the recent issue with SAMD21 not working with the lower_power PR was a TinyUSB thing, but it only manifested on my Dell i5, not on Scott's AMD box
#circuitpython-dev
tulip sleet
ivory yew
hehe doubt it
raven canopy
maybe i'll throw rasbian on an SD and test real quick. really wish they'd switch to a arm64 kernel; would make this easier than having to rely on Ubuntu's RPi images. π
tulip sleet
night, all, @ivory yew you're welcome, and @raven canopy good luck!
raven canopy
but i understand the backport issues they'd have with older Pis
nn @tulip sleet! thanks for the bug report link!
raven canopy
@ivory yew π (we had no doubt you'd get it... :D)
ivory yew
psh I did
I am well into uncharted territory right now
(for me, at least)
USB hosting, SPI slave, writing using Atmel/Microchip's silly C libraries.
raven canopy
ASF4. its... well, it is. hehe
ivory yew
Harmony is supposed to be different.
Maybe not "better", but different.
but meh, I'm invested here, so whatevs.
I've been super disappointed in the docs around this stuff
I mean, there's a literal 4 line example for SPI slave. A maybe a 2 sentence description.
raven canopy
The best part is when the docs say one thing, the errata says nothing, and it still doesn't work. I'm looking at you TC->PPW->direct_pin_capture. π
simple pulsar
@tulip sleet I've not made a note of the conditions but I've done some fairly normal things with Adafruit boards and managed to get Windows 8.1 in a state where 1) the device isn't recognised anymore by Windows and 2) a reboot is required and 3) shutdown of Window is borked and maybe pauses "forever" (at least 5-20 mins). It's very infrequent. I did have one just over a week ago. I might have seen this three times in total.
stuck elbow
that sounds like a problem with Windows or one of its drivers
simple pulsar
@stuck elbow That's what I would assume normally and was why I'd not mentioned it to date but Dan was talking about Linux keeling over due to possible USBisms.
idle owl
@tulip sleet @onyx hinge Am I correct in seeing there's no unstable release right now of CP?
onyx hinge
@idle owl I think that's right. @slender iron mentioned wanting to make an unstable release after lower power was merged so hopefully that'll happen soon
idle owl
Ok. Thanks.
manic glacierBOT
ionic elk
@lucid solar was your issue with UART reproducible (the BUSY issue you mentioned)? Or was it just a HAL question? If you were getting a HAL_BUSY usually that means something is wrong with the interrupts.
lucid solar
@ionic elk it's fixed, was because I was switching from interrupt based to polling, and it didn't reset properly.
idle owl
@prime flower You were tagged on this, in case you missed it. Thanks. https://github.com/adafruit/Adafruit_CircuitPython_MiniMQTT/issues/31
idle owl
@prime flower Thank you for following up!
tidal kiln
@raven canopy @tulip sleet yep - i'm seeing lock up's with arduino. it's a complete hardware freeze of the PC. mouse cursor doesn't even move. have to power cycle to recover. it happens with a basic sketch upload, right after upload is complete and it resets the board. of course it doesn't happen 100% of the time and i haven't been able to find a way to make it happen in a repeatable way for testing.
raven canopy
@tidal kiln i'm at 100% repro on a RPi headless with Ubuntu 20.04. its almost immediate once i go to bootloader. ssh pipe breaks, and activity LED is steady. still trying to get better traces, but the lockup isn't helping. π
tidal kiln
yep. it's not an easy thing to debug. and i always run into it when i'm trying to get something else done. so just punch my way through.
slender iron
I mean, there's a literal 4 line example for SPI slave. A maybe a 2 sentence description.
@ivory yew That's one reason I prefer register level. The data sheet is pretty thorough on what to do with the registers.
idle owl
@tidal kiln Hmm, do you have any suggestions for wording on the note for saying lux isn't calibrated? Is "Note: lux isn't calibrated!" enough or is there some better way to word that with more information?
tidal kiln
@idle owl that seems fine
ivory yew
@slender iron I'm not quite ready for that relentless experience just yet, but I certainly came close. π
tulip sleet
@raven canopy if you give a recipe for what you're doing on the RPi, I can try to reproduce and get a USB trace. Which boards do you see this on?
raven canopy
@tulip sleet Metro M4, bootloader 3.6 & 3.9. able to repro with a double-reset and screen session via reset(BOOTLOADER). again, this is only on the RPi w/20.04 which uses the dwc_otg driver. my laptop uses 18.04 & xhci and hasn't presented any issue for me.
idle owl
@lone axle Do you have a CLUE?
tulip sleet
do you see this only on M4?
raven canopy
well, so far yes. I tested a Grand Central yesterday (bootloader 1.x something) and it didn't have any issues; haven't tested newer bootloaders yet. i can try a M0 and nRF later.
i've been assuming that the issue was with my software/dependencies this whole time. its usually the case... π
tulip sleet
do you mean you call reset(BOOTLOADER) in the arduino program? Does it do anything else before that?
idle owl
@prime flower Thank you for labelling library issues - I really appreciate it!
raven canopy
no, in CircuitPython. connect board, CIRCUITPY enumerates. start screen session, and do:
import microcontroller
microcontroller.on_next_reset(microcontroller.RunMode.BOOTLAODER)
microcontroller.reset()
@gilded cradle Is this something you're able to look into? https://github.com/adafruit/Adafruit_CircuitPython_BLE/issues/77
lone axle
@idle owl Yep I do.
raven canopy
@tulip sleet and double-tap triggers the same. no diff in the dmesg/udev that i can see so far.
idle owl
@lone axle Interested in a coding quest?
tulip sleet
@raven canopy @tidal kiln I'm trying to figure out if it's TinyUSB in CircuitPython or the bootloader USB impl (which is not TInyUSB except on nRF), or the Arduino USB.
Try Arduino Blink and see if double-tap does the same thing. That's yet another USB impl.
gilded cradle
@idle owl yep, I can.
idle owl
@gilded cradle Thanks!
tidal kiln
@tulip sleet @raven canopy sry, can't help test right now.
raven canopy
now i have to remember where i put my Arduino... π€£
lone axle
@idle owl For sure!
tulip sleet
@tidal kiln np, if the RPi is reliably bad, we can get traces from that, and I don't have to keep waiting for my regular dev machine to reboot π
@raven canopy I meant Arduino Blink on the Grand Central
idle owl
@lone axle Keen. I'll tag you on a GitHub issue I created, and we can start a discussion there.
raven canopy
@tulip sleet badly worded joke. i haven't loaded Arduino IDE in like 2 years. π
tulip sleet
OK π good luck updating it all
raven canopy
so use Grand Central (which didn't have issues), not Metro M4?
idle owl
@lone axle Mentioned you here https://github.com/adafruit/Adafruit_CircuitPython_APDS9960/issues/25
tulip sleet
oh, I thought you were varying the bootloaders, no whichever one is bad is fine
raven canopy
kk
idle owl
@tulip sleet Is this a bug? https://github.com/adafruit/Adafruit_CircuitPython_BLE/issues/55 Trying to label things and am not sure if any labels fit this one or not.
tulip sleet
@idle owl it appeared to be a bug at first, but the underlying code is simple and seems to be doing the right thing. You or I could ping the OP and ask if this was resolved, since it's been a while
raven canopy
@tulip sleet no lockup with Arduino Blink -> bootloader via double-tap.
tulip sleet
@tidal kiln and your issues may have different causes π¦
prime flower
@idle owl np! I do that with the IO issues too, will keep doing on cpy-rel.
timber mango
@raven canopy I have a real true blue Arduino UNO R3 and an Arduino Mega R3.
onyx hinge
snagging a bite real quick before the meeting
tulip sleet
@slender iron do you want to talk about PacketBuffer for 10 mins now, or can be after you have lunch after the meeting
slender iron
@tulip sleet I can now
modern wing
Good afternoon -- happily lurking for the meeting.
onyx hinge
om nom nom complete
lucid solar
@onyx hinge something nice?
ivory yew
I'll be text-only today.
sterile bronze
lurking today
modern wing
Loud and clear Scott
thorny jay
lurking or text only. My notes are in the doc.
uncut nexus
lurking
old smelt
Lurking
lucid solar
oops, I forgot to go into the voice room (facepalm)
idle owl
timber mango
OK. It should be one topic for This week with two items below it.
onyx hinge
https://twitter.com/matt_trentini/status/1256908219307323392 https://melbournemicropythonmeetup.github.io/March-April-News-Roundup/
Twitter
thorny jay
Not bias, it's a great news roundup.
onyx hinge
https://twitter.com/MIT_CSAIL/status/1257323849131134976 https://link.springer.com/search?facet-content-type="Book"&package=mat-covid19_textbooks&facet-discipline="Computer+Science"
raven canopy
@timber mango i indented your bullets. is that what you wanted?
onyx hinge
tidal kiln
lurking
timber mango
Whom ever fixed my status update, thank you. π π
raven canopy
iirc, libraries include all reviewers/commenters. core & blinka might be different. i'd have to re-dig into it.
old smelt
@onyx hinge - thanks for posting those links. I'm on another call so just watching out of the corner of my eye. π
onyx hinge
@old smelt you're welcome ! PT would talk and paste links at the same time, but lately I've been doing it to help out whoever's running the meeting
lucid solar
I used blinka for the first time this week, thanks @gilded cradle
gilded cradle
That great to hear @lucid solar
onyx hinge
@unkempt dove we're doing "hug reports". Will you have any to share today, or are you just listening (lurking)?
ivory yew
re: converting a m0 feather to a hostess featherwing - maybe. No promises.
lucid solar
I just ordered a Serpente today! Looking forward to it arrive.
timber mango
@ivory yew What is a hostess featherwing?
unkempt dove
@onyx hinge i am just listening thanks
slender iron
welcome!
ionic elk
LOL
onyx hinge
(guides github repo are still on travis at the moment)
raven canopy
TravisCI gets no love... π¦
slender iron
ya, CI's are under appreciated
raven canopy
to be fair though, they can cause some frustration. hehe
lucid solar
and github actions is being annoying now...failing on my PR checks.
slender iron
ya, actions has been unstable
solar whale
π
unkempt dove
hi
idle owl
@slender iron Can we do my status update after yours? I need to head out in 10 minutes.
Thanks!
ivory yew
@slender iron have you tweeted about that? Happy to retweet it and see if any of my C++ friends have ideas.
slender iron
ivory yew
@timber mango a featherwing that handles USB hosting and communicates device data (like HID reports) to a host feather via SPI. No where near ready yet, I'm working through proof of concept stuff right now: https://twitter.com/theavalkyrie/status/1256449209609248768
Twitter
motorin'
ivory yew
yessss match cpython sockets
slender iron
@inland tusk don't forget to mute again
modern wing
+1 terraria -- really fun game, and the updates add a lot.
lucid solar
Love Terraria....I have my own server....have to invite you @lone axle
timber mango
@slender iron I absorb new information like a sponge. π
slender iron
π
onyx hinge
when compressed does the information leak back out?
timber mango
@ionic elk I am waiting for two Toasty boards. π
onyx hinge
check with limor for exactly what she's looking for
ionic elk
Quick question, do we have international font support in DisplayIO?
ivory yew
Out here making mp3 players like it's 2004. I love it.
prime flower
yay rfm9x fixes!
ionic elk
if I wanted fully translated debug information, for instance
solar whale
Done. Just add wood!
onyx hinge
@ionic elk depends what you mean exactly. Non-ASCII can be displayed. Terminalio has more limitations than the generalized Label.
gilded cradle
Broom is a nice touch
lucid solar
@solar whale gorgeous shed, fancy coming to Sweden to build one for me?
onyx hinge
advanced rendering like shaping and ligatures is way out there, right-to-left text layout is probably something you could code in Python as needed
ionic elk
Do we have character support for japanese/chinese?
ivory yew
Real on the nose for you to use your wood shed to store wood.
onyx hinge
and of course just the problems of storing, say, a repertoire of CJK characters in RAM, which is how fonts work now
ionic elk
I see stuff get spit out as unicode so I figured I just needed to install something
but yeah given the sizes I wasn't sure if it was supported
solar whale
@lucid solar would love to visit Sweden again, but I think I'm " one and done " for sheds ...
ionic elk
It almost certainly would be for the STM32-grade flash sizes but maybe not for other stuff
slender iron
check joey about the international font stuff too
ionic elk
@slender iron is joey a person or a software
slender iron
person
ivory yew
a "joey" is a baby kangaroo, obvs.
onyx hinge
@ionic elk do you want to write up an in the weeds for what you're thinking about? or cover it outside the context of the meeting?
ivory yew
swdio sounds awesome
ionic elk
I was just curious if it was a quick answer I don't mean to drag everyone into a discussion!
slender iron
joey castillo who is doing open book
ivory yew
Twitter
onyx hinge
sometimes
timber mango
@ivory yew Uh oh! You are on @slender iron's radar!
ivory yew
I stay on his radar. I live like a mile away from him. π
gilded cradle
ChipSnek
raven canopy
@slender iron you're referring to this, correct:
Adafruit CircuitPython 5.2.0 on 2020-04-09; Adafruit Grand Central M4 Express with samd51p20
Yes, the Circuitpython brand needs to be protected.
solar whale
Shrubery
onyx hinge
eel theme here
thorny jay
There is already an alternate name in the recommendation on CP trademark.
ionic elk
CIRCUITSHRUBBERY
crimson ferry
No. 2 The Larch
lucid solar
EmbeddedPy
thorny jay
I cannot find the name, but I read it while preparing my talk.
ionic elk
π³ π³ π³ π³
lucid solar
Can you not write something like "Based on CPY" ?
thorny jay
If you choose not to meet these requirements, then we ask you call your version of CircuitPython something else (for example, SuperDuperPython) and not use the Blinka logo. You can say it is βCircuitPython-compatibleβ if most CircuitPython drivers will work with it.
So I vote for SuperDuperPython
ivory yew
DefinitelyNotCircuitPythonISwear
timber mango
The word "Circuit" should not be in any new name, but "Python" can be in the name. So it could be SHRUBBERYPYTHON or something of the same form. π π
ivory yew
Adafruit CircuitPython 5.2.0 (Non-Adafruit build) on 2020-04-09; Adafruit Grand Central M4 Express with samd51p20 or something?
lucid solar
Based on Adafruit CircuitPython 5.2.0 on 2020-04-09; Adafruit Grand Central M4 Express with samd51p20
ivory yew
@lucid solar I like that too
lucid solar
@ivory yew that's nice
thorny jay
Branding
While we are happy to see CircuitPython forked and modified, weβd appreciate it if forked releases not use the name βCircuitPythonβ or the Blinka logo. βCircuitPythonβ means something special to us and those who learn about it. As a result, weβd like to make sure products referring to it meet a common set of requirements.
lucid solar
CircuitPython 5.2.0 (Non-Adafruit build) on 2020-04-09; Adafruit Grand Central M4 Express with samd51p20 I removed the Adafruit at the start @ivory yew
timber mango
ShrubberyPython, based on Circuitpython, created by Adafruit Industries.
tulip sleet
in REPL message: "CircusSnake (fork of CircuitPython) ...."
lucid solar
Probably remove the version # also
ivory yew
I like @lucid solar's + this:
Python based on Adafruit CircuitPython 5.2.0 on 2020-04-09; Adafruit Grand Central M4 Express with samd51p20
Note: This is a non-upstream build of CircuitPython. This build is not supported by the CircuitPython community.
This notice can be removed by setting the INTERPRETER_COMMON_NAME build variable.
I wouldn't put the instruction there
lucid solar
I get the point of a different name, I just feel that we shouldn't call it something new....just write Python, for example.
ivory yew
@onyx hinge something like:
Python based on CircuitPython 5.2.0 on 2020-04-09; Adafruit Grand Central M4 Express with samd51p20
Note: This is a non-upstream build of CircuitPython. This build is not supported by the CircuitPython community.
See {short_url} for more details.
I like that better @ivory yew
ivory yew
we can have a page explaining the branding stuff and how to remove the warning if you're establishing a real fork.
timber mango
ShrubberyPython, based on Circuitpython from Adafruit Industries
ivory yew
I honestly don't think we need a default separate name if we do that^
tulip sleet
"based on" -> "forked from" ??
ivory yew
"based on" has some precedence I think
slender iron
"circuitpython-compatible"
raven canopy
ArduPy..."forked". kind of. trying to be diplomatic π
tulip sleet
i also wonder about what the micropython folks will think about "based on", since we are "based on " MicroPython
ivory yew
We're not using any of their branding or trademarks, though
and that's the distinction
ionic elk
I'd like to see the warning contain something about tutorial compatiblity
ivory yew
like if they set the build variable for the name that could change to:
NewNamePython 5.2.0 on 2020-04-09; Adafruit Grand Central M4 Express with samd51p20
we'd drop the "based on CircuitPython" altogether if they set a new interpreter name
thorny jay
Sorry, I have to leave here the meeting.
ionic elk
Python for microcontrollers based on CircuitPython 5.2.0
onyx hinge
* enable pull request creation (DONE, but re-disabled temporarily)
* enable google Cloud Translation (MT) (DONE)
* some sort of backups (DONE; digitalocean weekly backups)
---- soft launch here?
* Move the github authentication API keys to adafruit organization
* Move the github PR creation to adafruit organization (adabot?)
* enable weblate.circuitpython.org via DNS
---- launch here?
* Move the Google Clould Translation (MT) to adafruit organization
cost: $20 / 1 million characters translated (probably not a lot of $$)
* Move the digitalocean droplet to adafruit organization
Python for microcontrollers based on CircuitPython 5.2.0 on 2020-04-09; Adafruit Grand Central M4 Express with samd51p20 Note: This is a non-upstream build of CircuitPython. This build is not supported and may break compatibility with libraries and tutorials. See {short_url} for more details.
I'd like to somehow imply that using the Circuitpython trademark is also not ok but I dunno if that makes it bloated
onyx hinge
ivory yew
Note: This is a non-upstream build of CircuitPython. This build is not supported and may break compatibility with libraries and tutorials and may not be distributed as an official CircuitPython build. ?
ionic elk
@ivory yew Noice. Concise
ionic elk
Python for microcontrollers based on CircuitPython 5.2.0 on 2020-04-09; Adafruit Grand Central M4 Express with samd51p20
Note: This is a non-upstream build of CircuitPython. This build is not supported and may not be distributed as an official CircuitPython build, as it may break compatibility with libraries and tutorials
See {short_url} for more details.
I like it
ionic elk
got rid of the double "and"
tulip sleet
"non-upstream build" -> "build of a fork of" ??
lucid solar
Note: This is a non-mainline build of CircuitPython. This build is not supported and may not be distributed as an official CircuitPython build, as it may break compatibility with libraries and tutorials
See {short_url} for more details.
Yeah I think a lot of custom builds won't consider themselves forks
which would be confusing
tulip sleet
how about just "unofficial"?
ivory yew
fine with me
ionic elk
sure
tulip sleet
it's symmetric with the mention of "an official CircuitPython build"
ivory yew
yup
raven canopy
think: USB logo/cert. at least, that's how i always analogize it. π
ionic elk
Python for microcontrollers based on CircuitPython 5.2.0 on 2020-04-09; Adafruit Grand Central M4 Express with samd51p20
Note: This is an unofficial build of CircuitPython. This build is not supported and may not be distributed as an official CircuitPython build, as it may break compatibility with libraries and tutorials
See {short_url} for more details.
onyx hinge
+++ b/_board/feather_m4_express.md
@@ -6,7 +6,9 @@ name: "Feather M4 Express"
manufacturer: "Adafruit"
board_url: "https://www.adafruit.com/product/3857"
board_image: "feather_m4_express.jpg"
+oshwa_certification: https://certification.oshwa.org/us000242.html
features:
+ - OSHWA Certified
- Feather-compatible
- Battery Charging
@tulip sleet Can you review this and merge it? I think it's set to go after my changes. https://github.com/adafruit/Adafruit_CircuitPython_GPS/pull/42
lucid solar
@slender iron ping me if you can't solve it within a day.
solar whale
π
modern wing
π
onyx hinge
in about 3 weeks we DO have a deviation from Mondays. Check our calendar!
raven canopy
thanks everyone!
gilded cradle
bye
modern wing
It's gotten warm on the east coast too
onyx hinge
oh oh, I forgot one other thing we'll need, a webhook from github to weblate so it updates when circuitpython master branch is pushed to
slender iron
@onyx hinge I can help with all of that. I'm a github admin
onyx hinge
@slender iron can you stick around now to work that out?
https://docs.weblate.org/en/latest/admin/continuous.html#automatically-receiving-changes-from-github
idle owl
@tidal kiln If you get a moment to take a look at this PR - you commented on the issue this stems from. https://github.com/adafruit/Adafruit_CircuitPython_Motor/pull/42
orchid basin
slender iron
@onyx hinge did you mean to include your checkout v2 change in the swedish PR?
onyx hinge
@slender iron it should probably be its own thing but I don't think it'll hurt. Just bringing the new build up to standards
slender iron
I sorted out the error when linking esp32s2. I needed -fno-rtti for linking and also -u newlib_include_pthread_impl
ivory yew
huzzah
raven canopy
@tulip sleet @tidal kiln of course. spend an evening reading/setting up kdump, put 3.9 bootloader and CircuitPython back on the Metro M4...and now it won't cause the lockup.
(β―Β°β‘Β°οΌβ―οΈ΅ β»ββ»
lucid solar
I sorted out the error when linking esp32s2. I needed
-fno-rttifor linking and also-u newlib_include_pthread_impl
@slender iron is it linking 100% now?
slender iron
Yup! Havenβt tested the binary though. I think I pushed the change
Will go heads down on it tomorrow
manic glacierBOT
half geyser
Can anyone tell me the story behind SPIM3 on NRF? It seems like it's the preferred SPIM controller, but it also seems like it requires an additional 8 kB of RAM to work around a hardware bug. Is that correct? Why is it more desirable?
timber mango
Hi Everyone, I'm a noobie in coding but very exciting by all the upcoming projects i can accomplish. I really would like to be able to code on circuitPython o my own, or at least modify the code the way i want. So i decided to start learn Python. As persons of experience, is it a good idea? do you have any recommendations? Thanks in advance and have a nice day !
lucid solar
Hi @timber mango and welcome 
Do you have a CircuitPython board?
timber mango
Hi @lucid solar
i do, but maybe a little old now, i have a Playground Classic. which is not python language. I want to at least go to Express version
lucid solar
Yes, that's correct, it doesn't support CircuitPython. The Playground Express is a great board to get.
The best way to get started is to follow the adafruit tutorials. https://learn.adafruit.com/adafruit-circuit-playground-express/circuitpython-quickstart
It's always fun to get the Neopixel LEDs flashing first.
timber mango
and i love LED π
lucid solar
tulip sleet
@half geyser
Can anyone tell me the story behind SPIM3 on NRF? It seems like it's the preferred SPIM controller, but it also seems like it requires an additional 8 kB of RAM to work around a hardware bug. Is that correct? Why is it more desirable?
@half geyser SPIM3 can run at 32MHz; the other SPIMs can only run at 8MHz. The hw problem is that SPIM3 doesn't do contention management properly on the RAM block it's using, so any access to that RAM at the same time it's doing a transfer can cause corrupted data.
There is workaround code in nrfx that doesn't use a whole 8kB block, but it doesn't work with the soft device. (NRFX_SPIM3_NRF52840_ANOMALY_198_WORKAROUND_ENABLED)
We thought it was worth wasting most of the 8kB block (the buffer is potentially a lot smaller) to allow for higher SPI speeds, which is particularly important for driving a display.
the RAM is segmented into blocks or regions (I forget the exact terminology), and 8kB is the size of one of those
We allocate the SPIMs in busio.SPI() in order of preference, so SPIM3 gets allocated first.
lucid solar
@tulip sleet do you have some time to help me with the debug_uart issue? I've got it to create it, but it HardFaults when trying to output the 'banner' (Adafruit CircuitPython 5.2.0 on ....) So I don't think it creates the object properly.
tulip sleet
@lucid solar do you have a backtrace from the call to HardFault_Handler? Did you set a breakpoint there? Is your code committed?
lucid solar
yes, yes, no ;)
I've made MP_DEFINE_CONST_FUN_OBJ_0(debug_uart_obj, debug_uart); I assume that I have to call that at some point manually to init the object? As it's outside of the python VM?
tulip sleet
is debug_uart a function? That is just a macro
lucid solar
Yeah, it is a func
tulip sleet
right, you need to create it somewhere. I am guessing it creates the UART object in a static place, not on the heap.
tulip sleet
do you want this to be usable from Python, or is it some compile option for doing debugging prints from C code?
i thought it was the latter but may be wrong
lucid solar
When running code.py, it needs to catch ctrl+c, and outputs when print() is called
When waiting for "any key to enter repl", that's outside python, but needs to catch any key
When inside repl, it needs to catch the keys and print the output.
tulip sleet
you mean it's a duplicate of the REPL serial connection?
tulip sleet
so it doesn't need to be accessible from Python as board.DEBUG_UART(). Have you seen multiterminal? https://circuitpython.readthedocs.io/en/latest/shared-bindings/multiterminal/__init__.html
in micropython this is dupterm or something like that
lucid solar
but won't that stop working outside of code.py/repl?
tulip sleet
yes, I'm just suggesting that it might be code to study
lucid solar
This should be available always...even if the py VM isn't running
tulip sleet
so do you want the debug-uart to be created when CPy starts? Then it should be created in main.c or something main.c calls. And should it always be created? Are you dedicating some pins to it?
i'm thinking you could piggyback on the multiterminal code to do the input handling, since the mechanism is already there
it's not clear to me it does output too, but it could be extended to do that. I don't know of anyone using multiterminal now, so not sure if it definitely works
lucid solar
I was thinking of creating it in serial_init, as that's where usb_init is called
tulip sleet
that sounds fine, but it's dedicated pins?
lucid solar
yes
tulip sleet
so just call common_hal_busio_uart_construct() in serial_init(). You don't need the pins.c addition.
i see that common_hal_debug_create_uart() basically does that
lucid solar
yeah π
tulip sleet
MP_OBJ_FROM_PTR(&debug_uart) in serial.c serial_init() doesn't do anything, that's just a type-casting macro
lucid solar
Yeah...ignore that...shouldn't have been committed
tulip sleet
so if you call common_hal_debug_create_uart() in serial_init(), I think that should do it
lucid solar
ok...am testing on the board now...
n_bytes=<optimized out>,
has_finaliser=<optimized out>,
long_lived=<optimized out>)
at ../../py/gc.c:489
#2 0x08066d0c in common_hal_busio_uart_construct (self=0x2000a2a4 <debug_uart_obj>,
tx=0x809c628 <pin_PD08>,
rx=0x809c630 <pin_PD09>, rts=0x0,
cts=0x0, rs485_dir=0x0,
rs485_invert=false, baudrate=115200,
bits=8 '\b', parity=PARITY_NONE,
stop=1 '\001', timeout=1,
receiver_buffer_size=64)
at common-hal/busio/UART.c:218```multiterminal was only ever implemented on esp8266
lucid solar
it goes into safemode after that
tulip sleet
The trouble is that the UART constructor creates a buffer on the heap, and there's no heap yet
so you need to add some way to have a statically allocated buffer
(I think, still looking to confirm)
right, the _construct calls ringbuf_alloc()
lucid solar
Yeah...that's why I didn't think it could be done...(easily)
tulip sleet
the _construct also looks for a free UART, but you want a particular one in mind. So don't call _construct, but do most of the work it does, using the dedicated UART and static buffer and object. Or add some args to _construct()
I'm wondering if this is useful on other than STM, as an option. There are times when we want to do debugging and USB is broken, but we'd want it to be compile-time optional and usually off, since we don't want to use up valuable pins
it's just the STM dev board that has these not-very-useful pins available?
slender iron
@tulip sleet I asked for _construct to be used
tulip sleet
well, we need an optional arg to pass in the ringbuf
slender iron
and suggested adding a pointer parameter to _construct
tulip sleet
ok, I didn't read the comments
lucid solar
and suggested adding a pointer parameter to _construct
@slender iron I missed that comment...or didn't understand it maybe π
lucid solar
Nah, it was because I read that at ~02:00 and didn't re-read it today
tulip sleet
ok, great minds think alike π
slender iron
π
tulip sleet
also you would need something to specify the specific UART periph you are using
lucid solar
do I need to have common_hal_debug_create_uart (shared_module)? or just call common_hal_busio_uart_construct directly myself?
tulip sleet
if it's a convenient abstraction, then keep it. If you are not exposing this UART to Python, then it doesn't need to be in shared-module. It can just be some extra code in serial.c or somewhere in supervisor. share-module is for implementing things that have a Python interface
slender iron
you don't need to pick the periph. just provide the pins to construct
tulip sleet
ah good point
lucid solar
ok...I think I get it now...will test it out...
tulip sleet
i thought the periph was special somehow
slender iron
nope, it's just a UART connected to another chip
we might need a baudrate but that should be all I think
it'll simplify all of the init
tulip sleet
do we need all the UNAVAILABLE stuff or is it just another UART?
or are there other reasons for that
slender iron
we shouldn't
but we need to add something to uart to alert when a ctrl-c is entered
tulip sleet
is there latent support for that via multiterminal?
slender iron
no
multiterminal was only implemented on esp when all of the serial processing was port specific
tulip sleet
it's still in the doc, maybe we should remove it
lucid solar
tx=0x809c690 <pin_PD08>, rx=0x809c698 <pin_PD09>, rts=0x0, cts=0x0,
rs485_dir=0x0, rs485_invert=false, baudrate=115200, bits=8 '\b',
parity=PARITY_NONE, stop=1 '\001', timeout=1,
receiver_buffer_size=64) at common-hal/busio/UART.c:181```GPIO_InitStruct.Alternate = self->tx->altfn_index;
tulip sleet
@lucid solar if you look for uses of MICROPY_KBD_EXCEPTION, that will help you find where the interrupt char (default is ctrl-c, but could be something else) is handled
lucid solar
@tulip sleet thanks, I think I understand that bit ok.
slender iron
π
idle owl
@timber mango Consider the Circuit Playground Express or Circuit Playground Bluefruit as the best place to get started. Both have all kinds of sensors, LEDs, buttons etc built in. The Bluefruit includes Bluetooth. There is a library designed to make getting started super simple. Here is the guide for it: https://learn.adafruit.com/circuitpython-made-easy-on-circuit-playground-express/circuit-playground-express-library
lucid solar
@slender iron @tulip sleet does it make sense I also need to construct the TX/RX pins in serial_init?
tulip sleet
the pin objects themselves are static; they aren't constructed
pin_PA_00 or whatever
lucid solar
When I try to write to the UART, it calls this mp_raise_ValueError(translate("No TX pin"));
tulip sleet
what are your arguments to _construct()?
lucid solar
115200, 8, PARITY_NONE, 1, 1.0f, 64, buf_array);``` const mcu_pin_obj_t* tx = MP_OBJ_TO_PTR(DEBUG_UART_TX);```$9 = (const mcu_periph_obj_t *) 0x0```that is weird ... looking
I think you'll need to step through the complicated uart assignment code in _construct and see why self.tx is not being set. Check that the arg coming in is not NULL too , but it shouldn't be.
manic glacierBOT
lucid solar
@tulip sleet it's calling common_hal_busio_uart_write before common_hal_busio_uart_construct it seems
tulip sleet
OK π
lucid solar
run_boot_py before serial_init
tulip sleet
the idea is that boot.py is run before any USB setup, so it can modify it if necessary
lucid solar
Yeah, I see that...where is best for this code then?
manic glacierBOT
tulip sleet
if you want the debug uart to be functional then, then its initialization should be moved further back. It's fine to create a new routine if you have to
lucid solar
ok...thanks
tulip sleet
a reminder that the code is not sacred, don't be afraid to change it, but consider the larger structure and keep the style, etc. this is probably obvious to you
lucid solar
Yeah, I try to not change too much...but sometimes you just have to.
void serial_init(void);
``` Just made this...simple I think@tulip sleet I've got it writing just fine now...still have to fix the reading, but that's something I can fix.
tulip sleet
yay!
zinc maple
EduBlocks for CircuitPython has got a refresh in the beta. Some new features are:
- Account system for file saving and sharing
- Split view (Blocks and Python on same screen)
- Few ui updates
- Translations
onyx hinge
@slender iron if you have some time to work on github settings now that we're shifting to hosted weblate .. the recommendation is to install the "weblate app". Apparently what I can do is "request" this, which I guess is going to be reviewed by a github project admin such as you?
https://docs.weblate.org/en/latest/admin/continuous.html#github-setup https://github.com/apps/weblate
" A request to install Weblate has been submitted on the @river quest account. "
(sorry for the unintentional mention, woops)
lucid solar
@slender iron can you create a doc for next weeks CPY meeting?
slender iron
@lucid solar yup! thanks for the reminder
lucid solar
No problem...just ready to start my hug_reports/status_updates
slender iron
Here is the notes document for Mondayβs CircuitPython Weekly meeting. Everyone is encouraged to attend! Please add your hug reports and status updates even if youβll be attending the meeting - itβs super helpful! If you are unable to attend but would still like to include updates, feel free to include them in the notes and weβll read them off during the meeting. Hope to see you there! <@&356864093652516868> https://docs.google.com/document/d/1bzaYoTmWls07acn7JB4z0CX9BiTymQH-22rwZzXIObE/edit?usp=sharing
slender iron
Here is the notes document for Mondayβs CircuitPython Weekly meeting. Everyone i...
tulip sleet
@slender iron I was going to make a new release of ble MIDI. Your last release was 0.8.0. Do you want to keep the sub 1.0.0 releases or should I bump it to 1.0.0?
tulip sleet
i'll do 1.0.0, since it's functional
slender iron
π
tulip sleet
will do in a bit; just going out
slender iron
k, np
onyx hinge
@slender iron great, I'll see what happens if I have it try to push changes back to github...
hmmm there's an option in weblate to update .po files from .pot files. I wonder if we could change 'make translate' to just update the .pot file.
On branch master
Your branch and 'origin/master' have diverged,
and have 1 and 15215 different commits each, respectively.
(use "git pull" to merge the remote branch into yours)
This seems really wrong
slender iron
I didn't want that in case folks edit .po files manually
timber mango
Circuitbrains is almost funded! The only need $671.00 more. I know we can help them. Several of us have already pledged for two of them. Are are others here who would like to see Circuitbrains become a reality?? https://www.crowdsupply.com/null-byte-labs/circuitbrains-deluxe/updates/reduced-funding-goal
ionic elk
I feel for those folks but I was worried about it from the moment I saw that it didn't come with USB. That's just such a big barrier to a board that seems Circuitpython focused.
raven canopy
being end-product/project focused, i can see why they only broke out lines. def not the "standard" 
board, though.
ivory yew
Yeah it works great for things like Sol which have the USB port in a specific spot and us a different connector (USB B)
I'd love to see a breakout/programming board for it.
kind river
It probably wouldn't be too hard to make a USB "cover" that could be soldered or just pressed on top of it for programming. It would have been nice if the USB and power pins were together though.
lone axle
on a busio.SPI object is there some way to access to baudrate that it was configured to use?
I found some old code trying to access that value like this spi._spi.baudrate. I'm not certain if it ever worked but now it does not because there is no _spi attribute.
tulip sleet
It is the actual frequency, not the requested frequency, so you can tell how it got altered by the granularity of the peripheral setting
lone axle
Thank you
crimson ferry
@lone axle That's in a library? I think it could be useful to expose that from within libraries since user code accessing .frequency just get the default, 24MHz I think, since the bus is typically unlocked in user code.
lone axle
It was not directly in a library, but rather was some code illustrating a problem in a library on this issue: https://github.com/adafruit/Adafruit_CircuitPython_TLC59711/issues/8
simple pulsar
FYI https://learn.adafruit.com/adafruit-circuit-playground-bluefruit/circuitpython mentions 5 is still in beta.
lone axle
They were printing the value of the baudrate to show that the library was not respecting previously set values.
.frequency does return 250000 for me if I don't configure the spi for anything specifically myself.
crimson ferry
yeah, the only way to get the actual frequncy for a given SPI device now is to be within it's unlock code
But it could be useful to know the actual
lone axle
it does also show new values if I set them with configure
crimson ferry
but I believe that's configuring a global default, unless you are within the code that is accessing a specific device unlocked. for example, If I do .frequency in my user code, I'll get 24 or 25MHz, whatever it is, but as each library is invoked, those devices will use their own specific configured SPI frequencies (but the user can't find out what they are at the library API level)
because user code running devices through libraries only sees the SPI bus unlocked (corrected my just prior statement)
crimson ferry
just realized I can cheat sometimes: with esp._spi_device as espspi: print(espspi.frequency)But for displayio.FourWire...?
crimson ferry
LOL```spi = board.SPI()
while not spi.try_lock():
pass
spi.configure(baudrate=microcontroller.cpu.frequency)
spi.unlock()
Why not just go ahead and multiply it by two, just to be sure you get the max.
crimson ferry
π
engineers always multiply by two, right?
just in case
that did get me up to 45MHz on Teensy 4.0
timber mango
No, they multiply by a power of 2 :wink: :stuck_out_tongue_winking_eye:
crimson ferry
2 is a power of 2 :p
timber mango
Of course, but that is finite. :p
crimson ferry
SAMD51: cpu=120MHz, max spi=24MHz, esp=8MHz. Teensy 4.0: cpu=600MHz, max spi=45MHz, esp=7.5MHz
it would be useful to know what actual speed spi displays are running at
crimson ferry
Is there a way to shift another ~512K of flash on the Teensy 4.0 to the CIRCUITPY volume? It seems it only gets 1MB of the 2MB, whereas SAMD51 chips with 512K+2MB get 2MB.
Itsy Bitsy
(1024, 1024, 2024, 544, 544, 0, 0, 0, 0, 255)
Teensy4.0
(512, 512, 2001, 8, 8, 0, 0, 0, 0, 255)```ya, it shouldn't be too hard
crimson ferry
should I make an issue?
slender iron
most imxes will have more than 2mb though
crimson ferry
so on 51J20 (like PyPortal) I'd assume only half of the on-chip 1MB is full?
Teensy's all I have for now (dev boards are overkill for me)
crimson ferry
It's not critical right now, I can delete some libraries for peripherals I can use on other boards, I just got caught off-guard when I hit it sooner than expected compared to SAMD boards. Another 512K would be really nice though.
slender iron
ya, we could adjust it. it's just hard to change in the long term without erasing the existing fs
crimson ferry
ah, so someone loading 5.4.x or w/e would get zapped, makes sense. No issue then, I'll keep pushing with what's here π
slender iron
π
half geyser
Thanks for the explanation, @tulip sleet ! That really helps a lot.
blissful pollen
I managed to convert one of the AHRS algorithms (Mahony the easiest) to circuitpython to see if it would run. Happy to report it seems to do okay on a Feather M4 express. I'm getting the same results as the original C arduino code and have it running on a LSM9DS1 sensor via I2C
https://github.com/gamblor21/CircuitPython-AHRS
Interestingly enough either North/South or East/West are reversed in the original and in my version - or I am doing something wrong with the sensor setup (quite possible)
idle wharf
I managed to convert one of the AHRS algorithms (Mahony the easiest) to circuitpython to see if it would run. Happy to report it seems to do okay on a Feather M4 express. I'm getting the same results as the original C arduino code and have it running on a LSM9DS1 sensor via I2C
https://github.com/gamblor21/CircuitPython-AHRS
@blissful pollen I had no idea what AHRS is, so I looked it up and gave you a PR. π
half geyser
Here's a strange one: I'm debugging Circuitpython on an NRF52833. My debugger is a Raspberry Pi running openocd. For whatever reason, reset halt doesn't work. reset works, and halt works, but it just won't reset halt.
This is particularly frustrating because the only reliable way I've found to reprogram it is to do nrf51 mass_erase, followed by reloading the softdevice and bootloader.
Here's an interesting hint as to what could be going wrong: it can't reboot into the bootloader anymore. That is, resetting the device appears to reset even the POWER block.
stuck elbow
I thought NRF52833 is not supported? only NRF52840
half geyser
I've been adding support for it. The PCA10100 is supported, as is Simmel which is a thing I'm working on.
Ooh, I'm glad I mentioned something. That caused me to look into openocd. For whatever reason, a "sysresetreq" isn't working, but "vectreset" is. So in openocd if I do cortex_m reset_config vectreset then I can flash it.
blissful pollen
@idle wharf Hey thanks! By the time I was committing things I was half asleep so some more explanation is good
slender iron
any objections to me releasing 5.4.0-beta.0 today?
lucid solar
@slender iron I've made a bit of progress with SWDIO, I've got an 'empty' module working (I can import swdio), so I understand what needs to be created. What I'm working on now is a 'spec' for the Python module itself, what methods/properties etc etc it needs.
prime flower
@slender iron no objection, only encouragement!
slender iron
nice @lucid solar !
idle owl
@simple pulsar CPB guide updated. Thanks for catching that.
@tulip sleet If you have a moment: https://github.com/adafruit/Adafruit_CircuitPython_GPS/pull/42
simple pulsar
@idle owl np - i've been writing some code for the CLUE, then realised most of it would work on CPB+Gizmo so I've made it run on both, hence the interest
manic glacierBOT
ionic elk
@slender iron would you consider problems with getting japanese/chinese characters to display in Circuitpython to be too esoteric to make issues of?
idle wharf
@slender iron would you consider problems with getting japanese/chinese characters to display in Circuitpython to be too esoteric to make issues of?
@ionic elk FYI: There are blog posts and learning articles about CP supporting Unicode Fonts... https://blog.adafruit.com/2019/05/29/new-guide-pybadge-conference-badge-with-unicode-fonts-conference-badgelife-circuitpython-pybadge-adafruitlearningsystem-adafruit/
My_Point: So I would think its expected to work.. π
slender iron
@ionic elk an issue is always good. Not sure itβs worth the effort to fix though
ionic elk
Seems like it's an Adafruit_CircuitPython_Display_Text issue anyway
not really with anything core
lucid solar
@slender iron @ionic elk Could you advise what would be the minimum RAM/Flash for a STM32 CircuitPython compatible mcu?
slender iron
for SAMD21s its 32k ram and 256k flash
192k flash would be equivalent if it had external spi flash
lucid solar
ok...I'm looking at the F411RE, that's got 512/128...so plenty π It's nice when I'm producing only a few boards, so an extra $ doesn't matter.
slender iron
π
onyx hinge
@slender iron @tulip sleet do you want to weigh in on the naming of this new module for interfacing to SD cards? sdcard_spi has been suggested. I only want to rename it once π
onyx hinge
I think because
A SDIO (Secure Digital Input Output) card is an extension of the SD specification to cover I/O functions. SDIO cards are only fully functional in host devices designed to support their input-output functions (typically PDAs like the Palm Treo, but occasionally laptops or mobile phones). These devices can use the SD slot to support GPS receivers, modems, barcode readers, FM radio tuners, TV tuners, RFID readers, digital cameras, and interfaces to Wi-Fi, Bluetooth, Ethernet, and IrDA. Many other SDIO devices have been proposed, but it is now more common for I/O devices to connect using the USB interface.
SDIO is overloaded and it doesn't do any of these things
tulip sleet
sdcardio is too much like "cardio?
there are already a couple of GitHub repos called "SDCardIO", which is actually good, I think, since this is related
what would be in the namespace below "sdcardio" (or whatever name)?
timber mango
Sdcardio sounds like a heart exercise.
onyx hinge
@tulip sleet right now a single class SDCard that is constructed almost like an adafruit_sdcard.SDCard (except that it takes a Pin, not a Digitalio, for the chip select line right now)
tulip sleet
I kind of like sdcardio, despite the accidental "cardio" part, since it matches all the rest, and it's different than "sdio", which I agree would be confused with a widely (?) known acronym.
strange adder
Curious if there are any thoughts to add a dev board+circuitpython support for the Dialog Seminconductor Smartbond DA145XX or DA146XX product lines?
lucid solar
"DA14695 Development Kit β USB" looks nice...it has a built-in debugger and hub so you can use the USB from the DA14695 and debugger at the same time.
strange adder
I just got a DA14531-00FXDEVKT-U 2 days ago for a particular project I have in mind. Pretty cheap.
lucid solar
I would say DA14695-00HQDEVKT-U is the minimum you'd need...I might get one, I like the idea of the debugger/hub
strange adder
Yeah, Keil debugging is a cinch. You can use JTAG or debug over UART
But if I remember correctly the newer DA145XX series consumes around half the power of the DA146XX
lucid solar
Shame they only do BGA packages...or it'd be nice to add to some projects I have
slender iron
@onyx hinge sdio is my preference actually because we can make it a generice API for both SPI and SDIO modes. SDIO can then be supported with an available SDIO peripheral then
also consider it would be cool to allow boards to construct them and have them long lived
onyx hinge
and substitutible for the "/" drive even
slender iron
maybeeeeeee
slender iron
@strange adder native USB is a must too. doesn't look like the DA14531 has it
strange adder
oic, thanks!
slender iron
π always interesting to see different micros
ionic elk
Is it possible to load a font from an SD card?
crimson ferry
@ionic elk absolutely
ionic elk
@crimson ferry cool.
also, though, I'm finding that making a foreign-language text group takes a full 40 seconds to create? On both STM32 and Atmel M4
crimson ferry
I always check "/" first, then if SD Card is present, check that if not found on "/" π
ionic elk
Seems crazy slow
crimson ferry
fast SD card? i tend to use Class 10 / A1 or better (probably overkill), load font and load glyphs at startup doesn't seem to take much time
time is relative I supppse π
ionic elk
I'm afraid if I go trying it on a 2.5MB font (chinese) it'll take all day
By which I mean 6 minutes. which is still a long time to make one text box
crimson ferry
1.4 seconds for a 32-point font (220K), loading ~15 glyphs. About the same time to load ~45 14-point glyphs (SAMD51). I expected more variation.
ionic elk
That's a LOT faster than my setup
crimson ferry
What's the alternative for lots of glyphs and / or large point sizes
ionic elk
Is it impacted by font size?
crimson ferry
I would think so
raven canopy
@tulip sleet @tidal kiln just so i can jinx it, i think i've found the issue/solution to the bootloader woes. after comparing 3K lines of udev debug, tracking block setup/teardown, and testing a rule, i think we're running into this: https://github.com/systemd/systemd/issues/7587
i can't say definitively, but i think it may be either the HID objects or block device isn't getting torn down completely which throws the cycle of death when it re-enumerates.
ionic elk
What's the size of the font you're using? I've been trying it on 250kb-350kb fonts that are 12pt, and I'm getting about 40s for 12 glyphs.
crimson ferry
The 32 is 220K, the 14 is 114K
ionic elk
Seems like my performance is clearly quite a bit worse
over an order of magnitude
1-2 seconds would be totally acceptable for my application but 40s is, uh, not
slender iron
@ionic elk loading glyphs from a bdf file is a linear search through a text file
ionic elk
@slender iron yeah it makes sense to me that it wouldn't be super fast, but anecdata is reporting a load time of 1.4 seconds for a library and number of glyphs just slightly smaller than mine
so I'm guessing I've got something else going on
slender iron
the glyph position in the font matters
ionic elk
ah, that could be it.
lucid solar
@slender iron what does FREEZE ../../frozen/circuitpython-stage/pygamer mean?
ionic elk
I don't suppose it'd be possible to give it an offset to start counting from? Hiragana starts at 12353, if it does that for every glyph, makes sense it might be bogging down
slender iron
@lucid solar it means that the stage python code is parsed and stored into flash
@ionic elk you really want a cache of byte offsets into the bdf
utf-8 offset won't help you
lucid solar
aha...just wanted to check that it wasn't meaning that it was freezing the code at some point...as I'm checking that my changes build on a out-of-port board before I commit.
ionic elk
@slender iron I assume this would have to be a manual rewrite of Label then
slender iron
not necessarily, the font loader that Label uses
are you doing this on adafruit time?
ionic elk
no of course not
slender iron
π
ionic elk
personal project that I was hoping to get done in an hour this morning and didn't quit on because I'm stubborn
slender iron
@ionic elk I understand that. π storing an offset cache in the BDF font would speed it up
also having label bulk load new glyphs would too
ionic elk
I don't see offset cache as an option in otf2bdf π¬ : is that something I need FontForge for?
slender iron
there is a binary version of bdf called pcf that would also be faster to search since it's binary
you'd do it manually in bdf
ionic elk
How does label bulk load? Is that documented somewhere?
slender iron
it doesn't bulk load "load_glyphs" to the font iirc
this is where it loads each character individually: https://github.com/adafruit/Adafruit_CircuitPython_Display_Text/blob/master/adafruit_display_text/label.py#L124
ionic elk
So, some rewrites but not super crazy ones
slender iron
optimizations
ionic elk
ye
Might need a pretty big table for Kanji
But if it's designed for SD use that might not be a big deal
From working today it seems like fitting a Kanji font of any readable size on 2MB internal flash is basically impossible. It's tantalizingly close but just barely doesn't fit.
slender iron
did you talk with joey yet? he has his own format for storing a bunch of characters
ionic elk
I have not, maybe he could have saved me some pain today
FontForge is the devil by the way, I would recommend not recommending it.
slender iron
do you know of something better?
ionic elk
otf2bdf works great
MakerMelissa recommends it in the badge tutorial.
If you just want to view them BirdFont is clunky but at least it doesn't segfault every other minute.
ionic elk
Sorry if I've made some grumpy messages here, I spent a lot more time than I expected on FontForge today, sorry.
slender iron
np, I'm heads down on esp32s2
timber mango
@ionic elk I am sure anyone who has tried to use FontForge will understand. I installed and tried it once, got several segfaults, and immediately removed it. π
ivory yew
making some good progress on my USB host featherwing
got this tiny little bit of code working today:
hostess_midi_in = HostessMidiPortIn(spi)
upstream_midi_out = usb_midi.ports[1]
while True:
# Get a MIDI event from a USB
# MIDI device connected to the
# Hostess Feathering
midi_event = hostess_midi_in.read()
if midi_event is not None:
# Send the MIDI event to the computer
# over this Feather's USB MIDI connection
upstream_midi_out.write(midi_event)
print(midi_event)
slender iron
@vale spire does the xtensa stack grow up instead of down?
onyx hinge
according to some doc I found, it grows down as we're accustomed to [Xtensa Instruction Set Architecture (ISA) Reference Manual p.588]
manic glacierBOT
mental nexus
@lone axle Thanks again for pointing me in the right direction on a few things (USB host controller for UART and write access to the CircuitPython file system) and for your encouragement!
I can now present to you cyberDΓCK:
https://hackaday.io/project/171269-cyberdck-a-circuitpython-anatidae
You can now edit and run CircuitPython code all on the same hinged duck.
It was a fun couple of weeks and I learned a ton! Thanks to all the community members for their βcan doβ attitude and willingness to help us less knowledgeable folks get better!
half geyser
A quick update on my reset issues yesterday: It seems that, for whatever reason, resetting this board (a Simmel board base on an nRF52833) now causes a brownout event. That's the source of my reset woes. The RESETREAS is all zeroes, which indicates a brownout. The GPREGRET and GPREGRET2 retained registers are also zero. Doing a core reset does not cause this problem. So it's not a problem with circuit python, it's a problem with how I'm using this chip.
I'll keep looking for the cause of the brownout. I don't think it's caused by VDD drooping during reset, since it's connected to a relatively beefy power supply. And the thing only draws a few mA.
ivory yew
@mental nexus that's rad.
mental nexus
Thanks @ivory yew , it was a labor of love.
In the process I created a simple text terminal class that can write to a display using displayio tilegrids. It has basic scrolling function and keeps track of the entry point. For the editor running on the crberDΓCK, I used three terminals, one for main text area, a highlighted one for the status row and one single tile terminal for the cursor. Iβve never released anything on github but if this could be useful to someone I can figure it out.
half geyser
I'll keep looking for the cause of the brownout. I don't think it's caused by VDD drooping during reset, since it's connected to a relatively beefy power supply. And the thing only draws a few mA.
Interestingly, reset works just fine until circuit python is started, then it stops until I reflash the board. The plot thickens...
half geyser
Huh. It's the write to NRF_UICR->REGOUT0 that's causing it. Interesting...
Setting it to 3.0V (instead of 3.3V) seems to fix the issue.
broken kettle
The voltage from the battery IS 9V
There are 6 number of globes.
How much voltage are across each globe
There is 1 glove and then 2 globe and then 3 globe. What are the brightness of the globes
Help
low sentinel
Can I not retain an arg pointer that I get in a shared-bindings _make_new function?
I have an arg like { MP_QSTR_some_list, MP_ARG_OBJ, {.u_obj = MP_OBJ_NULL} } and when I inspect it during _make_new it is sane (mp_obj_list_get shows lengeth=5) so I hold that .u_obj on a shared-module's some_type_t structure. Later when I go to use it, mp_obj_list_get shows length 1393 and I crash because there were only ever 5 items in the list.
edit:
self:20004570 calling action
self:20004574 doing action on self
Logging pointer addresses made it obvious that I was calling a function on the address of a member in another "class" function. :sigh:
broken kettle
half geyser
Final update on my NRF52833 woes: VDD got disconnected, and I was powering it off of VDDH. This apparently works fine, except reset is all wonky as the regulator resets itself causing a brownout. I re-shorted VDD to VDDH, and now it's behaving like it should. Thanks everyone for hearing my stream of consciousness debugging session!
main meteor
@broken kettle This doesn't really belong in #circuitpython-dev . However, did you figure it out?
broken kettle
Yeah I did mate @main meteor thanks for caring man love you β€οΈ
lone axle
@mental nexus that is awesome!
mental nexus
Thanks @lone axle I really appreciate your help.
slender iron
@sour lynx do you know the best way to get the current stack pointer? I need it for garbage collection
uncut nexus
Anyone have experience building older versions of CircuitPython? I have run 'git checkout' on 5.0.x and 5.1.x and they build; but when I load them they both end up in an infinite loop in DummyHandler.
hushed hedge
@mental nexus Well, that's a hack that quacks! We may be only a small amount of 3D printing away from a CircuitPython TRS-80 Model 100.
mental nexus
Quack-hack, thatβs a good one. Iβm thinking itβs more VIC-20 or Commodore 64 but I picture it powered by a python that quacks running on a hamster wheel.
hushed hedge
I'll have to ask one of my python-owning friends how successful they are at getting it into a hamster wheel.
slender iron
@uncut nexus they should work. make sure you update submodules as well
uncut nexus
yes, did that. Thanks!
@slender iron - does the platform you're needing a stack pointer for support '__builtin_frame_address(0)' in the compiler?
slender iron
not sure. I didn't know that was an option
tries it
looks like it does! thanks @uncut nexus !
sour lynx
@sour lynx do you know the best way to get the current stack pointer? I need it for garbage collection
@slender iron There iscpu_hal_get_spfunction in#include "hal/cpu_hal.h"
however... keep in mind that with Xtensa Window ABI, not all previous stack frames are guaranteed to be on the stack.
at most 48 physical registers from the previous stack frames may still be in CPU registers, not flushed onto the stack.
I remember explaining to micropython folks how to handle garbage collection because of this, lemme find the bit of code...
Okay, totally not what i suggested, but also works: https://github.com/micropython/micropython/blob/e2def200bfb6f68801568c6b26f73ee86b549be2/ports/esp32/gccollect.c#L42-L47
basically, make a recursive function which uses 8 physical registers per call βΒ call it 64/8 times and all the registers are guaranteed to be on the stack.
slender iron
Ok interesting! Will look a bit later. Mainly now Iβm not getting the right range with the getstartofstack method from free rtos
sour lynx
pxTaskGetStackStart(NULL) should be okay, i think
and the size should be about the size of the stack size you specify during task creation. For the main task (where app_main runs) it is given by CONFIG_ESP_MAIN_TASK_STACK_SIZE
slender iron
Ya. Thatβs what Iβm doing for size. When I get the current frame itβs a higher address than the start of the stack though
onyx hinge
is the "start" the initial SP address, or just the lowest byte of the allocated memory region?
n/m the docs say what it's supposed to mean
Returns the highest stack memory address on architectures where the stack grows down from high memory, and the lowest memory address on architectures where the stack grows up from low memory.
simple pulsar
@slender iron Is using assembler a no-no here?
sour lynx
Ya. Thatβs what Iβm doing for size. When I get the current frame itβs a higher address than the start of the stack though
@slender iron makes sense β start of the stack is literally the start of the stack memory region, i.e. its lowest address
Not the logical start of the stack, i.e. address of the first frame.
slender iron
@sour lynx thatβs not what the doc implies
@simple pulsar assembly is ok if thatβs the only way
sour lynx
Indeed! The doc says it is the other way around. Let me check what upstream freertos does.
slender iron
I couldnβt find it in freertos 10
(Seems like itβs always the lowest address though)
sour lynx
yes... the definition we are using seems to match https://www.freertos.org/FreeRTOS_Support_Forum_Archive/January_2016/freertos_How_to_see_upper_lower_bound_of_a_tasks_stack_1e322e0bj.html
slender iron
So itβs always the lower address?
sour lynx
Yes, lower address.
slender iron
Kk. Thanks!
sour lynx
It seems that this function comes from Segger SystemView patch: https://github.com/sifive/Segger_SystemView-metal/blob/master/SystemView/Sample/FreeRTOSV10/Patch/FreeRTOSV10_Core.patch
slender iron
I made lots of progress yesterday
sour lynx
we have pulled it in the commit which has added SystemView tracing support: https://github.com/espressif/esp-idf/commit/8d43859b6ab
i guess we didn't check the docs. Have filed an issue internally to get this fixed.
is there a way to avoid rendering gigantic link preview images here?
sour lynx
tnx
onyx hinge
fwiw this is where I pasted the docs from, I see now that it's "esp32" not "esp32-s2", sorry for any confusion.
https://docs.espressif.com/projects/esp-idf/en/latest/esp32/api-reference/system/freertos.html
slender iron
Only other weird thing Iβm seeing is that my makefile canβt build bootloader.bin but I can if I call ninja directly
Will keep poking stuff today. The partition api is working well for our file system
@onyx hinge the esp32 docs generally apply to the s2
slender iron
welp, I have the right range now. circuitpython caused a freertos "stack overflow" with it's own canary value
@sour lynx is there any way to provide idf.py monitor an .elf?
tulip sleet
@idle owl @slender iron I have some library code that uses time.monotonic_ns(). pylint on GitHub actions is complaining about this, because it's based on Python 3.6, which doesn't have monotonic_ns(). That I can understand, but I don't understand why https://github.com/adafruit/Adafruit_CircuitPython_BLE_MIDI doesn't have this problem also, because it also uses monotonic_ns(). Did either of you special-case that somehow?
the other libraries that uses monotonic_ns() have backup code to simulate it. On the nRF boards, that's not necessary, since it's always present
slender iron
what is the complaint you are getting?
tulip sleet
that the method doesn't exist
i don't get this when running pylint locally, because it's based on 3.8 locally, which makes sense
the latest GitHub Actions for ble_midi are fine, and don't show that error, which is the weird part
i don't see it special-cased in .pylintrc for the library
tulip sleet
yes, there may be a diff for ble_midi. tnx
@slender iron oh, adafruit_ble_midi disables "no-member" at the very top; that's why
slender iron
ah
tulip sleet
i also used ulab in adafruit services, and blinka doesn't know about it, so it complained about import order. that's another problem, not a big deal
slender iron
@onyx hinge looks like your stm rgbmatrix PR may have broken the build
idle owl
@tulip sleet Did you get sorted?
tulip sleet
@idle owl, yep, all set, though in the long run there are some things to fix, like adding dummy ulab to blinka and narrowing "no-member" in ble_midid
tnx
idle owl
@tulip sleet Ok, sounds good.
onyx hinge
@slender iron I just saw your note from an hour ago. I'll look into it. Wasn't it green when it went in!?
tulip sleet
@onyx hinge I thought so too, or maybe there was one red-herring bad build
manic glacierBOT
tulip sleet
@onyx hinge @slender iron I have been building locally from the tip with no trouble. I just pulled and updated submodules (no changes) and built one each of stm, atmel-samd, and nrf with no errors
onyx hinge
I see what's going on, I think
tulip sleet
oh wait no, not on m4
onyx hinge
@tulip sleet did you build stm32 feather or one of the stm boards that actually failed?
wait, is it m4/samd51 that's failing? Then I don't know what's going on after all
tulip sleet
i didn't know what was failing, just feather, so no protomatter. Metro M4 is failing
../../lib/protomatter/core.c: In function '_PM_row_handler':
../../lib/protomatter/core.c:442:13: error: dereferencing 'void *' pointer [-Werror]
442 | *core->addrPortToggle = newBits ^ priorBits;
| ^~~~~~~~~~~~~~~~~~~~~
../../lib/protomatter/core.c:442:35: error: invalid use of void expression
442 | *core->addrPortToggle = newBits ^ priorBits;
| ^
STM32 has a hang at startup right now, just put in a revised issue and PR
onyx hinge
https://github.com/adafruit/circuitpython/runs/633447824?check_suite_focus=true says in one spot it ran on ref https://github.com/adafruit/circuitpython/commit/e5be728c1c8a3a455b7095c9dab9e758c5016191
but if you look within the logs of build-arm capablerobot_usbhub, you see that it actually built:
e7eef6f0c Merge 6dfe9ab606352b36f8d295b250ac7118c459618b into 2d7cf4b792afa1b6755b9575b84cae4b65ae7ab9
which is not at all the claimed ref. The CI built some other ref than the one that the PR merged.
tulip sleet
and this is checkout@v2, right? oy
onyx hinge
yeah
tulip sleet
PRs were flaky for at least a few days recently. I had a PR that didn't show two new commits until I pushed a third
onyx hinge
I filed https://github.com/actions/checkout/issues/237 too, it feels weird enough that it's worth bringing to checkout maintainer's attention
tulip sleet
@onyx hinge thank you figuring that out and filing the issue
sour lynx
@sour lynx is there any way to provide
idf.py monitoran .elf?
@slender iron no, looking at the code i don't see any. But you can runtools/idf_monitor.pydirectly β see --help for options.
Edit: idf.py finds the ELF file name from build/project_description.json β this file is generated by the regular CMake build. If you can overwrite project_description.json and put a custom ELF file name in there, then idf.py monitor will pick it up.
timber mango
@gilded cradle Is there a way to tell which type of display is connected to the HT16K33 controller?
slender iron
@sour lynx ok nice! Iβll try that
manic glacierBOT
onyx hinge
@tulip sleet you're welcome. I wish I had also done more thorough local build testing but .. I didn't.
gilded cradle
@timber mango I don't believe so. It's the same chip but just hooked up to LEDs differently, so there's not really a way to automatically ID it.
timber mango
OK.
simple pulsar
@idle wharf The channels correspond to different frequencies which propagate in very different ways in non free space plus possible effects on tx/rx in each device. Knowing them is at least beneficial in grouping the different RSSI data. How much information can be gleaned from them, particularly in a non-static environment, is an interesting topic. I've been sticking updates in https://forums.adafruit.com/viewtopic.php?f=53&t=164851 .
idle wharf
Nice job @simple pulsar or as we say in my family.... "Like a Walters!" π
@slender iron I watched your deep dive from last week... and wanted to let you know they're not boring.
I don't work down at the C\Hardware level and hearing you explain how things are mapped across each layer is very interesting and helps deepen our understanding of what's going (and what could possibly be going wrong) when we're up in the CP runtime.
slender iron
thanks @idle wharf ! glad you are enjoying it
@lucid solar just pushed to my branch
crimson ferry
@prime flower This is an interesting fix in Espressif IDF, was causing blocking in certain circumstances https://github.com/espressif/esp-idf/commit/0c7204e934b8abda433026f9ec9800c5380a657d
old smelt
Does anyone know how Mu determines if it sees a CircuitPython device?
I am running a custom build for our board, and it appears to be working:
Adafruit CircuitPython 5.0.0-beta.2-13-g7387f6092-dirty on 2020-05-07; Alorium Technology Evo M51 with samd51p19
But when I try to use Mu for the Serial connection, I get this error:
Is it looking for a specific VID/PID?
raven canopy
@old smelt yep, VID/PID: https://github.com/mu-editor/mu/blob/master/mu/modes/circuitpython.py#L38
old smelt
Ah. Ok.
Thanks.
Should that file also show up in the local install directory for Mu?
raven canopy
i'm not sure. i haven't dug into how they ship binaries (or not binaries) for each OS.
old smelt
Ok. Don't see it in the app package.
I'll do some more digging. I can interact with the REPL via other means, so I'm good for now
Thanks
for your help!
raven canopy
yw!
crimson ferry
@old smelt Mu serial breaks for me when I have more than one CP device plugged in (contributing factor possibly renamed CP volumes, macOS)
manic glacierBOT
fathom trellis
Just in case anyone here is interested, I'm doing a live-streamed Q&A tomorrow at 9am PDT and talking about CircuitPython, the VS Code device simulator express plugin (currently supports cpx, microbit, and clue) and the PyBadge project I had put together for PyCon.
more details and a link at:
https://twitter.com/mbcrump/status/1258411407051116544
half geyser
What causes OSError: [Errno 5] Input/output error in the auto-reload terminal? Is there any way to get a more descriptive error?
Right now I'm seeing that happen when I save, but after about ten-twenty seconds it re-evaluates, even though I haven't saved it. Is the error due to a file that hasn't been fully flushed?
stuck elbow
more descriptive errors take up flash space
half geyser
If I do import io; io.open("code.py").read(), then exit the REPL, it works.
half geyser
code.py is just print("Start")
It reloads itself after about 20 seconds and actually works, but it immediately says OSError: [Error 5]. Makes me think there's a synchronization issue somewhere. Where would be a good place to add a breakpoint to try and trap the error as it's generated?