On mine I get 81K free after reload ( that is with wifi connected via settings.toml)
#circuitpython-dev
solar whale
crimson ferry
ok, mine was without connection... wow that's a lot
I guess the wifi module is pretty big imported
solar whale
I'll try w/o wifi
tulip sleet
it's not the importing per se, it's a native module, it's the setup it does to reserve RAM for its purposes.
solar whale
138K free w.o wifi !!
crimson ferry
Adafruit CircuitPython 9.1.0-beta.0 on 2024-03-28; Raspberry Pi Pico W with rp2040
>>> import gc
>>> gc.mem_free()
126160
>>> import wifi
>>> gc.mem_free()
125808
>>> import os
>>> gc.mem_free()
125504
>>> wifi.radio.connect(os.getenv("WIFI_SSID"), os.getenv("WIFI_PASSWORD"))
>>> gc.mem_free()
124624
``` # correctedhmm now I am getting 138K free even with wifi....
crimson ferry
is that with thw two modules removed? I have to correct my output above, I had a repl.py active
solar whale
Yes -- with PICODVI and USB_HOST removed
solar whale
From the build -- removing them reduces the RAM usage by about 14K
crimson ferry
so maybe there's 2K of other 8-->9 diffs
solar whale
a basic wifi_test works OK but I cannot connect to AIO to access a feed -- even without the modules.
short tendon
Sounds good. I have time this week
Is there any technical reason the ESP32SPI can't connect to a ntp server?
crimson ferry
shouldn't be... it can be a UDP client
wraith crow
Is there any technical reason the ESP32SPI can't connect to a ntp server?
I have connected to a ntp server using ESP32SPI
crimson ferry
tthere's an old version of the adafruit_circuitpython_ntp library that works with esp32spi
https://github.com/adafruit/Adafruit_CircuitPython_NTP/releases/tag/2.2.6 any v2 should work; v3 was the switch to native
short tendon
The old one used a built in method
wraith crow
I guess I was actually using adafruit_esp32spi.ESP_SPIcontrol.get_time()
short tendon
I was looking to see if I could get it to work for both native and esp32spi. I can cheat and use:
self._pool._interface.get_time()
But can't tell why the normal code doesn't work
If fails at connect
crimson ferry
@short tendon something like this, but with the NTP specifics? https://github.com/anecdata/Socket/blob/main/examples/udp_client_CircuitPython_ESP32SPI.py link corrected
wifi.connect, or socket.connect (a misnomer due to the non-standard NINA implementation)?
I can try playing with that again to make sure nothing broke
oh, here's one ready to go: https://github.com/adafruit/Adafruit_CircuitPython_ESP32SPI/blob/main/examples/esp32spi_udp_client.py
just tested it on PyPortal
short tendon
Well cool. Why is the connection different in the official ntp library...
crimson ferry
wifi or socket connect? are you meaning v2 of NTP lib?
short tendon
No v3 of the ntp lib
crimson ferry
ESP32SPI / NINA doesn't comply with that API
I tried mapping NINA/Arduino to standard socket calls once upon a time, but it was... not clean
crimson ferry
oh, here's one ready to go: <https://github.com/adafruit/Adafruit_CircuitPython_...
oh lol I forgot I wrote that
short tendon
Hahaha
crimson ferry
I don't know how people who do real development all day remember all. the. things.
short tendon
Where do most people get examples? GitHub or from the bundle?
crimson ferry
(I really don't know ...I tend to look for learn guide or github examples)
short tendon
I don't know how people who do real development all day remember all. the. thing...
Oh that's easy. I've been at the sample place for 6 years. I remember everything
I have this idea - for Internet libraries, templating the examples. And so:
code
code
{{ connection_method }}
code
code
Makes multiple examples replacing the {{ connection_method }} for each supported method...
crimson ferry
we could even add dummy connect into WIZnet ...but maybe that's just too painful
(then all three could be the same)
short tendon
I actually wish it had it. I find it frustrating that I can't create the object if it's not plugged in
short tendon
I have 😉
crimson ferry
it might break user expectations, where plugging in is magical
short tendon
If it were me, I would add a param to creation like require_link=True and so if you didn't want to require it, you could set that to false. And then connect would do that if not already done...
crimson ferry
that makes sense... conceptually it should be straightforware to split that stuff out into a connect function, and call it from either init or userland
short tendon
Yup
short tendon
I was able to get the code done today, but found out my trial expired. So I need to re-setup everything. I should still have it up tomorrow(ish)
orchid basinBOT
short tendon
@crimson ferry thank you! I have merged the 2 NTP methods and will be opening up a PR tomorrow. 1 more library that doesn't care!
solar whale
@danh FYI -- I am able to get the picow with the PICODVI and USBHOST removed to connect to AIO by switching to the .mpy versions of all the libraries. I've even sent my first images from the camera... progress
short tendon
@tulip sleet on 9210 espressif: add user-specified socket connect() timeout, did it work for you on an S2?
tulip sleet
OOPS, I forgot about that. I'll test and fix if needed.
tulip sleet
<@329766224093249548> on `9210 espressif: add user-specified socket connect() ti...
So... the timeout problem you found with ESP32-S2 is actually present in 9.0.4. It is not present in 8.x.x.
I opened the issue above ☝️
short tendon
I opened the issue above ☝️
Awesome. Glad to know I'm not crazy
short tendon
@tulip sleet I really want to dig into this Nina + Google SSL thing. Working sometimes and not others. Is there a way to know what cert it's validating against?
tulip sleet
<@329766224093249548> I really want to dig into this Nina + Google SSL thing. Wo...
you can just put the base url into a browser and look at the cert chain. It's a little easier to read in Firefox than Chrome. Click on the lock icon
short tendon
I think part of the issue is I'm getting different ones from different servers
And so wanted to compare that against when code works
tulip sleet
but I think it is a memory problem or something on NINA-FW. I need to do some testing on https://github.com/adafruit/nina-fw/pull/62
slender iron
@tulip sleet do you think anyone is relying on the storage extend stuff? I'm tempted to rip it out. (I just had a warning when I switched to one ota partition on c6)
tulip sleet
<@329766224093249548> do you think anyone is relying on the storage extend stuff...
it defaults to extended, is that right? I don't know, the author probably is. non-extended people would be using OTA
slender iron
it makes it way harder to protect where the user filesystem is
tulip sleet
reading...
https://github.com/adafruit/tinyuf2/issues/227 is also related
i think you just need to disable dualbank and the warning should go away. Are you saying that we shouldn't bother to extend CIRCUITPY into the OTA partition, or are you saying we shouldn't bother to have OTA?
slender iron
I'm going to remove ota1 in favor of extending ota0 for more cp code on c6 and h2 (and probably c3)
that has the nice property of leaving user flash as-is
short tendon
but I think it is a memory problem or something on NINA-FW. I need to do some te...
Is this ready in theory? I can also test
tulip sleet
Is this ready in theory? I can also test
it's ready to test, it does have some issues
tulip sleet
I'm going to remove ota1 in favor of extending ota0 for more cp code on c6 and h...
this makes sense, but then you don't need to remove extend, it should just work if the compile options are set up right. Maybe the makefile option setting/testing needs a little interking
short tendon
it's ready to test, it does have some issues
Okay, I'll play with it and see what happens
slender iron
this makes sense, but then you don't need to remove extend, it should just work ...
Ya, it does but it'll mess up the fs for folks who inadvertently used it
tulip sleet
the default is it's on, and it remembers the current state. It will only mess things up if people use the second arg in erase_filesystem()
and I have never heard of anyone doing that
slender iron
and I have never heard of anyone doing that
extend is default on though right?
tulip sleet
right
short tendon
@tulip sleet will the debug version give me any more detail by default?
tulip sleet
<@329766224093249548> will the debug version give me any more detail by default?
i don't remember!
short tendon
Fair. I'll try it and see
slender iron
<https://github.com/adafruit/tinyuf2/issues/227> is also related
The "my spi flash isn't all just CIRCUITPY" isn't a problem we can solve
tulip sleet
this is the situation as I understand it:
- on boards with two OTA partitions, dualbank is compiled in, but the second OTA partition is by default made part of CIRCUITPY. In that state,
dualbankraises an error that you can't do OTA. It will only become an OTA partition if you callstorage.erase_filesystem(extended=True) - So OTA support is latent, but it's there, for anyone who wants to use it, and the user gets a nice large CIRCUITPY.
- If we turned that off, then CIRCUITPY would be noticeably smaller, and some people would complain.
- If we just ditched OTA, then the few people who need OTA would complain and give up on us.
- For 4MB or smaller boards, makes sense to say "no OTA for you!". For larger boards, a few but non-zero number of people may want OTA. If the number of people who want OTA is < 10 on earth, well maybe it's not worth it.
slender iron
I think OTA for 8MB + makes sense
but it doesn't make sense to limit features for 4mb boards
tulip sleet
right, I agree for that, and so dualbank and extend should be disabled on 4MB boards, and enabled on >= 8MB boards
slender iron
that'll break anyone using extended storage already though
so I'll probably only do C6 and H2
tulip sleet
using extended storage on which boards?
almost everyone is using extended storage now on boards that support it
slender iron
right, which is why I won't do C3
tulip sleet
ok, so we might drop the OTA partition on 10.0, on those 4MB or smaller boards
slender iron
for C3 yes
and probably S3
if it weren't for this extended storage, we could drop it without disrupting the user fs
tulip sleet
yes, i agree with that. we have already done partition size changes, and people lived with that through alphas and betas. We just warned them.
tulip sleet
if it weren't for this extended storage, we could drop it without disrupting the...
I don't understand that - which boards do you mean?
slender iron
I'm looking to change the partition layout for the C6
but I'll leave the "user fs" one as-is
tulip sleet
do you mean like an 8MB board is like (made up numbers) 2MB OTA0, 2MB OTA1, 4 MB non-extended CIRCUITPY, but 4+2MB extended CIRCUITPY
slender iron
right so I'm removing OTA1 and replacing it with a 4mb OTA0
tulip sleet
so on C6, we haven't shipped anything yet, so it can just be OTA0 + CIRCUITPY, and extended is not functional
tulip sleet
i don't think we are disagreeing.
slender iron
I'm lamenting the extended storage making things complicated
tulip sleet
boards/adafruit_feather_esp32c6_4mbflash_nopsram/mpconfigboard.h
boards/espressif_esp32c6_devkitc_1_n8/mpconfigboard.h
boards/espressif_esp32c6_devkitm_1_n4/mpconfigboard.h
boards/unexpectedmaker_tinyc6/mpconfigboard.h
boards/weact_esp32c6_n4/mpconfigboard.h
boards/weact_esp32c6_n8/mpconfigboard.h
^^ these are the C6 boards:
i think the mistake was making OTA partitions for these small boards in the first place.
tulip sleet
so you are just correcting that. I don't mind telling people, watch out, your CIRCUITPY is going to be smashed
but we need to update tinyuf2 for all these also
slender iron
ota0 + ota1 may be bigger than 8mb ota0
no tinyuf2 because they don't have native usb
so it's up to us
tulip sleet
OH, yes, sorry
I am fine with "hey, in 9.1, these boards will have smashed CIRCUITPY, watch out"
slender iron
ota0 is 2048k for 8mb
tulip sleet
so it's not ripping out extended, it's just setting the compile options right, and fixing a non-far-sighted partitioning decision
slender iron
ota0 for 4m is 2816m without ota
tulip sleet
on like S3?
short tendon
Side question, I hadn't seen this extended storage before. Will I get more space if I set that to true? Or is it configured on the board by default?
tulip sleet
it defaults to true on boards with two OTA partitions (that is one partition for firmware and one for OTA)
so you are already getting it
slender iron
on like S3?
partition layout is agnostic to the chip. only* depends on flash size
gotta run now though
onyx hinge
@slender iron so far 2 c6 boards failing for lack of space; maybe once I can merge or rebase onto your c6 partition fixes it'll all be OK, but of course there might also be more failures to come. fingers crossed, and will look at it tomorrow. goodnight.
short tendon
@tulip sleet when you have a chance, can you please look at https://github.com/adafruit/Adafruit_CircuitPython_AzureIoT/pull/65? I want to make sure I get the requirements added correctly
stuck elbow
I'm trying to compile the uf2 bootloader for my board, and I'm getting a weird error, anybody knows what's wrong?
/home/sheep/dev/circuitpython/gcc-arm-none-eabi-10-2020-q4-major/bin/../lib/gcc/arm-none-eabi/10.2.1/../../../../arm-none-eabi/bin/ld: build/pewpew_lcd/bootloader-pewpew_lcd-v3.16.0-4-gd728aa1-dirty.elf section `.text' will not fit in region `rom'
/home/sheep/dev/circuitpython/gcc-arm-none-eabi-10-2020-q4-major/bin/../lib/gcc/arm-none-eabi/10.2.1/../../../../arm-none-eabi/bin/ld: section .binfo LMA [0000000000001ff0,0000000000001fff] overlaps section .text LMA [0000000000000000,000000000000202b]
/home/sheep/dev/circuitpython/gcc-arm-none-eabi-10-2020-q4-major/bin/../lib/gcc/arm-none-eabi/10.2.1/../../../../arm-none-eabi/bin/ld: region `rom' overflowed by 0 bytes
collect2: error: ld returned 1 exit status
make: *** [Makefile:161: build/pewpew_lcd/bootloader-pewpew_lcd-v3.16.0-4-gd728aa1-dirty.bin] Error 1
looks like there's just too much code. the text section would have to fit under 0x1ff0 but it doesn't.
I can't tell for sure but it seems like uf2-samdx1 is still using travis-ci !? and if so it's using gcc 9 still https://github.com/adafruit/uf2-samdx1/blob/master/.travis.yml#L18
stuck elbow
looks like there's just too much code. the text section would have to fit under ...
yeah, but is says it's overflowed by 0 bytes?
onyx hinge
I don't know why it says 0 bytes. I'm looking at these lines:
section .binfo LMA [0000000000001ff0,0000000000001fff] overlaps section .text LMA [0000000000000000,000000000000202b]
tame creek
maybe the linker script declares a rom region that’s all of the rom? text is still too big, and overlaps bininfo
tulip sleet
I'm trying to compile the uf2 bootloader for my board, and I'm getting a weird e...
can you compile other existing boards?
stuck elbow
yeah, it looks like it doesn't like the HAS_CONFIG flag
I don't see any samd21 using it, so it's probably too small
onyx hinge
@tulip sleet do you know if it's really still using travis ci? should it be updated to use GA?
tulip sleet
<@329766224093249548> do you know if it's really still using travis ci? should i...
travis doesn't run, as far as I know. When I make a new release I build the assets and upload them by hand. It would be nice to update it. Upstream doesn't have CI either.
should just remove the .travis.yml
short tendon
@tulip sleet for https://github.com/adafruit/nina-fw/pull/62, is it possible to retrigger a build? The artifacts have been deleted...
tulip sleet
<@329766224093249548> for https://github.com/adafruit/nina-fw/pull/62, is it pos...
I pushed an empty commit to re-run. It appears that once the artifacts and logs have expired, there's no way to re-run.
burnt tangle
Hi all, I'm learning to develop new modules and I'm wondering is the source to guides like https://learn.adafruit.com/building-circuitpython and https://learn.adafruit.com/extending-circuitpython in a repo somewhere? So far I've looked in circuitpython, Adafruit_Learning_System_Guides, and circuitpython-org. Thanks!
slender iron
Hi all, I'm learning to develop new modules and I'm wondering is the source to g...
The source isn't on github unfortunately. Learn has its own database
slender iron
<@252717193496756235> so far 2 c6 boards failing for lack of space; maybe once I...
We can merge your changes and I can enable it with the partition table updates.
I'm thinking I may change the default partition table for everything and override it for existing boards
burnt tangle
The source isn't on github unfortunately. Learn has its own database
mmm, I was thinking that might be the case. Cut and paste it is! Nice meeting you yesterday BTW
slender iron
you too!
tulip sleet
Hi all, I'm learning to develop new modules and I'm wondering is the source to g...
The extending circuitpython is out of date in a number of ways
the overall idea is fine but there are many changed details
burnt tangle
The extending circuitpython is out of date in a number of ways
Yes, thanks! Right now I'm looking at the code for analogbufio and audiopwmio for the real truth. I just wanted to be aware of what documentation exists and where it lives.
manic glacierBOT
spare jacinth
im trying to test my changes to the setboard command, however im pretty sure i need to go through setup.py for this "resource" thingy to work...
running pip install git+https://github.com/elpekenin/circuitpython@feat/stub/list will take a year because it will clone every single submodule
can i install/test in some other way? 😅
tulip sleet
im trying to test my changes to the `setboard` command, however im pretty sure i...
I don't understand why it would clone the submodules. But would pip install --dry-run help?
lone axle
im trying to test my changes to the `setboard` command, however im pretty sure i...
I think when I was working on that I would just build the stubs locally and then install with pip install -e . in the directory of the generated stubs.
I'm not sure if / how many of the submodules would have been required because it was a repo I also use for building the core so I did have all of them at the time most likely, but it didn't download them specifically for installin the stubs AFAIK.
and if you use -e then you can just build new versions of the stubs as you make changes and the venv that you installed it into will automatically see the latest built version.
spare jacinth
I don't understand why it would clone the submodules. But would `pip install --d...
it does, that's all i can say 🙃 ```
$ pip install git+https://github.com/elpekenin/circuitpython@feat/stub/list
Collecting git+https://github.com/elpekenin/circuitpython@feat/stub/list
Cloning https://github.com/elpekenin/circuitpython (to revision feat/stub/list) to /tmp/pip-req-build-fmlubuhb
Running command git clone --filter=blob:none --quiet https://github.com/elpekenin/circuitpython /tmp/pip-req-build-fmlubuhb
Running command git checkout -b feat/stub/list --track origin/feat/stub/list
Switched to a new branch 'feat/stub/list'
branch 'feat/stub/list' set up to track 'origin/feat/stub/list'.
Resolved https://github.com/elpekenin/circuitpython to commit 933efe082e80e16f44d6125cb2e8d4553a9a4f0c
Running command git submodule update --init --recursive -q
^CERROR: Operation cancelled by user
i left it run for like 5/10 mins the first time after giving up on waiting 🤣
spare jacinth
I think when I was working on that I would just build the stubs locally and then...
oh, i had tried -e root/setup.py-stubs, lets see
manic glacierBOT
spare jacinth
that worked perfectly, thx foamy!
tulip sleet
@lone axle all (or nearly all) the library conf.py files contain this:
# The short X.Y version.
version = "1.0"
# The full version, including alpha/beta/rc tags.
release = "1.0"
This causes the HTML < title > tags to contain "1.0", which shows most noticeably on search engine results for various libraries.
This is wrong. Where do you suggest I open an issue for this? in the bundle repo or in the cookie cutter repo (where this comes from), or somehwere else. I found some suggestions for gettin the version dynamically
lone axle
<@382939733107408897> all (or nearly all) the library `conf.py` files contain th...
I think either cookie-cutter or bundle repo is fine. Ultimately we'll need a bit of work in both places. a PR to cookie-cutter to correct it for new libraries, and some kind of patch run on the bundle to update the existing ones.
tulip sleet
thanks, I'll open an issue there.
short tendon
sooo many cookie-cutter patches...
spare jacinth
@short tendon 🥳
$ circuitpython_setboard -l
Available boards are:
* 01space_lcd042_esp32c3
* 0xcb_helios
* 42keebs_frood
* 8086_commander
* ADM_B_NRF52840_1
* Seeed_XIAO_nRF52840_Sense
* TG-Watch
* adafruit_esp32s3_camera
... and the list goes on beyond infinity ...
any suggestions for another formatting (eg use - instead) and/or how to make the list friendlier than needing a > file for Ctrl+F'in that file? 🤣
short tendon
<@1152454697614069811> 🥳
```
$ circuitpython_setboard -l
Available boards are:...
You could regex on the __init__.py file. Example:
"""
Board stub for 0xCB Helios
- port: raspberrypi
- board_id: 0xcb_helios
- NVM size: 4096
- Included modules: _asyncio, _bleio, _pixelmap, adafruit_bus_device, adafruit_pixelbuf, aesio, alarm, analogbufio, analogio, array, atexit, audiobusio, audiocore, audiomixer, audiomp3, audiopwmio, binascii, bitbangio, bitmaptools, bitops, board, builtins, builtins.pow3, busdisplay, busio, busio.SPI, busio.UART, codeop, collections, countio, digitalio, displayio, epaperdisplay, errno, floppyio, fontio, fourwire, framebufferio, getpass, gifio, hashlib, i2cdisplaybus, i2ctarget, imagecapture, io, jpegio, json, keypad, keypad.KeyMatrix, keypad.Keys, keypad.ShiftRegisterKeys, locale, math, memorymap, microcontroller, msgpack, neopixel_write, nvm, onewireio, os, os.getenv, paralleldisplaybus, pulseio, pwmio, qrio, rainbowio, random, re, rgbmatrix, rotaryio, rp2pio, rtc, sdcardio, select, sharpdisplay, storage, struct, supervisor, synthio, sys, terminalio, time, touchio, traceback, ulab, usb, usb_cdc, usb_hid, usb_host, usb_midi, usb_video, vectorio, warnings, watchdog, zlib
- Frozen libraries:
"""
So you could grab what's after the Board stub for
spare jacinth
yeah, more info is nice, BUT! the list is already terribly long, and that info could be seen by user on the stub (or docs)
short tendon
yeah, more info is nice, BUT! the list is already terribly long, and that info c...
The other option is that in https://github.com/adafruit/circuitpython/blob/main/tools/board_stubs/build_board_specific_stubs/board_stub_builder.py, you could also build a file that is included with the stubs that it just prints out
spare jacinth
gotta love automod's false positives!
anyway... i already parsed the file structure, no need to write another file
slender iron
gotta love automod's false positives!
try again. I removed the rule you hit
short tendon
just thought that way you could make a file that was:
Available boards are:
* 0xcb_helios: 0xCB Helios
... and the list goes on beyond infinity ...
made it list by port 😉
if args.list:
port_boards = defaultdict(list)
for board in resources.files("board_definitions").iterdir():
# NOTE: For the hand-crafted finding of port and prettified name in
# the docstring, its format is assumed to be:
#
# <empty line>
# Board stub for ...
# - port: ...
# - board_id: ...
# - NVM size: ...
# - Included modules: ...
# - Frozen libraries: ...
#
lines = get_doc_or_raise(board).split("\n")
port = lines[2].split("-")[1].split(":")[1].strip()
port_boards[port].append(board)
sys.stdout.write("Available boards are: \n")
for port, boards in port_boards.items():
sys.stdout.write(
header(port)
+ " * "
+ "\n * ".join(boards)
+ "\n\n"
)
sys.exit(0)
also would be cool if you coud do --list m4 and get boards with that name...
spare jacinth
just thought that way you could make a file that was:
```
Available boards are: ...
and i thought about printing the .split("stub for ")[1] which is a prettier name, but i think it is counter productive since users will have a 50% chance of using the wrong name (real id vs friendly) 🤣
short tendon
fair
spare jacinth
you can see how my comment still has the "pettified name", oopsie
now grouped by port, and both the ports and the boards for each port are sorted ^^ ```
...
renode
- renode_cortex_m0plus
stm
- espruino_pico
- espruino_wifi
- feather_stm32f405_express
- meowbit_v121
- nucleo_f446re
- nucleo_f746zg
- nucleo_f767zi
- nucleo_h743zi_2
- openmv_h7
- pyb_nano_v2
- pyboard_v11
- sparkfun_stm32_thing_plus
- sparkfun_stm32f405_micromod
- stm32f411ce_blackpill
- stm32f411ce_blackpill_with_flash
- stm32f411ve_discovery
- stm32f412zg_discovery
- stm32f4_discovery
- stm32f746g_discovery
- swan_r5
- thunderpack_v11
- thunderpack_v12
search/filter will be later today or during weekend, leaving now. code here if anyone curious
short tendon
🚀
manic glacierBOT
tulip sleet
@onyx hinge I am debugging a storage issue with ssl sockets, and have some questions about the (now shared) SSL implementation, particularly about wrap_socket. Do you have time for a text or audio or video chat? I have to be afk in about 45 minutes
onyx hinge
@tulip sleet sorry, I was out of the house and didn't see your message. will you be back later
(I'm still 15 minutes away from home)
tulip sleet
won't be back online until tomorrow
no problem - i can talk about it later. The essence of the q is why mbedtls is re-inited on every socket wrap, instead of just being set up once when ssl context is created.
i can fix this if it is not what you meant, don't need for you to work on it yourself
onyx hinge
Ok, I won't be online much this weekend. We can catch up next week maybe?
tulip sleet
sure
onyx hinge
@tulip sleet you mean
mbedtls_ssl_init(&o->ssl);
? Yeah I was actually wondering about that myself while working on ssl-anything. It seems like some of those things that are in the ssl socket struct might belong on the ssl context struct in C.
tulip sleet
Yes. I am going to try moving them. Might fix some other storage problems we are seeing. I was getting Memory Error on S2 but not S3 and tracked it down to this. I’m already out
onyx hinge
OK. lmk if you want me to take it over, it looks like a problem of my making. I was confused what you meant by "storage" at first (thought you meant like CIRCUITPY drive storage) but this makes sense.
spare jacinth
search/filter will be later today or during weekend, leaving now. code [here](<h...
yay, that only took a couple lines of code 🤓
spare jacinth
heading to bed, will check feedack tomorrow 😜
manic glacierBOT
I noticed that root_group typo but idk why it wasn't giving error for that, it happened during copy pasting it here.
But its still not working, i saved the code as code.py and i also used print statements, even till the end those print statements are working which can be seen in the ss below and lcd isn't showing any signs.
What if this is some driver issue or some issue with cir...
manic glacierBOT
short tendon
@tulip sleet for Monday: I was updating a personal project and noticed 9.0.4 of circuitpython-stubs didn't release. Guessing a build problem?
Which is strange, since 9.1.0b1 did and it was released after 9.0.4...
tulip sleet
<@329766224093249548> for Monday: I was updating a personal project and noticed ...
There was a problem building the stubs, which was fixed post 9.0.4 by https://github.com/adafruit/circuitpython/pull/9186. I can't just re-run the job because the version would be different but maybe I could manually upload it. Where is it you're getting the stubs from? From pypi?
short tendon
There was a problem building the stubs, which was fixed post 9.0.4 by <https://g...
Yup. Although looking there aren't any new boards or pin changes, so It wouldn't be any different from 9.0.3
tulip sleet
maybe I'll jsut copy 9.0.3 as 9.0.4
short tendon
I'm sure would be fine, or just wait for 9.1.0
tulip sleet
if it's easy to find the latest version, then yes, probably won't bother to do anything
short tendon
yeah, pip install works just fine. I was setting up a requirements.txt and had assumed it existed
manic glacierBOT
grizzled raven
Any reason that pwmio might interfere with pulseio? I'm trying to talk to an IR receiver while driving motors with PWM. A 100% duty cycle has no problems. But anything less and I start to get errors from adafruit_irremote : "Pulses outside mark/space"; "Both even/odd pulses differ"; and "Too short";
I am doing a nonblocking decode from within an async task, and the IR receiver scanning's in a different task.
devout jolt
Any reason that pwmio might interfere with pulseio? I'm trying to talk to an IR...
Any time you have motors you have noise. Try running your code again but with the motors physically disconnected. If you get the same errors, then you're getting noise on your power bus from the motors
grizzled raven
Any time you have motors you have noise. Try running your code again but with th...
It looks like you may be right - the problem seems to disappear if I power the L9110 board from a separate set of batteries with a common ground. Any idea on other approaches that might fix it?
devout jolt
That’s typically how larger motor noise issues are solved: by having separate power. The easiest is two totally separate supplies like what you got. You may be able to get by with filter capacitors near the affected devices (try 10uF and 0.1uF). You also could try having a separate voltage regulator off your main power supply to create a “second” power source. It’s not an easy problem to solve generally, as it really depends on your setup
manic glacierBOT
lone axle
Would it be possible or desireable to include in boot_out.txt a list of frozen modules that are available in that build of the core system? I find myself not keeping up with which ones are frozen on which devices and sometimes getting bitten by thinking I'm using one version of a library that I've copied over to my device when it's actually not using the one I copied because it's got a frozen one.
If there were a quick easy place to check on the CIRCUITPY drive itself I would definitely check there as a first step in troubleshooting. I know where to find them in the core repo, but the quicker access (and availability to non developers) would be appreciated I think.
burnt tangle
Hello experienced devs, I see that micropython uses a blizzard of types, many of which are generated by intricate macros. It is starting to make sense, but I am wondering, are there any tools you use to help make sense of this, or do you just, you know, suck it up? For instance, wondering if the preprocessor output sheds any light, or if looking at the memory while the code is running is helpful in revealing the structure.
tulip sleet
Hello experienced devs, I see that micropython uses a blizzard of types, many of...
Most of the macros are not so terrible, except for one set having to do with building object types. The use of most of the macros is very idiomatic, so rather than try to understand exactly what they are generating, reviewing the skeletal structure of a particular module will get you pretty far.
I don't think looking at the preprocessor output will help you a lot. Many macros are casts or struct-creating macros
tulip sleet
Would it be possible or desireable to include in boot_out.txt a list of frozen m...
The frozen modules are included in help("modules").
lone axle
Ahh, I think I've seen that before but had forgotten about it. Thank you.
tulip sleet
we could mark the frozen ones more clearly. help("modules") on CPython has a bunch of extra verbiage (try it), so I don't think adding a few markers or making the frozen list separate would be terribly incompatible.
lone axle
I will try it in a more limited venv a little while later. I tried in my main python instance. It outputed several warnings about Future, Deprecation, and UserWarning things and then eventually crashed with Aborted (core dumped) I've got a decent list of stuff installed in that main python instance that could be causing issues
tulip sleet
i was just trying it not in a venv
burnt tangle
Most of the macros are not so terrible, except for one set having to do with bui...
Yes, that's the one, MP_DEFINE_CONST_OBJ_TYPE. Ok, I won't dwell, thanks!
tulip sleet
yes, that one is recent, and really bumped up into the limitations of cpp.
onyx hinge
@tulip sleet the ref is e8a4c1dd537c7ccfba92936c6eee9d5cf529be9e
tulip sleet
<@329766224093249548> the ref is e8a4c1dd537c7ccfba92936c6eee9d5cf529be9e
i didn't realize you were hosting, sorry for taking that time
I can put in the state statistics
onyx hinge
I didn't realize I was hosting either 😅
<@&356864093652516868> hey y'all! turns out the weekly meeting is in 20 minutes and I'm the host! please add your notes if you can, and I look forward to seeing you in the voice channel at the top of the hour.
tulip sleet
@onyx hinge want me to do commuinty news while you set up?
onyx hinge
@tulip sleet If you want to grab another item or two that'd be dandy!
@tulip sleet @gilded cradle @lone axle can y'all grab the usual "state of" sections today?
tulip sleet
sure
lone axle
Yep
gilded cradle
@onyx hinge my internet keeps going in and out, so if it's offline at the time, could you read it?
onyx hinge
@gilded cradle sure, or if it's easier on you I could just plan to read it no matter what
gilded cradle
I should be getting a new cable modem tomorrow.
onyx hinge
sorry you're having internet troubles 😕 and hope that it's solved soon
gilded cradle
Yeah, modem was supposed to be here Saturday, but shipping delays...
Well, they just rescheduled it for later today.
midnight ember
18 people opened 100 issues??? huh.
ornate breach
18 people opened 100 issues??? huh.
Some people just do so much with circuitpython 😊
midnight ember
If they're regulars then I can believe it sure. Just wondering if there was some kind of PyCon sprint for finding bugs.
ahhh
i moved you up dan hope that's ok
Justin 👏
sampling single pixels and returning them
Nice work Jerry!
thorny jay
Maybe the Raspberry Pi foundation is working on a new RP2040 like microcontroler... let's cross the finger. It could have more memory. 😉
plucky tulip
Thanks, I remember it used to work back in 8.x
on PicoW
I see, thanks for explaining 🙂
midnight ember
Thank you for hosting @onyx hinge
lone axle
Thanks for Hosting Jeff. Have a great week everyone!
short tendon
Thanks!
thorny jay
People add memory on DE-10 (FPGA) to have more memory for more "emulation" (That is MiSTer FPGA).
midnight ember
Where there is a will there is a way but the chip has to support it.
onyx hinge
someone sufficiently advanced could write a high performance ARM-to-ARM JIT that treats the internal RAM as a cache and has a software/PIO PSRAM drier....
just a matter of sufficient dedication right?
midnight ember
You'd have to completely hack the firmware to shreds I guess... if someone is willing to do that for a pico w go for it.
yeah
If you would have told me when the pico came out that it would eventually run HDMI I wouldn't have believed it.... stranger things have happened.
manic glacierBOT
onyx hinge
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! <@&356864093652516868> https://docs.google.com/document/d/11WCg1nhwS0UgcSXYWZjla7fmye8TP3Y8f_N9kDZCVHo/edit?usp=sharing
pallid creek
I just noticed that the GFX library is going to be deprecated. https://docs.circuitpython.org/projects/gfx/en/latest/
Does that mean there will no longer be an immediate mode graphics api for CP?
pallid creek
I realize that displayio is the right API for most cases, but there are times when it is useful to have more control. In particular I'm thinking about partial updates for eInk displays.
manic glacierBOT
misty garnet
anyone else get complaints about the build process with adafruit_ticks:
File "/opt/hostedtoolcache/Python/3.11.9/x64/lib/python3.11/site-packages/adafruit_ticks.py", line 31, in
_TICKS_MAX = const(_TICKS_PERIOD - 1)
TypeError: unsupported operand type(s) for -: 'const' and 'int'this is during the automated build process for a pull request
onyx hinge
anyone else get complaints about the build process with adafruit_ticks:
File "/o...
@misty garnet that's not ringing a bell, and the last CI in there (5 months ago) was green. the type of const(1) should be int, so subtracting 1 from it shouldn't be a problem. Where are you seeing this message?
_TICKS_PERIOD = const(1 << 29)
_TICKS_MAX = const(_TICKS_PERIOD - 1)
(that is, the last CI inside adafruit_ticks)
misty garnet
Pull request for adafruit_minimqtt
onyx hinge
OK
I can help with that
during the build process you're getting this:```
File "/home/runner/work/Adafruit_CircuitPython_MiniMQTT/Adafruit_CircuitPython_MiniMQTT/adafruit_minimqtt/adafruit_minimqtt.py", line 39, in <module>
from adafruit_ticks import ticks_ms, ticks_diff
File "/opt/hostedtoolcache/Python/3.11.9/x64/lib/python3.11/site-packages/adafruit_ticks.py", line 31, in <module>
_TICKS_MAX = const(_TICKS_PERIOD - 1)
~~~~~~~~~~~~^
TypeError: unsupported operand type(s) for -: 'const' and 'int'
this occuring during the build of the documentation
this occurs because pydoc creates fake versions of modules that can't be imported, and sometimes these fake objects don't behave in the necessary way.
because adafruit_ticks is installable on standard python, the solution is to list it in docs/requirements.txt, so that it is installed during the CI build process, as was done with adafruit-floppy for instance: https://github.com/adafruit/Adafruit_CircuitPython_floppy/blob/main/docs/requirements.txt
if that still doesn't get you un-stuck please @ me on the bug and I'll try to take a peek in tomorrow
er on the PR
misty garnet
Ok, I'll give it a try and let you know how it goes, thanks for the help
short tendon
So, I see that Adafruit has already allowed codecov into its organization. Any issues with me adding it to a repo for coverage to see how it looks (would probably try with ConnectionManager)?
tulip sleet
no objections
onyx hinge
@misty garnet second guess, remove micropython from the explicit mock list in docs/conf.py: docs/conf.py:autodoc_mock_imports = ["micropython", "microcontroller", "random"]
it's micropython.const that's getting incorrect behavior via being "mocked"
misty garnet
I'll give it a try later today
manic glacierBOT
misty garnet
@onyx hinge removing micropython import from docs/conf.py did the trick, thanks for the help
manic glacierBOT
alpine holly
When creating a new board inside espressif ports for circuitpython, is there a standard way to define the serial mapping of DTR and RTS pins to the ESP32 reset and GPIO0 ? While the build itself doesn't care, other tools definitely need to know this (e.g. the online flash tool, serial emulators, upload/download scripts, etc)... I'm doing the port for the popular "ESP32CAM": does anyone else know another board which has hardware serial which also wires DTR (data terminal ready) and RTS (request to send) to the ESP32 reset and gpio0 pins (i.e. so I can see if/how they supported this?)
tulip sleet
When creating a new board inside espressif ports for circuitpython, is there a s...
All the time I've been using serial on Espressif chips, we have never used DTR/RTS. We simply use TX/RX. Any flow control is handled by the protocol of what's going over the serial lines
for instance esptool.py doesn't need external flow control
onyx hinge
esptool can support a "custom reset sequence" https://github.com/espressif/esptool/blob/1deb1c65c140363bbfa16fbd889cff8e6d33b8c0/docs/en/esptool/configuration-file.rst#custom-reset-sequence via an esptool.cfg file. We have not needed to add support for such a file on a per-board basis, but putting esptool.cfg in the board AND setting the relevant environment variable ESPTOOL_CFGFILE for make flash and make flash-circuitpython-only would seem like one approach.
but yeah, to date we have not identified a need for this, as the "standard reset sequence" is so widely supported by esp boards in the wild.
ornate breach
Does circuitpython support BLE audio streaming on nRF52840?
onyx hinge
@ornate breach no
alpine holly
@onyx hinge thanks, great pointers and good idea. @tulip sleet some boards (ESP32CAM most notably) physically wire the hardware serial chip flow control lines to the ESP32 reset and GPIO0 lines (and do not use them for flow control of course)... this allows software programmers to automatically "hardware reset" the board, and optionally put it into firmware-upload mode, even when the last uploaded-sketch might be crashing. I'm not familiar with how other boards handle that (probably force you to manually press a button?), but the issue with these automatic ones is that there's a 50%/50% chance that your non-arduino terminal emulator (and/or command-line tools like "ampy" etc for talking to micropython) is not going to work for these boards, because the tools (or if not the tools, the windows/mac/linux/wsl1/wsl2/etc operating system itself when the tools do not specify otherwise) incorrectly (from the point of view of the ESP32CAM, which is really "correctly" from the point of view of something expecting hardware flow control) set the state of these flow-control pins (even when you tell the tools you're not using flow control). Long story short - this seriously "trips up" most people trying to use an ESP32CAM (and similar - I think Freenove and esp32-wrover-dev do this too)
stuck elbow
but that's not something that has anything to do with circuitpython, is it?
onyx hinge
or to put it another way, can you clarify what CircuitPython would do with this information?
manic glacierBOT
tidal kiln
would this need updating for latest CP UF2? (due to UF2 file size)
TinyUF2 Bootloader 0.14.0-5-g45bc2fc - tinyusb (0.15.0-331-ge3b3229d6)
Model: Adafruit MatrixPortal S3
Board-ID: ESP32-S3-MatrixPortal-revB
Date: May 18 2023
not finding the summary of what versions work for what UF2 file sizes, etc.
devout jolt
Hi! Both you and @tulip sleet have run into this problem. I'm trying to make a library using cookiecutter and getting this AttributeError: module ‘pkgutil’ has no attribute ‘ImpImporter error when committing. I'm using Python 3.12 and have many other dependencies on it. Is there a way to tell pre-commit and its tools to use Python3.11, which I also have installed?
tulip sleet
would this need updating for latest CP UF2? (due to UF2 file size)
```
TinyUF2 B...
it's only older nRF bootloaders that had a problem that I know of. UF2 file size is larger than the firmware size by a lot
devout jolt
Hi! Both you and <@329766224093249548> have run into this problem. I'm trying to...
To answer my own question, looks like the solution is to add this to the top of .pre-commit-config.yaml:
default_language_version:
# force all unspecified python hooks to run python3.11
python: python3.11
I think if you bump the pylint version in .pre-commit-config.yaml to 3.something, that may work
lone axle
The core module hashlib seems to only support the sha1 algorithm. Inside of the core code I found that it's including things from mbedtls/ssl.h https://github.com/adafruit/circuitpython/blob/main/shared-module/hashlib/__init__.c#L30 I found mbedtls inside of lib/ and what I'm guessing is the ssl.h it's referring to: https://github.com/Mbed-TLS/mbedtls/blob/981743de6fcdbe672e482b6fd724d31d0a0d2476/include/mbedtls/ssl.h Is anyone able to confirm that is actually the correct ssl.h? And secondarily I'm interested in trying to make a build that has sha256 enabled inside of the core hashlib. I see many references to sha256 inside that ssl.h file and the mbedtls include directory, is it just a matter of adding a bit more code to hashlibs shared-module and shared-bindings __init__.c files? Or is there more a lot more to it than that?
onyx hinge
There may be changes necessary to other build or configuration files, but I don't know. py/circuitpy_defns.mk names files related to the implementation sha1 and sha512. fwiw without a real good reason I wouldn't include md5 or sha1, they are both considered cryptographically broken.
wait, the one we have is sha1? oof
alpine holly
@stuck elbow yes, because (for example) https://code.circuitpython.org/ - giving people a board they can't use because no tools can connect to it isn't a nice thing to do to them 🙂 [ hence my original question: is there a standard way to define the serial mapping of DTR and RTS pins to the ESP32 reset and GPIO0 ]
tulip sleet
<@351136009922019331> yes, because (for example) https://code.circuitpython.org/...
I am not sure what you mean by "define the serial mapping". are you saying that external tools (e.g. a USB-serial converter, driven by some software) want to attach to those pins? Are you saying that the CircuitPython REPL should be paying attention to the DTR and RTS pins? When I use a usb-serial converter, normally do not connect the DTR and RTS pins. I know that the standard terminal block includes those. But suppose you don't connect them, and only connect TX and RX. What functionality are you losing by not DTR and RTS, and does the external software assume those connections (esptool does not, for instance).
If DTR and RTS are tied to ESP32 RESET and GPIO0, then the state of those pins will allow external resets, etc., as you say. But that is way before CircuitPython even starts. CircuitPython pays no attention to flow control for the REPL on any pins. And indeed, since one of the the pins is reset, it cannot be used for flow control, since toggling it will reset the board.
what is it that you can't do right now with these boards?
lone sandalBOT
alpine holly
@tulip sleet "CircuitPython" is an ecosystem not just a firmware image. Other things need to talk to the firmware, and those need to know when these DTR/RTS features are available, so they can (a) use them, and (b) not cause everything to "not work" because they're ignoring them. While the things that use these pins (mu, vonny, https://code.circuitpython.org/, ampy, vscode, etc) are what's supposed to honor this, the most logical place for where this info should be authoritatively recorded, is someplace inside the board definitions. It is because circuitpython (the company) also has a product which is not working on account of this info not being recorded anyplace that I'm asking questions about this in here. It's a chicken-and-egg problem. https://code.circuitpython.org/ isn't going to get perfectly fixed unless ports/espressif/boards/* documents when this is required (which it isn't doing, because https://code.circuitpython.org/ doesn't look) ...
alpine holly
It's actually a somewhat larger issue - some tools work (e.g. arduino itself, micropython online flash components, the https://code.circuitpython.org/ REPL when you pick "circuitpython" - which makes no sense since these boards do not support USB FS, so https://code.circuitpython.org/ isn't usable when you do that), but most other tools fail. I've fixed the bug everywhere I can find (ampy, mpuremote, and a bunch of others) and provided workarounds for the root cause of the problem (python itself, which on some (but not all) operating systems does not allow serial constructors to specify what to do with those pins, and defaults to pulling them low - which locks boards in either reset or DFU states - I filed the relevant bug reports and someone there seems to be looking at it... the workaround it to manually construct a serial connection before telling it what port to use, set the pins as needed, then specify the port.). Confusing things even more - some boards wire reset this way using a capacitor (thus always see momentary resets) and others use a transistor (so pulling reset low locks up the board indefinitely)
stuck elbow
circuitpython is not a company
I still don't see what you want us to do with that information
alpine holly
@stuck elbow sorry, I mean adafruit, the owners of the circuitpython trademark and who run circuitpython.org website I still don't see what you want us to do with that information => my original question: is there a standard way to define the serial mapping of DTR and RTS pins to the ESP32 reset and GPIO0 I
stuck elbow
no
alpine holly
WOO HOO!!! Got the ESP32CAM working almost perfectly. Camera - YES! SD-Card - YES! LED+Flashlight - YES,YES!
Huge marathon... I've sat here for 12 hours on that... Just gotta test wifi. What's the developer ettiquette on this channel? Do n00b geeks like me get flamed for asking dumb questions? I'm wondering a bunch of stuff (like whether or not I should use the "AI Thinker" trademark in my board - they deisgned it, but I'm working on clones and have no idea if an original would work...)
Is this even the right channel for board devs ?
stuck elbow
this is the right channel for the firmware development and adding support for new boards, but we might be a bit lacking in legal knowledge about trademarks
I'm not a lawyer, but it seems that only the owner of a trademark should use their trademark
if you use someone else's trademark without permission, they can sue you
alpine holly
What, if any, procedure exists for including "tools" for a board? e.g. code samples for taking photos, mounting SD card, illuminating the flashlight, etc ? While some existing things work ( https://learn.adafruit.com/capturing-camera-images-with-circuitpython/example-qr-code-scanner ) others do not, and most of those require unnecessary (internet accounts)/unsupported(LCD screens) features and don't make it easy for new users to get started.
Another question (sorry! getting my ducks-in-a-row before creating a new-board PR...) - does anyone know how to use SD_Card in 4-pin (SDIO) mode, instead of 1-pin SPI mode - or if it's even possible? Seems a shame to limit the speed to 25% when all pins are soldered...
stuck elbow
What, if any, procedure exists for including "tools" for a board? e.g. code sam...
I would say that the board usually has its own website somewhere, and you can link to it from the circuitpython.org page
for adafruit boards this is usually the shop page and/or the learn guide for that board
for third party boards this is the website of the manufacturer
alpine holly
Thanks... I find they've called it SDMMC in the source, so I'm working on that now.
empty salmon
Looking for guidance: Default state of enable pins
I am working on an RP2040 board with built-in epaper display, neopixels, and sound.
When using sleep modes, I want to reduce power. The three mentioned features each have an _EN pin.
When the board first boots up, should the enable pins be “active” or should the documentation and example code always include setting these pins to turn on the features?
When I look at the Adafruit MagTag, they default to not setting their enable pins at startup.
alpine holly
I am NOT an expert, but, I've been "saved" in the past when I need to connect lots of things to boards with not-many-spare by simultaneously wiring pins to multiple devices, and using software to carefully control them and not use things that need the same pins at the same time... if stuff was getting "enabled" by default, that would mess me up. On a philosophical note: users can "enable" (not not!) anything they want during boot in their code.py, however, if you did that for them, you've denied them the chance to have the option... Semi-relatedly, I notice that if my wifi settings are not inside settings.toml, the wifi does not aautomatically get set up... that's another possible option for you ?
empty salmon
There is the question of "usability". With the CircuitPython focus on "ease of use and learning", pre-enabling is one less thing to remember and means the existing learning guides "just work".
By contrast, not pre-enabling is more power conscious.
Using MagTag and a single point of reference suggests there is precedence to no enable in the firmware.
(I can talk myself into anything. I'm very convincing 😜)
alpine holly
Nobody can use any of these boards without creating settings.toml (usually to add wifi credentials), so one of (a) making enablement an option in there, or at least (b) making "do not enable at all" (taking care not to enable first, then disable, since that's going to break stufffor anyone who diodn't want it enabled in the first place) at least a possible option in there, seems to make the most sense?
tulip sleet
<@329766224093249548> "CircuitPython" is an ecosystem not just a firmware image....
The boards that use DTR and RTS for non-standard functions such as reset I believe are ill-designed. The reset problem you describe is not related to the firmware running on the board. The solution in this case is not to connect all the pins of the USB-serial adapter to the board.
We don't control what third-party software does in its "CircuitPython" mode.
we are working on making code.circuitpython.org work on boards that do not present CIRCUITPY, but instead only present a REPL through a serial port.
files will be read/written through the REPL, as with ampy, etc.
alpine holly
@tulip sleet Those pins are connected for a reason - so programming is possible without humans needing to manually push buttons. There's millions of these boards out there - possibly more esp32-cam boards exist than any other esp32 or even circuitpython compatible boards - they're crazy-cheap, and often sold in the hundreds.
tulip sleet
the reset pin is not under control of CircuitPython so it is not part of the board definition. GPIO0 could be used for GPIO and is sometimes included
yes but those board break the definition of what DTR and RTS are usually for
if you want features to be added the the host-computer tools that know about the non-standard uses of those pins, that's ok, but it doesn't really have to do with CircuitPython
alpine holly
"we are working on making code.circuitpython.org work on boards that do not present CIRCUITPY, but instead only present a REPL through a serial port." - that's browser-based scripting. It already works (for micropython). That script needs to know this option. That's the whole point of me writing here about it.
tulip sleet
it has to do with those tools. To the extent that we control any of those tools, we're happy to entertain requests for such features.
alpine holly
the micropython firmware-upload stuff works fine. the adafrut copy of it does not. someone broke this specific issue when they cloned that...
this works: https://esphome.github.io/esp-web-tools/
this does not: https://code.circuitpython.org/
I'm happy to fix that for them later, if that's something they accept PRs on.
Quick question: whose job is it to fix a lint error of a single missing space in my PR? I assume either some tool, or whoever merges it, does that ?
tulip sleet
the formatting is checked by pylint. This PR run fails if the formatting is not right. You can use pre-commit to check the formatting before you submit the PR. https://learn.adafruit.com/creating-and-sharing-a-circuitpython-library/check-your-code
we're happy to entertain PR's on anything
alpine holly
it's C, not python.
tulip sleet
"*we are working on making code.circuitpython.org work on boards that do not pre...
I am saying we are working on adding that feature to code.circuitpython.org
alpine holly
It (https://github.com/adafruit/circuitpython/pull/9231) actually shows the diff of where the extra space needs to go - is there something I can do to "accept" that and let it make the change ?
tulip sleet
you can make the change manually and push a new commit, or use pre-commit locally and have it clean up your code before you commit
it is an "error" from the point of view of the PR submission checks, so it's something for the submitter to fix, like a compilation error.
alpine holly
(p.s. "code.circuitpython.org" already has the feature you are talking about. You just select "micropython" instead of "circuitpython", and the extra buttons show up in the UI. All you need to do is give circuitpython users an option skip selecting a folder, and then show those buttons, and it's ready... except for the broken RTS/DTR issue on some boards of course)
tulip sleet
(p.s. "code.circuitpython.org" already has the feature you are talking about. Y...
where is the "micropython" mode button you are talking about on code.circuitpython.org? Could you show me a screenshot?
i am talking about https://code.circuitpython.org
alpine holly
LOL - SORRY - I got "Mu" mixed up with that (too many new tools in one day!!). ping me if you need help - I've re-written ampy, mpremote, and pyboard for those things (and made ampy compatible with circuitpython today) as well as written my own emulator too... I've got a fair whack of experience with the ugly mess that is serial management of these boards. They're horribly dumb, and time-out when anything unexpected or unusual happens... and I fixed all of that.
Do you happen to knw what "C" lint or formatter is being used? It's blocking my PR inside a .h file, and the messages don't give me any clue as to what program they used to do that.
The "check-your-code" link only mentions python and "black" - but "black" simply exits with no message (so I assume it's happy?)
tulip sleet
Do you happen to knw what "C" lint or formatter is being used? It's blocking my...
it is uncrustify. And you have to use a specific version. black is for python code
micropython uses uncrustify, and since we are downstream, we adopted their conventions
they are switching to ruff for Python code; we will probably go that direction eventually
tulip sleet
LOL - *SORRY* - I got "Mu" mixed up with that (too many new tools in one day!!)....
ok, now I understand. We no longer maintain ampy. Did you submit a PR to micropython for mpremote?
ugly mess that is serial management of these boards
🙂 yes, that's what I was despariing
alpine holly
Yes, I submitted PRs for mpremote, ampy, pyboard, and python itself.
Unrelated - I really don't want to mess this up: I've got uncrustify working and "fixed" (ruined the readability by erasing the indent... but that's another story...) my header file. What are the steps to fix the PR? Do I do another "git add" and "git commit" and finally a push? I'm guessing it's more complicated - do I need to withdraw the PR first, or something ?
tulip sleet
Unrelated - I **really** don't want to mess this up: I've got uncrustify working...
Yes, add and commit and push
it's fine to have multiple serial commits to a PR. That's part of the point.
a PR doesn't have to be perfect when it's first submitted
We have a guide on the process: https://learn.adafruit.com/contribute-to-circuitpython-with-git-and-github
by the way, the indentation change is because you were using tabs for indentation. Since there are multiple standards for how many spaces a tab is, we use spaces.
alpine holly
All checks have passed - thanks HEAPS @tulip sleet holding my hand through that jungle was a HUGE help.
I've got a stack of about 10 new ESP32 boards here (mostly different sized displays, but a couple of camera boards too, plus some bare chips inside programming holders)... I'd like to get a production-line going, so I can push out support for all these things. LVGL is too bloaty, and I've already got them working in micropython (which has some weird complex-number bugs)... sorting out that lint nightmare has totally made my day!
manic glacierBOT
For what it's worth - here's my "notes" (a cobbled-together copy/paste from assorted places which seems correct):
alpine holly
I'm reluctant to introduce a whole pile of non-standard non-portable aliases simply to drop the leading "G" off everything... while I can see where you're going with that, I would very much prefer NOT to encourage that kind of thing. I'm working currently on the ESP32-WROVER-DEV board which (minus the SD card) is almost identical. Code written for either will run fine, unchanged, on the other... so long as people don't use silly pin names. The latter drops the "IO" prefix on its silkscreen, which will add a whole extra level of extreme confusion if I go down this route ... { MP_ROM_QSTR(MP_QSTR_16), MP_ROM_PTR(&pin_GPIO16) will resolve to an int instead of a string in those cases...
That said - I'm happy to keep you happy. This is my first time doing this stuff.
tulip sleet
@alpine holly maybe you have seen this already: https://learn.adafruit.com/how-to-add-a-new-board-to-circuitpython. Includes discussion of pin names.
alpine holly
Excellent reference - that suggests that my hunch is what they want: "Try to make the names correspond closely enough to the board silkscreen that it is self-evident which pins correspond to which silkscreen markings. It's fine to expand abbreviations, such as LED for L. You don't have to match the silkscreen exactly, since the silkscreen names are often constrained by the space available." ... I think "GPIO16" is exactly what "it's fine to expand abbreviations" is talking about (since it';s clear the silkscreen simply left off the "G" because there's no room to fit that there) https://learn.adafruit.com/how-to-add-a-new-board-to-circuitpython/pin-and-device-names
wraith crow
I'm just sharing some pin naming hurdles I've encountered, don't worry about making me happy 😁 It seems like you've considered the issues. Just one more thought, I believe a lot of the ESP32 boards do use the IOxx pin naming standard (I'm not home to confirm that right now) so porting code between this board and other ESP32 boards may be impacted by your naming choice
manic glacierBOT
Be aware of a common hardware fault on some of these - the power regulators are backwards ...
You may need to swap the 2 power regulators (65BX and 54FK) which got soldered in the wrong way around! See pic - cheap copy at top, original at bottom :-
alpine holly
@wraith crow - yeah - you utterly nailed that one!!!
[cnd@cdc:~/.../ports/espressif/boards]$ grep -r -l -P 'MP_QSTR_IO\d' . | wc
110 110 3550
[cnd@cdc:~/.../ports/espressif/boards]$ grep -r -l -P 'MP_QSTR_GPIO\d' . | wc
4 4 114
I'll fix this tomorrow - it's 4am, I've been going for 20hrs nonstop and I'm getting vague...
alpine holly
nah.. did it now, before I forget, so it's in a state ready to be accepted if someone like it.
tulip sleet
Excellent reference - that suggests that my hunch is what they want: "Try to ma...
From the guide also:
GPIOn pin names are typically abbreviated as IOn on Espressif boards.
I wrote that page
alpine holly
@tulip sleet (you wrote in a different channel) ...or remove OTA instead... There's (0x7000-17568) 11,104 spare bytes after the end of the bootloader on ESP32 right now. I would love to try my hand at at a new kind of OTA which facilitates updating a running system without needing any OTA partition by (a) first storing the new firmware inside the existing VFS or SD file system, (b) halting circuitpython and switching the a new OTA stub in the bootloader which (c) SWAPS the current firmware partition with the new new one (i.e. overwrites (a) from the old firmware as it writes to flash), (d) sets some flags in the new OTA stub, (e) reboots into a new "firstboot" stub which updates those flags and sets a watchdog timer then attempts to run the freshly flashed firmware, which itself will update those flags and disable the watchdog when the boot works, and (f) the watchdog OTA stub which (if called) loops back to (b) in order to replace the new broken firmware with the original working one. new firmware can be stored inside the VFS (if the user makes enough space for it) or on an SD card (potentially allowing many other ones to exist at the same time - featurecreep - grub-like UI to let people pick...). In very-low-VFS-space-situations, some kind of web-helper to temporarily "move" VFS things to some cloud the the update, then replace them upon workingness might be nice. it might even be possible to code a standalone VFS binary that's far smaller than the existing circuitpython firmware which would be callable form the OTA and be able to connect wifi or ble and manage the firmware update using the same process, but using "cloud" (or some BLE or phone app) instead of VFS for holding the actual image ... so many cool possibilities. I got a long way down this planning path and started on experiments in micropython back in January (before mpy' instability made me realize that's unsuitable for my production use...)
hmm. the same idea would facilitate an "undo" option that does not use up any extra time for users when they're flashing locally as well (like over serial or from the web flashing UI in chrome etc).
manic glacierBOT
manic glacierBOT
alpine holly
Do we have to do something to request a review of our PR's ? e.g. https://github.com/adafruit/circuitpython/pull/9233
austere acorn
Do we have to do something to request a review of our PR's ? e.g. https://githu...
Yes. You have to try to be patient.
manic glacierBOT
alpine holly
board-porting question: How do the LCD drivers get into the firmware? I purchased one-each of every one of these eps32 LCD boards ( https://www.aliexpress.com/store/1100192306 )... 11 in total. They mostly use different LCD drivers ( EK9716, GC9A01, ILI9341, ILI9341V, ILI9485, ILI9485, ILI9488, St7262, ST7735S-G6, ST7789, ST7796 ) There's 3 already board ports for these done (sunton_esp32_2424S012 sunton_esp32_8048S070 sunton_esp32_2432S028) which have EK9716, GC9A01, or ILI9341 LCD drivers in them. Looking at boards/sunton_esp32_2432S028/* reveals no mention at all of the driver (besides a comment) and it implements only display_init() - yet somehow the compiled firmware works fine (show shows the REPL on the screen at boot), and I can draw circles etc using board.DISPLAY ... how did the plumbing work that out ??? I firstly want to fix the ridiculously slow scrolling ( ILI9341 has hardware assist for this, which is clearly not being used), and then build a port for all those other boards.
stuck elbow
board-porting question: How do the LCD drivers get into the firmware? I purchas...
they only differ in the init sequence, if you look at the python drivers you will see that they also only have init
alpine holly
interesting! So the SPI commands to set pixels, clear screen, scroll, etc etc are identical, regardless of which LCD you're using ? Seems slightly implausible?
I notice my ILI9341 (320x240) scrolls crazy-slow, but my EK9716 (800x480) is really fast, and I notice the latter scrolls sideways even faster... suggesting there's some hardware support for scroll in only one of those drivers ?
(by "scroll" - I mean how the REPL shows up when I boot and upload files etc over serial)
midnight ember
interesting! So the SPI commands to set pixels, clear screen, scroll, etc etc a...
Some drivers do not read/write registers to tie into any hardware acceleration the driver chip might have. This does vary display to display. The init sequences for displays are the basic data commands for writing to the display. The init sequences are different for each display but yes the way displayio uses them remains consistent across displays that use displayio. That's one of the advantages of displayio is consistency across displays. Some drivers are more refined than others so you might see different performance depending on the driver setup. There are tons of features a display might have that are never tapped into. Compatibility with displayio is prioritized over fine tuned performance per display. Even if hardware acceleration is available on a display driver chip doesn't necessarily mean it's being used. Someone can correct me if I'm wrong but this is my current understanding of displayio init sequences vs driver libraries that require register read/write.
short tendon
@tulip sleet with the ConnectionManager updates, we lost the Error connecting socket: error 2, first error: error 1. Are we okay with that? I haven't seen a case where the two are different, but I can imagine the case where it's a MemoryError the first time and then a different one the second...
tulip sleet
<@329766224093249548> with the `ConnectionManager` updates, we lost the `Error c...
I am OK with that. And it wasn't handling SSL wrapper recovery: that would have caused two different errors if the second one failed. But the second error is the one that is interesting. The MemoryError is an internal thing that is not really an error: it's an out-of-sockets thing.
rather than put it in the exception, I think adding some logging at some point would be more helpful when debugging a problem. e.g. "This succeeded; this succeeded; this failed."
short tendon
I am OK with that. And it wasn't handling SSL wrapper recovery: that would have ...
Awesome. Will update the test accordingly and push it up in the morning (it's a beautiful day and not spending anymore time inside)
short tendon
@tulip sleet also sorry for the double ping, but wanted to make sure you saw this comment: https://github.com/adafruit/Adafruit_CircuitPython_MiniMQTT/pull/212#issuecomment-2106351165
lone sandalBOT
slender iron
<@&356864093652516868> Reminder: Our community meeting is in 7 minutes or so. Please get any notes in you'd like me to read. Talk to you soon!
midnight ember
Will there be a live stream or recording?
onyx hinge
@midnight ember pycon talks are posted later in the year, at least have been in past years
turbid radish
midnight ember
uh oh no mic 😦
The struggle with pylint typing is real.
S2 can make multiple SSL connections now?
🙌
lone axle
This file is part of the CircuitPython project, http://circuitpython.org/ Sees like a good line to me, I can't think of anything to change with that.
Reuse (which is run by pre-commit for us) validates those SPDX headers for us on library repos.
onyx hinge
>>> math.acos(2)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ValueError: math domain error
```value error: 2 is not within the set of proper arguments for acos
```py
>>> math.abcdef
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
AttributeError: module 'math' has no attribute 'abcdef'
```attribute error: there's no attribute (property) on the object with this nameThank you for hosting @slender iron Have a wonderful week everyone!
gilded cradle
Thanks
onyx hinge
👋
short tendon
Thanks!
lone axle
Thanks for hosting Scott. Have a great week everyone 👋
slender iron
Thank you for hosting <@252717193496756235> Have a wonderful week everyone!
Did you consider using dotclock support instead of the ra chip?
midnight ember
No... I thought a display of that size would require a bigger driver chip? 🤦
slender iron
I don't think so. The ESP32-S3 should be able to do it directly via dotclock iirc
midnight ember
I would need to figure out the 40-pin connector directly to an MCU. With 100 pins wouldn't even know where to begin. Only reason I'm taking this approach is Lady Ada's RA8875 is a proven design.
midnight ember
I do like the idea of having the hardware accelerator for some uses too. It really is like 100x faster at drawing rectangles with the hardware accelerator.
slender iron
I believe that
midnight ember
Ahh, thank you will look into that!
I believe the qualia boards 40-pin isn't compatible with the 7" display. I would still have to wire it from scratch or make a qualia-like board specifically for the 7" display.
They're both 40 pin but in different orders.
in the ra8875 product page There are other 40-pin displays that have different pinouts or backlight management and these may not work - they may even damage the driver or TFT if the boost converter pushes 24V into the display logic pins! For that reason, we only recommend the displays we've tested and sell here.
slender iron
👍 It is good to check the pinout
onyx hinge
yeah that particular combo doesn't seem to be compatible
midnight ember
yup that's where i'm going to right now to pull up a side by side schematic.
onyx hinge
the great thing about standards 😕
midnight ember
can you post up next weeks notes doc so i can go ahead and give you an early hug report for your advice? 🤗
tulip sleet
S2 can make multiple SSL connections now?
Please try it on something that wasn't working before, with the latest ConnectionManager library.
slender iron
can you post up next weeks notes doc so i can go ahead and give you an early hug...
working on it 🙂
Here is the notes document for next Monday’s CircuitPython Weekly meeting. It is at the normal time of 11am Pacific / 2pm Eastern here on Discord. Everyone is encouraged to attend! Please add your hug reports and status updates even if you’ll be attending the meeting - it’s super helpful! If you are unable to attend but would still like to include updates, feel free to include them in the notes and we’ll read them off during the meeting. Hope to see you there! <@&356864093652516868> https://docs.google.com/document/d/19GmXK8VzVgchFclJTfG4C6ejqNsg9emYgCgpcv1PVZ4/edit?usp=sharing
midnight ember
yeah the dotclock qualia 40 pin is definitely different from the ra8875 40 pin. the 24V backlite pinouts are the same so it won't toast it with 24v thankfully but enough pins are different so that all communication would be broken. incompatible. will have to decide which route to take. i'll stick with the ra8875 for the hardware accelerators.
i've drawn a full 800x480 bmp on the display without hardware accelerators and it takes about 3 mins to draw. even though the s3 can dotclock it directly a display of that size would still be too slow for any appreciable use with displayio.
For a weather display that only updates every 15 minutes it's doable so it really depends on the usage. Because I only need slow redraws it's very tempting and I will likely design another PCB in the future to try it out.
midnight ember
Please try it on something that wasn't working before, with the latest Connectio...
I do have a project that wasn't working correctly before so I can definitely help beta test it.
short tendon
@tulip sleet what are your thoughts on this idea:
adding an empty class CPythonRadio: to connection manager, and then when passed in as a radio to get_radio_socketpool and get_radio_ssl_context it would return the native socket and ssl.create_default_context(), allowing connection manager to be used in CPython?
tulip sleet
<@329766224093249548> what are your thoughts on this idea:
adding an empty `cla...
In terms of making examples run everyhwere, sounds like a fine idea, and it's small. Maybe CPythonNetwork, because it's not necessarily wifi.
short tendon
In terms of making examples run everyhwere, sounds like a fine idea, and it's sm...
100%
slender iron
@tulip sleet do you want to take a stab at the asyncio microdot server stuff? I might be able to fix it
(if not)
I've been poking asyncio more with cpython
tulip sleet
if you'd like, sure. I don't know if there's something missing internally that it needs (stream support or something), but the naming stuff seems pretty easy to fix
i am working on redoing headers and include guards at this very moment
slender iron
I can take a stab at it tomorrow when I'm off and on
slender iron
@short tendon this may be a place to use connectionmanager in asyncio: https://github.com/adafruit/Adafruit_CircuitPython_asyncio/issues/35
short tendon
<@1152454697614069811> this may be a place to use connectionmanager in asyncio: ...
It sure looks like it
I can work on that this week
tiny peak
@slender iron remember how you made a comment during one of your recent Deep Dive streams about a patch-panel MIDI controller? I liked that idea and started planning some stuff out last week. Ordered some parts. Got some of them. Still waiting on others. Might need to order more. In the mean time, I've been planning out the software. Been taking some notes on github (patch-thing repo). If it goes well, I'll probably write up a guide on the new Playgrounds thing. Also possible I might be able to generalize some of it into a patch-panel module.
midnight ember
That sounds neat, I'd read that.
manic glacierBOT
manic glacierBOT
small comet
https://octodon.social/@stylus/112411930220861884
Best of luck with the talk 🙂 (would love to see any slides when you're able to share them.)
manic glacierBOT
onyx hinge
Best of luck with the talk 🙂 (would love to see any slides when you're able to ...
Thanks! I think that at some point a link to my slides will be added on the talk page, https://us.pycon.org/2024/schedule/presentation/15/ -- if not, I'll share them as a google doc after the conference.
manic glacierBOT
slender iron
I can work on that this week
thanks!
slender iron
<@252717193496756235> remember how you made a comment during one of your recent ...
nice! let me look
I was thinking the 1/8th inch cables would carry uart signals that carry an id number
so it can work across "modules" and be plugged one to many as well
orchid basinBOT
lone sandalBOT
devout jolt
I have sort of a general Python question. (this in the context of PyUSB compatibilty that usb.core is doing that @slender iron & @onyx hinge were talking about in issue #9226): Are exceptions not "expensive" in Python? Are they more just an alternative return value?
My previous experience with exceptions in other languages is they're pretty expensive (blow stack frames, kill pipelining, etc) and meant to be used for substantive errors and not for expected logic flow. But in the underlying use case needed to make #9226 work, a try/except block will be in the core read() loop because most of the time usb.core.Device will be throwing timeout exceptions
tulip sleet
I have sort of a general Python question. (this in the context of PyUSB compati...
raising an exception creates an exception object. The other parts are not all that expensive. The raising is done by nlr jumping usually
stuck elbow
setting up a try block costs a bit
devout jolt
okay so it sounds like (Circuit)Python exceptions are not the 100x as expensive as return values like in C++ & Java
tulip sleet
I would say not
stuck elbow
return values are practically free, especially if the function is inlined
what costs is the if branch that checks for the error
tulip sleet
i would say try a simple test of doing something simple that raises an exception over and over vs jsut returns. Check the timing and check the storage consumption with gc. Are you saying you are going to busy-wait on this with no delay?
The point of using the exception is to mimic pyusb, which I am not really familiar with.
devout jolt
i would say try a simple test of doing something simple that raises an exception...
The current implemention of usb.core.Device.read(), which mimics PyUSB, is to throw a timeout exception if a timeout is specified and there's nothing to read. There is usually nothing to read. Just like when waiting for UART read()s. But UART reads do not throw exceptions when there's nothing there
stuck elbow
catching an exception involves checking its inheritance chain to see if it qualifies for the given except block, so if you have a lot of nested try-except blocks, things can get expensive
devout jolt
(and we usually want to use a timeout on the read so we can do otherthings with our code like read buttons, update display, etc)
tulip sleet
so are you going to busy-wait on that?
devout jolt
so are you going to busy-wait on that?
no because we want to do other things in our code. otherwise the read() blocks
tulip sleet
then if you are going to call read() say 10-100 times second, maybe it's not an issue
I agree that maybe pyusb's choice is not the greatest
devout jolt
This seems to be not an issue as much in PyUSB for whatever reason. I cannot find many examples that do reads-with-timeout to compare against
tulip sleet
are you seeing a problem now, or is this a hypothetical?
right now the problem is the bug
devout jolt
it's a problem in that in #9226, the underlying max3421e driver appears to not implement read-with-timeout and the underlying TinyUSB call crashes. The PIO-USB host driver on RP2040 works as expected and throws the timeout exception. And that got me wondering if I'll be loading up another footgun and killing performance by doing read(timeout=0.001) like I do with UARTs, because I've been trained by decades of C++ & Java that exceptions are very expensive
tulip sleet
it's a problem in that in #9226, the underlying `max3421e` driver appears to not...
I think the way to confirm or allay your fears is to do measurement. We don't think of exceptions as very expensive, and we don't go out of our way not to raise them. There is some philosophy about exceptions in Python: "easier to ask forgiveness than permission" https://stackoverflow.com/a/66792207/142996.
But using exceptions to substitute for normal return values is not great API design. I once designed an API like that when I first started using C++, and it was a mistake.
devout jolt
I think the way to confirm or allay your fears is to do measurement. We don't th...
Thanks that's interesting. Since I came into Python mostly via CircuitPython, I'm missing some of this philosophy. I'm definitely going to make a little synthetic benchmark to see how how costly raises are vs returns too.
devout jolt
I think the way to confirm or allay your fears is to do measurement. We don't th...
I made this synthetic benchmark: https://gist.github.com/todbot/6a3c8e5633a7267465775aa3afda99fd and the results on both CircuitPython and CPython are that exceptions cost about 33% more in time. In CircuitPython, where I can measure it, exceptions cost about 46% more in bytes.
tulip sleet
I made this synthetic benchmark: https://gist.github.com/todbot/6a3c8e5633a72674...
Interesting results. If you do a gc.collect() when the run is done I think most of the memory difference will go away. You're seeing garbage that needs to be collected.
devout jolt
Interesting results. If you do a `gc.collect() `when the run is done I think mos...
yep
slender iron
Interesting results. If you do a `gc.collect() `when the run is done I think mos...
I wonder if those allocations are costing cpu time too
@tulip sleet new header looks good to me too. I'd just double check that reuse is ok with it
tiny peak
I was thinking the 1/8th inch cables would carry uart signals that carry an id n...
hmm... yeah, I've been wondering about how to deal with expansion to more patch points. The UART thing might work. Have you ever messed with 1-wire stuff? It might work to solder TO-92 or SOT-23 1-wire EEPROMs directly onto "output" jacks, and then use 1-wire scanning on "input" jacks (or vice versa). Seems like 1-wire EEPROMs are pretty cheap and relatively robust against ESD. Might save on a lot of soldering and GPIO pins. 1-wire parts have serials, the bus can handle multiple parts (e.g. piggyback cables or splitters), plus, the EEPROM would allow for custom tagging.
I'm not sure how it would work to do piggyback connections (n outputs to 1 input) with regular TTL async serial. To handle multiple patch points joined together, seems like you might need pullups with open-drain transceivers, plus some kind of protocol to handle contention and addressing? (e.g. 1-wire)
slender iron
I'm not sure how it would work to do piggyback connections (n outputs to 1 input...
I'd only do one output to multiple inputs. I would do open drain to ensure no double output failures
I was thinking of using a cheap stm32 micro for output and input
then you'd read the input ids via i2c
and maybe set the output ids
slender iron
I was thinking "per module"
i2c would be the backplane via stemma connectors
the trs would be just gnd and uart
I may be getting ahead of myself 🙂
tiny peak
oh, so kinda like how the attiny seesaw breakouts work? https://www.adafruit.com/product/5681
slender iron
yup
tiny peak
I could see that. might work.
slender iron
https://github.com/tannewt/StemmaG0 stm32s have i2c bootloaders in rom so I was going that way
tiny peak
is that the thing you were talking about in relation to some kind of dynamic compilation of IO coprocessing tasks?
Can't remember how you phrased it, but I think you mentioned the StemmaG0 thing in relation to build system stuff
slender iron
ya, that's the hardware where a library would load the c binary onto it
the build system would theoretically make it easier to create those libraries
tiny peak
interesting. I'd been thinking of a small scale thing (<= 16 GPIO pins) where there would be maybe 8 outs, 8 ins, and you'd just raise one of the outs, scan the ins, then repeat that for each of the other outs.
It wouldn't expand well, but, you'd end up with a byte for each ouput, which would translate pretty nicely to sending as a MIDI CC value (sorta... only 7 bits to work with for MIDI CC)
For patch points, if you start getting into n:m mappings for n or m > 7, then it becomes potentially a lot more challenging to translate the graph of patch cables into something that can be represented as MIDI CC.
short tendon
@tulip sleet tell me if this is too much.
I love tests. I love coverage. Many of the CP libraries can't really be tested without a lot of mocking.
For something like MiniMQTT, we could have the tests spin up a mqtt server and test most methods for real. Is this something that would be interesting?
tulip sleet
<@329766224093249548> tell me if this is too much.
I love tests. I love coverag...
I think that's a fine idea. Consult with brentr about that maybe -- he might have suggestions about which server to use, etc.
midnight ember
@tulip sleet Tested ESP32-S2 Feather 4NR2 (no bme 280) with multiple sequential SSL connections successfully with the latest updates. Installed new 9.0.4 stable and 9.x bundle from today. Is there an open issue to confirm? I have one open for the S2 which can now be closed and I imagine a lot of S2 socket/SSL errors are now fixed. Very nice!
tulip sleet
<@329766224093249548> Tested ESP32-S2 Feather 4NR2 (no bme 280) with multiple se...
I closed https://github.com/adafruit/circuitpython/issues/9219. Which one did you have in mind?
I am glad this was a simple fix -- it wsn't something internal
midnight ember
The one I opened to track this issue is https://github.com/adafruit/circuitpython/issues/9115?
I think there are many S2 bug reports that will now be cleaned up.
midnight ember
Will leave it up to you or Scott to close. You've earned it. Well done. 👏
midnight ember
S2 is back on the menu.
manic glacierBOT
lone axle
Is it possible to customize the Bootloader section for devices on circuitpython.org page?
I've noticed some devices have different instructions for accessing bootloader (M5stack cardputer is press/release reset once, then press / release boot btn once. Double press reset does not work, and hold boot + press reset goes to ROM bootloader not UF2.). I'm pretty sure I've run across other devices like this before too, but don't recall which ones specifically.
I looked in https://raw.githubusercontent.com/adafruit/circuitpython-org/main/_board/m5stack_cardputer.md but the bootloader section content doesn't appear in here.
tulip sleet
Is it possible to customize the Bootloader section for devices on circuitpython....
it's configurable by port right now (that's all mushed into the bootloader text file now), but I don't know that we've done it by board. There could be an include or something that came from the board directory. Sounds like an issue for the circuitpython-org repo.
This gets into how much doc to provide and where for third-party boards where the mfr is not doing support for CIrcuitPython.
lone axle
I do agree it's tricky for 3rd party devices that don't use the defaults. I think it'd be great to at least have a mechanism that is a boolean "uses default bootloader behavior" flag that can be switched to false for ones that don't follow the norm so that the page at least won't contain incorrect instructions.
But having a way to for people to submit PRs that change or include updates to the text on a per board basis would be best IMO, then anyone who notices can submit a change with the correct info for that device.
manic glacierBOT
slender iron
For patch points, if you start getting into n:m mappings for n or m > 7, then it...
I didn't think so far as MIDI CC. I was mostly thinking synthio
slender iron
@devout jolt want to make a PR to update tinyusb?
tiny peak
I didn't think so far as MIDI CC. I was mostly thinking synthio
yeah, if you wanted to do all the synthesis in CircuitPython, handling more patch points would probably be easier compared to MIDI CC. That, or perhaps using a network stack to send OSC messages. For me, the interesting thing would be to have a USB MIDI controller that lets me play with Faust or SuperCollider patches in a way that's more tactile.
slender iron
yeah, if you wanted to do all the synthesis in CircuitPython, handling more patc...
Do they have an API for setting patch connections?
orchid basinBOT
tiny peak
Do they have an API for setting patch connections?
SuperCollider lets you write arbitrary code to implement event handlers for incoming CC messages. Faust is more about functional DSP data flows, so it lets you attach CC control channels to numeric UI elements (e.g. hslider), which you can then use in functional declarations to do whatever you want (subject to your ability to articulate what that is in functional style)
In my limited experience so far, playing with CC controls in Pd, Faust, and SuperCollider, implementing patchboard style functionality is not easy. The limiting factor for me is keeping straight in my head what's going on. For small numbers of controls, CC is convenient. But, particularly with Pd's graphical box and wire data flow stuff, things get really complex, really fast, as you start trying to scale up the number of controls.
slender iron
I believe it!
My brain was thinking only about virtual eurorack sims where you could make physical equivalents of the front panel for hopefully less money than the original
back in a bit, going for a short walk
slender iron
(back)
tiny peak
(back)
I think I'm gonna detour for a bit to try a breadboard prototype of a MIDI CC to synth patch thing
slender iron
I have this tab open as a reminder to get back to it: https://jlcpcb.com/partdetail/ShouHan-PJ242/C431536
tiny peak
I've got a Feather M4 on hand. If I leave A0-A5, SCL, SDA, and 13 free, that gives me 6 pins on each side for making a tiny little breadboard and DuPont wire patch panel. I want to see if I can get something useful happening with a 6:6 routing matrix on the Feather talking MIDI CC to Pd or Faust.
midnight ember
@short tendon I upgraded my main project (feather weather) with 9.0.4 stable and uses minimqtt. Everything working fine so far. I do have a lot of error handling so I think that issue report is just a matter of the person not having error handlers. In an ideal world error handlers shouldn't be necessary because they're beyond the ability of beginners. That would be a nice goal to set. I would love to be able to rip out all error handling from my script and just have it work.
I can get it to manually republish by switching pages then back to the home screen. Confirmed it's publishing to AdafruitIO broker.
short tendon
<@1152454697614069811> I upgraded my main project (feather weather) with 9.0.4 s...
Quite possible. Just trying to get resolotion so we can close them out...
midnight ember
Yeah, just wanted to let you know there's no problems that cropped up with a known good working minimqtt script updating to 9.0.4 and latest connection manager.
tiny peak
@slender iron I got a little breadboard patch-panel thing working. Code, wiring chart, Faust dsp patch, and preliminary documentation are up at https://github.com/samblenny/patch-thing/blob/main/notes/06_prototype/README.md
The code is a bit involved because I switch back and forth from everything pulled low (drain stray charge) to most the pins in high-z (avoid unintentional resistor networks making voltage dividers). I've got it wired up with 10k resistors on all the ins and outs as a cheap and dirty attempt of ESD and short protection. I think with the resistance that high, perhaps the capacitance of the DuPont wires and breadboard actually matter? Anyhow, the current code seems to work pretty well for mapping connections without glitching. Touching one end of the DuPont wire seems okay. Hooking up more than one input to the same output seems okay.
The Faust DSP patch doesn't do much other than just displaying the incoming CC values and using them to adjust the volume of some pink noise
...but, I think this breadboard circuit and code.py should be enough to enable some experimentation with better software synth patches that might hopefully kind-sorta resemble a modular synth setup.
manic glacierBOT
slender iron
...but, I think this breadboard circuit and code.py should be enough to enable s...
Nice! You are being more thorough than me
orchid basinBOT
slender iron
@devout jolt are you looking into the timeout issue?
manic glacierBOT
tulip sleet
@slender iron how are you running an older vesion of uncrustify on Arch? Did you build it yourself? The ubuntu 24.04 version is too new (lots of config file changes). pybricks repo has no builds for 24.04.
I see there is an older version in an Arch archive. I may just build it myself. And will contact the pybricks maintainer.
slender iron
<@252717193496756235> how are you running an older vesion of uncrustify on Arch?...
I have the old version and don't upgrade. I'd love to update to a newer version (or switch to clang-format)
tulip sleet
I just posted this: https://github.com/orgs/micropython/discussions/14496
for now I will just build it, I guess
oh wow, https://github.com/uncrustify/uncrustify doesn't even have tagged versions before 0.73
plucky tulip
Does circup support installing .py versions of libraries? I cannot find anything about it.
slender iron
Does circup support installing .py versions of libraries? I cannot find anything...
I think it is --py
it definitely does
plucky tulip
I think it is `--py`
Oh my god, you are right, I didnt see the option under circup --help, but it is present under circupt install --help. I'm sorry, it's 1am where I am at. Maybe enough for today...
Thanks👍
manic glacierBOT
manic glacierBOT
short tendon
@tulip sleet do you have a rough idea on when CP 9.1.0 will release? I wanted to try to get the AttributeError/ValueError fixes in requests before it does so it would be frozen in
tulip sleet
<@329766224093249548> do you have a rough idea on when `CP 9.1.0` will release? ...
it's not imminent. I'd say at least a couple more weeks, and probably longer.
empty salmon
@slender iron - I committed the change. Do I need to resubmit the PR?
slender iron
nope, if you commit through the UI then it'll update
PR's track the state of a branch
slender iron
@tulip sleet want me to merge your change? it looks ok to me
tulip sleet
<@329766224093249548> want me to merge your change? it looks ok to me
I was just eating dinner. I just merged. I'll update your files in another PR. I have a simple way of checking for files that need updating.
manic glacierBOT
alpine holly
Anyone able to shed light on what might be causing this build-check fail? https://github.com/adafruit/circuitpython-org/actions/runs/9137944914/job/25128516373?pr=1408 - is it me, or is the build system broken ?
vocal bridge
Anyone able to shed light on what might be causing this build-check fail? https...
I think it might be because you have an extra blank line in your YAML header: https://github.com/adafruit/circuitpython-org/pull/1408/files#diff-2e836a8a0947572aa502a38586111cfcc2f7a4e19d9aa5627b9d4c90ca9fdf39R17
alpine holly
Possibly not - the template we're told to copy from, "unknown.md", and many other boards also do as well?
vocal bridge
I only checked a few but they ones I saw didn't have it
Ah, there is another one that has it https://github.com/adafruit/circuitpython-org/blob/main/_board/raspberrypi_pi4b.md?plain=1
alpine holly
Yup - failed despite fixing that.
Any idea how/where I can look at the code that's running that check? Someplace there must be defined a file of acceptable features, but I can't find it...
vocal bridge
This is the code that is failing: https://github.com/adafruit/circuitpython-org/blob/main/tools/check-boards.py
Do you need to have 2 spaces before each - in the list of board URLs?
alpine holly
Thanks for the clues... I Found it: - I forgot to close the quote on the end of the board name: name: "Ai Thinker ESP32-CAM
vocal bridge
Aha
It's so easy to miss stuff like that
I once spent the better part of a day chasing down a :/; typo
Oh yeah, also it looks like the other files in that directory are named with _ but yours is with -
alpine holly
I agonized over the _ or - issue myself; all the other ai-thinker boards use hyphens in their build folder names, and most places online do that as well, so that's what I picked; but yes - I did (later) notice that most of the other ports don't do that.
manic glacierBOT
tulip sleet
I agonized over the _ or - issue myself; all the other ai-thinker boards use hyp...
in general we would prefer underscores, all lower case, mfr name first
when I review PR's, if I notice, I steer people in that direction
manic glacierBOT
short tendon
@tulip sleet can I lower the pystack value to cause it to error?
tulip sleet
Let’s wait and see what the person with the issue says about reproducing.
alpine rapids
Hi again! Question for y'all: I'm working on adding this board, an ESP32-S3, and I'm unsure about the USB PID/VID thing - I've requested one from espressif in the past and understand that process, but is that needed for this specific board? The reason I ask is because the board doesn't have a typical USB port, however I believe native USB is still supported when using the usb add-on. From the factory it uses a vid/pid of 303a:1001, which is a generic ESP32 identifier. Should I still request a fresh one from espressif for circuitpython?
ornate breach
Hi again! Question for y'all: I'm working on adding this board, an ESP32-S3, and...
For custom board definitions, it’s going to require a VID/PID unique to it.
urban knoll
Hey, wondering if y'all have a feeling for which CP ports have the most users, vs. ones that are less active? Want to focus on where the action is
ornate breach
Hey, wondering if y'all have a feeling for which CP ports have the most users, v...
Last statistic I saw was the RPi Pico (RP2040) had the most downloads
Not sure about least though
urban knoll
Are those stats accessible somewhere?
ornate breach
I’m not certain, I think Scott pulled them from aws or the circuitpython.org backend
devout jolt
Are those stats accessible somewhere?
I believe the list at https://circuitpython.org/downloads is displayed in order of download popularity, and I thought there was a repo for the circuitpython.org website (which might have the script or data to sort that list), but I can't find it right now
midnight ember
Just me or does that ribbon cable look like it's going right on top of the S3? With such a tiny board, the S3 might melt that cable. Looks like they assembled the ribbon connector backwards. :/
vocal bridge
I believe it is sorted by download count: https://github.com/adafruit/circuitpython-org/blob/main/downloads.html#L96
manic glacierBOT
tulip sleet
@urban knoll @ornate breach @devout jolt @midnight ember Yes, the boards are sorted by download count. The download counts are in here: https://github.com/adafruit/circuitpython-org/blob/main/_data/files.json. When a release is made, we generate download counts by doing a query on the AWS download logs. The release CI script generates a automated PR to circuitpython-org with zero downloads. Then we (usually I) run a small script that updates the download counts, push that commit, and merge the release PR.
Right now the query counts 8.x and 9.x releases. As the releases progress, I change the query slightly. I also clean up the logs from time to time. I think it's 2023-present right now.
urban knoll
I believe it is sorted by download count: https://github.com/adafruit/circuitpyt...
Ah, handy, thanks
burnt tangle
@danh thanks for the data. Here's what it looks like.
I do have some questions though, for instance how does it make sense that there are over 3000 downloads for https://circuitpython.org/board/trinket_m0_haxpress/?
stuck elbow
bots
burnt tangle
Yeah probably can't read much into any but the big numbers. There's a very long tail
tulip sleet
yes, don't take this as anything all that indicative of popularity
devout jolt
@danh thanks for the data. Here's what it looks like.
Great chart! I’m pleasantly surprised to see nRF so high
manic glacierBOT
short tendon
@crimson ferry when you have a moment, I would be curious your thoughts on this: https://gist.github.com/justmobilize/f236f332fa798fce520d01154b624f5e
I haven't fully tested it, but threw this together as a thought I've had in the back of my head.
one could just:
radio = get_radio()
pool = adafruit_connection_manager.get_radio_socketpool(radio)
ssl_context = adafruit_connection_manager.get_radio_ssl_context(radio)
requests = adafruit_requests.Session(pool, ssl_context)
And not need to think about anything else.
It defaults to FeatherWing pins, but one could set then real pins in settings.toml...
Regardless, I'm going to finish it up and use it (it's from a few helpers I already use)
crimson ferry
@short tendon Looks nice, helps to encapsulate the often-messy non-native radio setup. Are you envisioning (people) using old versions of ESP32SPI (there's a bit of code associated with those conditionals)? But yeah, pins get to be the tricky part for ESP32SPI and Ethernet, with so many form factors for those peripherals.
You and I already talked about the idea of WIZnet init not requiring a network connection, so that could obviate all the connect_needed checks, and connect then happens regardless of radio type.
And if we design each connect() to be negligible processing time if already connected (like for native wifi), there's not a lot of need for some of that checking.
Some of the other init kwargs add complexity too for some projects... MAC address, static IP, DHCP, debug, etc.
short tendon
<@1152454697614069811> Looks nice, helps to encapsulate the often-messy non-nati...
So, I thought about requiring CP9.1.0, because until that's released the esp32spi frozen library doesn't have them (and I've been doing a lot of pyportal help...).
Right now, this is very much a draft/proof of concept and will get cleaned up.
Ideally I'd like to require/warn for an old Nina firmware as well
It might be worth bringing up / drafting a PR for the wiznet. Although it isn't frozen, getting it closer (if people are good with that) would be good.
crimson ferry
right, forgot about the frozen thing
short tendon
But the release should happen eventually... 😉
manic glacierBOT
crimson ferry
I can open an issue on WIZnet, not sure about getting into the weeds on DHCP and all that... that library is kinda scary to me
short tendon
if you get bored, you should look over the issues I help with on #help-with-wippersnapper-and-adafruitio
short tendon
I can open an issue on WIZnet, not sure about getting into the weeds on DHCP and...
I'm happy to tackle the PR if I get a thumbs up
crimson ferry
I do lightly scan that channel, but don't use those services so don't usually have much to add
short tendon
There "bug" we found makes me want to cry
crimson ferry
do you remember where we talked about wiznet "connection"?
found it #circuitpython-dev message
short tendon
Dang, was just about to share that
crimson ferry
just glancing at the library, maybe it's just set_dhcp(), and if that's the case we have it now... call the init with is_dhcp=False, then user code calls set_dhcp() ...maybe?
I can test it out tomorrow
crimson ferry
@short tendon I think it's already baked in...py radio = WIZNET5K(spi, cs, reset=rst, is_dhcp=False) radio.set_dhcp() this works if the cable is connected, doesn't fail until radio.set_dhcp() if the cable is unconnected (ConnectionError: The Ethernet connection is down)
but maybe a connect() alias for consistency, and .connected and .ipv4_address
manic glacierBOT
short tendon
<@1152454697614069811> I think it's already baked in...```py
radio = WIZNET5K(sp...
Also, if there isn't a cable connected, you would get a 5 second delay here, wouldn't you?
https://github.com/adafruit/Adafruit_CircuitPython_Wiznet5k/blob/main/adafruit_wiznet5k/adafruit_wiznet5k.py#L251
crimson ferry
oh yeah, but does it really take 5 seconds to detect link, that seems like a lot