the teensy4.x are cheaper but don't break out SWD
#circuitpython-dev
gilded cradle
Correct @raven canopy
raven canopy
๐ just wanted to make sure you knew. its one of those sneaky details. ๐
gilded cradle
I generated a json file locally that I'm just using for my test code
onyx hinge
@raven canopy it's nice to see you dipping your toes back in!
orchid basinBOT
raven canopy
hai hai. its nice to be dipping again. ๐
ionic elk
@tulip sleet
lone sandal
orchid basinBOT
orchid basinBOT
mental nexus
Looks like the imx EVK 1010 evaluation board doesnโt have many pins broken out (for example to develop ParallelBus). And actually ParallelBus may not get much use on the early boards for the RT1011. Iโll keep my eyes open to see how the imx development proceeds and see if thereโs a topic that makes sense to get involved with.
Looks like I spoke too soon, I looked one more time and found an app note on an 8bit parallel demo on the EVK1010 board (6800 bus, not 8080 like on our typical LCDs): https://www.nxp.com/docs/en/application-note/AN12813.pdf using the FlexIO. So it looks like thereโs room to explore on the board you mentioned.
Just for completeness (and for future me when I forget where I found this), hereโs an 8080 parallel LCD demo (ParallelBus) on an older NXP chip and demo board. Seems like the โFlexIOโ configuration is similar to that on the RT1011: https://www.nxp.com/docs/en/application-note/AN5313.pdf
onyx hinge
@mental nexus the I/O layout of the 1010-evk is a bit weird and limited. Be especially careful if you're using the schematic because some things that look like they are connected are NOT because a 0 ohm resistor is marked "DNP" or "do not populate"! On the doubled Arduino-style headers there are some connections on the inner half but far from all connections are used.
mental nexus
Thanks for the warning. In looking at the docs, many of the pins arenโt broken out to headers. For 80 pin package I guess they use them for other purposes on the board. Will stew on this a couple more days and decide whether to order one of these.
jaunty juniper
say, I was looking at the _data/files.json file on the circuitpython-org repository and noticed it's technically not correct json (there's a , at the end of some lists), but I don't know where the code generating it is to maybe issue a fix ?
onyx hinge
@jaunty juniper I think they are usually generated by code but were hand-modified the last time around due to a bug
tulip sleet
yes, that might be it, I just grep -v'd some stuff out of the json
idle owl
@slender iron Ping me when you're around to chat about the pre-commit page. No rush.
onyx hinge
@jaunty juniper if you want to take this on ^^ it would be helpful. If the issue is too terse I am happy to expand on it
tulip sleet
looks like my hand edit did add trailing commas:
onyx hinge
I updated issue #4470 with more information so someone else may have a chance of understanding what is needed ๐
(it says "issue opened" above, but it was transferred from circutpython-org where I'd mistakenly opened it initially)
jaunty juniper
ah I see
I was investigating ways for circup to get info on the latest beta without doing parsing on its own
tulip sleet
#4470 needs to be fixed before the next release of any kind
mental nexus
I feel like I'm being dense, but I can't find an example in shared-bindings that takes a string as an input parameter. I'm looking to add a filepath input string to a function, and can't seem to find how to track down an existing example.
onyx hinge
@mental nexus a lot of things take opened files rather than filename strings ..
mental nexus
Yep, that's exactly the issue! : https://github.com/adafruit/circuitpython/issues/4197
onyx hinge
@mental nexus try ipaddress_ipv4address_make_new in shared-bindings/ipaddress/IPv4Address.c
mental nexus
Cool, will look there.
onyx hinge
It's parsed as a generic object { MP_QSTR_address, MP_ARG_OBJ | MP_ARG_REQUIRED }, then checked for being a string by shouting at it MP_OBJ_IS_STR(address) and then the bytes and (byte-?)length of the string can be extracted by additional shouting GET_STR_DATA_LEN(address, str_data, str_len);
mental nexus
Exactly what I was looking for. Thanks jepler!
tulip sleet
@slender iron (when you're up) and @onyx hinge too if you're interested. I would like to have a little release planning meeting about what should go into 6.2.0 and when we should branch it. @ionic elk has STM sleep code that will be ready to go soon, but it has some internal refactoring and renaming and will require revision of the nRF sleep PR. The nRF sleep PR is blocked by a USB issue I need to look at. We can also talk about whether the new flash stuff and any I2C fixes should go into RP2040 before 6.2.0.
ionic elk
@tulip sleet I'm around all day
idle owl
Does this really benefit from the debounce delay? Or could it be simplified to not include it. I haven't tested it without yet. Simply wanted opinions. https://github.com/adafruit/Adafruit_Learning_System_Guides/blob/master/CircuitPython_Essentials/CircuitPython_Digital_In_Out.py
Seems like spamming the result is irrelevant here, because the LED is on or off.
tulip sleet
no, because who cares whether the light goes on and off due to bounces. It may not even be visible
tulip sleet
it might be interesting to come up with an example where debouncing is crucial: a simple one would be counting button presses
like printing out a character on each press
idle owl
There's quite a few already. I usually do simple state machines to handle it though, so you can track press and release without a separate module.
Not sure we've done anything with the debounce being the focus though. Maybe Dave Astels did at some point since he wrote the debounce lib?
No idea.
onyx hinge
@tulip sleet happy to chat or get on a video call about it but I don't feel like I have a high level overview of the state of things.
orchid basinBOT
cobalt grail
Hej weblate GANG - Got my matching tattoo today! @fossil gorge @still zephyr 
jaunty juniper
@tulip sleet trying to run build_board_info.py locally I run into Expecting value: line 71 column 5 (char 979) because it loads the current files.json from circuitpython-org, should I PR a fix for the file ?
tulip sleet
certainly if it's easy clean up the current file ; thank you
onyx hinge
just to share some frustration .. I was trying to check out how I2C works on the 1010-EVK with some 6dof sensor I wasn't familiar with. didn't work. Started checking the schematic and .. after much hair-pulling .. noticed that THE WHOLE SENSOR is a "DNP" part. Oh, that's why it didn't show up on bus scan...
fossil gorge
What's "DNP"? I assume not "Do Not Pull-hair"?
tulip sleet
do not place
analog bridge
do not populate
tulip sleet
see last paragraph here: https://en.wikipedia.org/wiki/Bill_of_materials#Configurable_BOM
slender iron
@tulip sleet I'm around if you want to talk 6.2
tulip sleet
now is good
onyx hinge
I'm around but it'd be slightly inconvenient to hop in a voice channel right now
gloomy shuttle
I am playing around with the new Touch Deck guide but I am trying to use a FeatherS2. Unfortunately it doesn't use the same pins as the tft_featherwing expects. Now the problem is that tft_featherwing_24 and tft_featherwing_35 don't expose the ts_cs and sd_cs pins. I was able to get it to work, but should these be exposed so users can use other feather boards with those libraries easily?
mental nexus
just to share some frustration .. I was trying to check out how I2C works on the...
This sounds awful familiar...rewinds Discord a mere three hours.....
onyx hinge
what, when I told you to watch out for DNPs? yeah. ๐
lone axle
I am playing around with the new Touch Deck guide but I am trying to use a Feath...
You should be able to pass some keyword arguments to set them: https://github.com/adafruit/Adafruit_CircuitPython_FeatherWing/blob/c9e46e5680237d91cb66b70b63a1399d57e5cf85/adafruit_featherwing/tft_featherwing.py#L34
gloomy shuttle
You should be able to pass some keyword arguments to set them: https://githu...
Right, except that is only in the parent class and not the subclasses.
lone axle
The sub classes need to accept and then pass through kwargs here: https://github.com/adafruit/Adafruit_CircuitPython_FeatherWing/blob/c9e46e5680237d91cb66b70b63a1399d57e5cf85/adafruit_featherwing/tft_featherwing_35.py#L36
gloomy shuttle
I created a pull request if the keyword arguments should be exposed in the subclass.
lone axle
To answer your original question though, yep it definitely should allow for that. Excellent thanks for making that PR. I will review that tonight or tomorrow.
Using the actual argument names like you did instead of kwargs like I said is actually better I think. Good call on that. It will make the documentation more clear.
thorny jay
I am playing around with the new Touch Deck guide but I am trying to use a Feath...
Don't tell me about FeatherS2 pinout... I gave up on that fight, but @slender iron did continue trying to convince for a change.
gloomy shuttle
Don't tell me about FeatherS2 pinout... I gave up on that fight, but <@!25271719...
I am not trying to carry the torch on that fight. I did chuckle a little when I ran into this issue.
stuck elbow
How can I test the 8-bit parallel display with a Feather RP2040? I don't see a set of 8 consecutive gpios on it, is there some trick?
manic glacierBOT
The SPI clock itself will go fast, here's 30MHz:
but the gap between bytes gets relatively longer and longer at higher baud rates, limiting the effective throughput:
ionic elk
@tulip sleet quick question for you about power, got a sec?
onyx hinge
@stuck elbow I think they have to be consecutive but need not be aligned, so try D4 as the base
tulip sleet
<@!329766224093249548> quick question for you about power, got a sec?
I was talking to Scott about releases; free now
slender iron
I am not trying to carry the torch on that fight. I did chuckle a little when I ...
I'll be giving a short talk at the open hardware summit where I'll definitely touch on this
ionic elk
I think I basically have it figured out but just wanted to check - when the USB delay is called, we want to delay the interval for a deep sleep alarm until the deep sleep actually starts, right? In ESP32-S2 I think it does this automatically because esp_sleep_enable_timer_wakeup doesn't actually set an interrupt itself, the actual deep sleep command sets that up. But it's relevant for STM32 because the alarms that alarm_time_timealarm_set_alarms creates can go off in the middle of the USB delay.
tulip sleet
when we start the USB delay we have no idea whether we should be real or fake deep sleeping
ionic elk
Right now I've made it so that when common_hal_alarm_enter_deep_sleep is called, it calls alarm_time_timealarm_prepare_for_deep_sleep which sets the alarm again, which will either replace it if it was too short (went off during USB delay) or too long (in which case it overwrites it)
tulip sleet
are you saying like if we want to deep sleep for 10 seconds but have to wait for 5 seconds for USB that we actually sleep for only 5 seconds after the first 5 seconds (i.e., make it more accurate?)
ionic elk
I'm saying that if I sent a command to sleep for 2 seconds on STM32, unless I have a system to catch that, the alarm will go off during the USB delay and it will never be able to wake up
tulip sleet
yes, got it, so on the ESP32-S2 this was not a problem because the alarm time was only considered when we actually started sleeping. Maybe the name of the function should be more accurate
stuck elbow
@onyx hinge oh, I missed gpio6 because it wa son the other side, thanks
tulip sleet
I remember noticing that
ionic elk
There are two ways to solve that - either we catch the problem before deep sleeping and just restart, or I can make it actually only start the timer once it's decided to fake/real deep sleep. I picked the second because it's closer to the ESP32
tulip sleet
I agree with the latter
ionic elk
In the majority of cases, alarm_time_timealarm_set_alarms does actually start the alarm. We just need an extra check to make sure it doesn't expire before deep sleep, so the code can always wake up.
Now that I think of it, this could actually happen with user code... if someone (foolishly) puts another form of delay or long calculation between alarm_time_timealarm_set_alarms and common_hal_alarm_enter_deep_sleep, this could still happen
It might actually be a clearer API to always start all time alarms only when actually sleeping? Or does that interfere with the ambitions to have alarms function as a proxy ISR?
tulip sleet
@ionic elk The alarm times are set as a time in the future, as an absolute time.monotonic(). That is, I want to wake up at time x, not x seconds from now. So whatever makes sense to preserve those semantics is best. If the alarm time has already expired, then we wouldn't sleep at all (or deep sleep would wake up immediately after exiting the VM).
ionic elk
@tulip sleet the current system is not sufficient, then. If that is the intent, we need to change both the ESP32-S2 and upcoming ports to double-check before deep sleeping that their wakeup time interval has not already elapsed, and restart (possibly with a warning saying they were late)
Ah I seem I'm kind of repeating you there
tulip sleet
we should probably do that; it was not a big deal because we were envisioning sleep times of minutes or hours, not a few seconds
this was all directed toward the magtag, waking up very occasionally to update things and then go back to sleep
ionic elk
Yeah I mean the super trivial way would be to just throw a value error if you try to sleep under 5s
deep sleeping for 2s in any case is kind of silly (even though it will work just fine on every sleep restart after the first plug-in)
tulip sleet
that might make sense. We were just being sloppy/urgent.
ionic elk
Yeah it's no biggie, just digging these things up as I go ๐
tulip sleet
it's worth the clean-up thinking, thanks.
if you don't want to bother with it now, it can be a long-term issue
ionic elk
I may go with my current system as of right now just because it's working and I want to get my stuff in
(ie both the ESP32-S2 and STM32 will push out the wakeup date by 5s after first plug-in)
ionic elk
other trivial question, do we want to have a shortcut for setting the timealarms to relative time?
Seems like relative time is used a lot in the examples anyway and makes sense for many applications
tulip sleet
we decided not, to make the semantics clear
since it was "relative to what start time"?, and that wasn't obvious
but you could propose it
again, the magtag examples were more like: wake up at 9am and get the weather, as opposed to wake up 24 hours from now
so for longer periods it was an "alarm", not a "timer" (comparing your phone functionality, for instance)
ionic elk
Probably would want to include it in the call itself, maybe, alarm_time_timealarm_set_delay_after_sleep
specifying that it's after sleep actually starts
or something
tulip sleet
or it could be a different keyword arg, I think that's the idea of the monotonic_time keyword arg name. There may be some discussion of this in the original issue or PR.
ionic elk
Seems like something to just propose in an issue. I'll do some research to see if there's existing basis.
tulip sleet
btw, we decided to defer these new sleep impls to past 6.2.0 to get the current beta to stable quickly
onyx hinge
@ionic elk how much imx-brain do you have active at the moment?
ionic elk
@tulip sleet one more thing for you. Do we care about second-level resolution re: alarm times? For short alarms they're often off by 1s, but as per the above I don't know that we care
@onyx hinge I can re-brain some i.MX, just finished an STM32 task I'm out of the weeds mentally
what's up
onyx hinge
I am trying to figure out how to phrase the question ๐
onyx hinge
the context is https://github.com/adafruit/circuitpython/issues/3062
ionic elk
more like i.MESS amirite
onyx hinge
SPI transfers have absurdly long pauses between bytes
but all I do is, call their SDK function
and I don't know a better question to ask than, is that just how it is?
maybe, is there some other SDK function I overlooked for doing this with DMA, which is what someone in the issue mentioned? The lpspi API just has blocking and non-blocking versions, but neither uses DMA to the extent I can follow what they do
ionic elk
Have you poked around inside the SDK function to see if it's doing weird stuff? Maybe it wants you to do some kinda DMA thing?
onyx hinge
ionic elk
ah ninjad
onyx hinge
drivers/fsl_lpspi_edma.h oh maybe the dma related stuff is in a different file entirely
ionic elk
my question might be are the blocking calls slotting something weird in there, like a flash access or something that's taking lots of time
onyx hinge
shouldn't be?
ionic elk
SDK manual shows both Peripheral and eDMA sections
so they're considered separately there too
onyx hinge
OK, well, now I know why I couldn't find DMA
onyx hinge
I don't know if I found that one yet
ionic elk
I don't think NXP docs are very good but still probably nice as a reference
onyx hinge
.. is there no html version ? ๐ข
ionic elk
idle owl
@tulip sleet Mu works with Pico now? Or will..? Guide feedback mentioning the 2 COM ports.
onyx hinge
@idle owl 2 COM ports has been disabled for now
ionic elk
@onyx hinge ๐คทโโ๏ธ
onyx hinge
largely because of mu
idle owl
Oh I thought that was still a Pico issue. Fair enough.
Thanks.
I knew it had been disabled. Was conflating things.
onyx hinge
@ionic elk so far the main thing I'd tried was identifying the routines used during SPI'ing (in the sdk) and put them in itcm but that didn't make any difference
ionic elk
@onyx hinge apparently there are also examples at <SDK_ROOT>/boards/<BOARD>/driver_examples/lpspi you could try those and see how the speed comapres
sorry am I flaking on what itcm is
oh right address ram
onyx hinge
I don't exactly know, but my understanding is it's a region of memory that is faster for fetching instructions from -- instruction tightly-coupled memory
ionic elk
yep just forgot momentarily, google was giving me silly company names
Well, sorry I'm not more helpful here, I didn't write i.MX busio so I'm not super strong on it.
But I think DMA is probably your best shot, especially if you can find SDK examples for it
onyx hinge
thanks for talking through it, it is nice in itself
manic glacierBOT
jaunty juniper
@onyx hinge I'm looking at https://github.com/adafruit/circuitpython/issues/4470 how would one test it locally though ? (outside of a bunch of prints I guess)
onyx hinge
@jaunty juniper good question!
Comparing to existing files is tough, especially due to the unsorted nature of the files. There are json diff programs but I've found that the files of that size tend to give them trouble, particularly the browser-based ones
I'd just make a plausible looking change and manually check the language list.
orchid basinBOT
onyx hinge
@jaunty juniper honestly i'm not sure if/how to run build_board_info locally... hm
jaunty juniper
I can test it by adding a print at the end
and then there's moving language_allow_list and the related daunting prospect of testing build_release_files.py, I guess CI will have to do that
onyx hinge
CI does that everytime so I wouldn't worry there
idle owl
Is the process to add a new board to Mu documented anywhere? I received guide feedback asking for it to be added to the "adding a board to CircuitPython guide" and I don't know the process well enough to add it.
onyx hinge
@idle owl @tulip sleet is probably the expert about that at the moment. As I understand it, the process will be greatly simplified (or totally eliminated) once they take some enhancements he's been working on, but the schedule for that on mu's end was unknown last I heard.
idle owl
Thanks. I'll talk to Dan about it when he's around next.
onyx hinge
the status quo is that somewhere inside mu is a list of USB IDs and if your product is not an Adafruit product then you need to add it there. Soon, though, it'll just look for some other things in the USB information that say CircuitPython, is approximately the thing
idle owl
The guide feedback linked the file that needs to be changed, and what needed to be added. But I wasn't going to simply take that as valid.
tulip sleet
@idle owl the Mu file containing the VID/PIDs for boards now wildcards all Adafruit boards. I have a PR pending to Mu that eliminates the need for this entirely: it looks at the interface description string for the REPL connection and decides it's a CircuitPython board that way
is the board an Adafruit or non-Adafruit board?
idle owl
This would be a non-Adafruit board. The guide is how to add your board to CircuitPython.
So ostensibly it's other folx adding boards.
tulip sleet
They haven't yet accepted my PR, but they haven't even started reviewing it. They were actually stuck on some CI problems.
idle owl
@tulip sleet So for non-Adafruit boards, this still stands? they can submit a PR to Mu Editor, editing this file: https://github.com/mu-editor/mu/blob/master/mu/modes/circuitpython.py And adding their board to valid_boards. The format is like this: (VID , PID, "Manufacturer Name", "Device name")
tulip sleet
yes, though my PR will supersede any new PR for a single new board
orchid basinBOT
tulip sleet
<@!329766224093249548> one more thing for you. Do we care about second-level res...
Are the only-1s-accuracy alarm times just due to overhead or something like that? As long as they aren't being rounded to truncated to the nearest second, we could potentially fix that later.
idle owl
@tidal kiln You around?
idle owl
Do you actually have one of these nunchuck adapters? https://www.adafruit.com/product/4836
tidal kiln
i do
idle owl
no. i'm a square!
Aw come on.
idle owl
i do
Ok I need a favor. Can you check the I2C address?
Because the text on the back says 0x52, and the product copy and guide say 0x53
So something isn't right there.
tidal kiln
interesting. yah. let me check. the adapter itself really doesn't have an address. it's just an adapter. but let me see what a nunchuk reports as...
gloomy shuttle
I am 99% confident they are 0x52.
tidal kiln
>>> import board
>>> i2c = board.I2C()
>>> i2c.scan()
[]
>>> i2c.scan()
[82]
>>> hex(82)
'0x52'
>>>
first scan with just adapter
second with nunchuk plugged in
nunchuk = original nintendo model
not the one in shop, if that matters?
tidal kiln
np. mario out!
fossil gorge
second with nunchuk plugged in
I'm going to have to make note of that. I remember not seeing it when I did a scan, but working when I tried the code. I bet I didn't have a nunchuk plugged in when I scanned...
tidal kiln
worth checking, but still not 100%. there can be some odd things that don't scan well, but work fine otherwise.
manic glacierBOT
I graphed the data
Would be good is someone else could check it in case my SQ25 is misbehaving but that seems unlikely.
jaunty juniper
is this the correct list of excluded languages ? {'el', 'cs', 'ko', 'en_GB', 'hi'}
wait I could just look in the locale directory
(I looked, and it is)
lime trellis
Hi all! Linked below is a WebSerial demo with some tweaks for circuitpython.
This <200 line html file lets you access the serial terminal of your circuitpython boards using your browser!! Confirmed working in Chrome and MS Edge (v89+). Firefox and Safari will likely not work.
copy the text into a .html file and give it a go ๐
https://github.com/maholli/serialTerminal.com/blob/main/index.html
I'm always looking for easier ways to have my students connect to their circuitpython boards and this will be a game changer! File is small enough that I'll likely distribute boards to students with a similar html file onboard.
tulip sleet
@lime trellis this is fantastic! I had tried the Google demo and some others, and they didn't quite work.
lime trellis
<@492049388152029185> this is fantastic! I had tried the Google demo and some ot...
ditto! all credit goes to mmiscool. I just forked and added buttons so we can easily send some control characters to our boards
tulip sleet
can the index.html be served or must it be local
lime trellis
can the index.html be served or must it be local
it can be served
signal pasture
doesn't google chrome force pages to be accessed over https to use the serial port?
lime trellis
doesn't google chrome force pages to be accessed over https to use the serial po...
yes, but MS edge doesn't. Adafruit should host some version of the page so it's readily available
signal pasture
yeah, that would be really cool
lime trellis
update: works great hosted via github: https://maholli.github.io/serialTerminal.com/
tulip sleet
On Ubuntu 20.04 with Chrome 89.0.4389.90, I am getting "Serial Connection Failed"
after choosing and hitting Connect
lime trellis
On Ubuntu 20.04 with Chrome 89.0.4389.90, I am getting "Serial Connection Failed...
yeah I'm seeing that too from another device. Try connecting from https://www.serialterminal.com/. There appears to be something goofy as a result of github hosting.
tulip sleet
The Google demo has the typing in-line, instead of a separate line. That can't be too hard. Let me find it...
THis does a weird disconnect when I try to type, though, that was the problem
lime trellis
The Google demo has the typing in-line, instead of a separate line. That can't b...
agreed it should be tweaked to feel more like a terminal. props to the creator for making a simple template for us
lime trellis
THis does a weird disconnect when I try to type, though, that was the problem
https://googlechromelabs.github.io/serial-terminal/ works great for me in chrome and edge. able to halt, use REPL, etc...
tulip sleet
are you on windows or linux
lime trellis
windows
tulip sleet
that may be the difference. Try resetting the board and reconnecting with the Connect button and see if you get spurious DISCONNECTED
If I was previously connected, it works for me, but it does not work when started from scratch with a board that wasn't connected to web serial previously
lime trellis
If I was previously connected, it works for me, but it does not work when starte...
both scenarios seem to work on windows. I'll try my linux machine tonight
tulip sleet
would be interested to hear, thanks
gloomy shuttle
Hi all! Linked below is a WebSerial demo with some tweaks for circuitpython.
Th...
This is awesome! Does Chrome have a way to edit files on the CIRCUITPY drive? It would be awesome for Chromebook users to have a web editor like MU to do all their work in.
thorny jay
https://photos.app.goo.gl/RrA6ntgNwnJnfuS16 My feather RP2040 driving the 128x32 oled has a glitch. Same code works with all of my other MCU. I believe I had the same issue with a Pico. That is Liz code without modification (AFAIR).
thorny jay
https://photos.app.goo.gl/RrA6ntgNwnJnfuS16 My feather RP2040 driving the 128x32...
Found device at D:, running CircuitPython 6.2.0-beta.4-84-gc81007afb.
adafruit_debouncer==1.3.10
adafruit_displayio_ssd1306==1.3.1
adafruit_bitmap_font==1.3.7
adafruit_display_text==2.17.0
All is up to date.
thorny jay
I don't know why I did believe it was from Liz. Maybe because of the "voting intro" in her video???
bitter bronze
Using CircuitPython extension in VSCode. View>Command Palette>CircuitPython:.. fail with an error message that the command resulted in an error. That is true for all CircuitPython commands like Show Available Libraries...
cedar moth
Using CircuitPython extension in VSCode. View>Command Palette>CircuitPython:.. ...
Have you opened the D:\ folder in VSCode? The CP extension uses the boot_out.txt file found in the root of the D: drive and without it I have found the extension gives errors like what you describe.
manic glacierBOT
manic glacierBOT
solar whale
Just a "heads up" to others who have not been keeping up with the CP development tools... I updated to the tip of main today as I do most days and tried a build the adafruit_feather_rp2040. It failed because it now apparently needs cascadetoml to have been installed. No big deal, but a surprise. I went back to the building CP guide and noted what I think is a new step pip3 install -r requirements-dev.txt https://learn.adafruit.com/building-circuitpython/build-circuitpython I ran this and found I was missing several packages ... Everything is working now but it was a confusing start to the day.
mental nexus
I've got a pre-commit question. I've staged only a few files for a commit. When I run pre-commit I get this:
margaret@iMac circuitpython % pre-commit
[WARNING] Unstaged files detected.
[INFO] Stashing unstaged files to /Users/margaret/.cache/pre-commit/patch1616685107.
Check Yaml...........................................(no files to check)Skipped
Fix End of Files.........................................................Passed
Trim Trailing Whitespace.................................................Passed
Translations.............................................................Passed
Formatting...............................................................Failed
- hook id: formatting
- files were modified by this hook
[WARNING] Stashed changes conflicted with hook auto-fixes... Rolling back fixes...
It's not saving the formatting fixes on my files, maybe because I have un-staged files? How can I get the formatting to update the files that I have staged and make a commit?
manic glacierBOT
analog bridge
@mental nexus you can run formatting on a specific set of files by pre-commit run --files file1 file2...
mental nexus
Ok, I'll try that. In the meantime I messed up my branches and misplaced my files. Bleh. Will get help in the git channel.
slender iron
Just a "heads up" to others who have not been keeping up with the CP development...
Thanks Jerry! I should have let folks know in the meeting
tulip sleet
I did revise the Building guide very recently after someone reported the issue. I will try to make an announcement in-channel if I do revise the instructions.
fossil gorge
I've got a pre-commit question. I've staged only a few files for a commit. Whe...
What that was saying was that some changes you had weren't staged, so they got stashed. However, some of the files that were modified by the pre-commit hooks changed files that were stashed when pre-commit was started, and git wasn't able to reconcile the difference between the two, so it couldn't "undo" the stashing.
slender iron
@onyx hinge we really shouldn't need dma for spi
the peripheral has a 16 word transmit FIFO
(though dma insulates the transfer from interrupts)
onyx hinge
@slender iron my gut agrees with you, yet using the FIFO-based API in the sdk gives the results seen on the PR: big gaps between bytes
If you have something I could look at first before doing DMA I'm open to it
I considered seeing whether by bypassing the SDK and just touching the registers I could do better
onyx hinge
welll looks like my fix for flash files wasn't right & complete. I hope it's still a step in the right direction @slender iron
slender iron
No worries @onyx hinge. Iโll look at the flash stuff after breakfast. Is the sdk waiting on a fifo empty flag instead of fifo full?
slender iron
@onyx hinge got a link to the spi code in the nxp sdk?
onyx hinge
@slender iron https://github.com/adafruit/MIMXRT10xx_SDK/blob/master/devices/MIMXRT1011/drivers/fsl_lpspi.c#L778
while (LPSPI_GetTxFifoCount(base) == fifoSize)``` the most basic thing you suggested to check seems to be as expectedya, that looks correct
onyx hinge
yet from the traces it never successfully gets a 2nd byte into the FIFO before the first byte is shifted out.
slender iron
there is a retry loop in common-hal
onyx hinge
if (((NULL == rxData) || (rxRemainingByteCount - txRemainingByteCount) < rxFifoMaxBytes))```This was a bit confusing to wrap my head around, but I think it's right
the retry loop in common-hal is only for the "is busy" case, if it's not busy the single call does the whole transfer. /* Check that LPSPI is not busy.*/ if (LPSPI_GetStatusFlags(base) & kLPSPI_ModuleBusyFlag) { return kStatus_LPSPI_Busy;
slender iron
didn't nxp post the code on github now?
onyx hinge
I dunno, the one we use was just imported by arturo into a repo with empty history
slender iron
right, I feel like greg pointed me to it on github
onyx hinge
oh that would be useful. It didn't seem to be any of them here https://github.com/NXP
slender iron
yup, that's what greg pointed to
onyx hinge
It's not obviously, hugely different than what we have
onyx hinge
https://www.nxp.com/docs/en/errata/IMXRT1010CE.pdf nothing obvious
onyx hinge
yes
slender iron
ยฏ_(ใ)_/ยฏ
onyx hinge
switching to that is basically what I was intending to do
slender iron
it all looks right to me
onyx hinge
but .. how is it spending 8000 CPU cycles between putting a byte in the FIFO !?
slender iron
well its likely on a slower memory bus
so checking all the registers could be slow
have you logic analyzered the spi flash?
I'm trying to remember where I saw the memory structure map
onyx hinge
a/devices/MIMXRT1011/drivers/fsl_lpspi.c
+++ b/devices/MIMXRT1011/drivers/fsl_lpspi.c
@@ -777,2 +777,3 @@ bool LPSPI_CheckTransferArgument(lpspi_transfer_t *transfer, uint32_t bitsPerFra
*/
+__attribute__((section(".itcm.LPSPI_MasterTransferBlocking")))
status_t LPSPI_MasterTransferBlocking(LPSPI_Type *base, lpspi_transfer_t *transfer)
@@ -1813,2 +1814,3 @@ void LPSPI_SlaveTransferHandleIRQ(LPSPI_Type *base, lpspi_slave_handle_t *handle
+__attribute__((section(".itcm.LPSPI_CombineWriteData")))
static uint32_t LPSPI_CombineWriteData(uint8_t *txData, uint32_t bytesEachWrite, bool isByteSwap)
@@ -1897,2 +1899,3 @@ static uint32_t LPSPI_CombineWriteData(uint8_t *txData, uint32_t bytesEachWrite,
+__attribute__((section(".itcm.LPSPI_SeparateReadData")))
static void LPSPI_SeparateReadData(uint8_t *rxData, uint32_t readData, uint32_t bytesEachRead, bool isByteSwap)
``` I tried putting MasterTransferBlocking and everything it calls that is not inlined into itcm and it did not helphrm
onyx hinge
it's possible I got it wrong since I had to write it out rather than use our macros
0x00000000000019e8 0x4c build-imxrt1010_evk/supervisor/flexspi_nor_flash_ops.o
0x00000000000019e8 flexspi_nor_flash_page_program
.itcm.LPSPI_CombineWriteData
0x0000000000001a34 0x50 build-imxrt1010_evk/sdk/devices/MIMXRT1011/drivers/fsl_lpspi.o
.itcm.LPSPI_SeparateReadData
0x0000000000001a84 0x68 build-imxrt1010_evk/sdk/devices/MIMXRT1011/drivers/fsl_lpspi.o
.itcm.LPSPI_MasterTransferBlocking
0x0000000000001aec 0x22c build-imxrt1010_evk/sdk/devices/MIMXRT1011/drivers/fsl_lpspi.o
0x0000000000001aec LPSPI_MasterTransferBlocking
.itcm.common_hal_neopixel_write
0x0000000000001d18 0xd8 build-imxrt1010_evk/common-hal/neopixel_write/__init__.o
0x0000000000001d18 common_hal_neopixel_write
```but the map file looks rightya, seems right
what is you change the wait to read the current amount once and set the difference?
instead of checking to see if it's full each write
I'm sure the low speed bus is still much faster than the spi transmit
onyx hinge
does the data need to be coming from the right RAM area?
slender iron
for dma yes
dma can't access TCM
afaik
an interrupt could be slowing down the cpu core
maybe usb
do you get the gaps when not connected to usb?
onyx hinge
@slender iron figured it out ๐
slender iron
nice!!! what was it?
onyx hinge
30MHz transfer yay
there are several delays which are configured by default ``` masterConfig->pcsToSckDelayInNanoSec = 1000000000 / masterConfig->baudRate * 2;
masterConfig->lastSckToPcsDelayInNanoSec = 1000000000 / masterConfig->baudRate * 2;
masterConfig->betweenTransferDelayInNanoSec = 1000000000 / masterConfig->baudRate * 2;
is it toggling CS every byte?
onyx hinge
If I set them all to 0, the gaps go away. I think it's the betweenTransferDelay
this flag also may have helped: xfer.configFlags = kLPSPI_MasterPcsContinuous
anyway, I'll tidy it up but this looks very good at the moment
onyx hinge
No, we never enable the "internal" CS output
we always use our software CS in CircuitPython
.. right?
onyx hinge
no, Continuous on its own tightened things up a bit but it was not the resolution.
slender iron
hrm, weird
onyx hinge
// SPI bus noise before they are put into SPI mode.
config.baudRate = 250000;
hmmm, these initial settings were also made with a baudRate of just 250kHz, so they ended up being very big numbers when you switched to 30MHz instead of 250kHz. the settings were relative to some default baudrate, which we then set to another rate
those default delays were hidden inside SDK, not in CP
slender iron
๐
orchid basinBOT
onyx hinge
last thing to check is whether the sdk code needs to be in itcm to work at 30MHz
nope, fine without it
slender iron
๐
manic glacierBOT
It's possible to configure very large gaps between transfers (bytes), and indeed this is what we ended up with by using LPSPI_MasterGetDefaultConfig.
Set all the delays to zero, enable a bit for "continuous transfers", and cap the speed at 30MHz per the datasheet's restrictions.
After this change, here is a view of a transfer at 30MHz:
I tested on the MIMXRT1010-EVK both wi...
ionic elk
@tulip sleet I have a tricky problem.
bitter bronze
Have you opened the D:\ folder in VSCode? The CP extension uses the boot_out.tx...
Thanks @cedar moth. In VSCode,I have opened the drive on the Pico (F: in my case ) and then opened the code.py file (which is a simple flash the LED program), and tried running it. If I try View>Command Palette>CircuitPython:... it still fails with the "resulted in an error message". Other commands also fail. I have similar failures with the Pico-go commands. I'll keep poking around. Any ideas about a good help sight for VSCode.
ionic elk
@tulip sleet On STM32, the WKUP pin function just fine when in true deep sleep, but it cannot actually wake from WFI, and it in fact disabled all other functions of that pin when not in sleep mode. So it's impossible to have a program that can handle both true and false deep sleep without making a judgement on whether the program is connected to usb when the alarm is created
In the ESP32 case, the program only makes that judgement when it actually starts deep sleeping, since a pin can function as both a false and true deep sleep wakeup.
Or maybe I should dynamically reconfigure the pin only when true sleep actually starts, based on whether it was requested earlier... that's probably the way to go, nevermind.
tulip sleet
yup, the last thing is what i was going to say; you just have to remember the pin and do the work when you know
ionic elk
@tulip sleet yeah STM32 only has one viable wakeup pin so it's actually pretty easy
stuck elbow
displayio.ParallelBus on the Feather RP2040 works!
slender iron
๐
stuck elbow
however, I expected it to be a bit faster, is there a frequency setting somewhere?
stuck elbow
I see there is, just undocumented
mental nexus
If you have any speed comparisons vs other boards, I'm interested. Here's my benchmark that I've used to compare a couple of boards: https://github.com/adafruit/circuitpython/pull/4047#issue-560212220
blissful pollen
I see there is, just undocumented
It should default frequency to the fastest I got it to work
but I thought it may be faster too but cannot see anything that is slowing it down compared to other ports. Let me know if you notice anything
stuck elbow
@blissful pollen it defaults to 30_000_000, I got it to work with 62_500_000
I'm really not familiar with pio yet
blissful pollen
which display are you using?
stuck elbow
Z320IT002 โ that is a 3.2" tft with ILI9341 COG
blissful pollen
Hmmm yeah I will have to play around again to see. I ran into issues going any faster on the display I have. But that's why frequency is there ๐
stuck elbow
I appreciate its presence
though I think it might go even faster, if the pio allowed
I mean the display could probably handle it
blissful pollen
It takes 2 PIO cycles to control so I know that maxes at 60Mhz (120 divided by 2 instructions). The frequency setting takes that into account... or should unless I made a mistake hmm
stuck elbow
well, I did a binary search looking when I get the fequency out of range error, and I found exactly 62_500_000
hehe, the datasheet says write cycle is 33ns, so 30Mhz sounds about right
stuck elbow
I was really expecting the parallel interface to be faster than SPI, but it isn't
blissful pollen
I was really expecting the parallel interface to be faster than SPI, but it isn'...
Maybe it hits a CP limit elsewhere? I'd have thought so too
stuck elbow
@blissful pollen possible, I will look closer with a logic analyzer over the weekend
still zephyr
mmmm maybe now we would need an equalizer widget ๐ค
manic glacierBOT
manic glacierBOT
Looking at the datasheet for the RT1062, it looks like it includes a touch screen controller for resistive touch screens:
serene briar
I think I found a mistake in the documentation for the stm32f405 feather. In https://learn.adafruit.com/adafruit-stm32f405-feather-express it says the following ```Only MicroPython has support for CAN bus at this time. Arduino has an open issue, no support. CircuitPython does not have CAN support.
simple pulsar
FYI, the docs for aesio functions don't mention padding: https://circuitpython.readthedocs.io/en/latest/shared-bindings/aesio/index.html. This might be a useful addition to reduce confusion either for those that have never used it or those that are used to the underlying nRF52 functionality. Came up in https://forums.adafruit.com/viewtopic.php?f=60&t=177264
manic glacierBOT
mental nexus
<@!228322543507865601> possible, I will look closer with a logic analyzer over t...
Thanks for your detailed display update on hackaday. Itโs good to see the details since it doesnโt always โjust workโ. Anyway, looking at the ILI9341 it looks like a minimum 20ns write time in parallel mode, so is 50Mhz the top limit?
stuck elbow
@mental nexus yeah, that would be the maximum guaranteed to work, but in practice you can usually go faster
@mental nexus also, I think the slowness is not due to the slow speed of the clock, but pauses between transmissions
I plan to work on this over the weekend a bit, and the first thing I want to do is to compare with samd51 to see if there is any noticeable difference in speed
mental nexus
Compared to the pyportal I have seen slightly faster pure redraw rates from the ESP32-S2 so Iโm interested in how the RP2040 compares. I just ordered a Teensy 4.1 so Iโm interested to see how fast the iMX. I seem to remember seeing a limit somewhere in the imx datasheet of 75MHz but Iโll need to dig deeper.
tulip sleet
I think I found a mistake in the documentation for the stm32f405 feather. In htt...
Thanks - I have fixed that. canio is available on the STM32 Feather. @onyx hinge The same guide remarks that CircuitPython has just added SDIO support! However, it is still considered unstable and may see further development. Would you say it is more stable now or should we leave that.
serene briar
Thanks - I have fixed that. `canio` is available on the STM32 Feather. <@!42620...
Thanks. I am working on a project that will use a can bus and wanted to clarify and make things clearer for myself and others.
tulip sleet
FYI, the docs for `aesio` functions don't mention padding: https://circuitpython...
tnx - I have opened an issue for this
onyx hinge
@tulip sleet I don't think we need to say SDIO is experimental on stm32f405 feather anymore
tulip sleet
<@!329766224093249548> I don't think we need to say SDIO is experimental on stm3...
๐ I will update the guide
onyx hinge
appreciate it @tulip sleet
crimson ferry
@prime flower I looked at the Wiznet5k+Requests issue, but don't know how to fix the incompatibility
timber mango
hi, I'm looking a way to get nice animations for Dotstar rbw 2328 stripes
like you can find with processing NEopixels through fadecandy
onyx hinge
@timber mango #help-with-circuitpython is better for "how do I" questions, but check out the LED animations guide from Adafruit: https://learn.adafruit.com/circuitpython-led-animations
timber mango
ah sry, 1st time here
onyx hinge
welcome!
timber mango
thanks
meager fog
@lone axle hihi can you look at https://forums.adafruit.com/viewtopic.php?f=60&t=177110&p=863371#p863371 - it seems a little wierd that a titano with 256KB of RAM cannot run this program. like it seems that some change in the text/label system is hogging all the RAM
lone axle
<@!382939733107408897> hihi can you look at https://forums.adafruit.com/viewtopi...
I can take a look later tonight or over the weekend. But I don't have a Titano to test with. I do think the theory posed about the larger screen size is likely on the right track though. At first glance it looks like the code makes a Bitmap object based on the size of the display. One thing I would recommend is using bitmap_label instead of label in all cases that I have come across it leads to less memory usage, and with longer strings the savings can be fairly significant. It should be easy to try it by switching the import at the top of the file. I'll post a comment on that thread once I have a chance to dig in further.
lone axle
Possibly the Bitmap could be reduced to half the screen size and then scaled by 2 using a Group. That would save some memory I believe at the cost of slightly "chunkier" resolution.
There may also be some savings to be had by not using the PyPortal helper library. It looks like that is being used only for the display and playing sounds at first glance. Those could possibly be done with core APIs only and it would save a few imports.
meager fog
@lone axle i think there's something else, 256KB is a massive amount of RAM.... is the bitmap being made thats 16-bit color?
lone axle
looks like only 4 colors to me:
plot = displayio.Bitmap(GWIDTH, GHEIGHT, 4)
lone axle
@onyx hinge or @mental nexus do you know if using a PCF font would be expected to consume less RAM than the same font, but BDF format? I know there was a loading speed boost, but unsure about other RAMifications ๐
idle wharf
@prime flower are you around ?
tulip sleet
@onyx hinge https://github.com/hathach/tinyusb/discussions/753#discussioncomment-535334 mentions a conflict between the CAN peripheral buffer and USB. Is this some kind of issue for us?
prime flower
<@!311339734339289090> are you around ?
in meeting, ill be free in 25m
idle wharf
in meeting, ill be free in 25m
No rush... I've blocked out time to work on AzureIoT today focusing first on the ESP32SPI
prime flower
@idle wharf ping!
onyx hinge
@lone axle I would expect the final memory usage to be about the same
idle wharf
Hey @prime flower I'm trying to get just the SPI version working again.
https://github.com/askpatrickw/Adafruit_CircuitPython_AzureIoT/compare/fix24-spi
The send is failing... (let me get you a screen grab)
1967.96: DEBUG - Sent!
1968.83: INFO - - iot_mqtt :: connect :: iotc-738e1892-03dd-4ad0-be4a-c69e3c9a5182.azure-devices.net
1968.83: DEBUG - Attempting to establish MQTT connection...
1968.83: INFO - Establishing a SECURE SSL connection to iotc-738e1892-03dd-4ad0-be4a-c69e3c9a5182.azure-devices.net:8883
Traceback (most recent call last):
File "code.py", line 106, in <module>
File "/lib/adafruit_azureiot/iotcentral_device.py", line 160, in connect
File "/lib/adafruit_azureiot/iot_mqtt.py", line 387, in connect
File "/lib/adafruit_azureiot/iot_mqtt.py", line 133, in _create_mqtt_client
File "/lib/adafruit_minimqtt/adafruit_minimqtt.py", line 437, in connect
File "/lib/adafruit_minimqtt/adafruit_minimqtt.py", line 271, in _get_connect_socket
RuntimeError: Repeated socket failures```
I'm using the iotcentral simple example as my test.
I can't see what is wrong with iot_mqtt connect or _create_mqtt_client
I was just about to use the minimqtt spi sample to verify it works first. But if you see something obvious that would be helpful.Correct, bad choice of words before I got the screen ouput. The registration steps do suceed
onyx hinge
@tulip sleet That USB vs CAN memory doesn't ring a bell for me.
tulip sleet
yeah, I don't know what to make of it
prime flower
Correct, bad choice of words before I got the screen ouput. The registration ste...
Your credentials and scope and everything else are correct?
Is the root cert for AzureIoT in nina-fw still correct?
idle wharf
Your credentials and scope and everything else are correct?
That is interesting,... I did have to update the Nina FW. I just got the airlift, so if latest firmware is good then I'd say so.
prime flower
That is interesting,... I did have to update the Nina FW. I just got the airlift...
this is with Azure IoT Hub or Azure IoT Central?
idle wharf
IoT Central, on a known working instance. Verified the scope, id and key are good
prime flower
lol i am confused...hol don
So, registration.py first file is OK?
Or are you running a file which interfaces with adafruit_azureiot/iotcentral_device.py
idle wharf
I'm using the IoT Central simple example to validate the library changes as I go.
https://github.com/adafruit/Adafruit_CircuitPython_AzureIoT/blob/master/examples/azureiot_central_simpletest.py
prime flower
IoTMQTT looks like it matches latest MiniMqtt... https://github.com/adafruit/Adafruit_CircuitPython_AzureIoT/blob/master/adafruit_azureiot/iot_mqtt.py#L111
prime flower
I'm using the IoT Central simple example to validate the library changes as I go...
It looks like it's failing during get_connect_socket in MiniMQTT, I'd start by looking here: https://github.com/adafruit/Adafruit_CircuitPython_MiniMQTT/blob/master/adafruit_minimqtt/adafruit_minimqtt.py#L243 and validating the: host, port, ssl context, and interface
idle wharf
Thanks @prime flower its helpful to have someone validate you're headed on the right path.
I'm also going to run the minimqtt spi sample once to verify nothing in my firmware(s) or dependencies is involved. Not likely I think, but worth checking.
prime flower
Agreed
I'd try to test with only MiniMQTT as much as possible to narrow down the issue if you can
tidal kiln
achievement unlocked! ๐
Adafruit CircuitPython 6.1.0 on 2021-01-21; Adafruit QT Py M0 Haxpress with samd21e18
>>> print("i am not cat")
i am not cat
>>>
Thank for that @tidal kiln
still zephyr
@lone axle talking about pcf fonts .https://github.com/endico/breathable/issues/1
idle owl
@tulip sleet I'm trying to get https://www.serialterminal.com/ to work. It connects only when using that URL, none of the other URLs from maholli, or trying from a local version of the index.html. But I'm not sure it's doing anything once I'm connected. I'm using Chrome.
I can type something in, and it shows up below n next to >. But it doesn't appear to be a REPL or anything.
tulip sleet
maholli only tested it on windows
tulip sleet
did you click the "Connect" button in the upper left?
idle owl
Yes.
tulip sleet
i have similar issues on Linux
idle owl
And it popped up a window for me to click on a device. Chose the CPX, and then kinda nothing.
Hmm ok.
tulip sleet
yeah, I get this:
tulip sleet
i am using Chrome, but I get it on Linux after I try to type somethign
https://googlechromelabs.github.io/serial-terminal/ also works in windows, according to maholli, and it doesn't have the Arduino-style single-line input
but it keeps disconnecting on Linux
idle owl
I have a Windows VM, but that's a whole other bucket of fish.
That works for me on MacOS!
tulip sleet
close the browser and reopen it; sometimes it works after it fails for me, but it doesn't work the first time; it's weird
tulip sleet
yeah, sometimes I get it to work without knowing why
like, what it I do that's it now working?
someone needs to spend some time on this. it may well be a Chrome bug in the implementation of Web Serial.
but the potential is great
idle owl
hmm yeah
tulip sleet
i'd expect the low-level implementations to be different for each OS. I saw something similar (works on some platforms, not others), when trying their gamepad support
idle owl
Restarted browser, still "connects" but then nothing happens.
tulip sleet
if you type ctrl-C or return, then nothign?
idle owl
No response.
The googlechromelabs one works. on MacOS.
the serialterminal.com one doesn't.
tulip sleet
try the googlechromelabs one with a fresh browser
idle owl
I did, it still works
tulip sleet
well, that's as much news as serialterminal.com, maybe
maybe it's a paragraph that says: this is flaky, but try it out; if you're a JavaScript/browser expert, maybe you can help
idle owl
hmm maybe yeah
tulip sleet
the serialterminal.com one looks inspired by Arduino Serial Monitor
idle owl
Tried serialterminal.com in my windows VM, but I'm unconvinced it's not working other than because USB is flaky in a VM.
tulip sleet
i'll try it on a real windows machine
idle owl
It's ok. We don't need to spend more time on it right now. I'll ping Limor with the results and see what she thinks
tulip sleet
@idle owl i just had to power up my desktop Windows machine. the googlechromelabs one works. serialterminal.com is flakier
idle wharf
@prime flower You'll be happy to know that the minimqtt sample worked flawlessly so it is definitely something in the config or he way the library is calling Minimqtt.
I'm going to setup an IoT Hub and work that side as well so I can verify against the IoT Central path. Thanks again !
onyx hinge
@tulip sleet "RM0091" is about STM32F0x1/STM32F0x2/STM32F0x8, so it doesn't relate to STM32F405/407. The latter are the only ST MCUs that have canio enabled
I think CircuitPython is only on F4/F7/H7 so it's not relevant for us
tulip sleet
<@!329766224093249548> "RM0091" is about STM32F0x1/STM32F0x2/STM32F0x8, so it do...
ah, good, whew
marble hornet
@tulip sleet I don't think it's pertinent to the help channel discussion but, regarding reload() vs reset() , reload has some memory it doesn't free (under the hood )right?
tulip sleet
it restarts the VM; it's the same as typing ctrl-d. But the heap starts from fresh
thorny jay
@tulip sleet Super happy to could have help... There is something amazing about that OLED. It was going in "autonomous scrolling, partial or full screen". And if not "powering off" the scrolling from the I VOTED (with old firmware) was able to scroll your Hello World. That screen/driver have some "hidden" feature and seems to do things on it's own. At one point with the garbage random pixel on the screen it was like a starfield effect (ร la StarTrek or StarWar).
jaunty juniper
uuuuuuh USB keyboards are so complicated
stuck elbow
why?
manic glacierBOT
When connecting an SH1107 128x64 OLED FeatherWing to a Feather RP2040 via STEMMA QT I2C, the display does not blank on initialization and displays pixelated "noise" when updated.
This does not occur when using an [SSD1306 1.3" 128x64 OLED Display](https://w...
ornate breach
Does the RP2040 have a preconfigured USB PID/VID?
The chip itself not any of the current dev boards
jaunty juniper
why?
Different layouts introduce multiple types of changes, and I want a code that can for example select all (ctrl-a) and say switch to the 5th tab of firefox (ctrl-5).
5 on a french keyboard is shift-( but the position is the same, and the shortcut is actually ctrl-( not ctrl-shift-( so the keycode is unchanged. That is often the case, where number shortcuts will be based on position rather than character, otherwise it would add modifiers.
a on the other hand is switched with q, so the keycode changes, but the actual shortcut is the same. That is also usually the case, where the letter shortcuts are based on the letter.
So if I want to make a macro keypad that works on multiple setups by only configuring the layout it's a little complicated.
onyx hinge
@ornate breach in bootloader mode it does have a particular PID/VID for sure. after that it gets whatever the loaded software chooses.
onyx hinge
https://datasheets.raspberrypi.org/rp2040/rp2040-datasheet.pdf see 2.8.4.1 "identifying the device" for info about the rom uf2 bootloader
you might also take a look at https://github.com/raspberrypi/usb-pid
onyx hinge
It may not be the fastest way, I see there's only one allocation to date.
ornate breach
I has a company now so many Iโll eventually be purchasing a VID once Iโm making the big bucks ๐
Might be a few years but thatโs okay ๐
manic glacierBOT
jaunty juniper
btw, I tried serialterminal.com (online and locally) on Mac (Chrome) and it was able to connect, but only if I set a baud value. That's a bug in the page, it's undefined if never set (even though the menu says 9600)
stuck elbow
@jaunty juniper just configure keycodes instead of characters, and it becomes simple
manic glacierBOT
manic glacierBOT
mental nexus
Just learned this week about make -j to use all my processor cores for compiling CircuitPython. Just think how productive Iโll be after learning that. ๐คทโโ๏ธ
blissful pollen
Just learned this week about `make -j` to use all my processor cores for compili...
I saw Scott do that on stream a couple weeks ago, same with me!
manic glacierBOT
iron aspen
FYI - I have created a pull request (https://github.com/adafruit/CircuitPython_Community_Bundle/pull/62) for the Community Bundle to add a new uschedule library (https://github.com/cognitivegears/CircuitPython_uschedule) - please let me know if there is anything that I can help with!
manic glacierBOT
thorny jay
Low priority message for @idle owl: I just noticed that I don't have a board.LED thing for the CLUE. So I use led = DigitalInOut(board.WHITE_LEDS). Maybe that could be an issue for your learn guide thing about blink. Regards.
crimson ferry
@thorny jay It does seem like if there is only one LED pin (regardless of color), it should be board.LED.
The trouble comes when there are multiple colors of LEDs on a board... in addition to board.RED_LED and board.BLUE_LED, should perhaps the red LED also get an alias of board.LED so thre's always a board.LED?
jaunty juniper
I would say so
thorny jay
<@!681528999520108564> It does seem like if there is only one LED pin (regardles...
Kattni said one of the LED need the name LED, if there is only the blue it will be the blue. So I guess we need a LED alias for the WHITES in CLUE usecase.
tulip sleet
Just learned this week about `make -j` to use all my processor cores for compili...
I added some advice about make -j to the Building CircuitPython guide a few days ago.
lone axle
Wow. I never knew about that either. That cut a build of mine down from 1m32s to only 39s
idle wharf
Can -j be the default?
onyx hinge
there's not a super great way to make -j the default that I know of
I mean, you sure can do it with a line like this in a Makefile: MAKEFLAGS += -j4 except 4 isn't the right number for everyone
fossil gorge
Does it have to be -jx or can it be just -j?
lone axle
I did just -j no number
onyx hinge
@fossil gorge Generally it's recommended to use -j# where # is the number of cores or threads in your system.
Using no number or a very large number uses more memory (potentially including swap space) and provides no performance benefit. It can even run your computer out of memory entirely.
For instance, the raspberry pi build has over 500 object files in it. "make -j" will try to invoke the compiler in parallel up to that many times!
fossil gorge
Ah, I see. I thought that -j defaulted to the number of processors/cores, not "all the things"
onyx hinge
Maybe make has been changed in the last 15 years, weirder things have happened ๐
fossil gorge
Maybe make has been changed in the last 15 years, weirder things have happened ๏ฟฝ...
Really? Name 42 ๐
onyx hinge
(literally not joking, that's just about the last time I updated my idea about how make works)
fossil gorge
Fair enough
crimson ferry
i tried -j before just reading about -jx, and it reduced make clean from 3m to 1m. nice!
onyx hinge
If the '-j' option is followed by an integer, this is the number of
recipes to execute at once; this is called the number of "job slots".
If there is nothing looking like an integer after the '-j' option, there
is no limit on the number of job slots. The default number of job slots
is one, which means serial execution (one thing at a time).
fossil gorge
-j [jobs], --jobs[=jobs]
Specifies the number of jobs (commands) to run simultaneously. If there is more than one -j option, the last one is effective. If the -j option is given without an argument, make will not limit the number of jobs that can run simultaneously.
From my local man make. on macOS 11.2.3 with M1 proc
kind river
I've also seen make -j$(nproc) used in examples to automatically use the number of cores.
crimson ferry
-j 16 was 5 seconds slower than -j for me
onyx hinge
nproc as a shell command is available on most linux but not on mac
and since we all use Python here, another alternative is: ```NPROC=python3 -c 'import multiprocessing; print(multiprocessing.cpu_count())'
Is that also true on the GitHub build system?
When building core CP that is
onyx hinge
I hope so -- I grabbed that line from micropython-ulab's build which runs on github's build system (linux & mac)
fossil gorge
Worse case, you follow it up with a || # for a default number
onyx hinge
for circuitpython we've just hard-coded -j2 in a number of places, and there's even this ... curious ... ```PARALLEL = "-j 5"
if "GITHUB_ACTION" in os.environ:
PARALLEL = "-j 2"
Hmm
tulip sleet
Re -j: This is what I wrote up a few days ago: https://learn.adafruit.com/building-circuitpython/build-circuitpython#use-all-your-cpus-when-building-3086075-16.
fossil gorge
For macOS, you can use sysctl -n hw.ncpu
mental nexus
On my machine: 3.2 GHz Quad-Core Intel Core i5 I didn't see any real difference in compile time no matter if I used -j or whether it was set to -j, -j1, -j4 or '-j8`. But I perceived it to bog down my machine with larger values. It took between 140-146 seconds for all of these builds (for pyportal). This machine is from 2015, so maybe it's time for an upgrade.
meager fog
@lone axle hihi i see a green dot, i have a req about this guide https://learn.adafruit.com/circuitpython-display_text-library/overview
lone axle
<@!382939733107408897> hihi i see a green dot, i have a req about this guide htt...
Yep, whatcha got?
meager fog
can ya update that guide to talk about fonts too
like bitmap_font
does have to be a lot but folks dont see how to use custom fonts
wait
now im confused, there is a page
@tidal kiln when yr around can ya explain the thing you were asking about
now im confused >.<
tidal kiln
this guide for fonts:
https://learn.adafruit.com/custom-fonts-for-pyportal-circuitpython-display/overview
i think there's enough to dealing with fonts - creating and using, that more info in that guide would help
lone axle
One thing that does seem to be missing is a page or section with a basic example that actually shows the code for loading and using a custom font.
tidal kiln
that display text guide above is good - it covers https://learn.adafruit.com/circuitpython-display_text-library/overview well
think in terms of 3 parts:
- displayio
- text
- fonts
i think we have the first two covered well, but need a little more for the third one
lone axle
What kinds of things would you like to see added or expanded upon in https://learn.adafruit.com/custom-fonts-for-pyportal-circuitpython-display/overview ?
a section showing example code would be a good addition I think. Maybe some "outside the box" font uses like the FontAwesome stuff that Jeff E got into at one point?
tidal kiln
yep. that sort of stuff. like you mention above, not sure theres a good basic example that focuses on just the font aspect
meager fog
ok thanks for clarifyng - i wanted to make sure i got it right ๐
foamy, dya wanna come up with a few things we could do to update that guide and call it a 'project'?
lone axle
Yep, I can do that for sure
tidal kiln
i'm assuming that guide is the best place? the URL is a bit misleading (pyportal) but the actual guide name seems more fitting
lone axle
Yeah, a more general name might be good. Don't know if changing it causes too much trouble though.
manic glacierBOT
stiff pelican
Is this a website issue? Or is the Pynt missing modules it should have?
jaunty juniper
hmmmm there's "py" in the name, I wonder...
nah it's not the only one
ah it's aliases
manic glacierBOT
solar whale
@stiff pelican If I recall correctly, the Pynt actually uses the same build as the PyPortal. It is just renamed.
jaunty juniper
yep
jaunty juniper
there's a PR two messages above to fix it ๐
solar whale
Ah -- sorry - I missed that !!
jaunty juniper
(I was working in that file earlier last week, it's an easy fix)
but it will only apply to future releases
oh alias boards are also not in the support matrix on readthedocs
and I remember noticing that it (the matrix) also doesn't identify the internal adafruit_bus_device due to the CIRCUITPY_* define that enables it not following the format of the others
I guess I'll set my sights on https://github.com/adafruit/circuitpython/blob/main/docs/shared_bindings_matrix.py next
solar whale
I have noted some guides still refer to adding the bus_device library to boards .... I guess it does no harm other than taking up space. I think the core module is used even if the library is on the file system
jaunty juniper
seems so
should I add to this ? https://github.com/adafruit/circuitpython/issues/2021 it's also about missing modules in the support matrix
stiff pelican
Cheers @jaunty juniper and @solar whale . That's a very quick turnaround.
jaunty juniper
to be fair I was a little quick to push it, a minor detail of coding style but I should have sit on it a little bit
plucky flint
๐ Quick heads up folks... the next beta release of Mu is out, feedback most appreciated, blog post about the changes here: https://madewith.mu/mu/users/2021/03/29/beta3.html
We expect there to be some CircuitPython related updates in betas 4 and/or 5. We hope to maintain a fortnightly beta release schedule as things "come together" for a final release soon-ish. (It's ready when it's ready). Thanks for the help and support.
stiff pelican
Hi. Could someone add me to the circuitpythonistas role please? Many thanks.
solar whale
@stiff pelican Done
jaunty juniper
@onyx hinge ok so, I have never responded to a review like that before, github tells me I can click on "insert a suggestion", make your change and that will automatically make a commit once approved ?
onyx hinge
@jaunty juniper I have seen other people use "suggestions", I just create a new commit, push it to the pr, and mark the conversation resolved. Mentioning the person who made the original review is good etiquette too.
onyx hinge
@idle owl if you're not utterly sick of the topic I would like to talk to you about pre-commit and bundle libs, https://github.com/adafruit/cookiecutter-adafruit-circuitpython/issues/121 -- short version of the question is, we want duplicate code detected within which one or more of the following? (A) the library's actual code (B) the examples (C) the tests
idle owl
Low priority message for <@!330227457296957440>: I just noticed that I don't hav...
I need to think about this.
idle owl
The trouble comes when there are multiple colors of LEDs on a board... in additi...
The little red LED should always have LED. I'm working on adding that as I go.
jaunty juniper
I am deep into shared_bindings_matrix (which provides the support_matrix dict) to add aliased boards to the readthedocs Support Matrix page (as well as adafruit_bus_device), and that might actually replace PR#4507 since support_matrix[alias] will then work as expected
idle owl
<@!330227457296957440> if you're not utterly sick of the topic I would like to t...
I'm never sick of talking about pre-commit!
idle owl
<@!330227457296957440> if you're not utterly sick of the topic I would like to t...
Duplicate code checking should ONLY be done on the library code proper. Examples and tests should NOT be checked for duplicate code.
Good morning ๐
onyx hinge
Is there an effective commandline flag to entirely disable duplicate code checking?
idle owl
Low priority message for <@!330227457296957440>: I just noticed that I don't hav...
The NeoPixels are always on board.NEOPIXEL regardless of the number. I think in the case of CLUE, it should be board.WHITE_LED.
ornate breach
Going to attempt to lurk and work this morning
idle owl
Lurk and work. It's like Lunch and Learn only lurkier and workier.
thorny jay
The little red LED should always have `LED`. I'm working on adding that as I go.
Did not notice the little RED near the microUSB.
idle owl
Did not notice the little RED near the microUSB.
Very few boards don't have it. ๐
ornate breach
For the meeting, need to get back into doing more for Circuitpython which is my goal for my new Thursday stream ๐
thorny jay
The white one are "aggressive"... maybe usefull for color detection, but not super friendly.
onyx hinge
@idle owl is there an effective commandline flag to entirely disable duplicate code checking?
idle owl
@thorny jay Exactly what they're for.
idle owl
<@!330227457296957440> is there an effective commandline flag to entirely disabl...
No.
That is the whole thing.
You can't disable it.
You can only increase the threshold to make it stop running on the whole repo. Or force Pylint to run on files separately.
@onyx hinge Do you understand now why there are shenanigans?
onyx hinge
To me it looks like pylint --disable=duplicate-code,... effectively disables the check
locally I have pylint 2.6.0, is that the current?
idle owl
Unless that was added in the last few weeks, it is a red herring.
2.7.1 is what we're using
onyx hinge
upgraded to 2.7.2, and the commandline --disable= also works for me
idle owl
2.6.0 had not fixed the duplicate check yet.
onyx hinge
pylint 2.7.2
$ pylint --disable=pointless-statement,missing-module-docstring *
************* Module b
b.py:1:0: R0801: Similar lines in 2 files (long spew omitted)... (duplicate-code)
Your code has been rated at 9.95/10 (previous run: 9.95/10, +0.00)
$ pylint --disable=duplicate-code,pointless-statement,missing-module-docstring *
Your code has been rated at 10.00/10 (previous run: 9.95/10, +0.05)
Checking change logs.
onyx hinge
I recall you checked putting the disable in as a comment in a .py file
idle owl
We also tried adding it to .pylintrc but I swear we tried it on command line as well.
I guess we didn't. So that was all pointless.
thorny jay
@idle owl I wanted to use led_animation on the QT-Py, just an I2C sensor, a bit of code and blinking RGB using it. But I ran out of flash space. Works fine on the haxpress. I just needed colorcycle that only needed color for the Rainbow. So I started by removing everything except the two__init__, colorcycle.mpy and I replaced color.mpy by a stripped down version as color.py with only the needed color for the Rainbow (the MPY version was running out of memory). So now I can do rainbow, blinking, blinking between two color and still color. I could strip down a littly bit the init too as it contain a lot of thing I don't use. Would you consider a small version of led_animation?
idle owl
@onyx hinge Whatever. If you made it work, then PR to cookiecutter, and create an Adabot patch. We have to fix all the libs if that's the case.
onyx hinge
@idle owl yeah the second part I'm not thrilled about
I'll update my comments on that cookiecutter issue and if after reading it you feel like we've ended up on the same page then we can initiate actually doing all the stuff...
idle owl
<@!330227457296957440> I wanted to use led_animation on the QT-Py, just an I2C s...
We don't want to support two versions of libraries. It's a support nightmare. You're welcome to submit it to the Community Bundle. When I wanted animations on a QT Py, I put the code in code.py - https://learn.adafruit.com/qt-py-and-neopixel-leds/basic-animations-for-qt-py
thorny jay
We don't want to support two versions of libraries. It's a support nightmare. Yo...
I kind of needed "blink.animate()" way of working as animation is not the sole goal but the indication of the sensor value. Here a CO2 sensor. I try to find a "language" where the color, or blinking speed or alternating of color is an "universal" or "understandable" indicator. Like GREEN/YELLOW/RED for traffic light. "Rainbow" could mean the air quality is perfect. "RED/BLUE" like police light mean emergency to open the window.
idle owl
Fair enough.
thorny jay
I'll to incorporate what I need in the body. That library inflated to many usecase, but too much for QT-Py. I believe the QT-Py RP2040 will solve all the space and memory issue. ๐
But as this is for COVID refresh air indicator in school... availability is more important than a lot of other criteria.
orchid basinBOT
idle owl
I'll to incorporate what I need in the body. That library inflated to many useca...
That is the plan ๐
idle owl
But as this is for COVID refresh air indicator in school... availability is more...
That makes sense.
onyx hinge
if coding in arduino would let you attain your other goals, it's worth considering too
thorny jay
Yeah, that's an option. Now I have a PoC and it is working... it's for a friend that does not seems to be afraid to solder and program. So if he want to do better, he can.
orchid basinBOT
onyx hinge
Just realized yet another wrinkle with pre-commit & pylint ... ugh, it's complicated!
idle owl
Yes.
onyx hinge
I think there is a possibility that duplicate code checking within the main package is not correct under pre-commit because of reasons
idle owl
I believe it.
onyx hinge
@idle owl OK I wrote a long new comment on https://github.com/adafruit/cookiecutter-adafruit-circuitpython/issues/121 but ultimately I'm concluding that the status quo isn't super broken, so maybe not touch it right now. But I do want the thoughts written down while I've got a good grasp of them.
idle owl
Makes sense.
The getting the thoughts down. I haven't read it yet, no idea if your thoughts make sense to me yet.
onyx hinge
fair enough. you have other priorities today anyhow
fossil gorge
Lurk and work. It's like Lunch and Learn only lurkier and workier.
But less worky
onyx hinge
<@&356864093652516868> hey all! the weekly meeting will be taking place in about an hour. Scott will be your host today. Please get your notes in the doc! No need to do anything if you'll just be listening in. https://docs.google.com/document/d/1CXqf-nL0NEFt2aXH2ouDlDym6o2VWNz8JRLPNmWUnQQ/edit?usp=sharing
thorny jay
Anybody speaking?
idle owl
Nope.
fossil gorge
Listening only for now
idle owl
@fossil gorge Still listening only?
fossil gorge
For now, yes
still zephyr
Listening only
thorny jay
Is that going to be iOS only? The BLE workflow?
fossil gorge
Will BLE include desktop support as well?
slender iron
idle owl
@fossil gorge Still listening only?
fossil gorge
No, I'm "fully here" now
idle owl
(Also you're blinking intermittently like you have audio coming through.)
fossil gorge
(Also you're blinking intermittently like you have audio coming through.)
Thanks. Now sure why that is, since I'm on Push To Talk, but muted to be safe
idle owl
Happy early birthday, @blissful pollen!
fossil gorge
That looks SO slick!
blissful pollen
Happy early birthday, <@!228322543507865601>!
Thanks! Would be more fun if it wasn't going to be -8C ๐ฆ
idle owl
Thanks! Would be more fun if it wasn't going to be -8C ๐ฆ
Oi, yeah for sure.
fossil gorge
Thanks! Would be more fun if it wasn't going to be -8C ๐ฆ
May not be fun weather, but if that's the biggest concern, you're in good shape ๐
๐ฐ
thorny jay
There was the "use bytearray" advice.
And use binary font (unsure).
Reduce color depth.
Avoid recursivity?
Use generator rather than build big list.
mental nexus
Thanks all!
still zephyr
Thanks!
blissful pollen
Have a great week everyone!
thorny jay
Thanks.
gilded cradle
Thanks
lone axle
๐
slender iron
idle owl
Cheers!
manic glacierBOT
@kevinjwalters The build artifacts for this PR are here: https://github.com/adafruit/circuitpython/actions/runs/688127024. Just scroll down a bit and download the .zip file for the board. Or in the "All checks have passed" section above, click "Show all checks", then choose a board, and then find the "Artifacts" dropdown just above the white-on-black logging part of the page.
slender iron
Here is the notes document for Mondayโs CircuitPython Weekly meeting. It is at the normal time of 11am Pacific / 2pm Eastern here on Discord. Everyone is encouraged to attend! Please add your hug reports and status updates even if youโll be attending the meeting - itโs super helpful! If you are unable to attend but would still like to include updates, feel free to include them in the notes and weโll read them off during the meeting. Hope to see you there! <@&356864093652516868> https://docs.google.com/document/d/1sRuIt_eSshrvOPdPVI2f9sZlvCzpVltRkLHKIa3MCHY/edit?usp=sharing
slender iron
Here is the notes document for Mondayโs CircuitPython Weekly meeting. It is at t...
alpine nimbus
i can see items in this channel in the livestream but I don't see them here in discord. hmm
slender iron
what livestream?
alpine nimbus
the weekly meeting. wasn't i watching it live?
slender iron
It was three hours ago or so
alpine nimbus
oh, doh!
slender iron
๐
we don't stream to youtube live
but you can jump in here and join the circuitpython voice channel to listen live
alpine nimbus
that's ok. i didn't realize it wasn't live and was very confused
slender iron
np ๐
alpine nimbus
I just wanted to make a comment about memory usage. Last week I ran into a memory issue last week that kmatch and jose david were kind enough to help with and I finally got things to work.
I wrote some text to the screen on a Clue and replaced it a few times and then drew an emoji bitmap (not a huge one)
It seems like it should have placed only modest demands on the board. but that was enough to cause fragmentation and I couldn't load the bitmap
alpine nimbus
My main point is that the first thing everyone's first thought was to stop using the clue library. Libraries like that are supposed to make things easy for people like me. I did something small and the first thing I'm supposed to do is stop using the library? LOL
I could understand If I was doing something big
slender iron
ya, it's the challenge of having a library to do everything, even the things that you don't want to do
alpine nimbus
Also, using memfree, it looks like the clue library was barely using any memory
๐คทโโ๏ธ
slender iron
my preference would be to move away from those monolithic libraries in favor of smaller composable components
if the failing allocation was large, it would be a fragmentation issue
my oldest deep dive covers this: https://www.youtube.com/watch?v=baa5ILZTRkQ&list=PLjF7R1fz_OOXBHlu9msoXq2jQN4JpCk8A&index=49&t=4s
alpine nimbus
I moved the bitmap allocation to the beginning and things seem to be fine now. and made a few other tweaks
slender iron
๐ makes sense
left mirage
are there any known issues with enabling the uctypes module (#define MICROPY_PY_UCTYPES (1)) when building for the trinket m0? i was able to compile (by changing the warning flags for one step) and haven't run into any issues yet
alpine nimbus
@slender iron Thanks, that link sounds helpful!
slender iron
are there any known issues with enabling the `uctypes` module (`#define MICROPY_...
probably not. it's likely off to save space