Does circup use the readme or the requirements? Or does anything use the readme in an automated way?
#circuitpython-dev
short tendon
lone axle
Does circup use the readme or the requirements? Or does anything use the readme ...
circup doesn't use the readme for anything that I am aware of. I think the two places it can look for requirements are requirements.txt and in pyproject.toml under a specific argument name (that I don't know off the top of my head)
short tendon
I'm pretty sure by the time I'm done with ConnectionManager, I'll have touched every repo but 5...
Should the requirements.txt and README.rst be recursive? I see things like the MegTag requirements that has both portalbase and requests (and portal requires requests)
midnight ember
Should the `requirements.txt` and `README.rst` be recursive? I see things like t...
I've noticed that happens a lot. Things can get double imported through various levels of abstraction between code.py and multiple libraries requiring the same base library. Requests, PyPortal, & Portalbase is a good example. If you have multiple helper libraries that call requests it'll still work. Circuit Python seems to work fine no matter how many helper libraries call upon the same base library in different instances.
The multiple levels of abstraction buried in PyPortal is intended to make things easier for beginners but I found it actually made it far more complicated. That's why I started using Requests only for all my projects, it's much easier to use and removed a ton of overhead from PyPortal & Portalbase.
tulip sleet
@slender iron basic PWMOut is fine in terms of finalizers. espcamera allocates a PWMout and has a deinit(), but it does not have a finalizer. I think I need to add that.
BusDisplay.c still has
common_hal_pwmio_pwmout_never_reset(&self->backlight_pwm);
should this be removed?
onyx hinge
@crimson ferry ```Auto-reload is on. Simply save files over USB to run them or enter REPL to disable.
code.py output:
Wiznet5k WebClient Test
http text test
This is a test of Adafruit WiFi!
If you can read this, its working :)
https json test
{
"args": {},
"headers": {
"Host": "httpbin.org",
"User-Agent": "Adafruit CircuitPython",
"X-Amzn-Trace-Id": "Root=1-65dfa1ee-610a5eea3653af38294dd85f"
},
"origin": "162.210.7.8",
"url": "https://httpbin.org/get"
}
Code done running.
MemoryError:
Press any key to enter the REPL. Use CTRL-D to reload.
That's interesting, it manages to throw a MemoryError after the code finishes running 🤔 but .. it works! ssl request on wiznet (with modified core code & modified wiznet library).Yeah, python doesn't care (and once it's imported it doesn't import again), but just trying to figure out what the "policy" is, so I can get things correctly updated
onyx hinge
short tendon
OMG @onyx hinge that's amazing!!! (and if we did that first, I would have done CM differently....)
onyx hinge
what part of that is related to CM? did it need wiznet workarounds or something?
I didn't change much in wiznet
and it doesn't change esp32spi which is still its own thing
short tendon
Few things:
- wouldn't need
create_fake_ssl_context(assuming it also works with esp32spi) adafruit_connection_manager.get_radio_socketpoolwould be much simpler
In the end, it will just be removing code if it ends up working
onyx hinge
we probably can't just add CIRCUITPY_SSL on most boards, it adds a lot of code
so in summary I bet what you did is good and you shouldn't second guess it just because of this proof of concept code
onyx hinge
mmmaybe? the only requirements are to obey enough of the socket "protocol" (duck typing) & use SOCK_STREAM=1
but esp32spi's SSL is maybe accelerated by esp-idf code and this is generic unaccelerated code, and it takes away flash space we might not have to spare on existing boards, so it's probably not just a "switch & forget with no trade offs" kind of change
👋 I'm headed out again, ttyl!
ssl-anything branch updated with my latest code
lone axle
Yeah, python doesn't care (and once it's imported it doesn't import again), but ...
I don't know that we have any official policy around it or anything. I think some libraries have a more thorough list in the readme than others. I'd say generally I'm in favor of including the recursive ones in the list though so in that case list both portalbase and requests even though portalbase itself does require requests too.
crimson ferry
<@360465312518504448> ```Auto-reload is on. Simply save files over USB to run th...
That's awesome! I worked on it for a few hours last week but didn't get past Invalid socket for TLS. Proof of concept opens up a lot of new possibilities.
short tendon
we probably can't just add CIRCUITPY_SSL on most boards, it adds a lot of code
If you open a draft PR, we could see what boards fail on space, and I can try it on as many boards as I have to see if the errors are the same everywhere...
crimson ferry
I'd be happy to test too. The WIZnet RP2040 EVB Pico 5100/5500/6100 are step one. The Adafruit EtherNet FeatherWing could potentially go on a variety of processors if space permits. There is also a hat version of the WIZnet EVB Pico that could go directly on a Pico, or Pico W which already has ssl 🙂
I hope it will fit on some S2 or S3 feathers.
unrelated Q: is the CP I2C implementation intended to support multi-controller (e.g., two MCUs talking to a shared peripheral)?
short tendon
I'd be happy to test too. The WIZnet RP2040 EVB Pico 5100/5500/6100 are step one...
Oh cool, didn't know about those boards
crimson ferry
Oh, duh, S2 and S3 already have ssl ...I was testing with an S3 TFT + Ethernet FeatherWing. Pico W + 5100 hat is also ready out of the box.
tulip sleet
_unrelated Q:_ is the CP I2C implementation intended to support multi-controller...
There is no deliberate support for multi-controller in CircuitPython
slender iron
should this be removed?
Ya, I'm surprised I didn't delete that common_hal function completely. Does espcamera do bulk reset? You could call deinit on the pwmout then
crimson ferry
There is no deliberate support for multi-controller in CircuitPython
it works! I connected one SHT31 to two ESP32-S2 QT Py boards with Stemma QT cables (cut the power wire on one of them), and both processors get looping results successfully from simpletest.
candid sun
last night i found that 9.0 beta2 was causing the pycamera library to be incredibly slow while beta0 was normal and speedy. using absolute newest this morning and PR8990 seems to have fixed it 🎉
short tendon
@tulip sleet and @slender iron I opened up 9 Fix README requirements PRs. There are a few others that don't need Requests but need MiniMQTT, but the MiniMQTT PR hasn't been reviewed/merged so will wait on that
tulip sleet
<@329766224093249548> and <@252717193496756235> I opened up 9 `Fix README requir...
Approved all of these with some very minor tweaks. Asked Brent to review MiniMQTT.
I haven't looked at the Learn Guide one yet
short tendon
Approved all of these with some very minor tweaks. Asked Brent to review MiniMQT...
Thanks. Sorry for the ABC's I missed
tulip sleet
np 🙂
short tendon
Scott mentioned that the corresponding guides from my Learn Guide changes need to be updated. Is this something I can do?
ebon kayak
hi @slender iron, i saw your live on youtube on usb host debugging
i want to know how can i develop by myself the vendor specific host, because i have a board with stm32s working in device mode as vendor specific device and i want to connect it to the usb of the pico w to communicate through wifi
lone axle
Scott mentioned that `the corresponding guides` from my `Learn Guide` changes ne...
It depends on two things essentially. If the part within the guide that needs updating is actually an embedded code file from Github. Then the change needs to get made in a PR to https://github.com/adafruit/Adafruit_Learning_System_Guides repo, and you can do that. If the change is needed in the text of a guidepage or in some code snippet that is not embedded from Github then you can only change it if you have editor access in learn.adafruit.com.
tulip sleet
It depends on two things essentially. If the part within the guide that needs up...
This about https://github.com/adafruit/Adafruit_Learning_System_Guides/pull/2743. So the code will get updated automatically, but we'll need to look to see if there's related text that becomes obsolete
I think in many cases it's a mirrored page, so it will only need to be changed once.
lone axle
Ahhh, I see
tulip sleet
@short tendon Is the Learn Guide PR above dependent on the MiniMQTT library being updated first?
short tendon
<@1152454697614069811> Is the Learn Guide PR above dependent on the MiniMQTT li...
It is not. This is all just switching away from set_socket similar to some other PRs I opened. Technically those guides will stop working once MiniMQTT is released
tulip sleet
great - I'll work on that PR and look at what Learn Guide changes are required
short tendon
Cool. If there is a way I can help, let me know. My goal isn't to give people a bunch of work
crimson ferry
short tendon
https://docs.circuitpython.org/en/latest/ is Welcome to the ulab book!
slender iron
hi <@252717193496756235>, i saw your live on youtube on usb host debugging
i wan...
I've never done vendor specific stuff. The low level pyusb API that circuitpython implements shouldn't care though. It just reads and writes endpoints.
lone axle
archive.org shows the most recent two snapshots were on 2/14 (or perhaps 2/15, the page header disagrees with the calendar). It looks like it was loading the appropriate docs page at that time.
tulip sleet
Cool. If there is a way I can help, let me know. My goal isn't to give people a ...
The pages in question are mostly a single mirrored page. It's actually out of date in a lot of ways (talks about secrets.py, etc.). I started to edit it, but something broke and some of my changes were lost. It needs considerable rework and someone (maybe me or not) will be doing it. ex: https://learn.adafruit.com/adafruit-pyportal/internet-connect
also there is a lot of redundant info in there
tulip sleet
<https://docs.circuitpython.org/en/latest/README.html> is 404?
checking on this - we didn't change this recently, as far as i know!
it is broken
something bad is going on with the docs build, not sure what. make html locally doesn't even work
this may be some tool update that broke things. I can't even build 9.0.0-alpha.6 docs locally
slender iron
what is the error?
tulip sleet
a bunch, hold on...
tulip sleet
what is the error?
also ERROR setuptools_scm._file_finders.git listing git files failed - pretending there aren't any but I think that may be old and innocuous
I didn't deliberately update my pip3-installed tools recently
but the ulab-replacing-circuitpython above may be something else, or may be a different manifestation of this
the intersphinx inventory thing may reflect some change at rtd. I should look at their news
slender iron
ya, feels like a tool update
tulip sleet
ya, feels like a tool update
yes, getting same error even on 8.2.x branch
slender iron
let me know if you want me to look too
tulip sleet
let me know if you want me to look too
I updated aeveryting, including Sphinx, to 7.2.6 from 6.something and am now getting different errors. Could you just do a make html locally and see if it works for you?
slender iron
yup
tulip sleet
there is now something broken in make stubs --python is not an option
subprocess.CalledProcessError: Command '['/usr/bin/python3', '-Im', 'pip', '--python', '/tmp/build-env-fuqkaxvi/bin/python', 'install', '--use-pep517', '--no-warn-script-location', '-r', '/tmp/build-reqs-oftqxpb9.txt']' returned non-zero exit status 2.
slender iron
/home/tannewt/repos/circuitpython/shared-bindings/busdisplay/index.rst:447: WARNING: more than one target found for 'any' cross-reference 'ParallelBus': could be :py:class:paralleldisplay.ParallelBus or :py:class:paralleldisplaybus.ParallelBus``
tulip sleet
i didn't even get there yet, I'm failing further back
slender iron
lemme update to main
tulip sleet
@slender iron maybe create a temp venv and pip3 install -r requirements-*.txt and then see how that goes. I basically did that (without the venv)
@lone axle are you looking at the connectionmanager underscore issue or don't have time at the moment? I was going to but I am about three interrupts deep at the moment
slender iron
<@252717193496756235> maybe create a temp venv and `pip3 install -r requirements...
I'll run down the breaking package
I dumped my working venv state before switching
lone axle
<@382939733107408897> are you looking at the connectionmanager underscore issue ...
I'll look into it. I'm digging a bit off and on. Working on something different but it has long periods of waiting for things to load and save so I'm poking around on that circup / connectionmanager thing during those times.
tulip sleet
i thought maybe circup had a special-cases list
lone axle
That is odd that adafruit_bus_device seems unaffected, good catch on that. I first I thought maybe it had the same problem but wasn't noticed because it's a core module now too on most devices. But that isn't the case. Circup does correctly convert its pypi name to import name.
That could be. I do think there are some JSON data files that it pulls some of the meta info from. Perhaps in there or hardcoded somewhere within circup it's got handling for it.
slender iron
@tulip sleet It builds for me just fine after installing requirements-docs.txt
(and fixing parallelbus)
slender iron
If the value is null, Read the Docs will try to find a conf.py file in your project.
will PR to add it explicitly shortly
tulip sleet
find . -name conf.py
./extmod/ulab/docs/manual/source/conf.py
./lib/tinyusb/docs/conf.py
./conf.py
...
no idea why the order changed
slender iron
¯_(ツ)_/¯
tulip sleet
fail_on_warning: true -- we'll see if that uncovers other things
slender iron
I think that's what make html does already
looks like the scheduler is skipping docs builds too
ah, it only does as needed it looks like
mortal kernel
@tame creek Please give me a ping, I'd like to discuss #8824.
slender iron
watches RTD build
tulip sleet
The --python error I am seeing is due to this: https://github.com/pypa/build/issues/745
unfortunate coincidence
undone 39 minute ago
tulip sleet
I just was fiddling with the webhooks a few day ago
but I don't remember adding one, maybe I did
thanks
slender iron
¯_(ツ)_/¯
manic glacierBOT
tulip sleet
@slender iron my local doc build is fine now
i may have updated something with pip at the exact time build was wrong in pypi
popping my stack a few levels
tulip sleet
fantastic!
slender iron
@gilded cradle did you see my reply to you? https://github.com/adafruit/circuitpython/issues/8980#issuecomment-1965522826 I'm not sure you've found a bug
short tendon
@tulip sleet MiniMQTT was merged, so opened 3 more README fixes
And at this point, I have a bunch of examples to fix and a call in Adafruit_CircuitPython_ESP_ATcontrol. Should I fix ESP_ATcontrol? It would be similar to ESP32SPI, but I don't have a way to test
tame creek
@argonblue Please give me a ping, I'd
short tendon
all the above are set_socket
tulip sleet
_all the above are `set_socket`_
all merged
tulip sleet
And at this point, I have a bunch of examples to fix and a call in `Adafruit_Cir...
are there guides that use ESP_ATControl
short tendon
The code is in the ESPAT_WiFiManager
If I create the PR would someone be able to test?
tulip sleet
I would say fix it and we'll merge it and wait for PR's to fix anything wrong. We aren't supporting it anymore.
we aren't going to test it
short tendon
Sounds good
manic glacierBOT
short tendon
Adafruit_CircuitPython_ESP_ATcontrol PR open
lone axle
@tulip sleet you were right, cicup has a dictionary that maps the non-standard names which includes adafruit_bus_device. I'll submit a PR that adds Connection Manager.
manic glacierBOT
orchid basinBOT
short tendon
@lone axle, when do you usually go through and create the releases for the CP helpers?
lone axle
<@382939733107408897>, when do you usually go through and create the releases fo...
Monday's are typically when I go through the report and make releases for libraries that have new commits since their last release.
(I work a full day on CircuitPython on Monday's, the rest of the week my activity is more limited to evenings and weekend.)
short tendon
Great. All I have left for set_socket removal is some examples. I'll get them knocked out tomorrow (and hope they get merged)
And apologize in advance for all the extra work...
lone axle
Sounds good, and no need for apologies. I appreciate the work going into improving the libraries!
short tendon
I think I'll surpass 30 PRs...
Oh I see, you're fixing things behind the scenes for me... 😉
Also, I see requests in Mismatched versions in .pre-commit-config.yaml, so it doesn't run daily, but my changes are clearly the problem
manic glacierBOT
short tendon
It is done! 8 final set_socket PRs to update examples
manic glacierBOT
tulip sleet
@short tendon re the create_fake_ssl_context() additions. I don't understand this exactly. Would SSL work, but without actually checking authenticity or ? If Seesion was given a None ssl_context` instead, how would things behave differently?
I am not objecting, I just do not know.
short tendon
<@1152454697614069811> re the `create_fake_ssl_context()` additions. I don't und...
It's the same as what requests.set_socket did, just under the covers. So if SSL is supported, like on an ESP32SPI it works fine. On a WIZNET5K it will throw a proper exception.
tulip sleet
so what is "fake" about it? (I know that's the terminology that's used.) Is it becuase the SSL is set up later?
short tendon
It just returns the same socket back
tulip sleet
it's fake because SSL is handled in a different way instead of with an actual SSL socket, is that right?
I'm trying to understand how I would explain this in, say, a Learn Guide
short tendon
Ahhh. Thinking
tulip sleet
In *ix CPython, the context can wrap an actual SSL-capable socket, like https://docs.python.org/3/library/ssl.html#ssl.SSLContext.wrap_socket
short tendon
So that's what it's doing too
tulip sleet
ok, looking at this more, the context itself doesn't have a socket, but it has info about doing SSL (cert info, etc.). The "fake" context doesn't have that info: the device handles that itself
ok, I have never really paid attention to this boilerplate before, and now I have to.
short tendon
The naming was the same in both requests and minimqtt. I can rename it to something better...
short tendon
Sounds good
If Jeff gets wiznet5k working with ssl, we can revisit to see what we want to change
Since the helpers will need to be updated
Also @tulip sleet since I don't get notifications on issues. If something pops up from my work - feel free to ping me. I'll get to it quickly
crimson ferry
what's the usage difference between create_fake_ssl_context() and get_radio_ssl_context()?
end users only ever use get_radio_ssl_context()?
tulip sleet
Also <@329766224093249548> since I don't get notifications on issues. If somethi...
you mean you haven't enabled notifications (email or otherwise) on all issues? Or just that you get it in email only?
short tendon
Oh I guess I could do that... Normally I have it set on the organization level. Forget that I can do it individually
tulip sleet
@short tendon - @crimson ferry 's q is for you, I think
crimson ferry
@short tendon the ConnectionManager example (which works on ESP32SPI) haspy pool = adafruit_connection_manager.get_radio_socketpool(radio) ssl_context = adafruit_connection_manager.get_radio_ssl_context(radio) connection_manager = adafruit_connection_manager.get_connection_manager(pool)but the Requests ESP32SPI example has```py
ssl_context = adafruit_connection_manager.create_fake_ssl_context(pool, radio)
requests = adafruit_requests.Session(pool, ssl_context)
Oh, I should probably double check some of my changes...
Users should use get_radio_ssl_context, unless it's for something unsupported like the fona or ESP_ATcontrol
crimson ferry
ah, ok, thanks
tulip sleet
would it help if ESP32SPI, etc., had "get context" operations added, instead of ConnectionManager having to know about the boards?
crimson ferry
@short tendon if you revisit the docs, these I think would have return values: https://docs.circuitpython.org/projects/connectionmanager/en/latest/api.html#adafruit_connection_manager.get_radio_socketpool & https://docs.circuitpython.org/projects/connectionmanager/en/latest/api.html#adafruit_connection_manager.get_radio_ssl_context
short tendon
Yup. I did work on an unsupported board, and then didn't change back. Fixing now
short tendon
would it help if ESP32SPI, etc., had "get context" operations added, instead of ...
yes, but we would need to add that to all "radios"
tulip sleet
it could be a new requirement for radios
in the long run
ConnectionManager could check for the attribute for now
short tendon
ConnectionManager could check for the attribute for now
True, we could added and remove the special casing as we go...
short tendon
@tulip sleet I'm going to open a bunch more PRs. They will look like this: https://github.com/adafruit/Adafruit_CircuitPython_AdafruitIO/pull/118
Which removes the whole fake wording. except for the 2 old radios
tulip sleet
<@329766224093249548> I'm going to open a bunch more PRs. They will look like th...
So, suppose each "radio" had something you could import as ssl, and could do ssl.create_default_context(). Then you could remove knowledge about specific radios from ConnectionManager.
as with pool right now
since you can import blah as pool from various radios
short tendon
So, suppose each "radio" had something you could `import` `as ssl`, and could do...
Although at that point the spi radios would need to import connection manager to get the wrapper
short tendon
the create_fake_ssl_context one for wrapping the socket
tulip sleet
the `create_fake_ssl_context` one for wrapping the socket
yes, I'm saying that code could be in the radio class instead. It could be duplicated, or could be imported from ConnectionManager
short tendon
I can look at esp32spi and draft a PR to see how it looks
I just finished fixing the examples for the ones I pushed last night. 7 new PRs (Fona didn't need one)
tulip sleet
I have been out of the loop on this. @slender iron may have an opinion
short tendon
Okay, I found all the incorrect uses of create_fake_ssl_context. Should have 10 PRs. Sorry about that
and thanks for catching that @crimson ferry
tulip sleet
Okay, I found all the incorrect uses of `create_fake_ssl_context`. Should have 1...
some are drafts and some are not. Did you mean to distinguish some?
short tendon
Fixed. Was waiting on tests to run. With the GitHub issues, it's been slow
for the helpers, do you think keep the names the same:
pool = esp.get_radio_socketpool()
ssl_context = esp.get_radio_ssl_context()
?
tulip sleet
I thought that's what you did in the new PR's?
I am willing to wait for @slender iron 's comments here, since these are just in the examples.
The IOT repo examples make me think they really need to have wifi examples as well as ESP32SPI examples
I"m not asking you to do that!
e.g. there should be multiple groups of examples. There's a lot of boilerplate for ESP32SPI, unfortunately.
short tendon
So that was the agreed upon methods originally. What happened is I switched to fixing things for Fona and ESP_ATcontrol and started to grab the wrong helpers
If we moved things into the "radios", the code would be even cleaner and you wouldn't need to import adafruit_connection_manager
and then the next set of PRs could go from:
pool = adafruit_connection_manager.get_radio_socketpool(esp)
ssl_context = adafruit_connection_manager.get_radio_ssl_context(esp)
to:
pool = esp.get_radio_socketpool()
ssl_context = esp.get_radio_ssl_context()
And would it be hard to add that to wifi.radio?
crimson ferry
does the ethernet TLS case complicate things... ssl is independent of the 'radio'
tulip sleet
yes, so that's why I was thinking of an ssl to import from the non-wifi radio classes
is the ethernet library different? @crimson ferry
I mean, ConnectionManager is still useful as a library because of the socket managment, is that right?
short tendon
It doesn't. Because it could attempt to import ssl, and if works use it. if not it wouldn't
And yes ConnectionManager is still very much needed
crimson ferry
I think ConnectionManager is useful with any library or combination of libraries drawing sockets from a common pool
tulip sleet
i am thinking just esp.get_socketpool() and esp.get_default_context() or esp.ssl.get_default_context, trying to make it as close to CPython API as is reasonable. Do you need esp.get_socketpool(), or isn't that just a class?
crimson ferry
It doesn't. Because it could attempt to import ssl, and if works use it. if not ...
I think that makes sense, native ssl would get imported by the WIZnet library
short tendon
Yup
tulip sleet
you were already doing from blah import blahblah as pool
so that pool is the same as blah.get_socketpool()?
short tendon
yup
the point of get_socketpool is that you can write code for multiple devices, and not need to think about the imports
if we had radio.get... it would do the same
tulip sleet
yes, could try to make the boilerplate to getting the socketpool and the ssl context be the same in the example. So that goes back to using connection manager rather than hiding it by calling it on the radio. Either you use import to make the names the same or you use connection manager to make fetching them the same
so my idea of pushing it down to the radio is not necessarily better
short tendon
It depends. If we were to start adding other external radios, we would need to update the helpers... Not sure how frequently that happens
I think what we have now is good, and not needing to change it
tulip sleet
if you push it to the radio, then it's the radio's responsibility to add an implementation
that was my motivation
short tendon
and we can always update adafruit_connection_manager.get_radio_socketpool(radio) to try to call radio.get_radio_socketpool()
Probably best to roll out what we have and see what people say
tulip sleet
so would you like the current crop of PR's to be merged to keep it all consistent?
short tendon
I would
tulip sleet
on it
tulip sleet
I think you're paying plenty of attention 🙂
not sure what you meant, you mean radio instead of fake?
short tendon
not sure what you meant, you mean radio instead of fake?
where?
short tendon
is Adafruit_CircuitPython_AWS_IOT?
tulip sleet
is `Adafruit_CircuitPython_AWS_IOT`?
oops! sorry, now merged. They will all get new library versions in the next release sweep. but I won't do that now, since it's only example code, right?
tulip sleet
@mortal kernel of historical interest to you concerning BLE on Pico W: <#1012732018183913513 message>
short tendon
oops! sorry, now merged. They will all get new library versions in the next rele...
Some of them need new releases (for actual code changes), but they can all just wait for the release sweep
slender iron
I don't have strong opinions about the 'fake' changes
short tendon
@tulip sleet forgot to respond to your examples comment... One of the things I tried to do in requests (and partly taken from minimqtt) was to make a base example for each (onboard, esp32spi, wiznet5k, fona) that had the same calls, and different init for the radio. And then since it's cleanest, the bigger examples are all WiFi based.
manic glacierBOT
CircuitPython version
Adafruit CircuitPython 8.2.9 on 2023-12-06; Seeed Studio XIAO ESP32C3 with ESP32-C3FN4
Code/REPL
# web workflow, not Python, not the REPL:
http://10.0.0.1/code # but use a valid IPv4 address for your unit.
Behavior
Instead of going to the device, I am presented with the
dialog which sends me to circuitpython.local if...
manic glacierBOT
short tendon
I thought I would throw some thoughts together on ConnectionManager (and at the same time try out the new adafruit-playground. If anyone wants to take a peak, it's here: https://adafruit-playground.com/u/justmobilize/pages/connectionmanager
vital matrix
A few things that puzzled me as I was looking through how ports/unix is getting built:
a) it seems like it maybe predates the conventions suggested in https://docs.circuitpython.org/en/latest/docs/porting.html ? So I shouldn't read to much into that organization for the unix port at least?
b) I'm also assuming this is pretty much only used to run basic tests, and nothing here really relates to circuitpython on a linux SBC?
c) I'm trying to follow why many of the modules aren't included in the current build here: https://github.com/adafruit/circuitpython/blob/54e78c89fc9e0d2b31544dfd03c30a159e8540a4/ports/unix/Makefile#L197-L210 - e.g. modmachine.c and modffi.c aren't included (the only place where SRC_MOD is initialized just includes the ulab module)
woeful briar
Hello all. It has been a minute since I have posted to this discord. I had retired several years ago but kept up with the Adafruit YouTube. I have ended up contracting for a company in their Reasearch and Development department. I have become the bridge between Project Technical Design and Software Development that supports what ever the business and the design teams come up with. So much fun that I get to do what I have a passion for and get paid.
So, during some research to solve a problem, I found a ready-made device that fits the bill. The idea is to remotely server files to a device that makes a request based on an alert signal, then have the device present these files via USB to a connected host.
What I came up with is the Lilygo t-embed. I saw that @lethal abyss created a port for this device and I am anxious to get busy. So, I have downloaded the latest version of CP 9.0 beta, from the lasty build.
I flashed the ESP32S3 chip and I am able to see the "Hello World" on the screen. Getting exited now.
Then, when I looked at the code provided, I saw the '/SD' folder and read the text inside.
So, my question is, can I use the Wi-Fi to download files to the built in SD card reader and then resent them to the USB host as a simple file list?
Any feedback will be appreciated.
This question is about CircuitPython not the t-embed.
fringe yarrow
I'm new, so forgive me if this isn't the right spot to ask. I'm interested in working towards getting a currently unsupported ESP32 dev board device working with CircuitPython. While I have no doubt I can make it work for myself, I'm curious if there's any way to include my work in the official port for others to enjoy, or if that is reserved for the board makers themselves to handle.
short tendon
I'm new, so forgive me if this isn't the right spot to ask. I'm interested in wo...
You can make a PR into the circuitpython repo: https://learn.adafruit.com/how-to-add-a-new-board-to-circuitpython
We all benefit from users adding new ones
fringe yarrow
You can make a PR into the circuitpython repo: https://learn.adafruit.com/how-to...
Perfect. I had read that but the language therein was unclear if community could contribute or just the board makers, as it starts with "You've designed a circuit board ..."
short tendon
I've seen many of us ordinary folks open up PRs 🙂
crimson ferry
@fringe yarrow Adding a board, especially one similar to others out there, is pretty straightforward
choose the board closest to yours (same SOC or module, etc.), and tweak what's necessary for special pins, peripherals, etc
fringe yarrow
choose the board closest to yours (same SOC or module, etc.), and tweak what's n...
Thanks.. It wasn't so much the technical component I was unsure of, but the policy regarding who can add new boards.. I've done the preliminary pin reassignments from a similar board, now I'm at the whim of the postal service until I get the device.
short tendon
Thanks.. It wasn't so much the technical component I was unsure of, but the poli...
Curious, which board is it?
fringe yarrow
Curious, which board is it?
"MakerGo ESP32-C3 SuperMini". Very similar to the Seeed XIAO ESP32C3.
tulip sleet
A few things that puzzled me as I was looking through how `ports/unix` is gettin...
ports/unix is very close to the micropython unix port. Its main use is to run tests. jepler is working on a "posix" port that would be more in line with standard CircuitPython. However, it is still not necessarily a substitute for the Blinka library. modmachine and modffi are more MicroPython.
some mention of posix port here: https://github.com/adafruit/circuitpython/pull/8900
tulip sleet
I thought I would throw some thoughts together on `ConnectionManager` (and at th...
I have "Featured" your ConnectionManager Playground post.
Could you add something that makes clear some examples are "old way" and some are "new way"? Maybe with headings.
vital matrix
`ports/unix` is very close to the micropython unix port. Its main use is to run ...
Thank you for the pointers @tulip sleet ! I was casting about for some way to run basic automatic tests on some of the CP modules - if @onyx hinge is working on something that maybe was intended, or could do that - sounds perfect. I'll poke around at https://github.com/jepler/circuitpython/tree/port-posix in the meantime
onyx hinge
it's far from ready for prime time but you're welcome to look at the tree!
posix port isn't going to include "modmachine" or "modffi" you mentioned, as those aren't used in circuitpython.
vital matrix
Very cool! Were you intending to use it for also implementing some sort of virtual hal to test modules, or was it for a different purpose?
onyx hinge
My main goal right now is to be able to test everything that is implemented in shared-modules. I have added some of those things to the circuitpython posix port, "coverage" flavor, but each one is an ad-hoc addition, rather than using the circuitpython-specific py/*.mk files like the ports that run on hardware
I wouldn't rule out implementing things within its common-hal (like, say, some kind of mock digitalio) but that is out of scope for me right now, I'm more interested in just passing existing tests.
vital matrix
Fantastic - I'll continue looking at that branch, and best of luck with merging it back, sounds like it'll be really handy.
onyx hinge
right now the posix port status is 799 tests performed (49823 individual testcases), 177 tests skipped, 48 tests failed
vs the unix port 890 tests performed (53158 individual testcases), 890 tests passed, 86 tests skipped
So there are 48 failures + 91 skips to investigate and resolve
vital matrix
Ha! Congratulations that there are 600+ tests already passing 🙂
onyx hinge
and many of the skips are fine too, like vfs_lfs (circuitpython does not use littlefs)
at least 26 failures are due to differences in floating point behavior (test output includes printing floating point numbers)
short tendon
I have "Featured" your ConnectionManager Playground post.
Could you add somethi...
@tulip sleet done! I made them h2 which makes them linkable as well
vital matrix
at least 26 failures are due to differences in floating point behavior (test out...
Nice - getting quite close then.
onyx hinge
I'm not sure when I'll get back to work on it, I have had this last week off and I suspect there's lots of issues & new hardware stuff to tackle before I do this "fun stuff". goodnight for now, and thanks for the interest in what I'm working on!
midnight ember
@short tendon was working on porting my fitbit project from 8.2.2 to 8.2.10 and loaded up new libraries. Was pretty neat to see my code.py requests error with "no module named connection manager exists".
Even though I know connection manager is a dependency of requests it didn't affect my old code at all. Still works great. My guess is there should be some transitional code for 9.x in there somewhere?
and yes i found the lib and added it immediately
short tendon
Even though I know connection manager is a dependency of requests it didn't affe...
So anything using onboard WiFi, will be 100% seamless
I put this together yesterday: https://adafruit-playground.com/u/justmobilize/pages/connectionmanager
midnight ember
It definitely was. Great job. Didn't notice a thing other than the missing module error.
short tendon
I should probably add something about circup on that doc, since it would have taken care of that for you...
midnight ember
i don't use circup, just manual 8.x bundles drag and drop from circuitpthon.org/libraries
short tendon
And thanks, it was a fun project
I don't either, but I want to try it (especially for fonts)
Now I get to watch help-with and see if there's anything new to tackle, or if I want to work on some personal libraries
tulip sleet
Even though I know connection manager is a dependency of requests it didn't affe...
it isn't 8.x vs 9.x, should work on both. The latest circup 1.5.4 handles ConnectionManager (its naming without an underscore is special-cased: there is a list of such special cases). I was old-school and was not using circup, but then I discovered how much time it was saving me. It also does upgrades/downgrades properly with 9.x now.
short tendon
it isn't 8.x vs 9.x, should work on both. The latest circup 1.5.4 handles Connec...
I was just trying this and can't get it to work. Anything special I need to do to get it to detect the version? Everything looks like this with verbose:
03/03/2024 10:20:46 INFO: C:\Users\justi\AppData\Local\adafruit\circup\adafruit\adafruit-circuitpython-bundle-py\adafruit-circuitpython-bundle-py-20240301\lib\adafruit_requests.py
03/03/2024 10:20:46 INFO: Extracted metadata: {'mpy': False, '__version__': '0.0.0+auto.0', '__repo__': 'https://github.com/adafruit/Adafruit_CircuitPython_Requests.git'}
And so it's not upgrading from anything
tulip sleet
I was just trying this and can't get it to work. Anything special I need to do t...
The .py bundle files are not rewritten to include the version information when they are bundled. So they have the EDIT: This is wrong. They used to have the version number.0.0.0+auto.0 version number.
we could argue about whether this is good or bad, but they are the original source files, as found in the repo
short tendon
So I'm assuming I'm doing something wrong... I assumed circup update would update requests==2.0.3 to requests==3.0.0. Does it not do major versions? I assumed since it installed adafruit_connection_manager that it would
But it's not even doing 2.0.3 to 2.0.5
I am on 9.0.0-beta.2
tulip sleet
So I'm assuming I'm doing something wrong... I assumed `circup update` would upd...
It will not update the .py files. what is circup --version
short tendon
1.5.4. I don't have any .py files. They are all .mpy
03/03/2024 10:20:48 INFO: {'path': 'D:\\lib\\adafruit_requests.mpy', 'file': 'adafruit_requests.mpy', 'name': 'adafruit_requests', 'repo': 'https://github.com/adafruit/Adafruit_CircuitPython_Requests.git', 'device_version': '2.0.3', 'bundle_version': '0.0.0+auto.0', 'bundle_path': 'C:\\Users\\justi\\AppData\\Local\\adafruit\\circup\\adafruit\\adafruit-circuitpython-bundle-9mpy\\adafruit-circuitpython-bundle-9.x-mpy-20240301\\lib\\adafruit_requests.mpy', 'mpy': True, 'min_version': '9.0.0-alpha.1', 'max_version': None}
03/03/2024 10:20:48 INFO: {'path': 'D:\\lib\\adafruit_connection_manager.mpy', 'file': 'adafruit_connection_manager.mpy', 'name': 'adafruit_connection_manager', 'repo': 'https://github.com/adafruit/Adafruit_CircuitPython_ConnectionManager.git', 'device_version': '1.0.0', 'bundle_version': '0.0.0+auto.0', 'bundle_path': 'C:\\Users\\justi\\AppData\\Local\\adafruit\\circup\\adafruit\\adafruit-circuitpython-bundle-9mpy\\adafruit-circuitpython-bundle-9.x-mpy-20240301\\lib\\adafruit_connection_manager.mpy', 'mpy': True, 'min_version': '9.0.0-alpha.1', 'max_version': None}
which board are you doing it on? I will try to replicate
short tendon
which board are you doing it on? I will try to replicate
A UM FeatherS3 with ESP32S3
I just put an old version of requests.mpy in the lib folder and then copied a random example from the requests repo
tulip sleet
I just put an old version of `requests.mpy` in the lib folder and then copied a ...
i am seeing what you're seeing, but I thought I fixed this. I certainly tested it a while ago. I had to change how it finds the version number in .mpy files for 9.x
the bundle files look OK. They have a version number
this is wrong 'bundle_version': '0.0.0+auto.0'
short tendon
I also tried specifying the version and board, but got a cpv undefined error (assuming CP version). Happy to get the whole exception if needed (when back at my laptop)
tulip sleet
i will do some instrumentation locally of the current circup code to see what is going one
short tendon
Thank you. I'm also happy to dig in, now that I know I'm not crazy
tulip sleet
@short tendon The version info is extracted from the .py bundles. I was wrong when I said we don't rewrite the version number. We did, and it broke between the 2023-12-10 bundle and the 2023-12-15 bundle (there are no intervening bundles there, so there was some change in the bundle build code).
manic glacierBOT
short tendon
<@1152454697614069811> The version info **is** extracted from the .py bundles. I...
Ahhh. Hopefully a small set of changes to look at
Please let me know if I can help
tulip sleet
Something in here, I think: https://github.com/adafruit/circuitpython-build-tools/pulls?q=is%3Apr+is%3Aclosed. I will not be able to work on this for the next several hours.
proven garnet
Build tools not working? Sorry, catching up while I update my desktop.
The issue is that it isn't replacing the version text now, so it still has the placeholder in libraries?
Oh, it's not a library, but the bundle?
short tendon
@tulip sleet and @proven garnet found it
PR here: https://github.com/adafruit/circuitpython-build-tools/pull/115 waiting on tests before I mark it ready for review
And it's ready
short tendon
@tulip sleet is it possible to rebuild the bundle? Or just need to wait until it runs tonight?
tulip sleet
I did run the rebuild, but it said it was already there, so it didn't upload. I could try removing today's files and try again.
short tendon
Makes sense. I'll leave that up to you. Can easily wait until tomorrow to finish playing with it
tulip sleet
I can make a release of something in the bundle. It's actually rebuilding the March 1 bundle -- no changes since then.
released requests 3.0.1
short tendon
They'll be a bunch when foamyguy, runs through all my updates...
tulip sleet
rerunning again...
tulip sleet
@Justin this release does not have the versions in the .py, but I forgot to make a circuitpython-build-tools release with your PR change. Now I did, but I'm not sure I can do it again. I could try, but we might have to wait for the day rollover
short tendon
It's the weekend. I can wait 😉
I manually set the version in the downloaded files and that worked. So I'll just update my doc after the release tomorrow
manic glacierBOT
lone sandalBOT
manic glacierBOT
This PR resolves #8998 issue.
Eliminated unnecessary color order commands.
Please review and merge.
empty salmon
I’m trying to figure out if the recent 9.0.something “ePaper no longer shows error” is by design or mistake. To be specific, code errors and exiting to REPL no longer writes to my ePaper board.DISPLAY.
@tulip sleet @slender iron - we discussed the underlying behavior issue #8721 and #8675. (I am not using CIRCUITPY_TERMINALIO=0).
I have looked at recent commits and closed issues and can’t find what has changed.
In our prior discussions I was told that errors should show on ePaper displays just like other types of displays. The current behavior contradicts this conclusion.
I am fine with either behavior but I need to know is which behavior is correct going forward.
When a CircuitPython board …
- had an ePaper attached and 1. exposed the ePaper as board.DISPLAY and
- the display has been setup in the board firmware and
- the CircuitPython code has not made any changes related to board.DISPLAY
… and the code stops with an error, will CircuitPython attempt to render the error on the ePaper display or will it suppress the error on the ePaper display?
tulip sleet
@short tendon overnight 2024-03-04 py bundle looks good! There is a version for adafruit_requests.py. Thanks!
short tendon
@tulip sleet I saw. I could of sworn yesterday it still added ConnectionManager when it wasn't there, but now it's not. Should it? Or only on install?
old zealot
I have just downloaded CP 9.0 beta for a project and noticed that there is an /sd folder with a text file named placeholder. The file contains the text: "SD cards mounted at /sd will hide this file from Python. SD cards are not visible via USB CIRCUITPY."
Can someone explain the last sentence about SD visibility?
lone axle
Would it be expected that displayio.release_displays() should clear any RAM that was used to initialize the display that is being released? We've run into an issue where a device has a display initialized in the core, but in code.py if you release it and attempt to initialze it again it raises a MemoryError which makes it seem like perhaps the release() didn't give up some portion of the memory since it did fit successfully when the core initialized it. I can share some more specific info and reproducer stesp if it's helpful but I'm mostly trying to understand if that behavior is considered problematic or not, RAM usage and management is still a bit of a black box to me.
lone axle
I have just downloaded CP 9.0 beta for a project and noticed that there is an /s...
I think it means If your project uses an SDCard and your code initializes the connection with the SDCard CircuitPython code will be able to see and modify files on the SDCard (by looking inside of the /sd/ directory). But your PC that you have the device connected to will not be able to see any of the real files inside of /sd/ only the placeholder.
So you can't just connect the microcontroller it to a PC and copy files directly from the SDCard to the PC. You'd have to take out the SDCard and put it into a USB SD adapter or something similar in order to access it's files from the PC.
(Worth noting that web workflow does allow you to see and modify files on the SDCard if it has been initialized in code though. Which is very convenient compared to taking the card out and back in again over and over.)
old zealot
@lone axle Thanks. I am trying to make the device/board look like a UDISK but only see the SD card. But the CP code needs to make changes to these files on the SD card. I know that I cannot have both configurations in storage at the same time. Basically unless the device has not "turned on" the CIRCUITPY drive it just looks like a regular USB drive.
Is tat possible with CP 9.0?
lone axle
<@382939733107408897> Thanks. I am trying to make the device/board look like a ...
To the best of my knowledge it's not possible with CircuitPython to have a PC connected via USB be able to view or modify any files on an SDCard that is connected to the microcontroller.
old zealot
To the best of my knowledge it's not possible with CircuitPython to have a PC co...
So the SD card is for CP use only?
lone axle
So the SD card is for CP use only?
That is my understanding yes as far as USB connection goes.
stuck elbow
I think there is some ongoing work to make that possible, but it can take some time before it's working
tulip sleet
<@329766224093249548> I saw. I could of sworn yesterday it still added `Connecti...
if you did update, then maybe it doesn't look at the pre-reqs, though perhaps it should
short tendon
if you did update, then maybe it doesn't look at the pre-reqs, though perhaps it...
It doesn't... 😦
But I just pushed up a branch that does. Need to test it a bit more
Would you we willing to peak and see if anything looks off?
lone axle
Should we add settings.toml and /sd/ to these generated file screenshots?
tulip sleet
Should we add `settings.toml` and `/sd/` to these generated file screenshots?
That is a good idea, but I think we should do it coincident with 9.0.0 final being released.
lone axle
tulip sleet
Maybe we should make a "to do list" issue in circuitpython for 9.0.0 final. I can do that. Does the PR above need to have submodule update somewhere else, or just merging that PR will fix it?
lone axle
That would be good I think as well. For the screenshot maker I don't think it's used as a submodule on any repos. It gets executed from an Actions cron task I believe but I don't think it involves use as a submodule.
it looks like the actions just clone it from the public repo URL. So as soon as a change is made and merged the next time that task runs it should use the updated screenshot maker and create the screenshots with whatever modifications were in the changes.
manic glacierBOT
slender iron
Hello all. It has been a minute since I have posted to this discord. I had reti...
The files on the SD card aren't readable over USB. We hope to add that in the future.
slender iron
I’m trying to figure out if the recent `9.0.something` “**ePaper no longer shows...
I checked this recently and did see an error message appear on eink. I'm not sure why it stopped working for you
empty salmon
@slender iron - what hardware did you test with? I will compare my board to it.
slender iron
<@252717193496756235> - what hardware did you test with? I will compare my board...
iirc a magtag since the screen is built in
slender iron
<@&356864093652516868> Reminder that the meeting is in just under 90 minutes. The notes doc is here: https://docs.google.com/document/d/1NxTW_z8Ewhjbh9voOGli4Jt0PIRy9slbp2u51NzX02s/edit Talk to you soon!
onyx hinge
It's in May
turbid radish
☺️
blissful pollen
not voice sorry
crimson ferry
@onyx hinge could the new core ssl changes also be available on boards that already have ssl, like espressif and Pico W? ssl is big, but maybe not the surrounding changes? this would allow, for example, espressif feather + ethernet featherwing, or Pico W + WIZnet hat.
willow totem
love the wiznet rj45 pico evb board, thanks @onyx hinge
onyx hinge
@crimson ferry yes, the ssl changes would apply anywhere that the ssl module is available. It's just that enabling ssl on new boards is not likely to happen because of the amount of code it brings in, so you could do ethernet on a feather esp32s3 or similar. not that I tested this scenario.
candid sun
@short tendon haven't forgotten about testing your time library- just haven't had the time (ha)
short tendon
<@1152454697614069811> haven't forgotten about testing your time library- just h...
No worries!
onyx hinge
@short tendon these SSL changes I'm talking about might impact the speed of SSL socket operations slightly (because they replace what were C function calls with Python function calls) -- I have a feeling you couldn't measure it for (say) connect() which is once per socket, but it might add up for each internal recv() or send() call. Did you write anything that might help me benchmark this while working with socket manager? I could just find a 100kB file and stream it over https, which should be hundreds of individual calls.
In 9, the framebuffer is initially allocated in the "outer heap"
short tendon
<@1152454697614069811> these SSL changes I'm talking about might impact the spee...
Not really. I have one thing that might help. If you go to my write-up: https://adafruit-playground.com/u/justmobilize/pages/connectionmanager and search for session_id you'll see something I wrote for multiple sessions. Currently it grabs tiny chunks of data. If you added some timing and maybe a bigger file, it could be a good start...
candid sun
thanks folks!
short tendon
Thanks all!
onyx hinge
Note that the US changes to DST ("summer time") next weekend
so if you're not in an area that follows the same DST rule the time locally may be shifted by 1 hour
the time will become "2PM in UTC-4" which is a confusing way to talk about it 🙂
lone axle
Thanks for hosting Scott. Have a great week everyone!
onyx hinge
Thanks scott!
silver tapir
Note that the US changes to DST ("summer time") next weekend
It's y2k two times a year. Yey
gilded cradle
Thanks
blissful pollen
I am glad, for good or bad, Canada flipped to match the US when they changed to avoid mass confusion
onyx hinge
OK, wild and I hope to disprove this theory .. but is the failure not happening when at a repl prompt??
(web workflow reading /sd)
lone axle
Does anyone know if this syntax for string concatenation across multiple lines of code is intended to be supported by CircuitPython / Micropython?
str_val = (
f"firstline"
f"secondline"
)
print(str_val)
I'm noticing that CPYthon allows this but I'm getting SyntaxError: invalid syntax on CircuitPython
slender iron
OK, wild and I hope to disprove this theory .. but is the failure not happening ...
Listing it fails completely?
slender iron
Does anyone know if this syntax for string concatenation across multiple lines o...
I think its a known issue. Not sure what issue is for it though
onyx hinge
@slender iron specifically I'm testing getting the content of a file I know exists on my sd card, /sd/cmdline.txt via curl
slender iron
<@252717193496756235> specifically I'm testing getting the content of a file I k...
that should work as long as the sd cart is mounted via code.py
lone axle
Hmm, do you know if it's known issue with the slash syntax as well?
timestamp = f"firstline" \
f"secondline"
no, I'm not sure. could be f-strings only too
onyx hinge
@lone axle yeah f-string concatenation is not in circuitpython
also not in micropython unless they added it recently
lone axle
Or a better question might be what is the best or recomended way to shorten a very long string that pylint complains about?
onyx hinge
you'll have to use a pylint disable
if the problem is that your expressions are long, you can assign them on earlier lines so your ultimate f-string is shorter, like f"{a}{b}{c}"
short tendon
you can also use old-school .format()
onyx hinge
instead of f"{this is complicated}{extremely complicated}{I have dotted names or function calls}"
lone axle
yeah, they are basically a bunch of these: {time.localtime().tm_mon} they could be applied to variables with shorter names before being put into the string.
short tendon
@onyx hinge I would love (LOVE) to get isort, some update pylint global disables and a generic tox in cookiecutter. Asked a question on the current isort PR. and curious if (if it's wanted) would be better as 1 PR or multiple
onyx hinge
@short tendon I'm not likely to be working on that right now. I don't have positive experiences with tox but the rest sounds good.
lone axle
you can also use old-school `.format()`
do you know if it's possible to do the :02 type thingy (I don't know the proper name of this formatting thing) with .format()
from something like this {time.localtime().tm_hour:02}
short tendon
"{:.2f}".format(time.localtime().tm_hour)
lone axle
Thank you!
solar whale
I am shocked that adabot allowed this f"{this.... 😉 somehow the term f-strings always makes me chuckle....
short tendon
<@1152454697614069811> I'm not likely to be working on that right now. I don't h...
When you have a few minutes can you check out the existing isort PR (no rush)? You made a comment on needing another global and wanted to see why and if there's a way around it
onyx hinge
@short tendon send me the link here and I'll give it a few moments
this is not ringing any bells for me at the moment, sad to say
slender iron
Here is the notes document for next Monday’s CircuitPython Weekly Meeting. It is at the normal time of 11am Pacific / 2pm US Eastern here on Discord. Add your hug reports and status updates to the document before the meeting. If you are unable to attend but would still like to contribute, feel free to add your notes and we’ll read them off during the meeting. Hope to see you there! <@&356864093652516868> https://docs.google.com/document/d/1wyHuyLhHVop15ZEMZg2WSXNp-kdvO98wNbPvHk5b6WA/edit?usp=sharing
onyx hinge
@short tendon yeah I have no idea at this point. sorry 😕
short tendon
okay. I'll try a few and see
slender iron
I am shocked that adabot allowed this `f"{this....` 😉 somehow the term f-str...
Yup! Similar to https://twitter.com/mariatta/status/1051942829516189696
Twitter
onyx hinge
@slender iron ^^ I have no idea the fix, but I think I get the cause now.
I instrumented a bunch of code from workflow -> vfs -> oofat but you can easily just add a print for failing to lock the bus in sdcard read to demonstrate this is the source of the failure
slender iron
<@252717193496756235> ^^ I have no idea the fix, but I think I get the cause now...
I think you are on the right track. I don't know what the right fix is either...
I hadn't thought about sharing the SPI bus
onyx hinge
this feel like a moment when having an RTOS would be nice
slender iron
ya, I think we're trending that way. not sure how that'd save you here though
onyx hinge
to make the camera demo work better we could have a mode that stops writing to the screen, it should work then.. but you have to know to enter this mode.
slender iron
or just lock out web workflow during display updates
onyx hinge
ya, I think we're trending that way. not sure how that'd save you here though
yeah the specifics are unclear to me, except that with tasks maybe lock_and_configure_bus could just block (yielding internally) in this case, requiring that "the other lock holder" was in a different task.
slender iron
tcp connections can wait
onyx hinge
or just lock out web workflow during display updates
this seems feasible, I can look into it
slender iron
you could try to be a little fancier and only do it if an sd card was mounted
tidal kiln
any known issues using make_transparent with OnDiskBitmap? (not working for me, imageload does work with same bmp)
slender iron
any known issues using `make_transparent` with `OnDiskBitmap`? (not working for...
not off the top of my head
@tulip sleet you looking at rgbmatrix now?
tulip sleet
was going to do that soon. Just doing reviews
tidal kiln
@slender iron maybe it's me? is this the nominal approach?
>>> import displayio
>>> from adafruit_pyportal import PyPortal
>>> pyportal = PyPortal()
>>> bitmap = displayio.OnDiskBitmap("color_test.bmp")
>>> bitmap.pixel_shader.make_transparent(0xffff)
>>> tg = displayio.TileGrid(bitmap, pixel_shader=bitmap.pixel_shader)
>>> pyportal.splash.append(tg)
@lone axle hey. any chance you've used make_transparent with OnDiskBitmap? (see above for more context)
lone axle
<@382939733107408897> hey. any chance you've used `make_transparent` with `OnDi...
I'm pretty sure that I have done it successfully in the past. Can you share this .bmp I can try a few things on my end to see if there was some gotcha that I am forgetting.
blissful pollen
<@252717193496756235> maybe it's me? is this the nominal approach?
```python
>>>...
It's been a year, but I recall looking at that doing OnDiskGif and the code looks correct to me. I have to run out but I can glance through it later if no one has figured it out
tidal kiln
@lone axle thanks. there you go. pretty simple test check. just make one of the squares transparent. (i was trying for white)
lone axle
What color are you intending to make transparent?
okay, I'll give it a try, thank you.
tidal kiln
but also tested 0xf800 for red, which also worked with imageload
>>> import displayio
>>> from adafruit_pyportal import PyPortal
>>> import adafruit_imageload
>>> pyportal = PyPortal()
>>> bitmap, palette = adafruit_imageload.load("color_test.bmp")
>>> palette.make_transparent(0xf800)
>>> tg = displayio.TileGrid(bitmap, pixel_shader=palette)
>>> pyportal.splash.append(tg)
^^ that works as expected
lone axle
but also tested `0xf800` for red, which also worked with imageload
this one works to make the black transparent with ODB. I'm not sure about the white though still looking into that I would have expected yours to work, but perhaps the colorspace conversion doesn't come out quite to 0xffff
bitmap = displayio.OnDiskBitmap("/color_test.bmp")
bitmap.pixel_shader.make_transparent(0)
I believe it's ColorConverter, but I'm not sure. ODB creates it internally
tidal kiln
it's colorconverter. i checked via type.
lone axle
@tidal kiln I'm not entire sure how to check it, but I think that original .bmp might be in a weird color space perhaps? I opened it in gimp and re-exported it choosing 24 bit RGB888. with this one I can use bitmap.pixel_shader.make_transparent(0xffffff) and it makes the white part transparent
notably it's a few kb larger as well.
slender iron
does imageload also return a colorconverter?
tidal kiln
i think so. let me double check though.
tidal kiln
>>> import adafruit_imageload
>>> bitmap, palette = adafruit_imageload.load("color_test.bmp")
>>> type(palette)
<class 'ColorConverter'>
>>>
interesting, the annotation isn't quite right in that case. It does also allow you to pass a PaletteConstructor to it in order to control which type it will use.
stuck elbow
yeah the annotations in imageload are all wrong
I was looking into fixing them, but that requires some advanced stuff
and it would probably not be supported by ide autocomplete, so I see little point
tulip sleet
<@329766224093249548> you looking at rgbmatrix now?
ok, turns out what I was stuck on was not a hang, but a mysterious lack of printing when the web workflow is enabled:
import time
import microcontroller
print("sleep for five seconds")
time.sleep(5)
print("reset in five seconds")
time.sleep(5)
microcontroller.reset()
if the web workflow is enabled
CIRCUITPY_WIFI_SSID = "myssid"
CIRCUITPY_WIFI_PASSWORD = "mypwd"
CIRCUITPY_WEB_API_PASSWORD="something"
then the first print statement's output doesn't show up.
if the web workflow is not enabled (I renamed CIRCUITPY_WIFI_SSID, then the first print shows up
i am not sure that has anything to do with the original issue, but the original issue's test program does a reset, like the above, and I was confused by the missing print for a long time due to trying to instrument the test program
slender iron
shows up on usb?
maybe the web workflow wifi connect is delaying usb enum enough that it doesn't get going in the 10 seconds you are giving it
tulip sleet
shows up on usb?
right, showing up on USB. I am not connected to the web workflow. Your theory makes sense
slender iron
wooo, went down the tlsf rabbit hole and found a bug
made a PR to the esp tlsf implementation too
short tendon
I had a thought over the weekend.
In CPython, there is os.environ which returns a dictionary of the users env vars.
Since we are doing os.getenv(key) = settings.toml[key]. what if we returned the whole settings as a dict in os.environ?
This way it could help people move from secrets, since it could be the same dict object
manic glacierBOT
onyx hinge
@short tendon or we could make os.getenv some kind of "environment proxy" that calls getenv behind the covers (but it would need some kind of implemention of keys() then probably).
A lot of the current implementation is designed around the idea that the environment might be big compared to RAM, so it's all processed incrementally. That idea might be wrong.
short tendon
<@1152454697614069811> or we could make os.getenv some kind of "environment prox...
If there are already "ideas", I'm completely out of the know. The settings.toml I have are all just secure keys I wouldn't commit. I think the biggest I have is 10 keys. Everything else is either in code or a separate config.json that I load and parse. I would agree if settings.toml could be large, this wouldn't be a good idea...
onyx hinge
If there are already "ideas", I'm completely out of the know. The `settings.toml...
ah I think I have the story wrong 🙂 we wanted to be able to read the environment file from C code without doing any allocations at all, a requirement during the 8.x days for stuff "outside the VM", like setting up wifi workflow.
short tendon
ah I think I have the story wrong 🙂 we wanted to be able to read the environmen...
Ahh, but isn't it for both?
- knowing WiFi
- safely storing creds and not committing them
onyx hinge
I'm trying to tell you the story of why it's just fetching single values, not all of them
for context, not because that's the best way forever
short tendon
Ahhh. I think we can have both. I don't think adding os.environ would replace os.getenv or even that the later would use the former...
onyx hinge
another thing about os.environ is that it brings mutability. os.getenv is immutable. would a hypothetical CP os.environ be a read-only dict, would edits disappear at the end of a session, would they modify settings.toml?
short tendon
I know in CPython, the changes only exist in the session. And don't change the actual env vars. But it might be a learning curve with people thinking it would...
manic glacierBOT
manic glacierBOT
lone axle
I was not involved as much back when it occured, does anyone know if there was some automated utility used when pylint and black were first adopted by the libraries? The adabot automation utility can update the config and infrastructure files that are the same accross all libraries. But AFAIK it can't really make modifications within the code or examples of the library which are different in each of the libraries.
Did we have some utility that was able to run pre-commit after the config files were added and then commit and push the resulting changes? Or was that dones manually by people back then?
empty salmon
@slender iron - I have narrowed down my "epaper not showing errors" problem but I can't find a solution.
In my board.c board_init() I setup the epaper and I set its power pin HIGH. When I test the pin in CircuitPython I show HIGH and then show LOW. Something in the core is resetting the pin value.
What I am trying to achieve is to set a power pin for the display as output HIGH in the board_init() so the epaper has the normal default behavior but expose the pin to CircuitPython so the user has the option to set it to LOW if they want to disable the display.
onyx hinge
@empty salmon I recently had to handle a case somewhat like this: the Adafruit Feather RP2040 USB Host board has a GPIO to control the 5V power rail to the USB "A" connector. We wanted the regulator enabled at power on but still wanted to allow it to be controlled as a DigitalInOut from Python code. I added a new facility so that a board could override the state a pin is set to when it is "reset", including at power on. You can check out the PR https://github.com/adafruit/circuitpython/pull/8937 or where it's used https://github.com/jepler/circuitpython/blob/6fd89b7b13d68d052f1046d91c356c2971f8331a/ports/raspberrypi/boards/adafruit_feather_rp2040_usb_host/board.c#L37
So basically you can set the state for the pin to have whenever it's not in use from Python code, rather than "pad disabled, no pull up/down".
empty salmon
@onyx hinge - thanks! I have it working 👏🏼
I did find and odd anomaly in CircuitPython.
If I do nothing in code.py then the error appears on the ePaper as expected.
If I add control = digitalio.DigitalInOut(pin_number) then it's as if I have turned that pin OFF and there is no error output to the display.
This mostly makes sense since I assume CircuitPython has reinitialized the pin. However, if I then print(control.value) it returns True but no error output. It is as if the value managed by CircuitPython is out of sync with the actual value of the pin in the core.
manic glacierBOT
empty salmon
BTW: I should note that in board_reset_pin_number() I don't want it to do anything to the pin. I want it to remain in whatever state is was last in (either from board_init() or from the user).
onyx hinge
<@426200547024961539> - **thanks!** I have it working 👏🏼
I did find and odd a...
If the ePaper power pin has been turned off at any point, including any "glitch", then the display will be in its power on state and require the initialization sequence to be sent.
on rp2040 a DigitalInOut is initialized with gpio_init which operates in the following steps: ```c
void gpio_init(uint gpio) {
gpio_set_dir(gpio, GPIO_IN);
gpio_put(gpio, 0);
gpio_set_function(gpio, GPIO_FUNC_SIO);
}
This could be causing a glitchYeah. Your code comment was great. I have it working as needed.
It's just the value shown in CircuitPython which is wrong.
onyx hinge
I misread what you said
empty salmon
Mostly my fault there. I didn't tell the whole story in the first post.
I initially the pin in board_init(). Then I prevent changes to that pin using board_reset_pin_number().
onyx hinge
Is there an external pull resistor on this pin?
empty salmon
good point. NO
onyx hinge
(what HW is this? is it a public design?)
empty salmon
board design not yet published (hopefully in May)
onyx hinge
then you should probably set the internal pull resistor high or low before checking the value
empty salmon
it's an output pin.
onyx hinge
print(control.value)
```The two bits of code you told me were these lines, which will configure `control` as an input with no pull.DOH! yes, as soon as the CircuitPython code get the pin, it defaults to input 🤦🏼♂️
thanks. Now it makes sense. Easy to document.
onyx hinge
good, I'm sorry we were talking past each other for a minute.
empty salmon
Totally my fault. You have been a HUGE help ...
onyx hinge
you're welcome
empty salmon
Thankfully, you had already completed the PR for that change 👍🏼
tulip sleet
@slender iron are you going to make a new circup release with the web workflow support? I can do this quickly
all set - just need to push the button
slender iron
I was going to let someone else
but no concerns from me. I think it'll enable more dev on circup
tulip sleet
done!
slender iron
just about to get back into wdt debugging
tulip sleet
i have an nrf thing to debug
slender iron
idf 5.2 and 5.1.3 were released
short tendon
@tulip sleet / @slender iron did you have thoughts on my circup update PR? I have conflicts now. Happy to fix, but only if you want the change
tulip sleet
<@329766224093249548> / <@252717193496756235> did you have thoughts on my circup...
I definitely want the change. Sorry about causing conflicts. I thought I had merged yours already!
tulip sleet
i was just looking at the release notes
slender iron
I'm hoping the debugging fix in 5.1.3 will help
I've got a pretty good understanding of tlsf now if you want to talk it over
before it leaks out of my brain
short tendon
I definitely want the change. Sorry about causing conflicts. I thought I had mer...
Updated. I retested, but would appriciate another
slender iron
@tulip sleet I think I found the matrix issue. no need for you to look into it
basically the spi flash cache check is imperfect
I think we're getting an interrupt when the disable is starting and checking to see if it is disabled before it takes effect
and then calling mp_hal_delay_us which is in flash
tulip sleet
makes sense. That might also explain why pruning random things out of the program did or did not make much difference. should the delay routines be in RAM anyway because they might be used in an interrupt routine, like OneWire or somethign?
slender iron
that's my fix for now
and replying to may forum thread about it
updated to 5.1.3 too
tulip sleet
that's my fix for now
The "in case TinyUSB runs in a separate task" addition: did you encounter that case?
solar whale
I was going though my stuff and found a Metro M4 Express Beta board. Is it still supported? That is, will the Metro M4 Express build work on it?
I ask because I am having some issues.
tulip sleet
all the ones I have say "Beta" 🙂 what color is it?
solar whale
black
tulip sleet
update the bootloader if you have not done so already
solar whale
I'm happy to toss it. But wanted t see if it is worth persuing.
tulip sleet
but it might be bad flash
solar whale
ah -- I'll try the bootloader
tulip sleet
do the files actually run, or do they just appear to have been copied? and you did a storage.erase_filessystem() ?
solar whale
I was just trying to recall if there was a know issue with the betas.
Yes, it runs - but then if I RESET the board, it is nt there any more.
tulip sleet
they said "beta" for at least a few years. They were not really beta
yes, try the bootloader first
solar whale
OK -- thanks
tulip sleet
and also reformat CIRCUITPY
solar whale
@tulip sleet It seems to be happier now -- with the updated bootloader and the erased file system....
And I thought I was going to have an excuse to get rid of something 😉
manic glacierBOT
@anecdata Thank fro testing. HTTPS server works on ESP32-S3 fro me too, although I successfully get the response, maybe the problem is due to self-signed cert? When I added -k to curl if works, but this is not really the issue with CircuitPython or library.
The TypeError is also present, which is a result of (I believe) incorrect types being returned by SSLSocket.accept():
tulip sleet
https://adafruit-playground.com/u/justmobilize/pages/adafruit-connection-manager
Give a minimum version here: CircUp has recently been updated to add missing requirements on update make sure to install the most recent version (also fix run-on sentence)
I usually add empty parentheses when I mention functions or methods, e.g. requests.set_socket().
Change More to a more descriptive heading.
RIght now this is the "new" way, but it will quickly become the "normal" way. So maybe put the punch-line first, and describe the old way later.
and you can put comments in the "old" examples that say "deprecated way of ...."
so people don't just copy them blindly
This is a nice summary -- the above are minor!
short tendon
Give a minimum version here: _CircUp has recently been updated to add missing re...
Thanks! Updated
crimson ferry
did microcontroller.watcdog.timeout change from seconds to milliseconds on espressif between CP 8 and CP 9? readthedocs doesn't reflect any change
tulip sleet
did `microcontroller.watcdog.timeout` change from seconds to milliseconds on `es...
not intentionally! If the API that we call changed something might have gone wrong
crimson ferry
#7718 was a refactoring, but I don't see a mention of changing the scale
tulip sleet
the timeout is a float of seconds
crimson ferry
also```py
w.timeout=4294967295
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ValueError: timeout must be <= 4294967295
I'll test some more and file an issue if I can't find another cause
tulip sleet
also```py
>>> w.timeout=4294967295
Traceback (most recent call last):
File "<s...
This looks wrong both before and after: https://github.com/adafruit/circuitpython/pull/7718/files#diff-d40fe0f9108540afc8b776d5f3294b72ef8826917062ce458558bffb9b398119
it's setting a time in msec from a float secs. There's a cast on the left and no cast on the right, but it should be casting the float to a uint32_t in either case, and it should be multiplying by 1000, but it's not.
crimson ferry
it used to work (e.g., 8.2.8), but now a timeout of 300 resets in a few seconds
OK, I'll file an issue with a simple test case
vital matrix
[low priority, just for fun question] does someone happen to have a "skeleton" template (or something analogous to ports/minimal in micropython) to start off a new port? I.e., something that just builds but has stub C code for everything? (Would like to side-step the rabbit hole of understanding how the build makefiles are put together, if there's an easier path 🙂
tulip sleet
[low priority, just for fun question] does someone happen to have a "skeleton" t...
That would be an interesting thing, but do you mean stub code for digitalio, etc? Another thing to do is to take an existing straightforward port, like atmel-samd, and gut the implementation.
that is what I did in the past
vital matrix
That would be an interesting thing, but do you mean stub code for `digitalio`, e...
Thanks for the tip @tulip sleet ! - I picked litex as it had the smallest number of characters in the Makefile 🙂 and was lost in the Makefile voodoo that was generating the various genhdr dependencies; I'll give atmel-samd a go!
(and yes, this was mostly an experiment to see if I could create a small test-bed to conveniently test some of the native modules. I tried @onyx hinge s fork, but it was a WIP and wasn't able to get a successful build, and didn't want to bug them)
tulip sleet
all the ports have some idiosyncrasy or another. atmel-samd is the most straightfoward, but it is dependent on a rather messy very-compile-time-configured hw-specific layer, ASF4, from MicroChip. The nrf port is also pretty straightfoward, but has some unusual stuff to support the firmware blob library known as the "SoftDevice", and has some unusual stuff for BLE. STM has to support many chip variants that are slightly different, and has a lot of copied boilerplate from ST. mimxrt10xx has an unusual memory architetcure. raspberrypi has the pico-sdk, which is quite straightfoward. But, it depends on some cmake stuff. etc.
vital matrix
Ah, interesting - thank you for the overall guidance on these ports; that's very helpful to know maybe where to dig for info. I forsee a non-zero amount of Makefile trawling for this experiment...
vital matrix
(only if it's a quick answer off the top of anyone's head) any hints on where this QSTR actually gets defined? I didn't see it in the "standard" set added in makeqstrdata.py https://github.com/adafruit/circuitpython/blob/e64082d60052f92720031b5a17aec3732a6c4e26/main.c#L220
tidal kiln
are you looking for the def of the macro MP_OBJ_NEW_QSTR or the def of MP_QSTR__slash_lib ?
vital matrix
I'm having a compile error for a "minimal" port I'm trying to get working, with that symbol missing. I'm sure I'm probably missing a magic preprocessor or makefile variable somewhere, mostly looking for hints on how to debug something like this. (Happy to throw up the Makefile somewhere if helpful)
slender iron
I'm having a compile error for a "minimal" port I'm trying to get working, with ...
I'd suggest starting with a port most similar to the platform you are targeting
that way the compiler stuff would be the same
the problem with having a minimal port is that it won't get actively used at all
vital matrix
the problem with having a minimal port is that it won't get actively used at all
Ack - this is definitely something mostly for my fun time to be sure... though, I had some thoughts of eventually being able to run tests with it.
slender iron
what platform are you targeting?
vital matrix
just unix - I'd like to be able to stub out some of the hal pieces so I can test some of the native modules off device
slender iron
I was thinking that a renode port could be interesting
vital matrix
Ah, very cool - I wasn't aware of this. I'm guessing it also lends itself to a continuous integration workflow?
slender iron
ya, I think that's the idea
and aligns with your goal of testing native hardware modules
vital matrix
Fantastic - thanks for the pointer; that likely makes a lot more sense than what I was playing with.
slender iron
😄
renode could be really helpful for tracing and memory analysis too
qemu is another option but I think it isn't really microcontroller focussed
vital matrix
Yes, I'd actually gone some way towards qemu (and seeing if wokwi might be amenable to at least esp32 emulation) but the renode idea is a great pointer. More toys to play with 🙂
slender iron
I talked with the wokwi person about it too
and someone here was thinking about using it for capturing memory footprint of library code over time
vital matrix
Very cool. Uri seemed like a nice guy in the few interactions I had with him, hope there's some interesting collaborations to look forward to 🙂
slender iron
ya, unfortunately we haven't had the time to fully engage with him (or the renode folks)
mortal kernel
Renode looks very interesting. Browsing its source tree I see it has verilated (Verilog compiled to C) support for a few riscv variants. That should yield more accurate emulation than you'd get with something like emu.
slender iron
I like that renode is CI oriented too
slender iron
@tulip sleet are you doing #9010?
tulip sleet
<@329766224093249548> are you doing #9010?
I can - it's simple.
slender iron
I don't have any other bugs to work on
so I think its time for an rc
the usb ones aren't super critical
(unrelated) I wonder if switching to picolibc could free up space on samd21 builds
tulip sleet
just going to grocery store. The advertising storage problem is I think an issue with not locking ringbuf when advertisements are saved to become ScanEntry objects.
ringbuf was sort of thread-safe (not sure it really was) a long time ago. I took that away during a big cleanup of ringbuf. it was kind of an accident and I didn't realize it was used in that way for scan entries.
anyway probably not hard to fix
slender iron
that explains it
tulip sleet
so did 5.1.3 really grow the builds or was that due to the IRAM flags?
tulip sleet
so an rc wuld await a few more fixes
tulip sleet
the macOS serial bug is a nuisance for those who encounter it, but i guess macOS has multiple problems right now
slender iron
ya, and it'll get fixed
I was thinking we should release because of all the existing fixes
so it could be beta or rc
and I feel like we're close enough to call it rc
tulip sleet
i would jave said that an rc really is an rc, without further expected fixes. Usually I convert an RC to final. are you thinking we would wait for the tinyusb fixes before the actual final?
mortal kernel
The best we'll do on #8824 is a workaround. As @tame creek has found, there's a hardware issue at play.
slender iron
no, I think we could do final with those bugs
@mortal kernel are the pi folks aware of the issue?
tulip sleet
if you want to do 9010 for lack of anything else, go ahead
slender iron
we'll likely do 9.0.1 quickly anyway because the stable audience will find more things
tulip sleet
I told them about the issue. I'm not sure it affects them very much. Maybe it makes UF2 loading slower
mortal kernel
tulip sleet
sorry - i told them about the macOS issue, not the uSB issue
tidal kiln
what's preventing PWM on A2, A4,A5 here?
Adafruit CircuitPython 8.2.10 on 2024-02-14; Adafruit Feather M4 Express with samd51j19
>>> import pwm_test
No PWM on: A0
PWM on: A1
No PWM on: A2
PWM on: A3
No PWM on: A4
No PWM on: A5
they have TC's avail per pinmux
output generated from this test:
https://learn.adafruit.com/circuitpython-essentials/circuitpython-pwm#wheres-my-pwm-2984449
(and truncated)
stuck elbow
what's preventing PWM on A2, A4,A5 here?
```
Adafruit CircuitPython 8.2.10 on 20...
iirc one TC is used internally by CP
I think TC6?
slender iron
@tidal kiln channel 0 is used to adjust frequency
so only TC[1] can be used for pwm
tidal kiln
ah, ok, so anything channel 0 would not be supported?
slender iron
correct
tidal kiln
thanks!
crimson ferry
if tCI was re-run partially (some boards), where are the previous artifacts (other boards)?
onyx hinge
@crimson ferry you can look though all the pull request builds. there does not seem to be a way to filter based on PR#. https://github.com/adafruit/circuitpython/actions/workflows/build.yml?query=event%3Apull_request
crimson ferry
ah, the run# hides in a menu
mortal kernel
@tulip sleet Thank you for #9014. It may be the answer to some odd behavior I was seeing with BLE scanning on my RP2 work. Time for me to re-base my branch.
slender iron
<@252717193496756235> I will start working on release notes
Thanks! 🚀
mortal kernel
@slender iron The upstream cyw43-driver MAIN just picked up the two outstanding PRs we had in our local fork. Should I make a PR to revert to the upstream repo now, or wait until they've tagged an official release?
Just took another look and they're in the process of releasing 1.0.3. I'll submit a PR once they've tagged it.
manic glacierBOT
To supplement the original report, here is a screenshot of the (recurring?) duplicate transmission issue.
Thanks for the build instructions and pointer tannewt! I'll start poking at the this sender side code. Hopefully I can make a little contribution here :)
gusty siren
Hello CircuitPython people. I'm starting to poke at https://github.com/adafruit/circuitpython/issues/3644. I'm wondering how to attach a debugger or if I can add in "prints" into the C code? I don't have an ST link or connector for for the SWD port... maybe I should get one.
midnight ember
Attempted to submit a very simple code comment edit on the ht16k33 library and it gave me a weird error
https://github.com/adafruit/Adafruit_CircuitPython_HT16K33/pull/117
https://github.com/adafruit/Adafruit_CircuitPython_HT16K33/actions/runs/8186264263/job/22384315239?pr=117
Artifact bundles has been successfully uploaded!
Run echo docs=$( find . -wholename './docs' ) >> $GITHUB_OUTPUT
Run sphinx-build -E -W -b html . _build/html
Running Sphinx v7.2.6
making output directory... done
loading intersphinx inventory from https://docs.python.org/3/objects.inv...
loading intersphinx inventory from https://docs.circuitpython.org/projects/busdevice/en/latest/objects.inv...
loading intersphinx inventory from https://docs.circuitpython.org/en/latest/objects.inv...
Warning, treated as error:
failed to reach any of the inventories with the following issues:
intersphinx inventory 'https://docs.circuitpython.org/projects/busdevice/en/latest/objects.inv' not fetchable due to <class 'requests.exceptions.HTTPError'>: 504 Server Error: Gateway Time-out for url: https://docs.circuitpython.org/projects/busdevice/en/latest/objects.inv
Error: Process completed with exit code 2.
candid sun
```
Artifact bundles has been successfully uploaded!
Run echo docs=$( find . -wh...
i just re-ran it and it passed 🎉 i've run into that before where an http request fails. i'll wait a bit, run it again and it's usually fine
midnight ember
i just re-ran it and it passed 🎉 i've run into that before where an http reques...
Was hoping it was something simple like that. Thank you for looking into it.
tulip sleet
Hello CircuitPython people. I'm starting to poke at https://github.com/adafruit/...
You can print in the console (REPL) channel by adding , for eample, mp_printf(&mp_plat_print, "i: %d\n", i)
old zealot
The files on the SD card aren't readable over USB. We hope to add that in the fu...
What would it take to make the SD card readable over USB. But before you answer that, Does CP expose the ESP-IDF WiFi_provisioning component?
tulip sleet
What would it take to make the SD card readable over USB. But before you answer ...
we are working on this here: https://github.com/adafruit/circuitpython/pull/6555. The problem has been getting it to work the same on all three major operating systems.
slender iron
Just took another look and they're in the process of releasing 1.0.3. I'll submi...
Thank you!
mortal kernel
Thank you!
NP, the PR is in. #9016 ready for review.
tulip sleet
@slender iron I finished draft release notes for rc.0. Only PR's #9018 (Feather C6) and #9016 (cyw43-driver update) remain as possible to include.
I would say we could hold off on #9016 just in case it's a regression for non-obvious reasons. #9018 is not urgent and could be added post rc.0.
slender iron
I approved the C6 one already
I think there are C6 fixes in idf 5.2 that we may want
tulip sleet
limor hasn't tested the C6 on proto hw
but it doesn't really matter, it can be fixed
tulip sleet
@slender iron so in that case, I could go with the current set of PR's and could do an rc.0 right now
slender iron
up to you
tulip sleet
it's ok by me. I can make a 9.0.x milestone and move some issues to there.
I forgot, I didn't do the CPX testing. I'll try that for a bit
slender iron
I see cmake files like this and wish it was just python: https://github.com/ARM-software/LLVM-embedded-toolchain-for-Arm/blob/main/CMakeLists.txt
tulip sleet
totally agree with that. I don't understand why people keep inventing new weak programming langauges
with idiosyncratic syntax
slender iron
this is interesting to me: https://github.com/aappleby/hancho
uses asyncio to make a build system basically
cross compiling is such a pain
goes to lunch
tulip sleet
@slender iron I am having trouble saving larger code.py's on CPX after about 9.0.0 alpha.4 or 5. Trying to come up with a simple test case. It just hangs and then crashes hard. Fine on alpha.2 and before
tulip sleet
First save is ok, then if I remove a few lines or put them back, the second write fails badly. Just going to try it with a non-code.py text file of similar size, and will try it on Feather M0 Express or something else more debuggable with J-Link.
tulip sleet
@slender iron I was able to reproduce the problem on 8.2.10, so it's not a 9.0.0 showstopper. It seems to have to do with do PDMIn and doing a lot of writes to CIRCUITPY at the same time.
I was testing an old CPX program that did audio processing.
manic glacierBOT
orchid basinBOT
vital matrix
[low priority] noticed that the python bitbanged i2c doesn't check for ACKs from the peripheral after the initial address ACK, compared to the C bitbangio. Before I go down this particular rabbithole and also recognizing the python module is probably not used much 🙂 was there any other issue that avoids the ACK check here? (context)
tulip sleet
manic glacierBOT
manic glacierBOT
vital matrix
is tests/run-tests mostly a MicroPython thing to test the VM? ( I'm trying to see if there's something in there I could run manually, to verify on a connected device.)
tulip sleet
is `tests/run-tests` mostly a MicroPython thing to test the VM? ( I'm trying to ...
it's for running on the unix port. It is not for testing on a board.
fringe yarrow
our of curiosity, does anyone know if anyone is maintaining the creationid github / website anymore? I'm going through the steps to begin working on support for a board and at that step, I'm stumped as to what to do to get a creation/creator id for this board, as the link on the github page to creationid.org just returns ssl errors, and the last commit was several months ago. Is there a new procedure to getting a creation/creator id these days?
tulip sleet
our of curiosity, does anyone know if anyone is maintaining the creationid githu...
just submit a PR to https://github.com/creationid/creators
@slender iron are you doing something with creationid.org?
I'm not sure there ever was a website, or maybe it was a redirect to the github repo which broke or something like that
fringe yarrow
just submit a PR to <https://github.com/creationid/creators>
I thought it was something like that. I'm not particularly experience with the collaborative development process, so I wasn't confident in what was required to submit such a pr
tulip sleet
I thought it was something like that. I'm not particularly experience with the c...
who is the manufacturer
fringe yarrow
who is the manufacturer
"Sunton" its this thing: https://www.aliexpress.com/item/1005006556177475.html
tulip sleet
i'll get back to you, I have a meeting in two minutes. seems like there are many clones of this
fringe yarrow
i'll get back to you, I have a meeting in two minutes. seems like there are many...
There are, and no two models would be board compatible, as they all use different screens and pinouts. It's maddening. I found a json dump that has information on most of the variants.. but I'm only working on the one I currently have for now.
tulip sleet
There are, and no two models would be board compatible, as they all use differen...
so you're considering making a board definition for it?
fringe yarrow
so you're considering making a board definition for it?
Yes, I've started working on the pin definitions and such, but in order to do test builds, I need that creatorid.. though I suppose for my own internal tests I could just use random data, as long as i dont submit it that way
vital matrix
it's for running on the `unix` port. It is not for testing on a board.
Thx @tulip sleet ! Lmk if you can use a PR for https://github.com/adafruit/circuitpython/issues/9021 it seemed fairly small. (I was also able to re-purpose run-tests to run on device for testing fixes quickly, but I'll keep that local for now; happy to submit something for that later if that's more widely useful.)
lone axle
Is there a description anywhere of technically what happens when I run a "make flash" command like:
make BOARD=m5stack_cardputer PORT=/dev/ttyACM1 flash
I'm interested in how it interacts or relates to the UF2 bootloader. Like is this also flashing the bootloader? or just CircuitPython? Is this flash using the UF2 bootloader somehow? or something different?
After I run this command it seems to succeed and then I press the reset button on my device once and it then reboots and comes up in the UF2 boot mode with BOOT drive. Is that the expected behavior? Is there some command I can run to just reboot into the UF2 BOOT drive without flashing? I've not been able to figure out any successful sequence of reset and boot btns that lead to the UF2 drive on this device. But it does always come up after a reset after this make flash command.
tulip sleet
Thx <@329766224093249548> ! Lmk if you can use a PR for https://github.com/adafr...
i have a fix to test, and am checking the other ports as well
i just need to move the claim_pin(), I think
maybe you saw the asme thing
vital matrix
Yes, that seemed to be the only thing I could see at least for rp2. (I also initialized ->pin = NULL as I wasn't sure if mp was setting the struct to NULL initially.
(I moved the entire struct block init in addition to claim_pin(), as the de-init is looking for pin != NULL to see if it should do anything)
tulip sleet
Yes, that seemed to be the only thing I could see at least for rp2. (I also init...
gc_alloc() zeros out newly allocated blocks
vital matrix
gc_alloc() zeros out newly allocated blocks
Ah, cool - gtk!
tulip sleet
Is there a description anywhere of technically what happens when I run a "make f...
This is the flash target in ports/espressif/Makefile. It is not for UF2's. It's just a convenience for flashing the .bin
i don't know if there's a tinyuf2 build (or a compatible one) for this board
lone axle
I used the one from UM Feather S3.
tulip sleet
Yes, I've started working on the pin definitions and such, but in order to do te...
yes, the creation id is not that important for just testing. I see a grand total of two sales of this board. I am a bit reluctant to add a bunch of clone boards that all differ slightly if they are not popular.
slender iron
<@252717193496756235> are you doing something with creationid.org?
I bought it but never set it up iirc
tulip sleet
I bought it but never set it up iirc
just a redirect to the repo would be easy
slender iron
Is there a description anywhere of technically what happens when I run a "make f...
The ESP bootloader tracks which OTA partition is live I think. I generally use flash without the UF2 bootloader. udisksctl mount -b /dev/disk/by-label/RPI-RP2 && cp build-adafruit_feather_rp2040_can/firmware.uf2 /run/media/tannewt/RPI-RP2/ can copy a uf2 from the command line if you just want to use the uf2 bootloader