#circuitpython-dev
manic glacierBOT
manic glacierBOT
atomic summit
Hiya - I'm trying to commit my FeatherS3 Stemma I2C changes into the 8.2.x branch in github, and I'm having some pre-commit issues... everything has been installed correctly as per the learn guide for building CP. All requirements (dev + doc) have been installed. Everything looks fine... but I keep getting a fail with translations because of /bin/sh: xgettext: command not found
Full commit output is here...
https://gist.github.com/UnexpectedMaker/ce7d79bc0a2193d2f12bbeb850943ac3
This on an M2 Mac running 13.5.1, python version is in the gist
Any thoughts?
wraith crow
On linux I needed to run "sudo apt install gettext" at one point
atomic summit
I have the xcode command line tools installed - which should apparantly provide xgettext
Ok, I just brew installed gettext and got past that error, but now it's failing to translate a bunch of stuff I've not even touched 😦
Ok, I got it - wow, was a submodule conflict nightmare after checking out the 8.2.x branch
wraith crow
👍
manic glacierBOT
atomic summit
Well, that was fun! hahaha... dislike_git++ 😉
tulip sleet
Ok, I got it - wow, was a submodule conflict nightmare after checking out the 8....
I keep two different clones, one set to 8.2.x and one to main, so I don't need to switch a single clone back and forth.
atomic summit
I keep two different clones, one set to 8.2.x and one to main, so I don't need t...
That's a very clever idea ! The switching of branches with so many (and such large) submodules is a killer.
tulip sleet
Sometimes even more than that, if I have multiple PR's going 🙂
midnight ember
For the matrix panel height being implied might not work in all situations. a panel that is 64x32 rotated 90 degrees would have a height of 64 and width of 32. if it's implied prior to the rotation that would work but then there's also the 64x64 panels out there too. while height might be optional i still think it's best practice to specify all dimensions, just like you would do with a TFT.
manic glacierBOT
manic glacierBOT
midnight ember
I'm not sure where to put this. An issue, a feature request, etc..? The Metro M7 iMX 1011 does not include the RGBMatrix library. Apparently that library only exists in shared-bindings and common hal etc.. within circuit python.
Attempted to use the iMX with the Uno/Metro form factor RGB Matrix shield. Would likely work in Arduino but doesn't seem to be support in Circuit Python.
was just an experiment anyway. ```py
Traceback (most recent call last):
File "code.py", line 10, in <module>
ImportError: no module named 'rgbmatrix'
I know that kind of combination is likely out of leftfield. Was too tempting to try the iMX's speed with the panels.
onyx hinge
I'm not sure where to put this. An issue, a feature request, etc..? The Metro M...
@midnight ember yeah it's not supported at the moment on that chip. first move would be to file a feature request on the protomatter github repo, but I don't know if adafruit is inclined to devote developer time to it.
As to your other question/remark, unless I'm mistaken rgbmatrix doesn't support rotation. rotation comes into play when you add the displayio layer with a FramebufferDisplay. So rgbmatrix "height" and "width" are always about rows and columns in the panel addressing scheme. when you create a FramebufferDisplay you can get/set its rotation property and you can get its width and height properties which for 90 and 270 rotations are swapped compared to the rgbmatrix.
though .. there is a "teensy4" "core" for rgbmatrix, so maybe it's not as much work as you'd think (i.e.,does "teensy4" support all mimx10xx or just mimx106x or what..)
the code does imply it's 1062 only, unfortunately.
midnight ember
Could I use the matrixportal library? I was trying to use rgbmatrix itself as that's what I'm doing on the matrixportal s3. I'm not using the matrixportal library with the matrixportal s3... only rgb matrix.
It will do rotation np.
here's the current code i'm experimenting with 6 panels. https://github.com/DJDevon3/My_Circuit_Python_Projects/blob/main/Boards/espressif/Adafruit MatrixPortal S3/128x96 RGB Matrix/code.py
seems the matrixportal s3 has the rgbmatrix lib frozen in.
tulip sleet
RGBMatrix is native C; it is not frozen
onyx hinge
midnight ember
ah
onyx hinge
this section tries to explain how rotation works. You use the RGBMatrix constructor the same (so rows and columns are "un rotated" numbers), and then set the rotation on the FramebufferDisplay.
tulip sleet
The MatrixPortal library is a wrapper and convenience library around RGBMatrix, see, e.g. https://github.com/adafruit/Adafruit_CircuitPython_MatrixPortal/blob/main/adafruit_matrixportal/matrix.py
midnight ember
guess i should say i was using rotation with the framebuffer yes
was being lazy and didn't want to rewrite the whole thing around the matrixportal library. thank you for the directions, will do.
tulip sleet
onyx hinge
A pin on an I2C expander has a characteristic we never imagined before: maybe it can't be read or written RIGHT NOW because it's on the other side of a locked bus. 😕
wind kettle
Hello
Any news about BLE software support on Raspberry Pico W?
👀
Any updates or commits or pull requests?
random junco
Any updates or commits or pull requests?
There aren't any updates - you can follow progress on this issue: https://github.com/adafruit/circuitpython/issues/7693
wind kettle
There aren't any updates - you can follow progress on this issue: <https://gith...
Ah ;((
Alright
Thanks
stuck elbow
A pin on an I2C expander has a characteristic we never imagined before: maybe it...
I imagine there will also be all sorts of race conditions with setting its value
brazen hatch
Picow will probably not work very well with ble too..
It already is crippled by ram..
Reserving more for the ble stack isn't gonna be fun.
onyx hinge
@tulip sleet can you do final (hopefully) review on the dot clock display PR? Since scott's last review I fixed the typo and the copyright notice; and added the hack tablet board def. It'd be nice to merge it today since I'm out tomorrow; asking you because scott's out today.
tulip sleet
<@329766224093249548> can you do final (hopefully) review on the dot clock displ...
sure, I'll take a look
orchid basinBOT
onyx hinge
thank you
stuck elbow
Picow will probably not work very well with ble too..
It already is crippled by ...
you usually only use ble or wifi, though, so you could hopefully only reserve the ram for one or the other?
brazen hatch
you usually only use ble or wifi, though, so you could hopefully only reserve th...
I mean won’t cp reserve some just for the ble workflow?
I do not use picow, I don’t like it.
(Yes I bother making fixes and PR’s for a platform I don’t like)
But still I target ljinux to run on it.
And it’s seriously becoming hard.
It should be only using 70k but big allocations (10k+) fail a lot more often, rendering it pretty useless.
That’s fragmentation, even though I spam gc.collect every free slice of time
Until I learn how to execute code from storage in segments this is really hindering what picow can do
Cause currently to exec() something I need to load the whole thing.
And some .py’s are already 15k
wraith crow
Using the picow with Pimoroni's DVI base (or I imagine the PiCowbell) might demonstrate the case, it's possible to build the firmware with both WiFi and PicoDVI active but you're limited by RAM with what you do with it.
tulip sleet
@onyx hinge I approved the PR, but I think Scott would still like you to update the copyright notices in shared-bindings/dotclockframebuffer to be you. There was a spelling fix that it looks like you already resolved. I can re-approve after a push.
brazen hatch
Using the picow with Pimoroni's DVI base (or I imagine the PiCowbell) might demo...
rp4080 Ti with psram eta wen
stuck elbow
there was a some way to get the build artifacts for a particular pull request, anybody remembers how?
tulip sleet
there was a some way to get the build artifacts for a particular pull request, a...
do you mean after it's merged?
stuck elbow
do you mean after it's merged?
no, after the tests pass?
crimson ferry
Checks --> Build CI for open PRs
tulip sleet
in any case, if you click on one of the builds, you can then click "Summary" at the top of the left sidebar. Then scroll down to show the artifact .zip files
or click on the apppropriate build here: https://github.com/adafruit/circuitpython/actions
note that pre-merge PR builds don't build all the languages
brazen hatch
I will be continuing and probably finishing the ftp implementation today.
Can I stream it in one of the channels if anyone wants to see me working?
I have my v4l2loopback in obs setup so I can stream from wayland!
tulip sleet
I will be continuing and probably finishing the ftp implementation today.
Can I ...
No objections!
but you might need privs you don't have to do video
brazen hatch
It works.
wraith crow
Just curious how the streaming worked 😁 You're upside down though....
wraith crow
The video screen is upside down
brazen hatch
?!
wraith crow
that's better, fixed 😄
brazen hatch
So discord's preview is just wrong.. K, good to know, thanks!
slender iron
@tulip sleet I wonder if we need to make the Pin object smarter and encompass drive strength. maybe rename it Pad
tulip sleet
also pulls are usually independent of usage
manic glacierBOT
Everything working well for me, you know, very easy to check on Linux that the USB Serial Port is working as also the USB Mass Storage device:
slender iron
also pulls are usually independent of usage
ya, most micros do have a Pad concept that impacts how it behaves based on the incoming gpio or peripheral signal
brazen hatch
Is there a way to make sure a socket.send() has been ack'ed?
If I spam file parts at max speed, they arrive with missing parts, with no errors on the board side..
I had to add some nominal time.sleep()'s to make em arrive well.
time.sleep() however isn't the right way to ensure that.
Something like .send_with_ack() that checks the ack was good, like on the rfm stuff, doesn't exist for the socket stuff?
When sending small stuff it's all good, but as I begin sending files >20k, packets are lost..
512 byte chunks with sleep(0.02) is the best I could do reliable and with margin for hops
buoyant lagoon
Is there a way to make sure a `socket.send()` has been ack'ed?
If I spam file pa...
tcp or udp?
what does socket.send() return?
brazen hatch
what does `socket.send()` return?
self._socket = pool.socket(pool.AF_INET, pool.SOCK_STREAM)
self._socket.setblocking(False)
idk, isn't it tcp?
the no. of bytes sent
buoyant lagoon
and are you checking that the number of bytes sent, is the number you requested?
brazen hatch
it always returns the intended number, even if it fails?
lemme recheck to make sure
buoyant lagoon
if the buffer only has room for 500 bytes, and you try to send 1000 bytes
it will send the first 500 bytes, and then return 500
brazen hatch
Oh it does indeed report lower numbers on failures..
buoyant lagoon
so you need to retry the send, with the tail end of the buffer
brazen hatch
It is always the last bytes?
Well, I guess it's easy enough to just retry and check if the sums match
buoyant lagoon
when it returns 832, that means it sent the first 832 bytes of the buffer you gave it
so you need to cut the first 832 bytes off, and send the remainder
brazen hatch
lemme try that
buoyant lagoon
you likely also dont need to stick to 1024 byte blocks
just try to send your entire jpeg in one shot
and then discard however many bytes send sent
and keep trying with whatever is left
brazen hatch
it did around 2 pages of retries, all of which are 832 for some weird reason.
buoyant lagoon
there are ~3 layers of buffering that are going on
first, there is however big the buffer is on the sending device, it usually has a max size, to prevent using all ram
brazen hatch
dat = f.read(self.tx_size)
if not dat:
break
res = 0
while res != len(dat):
try:
if res:
print("Retry:", res)
res += self._data_socket.send(dat[res:])
except OSError:
pass
then there is the packets that are in-flight on the network
but an in-flight packet may be dropped, so it has to keep a copy in the tx buffer, until it gets an ACK
the more data in-flight, the higher the latency when you drop a packet and have to re-send all of the in-flight data
a normal TCP stack will adjust the "tcp window" to minimize the packet loss
and then there is the rx buffer on the receiving side
brazen hatch
I have never made it that deep into the esp source.
Tysm for sharing all of this info with me!
stuck elbow
```py
dat = f.read(self.tx_size)
if not dat:
break
res = 0
while res != len(...
don't read the whole file into memory at once
only as much as you want to send
brazen hatch
don't read the whole file into memory at once
the size I read is 1kb?
stuck elbow
ah ok
stuck elbow
but you can further improve it by using a memoryview for that slice
brazen hatch
and then there is the rx buffer on the receiving side
If modern linux has issue with 1mb.. I am fine with that..
brazen hatch
but you can further improve it by using a memoryview for that slice
memoryview is too complex for my poo poo brain
buoyant lagoon
the linux end is likely reading fast enough that the buffer never fills
brazen hatch
i assume thats inside another loop?
precisely so
stuck elbow
memoryview is too complex for my poo poo brain
it basically makes that slices without copying the data
brazen hatch
it basically makes that slices without copying the data
isn't python already passing it by reference to the funcs?
stuck elbow
isn't python already passing it by reference to the funcs?
slicing makes a copy
unless it's on a memoryview
brazen hatch
ah k
I will eventually try to learn memoryview
maybe once I finish the module and try converting it to C to add it to the core
buoyant lagoon
write() on linux behaves the identical way when in C
and most MCU stuff like circuitpython tends to try to copy that behavior
desert yarrow
Does anyone know if USB Host support is included in CircuitPython?
buoyant lagoon
depends on the board i think?
brazen hatch
I don't see it in the list
maybe it's frozen, maybe it's not
usb_host it is iirc
so if it's not in here, it prolly isn't on any other
desert yarrow
"Currently, support for the USB Host peripheral is only in Arduino. So check out the TinyUSB ‘dual role’ examples for some things you can do"
(from that link)
brazen hatch
Eh that's out of date..
brazen hatch
yea, lemme check the source
desert yarrow
thx
brazen hatch
answer: it does
I can see it init the usb port from boot
so it certainly does have it
It could be though, that 8.2 doesn't have it but the current main does
brazen hatch
No problemo, Idk much on the topic, but glad I could help!
desert yarrow
CircuitPython 8.2.5 Latest was just released 3 hours ago, so it's probably in there
buoyant lagoon
i recently build circuitpython from source, for the regular rp2040 pico, and wired a usb cable to the right gpio pins, and was able to do usb host, i think i just had to run 2 lines of python
wraith crow
Certainly worth checking, but CP is currently developing both the 8.x line and the 9.x line so if it doesn't seem to be in 8.2.5 grab the firmware from the "Absolute Latest" link
brazen hatch
CircuitPython 8.2.5 Latest was just released 3 hours ago, so it's probably in th...
the boot stuff that is in 9.x isn't in there
they prolly weren't backported
desert yarrow
This?
brazen hatch
This?
that's the board I am looking at
Well, it could be that just the boot stuff wasn't backported idk
I was busy studying for exams while the usb stuff was going on
so missed all the commits
desert yarrow
heh -- ok
brazen hatch
well, just go ahead and try it, if you want to try the 9.x pre-pre-alpha builds, that are 101% up to date, you can grab them from aws s3:
https://adafruit-circuit-python.s3.amazonaws.com/index.html?prefix=bin/
desert yarrow
Excellent!!!
Do you know when 9.x is "likely to be released"?
It is a few days or a few months-ish?
brazen hatch
months-ish more like
I have only been here since 7.0 so I have only witnessed the 8.x development cycle
9.0 will be a little bit messier, and we are right at the start
desert yarrow
Ok, good to know
brazen hatch
so, yeaaaa
desert yarrow
hah... ok, cool
Holy smokes the github diff between 8.2.5 and the main branch is massive
brazen hatch
Ya, it's like using debian vs arch in terms of updates.
I track main for that reason. (I use arch too btw)
manic glacierBOT
tulip sleet
https://blog.adafruit.com/2023/09/07/circuitpython-8-2-5-released/
Updated TLS certificates; RGBMatrix tuning; new boards
stuck elbow
I made a new clone of circuitpython, and I'm trying to run fetch-all-submodules, and both in main and in 8.2.x it errors with missing commits in a number of submodules. What am I doing wrong?
manic glacierBOT
brazen hatch
I made a new clone of circuitpython, and I'm trying to run fetch-all-submodules,...
Honestly no idea. I last fetched main 2 days ago just fine. I only fetch esp and rp2 modules though.
May I suggest remove-all-submodules?
Then start over.
I could have just been gith being weird
mortal kernel
I made a new clone of circuitpython, and I'm trying to run fetch-all-submodules,...
I've hit this problem a few times over the last couple of days too. It seems to be related to when I resynch the branches in my fork on github and receive a bogus "0 commits" message on the first synch attempt. Once that happens, the next clone to my local machine suffers the missing commits in submodules, but is OK on a subsequent attempt. Weird.
stuck elbow
May I suggest remove-all-submodules?
did that several times already, and tried with differtent branches
mortal kernel
did that several times already, and tried with differtent branches
Just cloned and setup successfully on main. My exact steps were:
mortal kernel
it suddenly works again
btw, do you know how to type a non-breaking return into discord?
stuck elbow
shift+enter
mortal kernel
shift+enter
thanks
very much!
orchid basinBOT
manic glacierBOT
I can replicate on test hardware here: Matrix Portal M4. Don’t even need to run code…issue is visible in the REPL display.
This is a photo of 8.2.5 but it’s basically the same as what’s seen w/8.2.2:
Meanwhile, an Arduino example w/current Protomatter lib (1.6.1), rock steady:
tidal kiln
anyone got any ideas on this one? it's not network. it's not power. it's not sd card or format.
https://forums.adafruit.com/viewtopic.php?p=985115#p985115
it's crashing on the file write line here:
https://github.com/adafruit/Adafruit_CircuitPython_PortalBase/blob/759c5c348878932adc5fcc9e4f3b3f570b43e17f/adafruit_portalbase/network.py#L310C19-L310C19
comment that line out and the code gets through there
orchid basinBOT
tulip sleet
anyone got any ideas on this one? it's not network. it's not power. it's not sd ...
is i a reasonable size?
tidal kiln
i did add a print(len(i)) at some point. i think it was like ~73 on first pass, and then 512 after that. chuck_size is 512.
strange spoke
@crimson ferry re the seeed esp32s3 board definition for Circuit Python, Matthew from Seeed is the person in charge of that wiki program I mentioned in the thread on their discord to potentially reach out re: as a contact for the vid and pid you need there. He is very responsive via email and includes his email in his github portfolio. He may be a good person to reach out to for the remaining board aspects.
I was hesitant to say more on their discord, as silly as it sounds, as I had seen an earlier ask they had from that program to implement CircuitPython for this board and it was marked as done / someone did it in a generic way vs the board definition and I didn't want to call out the deficiency in a public way shaming the dev (weird anxiety thing of mine) https://github.com/Seeed-Studio/wiki-documents/issues/327 given their callout specifically to support this board on this platform I think he'd be happy to have you reach out to discuss it further though to get the last bits in place.
tulip sleet
i did add a print(len(i)) at some point. i think it was like ~73 on first pass, ...
so does it only crash if it's writing to an SD card of any kind, or also crash if writing to a CIRCUITPY file or even sys.stdout?
manic glacierBOT
tidal kiln
so does it only crash if it's writing to an SD card of any kind, or also crash i...
only test with one sd card. same one that worked fine with 6.3.0 CP. did not test writing to CIRCUITPY, but poster did.
tulip sleet
only test with one sd card. same one that worked fine with 6.3.0 CP. did not tes...
It would be great to try to get a simpler example, but people write to SD cards all the time, so I'm wondering what the issue is. You could instrument a sequence of writes and data lengths that causes the problem and then try to reproduce that in a simpler example.
midnight ember
disable_concurrent_write_protection sounds dangerous. was that being used? if you pull the power during a file write yes it will absolutely corrupt files on the sdcard. it should not corrupt circuit python.
tidal kiln
It would be great to try to get a simpler example, but people write to SD cards ...
i haven't been able to recreate with simple example yet. it's something about that context. it also crashes writing some static bytes, i.e. replace file.write(i) with file.write(bytes([23]*512))
midnight ember
part of the problem might be related to running 8.2.4? try with 8.2.5?
tulip sleet
part of the problem might be related to running 8.2.4? try with 8.2.5?
there aren't any changes from 8.2.4 to 8.2.5 that should affect SD writing
midnight ember
since it's an online script if it failed halfway and didn't complete the chunks, it might be possible to corrupt an unfinished file?
tulip sleet
does it reset with an open file?
midnight ember
seems as if it's pulling from an api url though APIURL = "https://openaccess-api.clevelandart.org/api/artworks?cc0=1&has_image=1&indent=2&limit=1&skip="
with 8.2.4 the adafruitio part of the script would likely work but the clevelandart.org url might decide not to.
brazen hatch
I just got the ftp STOR command working (put)..
But writing is so slow!
---> STOR IMG_2023-03-29-14-56-39-732.jpg
150 Opening data connection for IMG_2023-03-29-14-56-39-732.jpg
226 Transfer complete
1067934 bytes sent in 60 seconds (17.4 kbytes/s)
ftp>
---> QUIT
That is on a dio flash board, and I cached the writes.
I made 2 bytearrays, one for rx and one for storing chunks of the file as to not write too often.
I think I will remove all the caching code, since the gain is pretty minimal from gathering many writes.
But I at least got to learn memoryview
tulip sleet
as well as 7.3.3
Suppose you disable the actual network fetch and just write fixed data?
tidal kiln
Suppose you disable the actual network fetch and just write fixed data?
do that in portal_base/network.py?
tulip sleet
or something like that
i'm trying to reduce the amount of code that leads up to that... Since you said a simpler example is not failing.
is the network code somehow interfering with the SD writes (which use SPI)
tidal kiln
looking...can comment the fetch...but then it needs something to iter on in the write loop....
not sure. don't think so? at this point it's done the fetch, and now it's just writing to sd.
tulip sleet
so you tried extracting the SD mounting, and the write loop, and it didn't fail?
tidal kiln
yah. a very simple sd card write example. that mounts, opens file, loops and writes bytes.
worked
tulip sleet
are the chunks determined by the fetch? Is the data coming back in chunks from the website, or is it all concatenated?
if the latter, you could replace the request.get() with a string assignment
tidal kiln
lots of layers. and a mix of frozen modules, etc.
i'm overriding the frozen portalbase for testing
by placing lib in root folder
(wasn't frozen in 6.3.0)
tulip sleet
is it using sdcardio or adafruit_sdcard?
tidal kiln
sdcardio
tulip sleet
could try substituting adafruit_sdcard and see if it's the same. Are you using one of the 512MB cards or some random 2GB,etc. card?
tidal kiln
using 16G card formatted FAT32
CP 6.3.0 had no issues writing to this SD card
could even remove and view BMP on PC
tulip sleet
is it writing to the display? Could disable the display
maybe some interaction between SD card SPI and display SPI
tulip sleet
disable display how?
just comment out anything that is writing to the display?
tidal kiln
not simple, since that's happening in side pyportal/portalbase libs
tulip sleet
how about no background image, don't add "circle" to splash; comment out the circle update code
response = pyportal.fetch(APIURL + str(itemid)) replace that with something static that's not a fetch
midnight ember
there was an issue in 7.1 with display shapes on the pyportal, think it turned out to have something to do with pinalarm. https://github.com/adafruit/circuitpython/issues/5705 old issue but somehow might be related if pinalarm has been revamped since then?
tidal kiln
`response = pyportal.fetch(APIURL + str(itemid))` replace that with something st...
not sure what you mean by static
tulip sleet
response = "some long bytestring the same length as what was sent"
simulate the fetch
midnight ember
replicate what the response should be so you can feed it data locally
tulip sleet
could just be [1] * reponse_length, kind of like what you did before
tidal kiln
that'll just not call the offending code?
tulip sleet
not call pyportal.fetch(). I want to see if the network fetch is causing a problem.
And if it still fails, that helps reduce the example.
and also take away messing with the display (the cicle thing). But try one thing at a time.
just divide and conquer
tidal kiln
been doing that since yesterday 🙂
i think i'm just going to label that guide as works with 6.3.0 only
tulip sleet
fer sure for now!
tidal kiln
there's tons of intermingled automagic happening in the pyportal/portalbase libs. it's a convoluted divide/conquer
tulip sleet
can open an issue with a summary of all this, just point to the forum thread, say the code in use, and the versions that work/don't work. It would be a circuitpython issue, should raise an error there
what error does it actually raise?
it doesn't sound like a hard crash
tidal kiln
oserror 5?
tulip sleet
sure include that
midnight ember
here's an example response url as far as i can tell https://openaccess-cdn.clevelandart.org/1962.154/1962.154_web.jpg
it parses the json for the path https://openaccess-api.clevelandart.org/api/artworks?cc0=1&has_image=1&indent=2&limit=1&skip=69
manic glacierBOT
manic glacierBOT
midnight ember
seems like it's first checking the width of the image so it doesn't grab one that's 5101 px wide for example
i don't have a pyportal. this one seems right up my alley though. :/
foamyguy would be a good one for this
tulip sleet
@tidal kiln ok, thanks for the issue!
tidal kiln
@tulip sleet thanks for taking a look and suggestions!
brazen hatch
The FTP server is almost done. Should I open an issue for us to discuss if and how it's getting in the core?
We could very well decide to just freeze it due to how big it turned out. But that would mean no ftp-web-workflow.
slender iron
The FTP server is almost done. Should I open an issue for us to discuss if and h...
Why not just have it as a library people can install?
brazen hatch
That's a valid option alright. I was just asking, cause it would be interesting if we could get it to be a web-workflow alternative mode.
jaunty juniper
for web workflow, I think it would be easier and much smaller to expand the current implementation with a few commands to make it webdav compatible
manic glacierBOT
brazen hatch
for web workflow, I think it would be easier and much smaller to expand the curr...
I mean, is there a reason other than lack of time that this hasn't been done?
manic glacierBOT
jaunty juniper
(modern UI FTP clients generally support webdav too)
brazen hatch
(modern UI FTP clients generally support webdav too)
(I mean webdav is poor man's ftp)
jaunty juniper
I mean, is there a reason other than lack of time that this hasn't been done?
and specifically my lack of time 😅
tidal kiln
@slender iron @onyx hinge just fyi - cool photo of someone using the swirly grid
https://forums.adafruit.com/viewtopic.php?t=204291
neat to see it in the wild 🙂
brazen hatch
Well I guess, it sounds pretty counterproductive, shoving 20k of code into core if webdav is a few command implementations away..
slender iron
the brass standoffs worry me a little
brazen hatch
I can take a look at webdav after this.
And perhaps fix the remote shell which never worked and made me implement telnet
Jokes aside, soon enough I will have almost all the common coreutils and programs a old linux distro would have running in circuitpython..
manic glacierBOT
wraith crow
Do you have the FTP client side built yet, with the direction boards are going to support local displays and keyboard input I'd actually be more interested in reaching out from the microcontroller for files rather than the other way around
brazen hatch
Do you have the FTP client side built yet, with the direction boards are going t...
If I get the server side going, it should be fairly easy to do the client too.
It's just that even having read RFC959 (twice), I still haven't understood all the commands.
The client is basically the same code in reverse.
Since the data sockets and stuff also need to made, all the same code would be there.
The messages would just be different
tbh, I made the ftp server just to be able to have "usb" on non-native usb boards.
I love my M5Stack timer camera.
But developing code without native usb is such a pain.
with filezilla I can just have a local copy and sync
Bruh.. bytes(memoryview(self._rx_buf)[:size]).decode(_enc).split(" ")[0].replace("\r\n", "").lower()
slender iron
anyone disagree with setting the default RTD version to 8.2.x for now?
brazen hatch
but you can further improve it by using a memoryview for that slice
Tysm! it's so fast now!
manic glacierBOT
manic glacierBOT
@PaintYourDragon when I tried to install the firmware, the circuitpython board is borked.
tulip sleet
anyone disagree with setting the default RTD version to 8.2.x for now?
Sounds like a good idea!
manic glacierBOT
@dhalbert @PaintYourDragon tested arduino simple.ino, seems to work properly.
#if defined(ADAFRUIT_GRAND_CENTRAL_M4)
uint8_t rgbPins[] = {8, 9, 10, 11, 12, 13};
uint8_t addrPins[] = {4, 6, 3, 5};
uint8_t clockPin = 1;
uint8_t latchPin = 2;
uint8_t oePin = 0;
manic glacierBOT
Scope trace, special CircuitPython build:
Same, Arduino lib 1.6.2:
The timing is identical, down to the clock cycle, even the every-8th-bit delay. I cannot explain why this would work with the Arduino library and not with CircuitPython _built against the same code and g...
crimson ferry
@blissful dragon what board?
Pico W?
what's the content of the boot_out.txt file on CIRCUITPY?
you need the Pico W UF2, that's the regular Pico... no wifi
crimson ferry
I don't know that stuff very well... but it works on the regular Pico UF2? that would be strange
the only differences are the wifi module, and a couple of subtle pin differences
2.1.0?
not too many people around right now, I'd suggest post the question in #help-with-circuitpython now or in the (U.S.) morning, and see if someone can help
it is, timing varies depending on the change and how extensive, who has expertise to review etc
yes, definitely odd... it's never a guarantee that a close firmware will work on a board with more features, but usually it's OK (some things might not be accessible or have the right names)
CP is in an awkward interval right now... working on 9.0.x (main), and having to backport critical changes to 8.2.x
you can always grab latest from the downloads page
midnight ember
Anecdata, it's 2am so not tagging. How did you get the markdown to do a collapsible arrow? That's a really nice feature I think should be added to the default issue report template. That helps make an issue report filled with code examples much easier to digest. https://github.com/adafruit/circuitpython/issues/8381
midnight ember
I will be doing this from now on. Would like to see it added to the base issue reporting template.
manic glacierBOT
mortal kernel
IEEE Spectrum ranks Python the number 1 programming language for 2023. A quote from the article for the hard working CircuitPython develpers to savor, "And although Moore’s Law is winding down for high-end computing, low-end microcontrollers are still benefiting from performance gains, which means there’s now enough computing power available on a US $0.70 CPU to make Python a contender in embedded development, despite the overhead of an interpreter. Python also looks to be solidifying its position for the long term: Many children and teens now program their first game or blink their first LED using Python. They can then move seamlessly into more advanced domains, and even get a job, with the same language."
onyx hinge
can someone double check me? This is not correct usage, right? you might write Union[A, B] or A | B but writing Union[A | B] is .. redundant?
in this docstring in the core: //| hostname: Union[str | ReadableBuffer]
lone axle
can someone double check me? This is not correct usage, right? you might write ...
Union[A, B] is the syntax I am most familiar with. I do believe comma's inside the sqaure brackets is correct, not pipes for seperation.
stuck elbow
I think both are correct, just mean different things
technically correct, anyways
Union[A | B] would be equivalent to Union[Union[A, B]] which basically means the same thing as A | B
onyx hinge
Yes, that is my understanding as well, and much better put/explained
slender iron
Sounds like a good idea!
Done!
thorny jay
Initially posted on the live channel by mistake:
Sorry, skipping the meeting without notes... my poor non CP excuses:
- There is/was a heatwave over Belgium
- School and sport restated with timing incompatible with the weekly meeting
- Part of the week-end I have been trying to be friend with the AI and the robot by carefully placing a perimeter cable around the garden, but the landmower robot has been sabotage it by cutting the wire (a luddite would not have done a better job)... and now it refuse to obey because it felt a bit of rain!
mortal kernel
Wondering if there are any outstanding CP issues that I could lend a hand with? I'm well tooled up for RP2040 and enjoy low-level detail oriented issues. Would really love to help.
slender iron
https://github.com/adafruit/circuitpython/issues/7990 is 8.2.x and rp2 but not super clear what the issue is
manic glacierBOT
slender iron
this should be easy https://github.com/adafruit/circuitpython/issues/7608
multicore, run C blob on second core could be interesting
bigger and less clear
mortal kernel
Thanks! I'll take a closer look at 4106 and 7990. 4106 looks especially interesting.
slender iron
<@&356864093652516868> We have around 15 minutes until the community meeting in https://discord.com/channels/327254708534116352/359422932545568799
idle owl
@slender iron Turns out the issue started on GitHub, and they suggested contacting them through their support page (https://readthedocs.org/support/). I ended up talking to Eric, and then another person took care of the needed changes. I would suggest using the support page as it definitely got me to the right folks.
👍🏻
mortal kernel
Link to meeting notes, please?
slender iron
turbid radish
If you have news, projects, updates, & more - please send them to cpnews@adafruit.com for the next newsletter
onyx hinge
turbid radish
Congratulations Kattni! We will miss you but it's exciting to see your next steps!
modern wing
🎉 Congrats, and the absolute best of luck in everything that comes next!
solar whale
Congratulations Kattni!! Best wishes! Thank you!
gilded cradle
Congratulations @idle owl
idle owl
Thank you so much, everyone! 😊
ember iris
Congratulations, and thank you for having been such a pillar in this community! A lot of how welcoming this community is is thanks to the effort you've put into it!
modern wing
Your guides have been an amazing welcome to so many newcomers to CircuitPython.
And helping run the newsletter, and the meetings, and all the other product guides, and building this Discord, and bringing in so many absolutely wonderful people to this community....you're an awesome community architect 💜
thorny jay
Whoa, that is breaking news...
I have so much to say.
I guess most of us have seen you create and improve this community... but we have also seen you make your first board, your first contribution to the core and so many other steps where you improved the range of your skills.
This place would not be what it is now without everything you did.
Good continuation @idle owl
idle owl
@thorny jay Thank you. I really appreciate your words. ❤️
onyx hinge
ooh gamma correction.
thorny jay
About @devout jolt in the weed... Is that for flashing from a computer or flashing from a smaller board like LadyAda testing board? Having a Feather RP2040 USB Host do that job would be great!
slender iron
computer I assume
thorny jay
About <@352910176736772096> in the weed... Is that for flashing from a computer ...
Educators may also be interested to setup a classroom "en mass".
onyx hinge
also yay it's important we have someone who has run an adabot patch actively involved
candid sun
thanks for hosting @slender iron
modern wing
Thanks for another excellent meeting!
gilded cradle
Thanks everyone
lone axle
Thanks for hosting Scott, have a great week everyone.
ember iris
Thanks all, have a great week!
slender iron
you all are welcome! thanks for meeting up!
onyx hinge
for some reason a very old pi -- is it a pi 1? -- has surfaced on my desk. (c) 2011.12
main furnace
It belongs in a museum.
slender iron
AC in Seattle -- who'd a thought ....
I think it might be my computers in this space. Not super sunny but still went from 78 to 82 during the meeting.
onyx hinge
@slender iron do I repeat the "Adafruit" name in both the USB manufacturer & product? ```USB_PRODUCT = "Adafruit-Qualia-S3-RGB666"
USB_MANUFACTURER = "Adafruit"
A Pi 1 is running my heating system automation... talking EnOcean to give the target temperature value to the valve that lets the hot water flow in the radiator.
Still working fine, obviously not developed by me. 😉
slender iron
<@252717193496756235> do I repeat the "Adafruit" name in both the USB manufactur...
I don't usually
onyx hinge
@slender iron is there any big problem changing the USB_PRODUCT string after the fact? There are apparently a handful of Adafruit boards that do this.
espressif/boards/adafruit_feather_esp32s3_4mbflash_2mbpsram/mpconfigboard.mk
espressif/boards/adafruit_feather_esp32s3_nopsram/mpconfigboard.mk
espressif/boards/adafruit_qtpy_esp32s2/mpconfigboard.mk
```plus the one I'm alreayd renamingI wonder if the first 3 were specifically due to there also being the Unknown Maker Feather S2/S3
devout jolt
About <@352910176736772096> in the weed... Is that for flashing from a computer ...
yep, full CircuitPython app restore via computer, preferably via UF2 (thanks @slender iron for clarifying for me, sorry I couldn't make the meeting)
onyx hinge
I don't think I'll change the others right now
mortal kernel
I'm seeing something weird when cloning the CP repo. Lots of "multiple configuration" messages like this:
warning: 01cabb0324186808c14ec1b02fb8a3be28837160:.gitmodules, multiple configurations found for 'submodule.lib/axtls.path'. Skipping second one!
warning: 01cabb0324186808c14ec1b02fb8a3be28837160:.gitmodules, multiple configurations found for 'submodule.lib/axtls.url'. Skipping second one!
gh repo clone eightycc/circuitpython circuitpython-82x
slender iron
I haven't seen that before
mortal kernel
It's just happening when I clone my fork. Think I'll try to delete it and fork again.
slender iron
I usually use raw git
Here is the notes document for next Monday’s CircuitPython Weekly meeting. It is at the normal time of 11am Pacific / 2pm Eastern here on Discord. Everyone is encouraged to attend! Please add your hug reports and status updates even if you’ll be attending the meeting - it’s super helpful! If you are unable to attend but would still like to include updates, feel free to include them in the notes and we’ll read them off during the meeting. Hope to see you there! <@&356864093652516868> https://docs.google.com/document/d/1J0mv2qVnfBwjg4DZj47JYn-WPQI68thUqp8ZYfr14fs/edit?usp=sharing
mortal kernel
Deleting my fork and re-forking fixed the problem.
manic glacierBOT
manic glacierBOT
manic glacierBOT
manic glacierBOT
manic glacierBOT
onyx hinge
I didn't see an I2CTarget guide on Learn...
onyx hinge
@tulip sleet I notice this in the git history: ec1e7a03b38c625664a4615e60d26956205d7a67 Reset UART on deinit, reduces power consumption
tulip sleet
<@329766224093249548> I notice this in the git history: ec1e7a03b38c625664a4615e...
ok, you saved me doing a bisect 🙂
onyx hinge
seems a likely candidate to introduce a subtle problem
mortal kernel
I need REPL redirect to a UART. I've found mention of DEBUG_UART in 4761, but other than a mention in ports/stm/supervisor/serial.c there's nothing else in CP. Before I hack up my own, wondering if anyone has redirect code gathering dust somewhere?
tulip sleet
I need REPL redirect to a UART. I've found mention of DEBUG_UART in 4761, but ot...
There is CIRCUITPY_CONSOLE_UART; I think that might do what you want. See supervisor/shared/serial.c.
mortal kernel
There is `CIRCUITPY_CONSOLE_UART`; I think that might do what you want. See `sup...
Thanks, looks like exactly what I need.
tidal kiln
@tulip sleet just curious - any idea why the singleton uart worked? (for that nrf issue)
tulip sleet
<@329766224093249548> just curious - any idea why the singleton uart worked? (fo...
i think it doesn't get deinit-ed on soft-restart
so the new code in the PR never gets run
tulip sleet
@cater the forum post user says they were using 1.7.5 ... hmm. I did circup update all my libraries too
tidal kiln
i had old fw. trying update now.
tulip sleet
the user said they had 1.7.5 but if they don't really that would cause the issue
we certainly should change the library code to prevent the red herring SD card message, which wasted a lot of time
tidal kiln
@tulip sleet what if you use libs taken from bundle zip? still work?
tulip sleet
<@329766224093249548> what if you use libs taken from bundle zip? still work?
circup update uses the libs from the latest bundle
so that is what I updated to, without having to do it by hand
tidal kiln
right. but wondering if it could be a lib issue. and we aren't sync'd on that? someone following guide would just use bundle zip.
tulip sleet
right. but wondering if it could be a lib issue. and we aren't sync'd on that? s...
i thought the program bundle downloads the latest libraries
I did not test the Learn example, just your simplified example
orchid basinBOT
tidal kiln
@tulip sleet one thing to keep in mind is that the portalbase lib is frozen in. so it isn't latest. and isn't using what is in /lib.
tulip sleet
<@329766224093249548> one thing to keep in mind is that the portalbase lib is fr...
oooh, hmm, I wonder if I have a non-lib/ version
@tidal kiln well, I only have .mpy libraries, and only in .lib
i just re-updated NINA-FW, and it's still working
testing 8.2.5 with your test program
does it work without an aio key?
tidal kiln
i'd think it'd fail trying to access it?
tulip sleet
yeah
tidal kiln
but on that point - my setup still has secrets.py instead of toml
tulip sleet
is there any possibility we are using different AIO servers which are set up differently
I have both secrets.py and settings.toml, updated with a new key. updating settings.toml did not help; I needed secrets.py
io.adafruit.com for me is 52.54.163.195
tidal kiln
64.190.63.111 for me
tulip sleet
hmm, hmm
could copy the library into the top level of CIRCUITPY and change io.adafruit.com to a numeric IP
i just did ping io.adafruit.com from the command line
tidal kiln
it pulls OK just copying URL into browser
tulip sleet
do you have a paid IO subscription? I don't
tidal kiln
no. just free one.
tulip sleet
@tidal kiln so nslookup io.adafruit.com returns 64.190.63.111 for you?
tidal kiln
@tulip sleet hmmm...not anymore. now i'm getting 52.54.163.195
tulip sleet
I have no idea how they manage servers; maybe it is dynamic
I tried using curl and it's working fine with the given URL.
You can get the time from Adafruit IO without any credentials, but it should go through the same TLS stuff: https://io.adafruit.com/api/v2/time/ISO-8601
here is a simple test program for ESP32SPI; could modify it to test the time URL I just gave
tidal kiln
tried numeric ip. looks like maybe nina fw doesn't support that? dumps with an esp32 not responding on the fetch.
tulip sleet
the numeric IP doesn't match the cert, so it complains (I saw that when I tried the numeric IP in curl). So that won't work 😦
tidal kiln
Fetching text from https://io.adafruit.com/api/v2/time/ISO-8601
----------------------------------------
2023-09-12T15:22:51-04:00
----------------------------------------
worked
had to comment out the wifi scan
but otherwise, just change TEXT_URL to the aio time URL
tulip sleet
the wifi scan works for me
tidal kiln
nothing but hidden here. so got a nonetype on return. and didn't like it.
tulip sleet
I erased CIRCUITPY completely, added back your test program, and the needed libraries. Still works.
how fast is your internet? Mine is 300Mb up/down. It's slower through the wifi, but I have good wifi access points
i'm grasping at straws a bit
could you maybe try your phone as a hotspot?
tidal kiln
oh...i'm sad slow...like 5mbps down/0.5mbps up. best option i have out here.
tidal kiln
@tulip sleet if you're suspecting something network, wouldn't it hit it also replacing the file write with a print?
tulip sleet
<@329766224093249548> if you're suspecting something network, wouldn't it hit it...
I'd think so, but it doesn't hunh? I was wondering if there were long delays on the retrieval
tidal kiln
lemme try it again. pretty sure i did that days? ago. whenever i was originally troubleshooting.
only portalbase is frozen in. pyportal is coming from /lib.
tulip sleet
we should be using the same libs now
tidal kiln
you're using ones from bundle zip?
tidal kiln
ohh....sry....that's what i meant by bundle
the bundleflyblob thing
from learn
and i'm not actually sure what i have loaded right now, since have tried many things. i'm going to storage.erase and start over.
tulip sleet
try what I am using: your short test program, and circup-installed libs. I'll compare the libs
tidal kiln
(i'll need to install circup)
tulip sleet
the libraries differ only in adafruit_esp32spi/PWMOut.mpy (!)
tidal kiln
ERROR: Failed building wheel for circup
worth chasing this down? or can just use bundle zip?
tulip sleet
use Learn bundle and if that fails, try library files from latest library bundle: https://github.com/adafruit/Adafruit_CircuitPython_Bundle/releases/download/20230912/adafruit-circuitpython-bundle-8.x-mpy-20230912.zip https://circuitpython.org/libraries
circup is really a time-saver; worth tracking down the issue at some point
maybe your desktop python is old
tidal kiln
it was something about setuptools versions. probably easy fix. but not now.
all of those taken from adafruit-circuitpython-bundle-8.x-mpy-20230912.zip
and then running simple test from GH issue thread
exact same issue
it consistently fails on the second pass through the iterator. note the single . that gets printed on first pass
going to try print instead of file write now
tulip sleet
it consistently fails on the second pass through the iterator. note the single `...
where is the iterator? I will instrument it the same way
are you using your short test program or the original program?
Never mind. I misunderstood. It's:
aving data to /sd/cache.bmp
.............................................................................................................................................................................................................................................................................................Created file of 144138 bytes in 9.0 seconds
using short test program from GH issue
@tulip sleet do you have both secrets.py and settings.toml on the PP?
tidal kiln
try removing toml
tidal kiln
ok. try running again.
tulip sleet
still works
tulip sleet
- Updated root certificates for Let's Encrypt sites
https://blog.adafruit.com/2023/09/12/circuitpython-8-2-6-released/
slender iron
@tulip sleet do you have an ssl test script?
tulip sleet
<@329766224093249548> do you have an ssl test script?
to run on CircuitPython?
slender iron
it'd be handy for idf 5 testing. yup
tulip sleet
You may need to comment out some of the fetches. I get memory errors sometimes if they are all executing. This is the usual test script, + adafruitIO + a Let's Encrypt site
tulip sleet
@slender iron I started to do that 8.2.x PR too, but it has merge conflicts
slender iron
want me to fix them or do you?
I want to rebase my idf5 changes on it since it has the pin in use fix
tulip sleet
i was going to do a local PR from my fork, yeah, I can do it if you just want to close it
manic glacierBOT
manic glacierBOT
manic glacierBOT
manic glacierBOT
crimson ferry
where do the CIRCUITPY_CREATION_ID come from for non-USB boards?
manic glacierBOT
here is a my version of the pin diagram.
Open for any suggestion.
Hi @tirtharajsinha, I noticed the drawing you published today. I have been working with the Waveshare ESP32-S3 Pico for over a week now (with great pain I might add)... I was not able to get the SPI interface to operate on the pins you indicated :
...
manic glacierBOT
onyx hinge
@tulip sleet like so? ```//| :param Sequence[digitalio.DigitalInOut] rgb_pins: The matrix's RGB pins in the order (R1,G1,B1,R2,G2,B2...)
//| :param Sequence[digitalio.DigitalInOut] addr_pins: The matrix's address pins in the order (A,B,C,D...)
yes, that would be great -- thanks. There's not complete standardization on the nomenclature, but that is a big hint
manic glacierBOT
slender iron
@analog bridge are you ok with me trying to force push to your update-idf branch? I'm just about ready to PR it. (I think) Pushing to your branch will update the existing PR.
slender iron
@brazen hatch your web workflow change also changed when we construct mdns
were you seeing issues with mdns as your ip changed? the mdns server should stay up
brazen hatch
hardfaults
at least if memory serves well
my brain is kinda cooked rn
Yea, if the network changed, it was producing a hardfault for some reason.
Not quite sure why, but I did decode the backtrace and it was mdns
So I kinda went, "sure, have a || reload".
Is this causing an issue?
I kinda assumed it's harmless and spared no more of my time on it.
slender iron
ya, its a bit weird because a second construct call will mark the object as deinit
so the web workflow then sees its deinit and doesn't think that mdns is active
brazen hatch
How do I check that?
I have very little experience with actually using it.
slender iron
I saw the issue on esp32
when I loaded the webpage, board name (really instance name) and hostname were empty
walk time. back in a bit
brazen hatch
I need time anyways to test some stuff, I have an idea on how to alleviate the issue.
But it could take many tries.
Enjoy your walk!
slender iron
removing the reload fixes the immediate issue
brazen hatch
But booting without web workflow, enabling and reloading will prolly be a hardfault
slender iron
But booting without web workflow, enabling and reloading will prolly be a hardfa...
enabling from user code?
I'll give it a shot as I test C3
brazen hatch
enabling from user code?
writing a settings.toml and reloading yea
brazen hatch
I'll give it a shot as I test C3
My condolences.
slender iron
writing a settings.toml and reloading yea
enabling web workflow via settings.toml. correct?
slender iron
after adding the web workflow password, it doesn't start up
doesn't crash though either
orchid basinBOT
manic glacierBOT
digital shoreBOT
onyx hinge
huh, weird/cool. I noticed today that this error isn't caught in circuitpython, but it is in micropython 1.20. Turns out it's newly detected by micropython, and misbehaved before: ```py
def f(**kw): return kw
...
f(x=1, **{'x':2})
{'x': 2}
i.e., micropython fixed the bug in 1.20
slender iron
it'll raise an exception now?
onyx hinge
let me double check but yeah
Use Ctrl-D to exit, Ctrl-E for paste mode
>>> def f(**kw): return kw
...
>>> f(x=1, **{'x': 2})
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: function got multiple values for argument 'x'
>>>
nice! better errors are better 🙂
orchid valve
Hello, sorry to bother, but I'm new looking to help debug issue i posted for supporting the STM32F7 disco board. Is there any documentation on how to setup to debug the core? I have experience with embedded firmware, so I don't need hand holding, but any standard practices would be helpful. I have also built circuit python for another project to add a custom board for a client using the RP2040, so vaguely familar with basic getting of code and building, but looking for any direction on how to build for debug. TIA
idle owl
@tulip sleet We did not fix the issue. https://adafruit-circuit-python.s3.amazonaws.com/adabot/bin/reports/circuitpython_library_report_20230915.txt
tulip sleet
Out shopping. I can make a new token and we’ll try that
idle owl
Out shopping. I can make a new token and we’ll try that
Sounds good. Thank you.
slender iron
Hello, sorry to bother, but I'm new looking to help debug issue i posted for sup...
Hi! The make command can take a DEBUG=1 parameter to include symbols in the build
and then a debugger should be able to know what is happening
for debugging safe mode, you can break on reset_into_safe_mode
@onyx hinge do you have a repo with test code for these dot clock displays?
onyx hinge
@slender iron no, just some gists e.g., https://gist.github.com/jepler/56758bd4bc0d9f9ff5548f52c5801cba
onyx hinge
oh the lcd driver?
slender iron
DotClockFramebuffer
onyx hinge
does it look like a tough change? I'll test of course.
onyx hinge
that's good for us! we can lose the hack to retrieve it
I think I also saw double/triple buffering
slender iron
yup, just deleted it
stuck elbow
Hello, sorry to bother, but I'm new looking to help debug issue i posted for sup...
there is also this guide: https://learn.adafruit.com/circuitpython-samd-debugging-w-atmel-studio-7/overview
@slender iron by the way, would it be possible to add the same tags to your article as are on that atmel studio one? it's much harder to find
stuck elbow
\o/
slender iron
🙂
blissful pollen
Unsure if I should raise an issue for this, but if anyone recalls there was an issue with the FunHouse having a display glitch (https://github.com/adafruit/circuitpython/issues/4775) that was fixed by lowering the display rate but it only occurred in CP an not in Arduino so there was more too it.
I am now seeing similar behavior on the ESP32-S2 TFT after I hooked my lightning detector board to it.
I know how to lower the display frequency/build a custom build to fix my porblem, and so far it has only occurred for me with a second SPI device hooked up. But this is showing that the underlying, unknown issue still seems to exist.
onyx hinge
humm is there a reason that these are not static const arrays as a rule? uint8_t display_init_sequence[] = {
seems like not making them const means they use up RAM in addition to flash
slender iron
humm is there a reason that these are not `static const` arrays as a rule? ```ui...
sometimes they get patched but I think that's usually from python drivers
manic glacierBOT
onyx hinge
I wonder why uncrustify INSISTS on the extra indentation here. ```c
- dotclockframebuffer_ioexpander_spi_bus spibus = {
-
.bus = i2c, -
.i2c_device_address = i2c_device_address, -
.i2c_write_size = 2, -
.addr_reg_shadow = { .u32 = 1 }, // GPIO data at register 1 -
.cs_mask = 0x100 << 1, // data payload is at byte 2 -
.mosi_mask = 0x100 << 3, -
.clk_mask = 0x100 << 2, - };
.bus = bus_obj,
.i2c_device_address = i2c_address,
.i2c_write_size = 1 + gpio_data_len,
// ASSERT(LITTLE_ENDIAN, "don't have to differentiate 8- vs 16-bits here")
.addr_reg_shadow = { .u32 = gpio_address | (gpio_data << 8) },
.cs_mask = 0x100 << cs_bit,
.mosi_mask = 0x100 << mosi_bit,
.clk_mask = 0x100 << clk_bit,
};
```it did here too@onyx hinge do you know what clock source it should be set to?
slender iron
I wonder why uncrustify INSISTS on the extra indentation here. ```c
+ dotcloc...
Is that using our post-processing?
onyx hinge
Is that using our post-processing?
that's how pre-commit changes it
slender iron
that's uncrustify + python post-processing iirc
so it might be the post stuff going awry
onyx hinge
<@426200547024961539> do you know what clock source it should be set to?
I don't know about the clocking. I guess I am relying on a default
you mean the function lcd_rgb_panel_select_clock_src?
examples/86box_smart_panel/main/bsp_board.c: .clk_src = RMT_CLK_SRC_DEFAULT, // different clock source can lead to different power consumption
examples/smart_panel/components/esp32_s3_lcd_ev_board/sub_board/sub_board_480x480.c: .clk_src = LCD_CLK_SRC_PLL160M,
examples/smart_panel/components/esp32_s3_lcd_ev_board/sub_board/sub_board_800x480.c: .clk_src = LCD_CLK_SRC_PLL160M,
examples/smart_panel/main/bsp_board.c: .clk_src = RMT_CLK_SRC_DEFAULT, // different clock source can lead to different power consumption
```some different demos in esp-dev-kits make different choices and I don't konw why.yeah the RMT_CLK_SRC must be for something else. Use PLL160M or LCD_CLK_SRC_DEFAULT then?
components/esp_lcd/test_apps/rgb_lcd/main/test_rgb_panel.c: .clk_src = LCD_CLK_SRC_DEFAULT,
welp, it compiles but doesn't work or error
tidal kiln
@tulip sleet im going to call this sd issue and close unless you want to keep open?
https://github.com/adafruit/circuitpython/issues/8382
tulip sleet
<@329766224093249548> im going to call this sd issue and close unless you want t...
I agree!
tulip sleet
np!
manic glacierBOT
plucky tulip
I am currently working on the templating engine for CP, a common feature that they offer are functions for escaping unsafe characters like < to < or " to " in HTML.
Although I can think of multiple formats that might benefit from this feature, only HTML seems to be useful in context of CP.
Markdown, SQL or XML cannot be used directly in CP, so implementing functions for them seems pointless as it increases lib size.
Can anyone think of any other format that would require this functionality while also making sense in CP?
shrewd nymph
I am trying to build a macro pad thing with the follow set up, is it possible to read the (big)switch if I wire it to the stemma port at the end of this chain? I currently have everything up to the big switch working as expected.
tulip sleet
I am trying to build a macro pad thing with the follow set up, is it possible to...
no, the big switch needs I2C connectivity, and it doesn't have it. Out of the box, it's just like any other pushbutton. The Trinkey RP2040 does not have any exposed pins you can use for that; it only has I2C, through the STEMMA port.
Consider using something like a QT Py RP2040 instead, which will let you connect the switch to one of its pins.
proper solstice
You could repurpose the rotary switch push button, if your not using it and bound to that hardware
manic glacierBOT
hidden rain
I create a single like this:
class SingletonClass(object):
def new(cls):
if not hasattr(cls, 'instance'):
cls.instance = super(SingletonClass, cls).new(cls)
return cls.instance
This is declared in fileA, and fileA is imported by many other files. It seems to initialize for each import!
WHy?
Its like i have multiple instances of my singleton...!
short tendon
class SingletonClass(object):
def __new__(cls):
if not hasattr(cls, 'insta...
I believe you want __new__
stuck elbow
__new__ is not supported in MicroPython or CircuitPython, iirc
short tendon
Ahh, there I go using my normal Python knowledge... this has bitten me too many times.
stuck elbow
just make a factory function
hidden rain
just make a factory function
Wasn't aware of the limitation. Can you point to a CPy factory function example that serves a simlar purpose?
stuck elbow
singleton_class_instance = None
def make_singleton_class():
global singleton_class_instance
if singleton_class_instance is None:
singleton_class_instance = SingletonClass()
return singleton_class_instance
Ah! global variables 101. Got it! Thanks man!
stuck elbow
a class variable is a hidden global
hidden rain
a class variable is a hidden global
Yes of course. Just never thought of it that way. OK.. I'm off to re-factor some code.
manic glacierBOT
manic glacierBOT
manic glacierBOT
manic glacierBOT
An actual 8mb sdcard, that has 6.8mb usable is reported to have fat12 content on a fat16 partition type.
My big boi n16r8 borb, reported to have fat16 on a fat12 partition type:
brazen hatch
I made a ramdisk module for circuitpy finally
Question
Why is:
def writeblocks(self, block_num, buf):
for i in range(len(buf)):
self.data[block_num * self.block_size + i] = buf[i]
Like 10 times slower than:
def writeblocks(self, block_num, buf):
area = block_num*self.block_size
self.data[area:area+self.block_size] = buf
it makes no sense..
stuck elbow
because the first one gets executed in python, while the second one is executed in C
brazen hatch
yea but the 2nd one is the slower one..
that's the issue
I tried to do it the smort way
stuck elbow
you mean the slice is slower?
stuck elbow
what is self.data and buf exactly?
brazen hatch
brazen hatch
what is self.data and buf exactly?
just plain bytearrays
stuck elbow
there is something weird going on there, the slicing should be much faster
brazen hatch
Here is the whole code:
from storage import VfsFat
class RAMBlockDev:
def __init__(self, blocks: int, block_size: int = 2048) -> None:
if block_size < 512:
raise ValueError(
"Block size must be at least 512."
)
if blocks < 64:
raise ValueError(
"Cannot be less than 64 blocks."
)
self.block_size = block_size
self.data = bytearray(block_size * blocks)
VfsFat.mkfs(self)
def readblocks(self, block_num, buf):
mvd = memoryview(self.data)
for i in range(len(buf)):
buf[i] = mvd[block_num * self.block_size + i]
def writeblocks(self, block_num, buf):
#for i in range(len(buf)):
# self.data[block_num * self.block_size + i] = buf[i]
area = block_num*self.block_size
self.data[area:area+self.block_size] = buf
def ioctl(self, op, arg):
if op == 4: # get block num
return len(self.data) // self.block_size
if op == 5: # get block size
return self.block_size
The code.py is:
import os
import storage
from ramdisk import RAMBlockDev
import time
a = time.monotonic()
print(time.monotonic()-a)
vfs = storage.VfsFat(RAMBlockDev(512))
storage.mount(vfs, "/ram")
storage.umount("/ram")
print(time.monotonic()-a)
btw, to make that code equivalent, it should be self.data[area:area+len(buf)] = buf
brazen hatch
the writes will always be of block size so its faster to use the stored val
stuck elbow
if buf is different size than self.block_size, your coode would make a copy of the whole self.data
using a memoryview might help
brazen hatch
if buf is different size than self.block_size, your coode would make a copy of ...
2k blocks.. I was on 512 and confused it.
so it's copying indeed
so wait
it's not writing blocks of the correct size???
stuck elbow
it will replace the self.block_size bytes with the whole buf
for that it will need to reallocate and copy the whole self.data
brazen hatch
k I fixed it, but it's still slow..
self.data[area:area+len(buf)] = buf
Now it should never copy??
I am actually gonna speedrun filesystem corruption
stuck elbow
try wrapping self.data in memoryview
brazen hatch
wait aren't mv's readonly?
brazen hatch
bb
apparently not.. you can write just fine..
still no speed improvement.. very weird
tulip sleet
@onyx hinge For the V1.20.0 merge, I am down to about eight files with remaining merge conflicts. Some of these are just tedious, but some I could use your expertise on (generators, type stuff). I think I would like to do some "pair merging" if you're up to it.
btw I like your new avatar on Discord. It's very crisp even tiny. Have you considered it for GitHub? The tiny pictures in Blame listings are hard to see.
onyx hinge
@tulip sleet good morning. I'd like to do that!
I'm out at 4PM ET to get my COVID vaccination, so we could do it before meetings today OR we could do it tomorrow OR we could do it for a short time after meetings.
tulip sleet
I am also trying to get vaccines around that time, though the store said the COVID one is not in stock. That's another story. in about 30-45 minutes would be good for me. I am running the meetings today so I'm pretty occupied 12pm ET on.
onyx hinge
OK, let me eat breakfast and then I'll ping you.
onyx hinge
@tulip sleet I'm available now.
tulip sleet
<@329766224093249548> I'm available now.
be right there - 1 min
onyx hinge
OK, take your time
manic glacierBOT
Turns out that FAT16 does indeed work on my Fairphone 4. I just took a MicroSD card from my Nintendo DSi XL (used in that with a MicroSD to SD convertor, in case you're confused) and it worked without needing to be formatted.
onyx hinge
@tulip sleet was that 2 files we got through? makes me appreciate your work all the more to do some of it with you.
tulip sleet
<@329766224093249548> was that 2 files we got through? makes me appreciate your ...
thanks! it was three files, five to go. After we do those I can finally commit everything as a WIP commit, and then maybe go back and review the diffs (and I might ask you more). I am more confident now about changing all the _type defs in shared-bindings etc. now too.
mortal kernel
@danh Big shout out of thanks for keeping the core of CP in synch with MP. Bravo!
onyx hinge
I'm trying to understand why an I2C bus needs a lock.
In the days of using "continued start", I can see why -- other interspersed activity on the bus would have been trouble. but now everything is writeto_then_readfrom.
Unlike SPI there's no settable frequency and no cs pin to manipulate.
Could the lock be dispensed with?
(broadcom and stm call RUN_BACKGROUND_TASKS during i2c transactions so that would have to change: that WOULD allow an i2c function to be re-entered recursively from background context and right now the locks prevent that. but most ports don't seem to)
blissful pollen
I'm trying to understand why an I2C bus needs a lock.
In the days of using "con...
There are a couple chips I've used where I could not use writeto_then_readfrom as you had to pause a small amount of time before the result was ready.
That aside, with CP controlling the i2c device though I'm trying to think of any case where a lock would be needed. Maybe something with DMA?
onyx hinge
@blissful pollen does access to a different device in between the operations have to be prevented?
or can you write(slow); write(fast); read(slow)
tulip sleet
I'm trying to understand why an I2C bus needs a lock.
In the days of using "con...
threads may have also been an initial consideration; I think @slender iron will have the historical background on the locks
we can discuss in In the Weeds
blissful pollen
<@228322543507865601> does access to a different device in between the operation...
I don't think reading/writing to a different device would matter while you are waiting for the result. But haven't tested it.
onyx hinge
I ask because the I2C lock is the main thing that made me decide to put aside the core I/O expander support work.
tulip sleet
i also found it intrusive in the past. I wonder the same about SPI
blissful pollen
I can only think maybe for large writes (displays?) if it is broken up for any reason and something else in background tasks jumps on the i2c bus and somehow that causes an issue. Maybe more for SPI. I'll keep mulling it over and let you know if I think of any cases.