#circuitpython-dev
manic glacierBOT
lone axle
@idle owl (sorry if you get double pinged, the bot ate my first one) RE: discord links in the repo readme files. I was able to re-use a few bits from adabot to check all of the repos and generate a report of ones that have the invalid discord links. There are many more with the invalid one, but there are some that have already been updated with the new correct one. https://gist.github.com/FoamyGuy/f913a47788fb8eede1747ffa4dc6ea7c
The cookie cutter repo has the invalid link as well: https://github.com/adafruit/cookiecutter-adafruit-circuitpython and it is not included in that report. I think only repos that start with Adafruit_CircuitPython plus the core repo.
idle owl
@lone axle Hmm alright. For sure update cookiecutter. We'll have to decide what to do about the libs.
charred flare
What's the difference between with ESP32 Airlift breakout and the HUZZAH32 ESP32 Breakout when used as a wi-fi co-processor in CP?
anything?
manic glacierBOT
The PCA10059 has one single color LED and one RGB LED, where as this has 3 physically separate LEDS and an empty set of pads for a 4th LED. If you think it's better to have it pretend to be a single RGB LED I'm not entirely against it, I just don't think it accurately reflects the hardware to treat it that way.
vs
tidal kiln
@charred flare airlift has nina firmware installed, which is what allows it to be used as a peripheral
solar whale
@charred flare you can use an ESP32 Breakout -- you have get the right pins and load the Nina Firmware to it.
charred flare
I figured, the product page says
Comes fully assembled and tested, pre-programmed with ESP32 SPI WiFi co-processor firmware that you can use in CircuitPython to use this into a WiFi co-processsor over SPI + 2 pins.
Which made me think it was the same nina firmware
slender iron
@onyx hinge sorry, got up early for a zephyr meeting and now just getting going. there is port_enable_ticks or similar that calls supervisor_tick. that could be shared for tasks that need periodic background triggers. we use the rtc for that interrupt and may use a second one to wake from sleep
onyx hinge
OK, makes sense.
we'll probably have to activate ticks when we have a display with auto-refresh, and when we might need to do filesystem background tasks .. I have the former implemented but only compile-tested. haven't gotten into the latter yet.
I remember now you said something about being up at 6AM for a meeting .. my condolences
slender iron
@onyx hinge Ya, activating ticks in that case is fine. Yup it's "face-to-face" meetings that are interesting, just early.
next wednesday too
idle owl
@lone axle I've asked @trim elm to work with you on getting the libs updated. No rush on it, he's finishing up another project first, but wanted to let you know.
lone axle
Is there something different about the cookie-cutter repo and how it uses Git compared to other library repos? I forked and cloned the cookie cutter repo. But my IDE thinks that none of the files within it have been added to the repo, which is not something I can recall ever seeing.
oh, looking back at the terminal I see now that I overlooked this error:
Receiving objects: 100% (702/702), 151.60 KiB | 2.44 MiB/s, done.
Resolving deltas: 100% (405/405), done.
error: unable to create file {{ cookiecutter.library_name }}/examples/{{ cookiec
utter.library_name | lower }}_simpletest.py: Invalid argument
error: unable to create file {{ cookiecutter.library_name }}/{% if cookiecutter.
library_prefix %}{{ cookiecutter.library_prefix | lower }}_{% endif %}{{ cookiec
utter.library_name | lower }}.py: Invalid argument
fatal: unable to checkout working tree
warning: Clone succeeded, but checkout failed.
You can inspect what was checked out with 'git status'
and retry the checkout with 'git checkout -f HEAD'
That is likely my issue.
slender iron
I'm not seen that. what OS are you on? maybe the file paths are too weird
lone axle
Windows
and yep, that was going to be my guess that it may not like the brackets used in filenames like that
or percents is probably the issue actually, not brackets
windows tries to resolve percents as shortcuts or something I think like %APPDATA%
I can try with linux in a bit.
onyx hinge
@gilded cradle hey, I'm adding a new board to circuitpython.org -- I created my new _board/boardname.md file. Building locally with jekyll, the boardname.html file gets generated, but I can't search it from "downloads". Did I miss a step? I can PR what I have in case you need to see it to answer.
hmm I kept fiddling with things and eventually it appeared ‽
prime flower
@onyx hinge thats pretty much my web dev experience
gilded cradle
@onyx hinge good to hear.
onyx hinge
thanks dan! Thanks dan!
slender iron
@prime flower https://wiki.python.org/moin/UdpCommunication
trim elm
@idle owl Can I push the discord link changes to master as long as I've verified nothing aside from the discord link was changed?
idle owl
@trim elm Hmm. Maybe PR a few first?
trim elm
@idle owl Yeah, that's probably smart
idle owl
@trim elm I mean it's probably fine, and obviously with version control we can roll back mistakes, but it's worth verifying first. PR a few and then we'll go from there.
trim elm
Yeah
idle owl
@trim elm I approved the last two, merge them once checks pass. And push the rest. But please verify each library once you've pushed, check the commit to make sure that's all there is to it. Basically so we're not finding out later that we borked something. PR to NeoPixel if it's needed, don't push directly.
Pypixelbuf as well. PR to that if it's needed.
trim elm
@idle owl Sure thing. I've already outputted the diffs to a file and verified that the changes were exactly the same, so I'm confident as far as that one goes, but I'll make sure they are passing the CI
Ok, both of those do need it so I'll do that first
idle owl
Thank you.
idle owl
@trim elm Ok approved pypixelbuf. Please merge it once it's passing.
trim elm
Pushing everything else now
@lone axle In about an hour (it's probably going to take a while for all the Actions to run) can you verify that half of the libraries I've pushed to are passing CI? I'll do the other half
lone axle
@trim elm yep can do.
trim elm
@lone axle Also, I'm starting to think we might want to make a directory in the adabot library that houses scripts we've written to do specific things using Adabot. It might be worth making a PR with the script you wrote and a new directory to house stuff like that
Ok, they're all pushed. I'll go through in a bit and verify they're passing
@lone axle If you want to do the first half (JWT to CharLCD on your list) I can do the second half
lone axle
Okay yep I will get the first half once I am wrapped up with what I'm working on now. And I'll make a PR for adabot to add that later on tonight.
trim elm
Awesome! Thanks
bronze shadow
Does anyone know if something about import depth was changed in cpy 5.3? I updated to that from an older version though i don't remember which, and had my project breakwith RuntimeError: maximum recursion depth exceeded. For fun I loaded the 6.0.0 alpha and it was fine yet again.
Project in question if it concerns anyone is this https://github.com/KMKfw/kmk_firmware as it's quite the large bit of code for these tiny microcontrollers.
trim elm
@idle owl I found 10 of my ~125 that are failing on either pylint or build docs, that were not previously. No clue why that's happening, and the pylint failures should be easy enough to fix, but I'm not as sure about the sphinx failures. Regardless, I'll try fixing them all tomorrow morning (unless you'd like me to do them tonight, in which case I can).
idle owl
@trim elm Nope, I do not want you to do them tonight. 🙂 You should be done for the day.
tidal kiln
@idle owl is the pyportal library suppose to be uploaded to pypi?
idle owl
No.
idle owl
Hmph
tidal kiln
idle owl
I imagine it did not work.
tidal kiln
nope
idle owl
Ok, I'll pull it and re-release it.
idle owl
Yeah that's bad.
Will do it tomorrow.
Need to give it a setup.py.disabled, release the CP lib, and delete it from PyPI.
prime flower
@tidal kiln @gilded cradle has been doing some work on the library, unsure if this is part of it
tidal kiln
cool. thanks. yah, no canadian rock band. tomorrows fine.
@prime flower semi-related, was trying to test something for that library
but pip installed the wrong one (my mistake)
gilded cradle
Yeah, the circuitpython version should not be on PyPI, but the Blinka version should. Perhaps in the release notes for the CircuitPython version, we can tell users to install the Blinka version instead.
idle owl
@gilded cradle Not sure I want to do that. It's not a bad idea, but I think we should stick to referencing only the lib in its release notes.
gilded cradle
Ok, then let's just not mention installing via PyPI (not sure if that's how we're already doing it).
idle owl
Yeah I'll get it fixed up.
tidal kiln
the CP version just shouldn't be there
idle owl
Right.
gilded cradle
Can we remove it?
tidal kiln
thanks @idle owl
manic glacierBOT
turbid radish
?SERVERINFO
digital shoreBOT
adafruit
manic glacierBOT
prime flower
@crimson ferry https://github.com/arduino/nina-fw/pull/50#event-3528897382 may be worth pulling into our fork?
crimson ferry
@prime flower Seems to correlate with one of the CP fork issues iirc? What's the process... pull/merge from Arduino, or manually make the (simple) change in the CP repo?
prime flower
@crimson ferry which issue?
there isnt a process for this (yet) as our fork has yet to pull anything from upstream.
crimson ferry
@prime flower I thought there was an issue on large responses, but looks like more of a mystery issue https://github.com/adafruit/Adafruit_CircuitPython_ESP32SPI/issues/9
prime flower
I'm curious if large responses are getting dropped due to the linked PR
crimson ferry
I often get incomplete responses with 200 HTTP, heven't gotten to the bottom of it
@charred flare It's a simple matter to load the NINA firmware onto a standalone ESP32 with USB https://learn.adafruit.com/adding-a-wifi-co-processor-to-circuitpython-esp8266-esp32/program-with-esptool. (fwiw, Airlift add-on boards can also run Arduino 😉 ) ...unfortunately, there are three different procedures for loading/updating NINA on standalone ESP32s (e.g., ESP32 Feather), Airlift add-ons (e.g., Airlift featherWing), and boards with integrated ESP32 (e.g., PyPortal). Standalone definitely the easiest.
charred flare
@crimson ferry Yeah I tried that on the last one, to update the firmware and I couldn't get it to load for whatever reason. I'm going to try again on this new one when it arrives. I'd like to stay away from Arduino both for the language and the IDE has always been super buggy for me since I started with these boards here in the last few months. Just tons of issues uploading sketches, duplicate ports showing up, etc.
crimson ferry
@charred flare Yeah, I get spoiled with the ease of CircuitPython. If you still have the old one, I'd be happy to work through it with you to try to get it working. If so, let's take it over to #help-with-circuitpython
charred flare
Thanks. It's all soldered up and in place though. It works well enough for a dev board. Just needs its own wifi from an old router to connect reliably - for some reason. lol (doesn't like my main Netgear b/g/n router but is mostly fine with the old Netgear running dd-wrt)
manic glacierBOT
turbid radish
?serverinfo
digital shoreBOT
adafruit
onyx hinge
@tulip sleet any idea what the sam d21 equivalent of this is, before I go snorkeling in datasheets? I need to acknowledge the DMA "transfer complete" interrupt for a channel. audio_dma.c:288:9: error: 'Dmac' {aka 'struct <anonymous>'} has no member named 'Channel' 288 | DMAC->Channel[dma_channel].CHINTFLAG.reg = DMAC_CHINTFLAG_TCMPL; | ^~ ../../py/mkrules.mk:55: recipe for target 'build-feather_m0_express/audio_dma.o' failed
tulip sleet
@onyx hinge I just searched for TCMPL in the DMAC registers:
onyx hinge
I guess it'll be multiple operations, similar to this? ```void dma_resume_channel(uint8_t channel_number) {
common_hal_mcu_disable_interrupts();
/** Select the DMA channel and clear software trigger */
DMAC->CHID.reg = DMAC_CHID_ID(channel_number);
DMAC->CHCTRLB.bit.CMD = DMAC_CHCTRLB_CMD_RESUME_Val;
DMAC->CHINTFLAG.reg = DMAC_CHINTFLAG_SUSP;
common_hal_mcu_enable_interrupts();
}
so you set the ID, and then write a 1 to TCMPL, it looks like
.bit.ID = channel_id, then .bit.TCMPL = 1
i would think
onyx hinge
er I guess this is enabling the interrupt, not acknowledging it. Still, thanks, I think this points me on in the right direction
tulip sleet
Writing a 1 to this bit will clear the Channel ID (ID) Transfer Complete interrupt flag
onyx hinge
yeah I've now confusingly referenced several different registers
but I think I got the answer I need, thanks
tulip sleet
is the code above for SAMD51? I think it has a different method of specifying the channel. When the bits are written, indirect access to the corresponding Channel Interrupt Flag register is enabled. (SAMD21, in that screenshot)
onyx hinge
the pasted dma_resume_channel was for D21
the error line was code that worked on d5x/e5x but was a compile error on d21
ah, I think that line is not needed anyway, it's a relic of previous confusion over whether to use dma interrupts or event system interrupts, which was itself an entire digression because the real problem was a bug that masked interrupts after a length of time. ⏪
tulip sleet
@onyx hinge Occasionally I got back and read the ASF3 drivers, which are considerably easier to read. If you checkout circuitpython tag 1.0.0 and go to circuitpython/atmel-samd/asf, you can find the ASF3 drviers (they're probably available from Microchip as a zip file too). Interesting line there:
sam0/drivers/dma/dma.c
225: } else if (isr & DMAC_CHINTENCLR_TCMPL) {
227: DMAC->CHINTFLAG.reg = DMAC_CHINTENCLR_TCMPL;
ionic elk
What's the best way to submit a change to a library like Protomatter? Should I just fork and clone it separately, or is there some slick method of submitting a PR out of a submodule?
tulip sleet
@ionic elk you can add your own fork as a remote to the submodule clone, then you make changes there, and submit a PR from there.
slender iron
@bronze shadow what microcontroller were you running on?
ionic elk
@onyx hinge when you have a sec can I ask you some protomatter questions?
onyx hinge
@ionic elk sure, give me 15 minutes?
ionic elk
Sounds good
bronze shadow
@slender iron It's the Itsy Bitsy M4 Express
slender iron
hrm, I wouldn't expect the stack size to change. I know the STM stack size did
there is a performance regression currently but Jeff is working on it
onyx hinge
@ionic elk I'm at your disposal
ionic elk
Coolbeans - I'm trying to figure out how _PM_timerStart get the actual STM32 timer to do the register stuff they do?
onyx hinge
OK, let me look around at the code and refresh my memory
ionic elk
I don't see any place where TIM6 is actually passed to the Protomatter object
So I don't get how they're doing things like turning SR on and off
onyx hinge
I think it's hardcoded
ionic elk
but where?
ionic elk
You've got stuff like tim->SR, where does the tim get assigned?
onyx hinge
here's where it's hardcoded in the core (ports/stm/common-hal/rgbmatrix/RGBMatrix.c): void *common_hal_rgbmatrix_timer_allocate() { // TODO(jepler) properly handle resource allocation including never-reset return TIM6; } and in protomatter (lib/protomatter/arch.h): ```
// Use hard-coded TIM6 (TIM7 is used by PulseOut, other TIM by PWMOut)
#define _PM_timerFreq 42000000
...
#define _PM_IRQ_HANDLER TIM6_DAC_IRQHandler
wait hang on yep just found that at the same time
onyx hinge
and t6_handle.Instance = TIM6;
but basically look for the number "6" in that source file arch.h I guess
I'm not proud, it was corner cutting
ionic elk
Ok. So, I need to replace all of this with a single call to the stm_peripherals_find_timer - but I guess that's how SAMD must do it so maybe I should just check out how that does it
thanks for the pointer
onyx hinge
in principle you get the "correct" timer pointer to void _PM_timerInit(void *tptr) in arch.h
but whether that happens yet or not I don't know
so anyway the right way is to pass the timer in to _PM_init, it gets to core->timer, which is then passed to _PM_timerInit cast to void*
manic glacierBOT
onyx hinge
normally common_hal_rgbmatrix_rgbmatrix_construct gets called with timer parameter as NULL, so it calls common_hal_rgbmatrix_timer_allocate and passes that to _PM_init where it gets stored to core->timer, and later core->timer is passed to _PM_init.
once you have all that happening, you can start to remove the hard-coding from arch.h in protomatter
manic glacierBOT
Board is a Metro M4 with UF2 2.0.0
Sensor is a prototype AS7341 but it will occur with nearly every sensor I use
I will update my UF2 version and try to get some info on the voltage levels
silver tapir
@gilded cradle It was so cool to see the pyportal electioncal on a biggger Raspberry screen. Looks great.
gilded cradle
Thanks @silver tapir. 🙂
idle wharf
That Fluff a great idea @stuck elbow !! I hope an ESP32-S2 one exists some day as well.
digital shoreBOT
Dylan H#2310 was unmuted
slender iron
@trim elm don't ping a bunch of people at once
manic glacierBOT
tulip sleet
@dherrada if those libs already have releases, no reason I know of not to release
you mean all the BLE ones?
idle owl
My bad, I asked Dylan to ping you four. Keep forgetting about the lockdown.
gilded cradle
Pretty much what @tulip sleet said. Though I can't think of any that shouldn't be released.
trim elm
Yeah, everything. I can link to the list if you'd like me to, but just assume that I'm releasing all of the libraries
tulip sleet
are there any that have never had a release?
trim elm
Not sure, but my script throws a TypeError when that happens so I look at those manually
manic glacierBOT
idle owl
@tidal kiln @gilded cradle Hah.... I went into my PyPortal lib directory and it was on a remove-from-pypi branch.... I had already done most of the work, but I neglected to delete it from PyPI. It's gone now.
slender iron
@prime flower can chat whenever you are ready
prime flower
@slender iron great, lemmie grab a water real quick
slender iron
kk
gilded cradle
Awesome @idle owl. It's possible that you deleted it and due to the cookie cutter naming everything Adafruit-CircuitPython-xxxx by default, I may have accidentally released to it when I was first creating the lib.
idle owl
@gilded cradle Who knows. I found the PR where I made it setup.py.disabled on May 4 I think. The last time PyPortal was updated on PyPI was April. Either way, it's gone now.
prime flower
@slender iron back! Amelia ok?
slender iron
yup
stuck elbow
@idle wharf esp fluff is hard to do, because antennas and RF stuff
idle wharf
@stuck elbow True. I guess the first decision would be on-board antennas or connectors.
stuck elbow
I will definitely want to do a PewPew with esp32-s2 at some point
but I will probably just use a module for it
turbid radish
?serverinfo
digital shoreBOT
adafruit
manic glacierBOT
idle owl
@gilded cradle Do you have a few minutes to talk about PWM on Pi? I'm trying to understand what I'm supposed to be doing to this servo guide.
gilded cradle
Sure
idle owl
Ok, here's the guide https://learn.adafruit.com/using-servos-with-circuitpython/overview.
Does it "just work" as is? Or does the code need to be updated to do something else to utilise Blinka's PWM
gilded cradle
It should just work as is.
idle owl
Fair enough, ok, so do you figure I simply need to include a wiring diagram using a Pi to indicate it works?
gilded cradle
You'll probably just need to make sure you have the latest version of Blinka.
gilded cradle
Yeah, it would probably be a good idea to have a wiring diagram.
I would test it of course. 😉
idle owl
Yah fair enough. I wasn't sure whether Limor wanted me to do that or you. I'll find out.
gilded cradle
Ok
idle owl
I was supposed to ping you about something anyway, not sure what specifically.
I have a Pi here, needs to be updated is all.
gilded cradle
Ok. Maybe she meant to ping me if it doesn't work or you get stuck?
Or you just have questions...
gilded cradle
That works 🙂
jovial wasp
New to circuitpython and discord
Looking to use pedometer methods in adafruit_lsm6dsL LSM6DS33
Tag: 3.0.0 has support
Tag 4.0.0 does not have support
lone axle
@jovial wasp lets move over to #help-with-circuitpython. This channel is more for development of the project.
manic glacierBOT
manic glacierBOT
manic glacierBOT
@dhalbert Indeed there is a pretty substantial dip, well below 2.7V:
Here's the capture from the screenshot above:
https://www.dropbox.com/s/gk8usbzqbtujw2x/sensor_insert_vcc_sag.sal?dl=0
Note that this is from the Logic 2 alpha so you'll have to have that installed to view it
I'll close this since it does seem to be unrelated to CircuitPython
onyx hinge
now I want this M60 keyboard 🙂
manic glacierBOT
onyx hinge
@gilded cradle On Monday during the meeting can you paste the newsletter links in? I'll be mostly running the meeting, and Kattni will be doing the recording so she doesn't want to be entering stuff in her discord client.
onyx hinge
huh, apparently common_hal_mcu_disable_interrupts is pretty expensive on nRF
tulip sleet
@onyx hinge Do you have a minute to chat about background tasks? I have to add a background task to the general-purpose HCI _bleio code I'm writing, and would have to add it to every run_background_tasks() in all the ports. There is a lot of duplicate code in the multiple impls of that routine.
In your recent work on background tasks, are you refactoring any of this into some shared routine?
onyx hinge
sort of
tulip sleet
huh, apparently common_hal_mcu_disable_interrupts is pretty expensive on nRF
@onyx hinge Yes, it has to keep the BLE stuff running, so it can't really just disable all inteerrupts.
onyx hinge
I am focused primarily on making certain things happen due to interrupts, and those things are then pulled out of the per-port run_background_tasks
I don't think that'll get me to them being 100% identical but a good thing to do while I'm there would be to pull out the common parts into a supervisor function
tulip sleet
right, I expect there to be some per-port stuff, but they all have to check for usb, display, filesystem, etc. background tasks. I will just continue not assuming that but will be happy to adapt when I have to merge in and you have probably already merge in
onyx hinge
what time frame are you operating in? nRF went fast (I think that after my addition to that PR just now, it is faster than 5.3 and probably works right 🙂 but I didn't test the audop peripherals)
tulip sleet
oh, at least a week before any PR, and that would be incomplete functionality. There is a lot to do
onyx hinge
but that leaves stm, esp32, cxd, imxrt, etc
and litex, which I don't have hardware to test on. bummer.
onyx hinge
OK, I think you should get your wish but I'm just not ready to guarantee it'll come in time
The bootloader allows you to load CircuitPython, Makecode, and Arduino programs. The bootloader is not CircuitPython. You can check the current version of your bootloader by looking in the INFO_UF2.TXT file when the BOOT drive is visible (FEATHERBOOT, CPLAYBOOT, etc.).
hm this boilerplate about what the UF2 bootloader enables is probably the same on all the boards, but .. I'm pretty sure https://circuitpython.org/board/same54_xplained/ doesn't support Makecode and idk about Arduino
tulip sleet
i wrote that boilerplate, I can caveat it.
onyx hinge
the device also doesn't ship with uf2 bootloader so the updater won't do many people any good ; but I think that since circuitpython is placed "above" the bootloader, it is probably needed before you load circuitpython on it. I did everything with the debug usb port of course
you need a different linker script to work on a bare board, right?
tulip sleet
you can turn off the bootloader updater box in circuitpython.org for that board if you want
only because the origin is different
i would not recommend changing the linker script, I would say CPy is not supported unless you install a bootloader
onyx hinge
How's this sound?
CircuitPython requires that the board be flashed with the UF2 bootloader. This can be done using the second USB connector, which provides a debug interface. After this has been done once, CircuitPython can be updated in the usual way using the UF2 bootloader, or by using the debug USB connector.
tulip sleet
would you add that to board description? The UF2 update boilerplate is shared across all the atmel-samd boards
onyx hinge
Yea I'm imagining that is going in the board md file above the bullet list of features
It also seems like the board doesn't actually run unless you have the debug interface going which is weird .. but maybe that's how they wanted it
tulip sleet
i think that's fine. We are talking about order-of-magnitude 10 people I think who will try this
onyx hinge
yeah, agreed
gilded cradle
Sure @onyx hinge Do you just have them lined up beforehand?
onyx hinge
@gilded cradle yeah the links will be in the notes doc, copying them from the newsletter draft is something I will have done shortly before the meeting
what I've done when I'm in the "pasting links" role is try to paste them when the speaker is starting to talk about that item. it's not critical.
If there are multiple links, I often pick just one, according to my feelings. If there's a non-twitter link, I use that one 🙂
gilded cradle
Ah, ok. Thanks
onyx hinge
none of it is critical, but I always liked the links appearing in the channel when PT did it, so I spontaneously started doing it when he had to step back
onyx hinge
@ionic elk okay, I was catching up on review requests and noticed it
ionic elk
👍
slender iron
@hybrid urchin hey! I got my orangecrab today!
hybrid urchin
\o/
slender iron
guess what I want to do? 🙂
hybrid urchin
I'm pretty sure I know what you want to see running on it 😛
slender iron
😄 I assume I want to start with the litex port we have
hybrid urchin
I'll write up a quick guide and make sure everything is still working. It's been a little while since I've had it running.
slender iron
ooooh, even better. I didn't know you've had it going
do you have it going with the ram and flash?
hybrid urchin
Yep! But that was basically all. Didn't have I/O or extra peripherials.
slender iron
kk, that's a start
digitalio and busio are the next things
I'm really interested in SoC design too
how fast can you clock the cpu?
you are using litex for the SoC right?
hybrid urchin
With the VexRiscv max speed is around 80MHz. I've gennerally had it set at 72MHz. There are lots of knobs to turn if you wanted to tune for performance.
Yep.
hybrid urchin
With the current configuration litex supports, the DDR3 memory runs at 2x CPU speed.
slender iron
haha nice!
hybrid urchin
So I've been spliting the 128MB DDR3 into two areas for firmware/data. And there is a small routine that runs on startup that inits the DDR3, and copies the FLASH into the DDR3 memory space.
slender iron
that makes sense! it's still way more ram than we have on any other CP board
slender iron
@hybrid urchin is your repo public in case I want to poke at it later?
hybrid urchin
Yep, I'm re-working the litex SoC side. It's bitrotted a bit. But that will be here: https://github.com/gregdavill/OrangeCrab-examples
And the small number of changes I've made on Circuit python will need to be rebased. But that's on my fork: https://github.com/gregdavill/circuitpython/tree/orangecrab
slender iron
ok awesome! Let me know when the SoC is available. I2C support will add support for most featherwings
it would be nice to have an iomatrix too. I know it's antithetical to fgpa design
hybrid urchin
Cool, iomatrix? is that for interfacing to pin-muxing?
slender iron
ya, to change peripheral to output pin mapping from circuitpython
k, I gotta make dinner. thanks! very exciting
hybrid urchin
Okay, I've pushed my test SoC. and some small changes to get it building. But it's from april and ~1000 commits behind the main branch.
Adafruit CircuitPython 5.2.0-24-gc88db8a4e-dirty on 2020-07-11; OrangeCrab with VexRiscv
>>>
manic glacierBOT
manic glacierBOT
kind river
With Circuit Python on a larger FPGA like the ECP5, you could probably make configurable IO "cells" which each contain one of each peripheral. This would allow you to uses as many of any peripheral as you have pins for.
hybrid urchin
That makes sense! You might still want/need some form of muxing with modules that would span over multiple io cells though? for example SPI/I2C
kind river
That's how the Parallax Propeller 2 smart pins work. The XMOS chips also use a similar method where pins/ports can can be in/out/both and can be clocked by another pin or an internal clock that can optionally be exposed on a pin.
Both of these methods use generic logic that can be configured to act as various peripherals though, so they require more software support.
slender iron
@hybrid urchin I wasn't able to build the soc
4.39. Executing DFFLEGALIZE pass (convert FFs to types supported by the target).
ERROR: Conflicting init values for signal 1'0 (\soc_basesoc_interface_adr [5] = 1'x != 1'0).
Ohh, yeah. There is bug that a recent Yosys change has un-covered. Perfect timing right 😛
slender iron
ok, perfect
hybrid urchin
I think the easiest solution if you want to be able to build the SoC yourself, is to roll back to a earlier Yosys build. Otherwise I've put some prebuilt files up in a new branch: https://github.com/gregdavill/OrangeCrab-examples/tree/prebuilt/litex
slender iron
ok, let me go snag an older toolchain zip
hybrid urchin
slender iron
kk, will try now
hybrid urchin
I'll try it here too, just to make sure I'm not sending you down a rabbit hole.
ivory yew
idk if anyone ever wondered about it, but CircuitPython on the SAMD21 should take about 12us (2 * 5.67) to do an ADC conversion, and it won't be accurate if the input impedance is higher than 7kOhm.
As far as I can tell, Arduino's config is almost identical but they bump up samplelen to 2 which increases their total conversion time to about 14us (6.67 * 2) but also increases their max input impedance to ~60kOhm.
might be worth considering for us, as 7kOhm is pretty low especially if someone uses a 100k pot as a voltage divider to feed an analog input.
tulip sleet
@ivory yew I found this useful, probably you've already seen it: http://ww1.microchip.com/downloads/en/DeviceDoc/Atmel-42645-ADC-Configurations-with-Examples_ApplicationNote_AT11481.pdf
Is our clock rate 1.5MHz as you set above in the calculator? I think we are using GLCK0 (48 Mhz) as the clock (ports/atmel-samd/peripherals/samd/samd21/adc.c): I don't see that we are setting the clock divider at all in common-hal/analogio/AnalogIn.c, so it looks like it might the default CTRLB.PRESCALER set to 0x0, which is divide by 4.
manic glacierBOT
ivory yew
@tulip sleet iirc the prescaler is set in asf_conf
tulip sleet
aha, forgot that
tulip sleet
// <o> Prescaler configuration
// <0x0=>Peripheral clock divided by 4
// <0x1=>Peripheral clock divided by 8
// <0x2=>Peripheral clock divided by 16
// <0x3=>Peripheral clock divided by 32
// <0x4=>Peripheral clock divided by 64
// <0x5=>Peripheral clock divided by 128
// <0x6=>Peripheral clock divided by 256
// <0x7=>Peripheral clock divided by 512
// <i> These bits define the ADC clock relative to the peripheral clock (PRESCALER)
// <id> adc_prescaler
#ifndef CONF_ADC_0_PRESCALER
#define CONF_ADC_0_PRESCALER 0x3
#endif
yes
when we were first doing all this we tried to be approximately like Arduino, so as not to surprise people migrating over. But I would not be surprised that we lost track of that when I was converting from ASF3 to ASF4.
ivory yew
I think Arduino changed things recently as well, but I'm not sure.
turbid radish
?serverinfo
digital shoreBOT
adafruit
manic glacierBOT
manic glacierBOT
manic glacierBOT
onyx hinge
What's the minimum I need to do to see an nrf device in the bluefruit connect device? Just having it powered is not enough?
ok got it
@tulip sleet okay, whee, it looks like I have successfully moved the BLE background stuff into supervisor code while working on this stuff
...
+ #if CIRCUITPY_BLEIO
+ supervisor_bluetooth_background();
+ bonding_background();
+ #endif
...
}```does that seem like it will fit your needs?
like you noticed the background functions are mostly similar, they call functions guarded by #if enables and bleio maybe fits into that mold
tulip sleet
I ended up doing this myself too, but I need to add a bleio-hci-specific background task as well, so I ended up create a bleio_background(). WIP Commit is here: https://github.com/dhalbert/circuitpython/commit/f6869c69c5fda701b57fbfca0eaf1aa03a2d164b. I don't need you fix whatever you have now: I'll adapt it as necessary when I merge from upstream before submitting a PR.
manic glacierBOT
onyx hinge
That sounds good, if I get there second I'll deal with the merge conflicts
ionic elk
@onyx hinge My question about the protomatter merge conflict is more that the merge conflict is actively wrong? It claims there's duplicated code when there is no such thing. It's suggesting I have two copies of #define _PM_pinOutput(pin_) do { when you can look at the code and see that's not the case.
onyx hinge
In my experience, it doesn't matter whether I think git is wrong, I still have to deal with reality as git presents it to me. Sounds weird though.
ionic elk
I don't mean to just complain about it I was simply confused about how to solve it
onyx hinge
I intended my advice to be constructive, so I hope you'll accept it in that vein even if it was useless
if it's that weird I'd probably just manually make my edits on a fresh branch at the tip of master
ionic elk
No no it's all good I didn't mean it like that. I was just wondering if you knew about why it was doing the weird conflict message. Your suggestion is fine and might solve it anyway - which commit should I try to target, though, if master isn't working?
onyx hinge
I'd just say fix it while you're in there
onyx hinge
I am PRETTY sure it's utterly trivial to fix
but y'know .. when I'm wrong, don't hesitate to raise the issue again
ionic elk
So ST, in their infinite wisdom, decided to name an enum value for their TIM6 IRQ two different things across chip families - TIM6_DAC_IRQn vs TIM6_IRQn. Is there any way for me to actually check whether an enum value like this is defined, like I can a macro?
Right now I get undeclared here (not in a function) errors but I don't know how to fix this for an enum?
lapis hemlock
@onyx hinge @tulip sleet do you have hardware that would be affected by this issue: https://github.com/v923z/micropython-ulab/issues/132 ? I think the solution is in upstream micropython. I will try to produce a fix today, and I just wanted to pass the information along, in case circuitpython has to be patched up, too.
onyx hinge
@lapis hemlock no, we don't use double-precision on any of our hardware
manic glacierBOT
lapis hemlock
@onyx hinge OK, thanks for the feedback!
onyx hinge
<@&356864093652516868> The weekly meeting is coming up in about 10 minutes! Hope to see you there. Please, if you can add your hug reports and status updates to the doc, it's super helpful. And if you'll be lurking (just listening in), it's also super helpful to note that too
@slender iron just a reminder -- I would like you to do the overall section when we get into the state of circuitpython (as well as the core)
@gilded cradle will you be able to paste relevant links out of the notes doc as we go along?
modern wing
Good afternoon all you wonderful folks -- happily lurking today, and I've updated the doc to reflect as such 🙂
turbid radish
?serverinfo
digital shoreBOT
adafruit
gilded cradle
@onyx hinge yes. All prepared.
onyx hinge
@gilded cradle appreciated very much
If anyone would like to do mic checks before we get started, now is a great time. Five minutes from now is too late
lone axle
🚫 🐿️
turbid radish
lurking
silver tapir
lurking
turbid radish
riiiiight
modern wing
I wish I had a GoXLR setup to pitch-shift my voice to sound like a chipmunk....because I'd never do that. That'd be mean 😛 🐿️
uncut nexus
Lurking today
stuck elbow
alurkin'
thorny jay
Lurking.
errant grail
Text only
gilded cradle
slender iron
@quartz patio @pine rampart are you just listening to the meeting? (aka lurking)
modern wing
Solar-powered blogs don't cast any shade 🙂
slender iron
💯
silver tapir
Happy to translate.
errant grail
Hug to @onyx hinge specifically for the ulab learning guide. It's an excellent guide for the newbee (me).
silver tapir
Yey, two @idle owls are better than one.
modern wing
ImitationKattni has been asked kindly to follow the Code of Conduct. 😉
turbid radish
The Kattni from the Alternate Universe ((c) Star Trek) is a bit rowdy
slender iron
influxdb and grafana are great for storing sensor data too
thorny jay
@lone axle Can I have Pybadger library support for CLUE too and Mini PiTFT 240x240? And then I don't have to do anything anymore.
lapis hemlock
I am only lurking today.
slender iron
welcome!
lone axle
@thorny jay Pybadger does support CLUE already. I think Kattni added it back around the time the CLUE came out to prepare for PyCon. https://github.com/adafruit/Adafruit_CircuitPython_PyBadger/blob/master/adafruit_pybadger/clue.py The CPB Port was largely based on the CLUE one.
thorny jay
ionic elk
@slender iron would probably make sense to do pulseIO since I've been working on it recently for i.MX/STM recently anyway
slender iron
sure! (mute your mic too)
ionic elk
whoops srry
slender iron
np
ionic elk
anyway, still need to get a control up and going for the ESP so I'll try and do that this week sometime
slender iron
kk, great!
here is one place featherwings are documented: https://learn.adafruit.com/adafruit-feather/community-feathers-wings
slender iron
yup 🙂
turbid radish
It is a lot of data
The 'Wing product pages have some info on compatible Feathers where things are a bit tricky
And third party boards can be very challenging
gilded cradle
Me too
turbid radish
Me too, no audio
thorny jay
Oups...
slender iron
@thorny jay can you hear us?
thorny jay
No
thorny jay
Ok.
gilded cradle
You may need to restart discord
thorny jay
close the meeting
slender iron
we're trying to note it better on the downloads page and PRs welcome where we miss it
turbid radish
Thanks all
modern wing
Thanks!
gilded cradle
Thanks
silver tapir
Thanks for hosting @onyx hinge, was excellent.
thorny jay
By
modern wing
Indeed, excellent job @onyx hinge!
onyx hinge
@thorny jay Thanks for noting the limitation of the Espruino, pull requesting a change would have been perfectly appropriate
gilded cradle
You did great
silver tapir
We need to neotrellis that.
thorny jay
And I really like the sound of your voice and how you have a sentence for everybody.
And also give a warning to the next next speaker.
That is really great @onyx hinge
Here is the compatiblity check: https://pinout.xyz/phatstack
It is cross reference Youtube link to Github and Github link to Youtube.
How do you go from GoogleDoc to MarkDown?
Ok
Chronological order.
Monotonic(?)
Yes but not reverse way!!!
Bye!
onyx hinge
@idle owl I don't have github permissions on https://github.com/adafruit/adafruit-circuitpython-weekly-meeting
I could still steam ahead, it would just make a pull request
You’re making changes in a project you don’t have write access to. Submitting a change will write it to a new branch in your fork jepler/adafruit-circuitpython-weekly-meeting, so you can send a pull request.
I added this as a prereq in the doc
slender iron
@onyx hinge added you to the circuitpython team and gave that team notes permission
onyx hinge
👍
idle owl
@onyx hinge Thanks for adding it as a prereq to the doc.
onyx hinge
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/19NJ_m-fbOBJ0c8wJegVMbbPcAgov1M5oxxr7oxwlxbo/edit?usp=sharing
orchid basinBOT
slender iron
Anyone object to removing displayio.Shape in 6.0.0? I don't think anyone uses it.
vectorio is probably better anyway
manic glacierBOT
slender iron
Ok, I did a decent issue sweep. Didn't go through all of Long Term. I bet there are more issues we can close
onyx hinge
@slender iron thank you for doing that! I don't know if it felt like a slog to you but from over here it is a great reminder that we've been getting things done. Go team! go community!
manic glacierBOT
These examples=are still unresolved, I believe, and a bunch more above in displayio and framebuffer. You have to open the "hidden" conversations to see them.
(screenshots below: don't try to click :) )
manic glacierBOT
orchid basinBOT
manic glacierBOT
@tannewt Yep, recreated. I went back to just the basic setup per the guide:
https://learn.adafruit.com/adafruit-max98357-i2s-class-d-mono-amp/circuitpython-wiring-test
and still see the issue. So may be something more fundamental and not related to MSA301.
Tested as follows. Libraries from 20200714 bundle release.
Adafruit CircuitPython 5.3.1 on 2020-0...
onyx hinge
this is a surprising hardfault.. ```#0 HardFault_Handler () at supervisor/port.c:327
#1 <signal handler called>
#2 0x000299c6 in sd_softdevice_is_enabled (
p_softdevice_enabled=0x200155a5 <sd_is_enabled.lto_priv> "")
at bluetooth/s140_nrf52_6.1.0/s140_nrf52_6.1.0_API/include/nrf_sdm.h:346
#3 0x0002aebe in common_hal_mcu_disable_interrupts ()
at common-hal/microcontroller/init.c:55
#4 0x00031b44 in background_callback_add_core (
cb=cb@entry=0x20014abc <callback.lto_priv>)
at ../../supervisor/shared/background_callback.c:38
#5 0x0003a532 in background_callback_add (data=0x0, fun=<optimized out>,
cb=0x20014abc <callback.lto_priv>)
at ../../supervisor/shared/background_callback.c:59
#6 usb_irq_handler () at ../../supervisor/shared/usb/usb.c:93
#7 USBD_IRQHandler () at supervisor/usb.c:93
#8 <signal handler called>
#9 memset (n=8, c=0, s=0x2003fa0c) at ../../lib/libc/string0.c:93
#10 memset (s=0x2003fa0c, c=<optimized out>, n=8)
at /opt/gcc-arm-none-eabi-9-2019-q4-major/arm-none-eabi/include/string.h:33
0x000299c4 <+0>: svc 18
=> 0x000299c6 <+2>: bx lr
End of assembler dump.
``` stack looks fine, the address the result is to be written to is fine (it's a stack address). is `sd_softdevice_is_enabled` OK to call from interrupt context? _goes off to read docs_no restriction on when it's allowed to call it is specified
@tulip sleet do you have any hints before I get too far in the weeds?
tulip sleet
did you erase flash or anything like that? If you just import _bleio does it crash. I'm wondering if the softdevice region of flash got erased
onyx hinge
I have been bootloading it and jlinking it all day. Let me check whether 5.3.0 has this problem and that would confirm that theory
it's a particle xenon fwiw
no, 5.3.0 is fine, it advertises and appears in bluefruit connect app
I assume it has to be my changes
it's perhaps unusual to enable/disable interrupts from an interrupt .. ?
tulip sleet
so with the new one if you import _bleio it crashes? The process of importing will call sd_is_enabled. you could also try 5.4.something or the 6.0.0 alpha
onyx hinge
I don't know for sure how much/little of my small script is needed to trigger. I can investigate further.
It was the smallest script I could run that got me to appearing in the bluefruit connect app ...
yes that is enough
though having done that the traceback is actually pretty different
327 reset_into_safe_mode(HARD_CRASH);
(gdb) where
#0 HardFault_Handler () at supervisor/port.c:327
#1 <signal handler called>
#2 0x00029a36 in sd_softdevice_is_enabled (
p_softdevice_enabled=0x20015592 <sd_is_enabled.lto_priv> "\001")
at bluetooth/s140_nrf52_6.1.0/s140_nrf52_6.1.0_API/include/nrf_sdm.h:346
#3 0x0002af06 in common_hal_mcu_disable_interrupts ()
at common-hal/microcontroller/__init__.c:55
#4 0x00031978 in background_callback_add_core (
cb=cb@entry=0x20014abc <callback.lto_priv>)
at ../../supervisor/shared/background_callback.c:38
#5 0x0003a56e in background_callback_add (data=0x0, fun=<optimized out>,
cb=0x20014abc <callback.lto_priv>)
at ../../supervisor/shared/background_callback.c:59
#6 USBD_IRQHandler () at supervisor/usb.c:99
#7 <signal handler called>
#8 0x00036da6 in tu_fifo_count (f=0x20012750 <_cdcd_itf+16>)
at ../../lib/tinyusb/src/class/cdc/cdc_device.c:119
#9 tud_cdc_n_available (itf=0 '\000')
at ../../lib/tinyusb/src/class/cdc/cdc_device.c:121
#10 tud_cdc_available () at ../../lib/tinyusb/src/class/cdc/cdc_device.h:178
#11 serial_bytes_available () at ../../supervisor/shared/serial.c:94
#12 mp_hal_stdin_rx_chr () at ../../supervisor/shared/micropython.c:40
#13 0x000709e8 in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)
``` same thing, arbitrary code -> USBD_IRQHandler -> ... -> sd_softdevice_is_enabledthe stack looks damaged. It should be clean all the way down
onyx hinge
I agree it's suspicious, but I'm not building with DEBUG= so I've shrugged it off. I feel like gdb does that to me a lot
tulip sleet
could you push your code to your repo?
onyx hinge
branch background-callback of jepler/circuitpython
I assume the problem is introduced at 4024e384d0a7 "nRF: Convert USB to run via callback"
tulip sleet
could a background callback you call add itself back on to the linked list or otherwise cause the linked list to grow?
onyx hinge
It's supposed to be safe to "add" a callback that is already added
+ CALLBACK_CRITICAL_BEGIN;
+ if (cb->prev || callback_head == cb) {
+ CALLBACK_CRITICAL_END;
+ return;
``` This test is supposed to stop without modifying the linked list if `cb` is on the list. It could be on the list at the HEAD (if callback_head == cb) or anywhere but the head (if cb->prev is not NULL)CALLBACK_CRITICAL_BEGIN is calling common_hal_mcu_disable_interrupts, so we aren't even getting to that test
just before 4024e384d0a7 usb is broken on nRF because I've stopped background tasks running all the time already, so it's not super-legit to test at that ref. You can sort of do it by calling supervisor_enable_tick() at the debugger prompt. When you do that, import _bleio works just fine.
Narrowing it down further ... With this change ALONE relative to origin/main@37e77b21c ("Merge pull request #3140 from dunkmann00/docs-version-conf") you get hardfaults at import _bleio: ```$ git diff
diff --git a/ports/nrf/supervisor/usb.c b/ports/nrf/supervisor/usb.c
index 3d2527faa..d8877fd08 100644
--- a/ports/nrf/supervisor/usb.c
+++ b/ports/nrf/supervisor/usb.c
@@ -88,6 +88,9 @@ void init_usb_hardware(void) {
}
}
+uint8_t some_global;
+
void USBD_IRQHandler(void) {
tud_int_handler(0);
- sd_softdevice_is_enabled(&some_global);
}
the softdevice will hard-fault if it misses a timing window, which can easily happne if you have turned off interrupts, instead of using the softdevice critical section turn-off/turn-on routines (which do their own fancy work of scheduling)
though I don't understand why it should fail on the test.
is it possible that you turned off interrupts when the sd was disabled and then turned them on when it was enabled? Then the pair wouldn't match and there would be all kinds of havoc
onyx hinge
that's hardfault and not reset_into_safe_mode(NORDIC_SOFT_DEVICE_ASSERT); ?
tulip sleet
it doesn't call the assert routine unless you compile debug
onyx hinge
in the "this change ALONE", I'm confident I didn't newly enable or disable interrupts
tulip sleet
it just hardfaults
i would double-check for accidental recursive interrupt-off, though I read the code and I don't see that
onyx hinge
I have for sure written those during this PR, it did give me a day of head scratching last time
tulip sleet
also, the change above shares a global for sd_softdevice_is_enabled(), so maybe two interrupts are checking and one is getting incorrect results
onyx hinge
I'm not sure I understand what you're saying by that. This is the only "use" of some_global. I assumed sd_softdevice_is_enabled was storing 1 or 0 in it. I'm not sure what would be "sharing" anything. but if I misunderstand totally what sd_softdevice_is_enabled is doing...
tulip sleet
i'm not sure why the interrupt handlers share the some_global.
onyx hinge
I just called the variable some_global when I declared it for purposes of demonstration, its name is purely arbitrary
tulip sleet
right, that's not what I mean, but why is the IRQ handler checking that? What does it do with the result?
onyx hinge
yes, I'm happy to explain what I'm getting at with my snippet above
I have a bunch of code in my branch, and when asking for help it's better to provide a simpler case for what's going wrong. That is what I'm trying to do.
What's "new" in my branch is: Now, USBD_IRQHandler [indirectly] calls sd_softdevice_is_enabled, plus a lot of other stuff
tulip sleet
the sd_is_enabled static is shared across two routines, and if common_hal_mcu_disable_interrupts() is called twice in a row, or the state changes between the two calls, then things will get messed up
onyx hinge
question: is the problem "a lot of other stuff" or "IRQ handler calls SD"
tulip sleet
I think the problem might be that the "is SD enabled" bool value becomes stale and wrong
tulip sleet
i'm just wondering if that's possible
onyx hinge
the minimal code I add to get the flavor of hardfault is NOT an additional use (read or write) of the circuitpython global sd_is_enabled, it's an additional call to the SD-provided function sd_softdevice_is_enabled which is expected to assign the distinct global variable some_global
tulip sleet
i'm just wondering if sd_softdevice_is_enabled is detecting that something has gone wrong
i looked in Nordic DevZone and some people have hard-faulted calling it, though it seems to be for unrelated reasons
what happens if you compiled DEBUG=1?
onyx hinge
same thing.
always on importing _bleio
so as soon as softdevice actually does become enabled
tulip sleet
it's really fishy the stack is smashed, not sure we are really seeing the correct backtrace
maybe the assumption you can call softdevice_is_enabled in an IRQ handler is wrong. Was that already the code?
onyx hinge
that is "new, for all I know"
tulip sleet
i mean, you added the SD call in the IRQ handler, is that right?
tulip sleet
what is the reason for checking the sd status there?
thinking about this, they wouldn't want you to do that, because an SD routine can do arbitrary stuff, including messing around with interrupts itself
onyx hinge
A reasonable question 🙂 this new linked list has to be manipulated in atomic fashion. The way I have learned to do those things is by having a critical section which is not interruptible, and I assumed it was OK to build that out of common_hal_mcu_{en,dis}able_interrupts
tulip sleet
there are a limited number of backround routines, right?
onyx hinge
No, MP3Decoders have them and the number of them is not static and fixed.
tulip sleet
suppose there was just a volatile boolean for each one, which you checked and called if true. Then you don't need to add or remove them. ... ok, suppose there was a boolean for each category of background routines. Some of the background routines are fixed, and they can be checked with a simple booelan. For MP3, there could be a single background routine that can maintain its own worklist. would that work?
onyx hinge
It's not the design I've been working on for weeks, and it still means you have to check a BUNCH of booleans instead of JUST ONE pointer. Getting to a single check is something I also view as important.
If we have to say "well, softdevice is what it is", something will have to give.
tulip sleet
there could be a master boolean that says if any of the other booleans are set
i think it's true that it's basically complaining about calling SD routines in an interrupt handler, but I'm trying to find doc about that
the USB interrupts are prio 0, is that right, which is higher than the SVC call:
It's fine to call them from the interrupt as long as the interrupt runs at lower priority (ie a higher number) than the SVC call. So you can call the sd* functions directly if you are running at priority 6 or 7 (the two APP LOW priorities) but not from 2 or 3 (the APP HIGH ones) as they are higher than the SVC call.
onyx hinge
what is the priority of USB?
onyx hinge
is it a static thing "USB is always high priority" or do we have a choice so that we can lower it to "so that softdevice works"?
tulip sleet
we changed the prio a long time ago during atmel-only days because we were losing interrupts, I think. I have to find that.
onyx hinge
there's a _NRFX_IRQ_PRIORITY_SET but it's not called by us
ports/nrf/nrfx_config.h:#define NRFX_POWER_DEFAULT_CONFIG_IRQ_PRIORITY 7
ah there are these
barren peak
Does anyone know if makecode arcade is supported on the nRF52840 express? I see a mention of it here that it will be added in the future, but not sure when the guide was updated, so perhaps it has support now? https://arcade.makecode.com/hardware/adding
onyx hinge
nrfx/templates/nRF52840/nrfx_config.h:#define NRFX_USBD_DEFAULT_CONFIG_IRQ_PRIORITY 7
supervisor/usb.c
46: NVIC_SetPriority(USBD_IRQn, 2);
@barren peak looks like it has been discussed here https://forum.makecode.com/t/makecode-for-nrf52840/1164/6
Update on testing nRF52840 boards:
1.) Feather nRF52840 Express - blinky works.
@tulip sleet thank you !
- NVIC_SetPriority(USBD_IRQn, 2);
+ // 6 is max priority that still allows calling SD functions such as
+ // sd_nvic_critical_region_enter
+ NVIC_SetPriority(USBD_IRQn, 6);
@onyx hinge thanks for the reply.I saw that thread but the conversation did not close.
experimental does not have nrf52840 yet as far as I can see
tulip sleet
@onyx hinge The reason we bumped the prio was https://github.com/adafruit/circuitpython/pull/901
onyx hinge
We've exceeded what I know about makecode, sorry I can't be more helpful
tulip sleet
i think it is a work in progress. The makecode folks might answer in #help-with-makecode eventually, or perhaps you can catch them when they are online
@onyx hinge I think it might make sense not to check for sd enabled all the time in the critical section routines, but instead to set a global that remembers the whether the sd is enabled. it only gets enabled/disabled one or two places. I'm not even sure it ever gets disabled programmatically once it's enabled.
onyx hinge
@tulip sleet when sd is enabled, it has to make sd calls to enter/exit critical sections so that won't help will it?
if (sd_is_enabled) {
sd_nvic_critical_region_enter(&is_nested_critical_region);
right, hmm, well, you could use a different locking mechanism for your linked list, instead of critical regions
maybe there are some test-and-set primitives you can use?
simple pulsar
Hello. I've just noticed that CircuitPython 5.3.0 doesn't reload code if code.py/etc is copied to the board (CLUE) if the running code is in an input() statement. It looks like it's deferred until after the input() completes, i.e. user (eventually) enters something. Is that regarded as a bug?
tulip sleet
@simple pulsar i agree that's unexpected; thanks for an issue!
onyx hinge
I agree that is the current behavior, I run into it all the time.
barren peak
@onyx hinge @tulip sleet thanks
simple pulsar
I rarely use input(). I've gone retro for tty stuff. I'll file it.
onyx hinge
with the IRQ priority thing reverted, it looks like sd_nvic_critical_region_enter/exit are safe to call
tulip sleet
you mean just on nrf?
onyx hinge
sorry I said something confusing again 🙂
without changing the USB IRQ priority (so it stays at 2), it looks like sd_nvic_critical_region_enter/exit are safe to call
they don't actually "call the SD"
manic glacierBOT
tulip sleet
@onyx hinge sorry, we got a food delivery. What did you change so that it works again?
onyx hinge
I have 2.5 approaches. I don't know which one to take.
- Change the IRQ priority of USB, and anything that needs to schedule tasks, so that it is 6 instead of 2.
- Track "sd is enabled" via a global and check that instead of calling
sd_softdevice_is_enabled
2.5. Ignore whether we are using sd, and always usesd_nvic_critical_region_{enter,exit}
I implemented 1 and 2.5, at least in terms of adding calls to USBD_IRQHandler to the required functions
onyx hinge
I see that's what it says there 🙂 but I also looked at the implementation, 😜
OK, won't go with 2.5 then
tulip sleet
i remember now trying to decide what to use. I also looked at the impl a year or two ago, and was confused.
onyx hinge
I squinted at it and said "I bet that's leaving interrupts (priorities?) 0 and 1 enabled, and masking the rest"
tulip sleet
yah, it's just a q of whether they might change it in the future..., i guess
onyx hinge
does soft reset disable softdevice?
tulip sleet
another q is if USB prio is made weaker again, will it get starved out as described in that PR I mentioned above. But we'd have to look at PulseIn for nrf
I think PulseIn was not the only case of stuff starving USB. audio might do it too
I think just checking a global is pretty easy
onyx hinge
what priority does PulseIn use on nRF?
nrfx_config.h:#define NRFX_TIMER_DEFAULT_CONFIG_IRQ_PRIORITY 7
tulip sleet
so that's safe
onyx hinge
with only 8 levels I guess you run out pretty fast
tulip sleet
sd_softdevice_enable and _disable are each called only one place, in common-hal/_bleio/Adapter.c. A global would be fine.
i guess there could be a race condition at the global setting/cleraing, though
onyx hinge
does soft reset go through Adapter.c to disable softdevice?
tulip sleet
bleah, no mutex mechanism in M4 architecture...
so the only issue would be if you were using levels 0 and 1 when SD was not enabled, then it wouldn't really be a critical section for interrupts at those levels
onyx hinge
but circuitpython would have no cause to use them
tulip sleet
right
so changing to those seems fine to me, with maybe a cautionary comment in the micrcontroller source file that calls them
onyx hinge
are you happier with that than with changing the USB IRQ priority?
tulip sleet
i'm thinking that there's a race condition anyway already because the checking of is_enabled() is not atomic with the branch that chooses which enter/exit to use
onyx hinge
I think there's another bug too, the counting isn't done when softdevice is enabled
but in this kind of code it's very easy to talk yourself into a wrong interpretation of what's going on
(I mean, for me to talk myself into it)
onyx hinge
@tulip sleet I don't understand why on samd common_hal_mcu_disable_interrupts always calls __disable_irq(). I guess __disable_irq() is a no-effect instruction if IRQ was already disabled...?
tulip sleet
@onyx hinge yes, i think so, why bother checking
onyx hinge
dinnertime, I will have to conclude this at a later date
thank you for your time and help
tulip sleet
us too, pad thai in process, yw, and thanks for your patience on the dead ends
onyx hinge
yum
lone sandalBOT
wise yacht
Zzzzzzzz
lyric cipher
how do i make it so When a customer inputs an ODD number, you should light up a single red LED. When the customer inputs an EVEN number, the green LED will light up.
I'm new to this whole arduino thing and my code is making my led keep blinking even without input
tulip sleet
@lyric cipher you need to be running a Firmata client on the board so it can communicate with the host. But do you want to use Firmata? What board is it?
a regular Arduino?
tulip sleet
Have you tried a tutorial like this? <> You need to compile and upload Firmata onto the board
this #circuitpython-dev channel is mainly for CircuitPython development. #help-with-arduino might have more people familiar with Firmata, and #help-with-circuitpython will have people familiar with Python, but not necessarily Python Firmata
lyric cipher
ahhhh ok thank you
tulip sleet
here's another link: https://realpython.com/arduino-python/#hello-world-with-arduino-and-python
I found that in the Python server in its #microcontrollers channel, after searching for "firmata"
@lyric cipher ^^
lyric cipher
I found that in the Python server in its #microcontrollers channel, after searching for "firmata"
@tulip sleet Ooooo thank youuu
woven shore
Is this is the right place to ask about circuitpython libraries?
Just wondering whether the board module is a dependency for busio
@gilded cradle I've been coding a lot on the Jetson Nano platform and the board module is tripping up some of my code
woven shore
ok got the cue! Thanks ! I'll put it into #help-with-circuitpython
onyx hinge
good morning .. still trying to get common_hal_mcu_disable_interrupts and _enable_interrupts working in the new way I need them to on nrf
onyx hinge
@tulip sleet OK here's what I ended up with, I'd really appreciate your eyes on it before I proceed. https://github.com/adafruit/circuitpython/commit/aa87b2fea89b303b28095d23d171be200d44dc70 -- with the rest of the background task list stuff on top of it, everything seems to work, _bleio imported or not
tulip sleet
@onyx hinge This looks very good, and I appreciate! Could you just add a comment at line 53 that warns that only levels 2-7 are disabled, and that levels 0 and 1 are reserved for use by the SoftDevice, and shouldn't be use by CircuitPython code? (I think that's right.)
onyx hinge
adding the comment in common_hal_mcu_disable_interrupts ?
tulip sleet
yah
onyx hinge
sure thing
+ //
+ // This only disables interrupts of priority 2 through 7; levels 0, 1,
+ // 5, and 6 are reserved for softdevice and should never be used, so
+ // this limitation is not important.
sd_nvic_critical_region_enter(&is_nested_critical_region);
i didn't know about 5 and 6, so our code should only use 2,3,4?
onyx hinge
oops I got the numbers wrong
let me double check
the 2 highest and 2 in the middle are taken by sd
but it's 4 and 5, not 5 and 6
eek I apparently have been getting SOME of my information from nrf51 series softdevice docs
OK, it's 0, 1, and 4 that are reserved. Not sure where I got that there were two middle priorities reserved.
+ // and 4, are exclusive to softdevice and should never be used, so
+ // this limitation is not important.
@slender iron there aren't any guides or notes or docs or anything about getting started with the esp32-s2 on circuitpython, right? I'm having trouble getting CMake to find the xtensa-esp32s2-elf-gcc compiler.
I think it's probably because I ran install.sh in my user folder? Not in the actual circuitpython directory? Do you have any notes on what I need to set my environmental variables to if I want it to build in Circuitpython?
solar whale
In the esp-idf folder run . ./export.sh
ionic elk
No changes. Still get
-- Found Git: /usr/local/bin/git (found version "2.24.1")
-- The C compiler identification is unknown
-- The CXX compiler identification is unknown
-- The ASM compiler identification is unknown
-- Found assembler: xtensa-esp32s2-elf-gcc
CMake Error at esp-idf/tools/cmake/project.cmake:307 (__project):
The CMAKE_C_COMPILER:
xtensa-esp32s2-elf-gcc
is not a full path and was not found in the PATH.
did all that.
solar whale
May need to rerun install.sh in the esp-idf folder
ionic elk
Wondering if I did it in the wrong place, though. Are specific environmental variables required?
solar whale
Yes. I think so.
ionic elk
Like does IDF_TOOLS_PATH need to change
could you tell me where it needs to go?
It doesn't specify any suggestions in the startup guide or the port
solar whale
I just followed that guide and it all works. I have to run the export.sh every time I open a new shell.
What OS are you on.
ionic elk
Hmm. So you followed the guide, didn't need to change any environmental variables (so presumably you have your tools in ~/.espressif), ran install.sh and export.sh from inside esp32-s2/esp-idf, and then ran make BOARD=<board> and it all worked?
I'm on mac OSX
solar whale
Ah . On Mac I also had to switch to bash shell.
ionic elk
Yeah I'm on bash.
Do you manually activate the virtualenv? I see that it creates one but I don't see instructions about actually turning it on so I assume the build process does that automatically if it needs it
ionic elk
Ok I re-ran install.sh within circuitpython and it seems to be working now. I guess it can't re-use the build tools across installations? Seems like it even re-downloaded everything.
solar whale
Yeah. I won’t pretend to understand all that it does. Just happy it works!
ionic elk
Well it isn't compiling lol so I'm not done yet
pyparsing.ParseException: Parsing sections info for library /Users/hierophect/Firmware_Projects/circuitpython/ports/esp32s2/build-espressif_saola_1_wrover/esp-idf/esp-idf/xtensa/libxtensa.a failed. (at char 0), (line:1, col:1) ninja: build stopped: subcommand failed.
solar whale
running on mine... just to be sure
ionic elk
this build process is also SUPER verbose, is there any way to trim it down?
@solar whale weird I tried just running make clean and built it again, and it worked?
solar whale
here is what I get -- worked ok https://pastebin.com/LittFDaF
ionic elk
It was clean the first time too so I don't know what's up with that
^looks about the same as what I have now. thanks for your help!
ionic elk
yeah new toolchains are always hairy.
ionic elk
@solar whale what are you using as your flash command? make BOARD=... flash?
solar whale
on the Mac -- that works for me
lone axle
Is there any known issue with github at the moment?
solar whale
on other OS I have to specify PORT=/dev/ttyUSB0 but the default on the Mac is good for me.
@lone axle I can access it.
lone axle
There is a PR whose fork/branch had more commits pushed to it. And those commits show up if I look at the fork/branch in question directly. But they never showed up in the PR.
silver tapir
@lone axle I pushed the re-request review to see if that triggered the second commit to get into the pr. But yeah, I'll try again in the afternoon.
Thanks for the review.
lone axle
I can test it out on my PyPortal later today as well.
lone axle
This PR: https://github.com/adafruit/Adafruit_CircuitPython_PyPortal/pull/82 is not pulling in this commit: https://github.com/fede2cr/Adafruit_CircuitPython_PyPortal/commit/f08b1019c31d0ccbce224b8838fa4fc074e9d055 Typically in my experience there is nothing "special" that needs to be done beyond pushing a new commit to have it show up in the PR. I've never seen it get in a state like this where the direct repo page shows the commits, but the PR page does not.
onyx hinge
just getting to my first flash on Linux but yeah I need to specify BOARD= PORT= flash
Wrote 480256 bytes at 0x00000000 in 14.2 seconds (270.5 kbit/s)...
Hash of data verified.
does it put the repl on the debug usb port, or do I have to hook up real usb now?
ionic elk
@solar whale I needed to do it on my mac
default was something like /dev/tty.SLAB_USBtoUART and mine was just a string of numbers
solar whale
ah -- well that makes sense
ionic elk
Ok, I guess I have something running? I get bootloader text on the USB to UART controller... but, uhh, now what?
Do we not have CDC USB running for the ESP32-S2 yet?
don't see a drive either
solar whale
mine comes up as /dev/tty.SLAB_USBtoUART
onyx hinge
The CircuitPython USB that provides filesyste etc, is not the connector on the board, you have to wire up your own usb connector
ionic elk
mine is tty.usbserial-1411120
onyx hinge
I'm just trying to find the info
lone axle
Looks like there are a few other people reporting similar issues with github PRs not showing new commits today: https://twitter.com/search?q=Github PR commit&src=typed_query&f=live
Twitter
silver tapir
There's a very nice guide for starting with the esp32-s2 here.
https://gist.github.com/askpatrickw/0a276c7e2d4f54e442b2cb6eaa0d32ea
ionic elk
Oh that's right I need the other connector
solar whale
I see tty.usbserial-1411120 as well...
onyx hinge
OK yes, white to "19" and green to "20" and now I see CIRCUITPY
Adafruit CircuitPython 6.0.0-alpha.1-84-g910f69c42 on 2020-07-15; Saola 1 w/Wroom with ESP32S2
thanks @silver tapir that link was exactly the info I needed
manic glacierBOT
silver tapir
BTW, a friend from centralamerica was able to load circuitpyton on a new board. The board is not his design.
https://twitter.com/N10_SoJu3/status/1283422121251799046
idle owl
@silver tapir What board? We can add it to the newsletter 🙂
onyx hinge
huh. this device comes up with different SNs. Usually it comes up as SerialNumber: 0000000000000000060000000EFCDF but at least twice it came up as 00000000000000000135DFF30EFCDF
silver tapir
@idle owl It's kinda like an m0 fona.
http://es.makerfabs.cc/product/maduino-zero-sim800c.html
idle owl
@onyx hinge It's a more precise version. 😉
onyx hinge
oh yes it must be floating point
anyway, goal achieved, I ran esp32s2 and it works after my background and usb changes. hooray.
and it's already 11AM
idle owl
@silver tapir Thanks! I emailed Anne with the Twitter link and the board link.
@onyx hinge 🎉
onyx hinge
fwiw I am NOT seeing the device come back after storage.erase_filesystem() ... hmm, I thought it was supposed to do an orderly reboot in fairly short order
idle owl
Hmm I thought you were suppose to reset after that.
solar whale
on other boards it does reboot -- not on esp32s2
but it will drop the USB connection on other boards -- still esp32s2 is different....
onyx hinge
OK thanks @solar whale nice to have it verified that it's not just me or my changes to the USB code
does anyone else work in a way that they would notice if the serial number was changing?
solar whale
never looked -- remind me how to read it?
onyx hinge
different ways on different systems.
onyx hinge
on linux this tells me all the different serial numbers used by espressif devices that I've plugged in today: grep -A1 "Manufacturer: Espressif" /var/log/kern.log | grep SerialNumber | awk '{print $10}'
I've got 5 different serial numbers logged, including just "0" 😕
main meteor
Sounds like a timing/synchronization problem
onyx hinge
this is a problem for my workflow, I like to automatically reopen the repl by using tio /dev/serial/by-id/[something long that has the serial number]