; wrover right)
and of course the "I" variants have the IPEX antenna connector and bypass the onboard antenna via different resistor placement
onyx hinge
@silver tapir If it's the string I'm thinking of, I asked @ionic elk about it, but it was a comment on a merged PR so I don't know if it'll get attention. Someone should simply fix it in the source.
mental nexus
@raven canopy Thanks for the help with the corrections on the Sphinx documentation!
lone axle
It looks like we might be subject to a new pylint check raise-missing-from
The newest CI actions are showing that they installed pylint version 2.6.0. The most recent actions that succeeded were 2.5.3
lone axle
This page lists the relevant changes: http://pylint.pycqa.org/en/latest/whatsnew/changelog.html including the new raise-missing-from check.
ivory yew
maybe a long shot, but since y'all have deal with tinyusb a lot: I've added tinyusb to my usb firmware - I got it to enumerate, but I can't seem to get it to send or receive usb midi data. :/
Any ideas?
ivory yew
nvm, I got it figured out. I had the clocks wrong.
thorny jay
I don't know who need to do what, but it would be great to release a new version of the CLUE library ( https://github.com/adafruit/Adafruit_CircuitPython_CLUE ) now that it contain a fix for the breaking change in adafruit_lsm6ds that is available in bundle "adafruit-circuitpython-bundle-6.x-mpy-20200821". So if next bundle could fix it, it will make the live of CLUE owner (Adabox) more fun. For support, right now they can stick with "adafruit-circuitpython-bundle-6.x-mpy-20200820" or earlier but not use the Pedometer example.
craggy galleon
Hey @tulip sleet how do you calculate the report_length in a USB HID Report descriptor?
craggy galleon
I think I worked it out after a bit of investigation. A high-level formula could be:
SUM OF (REPORT_SIZE * REPORT_COUNT) IN HID REPORT DESCRIPTOR
lone axle
Thanks @thorny jay I just made a new release for it that includes the latest fix.
thorny jay
Thanks to you. I have been frustrated that no Release Management was taken so that both lib with one relying on the other were not released simultaneously. I have not fix all of the Learn guide sample I found, but I made a few PR left and right, and except for the pedometer on CLUE, the other one are "blind change" based on the syntax change introduced (and only tested on one). So there is still a little bit of change to do. Most of those guide I cannot test, like Ukulele or what not. So I guess the author are best placed to check that.
lone axle
Is there some sort of change in a build/CI file somewhere when a library changes from being a single file to being a directory with multiple files inside? in order to get the PyPi uploading / installing to work properly specifically.
lone axle
ah yes packages vs. py_modules in the setup.py file.
thorny jay
It looks like we might be subject to a new pylint check
raise-missing-from
@lone axle Ok, now I understand what you mean. I made a minor unrelated change to a PR that was passing and now it is block on that pylint check. What should I do? (1) try to install pylint 2.6.0 on my side and make it happy? (2) make pylint ignore that line in my code? (3) change the pylint configuration to say it can ignore that (maybe that should be global thing for all the library)?
lone axle
@thorny jay I think the solution that pylint is looking for is to use from keyword when raising an exception from another one. In the case that I found we were catching a RuntimeError and re-raising it with a more helpful message. Using from like this got pylint to stop complaining:
except RuntimeError as run_error:
raise RuntimeError("Text length exceeds max_glyphs") from run_error
note the as run_error and then from run_error
I found this blog post to be fairly helpful in understanding it: https://blog.ram.rachum.com/post/621791438475296768/improving-python-exception-chaining-with
thorny jay
I used (3) by changing the rc file.
lone axle
I think it would be best if we do go ahead and use the from keyword syntax. I was unfamiliar with it before running into this pylint check. But it does seem like a positive thing. It can make the stacktraces a bit more detailed.
onyx hinge
@lone axle I think 'raise ... from' may not be supported in circuitpython. there's a test that is not run related to it: tests/run-tests: skip_tests.add('basics/exception_chain.py') # raise from is not supported
did you find otherwise?
oh, maybe it is. if so, nice!
Traceback (most recent call last):
File "zde.py", line 2, in <module>
1/0
ZeroDivisionError: division by zero
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "zde.py", line 4, in <module>
raise RuntimeError("ouch") from e
RuntimeError: ouch
jepler@babs:~$ src/circuitpython/ports/unix/micropython zde.py
Warning: exception chaining not supported
Traceback (most recent call last):
File "zde.py", line 4, in <module>
File "zde.py", line 4, in <module>
RuntimeError: ouch
``` sort ofso it'd be an improvement on blinka and a graceful degradation on circuitpython
lone axle
I can double check on it. We put it into the display_text library to resolve the pylint issue. I thought it was working as intented, but maybe I didn't take a close enough look.
onyx hinge
π
>>>
paste mode; Ctrl-C to cancel, Ctrl-D to finish
=== try:
=== raise Exception from None
=== except Exception:
=== print("Caught Exception")
===
Caught Exception
Looks like we should be good to go, sorry for raising an alarm unnecessarily.
lone axle
No worries
much prefer false alarm to having already broken something π
Unrelated to that specific error, but I wonder if it's worthwhile for us to target a specific version of PyPlint within the CI system so that we aren't taken by surprise when a new release drops and starts checking for new things.
onyx hinge
I'd check with Kattni, I think she had opinions. There are downsides both to being stuck on old versions, and to getting new versions without warning. I think that she consciously chose that we would get new versions without warning, but I may have events or identities wrong.
lone axle
Will do. For sure though it would be bad in other ways to let it get stuck on some ancient version as well. If it switches to manually we would have to be diligent about keeping an eye out for new ones.
It does look like I was mistaken in my belief that it makes the stack traces more detailed:
code.py output:
Traceback (most recent call last):
File "code.py", line 13, in <module>
File "/lib/adafruit_display_text/label.py", line 359, in text
File "/lib/adafruit_display_text/label.py", line 359, in text
RuntimeError: Text length exceeds max_glyphs
Looking into it more specifically now the only difference I can spot is that the File "/lib/adafruit_display_text/label.py", line 359, in text is duplicated when we use the Raise ... from syntax. Which does not seem all that helpful.
without Raise ... from I see the same stacktrace but with only a single copy of the line that is duplicated above.
onyx hinge
@lone axleyeah it only improves the quality of the backtrace on desktop python3
I think circuitpython accepts but ignores the syntax
lone axle
Maybe it would make more sense to disable that pylint check for our repos then unless / until we have that extra functionality supported.
onyx hinge
it does increase the size of the .mpy files though
lone axle
if we disable it inline with a comment in the file you mean?
onyx hinge
111 bytes: ```try:
1/0
except ZeroDivisionError:
raise RuntimeError("ouch")
137 bytes: ```try:
1/0
except ZeroDivisionError as e:
raise RuntimeError("ouch") from e
Oh, I see. the extra text to include it in the code.
onyx hinge
the comment shouldn't impact the size of the mpy file but I didn't check that
adding a bunch of comment lines to the top increases 111 -> 113 bytes, somehow. π€·
I wouldn't have expected that
lone axle
Yeah, I hadn't even considered mpy filesize. I know some of them get fairly tight on the non-express M0 devices already. I would probably shy away from making them bigger unless we get a clear gain from it.
onyx hinge
in current micropython raise ... from still doesn't improve the traceback message.
I don't know what the complexity of implementing it would be, but it might be considerable
raven canopy
for some history on pinning pylint to a particular version: https://github.com/adafruit/cookiecutter-adafruit-circuitpython/pull/69
we were pinned to 1.9.2 for like 2 years, and it was a beast of an effort to bring all the libraries up to date. i think kattni really wants to avoid that in the future, and handle new checks as they occur and through attrition.
onyx hinge
@raven canopy do we have any good options for turning off this new pylint check across everything in the bundle?
lone axle
Ah, I can understand that for sure. I just imagined 2 years worth of new pylint checks all hitting at once across hundreds of libraries and I can see how there would be a strong desire to avoid ever doing that again.
raven canopy
for the raise ... from we could probably ignore if circuitpython doesn't really support it. would be a "simple" adabot patch to all of the .pylintrc.
onyx hinge
hmmm now what
...
* bad-continuation and bad-whitespace have been removed, black or another formatter can help you with this better than Pylint
this makes modules which previously suppressed it via comment .. error. π¦ π¦
adafruit_ads1x15/analog_in.py:31:0: E0012: Bad option value 'bad-whitespace' (bad-option-value)
their release versioning annoys me...
onyx hinge
looks like about 150 instances of suppressing the tag via a comment-directive. There are more that suppress it via commandline, which seems to be OK (?)
raven canopy
the command line ones are in the Actions workflows for linting the examples, iirc.
they should all be the same across the libraries, so another "simple" update with adabot patching.
onyx hinge
yes I think that's right
tulip sleet
pylint 2.6.0 did update to a newer version of isort. I was finding that sphinx and pylint were requiring conflicting versions of isort [EDIT] before the update.
onyx hinge
oh dear, that's irritating.
tulip sleet
so now they DONT conflict
onyx hinge
oh that's good then
lone axle
Is it good (or bad) practice to change the name of an issue if/when it becomes apparent that it could be better and more focused on the actual root cause of the issue?
tulip sleet
@lone axle I change issue titles all the time; please do
@onyx hinge should I start doing wholesale reviews of the whitespace PR's?
onyx hinge
@tulip sleet if you feel like it. There are some that are failing and I'm going back through those. most are black wanting to make additional changes, some are import order errors that I didn't reproduce locally yet. But I think the "green" ones should be good to merge.
lone axle
I have also seen a few import order issues from the new version on some of the libraries
tulip sleet
I don't think I will bother to explicitly review. I will just merge the easy ones
onyx hinge
.. like this one https://github.com/adafruit/Adafruit_CircuitPython_LIS3DH/pull/61/checks?check_run_id=1018655807
others fail due to another pylint change, which I think will be adabot-patched(?) adafruit_mcp230xx/digital_inout.py:117:12: W0707: Consider explicitly re-raising using the 'from' keyword (raise-missing-from)
tulip sleet
i had those same import order things after fixing minor things in adafruit_ble
onyx hinge
thanks for getting the easy ones @tulip sleet
tulip sleet
@onyx hinge for failed PR's, I am assignging them to CP Librarians to review
solar whale
between black and pylint, when can we just write a paragraph describing the code and have it written for us. π
I guess makecode is close -- just draw pictures...
tulip sleet
i dunno, pylint is more like management than labor
solar whale
but at least pylint will find bugs, black just makes the code less readable....
tulip sleet
both are micromanagers π
solar whale
we have standards... they are low, but we have them...
slight mis-quote of Bette Midler
onyx hinge
the problem with code formatters is they disagree with me sometimes, and by definition I'm always right in matters of style.
thorny jay
Wat am I supposed to do with/against those new PyLint message? I make a simple change to fix a library, but I now need to re-order the import in example I did not even touch. π¦ Problem solved.
tulip sleet
it is a nuisance π
thorny jay
I am all for only checking the file I change and not the one I did not touch...
onyx hinge
What if we could generate shared bindings from pyi...
lone axle
can I force a device into safe mode? It's not connecting CIRCUITPY nor the serial console port. I suspect the code.py file on it might be causing trouble and auto-launching too fast to do anything about in normal mode.
I can double click reset and get to bootloader.
or maybe create a UF2 file that wipes code.py file?
lone axle
Managed to restore access by flashing back to 4.0.0 and my code crashed with ImportError: This release is not compatible with CircuitPython 4.x; use library release 1.x.x which allowed my device to connect normally so I can change code.py
crimson ferry
I think you can do it using microcontroller https://circuitpython.readthedocs.io/en/latest/shared-bindings/microcontroller/index.html
lone axle
I think I have to be able to access REPL or be able to change code.py somehow to use that though right?
crimson ferry
like maybe microcontroller.on_next_reset(SAFE_MODE)
oh, right
usually I get safe mode when there's an electrical disturbance, but... 
lone axle
ah, yeah I guess I could try to get it to brown out
would be nice to be able to make a "recovery" uf2 that is just like normal but will rename code.py to not_code.py (and similar for main.py) when it boots up.
that way as long as you can access BOOT drive you can clean anything bad out of code.py and then flash back to "normal"
crimson ferry
when I was trying out some disable write protection code I had to add some startup delay so I could get out if (when) things went wrong
or, might be worth it to put a 1-second startup delay in a boot.py
lone axle
Heh, earlier today when I was playing with something similar I did make sure to put a nice delay in so that I'd have ample time to connect and keyboard interrupt. Didn't realize I was in dangerous territory with the code I'm using now though. saved it without the delay and then couldn't do anything about it π
crimson ferry
maybe triple-click reset for safe mode, or double-click out of bootloader mode
arctic violet
What is the cheapest CircuitPython compatible board that I could wire 8 buttons to?
Is this thing not actually CircuitPython compatible? https://store.arduino.cc/usa/arduino-micro-without-headers
crimson ferry
@arctic violet Supported boards are here: https://circuitpython.org/downloads. Least expensive I've found are Adafruit Trinket M0 ($9, 5 pins; would need GPIO expansion), and Seeed XIAO ($5, 11 pins).
arctic violet
Thank you much
Saw page but was a little confused bc Arduino site says the MIcro was made in conjunction with Ada. Seems like a prime candidate for an inexpensive CP board?
onyx hinge
@lone axle My understanding is there's supposed to be a pattern to tap the reset button in to provoke safe mode, but I have had limited luck at it. It's slower than the double-press for bootloader. I also didn't find docs explaining it yet.
the function implementing it is wait_for_safe_mode_reset
tulip sleet
@lone axle the RGB LED goes yellow for 700ms during startup. During that time if you tap the reset button, it will reset into safe mode. It's a bit hard to react to the yellow light, so a slow double-click maybe more successful ( @onyx hinge )
@slender iron before you spend more time on the HCI _bleio review, I thought it might be good to have a chat (tomw) with you about its current state and how it got to where it's at.
slender iron
yup, that'd be good. gotta get the newsletter complete after the meeting
crimson ferry
ah, took me forever to find it @lone axle @onyx hinge it's in with individual board learn guides https://learn.adafruit.com/assets/75708
thorny jay
Someone may want to check my PR and release PyBadge library. The CLUE part is not working because of the lsm6ds.
Ok @lone axle did validate the change and Limor approved it. Now the lib need to be released: https://github.com/adafruit/Adafruit_CircuitPython_PyBadger/pull/38
lone axle
I can make a release for that.
lone axle
@crimson ferry thanks for that link. I had seen that flow chart before, but totally forgot that it included a safe mode path.
ionic elk
@tulip sleet what is process for obtaining a VID/PID, if you're a company looking to get your own legitimate codes? Maybe if we documented it, less people would ask.
tulip sleet
if you want to get them as your own co, you have to pay $5k to the USB org to become a member
that's prohibitive for individuals and some small co's.
ionic elk
Wow, that's pretty pricy yeah. No way I'm affording that for my stuff
lone axle
I think someone was discussing last week an 3rd party entity that allows smaller companies or individuals to obtain VID/PID without the fee, the website pid.codes seems to be at least one such service. Not certain if it is the one that was discussed.
or perhaps with a more reasonably sized fee. I don't actually recall if it was specifically stated as free or not. But certainly less than the $5k I believe.
onyx hinge
Yes, that was pid.codes. Their mission is to provide USB VID/PID pairs for open source hardware+software projects. They haven't approved any new PIDs lately (since april) but have a number of pending PRs so they may have shortage of administrative time. Tannewt had mentioned the possibility of reaching out to them but I didn't hear any result yet. I have a feeling a number of people would be happy to help them out.
idle wharf
FYI: Info on the USB-IF (Implementers Forum) Membership
https://www.usb.org/getting-vendor-id
ionic elk
@tulip sleet @slender iron I have a error reporting question. I just spent some time tracking down a bug with the LSE that I had very intentionally tried to mark with a reset to safe mode but it ended up just making it even harder to detect. What's a good way of crashing for something like a clock error that lets me send some kind of message about what happened?
ionic elk
Considering a new category to safemode for BOARD_CONFIG_FAIL, let me know if that sounds ok.
Would be an option for triggering if a board dev puts in the wrong settings for clocks and power, like saying there's an LSE when there isn't one, stuff like that.
slender iron
@ionic elk new safe mode is ok. usually we assume it's hardcoded correctly
ionic elk
I figured that was the philosophy but the clock bugs in this case are annoying and common enough that I really wanted to highlight them in particular. This is probably the third or 4th time I've had it come up.
Now it'll have a big flag to let you know you screwed up your clock settings
slender iron
np
thorny jay
I have an "electrical" question... can I use an nRF52 Feather Sense as a FeatherWing for another MCU? I mean, I would like to use the I2C sensor of it from an M4 Feather, just because I don't have any standalone lsm6ds33 to play with sand. [Why use the M4 and not directly the nRF52??? Because I only have the LEDmatrix connector for M4 pins] I did something similar with the CLUE using it's sensor from a Raspberry Pi.
solar whale
you'll have issues with the TX/RX and possibly with the power pins.
MCU feathers are not meant to be connected.
lone axle
Maybe you could connect just the I2C pins, and GND. And then power the Sense separately from the M4?
solar whale
I wonder if you can turn off the regulator on the Sense -- Ground EN -- then power the I2C devices via the 3V pin ??? Don't try this at home π
I guess that would still power the nrf52840...
modern wing
[well, where else are we going to try it? π ]
ionic elk
Finding myself stuck in cpu_get_regs_and_sp a lot today while debugging, anybody know what the cause of that might be?
might just be memory corruption or something but it seems like a funny place to get stuck
thorny jay
Thank you... so I need to give this a try... and I can try slowly adding one pin at a time until it work or the magic smoke get out of it. I am mostly interested into I2C things.
It seems that the challenging part is that both MCU will try to make 3V3 based on 5V.
modern wing
Good afternoon all you wonderful people. Happily lurking today, and I've updated the notes doc to reflect as such. 
turbid radish
Lurking
onyx hinge
<@&356864093652516868> The notes doc for today's meeting is https://docs.google.com/document/d/107jDI0oM4DuBHe4GuAPO_i5bVhOH-HDi1Q8YbNSjbak/edit?usp=sharing and we're starting very shortly
lone axle
Thank you
thorny jay
Also of interest is what code to run on the nRF52 to keep it "busy" but not disturbing. I had the same issue with the CLUE... what to run to make sure it does not do anything, I went for an empty code.py. But maybe I want to black the screen and turn off RGB and backlight.
tidal kiln
lurking
onyx hinge
if you can add your notes and "lurking" to the document it's super helpful!
stuck elbow
alurkin
onyx hinge
I didn't hear what Scott said at first so I wondered if it was a portmanteau of: camo + kawaii
turbid radish
raven canopy
lurking
blissful pollen
Just looked at her channel it is "Ka-Moo-Ee"
onyx hinge
Upcoming meetings: August 31 (Monday), September 8 (Tuesday) unless we want to make an executive decision to move it to CircuitPython Day September 9 (Wednesday)
turbid radish
Yes, should be 9/9/2020
solar whale
We should move it becasue 9/8 is NH Primary day ... π
onyx hinge
@slender iron I added it to In The Weeds π
inland tusk
lurking for status updates
turbid radish
Send projects, events, and interesting stuff via twitter hashtag #circuitpython-dev or @anne_engineer or via email at anneb(at)adafruit .com
idle owl
Definitely at least the second time.
old smelt
Lurking
onyx hinge
27 authors wow
- 0 issues not assigned a milestone
πΊ
We've got some remaining problems with pylint, in case anyone wants to help.
We need those cleaned up before we can release the affected libs
solar whale
pylint should have caught that.
onyx hinge
There's also the raise ... froms, I think there may be more problems are actually unknown currently, until a PR or release attempt occurs
idle owl
@onyx hinge Yeah I read through them. I'll see what I can do.
@thorny jay Are you text only for Status updates?
thorny jay
Maybe we need to check what PyLint would say if there was a change on all the library?
idle owl
It says you're lurking.
thorny jay
Fixed.
idle owl
Thanks for updating!
onyx hinge
@thorny jay I'm not aware of how to do that "automatically".
thorny jay
some adabot? or just making a PR that only add an empty file?
I wonder how much money does running all the GitHub CI cost?
lapis hemlock
I am only lurking today.
thorny jay
lapis hemlock
@idle owl Thanks for keeping the docs in sync with my status!
idle owl
@lapis hemlock Actually @onyx hinge beat me to it. We started typing at the same time.
lapis hemlock
Well, then @onyx hinge has earned some brownie points.
lapis hemlock
Yeah, I picked up the phrase while in the US.
stuck elbow
we do, but they are small mythical humanoids
onyx hinge
It's OK, mohnkuchen is pretty good
stuck elbow
lapis hemlock
"performing various chores" liking keeping the docs up to date...
stuck elbow
we don't usually eat them
lapis hemlock
The docs, do you mean? No, we don't.
stuck elbow
historically some doctors did get eaten, especially the ship docs
thorny jay
I remember those message about SPI on Discord... Maybe that was I2C !!!
modern wing
Wait, we're not supposed to eat the google docs?
ruh-roh
errant grail
Some of my projects rely on the SPI high baud rate setting for TFT displays. Setting it low by default for SAMD21 would be okay, though.
ivory yew
omg I missed the meeting
onyx hinge
@ivory yew want to do a special in the weeds status update or hug report?
we're still talking
ivory yew
Nah that's okay. I'm on the phone with the Department of Revenue right now π
lone axle
I am in favor of Circuit Python Day meeting
ivory yew
But hug report to all of the CircuitPython folks - digging through the CircuitPython source code has helped a lot with getting my own firmware up and running.
tidal kiln
yep. i like the idea.
idle owl
Nah
lapis hemlock
Why aren't the guides under version control?
lone axle
I have some code that I've used to search the Learn Guide repo a few times now that is posted in some of the recent Issues on that repo. Maybe there is some opportunity to integrate it with CI to see a list of potential projects that will be affected.
Would only help find instances of code that is in the Repo though, but not smaller snippets that are directly in the Learn Guides.
slender iron
@lapis hemlock https://github.com/adafruit/Adafruit_Learning_System_Guides/
the code is
lapis hemlock
Right, but it would make sense to expose the guides themselves to PRs, wouldn't it?
lapis hemlock
A couple of months ago, I updated the code for one of the guides, but then the text dropped out of sync, so I had to bug Jeff for a fix.
lapis hemlock
I think it would be worthwhile.
lapis hemlock
The thing is, and I agree with this, you are an advocate of integrating text (docs) with the source code. Exposing the code without exposing the guide is not really useful.
raven canopy
requirements.txt & setup.py, yes. its just versioning based with Dependabot.
slender iron
@lapis hemlock I'm not the one that needs to be convinced
lapis hemlock
OK, so I am barking up the wrong tree...
slender iron
ya, limor needs to be convinced
onyx hinge
@lapis hemlock Take this with a grain of salt, but I think one thing is that a lot of people who are writing content into the learn guides would state very strongly that they need the graphical editing tools vs "just writing markdown". I think the technical minded CircuitPython people are a minority of those writing guides, or at any rate perceived to be.
lone axle
I think that is true as well. If it could moved to be backed by Git or some other version control, I think it would definitely need some sort of very easy to use UI (ideally the exact same that exists today probably) it must be much easier than editing md code and then using git to push / merge / PR it.
lapis hemlock
@onyx hinge I think, such an approach would definitely lower my contribution threshold. I am very reluctant to even try to add content, if I have to trouble someone with that in the first place.
Markdown is OK.
tulip sleet
Rather than saying "markdown" and "GitHub", the goal is really to be able to suggest changes to Learn Guides, make it easy to approve and/or edit these changes, and give feedback to the proposers. Learn Guides do have a versioning scheme built in to the CMS, it's just not GitHub. It even shows diffs.
lapis hemlock
I don't care how it is done.
github, markdown etc. is OK with me, "ask Jeff to do something for me" is not.
tulip sleet
right, it just needs to be discussed in terms of functionality rather than implementaiton
lapis hemlock
Yes, but the snag is that if you accept markdown/github, then nothing has to be implemented.
lone axle
But then it's not easy enough to use for many of primarily intended guide authors.
lapis hemlock
But if you want to contribute to the code, you face the same difficulty, don't you?
I mean, this is really not a hurdle, or at least, not one that cannot be overcome.
lone axle
It can be overcome. But I think it's easy for folks like us who are knee deep in code and development tools all the time to underestimate how challenging and daunting it can be to someone with basically no traditional coding experience. Many of the guides are created by authors in that situation, and they still are able to create great and useful content even without having to learn those skills.
lapis hemlock
What I am saying is that there is already discrepancy in the reasoning: you can PR the code in the guide (you have already overcome the difficulties), but you cannot suggest changes to the text that would reflect the changes in your code.
I don't get it.
lone axle
There is a feedback link that can be used to suggest changes in guides.
lapis hemlock
That's true, but I cannot actually implement those changes. It is not efficient.
lone axle
(not saying that git or version control wouldn't be nice. I'd definitely like it as well)
lapis hemlock
@slender iron I am not trying. I have already accepted your answer.
OK, fair enough.
lapis hemlock
Oh, that's a very good point.
Yeah, but some of the requirements are not even clear.
kk.
idle owl
@onyx hinge I've replied to all of your PRs.
onyx hinge
@idle owl you are a treasure
thorny jay
Thank you all.
modern wing
Thanks!
lapis hemlock
See you all next timeπ
onyx hinge
See you @lapis hemlock
solar whale
π
onyx hinge
@idle owl OK I'll go ahead and try to finish the process of getting those PRs to "green" by fixing the other pylint errors. I'm not as confident as you that I touched most/essentially all libs with this set of PRs, there were under 100 and aren't we over 200 libs?
idle owl
@onyx hinge We are. I think in this case, the best option is to wait until they fail. It's not a great experience for a newbie, but we can handle the fixes ourselves on any newbie PRs that come in and explain the situation.
slender iron
<@&356864093652516868> any of you not have an esp32s2 board yet?
turbid radish
I don't yet