One "gotcha" that I remember from when we did it the first time to watch out for is make sure that mpy-cross is works with any typing syntax that you end up adding. In the past there were some situations where some syntax was supported in CPython but not by mpy-cross so we had to leave it out of the circuitpython libraries at the time. I know we've updated mpy versions a few times since then though.
#circuitpython-dev
lone axle
spare jacinth
Im trying to use from __future__ import annotations and if TYPE_CHECKING where/when posible, so im hoping it wont break anything (or very little)
spare jacinth
ran into another typing problem now, mypy complains about doing len(ReadableBuffer) because the TypeAlias does not "implement"(no idea what's the set-theory term) Sized...
im no typing expert, but this seems to do the trick:
ReadableBuffer: TypeAlias = TypeVar(
"ReadableBuffer",
Sized,
bound=Union[
array.array,
bytearray,
bytes,
memoryview,
"rgbmatrix.RGBMatrix",
"ulab.numpy.ndarray",
],
)
```it currently is defined as: ```py
ReadableBuffer: TypeAlias = Union[
array.array,
bytearray,
bytes,
memoryview,
"rgbmatrix.RGBMatrix",
"ulab.numpy.ndarray",
]
... and same treatment would go for WritableBuffer i guess
can anyone confirm whether that looks right, before i open a PR for it? 😄
steady mulch
Hi, I was wondering about unloading modules to reclaim RAM. Deleting from sys.modules does seem to work to some extent, but isn't a supported API, and doesn't work for modules in a package that's otherwise still in use. Has there been any discussion about this?
slender iron
I wouldn't do it. you can reset the VM instead
steady mulch
We can't do that because of wanting to maintain other hardware state
slender iron
unloading isn't really a thing
steady mulch
It's not really a thing in desktop Python either but would have uses for embedded
Though I can understand if it's architecturally impractical
Or maybe it would be micropython's business
slender iron
I think it will save some memory but not all
interned strings will still be kept for example
brazen hatch
Hi, I was wondering about unloading modules to reclaim RAM. Deleting from sys.mo...
I have been importing and deleting modules from globals() and sys.modules since forever, with very limited success.
For my stuff I tend to just.. exec(compile(read())) in a more complicated way.
This method is the clunkiest of clunky, but can yield net 0 memory usage.
Splitting a big file into tiny files run like this makes stuff run at virtually no ram.
It wont work for libraries really though.
You can of course rewrite em to work with this method if you want though.
Ah also did I forget to mention it's slow? Like really slow?
Nope, espcamera upsteam is still using old adc api, so it's not happening yet. Imma shelve the code.
Me: Misclicks in S3 artifacts
The circuitpython build:
Actually, why am I not using the pirate version for everything?
spare jacinth
down to from 184 to 67, it has been an exhausting evening 🤓
devout jolt
If I wanted to do some "printf debugging" out a UART on the raspberrypi and espressif ports, what's the magic make incantation? (if there is one, maybe I'm mis-remembering)
spare jacinth
If I wanted to do some "printf debugging" out a UART on the `raspberrypi` and `e...
As in: enabling CP core's debugging prints or making Python-level print() go into UART instead instead of REPL (or both?)
If former, i would guess it is just something like DEBUG=1
devout jolt
As in: enabling CP core's debugging prints or making Python-level `print()` go i...
Sorry, yes. I'm building CircuitPython and doing tests in C. I believe DEBUG=1 gets turned into CIRCUITPY_DEBUG which I don't see doing anything specific beyond printing a backtrace in espressif port. I'm looking for concrete, known solutions, not "well it should work like this"
spare jacinth
Yeah, didnt mean it as an actual answer, but trying to specify what the question meant
devout jolt
If I wanted to do some "printf debugging" out a UART on the `raspberrypi` and `e...
Okay looks like it used to be UART_DEBUG_TX & UART_DEBUG_RX and is now perhaps CIRCUITPY_CONSOLE_UART_TX & CIRCUITPY_CONSOLE_UART_RX from https://github.com/adafruit/circuitpython/issues/5904 and https://github.com/adafruit/circuitpython/blob/main/supervisor/shared/serial.c#L145
alpine holly
Has anyone else noticed that SSL (loading web pages) works on esp32-c3 and s2 etc, but not on generic esp32 ?
tulip sleet
Has anyone else noticed that SSL (loading web pages) works on esp32-c3 and s2 et...
This was tested, so if you are seeing a problem, please open an issue with a small example. Test using 9.1.1 or the latest main version (or both).
slender iron
Okay looks like it used to be `UART_DEBUG_TX` & `UART_DEBUG_RX` and is now perha...
Yup, this is the way to do it. There is a console_uart printf for it too
slender iron
@tulip sleet are you going to get the MP merge out for review today?
tulip sleet
<@329766224093249548> are you going to get the MP merge out for review today?
I have a couple of builds that are too big. I'm going to look at those, but I could also forget about that now and just un-draft the PR and fix those later.
I spent a long time on a test failure that was a double bug and was ultimately unimportant.
slender iron
Go ahead and undraft. I'll look today. Not sure how much I'll be around later in the week
I'd like to get it merged since its been pending for a while
tulip sleet
me too!
slender iron
we can bug fix after
slender iron
should I back port the fix to 9.1.x instead?
tiny peak
@slender iron I'm looking for advice on general approach to tracking down USB stack bugs... I'm having trouble with writing USB gamepad drivers where stuff that works in PyUSB on Debian doesn't work with usb.core + max3421e.Max3421E on ESP32-S3. I'm seeing 0 or None Python values that make me suspect unhandled error conditions somewhere in C-land.
Context: I've had good luck with reading button press reports from an 8BitDo SN30 Pro USB gamepad (wired version) on CircuitPython. But, when I try to talk to an 8BitDo Universal Wireless Adapter 2, various weird stuff happens. For example, sometimes core.find(idVendor=0x045e, idProduct=0x028e) will give me a device, but then when I try to read the .idVendor or .idProduct, they come back as 0. Or, sometimes I will catch a usb.core.USBError that has an empty description string and None for its .errno.
From Wireshark captures, I know that this wireless adapter does stuff that might exercise the Max3421E or usb.core error handling. For example, it changes its vid:pid abruptly when a bluetooth gamepad connects. Also, when polling for gamepad button reports, the first read always gives an error (on debian PyUSB, it's errno 75, "Overflow"), but a second read immediately following that error usually works. When I try the double-read approach in CircuitPython, it never works.
To me, this smells like certain USB error conditions, perhaps at the max3421e.Max3421E level, or maybe higher up in usb.core, are not getting handled. Since the Max3421E has an interrupt pin, I worry maybe there's some complex stuff going on at the IDF level. Or, maybe it's an easy fix. I'm interested in looking into this (maybe a PR if I can figure it out?), but I'm not sure where to start looking. Do you have any advice on a general approach to tracking this down?
slender iron
<@252717193496756235> I'm looking for advice on general approach to tracking do...
You might want to try it with the fix I just posted
You can try replicating it from tinyusb code without circuitpython
tiny peak
You might want to try it with the fix I just posted
you mean this one? https://github.com/adafruit/circuitpython/issues/9486
tiny peak
Will a CI build for that turn up automatically if I wait a day or two?
slender iron
what board are you on? I can build it for you
tiny peak
oh wow, thanks. there's no rush. but, I'm using Feather TFT ESP32-S3
it's one of the 4MB flash ones where the .BIN file works, but the .UF2 method has trouble because there's not enough space on the CIRCUITPY drive for the firmware file
tulip sleet
deleting a semaphore I held
Yay! Yes, backport to 9.1.x
slender iron
Yay! Yes, backport to 9.1.x
the PR is for 9.1.x
tulip sleet
excellent
slender iron
@tiny peak adafruit_feather_esp32s3_tft
tiny peak
You can try replicating it from tinyusb code without circuitpython
by "replicating from tinyusb code without circuitpython", do you mean making an Arduino version of my code?
oh wow, cool! I'll try the .bin file. Thanks!
slender iron
I was thinking plain C. There is an examples folder in the repo
tiny peak
I was thinking plain C. There is an examples folder in the repo
oh, okay. I'll see if I can figure out what's up with the examples folder
slender iron
or you could print from within circuitpython too
tiny peak
do you have any guesses about which part of the Max3421E or usb.core code might be doing something weird if it happens to be the case that there's some kind of unhandled error condition around reading from an endpoint on a device that's been configured okay?
slender iron
are you getting the product string? https://github.com/adafruit/circuitpython/blob/main/shared-module/usb/core/Device.c#L115
tiny peak
are you getting the product string? <https://github.com/adafruit/circuitpython/b...
sometimes?
the pattern to a lot of this stuff is that something weird happens on the first try, but then it behaves better if you repeat the request
slender iron
https://github.com/adafruit/circuitpython/blob/87f7bc9fcc74e8ada4f255a8d7824f46a039995c/py/circuitpy_mpconfig.h#L473 will print tinyusb debug stuff to a uart too
tiny peak
With some consistency, I can get to the point where the device is configured and I can read from the endpoint, but then I get weird exceptions with no error message string
slender iron
the Device.c code I linked to doesn't do much error checking
tiny peak
<https://github.com/adafruit/circuitpython/blob/87f7bc9fcc74e8ada4f255a8d7824f46...
oh wow, CIRCUITPY_DEBUG_TINYUSB looks like it might be really useful. that should send stuff out on the TX pin then?
slender iron
I think you need to set the uart console too
because you don't want it going out usb
tiny peak
okay. I can work with that. I'll try that build you sent me and see if I can get any interesting debug messages from tinyusb. Thanks!
slender iron
you'll need your own build to enable the debug message stuff
tiny peak
okay... I figured out how to do that for some of the SAMD boards a while back, so hopefully I'll be able to figure it out for espressif.
I'll give it a try
slender iron
the building circuitpython guide has instructions for esp too
esp-idf is a bit different from samd
tiny peak
are they current? seems like I recall discussions here about all the docs being wrong?
slender iron
are they current? seems like I recall discussions here about all the docs being ...
ya, the building guide is up to date. it's the extending one that is out of date
tiny peak
oh, okay, fabulous
thanks again. I'm gonna stop here to copy some of these notes out into a non-scrolling-away-by-the-minute document, then try some of your suggestions.
slender iron
np 🙂
fading stirrup
Hey, @slender iron pointed me here. i have a P4 devkit and would like to start trying to port CircuitPython over to it. I've gotten to the point where it builds, but it's not running yet (invalid header on boot, I think it's the flash setup). I have the fork here. I'll keep working on it, but should I post this as a WIP PR? Does anyone else here have the P4 devkit and want to pitch in ? https://github.com/adafruit/circuitpython/compare/main...bwhitman:circuitpython:main
tiny peak
<@1220648290941669377> `adafruit_feather_esp32s3_tft`
I tried your alpha build. The USB weirdness still happened, and it seemed to act pretty much the same as with a normal 9.1.1 build.
manic glacierBOT
spare jacinth
Ideas about the __init__ question i posted in the PR's description? 😅
slender iron
Hey, <@252717193496756235> pointed me here. i have a P4 devkit and would like to...
Nice job getting it building! Go ahead and post a draft. I’m out today but will look when I have a chance
tulip sleet
@spare jacinth re the return type of __init__(). As https://stackoverflow.com/questions/46779046/correct-type-annotation-for-init points out, __init__() actually returns None. The constructor returns an object with the type of the class. We have been saying __init__() returns a typed object, but that's not really what's going on.
@spare jacinth oh never mind, it is already -> None! I was misremembering.
you can't call the constructor directly, but I wonder if you can call __init__() on an already-constructed Characteristic directly.
it's a bit confusing because the signature is being overloaded with the constructor arguments
spare jacinth
yeah, and the problem is that, since ClassName() is (internally):
class_name = ClassName.__new__(*, **)
class_name.__init__(*, **)
``` we apparently can't tell (and it is reasonable, because this is not happening on a Python, but C level) the type system that the second part of that will fail (but first one wont, as we can instantiate by other means)
.... or something like that 🤣anyway, I think it's fine to leave it alone as None
spare jacinth
so, as said... leaving it be, no red underline when (if) user mistakenly does a ClassName() and they will simply crash at runtime
tulip sleet
there is no pure python analog to what we are doing
yes, so I don't see how to mark it as an error
spare jacinth
so, as said... leaving it be, no red underline when (if) user mistakenly does a ...
(then they read the message and fix their code, and everyone is happy)
tulip sleet
ok, so I will go ahead and approve the PR, unless you want to do anything else
spare jacinth
is good for me, worst case i can always open a second one
spare jacinth
dang, i just found another one 🤣
spare jacinth
Here, prefixes is definitely not a ScanEntry (or is it? 🤔 ) but what is a mp_buffer_info_t
ReadableBufferIterable[ReadableBuffer]- else
//| class ScanEntry:
//| """Encapsulates information about a device that was received during scanning. It can be
//| advertisement or scan response data. This object may only be created by a `_bleio.ScanResults`:
//| it has no user-visible constructor."""
//|
//| def __init__(self) -> None:
//| """Cannot be instantiated directly. Use `_bleio.Adapter.start_scan`."""
//| ...
//|
//| def matches(self, prefixes: ScanEntry, *, match_all: bool = True) -> bool:
//| """Returns True if the ScanEntry matches all prefixes when ``match_all`` is True. This is stricter
//| than the scan filtering which accepts any advertisements that match any of the prefixes
//| where ``match_all`` is False."""
//| ...
static mp_obj_t bleio_scanentry_matches(mp_uint_t n_args, const mp_obj_t *pos_args, mp_map_t *kw_args) {
bleio_scanentry_obj_t *self = MP_OBJ_TO_PTR(pos_args[0]);
enum { ARG_prefixes, ARG_match_all };
static const mp_arg_t allowed_args[] = {
{ MP_QSTR_prefixes, MP_ARG_OBJ | MP_ARG_REQUIRED },
{ MP_QSTR_match_all, MP_ARG_BOOL | MP_ARG_KW_ONLY, {.u_bool = true} },
};
mp_arg_val_t args[MP_ARRAY_SIZE(allowed_args)];
mp_arg_parse_all(n_args - 1, pos_args + 1, kw_args, MP_ARRAY_SIZE(allowed_args), allowed_args, args);
bool match_all = args[ARG_match_all].u_bool;
mp_buffer_info_t bufinfo;
mp_get_buffer_raise(args[ARG_prefixes].u_obj, &bufinfo, MP_BUFFER_READ);
return mp_obj_new_bool(common_hal_bleio_scanentry_matches(self, bufinfo.buf, bufinfo.len, match_all));
}
static MP_DEFINE_CONST_FUN_OBJ_KW(bleio_scanentry_matches_obj, 1, bleio_scanentry_matches);
it is a readable buffer. It's usually a bytes or bytearray
spare jacinth
mypy says: adafruit_ble/advertising/__init__.py:353: error: Argument 1 to "matches" of "ScanEntry" has incompatible type "bytes"; expected "ScanEntry" [arg-type]
spare jacinth
it is a readable buffer. It's usually a bytes or bytearray
trying with adafruit_typing.ReadableBuffer now, then
that did the trick, nice
spare jacinth
lets not merge this one until i have everything solved locally 😛
orchid basinBOT
random junco
@slender iron That LWN article you mentioned from a few weeks ago is free now: https://lwn.net/Articles/984216/
spare jacinth
would love some testing/feedbacks here, TIA 🙂
https://github.com/adafruit/Adafruit_CircuitPython_BLE/pull/200
there are a couple of places with typing hard/imposible (at least with my knowledge) to represent....
however, i didnt feel like breaking API and just made a couple of hacks to (eg if TYPE_CHECKING: @overload blocks, assert, and a couple of type: ignore), but it's hopefully not too gross for a merge 😅
tulip sleet
https://blog.adafruit.com/2024/08/22/circuitpython-9-1-2-released/
Highlights of this release
- Fix Adafruit FunHouse build, which was not working.
- Correct pins for Adafruit Feather ESP32-C6.
- Allow wider range of counts for Counter, FrequencyIn, and IncrementalEncoder.
- Prevent crash or hang on nRF boards when stopping VM and BLE is not running.
orchid basinBOT
tulip sleet
https://blog.adafruit.com/2024/08/22/circuitpython-9-2-0-alpha-2351-released/
Highlights of this release
- Fix Adafruit FunHouse build, which was not - working.
- Correct pins for Adafruit Feather ESP32-C6.
- Prevent crash or hang on nRF boards when stopping VM and BLE is not running.
- Many new RP2350 boards.
- Update to Espressif ESP-IDF V5.3.
- Merge MicroPython updates from v1.22.2.
spare jacinth
only this is left to finish my PR fixing typing```
adafruit_ble/services/nordic.py:82: error: "Characteristic" has no attribute "read" [attr-defined]
adafruit_ble/services/nordic.py:94: error: "Characteristic" has no attribute "readinto" [attr-defined]
adafruit_ble/services/nordic.py:103: error: "Characteristic" has no attribute "readline" [attr-defined]
adafruit_ble/services/nordic.py:108: error: "Characteristic" has no attribute "in_waiting" [attr-defined]
adafruit_ble/services/nordic.py:112: error: "Characteristic" has no attribute "reset_input_buffer" [attr-defined]
adafruit_ble/services/nordic.py:116: error: "Characteristic" has no attribute "write" [attr-defined]
as far as i can tell, the problem is here: (removed some useless details)```py
class UARTService(adafruit_ble.services.Service):
_server_tx = StreamOut()
_server_rx = StreamIn()
# FYI: these two classes inherit from ComplexCharacteristic
def __init__(self, service: Optional[_bleio.Service] = None) -> None:
super().__init__(service=service) # <--- there's some magic here with getattr and settatr
self._rx = self._server_rx
self._tx = self._server_tx
sorry for the ping @slender iron but the magic code there is (at least partially) yours, and it is better and faster to ask than trying to understand what is going on there...
can you confirm whether that code does some things that are not obvious, and _rx and _tx end up being CharacteristicBuffer's? (i think that's the actual type because of the methods being called on them, but im not too positive on that either).
instead of reading/understanding, may as well just run an example and call type(uart._tx) but i dont have any hw to do so 😦
manic glacierBOT
orchid basinBOT
spare jacinth
only this is left to finish my PR fixing typing```
adafruit_ble/services/nordic....
that typing is better but still doesnt pass all checks 🤔 🤔
slender iron
only this is left to finish my PR fixing typing```
adafruit_ble/services/nordic....
I think that’s right. Complex characteristics replace themselves with another type
spare jacinth
only this one left 🙂 adafruit_ble/services/nordic.py:119: error: "CharacteristicBuffer" has no attribute "write" [attr-defined]
spare jacinth
by the way, for the readinto(buffer, nbytes) syntax to work i had to tweak comments once again (commit) but there's a comment:
static const mp_stream_p_t characteristic_buffer_stream_p = {
.read = bleio_characteristic_buffer_read,
.write = bleio_characteristic_buffer_write,
.ioctl = bleio_characteristic_buffer_ioctl,
.is_text = false,
// Disallow readinto() size parameter.
.pyserial_readinto_compatibility = true,
};
but... looking for pyserial_readinto_compatibility (or even readinto_compatibility) i only see it on the typedef struct declaration, and some objects setting the flag to true... is it not accessed anywhere? 🤔
spare jacinth
Tx is probably a normal characteristic
that type doesnt have a write either 🙃 , maybe a BoundWriteStream which does pass mypy's checks
descriptors are really a pain to reason about, they are very cool and come handy, but once there's a non-trivial usage of them.... my brain leaves the room
spare jacinth
that type doesnt have a `write` either 🙃 , maybe a `BoundWriteStream` which doe...
oh yeah, that would match {StreamIn, StreamOut}.bind return type, which is Union[_bleio.CharacteristicBuffer, BoundWriteStream] (depending on direction of the conn)
LGTM
manic glacierBOT
orchid basinBOT
manic glacierBOT
spare jacinth
while testing my changes to doc-comments, i've run make stubs and noticed that within its output there were several warnings for some untyped stuff, and im trying to get those sorted
on this function am i understanding correctly that it modifies the bitmap in place, and also returns a reference to it? in other words... should should it be -> displayio.Bitmap, -> None, else?
midnight ember
attempted to submit an api example to adafruit_requests and getting a new ruff error never encountered before. any ideas?
spare jacinth
not that im a ruff expert, but to me that sounds like a warning similar to:
hey, you've told me to run the check named "PLC2401", but i'm not running it because that's an experimental(preview) feature and you haven't enabled the preview mode
... or something similar
(and same for the other code names, of course)
lone axle
attempted to submit an api example to adafruit_requests and getting a new ruff e...
The warnings about preview can be ignored. The ruff-format hook that failed is basically the new version of black. So that just means that the format got modified when it ran the formatting tool. You should be able to just run pre-commit again and it will not fail the second time since it does the formatting automatically when it fails.
thorny jay
I wannabe a deepdiver @slender iron
(I saw a YT notification, so I have no excuse)
midnight ember
yeah i tried that and failed with the same message. i just sync'd my repo with requests. posted the file in there manually in my github website repo and it went through.
i'm running a custom pre-commit with powershell so my guess is i need to change something with it. :/
slender iron
I wannabe a deepdiver <@252717193496756235>
(I saw a YT notification, so I have...
Done!
spare jacinth
i'm running a custom pre-commit with powershell so my guess is i need to change ...
not too sure what your environment/setup is... but would likely be a better idea to develop under WSL than messing with powershell 🙃 😅
midnight ember
not too sure what your environment/setup is... but would likely be a better idea...
Oh I went through a lot of trouble to get that powershell script working with pylint, black, and pre-commit. 😉 https://adafruit-playground.com/u/DJDevon3/pages/github-desktop-crlf-to-lf-powershell-workaround
spare jacinth
yeah, i remember reading some of the messages back then... my point being that if your windows is recent enough to be compatible with WSL (thus the "i dont know your setup" part), i would use it (it's a builtin Linux "virtual machine") rather than messing with Windows<->Linux conversion issues running stuff on cmd/powershell
midnight ember
I know ruff is a recent addition and likely what is throwing off my script since I'm not accounting for ruff.
i tried WSL and it presented issues of its own. i'd rather just use windows.
spare jacinth
im now curious what those were 🤣
and they might have been easier to solve than the powershell hackaround
anyway, i dont see anything that looks wrong on your previous message, with regards to pre-commit's output
as foamyguy said, the pre-commit check "failed" because ruff-format introduced changes (aka: reformated some of your files)
unless those changes were some CRLF/LF changes, you just git add the files (or maybe they already get added after being format, idk) + try and commit again?
manic glacierBOT
spare jacinth
@tulip sleet the typing PRs i've made to main won't get punlished on PyPi right? not even when doing $ pip install --pre circuitpython-stubs
which branch should i target instead? 9.1.x? do i open a cherry-picking PR after the 2nd set of improvements gets merged?
tulip sleet
<@329766224093249548> the typing PRs i've made to `main` won't get punlished on ...
I would be inclined to have you just push to main and we'll wait for 9.2.0 final for the fixes to show up in the stubs.
I don't think it's that critical to have the stubs fixed for 9.1.x. You may disagree, if you use them!
spare jacinth
follow-up PRs to libraries will have to be held back for a bit, then 🆗
tulip sleet
We try not to churn the stable release too much.
spare jacinth
by the way, i added the missing hint for codeop.compile_command's return value as -> "code" because the type doesnt seem to be exposed to Python (or at least, i couldnt find where)
CI failed because it's not an actual thing, unsure what to do
tulip sleet
by the way, i added the missing hint for `codeop.compile_command`'s return value...
can you just omit it? we don't have to annotate everything, or does mypy complain bitterly
you could make it Any
spare jacinth
you could make it Any
yeah, that was my line of thought
spare jacinth
can you just omit it? we don't have to annotate everything, or does mypy complai...
i haven't run mypy on it, not directly at least... i just noted some [WARN] lines when i ran make stubs and fixed them 🙂
no idea which tool is used for that 🤣
tulip sleet
see circuitpython/Makefile stubs: target. It runs some scripts
in particular tools/extra_pyi.py, which uses ast
spare jacinth
as far as i can tell, the codeop.compile_command function is a tiny wrapper for mp_builtin_compile_obj, whose return seems to be of type mp_type_code.... and im guessing all C/Py types should be somehow import-able from Python level, right?
spare jacinth
see `circuitpython/Makefile` `stubs:` target. It runs some scripts
yeah figured that much... i mean i dont care what is being run, i simply wanted to silence it
Anyway, i will just omit proper typing
tulip sleet
Well, there is such a type:
dafruit CircuitPython 9.0.5 on 2024-05-22; Adafruit Circuit Playground Bluefruit with nRF52840
>>> import codeop
>>> c = codeop.compile_command("1+2")
>>> type(c)
<class 'code'>
yeah, i've gotten that far on my REPL too, but dont know where code can be imported from
tulip sleet
you can't import it, any more than you can import int
spare jacinth
was not in the "implicit namespace" (whatever the proper naming is), nor in builtins
tulip sleet
It's not visible in CPython either:
ython 3.12.3 (main, Jul 31 2024, 17:43:48) [GCC 13.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import codeop
>>> c = codeop.compile_command("1+2")
>>> type(c)
<class 'code'>
>>> code
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
NameError: name 'code' is not defined. Did you mean: 'codeop'? Or did you forget to import 'code'?
>>>
so, Any it is
tulip sleet
I don't see any type annotations that use code in CPython either
nobody seems to have bothered to write type annotations for the builtin functions/modules
or at least they are not in the documentation
spare jacinth
PR ready to go from my end, lets see if CI is happy as well
time for me to go bed, 4AM already here (:
and thx for the feedback!
tulip sleet
time for me to go bed, 4AM already here (:
good night!!
spare jacinth
is there a reason not to generate stubs for the micropython module? namely: prevent typing issues with from micropython import const, but there might be other types/functions in there that im not aware of and would also be nice to stub
ok, i think i've answered myself
- it comes from
py/notshared-bindings/(i first thought it was a CP-specific extension to access MP things) - micropython doesn't add comments to document the python-exposed API
- adding them would (probably) cause merge conflicts
- ...and keeping a separate
.pyifile doesn't sound like the best of ideas, as it could easily end up out of sync
any ideas on how (if) to handle this?
onyx hinge
@spare jacinth @tulip sleet ```>>> c = codeop.compile_command("1+2")
isinstance(c, types.CodeType)
True
I think the correct annotation is CodeType
tulip sleet
is that in CPython?
onyx hinge
yes
tulip sleet
sounds good
spare jacinth
Nice, thanks
I thought types wasnt in CircuitPy and didnt even check it, should have done it
Though it doesnt even matter as the import is on a pyi file which CircuitPython wont ever see 🤦♂️
tulip sleet
right, since CircuitPython totally ignores all the annotations
manic glacierBOT
clear halo
is there a reason *not* to generate stubs for the micropython module? namely: pr...
Just FYI, there's the micropython-stubber project:
https://github.com/Josverl/micropython-stubber/
And some relevant discussion of documenting more of MicroPython inline:
wary talon
Hello everyone, I'm so glad to join you all. Excuse my ignorance at this point. I have spent last 48hs studying and making tests with my CLUE devide. I have managed to use several of the examples and do some changes. But now, I'm was trying to use the adafruit_ble_broadcastnet to broadcast the pressure. I managed to broadcast but there's no sense in the hexadecimal values I get from the Bluefruit Connect App. I should be receiving something like 1030.98 hPA in hexa, but I'm unable to find how to convert it. I know that 2208 0303 0002 0A0A is all part of the manufacturer data and company, but, after that I do get 54 E0 80 44. 4 octetcs that should represent my flot of the preassure. And I'm struggling so hard because I make no sense of it. Would you be so kind as to give me some tips?
I did resolve this with a Govee devide with no problems, but with this Adafruit CLUE it feels I know nothing 😦
stuck elbow
Hello everyone, I'm so glad to join you all. Excuse my ignorance at this point. ...
you might want to ask this on #help-with-circuitpython, this channel is for the development of circuitpython itself
spare jacinth
Just FYI, there's the micropython-stubber project:
https://github.com/Josverl/m...
Found the project while Googling around, didnt see the discussion, will read it now. Thanks
wary talon
you might want to ask this on <#537365702651150357>, this channel is for the dev...
Thank you! I'll do that, sorry for the confusion!
manic glacierBOT
Huzzah, it worked! From S3:
Sorry to bypass the approval checking, but I thought I was going to have to debug this and run the approval merge several times.
tiny peak
@tulip sleet Last night I managed to get make working for the espressif port. Was able to successfully build and flash a firmware image for Feather TFT ESP32-S3 with your 9.2.0 alpha stuff. Only other documentation glitch I encountered was that the esp-idf/install.sh fights with pip3 install --upgrade -r ../../requirements-dev.txt over the version of setuptools. ESP-IDF wants version 71.0.0, but pip upgraded to version 73.0.1. Once I re-ran the install script, it was fine.
So, the procedure that worked for me was: 1) run install.sh, 2) run the pip install stuff, 3) run install.sh again
tulip sleet
<@329766224093249548> Last night I managed to get `make` working for the espress...
Yes, I think I did something like that, though maybe no intentionally. I normally run with a venv all the time in Ubuntu 24.04, but when I am doing espressif builds, I deactivate the vent. I think i ran install.sh, did pip3 -r ../../requirements-dev.txt, and then did the export.sh
tiny peak
Yes, I think I did something like that, though maybe no intentionally. I normall...
yeah, I was a little surprised to figure out that the espressif scripts set up their own venv. But, they do it in such a way that it doesn't add the (venv) $ prefix to my shell prompt, so it's not obvious that a venv is active.
I initially assumed that I would need to make my own venv as is often the case with other stuff, but that didn't work. So, then I read more closely and figured out esp-idf has its own thing in $HOME/.espressif
onyx hinge
@lone axle @slender iron let me know if you'll be able to read your usual parts in the State of CircuitPython section today.
lone axle
<@382939733107408897> <@252717193496756235> let me know if you'll be able to rea...
I will be. 👍
slender iron
@onyx hinge I'll be missing the meeting because I have family in town.
orchid basinBOT
onyx hinge
@slender iron OK, got it. @tulip sleet let me know if you can read the state of the core section today!
tulip sleet
<@252717193496756235> OK, got it. <@329766224093249548> let me know if you can r...
sure, no problem
onyx hinge
thank you!
tulip sleet
@onyx hinge https://github.com/raspberrypi/pico-feedback/issues/401#issuecomment-2310001946 can be deleted now. GitHub removed those users and their posts
onyx hinge
<@426200547024961539> <https://github.com/raspberrypi/pico-feedback/issues/401#i...
oh good
brazen hatch
I'd say idf5.3 is stable
onyx hinge
howdy all and in particular <@&356864093652516868> -- I'm back and I'll be running the weekly meeting in a little under 30 minutes! Please add your notes at https://docs.google.com/document/d/1AlQdjcpzMwWRF_2UsoXykCqwJVCPx5kSR14VG4ozEOo/edit?usp=sharing and just leave a little note if you need me to read them out for you. TTY soon
tulip sleet
@spare jacinth is your contribution in "In the Weeds" meant to go in "Status Updates"?
tulip sleet
@spare jacinth mic sounded crackly and went in and out, like a bad connection
spare jacinth
rebooting the headset now
gilded cradle
Thanks
lone axle
Thanks for hosting Jeff. Have a great week everyone!
onyx hinge
Here is the notes document for next Monday’s CircuitPython Weekly Meeting. It is at the normal time of 11am Pacific / 2pm US Eastern here on Discord. Add your hug reports and status updates to the document before the meeting. If you are unable to attend but would still like to contribute, feel free to add your notes and we’ll read them off during the meeting. Hope to see you there! <@&356864093652516868> https://docs.google.com/document/d/1BXYMbOxPWUug-6WwltLdcvqv7i_de3059noqBx0Sn-M/edit?usp=sharing
brazen hatch
Oh.. I got very weird idf breakage.
Under Beryllium OS, specifically on the cardputer, when the wifi is not connected, the standalone tty (kb input, display output) hangs on artifact PR9526 (idf pr) while it does not on the one before..
Furthermore, it's undebuggable, cause the moment I attach usb or another serial (uart or telnet) it immediately fixes itself..
Why does it always have to be I find the weird bugs..
I'll nuke usb serial and send stuff out with print() I guess
onyx hinge
oh that reminds me, there was a question around how to declare the type of the __init__ of a class that can't be constructed. It can be typing.NoReturn or typing.Never. This passes mypy --strict: ```import typing
class K:
def init(self) -> typing.Never:
raise RuntimeError("Cannot construct K instances")
NoReturn is probably better, as it is in 3.6; Never was added in in 3.11.
spare jacinth
i tried several things and none seemed to integrate nicely with both mypy and vscode's intellisense
maybe i did something wrong though
onyx hinge
yeah I only tested mypy
spare jacinth
also, from my understanding of the documentation, even though they both mean the same "do not do it, you will blow up", seems like Never is more for attributes/arguments, and NoReturn should be used for funcs
i even recall seeing some example as in def func(args: Never) -> None to mark that you should never fed arguments into the function... which is quite a weird syntax imo
onyx hinge
def never_call_me(arg: Never) -> None:
pass```it's true that there's an argument of type Never shown in the Python documentation. (https://docs.python.org/3/library/typing.html#typing.Never)yeah, and i can't see how that differs from def never_call_me() -> NoReturn which looks a lot more obvious
onyx hinge
I think that's just some type theory stuff, I can't see why it's useful. (because once you take away the int and str type possibilities from the type annotation int | str you have an empty range of type possibilities, or the so called "bottom type")
never_call_me does always return None, though.
spare jacinth
yeah, but hinted as "never call me" perhaps means that "function" is not real code but a stub (.pyi) and needed some "fake" body for being valid syntax (pass)
anyway, i guess both ways get the thing done... no big deal
but i really didnt like how NoReturn/Never integrated with __init__ and left it as is 🙃
onyx hinge
a lot of type annotations are very practical. the Never argument type seems much less so, but clearly important to theoreticians. https://en.wikipedia.org/wiki/Bottom_type 💤 😴
When the bottom type is uninhabited, a function whose return type is bottom cannot return any value, not even the lone value of a unit type
spare jacinth
brain boom right here
stuck elbow
that's quite irrelevant for python, though, because python's type system is not consistent or sound anyways
so you can't really reason about it using the category theory
slender iron
@onyx hinge were you looking at that bug too?
onyx hinge
@slender iron That's the one I'd intended to look at first (I think I said rgbmatrix in the internal meeting but that's the issue I meant) but I grabbed that analog sampling one first so I had not spent much time on it. Toe status: absolutely not stepped on.
blissful pollen
@onyx hinge Quick question, I'm looking at synthio for a random idea and general learning and wondering if you had any references for the direct digital synthesis it uses? If you don't have anything handy don't worry.
onyx hinge
@blissful pollen "direct digital synthesis" is just a fancy word for how synthio generates frequencies that don't exactly divide the sample rate. I don't have a good reference about it offhand
say you are at 44.1kHz sample rate, and you want to generate a 440Hz note with a square waveform made up of 2 data points in the sample. Each output sample you need to advance by 440*2/44100 ~= .01995... input samples. If you scale this by some convenient factor like 65536 and then round it, you find you need to advance by 1308/65536 samples.
Then each time you want to generate an output sample, you just add 1308 to an accumulator, subtract off any excess part of the number, and divide by 65536 to find out which input sample to use. (which is a bit shift so it's quick to do)
blissful pollen
Oh cool thanks. That makes sense. I kept seeing many DDS example all with sine waves but that is basically the waveform in synthio (by default the square wave). That helps a lot!
onyx hinge
https://github.com/adafruit/circuitpython/blob/main/shared-module/synthio/__init__.c#L227 is the code that does the "add, subtract off excess, scale, & index".
the rate calculation is dds_rate = (sample_rate / 2 + ((uint64_t)(base_freq * waveform_length) << (SYNTHIO_FREQUENCY_SHIFT - 10 + octave))) / sample_rate; on line 180
which today-me is glad that it was yesterday-me who figured THAT line out.
blissful pollen
Coincidentally that is the line I have opened and highlighted in my editor now trying to wrap my head around all this!
onyx hinge
(for MIDI note numbers; synthio_frequency_convert_scaled_to_dds is called for Note objects)
do you have a goal or project with this beyond better understanding?
blissful pollen
I have been thinking if it would be possible to do more reverb/echo effects, but that requires buffering more of the output to feed back into the loop. The 2350 may have the power to pull it off
onyx hinge
that would be cool. It would probably be a post processing step similar to the rudimentary IIR digital filtering, I'm imagining.
you might also look whether audiomixer could do it, especially since timchinowsky has contributed some work towards real time audio processing. if it doesn't have to be limited to synthio but could be done with mp3 files, sampled audio, etc., it could be pretty versatile
https://github.com/adafruit/circuitpython/pull/9421 & https://github.com/adafruit/circuitpython/pull/9438
blissful pollen
Hmm I had not thought about that. I was poking through that code to see how the whole audio train of components works and makes sense.
Doing it on a per channel in synth would be nice but nothing lost to try a more general cause in mixer and then it could be used for any audio played that way too.
brazen hatch
Print output buffer gets stuck when usb is not attached
No wdt, just waits
resumes instantly when plugged
This is on repl so I should be able to replicate on a feather with just this for loop
imma make the bug report
Huh, no it doesn't happen on the feather..
Weird, maybe the cardputer keyboard thing is confusing it?
manic glacierBOT
Not sure what I need to do here, github says I can't commit the suggested change =)
onyx hinge
I hope that works .. that was an annoying bug and it'd be nice to lay it to rest.
tulip sleet
I hope that works .. that was an annoying bug and it'd be nice to lay it to rest...
nice soft fail. thanks!
@onyx hinge @slender iron should we maybe have an issue triage and assignment meeting today or soon?
onyx hinge
I'm out a chunk of the afternoon to catch up with a friend but I should be back by 1h before the evening meeting
lone axle
Is it desirable to update the buttons for the bootloader updater UF2 files on this guide page? https://learn.adafruit.com/adafruit-itsybitsy-nrf52840-express/update-bootloader-use-uf2 They're currently pointing to 0.8.2 version of the bootloader with 0.9.2 being the latest release https://github.com/adafruit/Adafruit_nRF52_Bootloader/releases. If there isn't a specific reason the are held back I can update them to say and point to the 0.9.2 versions
tulip sleet
Is it desirable to update the buttons for the bootloader updater UF2 files on th...
yeah, I was just waiting for https://github.com/adafruit/Adafruit_nRF52_Bootloader/issues/334
tulip sleet
I'm out a chunk of the afternoon to catch up with a friend but I should be back ...
that would be fine
wraith crow
Weird, maybe the cardputer keyboard thing is confusing it?
I haven't had a lot of cycles to look at CP recently, but this sounds similar to https://github.com/adafruit/circuitpython/issues/9253
brazen hatch
I haven't had a lot of cycles to look at CP recently, but this sounds similar to...
Similar, but not really..
The cardputer accepts Ctrl+Alt+Del which force-reloads the code regardless of the program state (the core handles that event), and that doesn't work there, even though it works in every other state, indicating it hang.
I'll try to debug it in the core.
slender iron
that would be fine
2pm pacific works for me
manic glacierBOT
lone axle
@gilded cradle do you know if there is a known issue that leads to this error when attempting to select the host folder on the beta code.circuitpython.org site?
Also I started working on a fix for the duplicate serial messages after disconnect/reconnect multiple times and think I found a solution. But didn't notice until after I started working that there are both main and beta branches. Which one is the preferred one to submit PRs against?
manic glacierBOT
manic glacierBOT
orchid basinBOT
manic glacierBOT
manic glacierBOT
onyx hinge
Darn, my #9460 reproducer has still not reproduced a 2nd time. There must still be an unknown factor.
onyx hinge
it did finally crash again but this time with hard fault, not wdt
slender iron
looks like the docs build is broken under 3.12
slender iron
I'm looking into the RTD issue
looks like an autoapi update that broke us
it removes stdlib methods
tulip sleet
@slender iron so autoapi change is a big one?
slender iron
not big. just breaks us
tulip sleet
Maybe merge to 9.1.x and then we'll merge to main. I have another thing from 9.1.x too
slender iron
kk, lemme switch it
tulip sleet
I am about to make a fix for ESP32-C6 UART for 9.1.x.
tulip sleet
@slender iron github is having issues. The "Actions" item has disappeared from the horizontal list of things at the top, at least for me. For you, too? https://www.githubstatus.com/ is reporting problems
tulip sleet
ack, I was trying to disable actions on MY repo.
my fault
i was getting spurious micropython builds
slender iron
👍
tulip sleet
i wonder how we can trigger them again for the PR's
I amended my last commit, just changing the message slightly, and the push -f
slender iron
@tulip sleet this usb issue is in tinyusb. think I should switch us to a fixed repo or just wait for thach to merge in the fix? it is simply turning on an interrupt on resume
tulip sleet
<@329766224093249548> this usb issue is in tinyusb. think I should switch us to ...
which issue is it? We could ask him to merge a PR tonight, if it's kinda obvious.
slender iron
ya, as long as he's working it should be an easy merge
tulip sleet
is it something that should go in 9.1.x?
slender iron
this: https://github.com/adafruit/circuitpython/issues/9454 and probably the MSC version too
tulip sleet
i would put that in 9.1.x since it's a significant regression: might break macropad, etc, Is it just Espressif?
slender iron
i would put that in 9.1.x since it's a significant regression: might break macro...
yup, just espressif
tulip sleet
@slender iron I think only q is whether there are other changes we should know about that we would pick up with an update to master that includes this
is it a regression in TinyUSB or did it never actually work?
tulip sleet
ok, so then I wouldn't worry for 9.1.x after all
slender iron
https://github.com/hathach/tinyusb/pull/563/files adds the interrupt handler support but doesn't enable it (added four years ago)
tulip sleet
9.1.x fixes are for significant bugs, especially regresions, or new boards that need a stable release, etc. -- that's been my thinking
tulip sleet
@slender iron rtd jobs are still failing (maybe in a different way) in 9.1.x builds; it's confusing to me whether it's just failing because it's failing to fetch the badge?? WARNING: Could not fetch remote image: https://readthedocs.org/projects/circuitpython/badge/?version=latest [502]. That is actually fetchable when I try
not holding me up
orchid basinBOT
tulip sleet
https://blog.adafruit.com/2024/08/29/circuitpython-9-1-3-released/
Highlights of this release
- Fix Espressif ESP32-C6 UART creation.
- Adafruit Qualia ESP32-S3 RGB666: fix display tearing.
- Pull request merge builds for stable branches, not just the
mainbranch, are now uploaded to AWS S3. - Fix Espressif board-specific pin initialization. This fixes
board.NEOPIXEL_I2C_POWERinitialization on the Adafruit Feather C6.
sturdy crest
Any chance this will be available in a beta release soon?
onyx hinge
oh git & github, it's so easy to not get everything exactly right.
tulip sleet
Any chance this will be available in a beta release soon?
we're waiting for a few more PR's and then will make another 9.2 alpha or maybe beta release
onyx hinge
not getting very far on #9460 -- haven't actually had any crashes so far today.
I tried to determine whether it's a problem up in Python object land or down in lwip internals land. For instance, I totally bypassed actually calling lwip_accept and got 0 crashes in 1 hour. But since then I've gotten 0 crashes in about 3 hours with the original code. so it's not a conclusive test.
blissful pollen
I think I have figured this out but if someone knows to confirm, for audio buffers if it is stereo the buffer object is LRLRLR... vs mono would just be MMMMM? Then I2S can just take L/R to transmit of double M to MM to transmit out both channels?
stuck elbow
the chip I use lets you pick which channel to use
onyx hinge
@blissful pollen roughly yes, I think "standard" I2S always has to transmit 2 channels, so one way to deal with a mono sample is to just replicate it before outputting. the C function audiosample_convert_u16m_s16s does this and is used e.g., on espressif when plaing a mono stream on i2s. I don't see where rp2040 does this.
blissful pollen
<@228322543507865601> roughly yes, I think "standard" I2S always has to transmit...
Thanks. I had seen in the 2040 if you just pass 16 bits to pio it replicates them to the top 16 as well (as it has to deal with 32 bits) so you get a free “copy”.
Been trying to think through how audio effects could be built to work for any type of audio channel. So if I want reverb could add it to a mixer or synth channel or wav, etc. Have a long drive this weekend to think about it more
onyx hinge
oh yeah that data replication feature would do it
manic glacierBOT
manic glacierBOT
nimble sparrow
Has anyone ever broached adding the ability to redirect sys.stdout to a valid alternative stream?
Just skimming back through I don't see a discussion, though I'm sure it'll be somewhere buried.
stuck elbow
Has anyone ever broached adding the ability to redirect sys.stdout to a valid al...
you can use the second usb uart for this
nimble sparrow
To redirect stdout?
stuck elbow
to use it in place of stdout
you don't get the repl and error messages on it too, so it's easier to handle
nimble sparrow
Ah, right I see what you mean - no, I specifically want to redirect the repl. exec() by longstanding and perfectly sensible design doesn't return output like eval(). In CPython you can redirect stdout to any valid stream (for example, a file, or an arbitrary serial object).
stuck elbow
hmm, webrepl does it, but it's C
nimble sparrow
Yeah, I imagine it would be relatively straightforward under the hood in the firmware and I expect it's far less simple at the higher level
tulip sleet
Has anyone ever broached adding the ability to redirect sys.stdout to a valid al...
MicroPython has something called multiterminal which we inherited from upstream but which we eventually turned off /removed
nimble sparrow
Can't blame you for streamlining, my use case is undoubtedly niche
tulip sleet
there are ways to redirect it to a UART in addition via compilation options
nimble sparrow
Most of the time when you want to view the REPL other than via web or cdc you can just use terminalio
nimble sparrow
there are ways to redirect it to a UART in addition via compilation options
Ah right, this would be via BLE - I know finishing up BLE Workflow is on the cards eventually anyway so I may as well just be patient.
tulip sleet
Ah right, this would be via BLE - I know finishing up BLE Workflow is on the car...
there is already code to send the REPL back and forth over BLE. It is enabled by CIRCUITPY_SERIAL_BLE, which is turned on for nordic boards and espressif with _bleio
nimble sparrow
I may have misread the active PRs and Issues, thought there were issues being worked out for Espressif BLE
tulip sleet
that is true, I think, but it was working for nordic
i have forgotten which apps or browser apps support BLE REPL, if any
ok, it's available via https://code.circuitpython.org
but I haven't tried it
@slender iron @onyx hinge there is a new PR for esp-camera to switch to the new I2C driver API: https://github.com/espressif/esp32-camera/pull/677
nimble sparrow
that is true, I think, but it was working for nordic
Yeah, that's what I remembered too
Righto, no worries - and certainly not worth contemplating unearthing a crusty bit of MP/CPython
onyx hinge
<@252717193496756235> <@426200547024961539> there is a new PR for esp-camera to ...
that's promising
manic glacierBOT
nimble sparrow
This was a terrifying find, very grateful to @brazen hatch for clocking the minor difference in reg values!
brazen hatch
If anyone here knows battery chemistry please let me know. I can set it to charge all the way up to 4.2V, stopping only when there 0mA current draw from the cell.
But this means imperfect readings could charge the battery to 4.21V.
Should I set it to charge to 4.2 or 4.185 like I have it with the 25mA stop limit?
Do batteries even have an acceptable tolerance at 4.2V?
manic glacierBOT
I forgot to include the PNG and BMP spritesheets.
This is sprites.png (it's smol, 8x16px):
GitHub won't let me upload the BMP, but you can get it from the reproducer repository I linked to above.
nimble sparrow
Do batteries even have an acceptable tolerance at 4.2V?
From a bit of browsing it seems as though 4.2 is the safe cut off for LiPo chemistry unless explicitly binned as high voltage
brazen hatch
From a bit of browsing it seems as though 4.2 is the safe cut off for LiPo chemi...
I did test it on my unit and it took about 3 hours to finish charging (it was trickle charging).
Meanwhile the 25mA limit only took like 20 minutes trickling.
So judging by the fact no devices trickle for 8 billion years it's best to leave it at 25mA cuttoff.
PR is good to go.
wraith crow
I probably missed this when the pico2 first came out but when I build the pico2 firmware locally my device doesn't boot, no serial device, no CIRCUITPY drive. If I grab the "absolute newest" from circuitpython.org the device boots normally, I've been building espressif boards without issue but it's been a while since I've built an RP2040 board, is there a raspberry SDK update I have to do locally or something?
tulip sleet
I probably missed this when the pico2 first came out but when I build the pico2 ...
There is a pico-sdk submodule update but I don't think it would even build without that. The board page is here: https://circuitpython.org/board/raspberry_pi_pico2/
stuck elbow
I probably missed this when the pico2 first came out but when I build the pico2 ...
also make sure you have the right version of gcc
wraith crow
I did a make remove-all-submodules and then make fetch-all-submodules again just to make sure, the firmware builds but doesn't boot. I did download the absolute newest from the board page and that works fine. My gcc version is 11.4.0.
tulip sleet
we are using gcc 13.2rel1
wraith crow
okay, I'll try getting that updated 😁
tulip sleet
wraith crow
okay, it has been a long time since I looked back at the learn guides 😁
wraith crow
@danh @stuck elbow That did it! Thanks 🎆
tulip sleet
@danh <@351136009922019331> That did it! Thanks 🎆
interesting, and glad that it did. I wonder what is better in the newer compiler
wraith crow
I was actually using version 10.3 of the cross compiler, 11.4 was the native gcc....
I guess it's really been a while since I've built raspberry images 😁
tranquil ether
hello! I'm pretty far along on a big project and I would like to add some error logging. Sometimes I get hangs/crashes after long run times and I thought if I periodically log to file, I could investigate the crashes. SO, I looked at the website and they show that you need to change the file IO to (a) PC can write files to (b) CircuitPython can write files. And their recommendation is is to use hardware pin and some code in boot.py. I don't like the idea of adding a pin-read. Has anyone actually worked with this file switcheroo functionality?
Actually, I could use the push button that is already on the system. boot.py would read only on start-up, so my procedure would be to hold that button down on boot. Still I would really like to hear if someone has worked with this functionality
spare jacinth
i dont think anything would be more convenient than the button, but you could eg store a bit on nvm that controls whether computer or circuitpython gets access to the filesystem (being toggled on each boot)
tranquil ether
I already have an NVM read on boot and it has been working reliably for months now. I quite like it
spare jacinth
if the PCB was designed that way, you could perhaps check whether power comes from USB port (likely PC, but might be some other thing with a USB cable in it), or from some other input... to then configure filesystem accordingly
but yet again, and specially when you already have a button, i think that's the best solution
that's just my personal preference, ofc
tranquil ether
yeah, I think you are right. Thatnks for kicking this around a bit. I think I have a good way forward now
manic glacierBOT
lofty terrace
I was told that I should at least mention my CP issue here, since it "might be a regression in the socket code" (I hope those words mean something) from 9.05 to 9.1.0. The linked post includes code to duplicate the problem.
onyx hinge
[answered in #help-with-circuitpython as I think it's an intentional change]
lone axle
hello! I'm pretty far along on a big project and I would like to add some error ...
I've used a digital pin to enable / disable writable mode for a handful of projects on several different micro-controllers and never had any issues.
I suppose if it lost power while circuitpython was trying to write it might lead to the drive being in a corrupted state though. Maybe best not to do it on a battery powered project unless you also have active battery monitoring and can disable "risky" behaviors like writing when the power is low.
Truthfully though if you can work in an SDCard to the circuit then it's more convenient than writing to CIRCUITPY drive IMO. You get way more capacity and don't risk corrupting the microcontroller storage + you don't have to set up and use the pin to switch states during boot since your SDCard can just always be writable.
tranquil ether
I've used a digital pin to enable / disable writable mode for a handful of proje...
Thanks for the response. Just what I wanted to hear was someone's experience over multiple projects. I'll keep the SD card in mind. Many of my projects are on the larger side and adding features that are for ME and for development just make sense. I had a great big push button in the project and so now I'm reading that during startup. So, to go into dat-logging and error-loggin mode, I just hold down the button during power-on. Seems to be working ok (<24 hrs)
How about the reset signal? as soon as I screw the enclosure shut, I seem to need to do a hard reset. Do you have experience using a reset button? perhaps less of an issue now that I'm using a powered USB hub with a push-on/off button.
manic glacierBOT
lone axle
How about the reset signal? as soon as I screw the enclosure shut, I seem to nee...
Some microcontrollers expose the reset pin on a header that you could then extend out to the outside of your case if you needed. The raspberry pi pico famously does not have an on-board reset button so if you want to reset it you must either attach your own to the exposed pin or just use a jumping wire to connect it to GND.
I use reset buttons frequently, but don't use enclosures that much, and when I do I typically end up with ones that leave the on-board reset pin exposed either on an edge or with a hole cutout to poke it with something skinny.
midnight ember
@tranquil ether When wiring up an enclosure external reset switch, for microcontrollers that only have a reset button and no GPIO reset (such as the QT Py) you can bodge some wire to the physical reset button. It won't be pretty but it will work. Any Feather is super easy in comparison as Feathers have 2 GPIO for hard reset (RST to GND or EN to GND).
onyx hinge
<@1112401636166488165> When wiring up an enclosure external reset switch, for mi...
@tranquil ether in addition to that, many Adafruit boards have undocumented test points for reset & boot mode that can be seen on the PCB designs. these are used for factory test. For instance, qt py esp32-s2 has SWC1 for rset and SWC2 for BOOT0. It is tricky but you can solder fine wire to these pads and bring it out to a physical button. There's no silk for these pads so you'll want to import the design files into an eagle-compatible PCB editor so you can locate it. https://learn.adafruit.com/adafruit-qt-py-esp32-s2/downloads
manic glacierBOT
tulip sleet
<@&356864093652516868> The weekly CircuitPython meeting (delayed one day due to a US holiday) will be in 1.5 hours, at 2pm US Eastern time. Add your notes to https://docs.google.com/document/d/1BXYMbOxPWUug-6WwltLdcvqv7i_de3059noqBx0Sn-M/edit?usp=sharing. ttyl!
slender iron
(was just going to ask about the merge) 😄
onyx hinge
👋
lone axle
Thanks for hosting Dan. Have a great week everyone.
slender iron
@tulip sleet need any help/assistance on the 1.23 merge? It'd be good to have that in 9.2
tulip sleet
<@329766224093249548> need any help/assistance on the 1.23 merge? It'd be good t...
I haven't worked on it since last week, but it's my next thing. I'll let you know!
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/13FWJDCzuchjYlYK_5Q_3zCgKl5-x6oX0YLFo-AkJMJM/edit?usp=sharing
wanton dirge
Getting our young port up and running for the first time...the MSC device mounts correctly but shows up as a USB drive (not "CIRCUITPY"). Windows wants to format it before I can use it...I'm doubting that's advised as I can't see that anywhere in documentation. Is there anything special I have to do for Windows to recognize the filesystem properly?
slender iron
@wanton dirge I suspect the formatting code that CP is doing isn't working
how are you storing the filesystem? CP has an "internal" mechanic or an "external" one
wanton dirge
Internal currently, but wanted to get the REPL up before I was really able to test my internal_flash.c/.h. It is totally possible that my flash FS driver is the culprit.
slender iron
do you have a way to do debug printing? you could dump the first few bytes of every read and write
to make sure the subsequent reads match what was written
wanton dirge
At this point I have the USB-CDC, so I could try dumping to that with mp_printf?
slender iron
that'll only work once USB is going
wanton dirge
Ah, right
slender iron
and the flash formatting happens before that. have you done busio.UART yet?
(you could also use your own uart code)
wanton dirge
Doesn't get set up until after boot, right. No haven't done busio.UART but I'm not too far off; I did add in a serial.c for the supervisor at least which I can perhaps mangle into that module.
slender iron
what does your current serial do?
busio.UART just means the standard "CONSOLE_UART" stuff will work
wanton dirge
Nothing in particular, I had it as a placeholder to test out after getting the REPL up. That happened about 30 minutes ago 🙂
Was intending to perhaps use it as a debug UART for me to develop with
slender iron
it may be quicker to just init a sdk UART for debug prints
busio.UART may get you into pin mapping earlier than you want to
wanton dirge
I do have a lot of the preliminary pin scaffolding done, but I think using a simple debug UART to debug the FS stuff should be pretty simple. Thanks, I'll give that a shot!
slender iron
👍 good luck!
I think your error is "usb is working ok but the MSC bytes back don't have a file system"
wanton dirge
Yep, that sounds about right. Just wanted to make sure that was the right path before diving in again
tulip sleet
Ah, right
there is normally a "safe mode" message that says "CIRCUITPY drive could not be found or created." which should show up in the REPL when you press reset.
blissful pollen
Making sure this makes sense, after doing some research and testing with audio effects best bet to make an issue to track my ideas? There is one (https://github.com/adafruit/circuitpython/issues/8974) but is specific to AudioMixer and I have an idea that may work more generic. Probably better this way to discuss any API issues before a draft PR?
slender iron
Making sure this makes sense, after doing some research and testing with audio e...
I think its fine to discuss there even if you don't think it'd be done in audiomixer
blissful pollen
I think its fine to discuss there even if you don't think it'd be done in audiom...
Thanks sounds good. My rough idea is maybe an audioeffects library but I'm trying to come up with a way it can then be used for any audio library if an interface is added.
manic glacierBOT
manic glacierBOT
onyx hinge
I noticed that this message is printed during a failed build. ```common-hal/rp2pio/StateMachine.c: In function 'common_hal_rp2pio_statemachine_get_pc':
common-hal/rp2pio/StateMachine.c:1102:13: error: unused variable 'offset' [-Werror=unused-variable]
1102 | uint8_t offset = _current_program_offset[pio_index][sm];
| ^~~~~~
cc1: all warnings being treated as errors
-e See https://github.com/micropython/micropython/wiki/Build-Troubleshooting
../../py/mkrules.mk:14:HELP_BUILD_ERROR ?= "See \033[1;31mhttps://github.com/micropython/micropython/wiki/Build-Troubleshooting\033[0m"
I was going to take that out, but hadn't done it yet. I can do it as part of the v1.23 merge
I think it snuck in, in the v1.22 merge
onyx hinge
-e whatever
```system that use dash as /bin/sh may print the -e because it's not a recognized flag by the built in echo command.@tulip sleet thanks
tulip sleet
I'm not sure it's all that helpful a message even if we pointed it to, say, the Learn Guide
slender iron
I'm not sure it's all that helpful a message even if we pointed it to, say, the ...
You could point to this discord channel
tulip sleet
You could point to this discord channel
I wasn't sure of that, but good idea, maybe the guide and discord
slender iron
@onyx hinge if you are in PIO land, you could look at the new additions in the rp2350
onyx hinge
@slender iron good idea. I'm also a bit interested in doing ps2io in pio, I don't think we have it on rp2xxx yet.
lone sandalBOT
uncut nexus
@onyx hinge What day(s) are planning to be at the Vintage Computer Festival? We'll be going up on Saturday, I think
onyx hinge
@uncut nexus I'm planning to be there both days, driving in from Nebraska friday & driving home monday.
uncut nexus
@onyx hinge OK - I'll be the tall old guy saying things like "haven't we got one of those in the basement?"
onyx hinge
I'd be happy to meet you. I think I'll mostly be wandering around seeing what people have at their exhibition tables.
uncut nexus
Us too. I'll keep an eye out for you. Hard to meet up with so many folks around
stuck elbow
Us too. I'll keep an eye out for you. Hard to meet up with so many folks around
blue hair help
onyx hinge
my hair has mostly faded/grown out since pycon
stuck elbow
oh well
manic glacierBOT
crimson ferry
uh oh... simply adding import wifi to a <200 line code.py running Adafruit CircuitPython 9.1.3 on 2024-08-30; FeatherS2 with ESP32S2 causes it to crash/hang (but no safemode) ...I'll see if I can narrow it down
slender iron
@fading stirrup how'd the P4 work go? I've got a devkit coming tomorrow
wanton dirge
Hey folks, I'm checking in again on the FS component for my young port: FS is mountable -- proper drive name and all! I can't write to it from the OS at the moment, but working on that.
- CIRCUITPY drive mounts, and includes a bout_out.txt and an empty code.py. REPL works and I don't get any safe mode messages in REPL or boot_out.txt, even after pressing reset.
- Quite a lot of FS artifacts seem to be in "CIRCUITPY:" on first startup...is this okay, or does it indicate some error condition I should be concerned about?
- I'm having issues writing files to the drive. For example if I try to modify code.py, the OS errors out and says it cannot be written. Seems to break the REPL once I do this. It comes back if I press RST though. Is there something about the workflow I may be missing when the OS tries to write this back to the CIRCUITPY drive?
manic glacierBOT
tulip sleet
Hey folks, I'm checking in again on the FS component for my young port: FS is mo...
The .tmp files are not ours, but all the rest are created when CIRCUITPY is created. They prevent macOS from creating junk files, suppress the trash can, etc.
Might the .tmp files be being created by your editor or by some utility program on your host computer?
wanton dirge
Might the .tmp files be being created by your editor or by some utility program ...
These may be created by Notepad++ when the attempt to write into code.py crashes. I was able to delete them, eject the drive, then reset via the button and the drive comes back without those files and without issue.
tulip sleet
These may be created by Notepad++ when the attempt to write into code.py crashes...
notepad++ does not do a flush when writing files, which means files may go unwritten for a number of seconds. See https://learn.adafruit.com/welcome-to-circuitpython/recommended-editors
wanton dirge
Okay, that did the trick; thanks very much for linking that article. I do get one strange error when trying to run code.py...I'm not trying to import anything and I don't see code.py files usually including any if " __name__ == '__main__' or similar. Is this something that should resolve as I add more modules?
tulip sleet
we don't need to use if __name__ == "__main__" generally, because code.py just runs. I'm not sure about your error, I don't see the underscores, and there are a lot of extra spaces.
Adafruit CircuitPython 9.1.3 on 2024-08-29; Adafruit Metro ESP32S3 with ESP32S3
>>>
>>> __name__
'__main__'
>>>
Hmmm okay, I will try and dig in on it a bit more then on my end. I get the same output for the __name__ variable as above.
tulip sleet
could you upload your code file using the + to the left. Also here is how to paste code: #welcome message
wanton dirge
As in my code.py? Here...
a = 5
b = 10
c = a + b
I tried a print("hello") as well but got the same issue.
tulip sleet
I don't know where the name ' ' is not defined is coming from
wanton dirge
No worries, I will try and inspect it a little more. Thanks for the help in any case!
glass light
MOVED TO HELP-WITH-CIRCUITPYTHON: Is there anywhere I can look to find information on how to connect BMP390 sensor with circuitpython to adafruit io? Thanks!
tulip sleet
MOVED TO HELP-WITH-CIRCUITPYTHON: Is there anywhere I can look to find informati...
ask in #help-with-circuitpython ; this channel is for core development questions, or ask in #help-with-wippersnapper-and-adafruitio
slender iron
nice work @wanton dirge !
wanton dirge
thanks @slender iron!
slender iron
@tulip sleet you around?
slender iron
the rp2350 issue looks like clock startup delay issue again
just wondering if you have any context on the 64 setting
it makes no sense to me that it uses the xosc to count the delay
tulip sleet
did you already set the MULTIPLIER
I tried several multipliers and 64 seemed safe. 32 was too low
i'll look at pico-sdk...
tulip sleet
i think a q is what's different about our component choices? do the other RP2350 boards use the MULTIPLIER?
tulip sleet
I made a minimal change to their logic and delay computation to be able to lengthen it by a lot. That was how I did my initial PR to pico-sdk
slender iron
the code now relies on the xosc peripheral to time the delay
based on the xosc input!
tulip sleet
all I did was include the multiplier in here: #define STARTUP_DELAY ((((XOSC_HZ / KHZ) + 128) / 256) * PICO_XOSC_STARTUP_DELAY_MULTIPLIER)
slender iron
my change set the multiplier back to 1 for everything
I see it wait 57 ms or so now
tulip sleet
that sounds about right. yes, it does count cycles. The original idea was to let the amplitude go up. What I saw was that it was oscillating at a higher frequency than the 12MHz, or was jittering. So I made it count a lot more cycles.
The XOSC is not immediately usable because it takes time for the
oscillations to build to sufficient amplitude.
slender iron
yup, that's what I'm seeing too. it is counting too fast during start up
tulip sleet
I saw that on the Saleae and/or an oscilloscope
we are not using the specified crystal
most other people are
wanton dirge
No worries, I will try and inspect it a little more. Thanks for the help in any ...
Surprise surprise...it was a file I/O issue. I had fudged some math on the write_blocks that gave me an issue which only cropped up deeper into the FS. So the file text was getting written over a different array and the lexer only saw "0xff"s all the way down.
All good now though. Thanks again for the help folks!
tulip sleet
@slender iron I will shoot for another alpha release tomorrow after I fix the board overflow in #9598 and maybe see about #9583. I'm just starting to work on getting the V1.23 merge to compile, but no big stumbling blocks. There are some calls I have to change pervasively due to signature changes or newly available obj creation routines.
after the merge we can move to a beta.0, I guess
slender iron
Sounds good!