And I recently noticed that I2SOut is restricted to 16-bit output (on rp2xxx at least). I'd love to keep the audio system open to other bit widths, but keeping it all at 16-bit definitely simplifies things.
#circuitpython-dev
fleet hollow
blissful pollen
I'll look at the distortion PR first for sure. And I agree it may be easier to squish the audio to mono at the end vs handle it everywhere. The smaller boards are not using any crazy audio processing anyways.
I do wonder also is 8 bit being use that much either.
onyx hinge
Howdy folks. I have some big news: In 2025 I'm finishing up my work at Adafruit. I'm not quite comfortable calling it "retirement", but we'll see what the future brings. https://social.afront.org/@stylus/113691513879680883
blissful pollen
Howdy folks. I have some big news: In 2025 I'm finishing up my work at Adafruit....
First off congratulations, I know how tough a decision ones like this can be. I am also jealous of the travels you will have in April.
Thank you for all your contributions and help over the years since I joined. I cannot count how many times I saw something you were working on and went "wow cool" to myself.
onyx hinge
I appreciate hearing that, @blissful pollen
manic glacierBOT
jaunty juniper
@willow totem I see you added Circuitpython and tinyuf2 PIDs for the Lilygo T-QT [pro] to the Espressif usb-pids repository (around 2023), but the board is not implemented in CP or tinyuf2 that I can see.
I have a T-QT pro (N4R2) and was preparing to PR it to both repos, I think I can use those PIDs then ?
I also implemented the N8 model though I can't test it, but there isn't much difference anyway (I followed the QTPY S3 which have the same variants). I believe they can share the same PID.
(these PIDs https://github.com/espressif/usb-pids/pull/80)
willow totem
Yeah think that should be totally fine, but verify they're correct, I was on a t-displayS3 and t-qt mission at the time (its a t-qt pro I have too)
oh sorry, vid/pid mad, I've no idea what I'm saying, it's 1am here and I've had a bit of pub time
Thanks for your websockets lib too, recently had to use it for an RC joystick to RC forklift 😁 (still unfinished like so many things - but works)
jaunty juniper
oh nice, I think I might want to look into using ConnectionManager in the future
willow totem
Just reread the PR, yeah that should be fine
I wondered about exposing some timeout stuff, but otherwise nothing came up, it just mostly worked
jaunty juniper
I'm getting back on the Circuitpython saddle after a long time away and I have so many projects I want to update 😛
willow totem
Oh exciting, you bring some interesting things usually, I often find myself down similar paths that you've already beaten/trodden
Oh I might have already done half the work for the t-qt, you might want to peruse these branches in my fork:
jaunty juniper
I'll take a look then, and I'll submit PRs with those PIDs tomorrow (it's past 2AM here, I try not to PR after midnight 😄 )
willow totem
Wise, very wise (he says having released at 3am on Friday night), but it was only delayed by some good testing 😁 and I wanted the pico2w out for holidays
onyx hinge
@errant grail "Luxurious Cedargrove Atmospheric Reading System (LCARS)"
errant grail
<@339922377746022402> "Luxurious Cedargrove Atmospheric Reading System (LCARS)"
Hilarious!
onyx hinge
Let this serve as a reminder that our next Discord meeting is not until January 6, at the normal 1PM ET/10AM PT time. Happy holidays!
manic glacierBOT
Nevermind, I have ruled out power rail issues by virtue of it having done... something?? Must be something funny about the startup sequence of display modes.
manic glacierBOT
manic glacierBOT
Also, I notice that the bit clock signal looks like this (example pulled off the net, not actual signal) instead of a true square wave :
On my other ESP32 board, it is a true square wave, with sharp corners.
blissful pollen
@fleet hollow Just to give you an update I started the review but have been fighting some stomach bug so got slowed down.
lone sandalBOT
manic glacierBOT
mortal kernel
A common CP debugging problem occurs when attempting to use intensive serial logging while reproducing a timing-related problem. The delays caused by writing log data to a UART at 115Kbps (3 to 12 ms per line depending on its length) are large enough to affect the problem under debug. On devices with large memories (ex., RP2350 with 8MB PSRAM), it's feasible to allocate a large log buffer that receives log output. In its simplest form the buffer wraps when its limit is reached, so the most recent log entries are available.
Logging to memory is a nice compromise between logging speed and simplicity of logging implementation.
tulip sleet
A common CP debugging problem occurs when attempting to use intensive serial log...
Good idea! The SAMD UF2 bootloader also does in-RAM logging, though I have yet to get it to work properly.
mortal kernel
I'm playing with it now. Do you think it would be worthwhile to expose it as a Python module?
(by "it" I mean logging to PSRAM, not the SAMD UF2 bootloader stuff)
tulip sleet
is this logging you are doing in Python or C?
tulip sleet
I'm playing with it now. Do you think it would be worthwhile to expose it as a P...
Could it be an additional handler added to https://github.com/adafruit/Adafruit_CircuitPython_Logging ?
mortal kernel
It's in C. I've added an additional "device" following the mp_*printf conventions.
That way I capture time-stamped REPL messages as well as log messages written specifically to the log device.
That logging module is nice. I think I can work with it.
tulip sleet
so what would the Python API be? To access the buffer and/or to make log entries?
Since it was in C a module that was normally off would make sense, since there's no use for it unless you add debugging printouts, I think.
mortal kernel
I'm thinking mostly about C logging for CP debug right now, so it's under a build-time flag CIRCUITPY_FAST_LOGGING. As it is now, when the flag is set, the buffer gets allocated and logging starts immediately. A Python module that allows configuration of buffer size, logging on/off, and buffer dump to a stream is the minimum I'd like.
tulip sleet
CIRCUITPY_RAM_LOGGING?
mortal kernel
Of course logging from Python would be desirable. Sure, RAM makes more sense.
I've also fixed a slew of bugs in serial.c. It no longer hangs up CP when a message goes out before the UART is initialized.
A given port can now move the initialization of the serial console earlier in its startup with consequences.
Also added general purpose memory dump. Here's a sample:
112.071(0.087): [ 112071] ETHTX itf=w0 len=1514 src=28:cd:c1:ff:b9:17 dst=00:1b:21:f0:77:ee subtype=0800
112.078(0.007): IPv4 tlen=1500 id=57 flags=0 frag_offset=0 protocol=6 src=192.168.001.219 dst=192.168.001.217
112.088(0.010): TCP src=80 dst=53596 seq=00004520 ack=5ae2db14 flags=5018
112.095(0.007): 0x0000: 001b21f0 77ee28cd c1ffb917 08004500 : 05dc0039 0000ff06 30dec0a8 01dbc0a8 : ..!.w.(.......E. ...9....0.......
112.107(0.012): 0x0020: 01d90050 d15c0000 45205ae2 db145018 : 2c363a04 00007d2e 70757265 2d666f72 : ...P.\..E Z...P. ,6:...}.pure-for
112.119(0.012): 0x0040: 6d202e70 7572652d 696e7075 742d312d : 347b7769 6474683a 3235257d 2e707572 : m .pure-input-1- 4{width:25%}.pur
112.131(0.012): 0x0060: 652d666f 726d2d6d 65737361 67652d69 : 6e6c696e 657b6469 73706c61 793a696e : e-form-message-i nline{display:in
112.143(0.012): 0x0080: 6c696e65 2d626c6f 636b3b70 61646469 : 6e672d6c 6566743a 2e33656d 3b636f6c : line-block;paddi ng-left:.3em;col
112.155(0.012): 0x00a0: 6f723a23 3636363b 76657274 6963616c : 2d616c69 676e3a6d 6964646c 653b666f : or:#666;vertical -align:middle;fo
112.167(0.012): 0x00c0: 6e742d73 697a653a 2e383735 656d7d2e : 70757265 2d666f72 6d2d6d65 73736167 : nt-size:.875em}. pure-form-messag
112.179(0.012): 0x00e0: 657b6469 73706c61 793a626c 6f636b3b : 636f6c6f 723a2336 36363b66 6f6e742d : e{display:block; color:#666;font-
excellent
another thing we have not had is a pervasive macro to do mp_printf(&mp_plat_print, ...) and handle multiple args. I haven't taken the time to figure out how to do that with the preprocessor. It would save typing time. You'd have a similar macro for the RAM logging.
mortal kernel
Yes! I've factored it somewhat, but there are still several different ways to write to the serial console, various flavors are used in various places to accomplish the same thing.
Since C doesn't do overloading, at least PRINTF and VPRINTF variants are needed.
tulip sleet
Since C doesn't do overloading, at least PRINTF and VPRINTF variants are needed.
I found some examples in specifc modules. for a macro, it is just:
#define MP_PRINTF_DEBUG(...) mp_printf(&mp_plat_print, __VA_ARGS__)
Here's the code that motivated the idea that VPRINTF was needed:
static inline void cyw43_printf(const char *fmt, ...) {
const uint32_t max_len = 132;
// leave room at end of buffer for possible \r\n and null terminator
char buf[max_len + 2];
uint32_t bidx = 0;
if (cyw43_last_printf_line_ending) {
uint32_t now = time_us_32() / 1000;
uint32_t delta = now - cyw43_last_printf_time_us;
cyw43_last_printf_time_us = now;
bidx += snprintf(&buf[bidx], max_len - bidx, "%lu.%03lu(%01lu.%03lu): ", now / 1000, now % 1000, delta / 1000, delta % 1000);
}
va_list args;
va_start(args, fmt);
uint32_t max_buf_len = max_len - bidx;
uint32_t print_len = vsnprintf(&buf[bidx], max_buf_len, fmt, args);
bidx += print_len;
va_end(args);
Here I'm taking a variable number of arguments, but to pass on those args I need to capture them as a va_list object and pass them on to a print specialized to handle it. In this example I'm printing to memory, but if I were using macros to access mp_printf I'd also need one that called mp_vprintf.
jaunty juniper
I have a Lilygo t-watch (ESP32 not S3) that doesn't have a REPL anymore in 9.2, but comes back when I empty the sdkconfig file which contained the following:
CONFIG_ESP_CONSOLE_UART_CUSTOM=y
CONFIG_ESP_CONSOLE_UART=y
CONFIG_ESP_CONSOLE_UART_CUSTOM_NUM_0=y
CONFIG_ESP_CONSOLE_UART_NUM=0
CONFIG_ESP_CONSOLE_UART_TX_GPIO=1
CONFIG_ESP_CONSOLE_UART_RX_GPIO=3
CONFIG_ESP_CONSOLE_UART_BAUDRATE=115200
I'm not sure if that means we lose the esp-idf output and if there's something that should be done about it or we leave it like that.
manic glacierBOT
jaunty juniper
oh but also the web workflow is unreachable in CP9
tulip sleet
I have a Lilygo t-watch (ESP32 not S3) that doesn't have a REPL anymore in 9.2, ...
Fairly recently we took away all those defs from a number of boards. I can't find the PR right now, but Scott made it all be default. You can see in mpconfigboard.h:
// UART pins attached to the USB-serial converter chip
#define CIRCUITPY_CONSOLE_UART_TX (&pin_GPIO1)
#define CIRCUITPY_CONSOLE_UART_RX (&pin_GPIO3)
Maybe that board was missed.
web workflow should be working
jaunty juniper
I might go try a debug build to see what the web workflow is doing
haven't had issues with other boards
except maybe I suspect some instability when web, USB and some other activity like neopixel animations are working together (on the Fun House) but I haven't investigated those yet
tulip sleet
you could try loading some stock ESP32 build (if the UART pins are the same)
manic glacierBOT
jaunty juniper
@tulip sleet it seems like my issue with the Lilygo T-Watch was that the web workflow was started as soon as the settings.toml file was uploaded in CP8, while it required a hard reset in CP9. Because it worked without it in 8 I didn't even think of that... Also on any other board I would have, but the watch having a battery and no reset button, it requires holding its button to turn it off then on (or microcontroller.reset())
tulip sleet
<@329766224093249548> it seems like my issue with the Lilygo T-Watch was that th...
got it... I don't remember this change specifically, but it was probably part of a re-org internally.
jaunty juniper
so I'll soon PR the serial fix and it should be good
austere acorn
I want a lil help
If the help is about core CircuitPython development, ask away.
If the help is about something else, pick a different channel (on of the "help-with-..." channels preferably), and then ask away.
wraith crow
I have the p4 ev board and did have the firmware running. I am not home at the moment but I seem to remember it took a couple tries to get things going. I'll take a look again tonight.
manic glacierBOT
manic glacierBOT
So something is different. It sounds like you are using the same ports, but here's a picture -- maybe you can both describe exactly which ports you are using:
(from https://docs.espressif.com/projects/esp-dev-kits/en/latest/esp32p4/esp32-p4-function-ev-board/user_guide.html)
What is the version number printed on the board?
What is the part number of the ESP32-P4?
What is the part number of the f...
mortal kernel
@onyx hinge mpconfigboard.mk for raspberry_pi_pico2_w sets CYW43_PIO_CLOCK_DIV_INT=3. Why was this value selected?
onyx hinge
@mortal kernel that's a reasonable question. I do not recall. I almost certainly copied it "from somewhere".
if it's wrong then let's change it.
mortal kernel
<@849815081251766292> that's a reasonable question. I do not recall. I almost ce...
Good to know. It probably came from MP. There are differences between the RM2 and the CYW43439 in the way that gSPI selects its phasing for RM2 to host responses that are causing stability problems. See https://github.com/micropython/micropython/pull/16057. If the raspberry_pi_pico2_w will tolerate /2 my opinion is we should change it.
onyx hinge
Removing a non-default setting so we get the SDK default setting seems like a good idea. do you have a Pico 2 W to test with?
mortal kernel
Removing a non-default setting so we get the SDK default setting seems like a go...
I don't, but I will order a couple today.
onyx hinge
Another factor that may affect the RM2's gSPI clock measurement leading to unpredictability is the PIO clock divider's insertion of "leap" cycles to maintain the specified clock rate. The resulting jitter may cause measurement of small clock samples by the RM2 to be inaccurate causing it to change phasing erratically.
Noticed this in your comments on https://github.com/micropython/micropython/pull/16057 @mortal kernel. Do 'leap cycles' apply to integer divisors or only int+frac divisors?
mortal kernel
(I live for that Adafruit THANK YOU on a complete order, always makes my day). My understanding is that any divisor that's not a power of 2 will jitter. The PIO divider accumulates fractional residue and inserts an extra "leap" cycle whenever it exceeds a full processor clock cycle.
onyx hinge
I'd check out https://datasheets.raspberrypi.com/rp2350/rp2350-datasheet.pdf section 11.5.5 "clock dividers" to see if it matches your mental model.
mortal kernel
Thanks! I'll be AWK for the next 6 hours or so.
manic glacierBOT
manic glacierBOT
orchid basinBOT
lone sandalBOT
orchid basinBOT
manic glacierBOT
ionic elk
@slender iron currently working on an alternative to FourWire called PreambleFourWire that replaces fourwire's use of a command/data pin with command/data preamble bytes to support the M5Paper's read-write SPI protocol. I think the existing EPaperDisplay class should be able to fit on top of that without too much craziness but I'm still learning the assumptions BusDisplay makes about the devices it supports.
FourWire could be modified directly to support preamble bytes but would require changes across the whole codebase.
tulip sleet
<@252717193496756235> currently working on an alternative to `FourWire` called `...
Scott is off until the New Year, btw
crimson ferry
some (espressif) boards have this in mpconfigboard.h, but many don't:```
#define CIRCUITPY_CONSOLE_UART_RX DEFAULT_UART_BUS_RX
#define CIRCUITPY_CONSOLE_UART_TX DEFAULT_UART_BUS_TX
@crimson ferry py/circuitpy_mpconfig.h:#if defined(DEFAULT_UART_BUS_TX) && defined(DEFAULT_UART_BUS_RX) py/circuitpy_mpconfig.h:#define CIRCUITPY_BOARD_UART_PIN {{.tx = DEFAULT_UART_BUS_TX, .rx = DEFAULT_UART_BUS_RX}} based merely on a text seach it seems those DEFAULT_ macros are used to define what serial connection board.UART() creates. If those are in use by the console instead, then I don't think board.UART will work, or else it will interfere with the REPL.
crimson ferry
console is typically over USB though right? so why in an ESP32-S3 would those be doubled up?
tulip sleet
these are usuallly on chips without native USB (e.g. plain ESP32), and some USB-capable chips with extra USB connectors (or no connection to the native USB!)
tulip sleet
which one? e.g. lilygo S3 watch has no USB connector
crimson ferry
it has a USB-C connector
tulip sleet
from 'mpconfigboard.mk` for that board:
CIRCUITPY_CREATOR_ID = 0x1BBB0000
CIRCUITPY_CREATION_ID = 0x00AB0001
# This board doesn't have USB by default, it
# instead uses a CH340C USB-to-Serial chip
CIRCUITPY_USB_DEVICE = 0
CIRCUITPY_ESP_USB_SERIAL_JTAG = 0
WHY they did that, I have no idea
crimson ferry
(I don't have that board, I'm just trying to find a good starting template for the Waveshare ESP32-S3-ETH)
tulip sleet
see if has a serial chip or not
crimson ferry
it (the -ETH board) has native USB
stuck elbow
WHY they did that, I have no idea
probably copied an esp32 project...
tulip sleet
(I don't have that board, I'm just trying to find a good starting template for t...
so in that case, pick a more conventional S3 board with native USB broken out to the connector
ionic elk
Scott is off until the New Year, btw
no problem, just hopin he can chip in once he gets back
crimson ferry
is this a recent addition to the process?
ModuleNotFoundError: No module named 'minify_html'
make: *** [build-waveshare_esp32_s3_eth/autogen_web_workflow_static.c] Error 1
no, it was added with the web stuff
the solution is to pip install the missing modules after activating idf
nimble sparrow
Has anyone mooted adding a HSTX flavour to FourWire for the rp2350?
stuck elbow
Has anyone mooted adding a HSTX flavour to FourWire for the rp2350?
I think someone played with that with pio on the rp2040 but it's a very vague memory
nimble sparrow
The hstx peripheral is new in rp2350 but I could imagine someone playing with more generic PIO versions too
stuck elbow
generally speaking there is no benefit in porting fourwire to hstx, I created an issue for it some time ago and thete was some discussion
on a phone right now so can't search for it easily
nimble sparrow
It's mainly that it can be a "free" faster gSPI peripheral without the loss of a PIO unit, but I agree the gain is marginal!
manic glacierBOT
manic glacierBOT
crimson ferry
Is there a way to capture the debug output on an ESP32-S3 if the debug UART pins aren't broken out?
edit: found TXD0 on the Feather S3 TFT
steep cove
HI! nice to meet you! And happy holidays to everybody. Im looking on the website (https://circuitpython.org/downloads) for the waveshare 2350 eth (https://www.waveshare.com/wiki/RP2350-ETH) micropython version, but I cant find it out. How may I request that version to be added? (or how may I compile by myself)
crimson ferry
@steep cove https://learn.adafruit.com/how-to-add-a-new-board-to-circuitpython often you can find a similar board and use that as a template
(I just added the Waveshare ESP32-S3-ETH)
steep cove
<@373963819506073600> <https://learn.adafruit.com/how-to-add-a-new-board-to-circ...
THANK YOU!
jaunty juniper
oooooh I'm encountering an issue in discotool manager and circup with figuring out version numbers from mpy files with MPY version 6:
adafruit_httpserver 0.0.0 4.5.10 Major Version
this is because it scans all the mpy files for any \d+\.\d+\.\d+ string and keeps the last result. But it turns out one of the httpserver files contains a string that looks like a version number: an IP address (0.0.0.0) in some default argument.
I think I'll make it scan __init__.py first, and also stop scanning for other files as soon as a version number is found in one of the files
or at least stop updating the version number, since we want to scan for bad files
(or at least I did when I wrote the original code I believe)
onyx hinge
@jaunty juniper modern circuitpython has a program called "mpy-tool" in the source tree. However, it is not designed to be used outside the source tree. Hypothetically, though, you could use it to disassemble an mpy file and accurately find the assignment of __version__: ```$ python tools/mpy-tool.py -d boo.mpy | grep -B1 "STORE_NAME version"
10:02 LOAD_CONST_STRING 1.2.3
16:03 STORE_NAME version
I don't know how much effort would be needed to make mpy-tool distributable separately (e.g., as pip install circuitpython-mpy-tool-9 or so)
jaunty juniper
at least it looks self-contained but yeah
steep cove
(I just added the Waveshare ESP32-S3-ETH)
apologize, maybe it is a bit off topic, but you showed me a piece of gold: is the wireless wifi/bluetooth possible to turn off completely?
crimson ferry
@steep cove depends how completely you mean... if you don't import wifi or bluetooth, they won't get init'ed. If you are looking for just ethernet, there are other options too: WIZnet Pico boards (including RP2350), or a small ethernet add-on that can be used as an SPI peripheral on any MCU (https://gist.github.com/anecdata/bccd0cf19518b2cce85170c95ddbfae9?permalink_comment_id=5029825#gistcomment-5029825)
onyx hinge
@jaunty juniper it uses at least one other file from the source tree, ```py
sys.path.append(sys.path[0] + "/../py")
import makeqstrdata as qstrutil
looking for eth+camera with POE opportunity. But honestly having the Wifi/Bluetooth option is a nice idea. Im just thinking about "they consume energy", so being able to shut them down completely if not used is interesting
so basically @crimson ferry if I dont load python module, bluetooth and wifi will not be init'ed and they will not turn on, and they will not consume energy. This sounds quite amazing as board. Practically what I was looking for, with the plus of bluetooth/wifi. Thank you very much!
crimson ferry
so basically <@360465312518504448> if I dont load python module, bluetooth and w...
not using radio energy, but it's not like the circuits are powered down
crimson ferry
region 'dram0_0_seg' overflowed by 15408 bytes is flash? Trying to make a (ESP32-S3) debug build, turned off ULAB and then 17 other seemingly-non-trivial modules successively (with rabbit holes along the way for modules that are needed by the core). Each time I remove more modules, the overflow only drops by a few-to-a-couple-hundred bytes. Any suggestions?
I suppose I could change the partition table, but pretty soon I won't have a good control group to compare behavior against a non-debug build
crimson ferry
Is there a way to build but force selection of an alternate partition table, like partitions-4MB-no-ota?
tulip sleet
Is there a way to build but force selection of an alternate partition table, lik...
See for example boards/adafruit_esp32s3_camera/mpconfigboard.mk
crimson ferry
that worked, thanks!
well, it built, but it doesn't run... invalid header: 0xffffffff repeatedly in the debug console
crimson ferry
it doesn't seem to be picking up the new partition table
crimson ferry
I just overwrote the default partition map, and it seemed to take, but still overflowing. I don't think I'm solving the right problem.```
Memory region Used Size Region Size %age Used
iram0_0_seg: 96112 B 152 KB 61.75%
iram0_2_seg: 1219644 B 7864288 B 15.51%
dram0_0_seg: 172064 B 152 KB 110.55%
drom0_0_seg: 341746 B 4128736 B 8.28%
rtc_iram_seg: 44 B 8168 B 0.54%
rtc_slow_seg: 0 GB 16 B 0.00%
rtc_data_seg: 4144 B 8168 B 50.73%
rtc_reserved_seg: 24 B 24 B 100.00%
extern_ram_seg: 0 GB 7680 KB 0.00%
d'oh! n/m, I cut-and-pasted something wrong hours ago and have been building for the s2 TFT variant, not the s3 ...all good now o_O
lone axle
if anyone ends up having a moment and can review https://github.com/circuitpython/CircuitPython_Library_Screenshot_Maker/pull/28 it's appreciated. This is essentially a followup fix for connection_manager including the settings.toml file in the screenshot the previous attempt to resolve that had omitted the .mpy extension which was needed.
onyx hinge
@fleet hollow @blissful pollen I hope to be able to review audio PRs today. There are several open. Which one should I start with?
jaunty juniper
if anyone ends up having a moment and can review https://github.com/circuitpytho...
I wonder if you should make it ignore extensions ? (like filter .mpy out from the list of files before testing without it). If a module moves from a single file to a package you would need to update the tool (but maybe that's ok)
onyx hinge
These "new commits" messages seem to be new. Is it something that's desired?
lone axle
I wonder if you should make it ignore extensions ? (like filter `.mpy` out from ...
That does sound like a better approach to me. Thank you! That would also prevent the possible mistake of having it incorrect in the list. I'll make that change.
onyx hinge
howdy circuitpythonistas (no ping) -- just a reminder that we're taking the holidays off and not holding the Discord meeting. Our next meeting is January 6! Hope to see you there. If you're around, though, you might want to work on your CircuitPython 2025 post! See pinned messages for how to do that.
blissful pollen
<@527668354568159232> <@228322543507865601> I hope to be able to review audio PR...
I was reviewing 9776. Did a code review gotta put it on hardware still.
fleet hollow
I should be able to address updates to 9776 this evening. Thanks for the reviews!
By the way, the bi-directional i2s pr is very experimental, and I would love some insight on it. It's functional but not reliable.
onyx hinge
you're welcome. thanks for sticking with it!
kind river
Are there any plans for a port to the Pimoroni Presto? I just got mine and I'm down to help with one.
stuck elbow
Are there any plans for a port to the Pimoroni Presto? I just got mine and I'm d...
I think that would be a question for pimoroni
kind river
I think that would be a question for pimoroni
They use micropython
stuck elbow
They use micropython
they actually have their own fork, just like circuitpython, just never renamed it
but nevertheless, they are in the best position to add support for their hardware to cp
ionic elk
Are there any precompiled target specific libraries used in Circuitpython already?
My buddy and I are trying to add MicroROS as an optional feature for some Espressif targets but we don't want to add obnoxious ROS build system stuff like their package manager etc. We were thinking maybe we could precompile some .a packages and have them get linked in with a build flag maybe? Or as an "alt build", does anything like that already exist?
crimson ferry
@kind river if Pimoroni isn't interested, Pico 2 W firmware would probably be fine for now, with the peripherals supported in python. Or, you could PR a board def, the most involved thing would probably be for the display init.
(I think it uses the standard Pico [2] W pins for wifi... if not: https://github.com/adafruit/circuitpython/issues/9777)
tulip sleet
Are there any precompiled target specific libraries used in Circuitpython alread...
ESP-IDF supplies some .a's (binary blobs) that get linked in as part of the Espressif-supplied part of the build.
ionic elk
Ah ok I see libespnow.a etc. It doesn't have a target specific path but I assume there must be some file filtering that happens beforehand, right?
nvm it must, i didn't see circuitpython is supporting the C series too now
crimson ferry
Has anyone seen the Espressif decode_backtrace.py just print a bunch of blank lines after entering the backtrace? Mostly I'm geting corrupted backtraces, but when they don't say corrupted, I just get blanks from decode.
manic glacierBOT
I can't find a schematic but this image:
from https://github.com/Xinyuan-LilyGO/T-keyboard/tree/main/frimware says to use dio (you have qio) and that's it's an 8MB flash chip.
tulip sleet
Has anyone seen the Espressif `decode_backtrace.py` just print a bunch of blank ...
it has worked for me in the past, but I haven't tried it recently.
crimson ferry
It’s worked for me in the past too
orchid basinBOT
manic glacierBOT
kind river
<@177230548236763136> if Pimoroni isn't interested, Pico 2 W firmware would prob...
My plan was just to copy the 2 W port and pull in the pin defs from the MicroPython port then port the ST7701 to CircuitPython
slender iron
@tulip sleet I'm back. Want to sync today?
tulip sleet
<@329766224093249548> I'm back. Want to sync today?
sure, could do it now or later
slender iron
now works for me
worthy briar
Is there any reason to have the safe mode wait enabled on boards with no reset button? e.g. Pico and other RP2040 boards. I'm currently building with CIRCUITPY_SKIP_SAFE_MODE_WAIT=1 in order to reduce boot time, but I'm wondering if there's a reason this isn't the default on these boards?
stuck elbow
Is there any reason to have the safe mode wait enabled on boards with no reset b...
I added that option because I needed it for my things, and I don't think anybody else actually used it?
jaunty juniper
Is there any reason to have the safe mode wait enabled on boards with no reset b...
the pico has a reset pin, you can use it for safe mode, stick a button to it, most if not all other RP2040 boards have a reset button. Safe mode is an important feature to not lock yourself out or have to resort to erasing the flash
worthy briar
I added that option because I needed it for my things, and I don't think anybody...
Ah, OK, I get it now. I hadn't realised the difference between hitting the boot button, and resetting during the safe mode window
jaunty juniper
oh you mean boards without a boot button then
worthy briar
yeah, but I now understand that there's two different ways into safe mode (reset or boot button)
jaunty juniper
there were discussion of alternatives and ways to disable the safe mode wait from user code, although it couldn't be code of course
I was thinking of having some persistent bytes in a dedicated location that could be written from a UF2 bootloader to configure the build without having to make an alternative build for things like that
(for things that couldn't be a config file on the drive, that is)
(and of course safe mode is something that should not be overrideable by the drive, or that kind of defeats the purpose)
(like you could lock yourself out of the board if it's a settings.toml)
worthy briar
From my point of view, doing a custom build wasn't a big issue, but it took me a while to track down the cause of the delay, and then figure out how to disable the wait. I eventually tracked down the github issue where CIRCUITPY_SKIP_SAFE_MODE was discussed and added.
I've used MP in the past, so I was confused by why it was booting slower.
It probably doesn't help that I'm on a custom board with no status LED setup - a flashing LED might have given me a clue.
jaunty juniper
fun fact the first releases for the pico did not have safe mode wait, it felt so fast, I was using a boot.py that would do the wait and read a button to reset into safe mode
(especially when trying things where you disable the drive or looking for bugs that cause boot loops)
worthy briar
My current project is boot-time sensitive because it needs to set a GPIO pin high in order to keep the power on after the user lets go of the power button.
stuck elbow
My current project is boot-time sensitive because it needs to set a GPIO pin hig...
I thought you were putting that in board.c?
worthy briar
Yes, I am now
stuck elbow
oh, that happens after the safe mode wait
slender iron
we could add a pre-wait board_init
worthy briar
actually, I put it in board_reset_pin_number which I think does get called before the wait, because it's called from port_init()
but I also want the safe mode wait disabled to make it more responsive
worthy briar
Never - it doesn't have a reset button.
slender iron
what needs to be responsive then?
worthy briar
The device turning on. i.e. minimal delay from pressing the button to actually doing stuff.
slender iron
pressing what button?
slender iron
@tulip sleet want me to do the idf update?
tulip sleet
<@329766224093249548> want me to do the idf update?
sure! I didn't really get very far with it.
slender iron
k, would be a good short task (hopefully) for today
worthy briar
pressing what button?
The power on button. It's a battery-powered kids game that turns off its own voltage regulator when not in use, so it does a cold boot whenever you use it. Initially I was seeing 2-3s from pressing the button to seeing any sign of life, and you had to hold the power button for most of that, which felt very sluggish. Removing the safe-mode wait, pre-compiling to mpy, and a few other optimisations have got it to comfortably under a second, which is good enough, and it now only needs a slightly long press of the power button (probably about 0.3s) to turn it on, which is much better.
worthy briar
I've just raised that PR with an addition to the README which would have helped me.
slender iron
Ok! Makes sense
manic glacierBOT
orchid basinBOT
manic glacierBOT
dawn rampart
Hello team - I'm wanting to add a new board to CP - the Waveshare ESP32-C6 with 1.47 LCD. It is a ST7789 driver: https://www.waveshare.com/wiki/ESP32-C6-LCD-1.47. Mostly I can get this included, but the issue I'm seeing is that I'm not sure how to get the LCD setup. I looked at other examples in the espressif boards folder as well, but I likely need help getting the LCD part of the board.c file setup. Likely a combo of https://github.com/adafruit/circuitpython/blob/main/ports/espressif/boards/lilygo_ttgo_t8_s2_st7789/board.c and https://github.com/adafruit/circuitpython/tree/main/ports/espressif/boards/makergo_esp32c6_supermini. Is there anyone who can help me with the LCD board.c part of it? Maybe there is documentation outlining how to set it up? I'm assuming I can look here: https://github.com/adafruit/Adafruit_CircuitPython_ST7789/blob/main/adafruit_st7789.py for some guidance.
manic glacierBOT
wraith crow
@dawn rampart It looks to me like the information you need is in the Display_ST7789.cpp and Display_ST7789.h files included in https://files.waveshare.com/wiki/ESP32-C6-LCD-1.47/ESP32-C6-LCD-1.47-Demo.zip. You'll need to convert the LCD_WriteCommand/LCD_WriteData statements to the board.c display_init_sequence values but by comparing the gamma setting strings you can work out the basics
I'm really not all that familar with LCD setups, but I'm thinking the first few lines would convert something like this:
0x11, 0 | DELAY, 120,
// Vertical
0x36, 1, 0x70,
0x3A, 1, 0xB0,
0xB0, 2, 0x00, 0xE8
slender iron
@dawn rampart I'd suggest starting with another ST7789 board_init(). It is likely very close to what you need. You may neeed to tweak size and colstart but that's it
slender iron
@lone axle I think you can change your CP25 gist to .md and it'll word wrap.
lone axle
<@382939733107408897> I think you can change your CP25 gist to `.md` and it'll w...
Thank you, just changed it, and indeed it does.
manic glacierBOT
stuck elbow
Sooo, the ch32v307 has 64kB RAM and 256kB flash, that's a CircuitPython material, right?
slender iron
sure!
stuck elbow
even tinyusb is already ported
sharp hollow
I'm using an MPU-6050 to measure vibration. I'd like to contribute my code, but am wondering what the interface should look like. Measuring vibration requires very fast sampling, so I end up enabling the MPU-6050's FIFO and need relatively performant code to read out the FIFO. Should I just expose something to read the FIFO, or should I also include e.g. my calculation of acceleration magnitude, frequency spectrum, etc?
stuck elbow
I'm using an MPU-6050 to measure vibration. I'd like to contribute my code, but ...
that's still over i2c, is it?
sharp hollow
I'm using an MPU-6050 to measure vibration. I'd like to contribute my code, but ...
My concern with not packaging up everything is that using the FIFO is somewhat... unfriendly. If I hide all of it inside an easy interface (e.g. a hypothetical mpu.get_vibration_spectrum), that could help. But on the flip side, my ESP32 takes on the order of a second to do the processing, so this call would block for quite a while. Even reading the FIFO takes >>100ms.
sharp hollow
that's still over i2c, is it?
Yes
stuck elbow
I would think python is fast enough for handling i2c, even with bigger transactions
sharp hollow
_fifo_data = Struct(_MPU6050_FIFO_R_W, ">hhh")
res = []
for i in range(0, self.fifo_count, 6):
foo = self._fifo_data
res.append(foo)
That takes >>100ms when reading 128 samples. Maybe I' mdoing something wrong?
stuck elbow
yeah, you will get much better results if you create the list of the desired size, and fill it, instead of appending to it
you could also use an array, or a pylab array even
slender iron
I imagine the i2c read is the slowest part
can you read from the fifo continuously?
instead of reading one tuple at a time, read as many as you can
stuck elbow
possibly an io stream interface could work?
slender iron
foo = self._fifo_data does an i2c transaction each time
do a longer read that pulls all of the values at once
sharp hollow
By declaring my Struct as wider than the 3 halfs of a single entry? Or how?
slender iron
don't use struct. read the register directly
stuck elbow
aren't esp32 i2c buffers limited to like 12 bytes per transaction or something like that?
slender iron
not that I know of
stuck elbow
oh, that was a limitation of the arduino library, sorry
so what does the i2c readfrom do when it gets a NAK before it reads all the bytes it was supposed to?
sharp hollow
Sorry for the dumb question, but what would the code look like to just read the same register N times, then?
slender iron
raises an exception I think
stuck elbow
but does it still fill the buffer up until the NAK? and is there a way to know how many bytes were read?
stuck elbow
Sorry for the dumb question, but what would the code look like to just read the ...
if it's a fifo register, you can just keep reading it, no?
sharp hollow
if it's a fifo register, you can just keep reading it, no?
just call i2c.readfrom_into with my address and large byte array to read into?
stuck elbow
just call i2c.readfrom_into with my address and large byte array to read into?
let me look at the datasheet to confirm my assumptions
sharp hollow
One thing that's been bugging me is how does the read know whether, when fetching two bytes, to read register N and N+1, or read register N twice.
stuck elbow
looks like they call it "Burst Read Sequence" in the datasheet, and yes, you can keep reading until you get a NACK, at which point you know the buffer is empty
slender iron
You read FIFO_COUNT and then read that many bytes
sharp hollow
You read FIFO_COUNT and then read that many bytes
with one readfrom_into?
slender iron
yup
sharp hollow
I'll try that, tahnks
slender iron
you can use an array.array instead of struct too
then you'll read 16 bit numbers directly
stuck elbow
One thing that's been bugging me is how does the read know whether, when fetchin...
the fifo data is just one register
sharp hollow
the fifo data is just one register
sure, but if you look at COUNT it's broken into high/low, and it seemed like I get the correct count by just declaring this:
fifo_count = ROUnaryStruct(_MPU6050_FIFO_COUNT, ">h")
so why did that know to move to register+1 for the second byte, intsead of building the half from two reads of the same byte register?
slender iron
it may be wrong
stuck elbow
I would expect it to be a special case, but I don't see any mention of that in the docs
slender iron
it looks like writes autoincrement but reads don't
stuck elbow
To write multiple bytes after the last
ACK signal, the master can continue outputting data rather than transmitting a stop signal. In this case, the
MPU-60X0 automatically increments the register address and loads the data to the appropriate register. The
following figures show single and two-byte write sequences.
sharp hollow
strange that my count is correct then, but it moves completely reasonably. A problem for another day. For now, this. I think the big read was quite a bit faster; it looks like ~70ns now to get 128 samples
stuck elbow
yeah, can't see anything about incrementing the register in reads
sharp hollow
(128 samples = 768 bytes)
stuck elbow
strange that my count is correct then, but it moves completely reasonably. A pr...
perhaps it never exceeds the lower byte?
sharp hollow
nope, it goes all the way up to the 1023 or 1024 it's supposed to
that was my initial thought yesterday too.
stuck elbow
did you look at the data? it will just repeat the same value when you read more than the count
sharp hollow
I know, but I' msampling at 200Hz and if I sleep 0.1 and check count, it grows 20*6 at a time.
stuck elbow
my guess it that they do the increment for some registers but not for all, that's what I would do, but I don't see it documented anywhere
sharp hollow
they even explicitly warn that they don't adjust the count until you've read both bytes
stuck elbow
incidentally, fifo data is register 74, and there are only 75 registers, so it's almost the last one
it would make sense that it doesn't increment if it was the last one
but then they had to add the "who am I" register and break that
in any case, try just doing readfrom with the count, and see if that works, that should speed things up a lot
sharp hollow
it did, by about 2x. nwo i'm at ~211ms to read the samples and also convert from raw bytes to accelerations. I think I can speed this up a bunch by eliminating multiple divides.
slender iron
what board are you on?
sharp hollow
Ok, down to 85ms to get 128 samples converted to acceleration. I broke something
ESP32-WROOM-32 on some aliexpress board ("30pin ch340 type-c", only the best product naming)
slender iron
floating point division should be fast on that board I think
sharp hollow
yeah, it didn't help as much as I'd hoped.
sharp hollow
Ok, so I'm back to my original question. Because using the FIFO requires picking which features go into the FIFO, ensuring you're reading at a sane time, sane number of samples, etc... what sort of interface makes sense here? For the moment, my changes to the adafruit_mpu6050 library just expose a way to dump the whole FIFO, blindly assuming it's configured to only store accelerations (and then my application code computes magnitudes and runs an FFT to report the frequency with the highest vibration + power). But this feels fairly specialized... likely to the point where it might even make more sense to simply expose a vibration power spectrum as the output, then have the library enable and configure the FIFO itself, then disable the FIFO after reporting vibration.
slender iron
I think you'll want to change the API to support your vibration math. I'm not sure we want it in the driver
maybe a fifo property that return an array of numbers
it is up to the caller to know what is in the fifo
sharp hollow
Then I think also expose the scaling too, so the user can call it rather than depend on knowing the details of the implementation? e.g. expose a "bytes_to_accel", "bytes_to_gyro", etc?
slender iron
sure
manic glacierBOT
keen timber
Hello all! does anyone know if CircuitPython supports MQTT over websockets? Having trouble finding information if it does or not? Thanks!
Also curious if there is a way to install a library onto an esp32-s3 chip using pip
wraith crow
Hello all! does anyone know if CircuitPython supports MQTT over websockets? Havi...
For MQTT and websocktes, this might have what you're looking for: https://adafruit-playground.com/u/justmobilize/pages/adafruit-connection-manager
and I don't think there's a way to install using pip but circup might be close (https://learn.adafruit.com/keep-your-circuitpython-libraries-on-devices-up-to-date-with-circup/overview) I don't know if the learn guide talks about it, but you can actually use circup and web workflow to install libraries over WiFi
keen timber
@wraith crow
I did look into the MQTT lib. I just wasn't sure if it suppot MQTT over websocket since the default MQTT port may be blocked in some cases 😦
wraith crow
I see, I've only used MQTT as shown in the examples. Have you tried asking in the help-with-circuitpython channel?
keen timber
I'd like to do a little more experiement first, but definitely will if I get stuck. Thank you!
manic glacierBOT
manic glacierBOT
Instead of the Nordic dongle, I tried it out on 2x Feather STM32F405 Express (pins D9 and D10 for data) as well, with less success, as none of the queries was answered successfully.
Additionally, on each second save in MU editor, a RunTime error occured - see the screenshot
manic glacierBOT
stark aurora
Someone working on efr32x24 or efr32x27 and have some functional manage SOC peripherals via Bluetooth (android/IOS) App?
crimson ferry
bitbangio doesn't have UART... are there technical issues, or just not a priority?
tulip sleet
`bitbangio` doesn't have UART... are there technical issues, or just not a prior...
it's a lot harder to do the timing
there are PIO versions of UART, if you are on an RP2 board
tulip sleet
@lone axle HI - it looks like you're on. If you want, could you review https://github.com/adafruit/Adafruit_CircuitPython_MiniMQTT/pull/235 ? Otherwise I will merge without a review, because #231 and related caused a regression with ESP32SPI socket code, and I want to get the change in kinda quick. Thanks.
lone axle
<@382939733107408897> HI - it looks like you're on. If you want, could you revie...
Yep, will do.
ionic elk
yeah I forgot the top level makefile isn't actually used or called by the port level ones, it's just a separate thing altogether
might submit a couple of minor corrections to https://learn.adafruit.com/building-circuitpython/espressif-build
you get messed up a bit if you do things in the order it suggests
ionic elk
right
and I think you actually need to run install.sh, then run export.sh, then install the requirements, then run install.sh again to have it revert some pip libraries that have a version that's too high, then run export.sh again
jaunty juniper
oh interesting
slender iron
<@&356864093652516868> Just a reminder that our first community meeting of the year is in an hour and half. I'm excited to chat with you all. Notes doc is here: https://docs.google.com/document/d/1f2Em2jVmoOt-aoOyxfzj5u1yUQXcxz6sGIjCWcWIAHY/edit?tab=t.0
onyx hinge
a lot of folks choosing to spend their holidays with us, for which 😍
5659 PRs since the dawn of time in the core repo, according to my math
onyx hinge
text data bss dec hex filename
105117 6272 0 111389 1b31d Modules/_sre/sre.o
```just as a note, there is a LOT of code in cpython's implementation of regular expressions too. Approximately 100kB when compiled for desktop computers (amd64)
you could see if there's a pure python implementation that does the things you need, too.
errant grail
Thanks all and to Scott for hosting. Happy New Year!
onyx hinge
have a fine week folks!
lone axle
Thanks for hosting Scott. Have a great week and happy near year to everyone!
onyx hinge
ugh so in Matter, passcodes are 8-digit decimal numbers (26 bits of entropy) but for "security" they use a technique called "pbkdf2", which lets you in some vague sense increase security by doing a lot of busy-work (the iterations Dan was talking about). But pbkdf2 assumes "the computer you might want to generate or check a password with" is not much slower than anybody else's (you have to use an agreed on number of repetitions, or iterations), a flawed assumption when dealing with embedded devices like these. [Though, other embedded devices implement the spec just fine so maybe there's some obnoxious limitation or anti-optimization of the implementation in CircuitPython]
tulip sleet
ugh so in Matter, passcodes are 8-digit decimal numbers (26 bits of entropy) but...
yes, this is the pbkdf2 logic. I haven't even checked yet whether this is per session or not, but, without checking, I was thinking this was in the initial key generation. The device already knows its key(s), so I think it does not need to do pbkdf2 fresh. But I haven't looked at that in detail.
did you find a writeup of this?
slender iron
I don't consider circuitmatter secure anyway
lilac folio
Hi folks; I'm bringing up a new ESP32 port/board and need to adjust some pins as early as possible. Is there a way I can get this done in the bootloader or something?
tulip sleet
Hi folks; I'm bringing up a new ESP32 port/board and need to adjust some pins as...
you can do it in board.c. There are other boards that do that. Is that early enough?
lilac folio
When this board powers up from battery it needs to set up an input to stay powered up, but if the button is held for 3s it powers down again, so it needs to happen within the first 3s after boot - but with enough time for a human to realise. Right now in board_init() that's about 2.9s after boot so the window you can let go of the button is very short!
slender iron
Here is the notes document for next Monday’s CircuitPython Weekly Meeting. It is at the normal time of 11am Pacific / 2pm US Eastern here on Discord. Add your hug reports and status updates to the document before the meeting. If you are unable to attend but would still like to contribute, feel free to add your notes and we’ll read them off during the meeting. Hope to see you there! (Markdown markup is now implicit in Google Docs formatting.) <@&356864093652516868> https://docs.google.com/document/d/1zy3Dm9Hb6pBi3phJY4S6hOdAvgtOnlOcx-dCMBeFvbo/edit?usp=sharing
tulip sleet
When this board powers up from battery it needs to set up an input to stay power...
You could add an early_board_init() functionality to main.c, near the call to port_init(), define it in your board.c ,and provide a WEAK function that does nothing. We'd probably take that as a PR, no problem, because it doesn't affect anything else.
slender iron
When this board powers up from battery it needs to set up an input to stay power...
Another person was doing custom pin reset that happened earlier.
stuck elbow
I think they just disabled the safe mode button wait
tulip sleet
yes, it depends on whether they're willing to disable that.
lilac folio
I'm willing to give that a try, how do I go about it?
I mean disabling the safe mode wait; I think I can figure out an early board init in main()
stuck elbow
I'm willing to give that a try, how do I go about it?
define CIRCUITPY_SKIP_SAFE_MODE_WAIT to 1 in your mpconfigboard.h
lilac folio
Thanks, giving it a go
It looks like even with that, a board should be able to request safe mode using board_requests_safe_mode() right? I think that's missing in the espressif port, so I might add it.
lilac folio
Thanks, that seems to do the trick! No wait makes the window much longer, and holding the button down makes it enter safe mode. Winner.
If anyone is up for critiquing my code, I have a PR: https://github.com/adafruit/circuitpython/pull/9939
onyx hinge
well that was unexpected but I think it's the answer 😕 without a 32kHz crystal you're a bit doomed in the RTC department.
jaunty juniper
can rtc.calibration help then ?
crimson ferry
I had trouble with drift on some SAMD51 boards a few years ago, and the ppm-scale rtc calibration was no match for losing minutes per hour
but if the drift is consistent, I suppose you could fudge something in code
modern sundial
but if the drift is consistent, I suppose you could fudge something in code
Unlikely to help for more than a couple minutes. Drift changes with temperature and voltage, but possibly light and humidity too.
I think it’s independent of barometric pressure
crimson ferry
Unlikely to help for more than a couple minutes. Drift changes with temperature ...
true; it all depends on the accuracy being sought
tulip sleet
@slender iron CircuitMatter uses the avahi --subtype arg to advertise additional services on mdns (from start_commissioning() in __init__.py). But the the native mdns doesn't provide this capability. I'm wondering if this is a problem.
slender iron
<@252717193496756235> CircuitMatter uses the avahi `--subtype` arg to advertise ...
I was thinking we'd need to add it
tulip sleet
I was thinking we'd need to add it
I've commented it out for now, to keep going, but then I guess the commissioning may fail. I'll look at it when i get there.
slender iron
ya, you may not be able to discover the device
tulip sleet
@slender iron I have finally gotten this far, and entered the manual code in Apple Home, but not proceeding, so I'll need to see what's up. I think indeed it may be mdns capabilities I need to add. However, this is a lot better than the Python errors, which had been the chief stumbling blocks:
>>> import cpblink
removing /sd/matter-device-state.json if present
Listening on UDP port 5541
QR code not available
Manual code: 0454-440-2633
nice! yup. you can use wireshark to sniff the mdns stuff
tulip sleet
✔️ will do -- thanks
onyx hinge
@tulip sleet you previously reviewed that rp2350b PIO PR. Would you prefer that I avoid rebasing it at this point?
tulip sleet
<@329766224093249548> you previously reviewed that rp2350b PIO PR. Would you pre...
rebase is fine! my review was mostly testing, less detailed code review
onyx hinge
OK. I finally found an obvious error and i2s is working now. but I'll test a bit more before updating the branch on github.
well it was working, hmm
must be two problems or else my previous test was wrong
it's not working on the first run after reset, so there's still a second problem. continuing to dig.
onyx hinge
```/build/gdb-kUlB6r/gdb-13.1/gdb/i387-tdep.c:957: internal-error: i387_supply_xsave: Assertion `tdep->st0_regnum >= I386_ST0_REGNUM' failed.
i had one of them i387's, I think
onyx hinge
https://stackoverflow.com/questions/54439919/debugging-a-simple-arm-64-bit-executable-causes-internal-error-in-gdb I'm not the only one who's hit it though
inner vapor
Anyone working on a CircuitPython port for the Pimoroni TinyFX W? (https://shop.pimoroni.com/products/tiny-fx-w)
tulip sleet
Anyone working on a CircuitPython port for the Pimoroni TinyFX W? (https://shop....
Not that I know of, feel free!
inner vapor
Not that I know of, feel free!
Seems like it should be easy— but I guess I should have asked before vacation was already over and I got over my head at my day job 😉 (Really like CircuitPython more than MicroPython)
Am I correct in assuming I'd basically double check the schematic, add board under ports and do what was done between Pimoroni Pico Plus 2 and Pico Plus 2 W?
tulip sleet
Seems like it should be easy— but I guess I should have asked before vacation wa...
it's an RP2040 board, so copy and modify some existing Pimoroni RP2040. Figure out what the USB PID is. Do you have one of these boards?
inner vapor
it's an RP2040 board, so copy and modify some existing Pimoroni RP2040. Figure o...
I do, got it for Christmas. Also have a TinyFX (no W) as well. Yeah, I zoned on the 2040 vs 2350.
proud ember
So, this isn’t something I need help with, but more of a feature suggestion for CircuitPython devs.
Has anyone considered making a plug-in for Home Assistant that would take a YAML file and generate CP code and push the code to an end device as an OTA update? (Like the way ESPhome works but skipping the compile step because it’s Python).
Just thought it might be an interesting piece of low-hanging fruit given the availability of functions like asyncio, existing OTA workflows, and the well structured nature of board and library support in CP.
I can ask somewhere else if this is the wrong place. Thanks for all that you do, CircuitPython is a delight!
jaunty juniper
I'm encoutering that with a repository I just updated to the latest version of cookiecutter - my first time using ruff - should I then remove all the offending rules ? Should the cookiecutter be updated to avoid that in the future ?
those:
warning: Selection `PLW2101` has no effect because preview is not enabled.
I have just left them alone so far, it seems that after running a few times they are supressed automatically. I'm not entirely sure what all of those different rules it's mentioning are, though I believe the list in that file was arrived at mostly by converting values from the old pylint config. I also am not entirely sure what it means by 'preview'.
jaunty juniper
ok, maybe they are not generated if there are no errors ?
man that thing likes adding empty lines between imports 🙄
Preview mode enables a collection of unstable features such as new lint rules and fixes, formatter style changes, interface updates, and more. Warnings about deprecated features may turn into errors when using preview mode.
ok so it's neceassry for those checks to be enabled I guess
which is done by adding this line in the lint section
preview = true
according to this https://docs.astral.sh/ruff/preview/
so basically all the rules listed are not actually active otherwise
jaunty juniper
ok my local ruff wants this:
import board
from rainbowio import colorwheel
import waveshare_LCD1602
but when running on github action it wants that:
import board
import waveshare_LCD1602
from rainbowio import colorwheel
it depends where the libraries are installed
it groups imports into standard library, system, and local, etc.
slender iron
So, this isn’t something I need help with, but more of a feature suggestion for ...
I thought about that for adafruit io. I haven't used home assistant much though
proud ember
I thought about that for adafruit io. I haven't used home assistant much though
Oooh! I love that idea! Even if you’re not planning to support Home Assistant directly, ESPhome has a really solid design for how it structures and parses the YAML files. Making the interface the same would be really cool because if someone were to add a CP integration in the future, you could copy sensor blocks from one framework to the other and it would still work!
lone axle
I am troubleshooting issues using the adafruit_ntp library and noticed that it has this default URL 0.adafruit.pool.ntp.org in it here: https://github.com/adafruit/Adafruit_CircuitPython_NTP/blob/a1acc411a2b4d5c23a565e9ee7842b052fc34d9b/adafruit_ntp.py#L45C24-L45C47
As a quick sanity check I loaded that in a browser and found that it seems to be returning the home page of a person named Hsing Lo. Is it known / expected for that default URL to load a persons page like that?
okay, so it's not always loading the same thing. I just loaded that URL again and got the rick roll video.
maybe that is just how it's set up to reply to HTTP requests?
In any case it seems like some server involved in handling NTP for 0.adafruit.pool.ntp.org might be having issues. I was able to change to a NIST ntp server and that one is working where-as I just keep getting timed out error with the default one.
stuck elbow
In any case it seems like some server involved in handling NTP for `0.adafruit.p...
looking at all the ip addresses that hostname resolves to, they are belong to different schools in Switzerland
lone axle
It must have been an intermittent issue. I switched back to the default after a few minutes and now it's working. I guess it grabs from a list of available NTP servers and they are allowed to return whatever they want on port 80 HTTP unrelated to the actual time service.
stuck elbow
one in Zug, one in Geneve and two in Zurich
sharp timber
trying to do this project
https://learn.adafruit.com/control-wiz-lights-with-circuitpython
when you get to the point of editing the settings.toml file — can you just open and edit it with MacOSX text editor?
jaunty juniper
yes
sharp timber
it’s doesn’t seem to be connecting
using a RP2040 Pico W, every few seconds it does two quick LED blinks
jaunty juniper
note that it's better to go in #help-with-circuitpython for that, more people for helping there
slender iron
Oooh! I love that idea! Even if you’re not planning to support Home Assistant ...
You could write up a #CircuitPython2025 post with more details. I don't know the HA or ESPHome side but Brent likely does.
onyx hinge
I wonder why delaying works. weird.
tulip sleet
I wonder why delaying works. weird.
and needed only for that chip
this is an "empirical" fix, but I'll eventually try to make an ESP-IDF example and submit an issue upstream
onyx hinge
nods
proud ember
You could write up a #CircuitPython2025 post with more details. I don't know the...
I’d love to! What and where is that? Is there a guide or guideline listed somewhere?
shut copper
I put up a PR for circup to align it with the adabot example naming changes I made. Please let me know if this is a good direction for the example command in circup. https://github.com/adafruit/circup/pull/239 @lone axle
stuck elbow
Was there a learn guide for porting CircuitPython to a new port, or am I imagining things now?
solar whale
Was there a learn guide for porting CircuitPython to a new port, or am I imagini...
https://learn.adafruit.com/how-to-add-a-new-board-to-circuitpython/overview is this what you are looking for for?
stuck elbow
<https://learn.adafruit.com/how-to-add-a-new-board-to-circuitpython/overview> is...
sadly no, I meant new directory in ports with a new architecture
but that thing I linked is good for a start for me
solar whale
Good luck!
lone axle
I put up a PR for circup to align it with the adabot example naming changes I ma...
Thank you! I've got it in my list to take alook at.
mortal kernel
I've implemented a new module that has no per-port specializations, so I've placed the __init__.[c,h] and ramlog.[c,h] in a folder ramlog in the shared-module folder. I've added a pattern to the filter in circuitpy_defns.mk like this:
ifeq ($(CIRCUITPY_RAM_LOG),1)
SRC_PATTERNS += ramlog/%
endif
and added the sources to SRC_SHARED_MODULE like this:
ramlog/__init__.c \
ramlog/ramlog.c \
This results in a
make: *** No rule to make target 'shared-bindings/ramlog/__init__.c', needed by 'build-pimoroni_pico_plus2w/genhdr/qstr.i.last'. Stop.
I'm misunderstanding something here. I don't think I should need a shared binding?
slender iron
Was there a learn guide for porting CircuitPython to a new port, or am I imagini...
I have one that I never released. What MCU are you looking at?
stuck elbow
I have one that I never released. What MCU are you looking at?
the ch32v208, but more generally the ch32v line
slender iron
any zephyr support?
slender iron
I've implemented a new module that has no per-port specializations, so I've plac...
You still need the binding for the Python -> C layer. shared-module is for the implementation
stuck elbow
any zephyr support?
I'm not familiar with zephyr, there is tinyusb support
looks like there is at least some work https://matthewtran.dev/2024/02/basic-ch32v-zephyr-support/
slender iron
Can you see the guide with https://learn.adafruit.com/implementers-guide-to-circuitpython-api/overview?preview_token=NXki5quOVZd7I7_ilEDxuw ?
stuck elbow
Can you see the guide with <https://learn.adafruit.com/implementers-guide-to-cir...
yes, I have an account
thank you!
tulip sleet
@slender iron @onyx hinge Now that the PIO and IDF 5.3.2 PRs are merged, I think that's a good place to do 9.2.2. I will merge the Translation PR that is currently running and then make a release.
slender iron
want to do an RC to give 5.3.2 a little soak time?
tulip sleet
I could, but I think very few people will try it. I tested wifi and BLE, which I think would be most likely subject to regression. I can do an rc until Monday if you think that would make sense.
we are not in a hurry, it's true
slender iron
meh, we can always do 9.2.3 if an issue comes up
tulip sleet
yah. I will be around to do any emergency fixes
manic glacierBOT
mortal kernel
You still need the binding for the Python -> C layer. shared-module is for the i...
Thanks, I think I've got it. Python bindings go in shared-bindings and the module's implementation (if any) goes in shared-module. Some implementation bits like protocol methods (ex., standard stream methods) go into shared-bindings, so there's a bit of discretion involved.
wraith crow
I think something may still be going on, but I'll wait until I have something more definitive and open an issue. Thanks!
onyx hinge
Thanks Dan!
tulip sleet
https://blog.adafruit.com/2025/01/09/circuitpython-9-2-2-released/
HIghlights of this release:
int.from_bytes()and.to_bytes()calls can omitbyteorderarg, as in CPython.DemuxKeyMatrix: add scan delay, addcolumns_to_anodesandtransposeargs.- Implement
audiomixer.MixerVoice.levelassynthio.BlockInput. displayionow supports 24-bit color depth.- Add
sideset_pindirstorp2pio.StateMachine. rp2pio: support on GPIO pins 32 and up, available on RP2350B.- Initial implementation of
audiobusio.PDMInon Espressif. - Update Espressif ESP-IDF to v5.3.2, which supports ESP32-P4 v1.0 chips.
- New
ports/analogport, initially supporting Analog Devices ADI MAX32690 microcontroller. - Over 20 new boards.
tulip sleet
has anyone used PyCharm with CircuitPython and Raspberry Pi Pico/Pico2 ?
There's some discussion here: https://learn.adafruit.com/welcome-to-circuitpython/pycharm-and-circuitpython. It was written a while ago.
tulip sleet
#help-with-circuitpython is better for this
bitter dew
I'll try and ask there, thank you
tulip sleet
Please don't cross-post. Let's move discussion to there.
bitter dew
of course....I haven't cross-posted...I've read that this channel was a better place to ask for development. That's wehy I've asked here first
tulip sleet
I just saw your "Hello all", sorry. This is for development of CircuitPython, rather than in.
bitter dew
ah I see
ionic elk
I'm hopefully close to wrapping up my microros PR. I haven't submitted a PR in a while so I wanted to refresh on some protocol stuff:
- are draft PRs still best for stuff that's not quite ready yet?
- is the best place to ask around for compilation problems here, in the issue, or on the PR itself?
tulip sleet
I'm hopefully close to wrapping up my microros PR. I haven't submitted a PR in a...
yah sure on drafts. If you do a PR, we can look at the build problems easily in the PR build or clone it to build locally
ionic elk
I also have some style questions, structural stuff. MicroROS is a big component (lots of includes), has onerous build requirements (ROS Colcon) so I've precompiled it, and will probably not be used by many people (best kept optional?). Should I ask those in more detail here, in the issue, or on the PR?
tulip sleet
what chip family is this for? You know that Scott is working on zephyr, right?
ionic elk
this is for espressif. It's an existing component implementing microros
I'm not up to date on Scott's zephyr work but I'm not sure if there would be incompatibility, since microros in esp-idf projects is implemented within freertos
it's just kind of a proof of concept anyway
tulip sleet
well, if CIRCUITPY_MICROROS = 0 for most or all boards, then would it affect the normal builds? ... I was just mentioning zephyr in case you were looking for some general rtos stuff that might be covered by it. I don't really know what's in ROS.
I think I made an incorrect assumption about it.
ionic elk
yeah "robot operating system" is kind of a misnomer it's basically middleware for robotics programs, and MicroROS is just a DDS system that pipes data up to an agent on a more sophisticated platform in the right format
I'm just hoping to let circuitpython boards talk directly to ros platforms over Wifi so people can send commands to drone swarms or get sensor data or stuff like that
tulip sleet
never mind that - misunderstanding
ionic elk
don't go down the ROS rabbithole hahaha
I'm partially doing this project to learn about it so I may bang my head on it up to some point and then realize somethings horribly incompatible, who knows
tulip sleet
you can turn on CI builds in your own fork if you want, if you want to wait before presenting it to us 🙂
ionic elk
I'd be happy to get it up for comments now. Writing some background on it in the PR
I haven't done any API design work yet, I want to make sure it just compiles OK and we've determined whether it will play nice with the WIFI
Are there any boards where we compile more than one version per board? I feel like this could either be optional, or an "alternative" build for platforms like the M5Cardputer, which would be a good host for MicroROS
tulip sleet
CPX has multiple builds, as an example
tulip sleet
circuit playground express
atmel-samd
there is a build with displayio and one without because we had to shoehorn it in
it's done as multiple boards
ionic elk
ah ok, figured that might be how to do it
stuck elbow
yeah "robot operating system" is kind of a misnomer it's basically middleware fo...
I wonder of it would make more sense to just reimplement the protocol in python?
manic glacierBOT
ionic elk
I wonder of it would make more sense to just reimplement the protocol in python?
possible but slower, and it would take a lot of disk space to store all the message formats ROS uses
orchid basinBOT
stuck elbow
possible but slower, and it would take a lot of disk space to store all the mess...
it doesn't necessarily have to be slower, and can't you just pick the one format that you need?
ionic elk
it doesn't necessarily have to be slower, and can't you just pick the one format...
my more ROS-knowledgeable buddy and I evaluated both options a while ago and decided MicroROS's implementation of xrce-DDS sounded easier than python-implemented DDS messaging but I'd have to go back and find my notes for specifics
stuck elbow
my more ROS-knowledgeable buddy and I evaluated both options a while ago and dec...
yeah, sorry, I had several attempts at ros but bounced off each time, so I only know generals
ionic elk
The compute half is extremely annoying haha. But it's super common in both robotics companies and research labs which is why I wanted to have a crack at this
stuck elbow
but how complicated an event propagation system can be? ;-)
jaunty juniper
anyone uses a Waveshare ESP32-S3 Pico ? The status LED on mine doesn't have the same color order as default so green status is red and red status is green (on UF2 too), I don't know if it's all of them or there are variants, so I don't want to change the board definition without checking
then again the only boards that use MICROPY_HW_NEOPIXEL_ORDER_GRB are waveshare boards, so maybe it's what they do
jaunty juniper
ah it's switched in their Micropython example so I'm gonna PR the change on both repos
by the way that board has a weird feature with a USB hub chip on board connecting on both the S3 USB pins and a serial chip connected to the default serial for programming (instead of having 2 USB ports like some boads do). But that USB hub chip is not recognized on my mac if connected to a hub instead of directly to the computer's port which raises so many questions on standards and drivers
kind of cool feature for developing when it works though I guess
crimson ferry
anyone uses a Waveshare ESP32-S3 Pico ? The status LED on mine doesn't have the ...
oddly, the Waveshare ESP32-S3-ETH uses RGB, even though a bunch of their other boards show as GRB in their wiki
jaunty juniper
yeah it seems it's only some of their boards 🤷
jaunty juniper
am I misunderstanding the naming logic or should MICROPY_HW_NEOPIXEL_ORDER_GRB be called RGB, since the order the bytes are sent by default is G then R then B, but R and G get swapped (to RGB) when the define is set ?
manic glacierBOT
remote light
If anyone is able to check my addition of the M5Stamp S3 to the CircuitPython site that would be very helpful.
I can't get the local preview working on Windows.
I've also added a code block to add the optional LCD accessory, but I'm not sure that belongs on the download page.
https://github.com/StellasFun/circuitpython-org/tree/M5Stack-StampS3
strange ravine
moving my issue here
so im trying to build circuitpython for my STM32H7 board, and ive succesfully built it and flashed it, however, it doesnt werk
below are the build files i used, which are based off the nucleo H743 and the OpenMV H7 modified to fit my board
the board itself does work so its likely an issue with the firmware
below are the files im using for the build, i made them based on the nucleo h7 and openmv h7
pins.c was made using parse_pins_csv.py provided in the stm/tools folder, using a pins.csv from a board with nearly identical pinout to my board
manic glacierBOT
strange ravine
Update, im getting somewhere
frail widget
Hey CircuitPy fam. I'm new to the community and don't know the protocol so feel free to point me in the right direction. Here's my PR against the typing library: https://github.com/adafruit/Adafruit_CircuitPython_Typing/pull/46
strange ravine
Update, im getting somewhere
Im going to try to build an earlier version of circuitpython
strange ravine
i tried 9.2.1 but nah it didnt work
strange ravine
Its weird coz i got the flashing light after disabling the LSE
And i know the LSE and RTC works i tested it
manic glacierBOT
jaunty juniper
hey, there has been many changes to adafruit_httpserver and adafruit_esp32spi since last time I used a board with that. I had a trick to make them work together by wrapping the ESP32SPI socket module with some compatibility layer, but it doesn't work anymore and it wasn't a quick fix.
by which I mean i can't seem to find how to fix it
tulip sleet
hey, there has been many changes to `adafruit_httpserver` and `adafruit_esp32spi...
I don't remember that this ever actually worked; if you had some trick, that would be great. There are issues: https://github.com/adafruit/Adafruit_CircuitPython_ESP32SPI/issues/203 and also multiple issues in the server library.
jaunty juniper
I don't remember that this ever actually worked; if you had some trick, that wou...
testing on a Matrix Portal, my socket wrapper works with CP 8 and its version of the esp32spi library (before adafruit_esp32spi_socket became adafruit_esp32spi_socketpool), as well as adafruit_httpserver 4.0.0.
It implements missing functions as calls to the underlying interface when possible, or spoof thems if that works.
Updating everything to latest I get errors with setsockopt and constants like SOL_SOCKET not existing and other problems.
this is code that works on CP8 with httpserver 4.0.0:
tulip sleet
testing on a Matrix Portal, my socket wrapper works with CP 8 and its version of...
I don't know why it broke after esp32spi socketpool, but I think that could be tracked down. I think the compatibilities provided UniversalSocket incorporated into ESP32SPI would be a goal.
jaunty juniper
I almost get there by adding some of the missing things, but I would probably have to dive deep to make it work, I'll make that a goal for this year I think
reviving that Matrix Portal was an interesting experience, I found a bug in PNG imageload (filed an issue) and another with some BDF fonts that used to load correctly but now have a large empty space on the top
or that last one is an issue with display_text, not sure ?
jaunty juniper
ah the issue appear with bitmap_label and not label
oh yeah it's issue 214 over there, setting background_tight=True works as some temporary fix
manic glacierBOT
arctic cliff
mimxrt10xx: NXP i.MX RT10xxx problem since CircuitPython 9.1.0
Up to CircuitPython version 9.0.5, a Metro M7 1011 delivered:
import microcontroller
dir(microcontroller.pin)
['class', 'GPIO_00', 'GPIO_01', 'GPIO_02', 'GPIO_03', 'GPIO_04', 'GPIO_05', 'GPIO_06', 'GPIO_07', 'GPIO_08', 'GPIO_09', 'GPIO_10', 'GPIO_11', 'GPIO_12', 'GPIO_13', 'GPIO_AD_00', 'GPIO_AD_01', 'GPIO_AD_02', 'GPIO_AD_03', 'GPIO_AD_04', 'GPIO_AD_05', 'GPIO_AD_06', 'GPIO_AD_07', 'GPIO_AD_08', 'GPIO_AD_09', 'GPIO_AD_10', 'GPIO_AD_11', 'GPIO_AD_12', 'GPIO_AD_13', 'GPIO_AD_14', 'GPIO_SD_00', 'GPIO_SD_01', 'GPIO_SD_02', 'GPIO_SD_03', 'GPIO_SD_04', 'GPIO_SD_05', 'GPIO_SD_06', 'GPIO_SD_07', 'GPIO_SD_08', 'GPIO_SD_09', 'GPIO_SD_10', 'GPIO_SD_11', 'GPIO_SD_12', 'GPIO_SD_13', 'USB_OTG1_DN', 'USB_OTG1_DP', 'dict']
this output.
Since version 9.1.0, the relevant information is lost:
import microcontroller
dir(microcontroller.pin)
['class', 'dict'].
Ditto for the Teensy 4.1:
import microcontroller
dir(microcontroller.pin)
['class', 'GPIO_AD_B0_00', 'GPIO_AD_B0_01', 'GPIO_AD_B0_02', 'GPIO_AD_B0_03', .........., 'USB_OTG1_DN', 'USB_OTG1_DP', 'USB_OTG2_DN', 'USB_OTG2_DP', 'dict'].
CircuitPython version 9.1.0 only returns: ['class', 'dict']. This also applies to the newer versions up to and including version 9.2.2.
Any hints?
ionic elk
Any ideas what this is from?
build-m5stack_cardputer/genhdr/qstrdefs.generated.h: In function 'MP_COMPRESSED_ROM_TEXT':
../../supervisor/shared/translate/translate_impl.h:39:49: error: implicit declaration of function 'strcmp' [-Werror=implicit-function-declaration]
39 | #define TRANSLATION(english_id, number) if (strcmp(original, english_id) == 0) { return (mp_rom_error_text_t)&translation##number; } else
| ^~~~~~
I don't really know how these translation generated files work, is it possible this is caused by a header file conflict or something?
tulip sleet
mimxrt10xx: NXP i.MX RT10xxx problem since CircuitPython 9.1.0
Up to CircuitPyth...
Could you open an issue about this? https://github.com/adafruit/circuitpython/issues/choose Thanks.
tulip sleet
Any ideas what this is from?
```
build-m5stack_cardputer/genhdr/qstrdefs.generat...
did you do a make clean first?
ionic elk
make clean is actually not working for me?
- Verbosity options: any combination of "steps commands rules", as `make V=...` or env var BUILD_VERBOSE
../../py/mkrules.mk:172: target `build-/peripherals' given more than once in the same rule.
rm -rf build-
I get this and then it doesn't actually do rm -rf build-
do I need to run it from root
(Worth noting that I did do it manually)
jaunty juniper
shouldn' it be rm -rf build-* ?
ionic elk
seems like a very reasonable hypothesis
this is the mkrules bit
clean:
$(RM) -rf $(BUILD) $(CLEAN_EXTRA)
.PHONY: clean
build-/peripherals kind of looks like a variable extrapolation error, which happened when using the mac's internal version of make, which was an issue with tinyuf2, but gmake still doesn't clean anything, just without the complaint
❯ gmake clean
- Verbosity options: any combination of "steps commands rules", as `make V=...` or env var BUILD_VERBOSE
rm -rf build-
did you specify a BOARD?
jaunty juniper
yeah I think it only supports cleaning a board, there's no clean all
or is there ?
stuck elbow
not that I know
stuck elbow
I mean I do "make clean all" but that's just doing "make clean" and then "make" but in one step
ionic elk
Maybe I can do a cheeky PR for a warning message there
stuck elbow
there should actually be an error when board is not specified
other makefiles do that
jaunty juniper
yeah, and you could implement clean-all: rm -rf build-* 😉
ionic elk
if someone else can validate it's actually missing and this isn't just me messing something up I'm happy to add that in
ok I can 100% verify my string issue happens immediately after an rm -rf build-*, validated that directory is gone, with no intermediate steps. Full error:
In file included from ../../supervisor/shared/translate/translate.h:26,
from ../../py/misc.h:332,
from ../../py/mpstate.h:33,
from ../../py/mpstate.c:27:
build-m5stack_cardputer/genhdr/qstrdefs.generated.h: In function 'MP_COMPRESSED_ROM_TEXT':
../../supervisor/shared/translate/translate_impl.h:39:49: error: implicit declaration of function 'strcmp' [-Werror=implicit-function-declaration]
39 | #define TRANSLATION(english_id, number) if (strcmp(original, english_id) == 0) { return (mp_rom_error_text_t)&translation##number; } else
| ^~~~~~
build-m5stack_cardputer/genhdr/qstrdefs.generated.h:1938:1: note: in expansion of macro 'TRANSLATION'
1938 | TRANSLATION(" File \"%q\", line %d", 0)
| ^~~~~~~~~~~
In file included from ../../supervisor/shared/translate/translate_impl.h:19:
build-m5stack_cardputer/genhdr/qstrdefs.generated.h:1:1: note: include '<string.h>' or provide a declaration of 'strcmp'
+++ |+#include <string.h>
1 | // This file was automatically generated by makeqstrdata.py
cc1: all warnings being treated as errors
-e See https://learn.adafruit.com/building-circuitpython; Adafruit Discord #circuitpython-dev
make: *** [build-m5stack_cardputer/py/mpstate.o] Error 1
supervisor/shared/translate/translate.h explicitly includes <string.h> at line 11 in my copy, before including "translate_impl.h" on line 26. So, I dunno why you're seeing this.
is it possible you've created/added a file named string.h that is being erroneously included in lieu of the system header? otherwise, I don't have any top-of-head guesses.
ionic elk
yes it's possible, that's what I was hoping to find out, is whether that was the likely cause
I'll see if I can dig out that file and exclude it somehow
onyx hinge
If I create the file ports/espressif/string.h with the content this ruins everything it does in fact ruin everything. This is because the current directory is listed as -I. in the Makefile INC variable, meaning that it is consulted for <headers> and "headers". -isystem is the same. -iquote only adds it for "header" searches, not all searches. So you could try changing the problematic inclusion path to use -iquote instead of -I or -isystem
https://gcc.gnu.org/onlinedocs/gcc/Directory-Options.html there's also -idirafter which can add the directory to the END of the include path instead of the start, according to the docs. I'm not familiar with it.
ionic elk
ok, I'll give that a shot!
and that's a good start for me to do my own research and learn more about these options
onyx hinge
you can also generate the preprocessed source for any file so you can inspect it: $ make BOARD=adafruit_metro_esp32s2 build-adafruit_metro_esp32s2/py/mpstate.pp creates build-adafruit_metro_esp32s2/py/mpstate.pp with about 48k lines in it
ionic elk
I'm going to cross my fingers here and hope for the best but if this problem just moves to other headers causing similar issues I may have to re-evaluate my strategy
MicroROS comes with a huge collection of headers
yeah sadly it's just moving on to new errors. I'll have to research how to try and cordone off header conflicts in a bulk way
../../supervisor/shared/web_workflow/web_workflow.c:1163:13: error: implicit declaration of function 'strcasecmp'
https://github.com/micro-ROS/micro_ros_arduino/blob/4d6ed1f902eaf3eb4eefe660ecac32639a963c69/src/rcl/publisher.h#L84 totally in guessing territory here, but if it's part of std_msgs I think the directory that should be listed in the compiler flags is the one containing std_msgs because the inclusion is written as #include <std_msgs/msg/string.h> here.
ionic elk
find shows these inclusions of string.h
./microros-ext/include/example_interfaces/example_interfaces/msg/string.h
./microros-ext/include/std_msgs/std_msgs/msg/string.h
./microros-ext/include/rosidl_runtime_c/rosidl_runtime_c/string.h
./esp-idf/components/linux/linux_include/string.h
i tried replacing those paths from isystem to iquote which fixed the first issue but resulted in the second. But I'm just cargo culting rn and have to learn up on what iquote actually does
onyx hinge
yeah that one is included inside one of their source files as #include "rosidl_runtime_c/string.h" so the inclusion directive (whether -iquote or -I or whatever) should be something like -Imicroros-ext/include/rosidl_runtime_c (or -iquote), not -Imicroros-ext/include/rosidl_runtime_c/rosidl_runtime_c
ionic elk
ohh i see what you mean
Ok yes that could be causing it I was pretty crude about my include paths
onyx hinge
similarly ./microros-ext/include/std_msgs/std_msgs/msg/string.h maybe should be -Iext/include/std_msgs in the compiler flags + #include <std_msgs/msg/string.h> in the source file that needs it
they show that one <diamond> style so -iquote won't cut it
ionic elk
I'm pruning down all my includes and I'll see if that still flies, hopefully the style is consistent
ionic elk
yeah sorry for being kind of a chucklehead the OG project uses CMAKE and I don't know it very well
I don't think CMAKE has explicit includes lists at all right? So I just made it with find...
Probably a way to get it to spit out what I need for a makefile
onyx hinge
beats me, I don't know cmake well either
ionic elk
Random style question while I'm compiling - since the ROS includes directory is so bloated, should I move it to my own repo and link it in as a submodule? I don't want to attach the original microros component because building it requires ros-specific build tools like colcon and ament and I'd rather attach a prebuilt library per-target so the CI doesn't have to deal with that
onyx hinge
That's not a question I can answer. my gut is to use source code via submodules when it's feasible, but this is more of a tannewt judgement call I feel like
ionic elk
yeah that's why im iffy on it cuz this isn't actually source code, just a single .a file and then 400,000 lines of includes
IT BUILT
RAHHH
onyx hinge
fwiw https://github.com/adafruit/circuitpython/commit/5db6db01287079b0314a5356cd9b5c40886e4d23 incorporated something already properly set up as an esp-idf "component" into our build.
ionic elk
it's not the component part that's the issue it's the extra python packages it installs
onyx hinge
[heading into a video call, ttyl]
ionic elk
I'm not even sure that colcon will work at all if not on a specific ubuntu version
anyway thanks for your help!! you got me through it
onyx hinge
another question I'd raise is whether the "binary .a" approach ties us to a specific esp-idf version, i.e., the one that the .a file was built for. I don't know whether this is typically the case or not, but we would probably think twice before taking things that make it harder to update a core dependency like that.
ionic elk
Honestly I'd be fine with a case where CPY_MICROROS simply gets turned off if it's ever holding anything back, restricting it to only specific versions of Circuitpython
Given that my plan thus far is to only include it as an optional build-it-yourself flag and maybe an alternative build for a couple of feathers or the M5Cardputer
If it actually is cool/gets used the discussion could pick back up then
jaunty juniper
I see supervisor.runtime.serial_connected being always True (both on a samd51 and an ESP32S3 board) regardless of whether I'm connected to the board (with tio). Meanwhile usb_cdc.console.connected does correctly report when I'm connected or not.
Is that expected behavior for supervisor.runtime.serial_connected ?
strange ravine
Ah
So it hasnt worked in a while apparently
I should probably make an issue on the H7 port no longer working
austere acorn
I should probably make an issue on the H7 port no longer working
There already is one https://github.com/adafruit/circuitpython/issues/5470
I've been working with stm32h7 ports, and might have a fix for it, i'll post it here in a few days
strange ravine
I've been working with stm32h7 ports, and might have a fix for it, i'll post it ...
Omg thanks
austere acorn
Omg thanks
It took a while to check my fix, since there are only 2 boards for the H7, and one of them isn't produced anymore, but I got my hands on an openmv camera now. I can ping and let you know if you want.
strange ravine
It took a while to check my fix, since there are only 2 boards for the H7, and o...
Oki
Feel free to ping me at any time, ill try out your fixes on my own H7 board. It seems very few boards still use the H7 like the arduinos portenta and giga, i guess its that way since its a high end and expensive MCU
jaunty juniper
oh it seems that supervisor.runtime.serial_connected is True whenever there's a display, whether the REPL is on the screen or not
tulip sleet
Random style question while I'm compiling - since the ROS includes directory is ...
Are those python tools or system-installed tools. If it's just pip3 installs, I think it's OK to install them. There could be a separate requirements-something.txt for them if you are worried. As Jeff mentioned, having a prebuilt .a is problematic in several respects.
Where do the ROS includes come from? Did you copy them from some existing repo, or are they in esp-idf as a submodule? It's fine to have a submodule.
manic glacierBOT
I bought this from Adafruit on November 10, 2023 and attached is the output from ESPTool before flashing.
lilac scaffold
It took a while to check my fix, since there are only 2 boards for the H7, and o...
Please please tag me if you figure it out
tiny peak
tannewt, you asked for suggestions on well supported Zephyr boards... I've been having pretty good luck trying out various samples on a QT Py ESP32-S3 8flash/nopsram. For example, this works for BLE NUS UART:
west build -p always -b adafruit_qt_py_esp32s3/esp32s3/procpu --sysbuild samples/bluetooth/peripheral_nus
west flash
west espressif monitor
I can talk to it with the BluefruitConnect app from iPhone. Only catch is that the app needs to be ready and scanning before you run west espressif monitor (which resets the board, triggering the BLE advertisement at the start of the sample's main())
source for the sample, showing how they use the Zephyr bluetooth APIs is at https://github.com/zephyrproject-rtos/zephyr/blob/main/samples/bluetooth/peripheral_nus/src/main.c
Zephyr API docs for NUS are at https://docs.zephyrproject.org/apidoc/latest/nus_8h.html
there's also this "nus-console" snippet thing, whose readme claims it can be used to redirect Zephyr's console output to a BLE NUS UART:
https://github.com/zephyrproject-rtos/zephyr/tree/main/snippets/nus-console
arctic cliff
Could you open an issue about this? <https://github.com/adafruit/circuitpython/i...
Done! THX for the link.
strange ravine
I hope that by fixing the H7 port we can add support for more boards like the one by weact studios and then add support for the arduinos portenta, nicla and giga
spare jacinth
i have the one of the left, it's been forgotten on a drawer for a few months, i could try and help with the porting work 🙂
my idea with it was to try out the LTDC peripheral, but it doesn't have external RAM, and the internal one is not big enough for the display i bought 🤣
slender iron
tannewt, you asked for suggestions on well supported Zephyr boards... I've been ...
nice! I'm excited to implement ble in the zephyr port
slender iron
I hope that by fixing the H7 port we can add support for more boards like the on...
My plan is to put more work in to the zephyr port and hope that the stm32 stuff comes with it
strange ravine
My plan is to put more work in to the zephyr port and hope that the stm32 stuff ...
lone axle
<@&356864093652516868> the weeking meeting will occur here on discord in just a bit over one hour from now at the usual time of 2pm US eastern / 11am US Pacific. You can add your notes to the document ahead of time if you like here: https://docs.google.com/document/d/1zy3Dm9Hb6pBi3phJY4S6hOdAvgtOnlOcx-dCMBeFvbo/edit?usp=sharing We look forward to hearing from everyone.
shut copper
PR added for List option for circup. @lone axle https://github.com/adafruit/circup/pull/240
