(if it exists!)
#circuitpython-dev
tulip sleet
... actually, there are no no-ota UF2 bootloaders? so it doesn't exist?
that's why we're waiting for 10.0 to do this
jaunty juniper
yeah I was under the impression that the UF2 bootloader needs to match, so I installed with the bin
tulip sleet
I didn't think it through 🙂
jaunty juniper
and I can do esptool.py erase_region 0 0x210000 to keep the drive when updating
(but still erase the flash that will be overwritten to be safe)
there is a partitions-4MB-noota.csv in tinyuf2 which matches, so I could compile that for my own use but eh
lone axle
<@&356864093652516868> the weekly meeting will be occuring at the usual time of 2pm Eastern (just under 1 hour from now). You can add your notes to the doc https://docs.google.com/document/d/1XVbpKF0WlpLap2RIqy4IgeEOFLTUaBdCRs_qWGoj1Qs/edit?usp=sharing We look forward to hearing from everyone!
tulip sleet
@slender iron https://github.com/dhalbert/circuitpython/tree/nrfx-update
This is from 2 years ago. There is a commit for the nrfx repo, not sure what version it is exactly.
- The original idea was to just to start using the
nrfx_*_init_check()routines which they added, kind of on my request, but several years too late. Right now we have our own patches to make the driver inits be idempotent. This would get rid of those adafruit-specific patches. - But as I got into it I found many other changes that made it look like significant work, and I ran out of time. I think maybe the I2C driver changed, or some other driver, and there were other significant changes that required noticeable refactoring on our part.
See https://github.com/NordicSemiconductor/nrfx/issues/43 re the idempotence
lone axle
I am in favor of discontinuing 8.x
Thanks for hosting Dan, have a great week everyone!
turbid radish
Thanks Dan!
tulip sleet
Here is the notes document for next Monday’s CircuitPython Weekly Meeting. It is at the normal time of 11am Pacific / 2pm US Eastern here on Discord. Add your hug reports and status updates to the document before the meeting. If you are unable to attend but would still like to contribute, feel free to add your notes and we’ll read them off during the meeting. Hope to see you there! <@&356864093652516868>
https://docs.google.com/document/d/1r4p9vARPphE1LlejdCZIOuOJ1eD_ZxGTvukjTnw3Ryg/edit?usp=sharing
slender iron
@tulip sleet you ok reviewing the zephyr port PR? I'll un-draft once the CI is ok
onyx hinge
@slender iron what graphics mode should feather rp2350 dvi come up in if it detects a monitor is detected? Right now I coded 320x240x8bpp but that's an arbitrary choice.
slender iron
I think limor suggested 320x240x16bpp
onyx hinge
OK, I can do that
slender iron
maybe not if it doesn't leave much ram left
onyx hinge
I'm testing with a feather I soldered psram to so that's not a good test
slender iron
maybe it should be less if there isn't psram?
onyx hinge
>>> import gc
>>> gc.mem_free()
335056
```If it's without psram, the 8bpp mode leaves 335k free@16bpp ```>>> gc.mem_free()
261328
Any objection to moving core python file lint and format to ruff? That matches MP
onyx hinge
is this something that changed since the last merge?
slender iron
is this something that changed since the last merge?
MP changed it in 1.22. My zephyr PR changes requirements-dev.txt which causes a cache miss for black and then fetches a newer black. So I'd rather move to ruff than a newer black
onyx hinge
huh, we merged 1.22.2 in before 9.2.0, so I don't know why we didn't get on the ruff train then
so yeah go for it!
slender iron
huh, we merged 1.22.2 in before 9.2.0, so I don't know why we didn't get on the ...
Probably different CI setups
mortal kernel
@slender iron I'd like to push a branch from my private lwIP repo to the Adafruit lwIP repo. This is an update of lwIP to top of main with a local Adafruit modification rebased and applied. How to proceed?
tulip sleet
<@252717193496756235> I'd like to push a branch from my private lwIP repo to the...
just make a PR to that repo
onyx hinge
@mortal kernel I assume you mean your public fork of lwIP? I don't think PRs can be made from private repos into public repos.
When I have done work in circuitpython that needs changes to a submodule, I've created a PR into the adafruit fork of the submodule at the same time as I created a PR into circuitpython. The circuitpython PR points the submodule at the unmerged PR.
After everything is approved, the submodule PR is merged and the circuitpython PR is updated to point to the tip of the submodule's main branch before it is also merged.
When applicable, I also create a PR upstream. For instance, if you're fixing an outright lwip bug that exists upstream, it's good practice to also PR the fix into the official lwip repo.
boy lwip doesn't make a lot of releases, do they. 2023, 2021, 2018...
oh and call out and link the submodule PR when you make your CP PR
mortal kernel
@onyx hinge and @danh Thanks for the tips! I've submitted my PR (#1) to adafruit/lwip. I re-synced my personal public origin repo to lwip_tcp/lwip , branched it, and then re-based the Adafruit secondary host support on top of the current top of master. Hopefully this will merge properly.
Once that's merged, I'll update my CircuitPython pull to point to the new circuitpython9 branch.
lwIP looks like it has one lonely dev.
onyx hinge
hug report to eightycc for working on that lwip stuff!
onyx hinge
@slender iron on feather rp2350 (with my changes?? could it matter?) I am seeing the display go blank during flash read/write. I didn't expect that to matter, since isn't it all the code supposed to be on core1 and the framebuffer in SRAM so nothing can stop it or slow it down?
this is one that I soldered PSRAM onto fwiw
onyx hinge
@devout jolt it randomly occurs to me you should be able to do 12x rotary encoder on rp2350 now, 4 on each of the 3 PIO peripherals.
devout jolt
<@352910176736772096> it randomly occurs to me you should be able to do 12x rota...
Nice! I can resurrect this project! https://github.com/todbot/pico8enc (I did get it working on RP2040 by creating a DualIncrementalEncoder pio-based class
onyx hinge
too few pins to do the switches though, unless you use rp2350b
I forget why you couldn't get to 8 on circuitpython without modifications
devout jolt
too few pins to do the switches though, unless you use rp2350b
could use 4051 mux for the switches 🙂
onyx hinge
with rp2040
devout jolt
I forget why you couldn't get to 8 on circuitpython without modifications
ran out of statemachines if I also wanted to do Neopixel
onyx hinge
oh a neopixel too. yeah.
and i2s audio needs one state machine as well. because how could you NOT do audio on the next one.
so 10 knobs + i2s + neopixel. you know, after you're done with all your other ideas
devout jolt
so 10 knobs + i2s + neopixel. you know, after you're done with all your other id...
sounds like a fun time ahahah
manic glacierBOT
slender iron
<@252717193496756235> on feather rp2350 (with my changes?? could it matter?) I a...
maybe an interrupt is blocked? I'm trying to remember if there is an interrupt between frames
mortal kernel
@onyx hinge Following up on your review comments, I'd like to create a PR to push a new circuitpython9 branch from my public repo (https://github.com/eightycc/lwip/tree/circuitpython9), but github won't let me do it.
slender iron
<@426200547024961539> Following up on your review comments, I'd like to create a...
PRs update automatically with the state of your branch
mortal kernel
PRs update automatically with the state of your branch
The last PR I submitted to adafruit/lwip wound up on the master branch where what I really wanted was for it to become a new circuitpython9 branch. I see that happened because github's pull request generator selected master as its destination. Trying to fix it with a new PR, I updated the circuitpython9 branch in my public repo to include the 2.2.1 release commits. but when I try to create a new PR github refuses to allow me to select circuitpython9 as the destination. I'm missing something obvious, but I'm not seeing it.
slender iron
The last PR I submitted to adafruit/lwip wound up on the master branch where wha...
PRs don't bring in a new branch name. They merge changes from one branch (your cp9) to another (adafruit/master). Why do you want a different branch? adafruit/master is the one used by circuitpython and there is a separate lwip master that we can reference.
mortal kernel
PRs don't bring in a new branch name. They merge changes from one branch (your c...
I was following the pattern I saw in the lwip repo: the Circuitpython mods were on branch circuitpython8. @onyx hinge called out the use of the name master, so I thought it would be logical to put the rebased local mods on circuitpython9.
slender iron
ok, I'll change it directly
mortal kernel
Thank you! Sorry for the kerfluffle.
slender iron
np
thanks for fixing it up! would you mind updating to 2.2.1 proper?
probably just the changelog update but it'd be nice to know it is a released version
onyx hinge
thank you @mortal kernel, I appreciate you being responsive even about minutae!
lone sandalBOT
orchid basin
onyx hinge
interesting, some espressif boards have audiomp3 enabled but no audio output enabled. This used to not be a build error, but in #10036 it is.
blissful pollen
That's interesting. I really like that PR (I'll take a look at it later in more detail). I was actually wondering earlier if it would be possible to also remove the memory buffers we use to buffer audio in many places to one object they could all share.
onyx hinge
I'm not exactly sure how that would work. Has the RAM size of these buffers been an issue in practice, or is it more a theoretical benefit?
the other thing I'd say in retrospect about the new audio functionality is to consider whether it REALLY needed to work on anything but 8-bit signed samples. but it may be too late for that now, until version 10.
blissful pollen
Just theoretical at the moment.
8-bit or 16-bit? I was thinking about that as well.
onyx hinge
8-bit 16-bit
blissful pollen
It would make the code cleaner, and it would not be hard to write a 8 to 16 bit converter effect or vice versa
manic glacierBOT
manic glacierBOT
manic glacierBOT
onyx hinge
aha I figured out why ci_set_matrix wasn't building what it should. 😰 that script gives me cold sweats whenever I look at it, but I think I can fix this problem.
tulip sleet
aha I figured out why ci_set_matrix wasn't building what it should. 😰 that scri...
is default___exit___obj a macro too? I guess I'd just make it consistent, one way or another
onyx hinge
No, default___exit___obj is a function object, just defined in a central spot.
tulip sleet
a macro is OK. It's nice that default___enter___object and default____exit___object have analogous names. Principle of least surprise. I'd use those names.
onyx hinge
a macro is OK. It's nice that `default___enter___object` and `default____exit___...
makes sense to me as well. I'll change the existing sites that are definining the __exit__ method from mp_identity_object to default___enter___obj for clarity & consistentcy, even though the latter is a macro that expands to the former.
slender iron
aha I figured out why ci_set_matrix wasn't building what it should. 😰 that scri...
yup, me too. I was surprised when a header file change didn't trigger any rebuilds
onyx hinge
yup, me too. I was surprised when a header file change didn't trigger any rebuil...
I was glad to get to that "aha" moment
slender iron
I wonder if 10 pre-releases should be minor branches off 9.x stable that changes the partition layout
I don't think we want a hard branch because there aren't that many additional changes
or maybe we do 10 as a small and quick stable release
tulip sleet
I wonder if 10 pre-releases should be minor branches off 9.x stable that changes...
we have to do the TinyUF2 builds with the new partition tables first, and give them some clear names. Not sure if they should be new boards or additional artifacts produced in a board directory. They could be given clearly specified names that point out the difference.
What do you not want to include in 10 that might be in main?
we could have a live discussion about this vs typing
tulip sleet
@mortal kernel this may of be interest?? https://github.com/micropython/micropython/pull/16623
at least it may point out a potential problem, not necessarily the fix we want
slender iron
@tulip sleet It may be better as a tinyuf2 issue so thach can weigh in
slender iron
thanks!
it's probably ok to just switch all 4mb boards to the noota version
release it as a major version bump of tinyuf2 and say it is required for cp 10
tulip sleet
but maybe the filenames should have "cpy10" in them so people don't smash their 9.x boards by accident
slender iron
9.x uf2 should still work with the newer tinyuf2. they just won't take as much space
tulip sleet
good point
slender iron
- as long as ota_0 just gets ota_1 added
tulip sleet
not sure what you mean by that
slender iron
I checked, noota has uf2 and the fat fs in the same location as the ota version
mortal kernel
at least it may point out a potential problem, not necessarily the fix we want
This is one of several statically allocated pools eliminated by going to dynamic allocations in lwIP. Maybe MicroPython should do the same?
tulip sleet
you could comment in that PR 🙂
mortal kernel
do all dynamic CP 9 allocations now use TLSF?
Ultimately, yes.
crimson ferry
so if I read right, in theory, a lot fewer of the memory allocation exceptions that people see would be attributable to fragmentation
mortal kernel
so if I read right, in theory, a lot fewer of the memory allocation exceptions t...
Not sure about that. The VM takes chunks from tlsf and manages its own object allocations inside them.
crimson ferry
ah, ok
slender iron
TLSF manages a heap outside of the VM. The VM requests chunks of the outer heap to use as its heap.
crimson ferry
naive question: why not allocate directly from TLSF?
tulip sleet
naive question: why not allocate directly from TLSF?
we'd then have to scan all of the TLSF heap for garbage collection.
slender iron
right, tlsf requires manual alloc/free
the MP VM does pointer collection to know what is still in use
crimson ferry
thanks, that's making some sense
slender iron
the MP split heap functionality where it grows with use actually speeds up the collect process I believe
mortal kernel
Another benefit I've seen is that VM heap can be completely purged without affecting lower-level allocations. I've taken advantage of this in my ramlog facility.
slender iron
since it can be small when using little memory
tulip sleet
@slender iron [I'm writing the issue] we can't do BLE on ESP32-S2, so should we get rid of ota_1 on ESP32-S2 also? I can't remember how short we are on S2 for space
slender iron
I can't either. I think it is fine to leave. I don't know what else we're going to add to it
tulip sleet
I can't either. I think it is fine to leave. I don't know what else we're going ...
The biggest ESP32-S2 Feather builds, for example, have about 20kB free. So we could leave them alone, I think.
slender iron
does the board def have stuff disabled?
tulip sleet
nope, nothing is disabled
onyx hinge
I thought -ggdb just controlled whether debug information was generated, not optimization. and that debug information was present in the .elf but not the .bin or .uf2. I wonder why just adding that flag, and not changing anything else, makes this espressif build overflow.
tulip sleet
I thought `-ggdb` just controlled whether debug information was generated, not o...
are you adding that directly or via DEBUG= or something?
I agree that's what the gcc doc says
onyx hinge
+++ b/ports/espressif/Makefile
@@ -174,7 +174,7 @@ ifeq ($(DEBUG), 1)
# You may want to enable these flags to make setting breakpoints easier.
# CFLAGS += -fno-inline -fno-ipa-sra
else
- CFLAGS += -DNDEBUG
+ CFLAGS += -DNDEBUG -ggdb
so maybe the/some debug info is in the .bin?? I certainly used to use strip a lot when doing native compiles.
i think it's a linker question, maybe
onyx hinge
lots of functions (according to nm --print-size) get bigger with -ggdb. for instance, mp_obj_str_binary_op grows by 84 bytes for .. some reason
(gdb) disas app_main
Dump of assembler code for function app_main:
0x4009fa10 <+0>: entry a1, 32
0x4009fa13 <+3>: or a1, a1, a1
0x4009fa16 <+6>: call8 0x4009e63c <main>
0x4009fa19 <+9>: retw.n
# debugging off
(gdb) disas app_main
Dump of assembler code for function app_main:
0x4009db80 <+0>: entry a1, 32
0x4009db83 <+3>: call8 0x4009c8d4 <main>
0x4009db86 <+6>: retw.n
End of assembler dump.
In this case it's inserting a nop-like instruction (I think that 'or' is OR'ing the a1 register with itself), only when -ggdb is turned on. In a bigger function like mp_obj_str_binary_op it's harder to find all the changes, but the first one's also an insertion of that same or a1, a1, a1 instruction, just before a call8.
but the whole function is actually organized differently and it rapidly diverges after that.
well, that's interesting to know.
mortal kernel
```# debugging on
(gdb) disas app_main
Dump of assembler code for function app_m...
I've seen the same with -ggdb, it does seem to interact with optimizations. If you're looking for debug with a smaller flash footprint, try -Og.
crude blaze
@slender iron if you are currently working with nRF5340 then you need to make sure you have a an image defined for the network core and have a transport defined to use for talking to the ble controller on the network core from the application core, if there is no firmware on the network core you can't do BLE, the app core has no access to the radio
https://docs.zephyrproject.org/latest/connectivity/bluetooth/bluetooth-arch.html might be helpful to get a better idea of the interplay
the app core firmware will always be a host only build and the network core a controller only build (can be a combined build but rare that you want that)
onyx hinge
anybody know why some tables use MP_ROM_QSTR and others used MP_OBJ_NEW_QSTR? e.g.,```c
{ MP_ROM_QSTR(MP_QSTR_TOUCH), MP_ROM_PTR(&pin_PA07) }, // neokey trinkey
{ MP_OBJ_NEW_QSTR(MP_QSTR_APA102_MOSI), MP_ROM_PTR(&pin_PA12) }, // cp32-m4
It seems like in the configurations where we build, `MP_ROM_QSTR` expands to `MP_OBJ_NEW_QSTR` but with different build choices extra curly brackets may be necessary to appear in a ROM table.... but I'm not too sure about this.This was a MicroPython changeover, as I think you know. Worth looking at whether there are remaining MP_OBJ_NEW_QSTRs in their code.
manic glacierBOT
slender iron
the app core firmware will always be a host only build and the network core a co...
I'm doing a sysbuild so I think both should be included. That's my intent for CP firmwares. I want all firmware needed in one.
manic glacierBOT
onyx hinge
This was a MicroPython changeover, as I think you know. Worth looking at whether...
No, apparently I missed this change at the time
manic glacierBOT
candid sun
@blissful pollen question about the audio effects modules- is the plan that they only work with synthio? i was experimenting yesterday with passing an analogbufio buffer but got this error TypeError: 'array' object does not support 'protocol_audiosample'
manic glacierBOT
blissful pollen
<@228322543507865601> question about the audio effects modules- is the plan that...
The plan is general use, I have tested it with wave files, MP3 and raw audio. Let me take a quick look at analogbufio
candid sun
okay cool. yeah basically my idea is to try to use it as a guitar pedal so i was using a 1/4" jack to an analog pin
blissful pollen
I think Cooper was doing some work on an RP2350 audio input that could be a source.
From what I see BufferedIn just goes into an array. Would need a layer to present that buffer to the audio "system" as such. Probably not too complex.
@candid sun Crazy idea time:
ain = analogbufio.BufferedIn(board.GP26, sample_rate=22050)
raw = audiocore.RawSample(mybuf, sample_rate=22050)
ain.readinto(mybuf)
And then play the raw sample on repeat. I don't have anything handy to provide an audio single in to test this, but it may work?
So something like mixer.play(raw) and then a loop reading audio forever
candid sun
@blissful pollen ohh okay, thanks! i can try it in a bit
candid sun
<@347203168121126912> Crazy idea time:
```mybuf = bytearray(1024)
ain = analogbu...
so this does seem to be working, i'm having audio coming through. i need to improve my circuit setup but this is a good starting point. thanks!
blissful pollen
Nice! Let me know how it goes. I was thinking through how to make an audio object that plays a buffer and realized we had one already.
short tendon
@tulip sleet quick question for you. Been out for a bit, so don't know what's right. I have some updates I want to make to the SSL test example in the ConnectionManager library, but have seen a few comments about not having examples like that (since it causes unnecessary releases). What's the right way now?
tulip sleet
<@329766224093249548> quick question for you. Been out for a bit, so don't know ...
It's fine, go ahead. I was trying to minimize having tons of examples added to certain libraries but was convinced otherwise. Some examples could be Playground notes
crude blaze
I'm doing a sysbuild so I think both should be included. That's my intent for CP...
so you have one of these in your sysbuild config?
https://docs.nordicsemi.com/bundle/ncs-latest/page/nrf/applications/ipc_radio/README.html#sysbuild_kconfig_options
They are responsible for pulling in the controller firmware sample for IPC or RPC based HCI.
You can also make sure that you have it setup by observing if both application and network core get flashed. They have separate flash banks and run independent from each other (though app core kicks off netcore with a signal). Zephyr generates a "merged_domains.hex" if its multi-image build.
fleet hollow
so this does seem to be working, i'm having audio coming through. i need to impr...
Here's a gist which goes over some of the usage with analogbufio. https://gist.github.com/relic-se/77e48df1920b20d3e50762cbc06d1bb9 You could also use the new pio i2s library to do the same thing, but you'd need a separate i2s bus for output (if needed).
candid sun
Here's a gist which goes over some of the usage with analogbufio. https://gist.g...
awesome, thank you!
fleet hollow
awesome, thank you!
If you're interested in the hardware I'm working on, let me know and I can give you more info.
zinc walrus
I realize that it's still extremely early days with the CP Zephyr port, but I have an nRF5340 Audio DK that I'm very excited to test.
I'm not sure that I can accommodate actually building CP yet, but can an expert share any advice on just loading the nRF5340 runtime as an experiment?
candid sun
If you're interested in the hardware I'm working on, let me know and I can give ...
yeah actually I’m curious. I’m thinking of using an mcp6002 for the guitar input
fleet hollow
yeah actually I’m curious. I’m thinking of using an mcp6002 for the guitar input
I've heard that's a decent op-amp, but I'm using a standard tl972
thorny jay
About "Fruit Jam" discussion on Function key in "Deep Dive".
This is what the Maximite Colour 2 family of BASIC computer use.
You start with a command-line (REPL but for BASIC).
From there the F1 key launch a file browser (text base) that let you edit file, run file and preview/run some format like image file and sound file (MOD). The usage of the file browser is also using function key to select what you want to do for the highlighted file.
You also have the command EDIT that go into a full screen editor, also with function key and a bottom line description of what each key is doing. You can RUN the program you are editing directly from the editor.
Some function keys in the REPL have pre-defined function such as F1=FILES, F2=RUN, F3=LIST, F4=EDIT.
Other are macro to quickly type some stuff such as F5=AUTOSAVE "", F6= XMODEM RECEIVE "", ...
Finally F11 and F12 are user customisable with custom text.
Maximite Colour 2 is very usable just with the keyboard and you don't need anything else to program and execute your code.
manic glacierBOT
mortal kernel
@crimson ferry I've been trying to get the https server here https://github.com/ide/circuitpython-https-server working to test https performance on RP2 without success. I've patched the missing SOL_SOCKET and SO_REUSEADDR CP8->CP9 issue in TLSServerSocketPool but I'm failing with a broken pipe (-32) error. I suspect the .pem files are bad. How have you been getting it to work? Maybe there's an alternative tool you can suggest?
crimson ferry
@mortal kernel I haven't tried that one in a while, but the Adafruit library is updated for HTTPS https://github.com/adafruit/Adafruit_CircuitPython_HTTPServer, and there's an HTTPS example https://github.com/adafruit/Adafruit_CircuitPython_HTTPServer/blob/main/examples/httpserver_https.py
onyx hinge
@slender iron I'm looking without success for the way to "disable low priority interrupts", which is what I thought you were discribing for fixing the blank DVI display during flash writes problem on RP2350. I found how to set IRQ priority in the pico sdk, but I didn't finda way to mask/disable interrupts based on priority. I only found the cpu-wide IRQ enable flag that common_hal_mcu_disable_interrupts is already using.
hints welcome
there's something called PRIMASK but it's also a single enable/disable bit https://developer.arm.com/documentation/ddi0419/c/System-Level-Architecture/System-Level-Programmers--Model/Registers/The-special-purpose-mask-register--PRIMASK?lang=en
mortal kernel
<@252717193496756235> I'm looking without success for the way to "disable low pr...
The ARM NVIC prioritizes presentation of concurrent enabled interrupts, i.e., which vector to invoke first. PRIMASK will boost the PE to execution priority 0, effectively masking interrupts at a level lower than 0. The Main Extension augments PRIMASK with FAULTMASK and BASEPRI adding more specificity to priority selection. The RP2350 implements the Main Extension. Aside from access to these registers via the processor core registers, I'm not finding any support in the Pico SDK.
onyx hinge
OK, BASEPRI seems like what I'm looking for maybe
thanks!
are BASEPRI numbers the same as the irq_set_priority numbers?
mortal kernel
Yeah, I think that what this inside Irq_set_priority does:
io_rw_32 *p = nvic_ipr0() + (num >> 2);
// note that only 32 bit writes are supported
*p = (*p & ~(0xffu << (8 * (num & 3u)))) | (((uint32_t) hardware_priority) << (8 * (num & 3u)));
(inserts the priority directly into the NVIC_IPRxx register corresponding to the interrupt number)
flint flare
Hi, I'm trying to burn a T-Display-S3R8 device with the firmware for the NerdQaxe+ following the steps below https://github.com/shufps/qaxe/tree/main?tab=readme-ov-file#installation-via-usb-bootloader-on-board-with-boot-button-qaxe-qaxe, but I can't get it. In the instructions it says: ‘now start the stm32 in DFU mode by pressing boot’, (I understand that the boot button is the one on the left if you have the buttons facing you) I have done it but it doesn't do anything. Also the AIs tell me I should have three binary .bin files:
esptool.py --chip esp32s3 --port /dev/ttyACM0 write_flash -z \ 0x1000 bootloader.bin 0x8000 partition-table.bin \ 0x10000 firmware.bin
However I only have one binary firmware.bin or qaxe.bin (depends on what I name it, but it's the same).
I need help from someone who has flashed a NerdQaxe+. Best regards and thanks in advance.
PD: Web related https://docs.espressif.com/projects/esp-idf/en/stable/esp32/get-started/index.html
onyx hinge
@flint flare I'd encourage you to reach out to that project if the installation instructions aren't working for you. fwiw I wouldn't put much stock in an AI answer about how to use esptool, because the instructions vary depending on the project & which espressif mcu it is. This channel is for circuitpython development.
slender iron
<@&356864093652516868> Sorry for the late notice. We have our weekly meeting now in this text channel and in the circuitpython voice channel. The notes doc is here: https://docs.google.com/document/d/1r4p9vARPphE1LlejdCZIOuOJ1eD_ZxGTvukjTnw3Ryg/edit?tab=t.0
onyx hinge
@slender iron I can read core and/or blinka just to give you a rest during that section if you like.
errant grail
onyx hinge
[rp2350.dap.core0] clearing lockup after double fault
Thank you!
onyx hinge
thank you scott!
random junco
thanks all, have a great week
lone axle
Thanks for hosting Scott, have a great week everyone!
mortal kernel
```
[rp2350.dap.core0] clearing lockup after double fault
```
Extra points for the double fault! You usually need an MMU to get one of those.
slender iron
onyx hinge
slender iron
onyx hinge
0x10027cde in port_idle_until_interrupt () at supervisor/port.c:507
507 __WFI();
(gdb) where
#0 0x10027cde in port_idle_until_interrupt () at supervisor/port.c:507
#1 0x100264be in run_code_py (safe_mode=<optimized out>, simulate_reset=<optimized out>) at ../../main.c:741
#2 0x10026aec in main () at ../../main.c:1108
r0 0x0 0
r1 0x20006150 536895824
r2 0x3 3
r3 0x0 0
r4 0x0 0
r5 0x0 0
r6 0x0 0
r7 0x0 0
r8 0x6b0 1712
r9 0x1 1
r10 0x190 400
r11 0x0 0
r12 0x1 1
sp 0x20080f58 0x20080f58
lr 0x20004dbb 536890811
pc 0x10027cde 0x10027cde <port_idle_until_interrupt+50>
xpsr 0x69000000 1761607680
fpscr 0x10 16
msp 0x20080f58 0x20080f58
psp 0x0 0x0
msp_ns 0x0 0x0
psp_ns 0xfffffffc 0xfffffffc
msp_s 0x20080f58 0x20080f58
psp_s 0x0 0x0
primask 0x0 0
basepri 0x60 96
faultmask 0x0 0
control 0x4 4
msplim_s 0x0 0x0
psplim_s 0x0 0x0
msplim_ns 0x0 0x0
psplim_ns 0x0 0x0
primask_s 0x0 0
basepri_s 0x60 96
faultmask_s 0x0 0
control_s 0x4 4
primask_ns 0x0 0
basepri_ns 0x0 0
faultmask_ns 0x0 0
control_ns 0x4 4
#include "src/rp2_common/cmsis/stub/CMSIS/Device/RP2350/Include/RP2350.h"
mortal kernel
@onyx hinge You might be faulting on a bad mrs destination. The RP2350 has the security extension, so there is a basepri_ns (non-secure) that you may want to try.
onyx hinge
The double fault was when trying to set the NVIC interrupt enable bits, which is not the approach I'm going for. On voice call, scott & I figured out what was going wrong with the BASEPRI code, and I think I'll have a PR soon. Appreciate the educated guesswork though!
mortal kernel
Apologies for jumping in half-baked, it's just that I find stuff at this level too tempting to ignore.
onyx hinge
@slender iron ```$ make TRANSLATION=ja
Only first 256 glyphs can be displayed, font subset has 520 glyphs
make: *** [../../supervisor/supervisor.mk:270: build-adafruit_feather_rp2350/autogen_display_resources-ja.c] Error 1
@onyx hinge is the PioMatter() constructor more strict on the framebuffer size than the prior ones like AdafruitMatrixBonnetRGB888Packed() were?
I am finding that
matrix_framebuffer = np.zeros(shape=(geometry.height, geometry.width, 3), dtype=np.uint8)
used to work with the old constructor, but with the new one throws
RuntimeError: Framebuffer size must be 16384 bytes (8192 elements of 2 bytes each), got a buffer of 49152 bytes
The 3rd dimension of the array shape seems to be the culprit, if I remove it then it initializes, but further on the copying breaks.
onyx hinge
@lone axle that looks like some kind of mistake on my part, 16384 bytes doesn't seem right.
lone axle
<@382939733107408897> that looks like some kind of mistake on my part, 16384 byt...
it's with --colorspace RGB565 if that factors in.
onyx hinge
oh, from what you said I thought the message you were seeing was from an 888Packed situation
RGB565 should have 2 bytes per pixel. The array should maybe be something like np.zeros(shape=(geometry.height, geometry.width), dtype=uint16)
is 8192 the right number of pixels? that'd be 64x128 or something?
lone axle
the root problem I'm trying to solve is these two lines https://github.com/adafruit/Adafruit_Blinka_Raspberry_Pi5_Piomatter/blob/dd77450f156d8617f6cef01f6fe33aa519950adb/examples/fbmirror_scaled.py#L77-L78 successfully initilize with the old style constructor, but the cooresponding ones under the new way raise an exception.
Correct, I have 128x64 total currently (4x 64x32 panels).
onyx hinge
matrix = adafruit_raspberry_pi5_piomatter.AdafruitMatrixBonnetRGB888Packed(matrix_framebuffer, geometry)``` That is correct for RGB888Packed, it makes 3 "uint8" elements for each pixel, 3 bytes each. But it's not appropriate for RGB565The 3rd dimension of the shape is important for the way the fbmirror_scaled works currently because the array that results from np.asarray(img) has a shape with that 3rd dimension.
Right, sorry. The switch to 565 was one attempt at resolving the issue
onyx hinge
since "the thing you want to display" is a PIL imagein RGB888Packed format, I don't think --colorspace RGB565 can be expected to work.
lone axle
the same issue is present with 888 as well (which I've gone back to). it just changes the numbers in the exception.
onyx hinge
this is a difference between the unscaled and scaled demos: In the scaled demo, I have to convert the image data into something PIL can handle, so I can use the PIL scaling routine. But that means the data being displayed is in a different format between the two programs.
mortal kernel
@slender iron The duplicate application of the secondary host changes to Adafruit's lwIP repo is happening because of the --force I'm forced to use because of the weird state of the CIRCUITPYTHON9 branch. What I thinks needs to be done is to delete the CIRCUITPYTHON9 branch, make a new main branch that tracks the upstream master, re-sync the new main branch, create a new CIRCUITPYTHON9 branch from main at tag STABLE-2_2_1_RELEASE, and then I can PR the re-based local updates cleanly on top of that.
lone axle
without --colorspace RGB565 the exception is
RuntimeError: Framebuffer size must be 32768 bytes (8192 elements of 4 bytes each), got a buffer of 49152 bytes
@lone axle can you send me what you have?
onyx hinge
thanks, I'll run downstairs and load it up and see what I can understand..
lone axle
thanks, I'll run downstairs and load it up and see what I can understand..
at least 1 problem with my current code is the wrong dtype value passed to np.zeros() when creating matrix_framebuffer I mistakenly used the var dtype instead of the proper value dtype=np.uint8
onyx hinge
@lone axle I think I'm close to having it working. Maybe it's a total mistake to have --colorspace even be a commandline argument. I'll give a better explanation once I can actually get a good display
slender iron
@mortal kernel I usually start a new branch for every change I make. That may simplify it for you
onyx hinge
I suspect most of the time it's up to the code what color depth it's written for, while it has a greater chance of adapting to differing width/height/etc
@lone axle try with these changes: https://gist.github.com/jepler/7ae5df5c3877b4c97edbf5f6681c430d
mortal kernel
<@849815081251766292> I usually start a new branch for every change I make. That...
Me too. Let me run through my steps again and I'll let you know where things go off the rails.
onyx hinge
by sending colorspace=None it removes the ability to customize the colorspace via the commandline (since any other value will not work).
the "matrix" is created with dtype uint8 which you already figured out
and I left a stray debug print in, just so you have something to do 🙂
slender iron
Me too. Let me run through my steps again and I'll let you know where things go ...
that also makes it easy to force push to that branch to update the PR
onyx hinge
if you contemplate whether ANY of the demos still work when the user specifies a different --colorspace than the intended one, we can decide whether to change the default so that --colorspace is not added as a commandline argument at all
lone axle
if you contemplate whether ANY of the demos still work when the user specifies a...
What would be considered the intended colorspace of fbmirror.py the non-scaled one? In order for that one to run for me currently I have to pass --colorspace RGB565 without it I get
RuntimeError: Framebuffer size must be 32768 bytes (8192 elements of 4 bytes each), got a buffer of 16384 bytes
Yeah, RGB565 is what it uses now.
but the default default with the argument parser is --colorspace RGB888
... colorspace=piomatter.Colorspace.RGB888, ...):
Here is the notes document for next Tuesday's CircuitPython Weekly Meeting. It is at the normal time of 11am Pacific / 2pm US Eastern BUT 24 hours later than normal here on Discord. Add your hug reports and status updates to the document before the meeting. If you are unable to attend but would still like to contribute, feel free to add your notes and we’ll read them off during the meeting. Hope to see you there! <@&356864093652516868> https://docs.google.com/document/d/1oM_x1t2iD_iStf4riPa5j_naS-yqZAR-Y8_qKyVpa7s/edit?usp=sharing
slender iron
I realize that it's still extremely early days with the CP Zephyr port, but I ha...
I can add support for the audio kit
mortal kernel
@slender iron I've re-synched my fork of the Adafruit repo which automatically killed PR #3. Cloned repo to my local machine. Created a new branch off circuitpython9. Now I'd like to cherry-pick the two upstream commits that officially update to v2.2.1. but the commits are not in my repo because they've not been synched to the Adafruit repo. I've tried git fetch upstream, but no help.
rabeles at debian12 in ~/D/c/i/lwip-repo
↪ git cherry-pick b6c60989645be22b9e91bcc7a81c07fff4dabaa1
fatal: bad object b6c60989645be22b9e91bcc7a81c07fff4dabaa1
rabeles at debian12 in ~/D/c/i/lwip-repo
↪ git remote -v
origin ssh://git@github.com/eightycc/lwip.git (fetch)
origin ssh://git@github.com/eightycc/lwip.git (push)
upstream ssh://git@github.com/adafruit/lwip.git (fetch)
upstream ssh://git@github.com/adafruit/lwip.git (push)
rabeles at debian12 in ~/D/c/i/lwip-repo
↪ git fetch upstream
rabeles at debian12 in ~/D/c/i/lwip-repo
↪ git cherry-pick b6c60989645be22b9e91bcc7a81c07fff4dabaa1
fatal: bad object b6c60989645be22b9e91bcc7a81c07fff4dabaa1
Do you have lwip as a remote? git remote -v
mortal kernel
see above
slender iron
I'd add another remote for lwip
buoyant lagoon
try git fetch origin b6c60989645be22b9e91bcc7a81c07fff4dabaa1
you can tell it to just fetch a given sha1 directly
slender iron
I name my remotes by the person, not origin and upstream
mortal kernel
try `git fetch origin b6c60989645be22b9e91bcc7a81c07fff4dabaa1`
bingo. did that and then the cherry-pick worked.
abeles at debian12 in ~/D/c/i/lwip-repo
↪ git fetch origin b6c60989645be22b9e91bcc7a81c07fff4dabaa1
remote: Enumerating objects: 3, done.
remote: Counting objects: 100% (3/3), done.
remote: Compressing objects: 100% (2/2), done.
remote: Total 3 (delta 1), reused 2 (delta 1), pack-reused 0 (from 0)
Unpacking objects: 100% (3/3), 1.62 KiB | 332.00 KiB/s, done.
From ssh://github.com/eightycc/lwip
* branch b6c60989645be22b9e91bcc7a81c07fff4dabaa1 -> FETCH_HEAD
rabeles at debian12 in ~/D/c/i/lwip-repo
↪ git status
On branch issue-9837
nothing to commit, working tree clean
rabeles at debian12 in ~/D/c/i/lwip-repo
↪ git cherry-pick b6c60989645be22b9e91bcc7a81c07fff4dabaa1
[issue-9837 f89bdfda] Update CHANGELOG in preparation for 2.2.1 release
Author: Simon Goldschmidt <goldsimon@gmx.de>
Date: Wed Feb 5 20:25:11 2025 +0100
1 file changed, 47 insertions(+)
ah, earlier you did a fetch from upstream (aka adafruit)
but the cmd i gave, does a fetch from origin (aka eightycc)
that might also play a role
mortal kernel
Now I've got the two commits I wanted on my branch. But, the SHA hashes differ from the upstream repo. Could that cause trouble down the line?
slender iron
you want to merge rather than cherry pick
mortal kernel
you want to merge rather than cherry pick
I'll give it a try. Here's my current log:
↪ git log
commit 2978c5b176f608e6779271fdf4c1fd021f73ba02 (HEAD -> issue-9837)
Author: Simon Goldschmidt <goldsimon@gmx.de>
Date: Wed Feb 5 20:38:17 2025 +0100
Prepare 2.2.1 release
commit f89bdfda57220490e1ec771fd2bd53280a8413d7
Author: Simon Goldschmidt <goldsimon@gmx.de>
Date: Wed Feb 5 20:25:11 2025 +0100
Update CHANGELOG in preparation for 2.2.1 release
commit 4097a0709b70343e6bef6d129537ba4eb474ce10 (upstream/circuitpython9, origin/circuitpython9, origin/HEAD, circuitpython9)
Merge: 239918cc 2396f85f
Author: Scott Shawcroft <scott@tannewt.org>
Date: Wed Feb 5 12:14:38 2025 -0800
Merge pull request #1 from eightycc/circuitpython9
Update to almost version 2.2.1
cherry pick and rebase are things you only want to do on your own commits since it changes the commit itself (and therefore its hash)
buoyant lagoon
yeah
mortal kernel
Ouch!
buoyant lagoon
ive had trouble before, because commit X contains a fix i need
and commit X isnt in my branch at all, because it was cherry-picked
so i spent a while trying to make sense of why i didnt have the fix
mortal kernel
It's all clear to me now. I think. I'll go through my steps again from scratch just to be sure...
Thanks a million!
buoyant lagoon
cobalt grail
https://xkcd.com/1597/
OMG that is SO true!
buoyant lagoon
once you get good enough, you become the guy they call 😛
i'm now rebasing several branches on a daily basis at work
and fixing botched merge conflicts that got commited, making a mess of the history
onyx hinge
I love git, I think I'm good at it, but helping somebody through git troubles via online chat makes me want to cry.
mortal kernel
So close, but using merge still yields a different SHA:
↪ git status
On branch issue-9837
nothing to commit, working tree clean
rabeles at debian12 in ~/D/c/i/lwip-repo
↪ git fetch origin b6c60989645be22b9e91bcc7a81c07fff4dabaa1
remote: Enumerating objects: 3, done.
remote: Counting objects: 100% (3/3), done.
remote: Compressing objects: 100% (2/2), done.
remote: Total 3 (delta 1), reused 2 (delta 1), pack-reused 0 (from 0)
Unpacking objects: 100% (3/3), 1.62 KiB | 416.00 KiB/s, done.
From ssh://github.com/eightycc/lwip
* branch b6c60989645be22b9e91bcc7a81c07fff4dabaa1 -> FETCH_HEAD
rabeles at debian12 in ~/D/c/i/lwip-repo
↪ git fetch origin 77dcd25a72509eb83f72b033d219b1d40cd8eb95
remote: Enumerating objects: 13, done.
remote: Counting objects: 100% (13/13), done.
remote: Compressing objects: 100% (4/4), done.
remote: Total 13 (delta 9), reused 12 (delta 9), pack-reused 0 (from 0)
Unpacking objects: 100% (13/13), 2.40 KiB | 613.00 KiB/s, done.
From ssh://github.com/eightycc/lwip
* branch 77dcd25a72509eb83f72b033d219b1d40cd8eb95 -> FETCH_HEAD
rabeles at debian12 in ~/D/c/i/lwip-repo
↪ git merge b6c60989645be22b9e91bcc7a81c07fff4dabaa1
Merge made by the 'ort' strategy.
CHANGELOG | 47 +++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 47 insertions(+)
rabeles at debian12 in ~/D/c/i/lwip-repo
↪ git log
commit abf8c210045b6bfac7a7913e2607bb82b749dfec (HEAD -> issue-9837)
Merge: 4097a070 b6c60989
Author: eightycc <eightycc@gmail.com>
Date: Mon Feb 10 14:47:11 2025 -0800
Merge commit 'b6c60989645be22b9e91bcc7a81c07fff4dabaa1' into issue-9837
commit 4097a0709b70343e6bef6d129537ba4eb474ce10 (upstream/circuitpython9, origin/circuitpython9, origin/HEAD, circuitpython9)
Merge: 239918cc 2396f85f
Author: Scott Shawcroft <scott@tannewt.org>
Date: Wed Feb 5 12:14:38 2025 -0800
Merge pull request #1 from eightycc/circuitpython9
Update to almost version 2.2.1
😭
wraith crow
I noticed that the DVI pins for the RP2350 are the same pins used on the DVI Sock but the D0/D1/D2 pins are in a different order. Does that make sense or should the order be standardized?
slender iron
So close, but using merge still yields a different SHA:
```
↪ git status
On bran...
It is ok that the merge commit has a different sha
the two new commits should have preserved commits
slender iron
I noticed that the DVI pins for the RP2350 are the same pins used on the DVI Soc...
I think we debated this and though it didn't matter since the software can change it
slender iron
so you have one of these in your sysbuild config?
https://docs.nordicsemi.com/bu...
I believe I'm flashing both. Two flashes happen. maybe it is the 7002 instead though?
crude blaze
I believe I'm flashing both. Two flashes happen. maybe it is the 7002 instead th...
ah hm, I have not worked with the 7002 yet but it does seem to have an externally residing firmware patch flash. Could be that in a dev setting that this gets always written.
In nrfjprog log you should see it explicitly logging the flash process to the app and network core
slender iron
-- runners.nrfjprog: Flashing file: /home/tannewt/repos/circuitpython/ports/zephyr-cp/build-nordic_nrf7002dk/zephyr-cp/zephyr/zephyr.hex
[ #################### ] 31.237s | Erase file - Done erasing
[ #################### ] 6.547s | Program file - Done programming
[ #################### ] 6.677s | Verify file - Done verifying
Applying pin reset.
-- runners.nrfjprog: Board with serial number 1050728232 flashed successfully.
-- west flash: using runner nrfjprog
-- runners.nrfjprog: reset after flashing requested
-- runners.nrfjprog: Flashing file: /home/tannewt/repos/circuitpython/ports/zephyr-cp/build-nordic_nrf7002dk/hci_ipc/zephyr/zephyr.hex
[ #################### ] 12.714s | Erase file - Done erasing
[ #################### ] 1.573s | Program file - Done programming
[ #################### ] 1.604s | Verify file - Done verifying
Applying pin reset.
-- runners.nrfjprog: Board with serial number 1050728232 flashed successfully.
flashing both
crude blaze
ok nice
slender iron
maybe I need to call bt_enable()
slender iron
is there really no way to get a random number
crude blaze
if that doesn't do it then maybe try a very simple BLE sample that can makes it easier to trace down
slender iron
without ble?
crude blaze
oh you want a random number? Yea that should be possible, it has a RNG peripheral, would be suprised if that relies on the BLE radio
slender iron
wifi is getting a random value for the nonce
crude blaze
could you point me to the code/sample that suggests using BLE for the random number?
slender iron
my final zephyr.dts has zephyr,entropy = &rng_hci;
not sure where it is coming from
crude blaze
oh okay
slender iron
crude blaze
ah okay, it looks like its not actually connected with BLE. Just that the RNG peripheral is connected to the radio core and the radio core needs to forward the data to the app core
crude blaze
that is not directly related, bt_enable is for starting the actual BLE radio.
The network core is just a normal CPU but has exclusive access to some peripherals, so if you want to get something from those peripherals on the app core you need communicate it to the appcore via some mechanism. In their sample they use their RPC implementation for that, it can be used for BLE but is not strictly for that.
though looking at the datasheet there also seems to be an RNG on the app core side
so I don't know if thats truly necessary to use the network core one
slender iron
the sample is for nordic's sdk
crude blaze
yea for a lot of things you won't find a sample in vanilla Zephyr, you can pull in individual libraries like that without using their whole fork
but not relevant for this particular issue I would say
checking what the app core rng peripheral is called in the DK
&psa_rng
maybe try that
slender iron
looks like bt_enable makes it work
crude blaze
nice
slender iron
even though the call fails 🤦
crude blaze
this seems to be the appcore side rng if you don't want to rely on the network core being up for this functionality
ah only for the non secure context, in secure context its then probably "cryptocell"
https://github.com/zephyrproject-rtos/zephyr/blob/main/dts/arm/nordic/nrf5340_cpuapp.dtsi#L101
which should work with this sample, might be down to just removing that entropy user override
CONFIG_PSA_CRYPTO_DRIVER_CC3XX=y
explicitly enables the app core rng driver, normally CONFIG_CRYPTO should pull it in thoug
crude blaze
sure thing
slender iron
@onyx hinge think it'd be ok to define these based on implementation values: https://github.com/adafruit/circuitpython/blob/main/shared-bindings/socketpool/enum.h#L12 Zephyr uses different values than what we have so I'd have to map them all
I hit a static assert about it that you added
onyx hinge
<@426200547024961539> think it'd be ok to define these based on implementation v...
sadly, no. In order to make wiznet (and other non-core socket implementations) work with core ssl, I had to promise circuitpython userspace the numeric value of those constants
wiznet can't import socketpool to get the integer values, it exists on systems that don't have it at all
onyx hinge
actually I'm less sure why I added the assert THERE, it seems to have to do with adding AF_INET6, not adding ssl support for wiznet sockets. so I might be wrong.
"""WIZNET5K SocketPool library"""
# These must match circuitpython "socketpool" values. However, we cannot
# depend on socketpool being importable, so hard-code them here.
SOCK_STREAM = 1
SOCK_DGRAM = 2
SOL_SOCKET = 0xFFF
SO_REUSEADDR = 0x0004
```these are the ones that I commented had to match the core. But I think the static asserts you were talking about were about `AF_` values```oh AF_INET is one of themIf I replace time.sleep(2.0) in the loop with the following:
tick = 40
while tick:
print(effect_delay.delay_ms.value)
time.sleep(0.05)
tick -= 1
I can generate the graph below (in LibreOffice Calc):
It looks as if the default waveform is functioning as intended, but it is somehow not affecting the effect parameter correctly unless a waveform is specified by the user.
short tendon
@tulip sleet are you going to bring the updated certificate submodule into circuitpython?
tulip sleet
<@329766224093249548> are you going to bring the updated certificate submodule i...
Yes, I'll do that too after a little more testing. I'll open an issue -- thanks.
short tendon
Yes, I'll do that too after a little more testing. I'll open an issue -- thanks.
Let me know when you have a build. I'll do some tests
onyx hinge
@slender iron I reviewed the state of things further and I think maybe it's almost-fine.
The ssl module actually only seems to check the socket.type against SOCK_STREAM. I can't see where it checks AF_ or uses SOCK_DGRAM. and it does call setsockopt but only with values passed down from a caller.
However, I still don't see a way around making SOCK_STREAM match. For a minute I thought we could effectively getattr() it from the socket object, but it's not a property of the socket object, it's a property of the socket module/socketpool and we don't have an effective way to pluck out that information: There's not a link back from a socket object to its socketpool type that I could see.
I'm not sure what the consequences are of trying to create an SSL session on top of a SOCK_DGRAM besides that it's not expected to work. So we could also just ... remove the check?
slender iron
<@252717193496756235> I reviewed the state of things further and I think maybe i...
we could add those constant to a socket module too
like cpython
manic glacierBOT
There's definitely a bug. Looking closely at harmonic content in spectrogram of the latest audio snippet, the modulation with WAVEFORM=False looks square-like:
and with WAVEFORM=True it looks exactly as it should:
also, the tiny slopes of the square-like transitions suggest that it once was a triangle wave....
tiny peak
tannewt, I figured out how to make a minimal Zephyr board definition for the Feather RP2350. It can do I2C and UART serial with the Zephyr shell sample app. Haven't tried it with anything else yet. Wrote up the board def in a Playground guide if you're curious. Not sure if the same technique will work for RP2350B, but I plan to try as soon as I can get my hands on a board.
mortal kernel
@slender iron Big hug to you and @buoyant lagoon for your assistance with git. I've submitted a new pull request that should pass muster. There's a warning compiling one of the tests (unchanged from the upstream repo) that's causing one of the CI tests to fail, but I think we're good to go...
manic glacierBOT
slender iron
tannewt, I figured out how to make a minimal Zephyr board definition for the Fea...
Thanks! We've started this repo for adafruit supported board defs: https://github.com/adafruit/adafruit-zephyr-support Mind making a PR?
tiny peak
Thanks! We've started this repo for adafruit supported board defs: https://githu...
sure. I may try to test it a little better first. my initial board def is a little rough. just got it working last night.
slender iron
thanks!
tiny peak
thanks!
for that repo, are you using adafruit-zephyr-support as a workspace directory with west init? Or, more specifically, I'm wondering about what the relative path is from the adafruit-zephyr-support/boards directory to the zephyr/boards directory. The way I did my board def is to include the rpi_pico2 dts file, then delete pinout related stuff I don't need, then add the proper pinctrl stuff. So, it depends on the relative path.
tiny peak
I feel like the relative path thing is a little fragile. But, it has the advantage of working without copying Apache-2.0 licensed code fragments into an MIT licensed repo. If you're cool with just using Apache-2.0, then I could make a board def that copied the devicetreee code without the relative path include.
slender iron
ya, I think it is to have both marked with spdx
thorny jay
TLS over UDP... Maybe this is part of HTTP/3 and QUIC.
TCP is so old fashioned.
tiny peak
@slender iron PR for the Feather RP2350 board def is up. I2C and UART serial work. SPI is probably busted. Tried a speculative devicetree config for the spi pins but I haven't tested it yet. Specifying the pins you want is weird and poorly documented. May be possible to use the PL022 driver with any GPIO for the CS line, or maybe not. If that doesn't work, there's also an spi-pio driver.
manic glacierBOT
manic glacierBOT
slender iron
@onyx hinge did you try usb host yet? mine isn't working OOB
onyx hinge
@slender iron no I didn't. you said something about needing to enable hubs.
I ended up over in arduino land working on hstx
slender iron
kk. one should work already I think. I was gonna bump it to 2 so an external one would work too
slender iron
I'm excited about 720x400
is #10028 close? looks like you updated it but it has conflicts
(the auto-configure display PR)
onyx hinge
oops I'll look at it
slender iron
@mortal kernel I tried to change it and it didn't work. I added the function back and am unblocked
mortal kernel
Not what I expected. What went wrong?
slender iron
don't remember. that memory was overwritten by usb debugging
mortal kernel
Hah! I know how that goes. As long as CIRCUITPY_CONSOLE_UART_RX/TX are set it should have worked. I'll clone a new repo and see what's up.
Which port/board are you targeting?
slender iron
@mortal kernel fruit_jam on the rp2350. trying to print out tinyusb host debugging
mortal kernel
Thanks! Hope you're enjoying your jam!
onyx hinge
@slender iron using the metro uf2 my i2c scans ... I assume that's the i2s amp's i2c personality. Yesterday it detected a monitor at 0x80 0x50 when it was plugged in. ```Adafruit CircuitPython 9.2.3-27-g8be1681493-dirty on 2025-02-11; Adafruit Metro RP2350 with rp2350b
import board
i = board.I2C()
i.try_lock()
True
i.scan()
[24]
(this is on the fruit jam proto)turns out they were the same
manic glacierBOT
mortal kernel
@slender iron I tried to look into the serial debug problem, but with CIRCUITPY_DEBUG_TUSB set the build fails because __ZEPHYR__ is undefined.
slender iron
<@252717193496756235> I tried to look into the serial debug problem, but with `C...
Change it to ifdef
mortal kernel
@slender iron Did that. The problem is that TinyUSB assumes it won't have a prototype for whatever CFG_TUSB_DEBUG_PRINTF is defined to, so it tires to make one in tusb_debug.h.
Since we're now sending in a macro, that doesn't go well. This method might be used elsewhere, so I'll submit a PR to add the function console_uart_printf back in.
lone axle
with usb.core.Device.read https://docs.circuitpython.org/en/latest/shared-bindings/usb/core/index.html#usb.core.Device.read is it "allowed" for user code to use a buffer that is smaller than the amount of data sent by the USB device? And if that is allowed, is it expected that the count returned could be larger than 8 which is the size hardcoded for the buffer?
I have this basic script to print incoming data, and I'm finding that count is 27 even though my buffer size is 8
device = None
while device is None:
for d in usb.core.find(find_all=True):
device = d
time.sleep(0.1)
device.set_configuration()
buf = array.array("B", [0] * 8)
# ...
while True:
try:
count = device.read(0x81, buf)
print(count)
except usb.core.USBTimeoutError:
continue
And I'm also getting Hard fault: memory access or instruction error. somewhat regularly, albeit changing seemingly randomly depending on the rest of the code.
So I am hypothesizing that device.read() is reading 27 bytes, but only 8 of them goto the buffer and the rest clobber some unlucky bits in memory.
mortal kernel
@slender iron Submitted the PR for console_uart_printf. I ran some tests with TinyUSB debugging activated. Due to the volume of messages and the slowdown induced by logging to the serial UART, it looks like an ideal candidate for RamLog. Let me know if you've got any interest in trying it out. I could use some feedback. I'll be AWK most of today.
onyx hinge
@slender iron I thinkhttps://github.com/adafruit/circuitpython/pull/10028 is ready to go in if you can take a quick look
and #10040 if you feel like it, but that one's not blocking anything
slender iron
<@252717193496756235> I thinkhttps://github.com/adafruit/circuitpython/pull/1002...
ok, will look now
just getting to the office
I suspect my i2c problem was my hdmi cable
onyx hinge
@slender iron just checking, do you want me to fully revert the board.DISPLAY part of the changes?
slender iron
I didn't understand that it'd be available on all boards then
onyx hinge
The first time you reviewed it, it wasn't.
slender iron
we don't have any defines that say if a board has a display do we
since board_init sets it up
onyx hinge
nope, we don't.
slender iron
Maybe it'd be better to add one that has the display type? so we can factor out the auto-init in the future?
onyx hinge
What specifically would you like me to do for right now, so that board.DISPLAY is available on these rp2350 boards that may or may not initialize a HDMI display at boot time? I think getting this PR unblocked is more important than realizing an ideal long range plan right now.
(I'm a bit sad, I felt like using this new technique made board.DISPLAY work strictly better than before, with very little code. Because you could for instance configure a display that you've permanently attached inside of boot.py and use it in code.py as though it was always part of the board. It's true that it's a bit "impure" since strictly speaking it makes displays that are not "part of the board" appear as board.DISPLAY too, but then by that logic NO DVI display could appear as board.DISPLAY because it's on the wrong side of a connector and is strictly optional)
slender iron
I don't care that it is board.DISPLAY on rp2350. I mainly want the console to auto appear. Maybe add supervisor.runtime.display now?
and make it work like your board.DISPLAY does now
slender iron
the goal of that work is to not need release_displays anymore
only the supervisor's display lives longer
and can be referenced across VMs
other displays could be init but would be stopped at the end of code execution
slender iron
@onyx hinge is that ok or am I being too much of a purist?
having to change libraries for the new behavior was a red flag for me
onyx hinge
@slender iron It's fine.
I think the existing library code was kinda wrong already, because it didn't check whether board.DISPLAY had been released via displayio.release_displays() either. As I'm sure some bug or other has discussed, in that case board.DISPLAY becomes a none-like object that is not None....
onyx hinge
@slender iron in your ideal conception, would the first created display become supervisor.runtime.display or does it require explicit action? Would auto dvi on rp2350 be an exception? With the "easy to do right now" code, the first created display would always become supervisor.runtime.display
slender iron
my intent is that it does require explicit action for user inited displays
instead of assuming that's where folks want the console
so board.DISPLAYS and auto dvi would be auto-init and set
(because CP does it)
explicit action means that release_displays isn't needed in examples
manic glacierBOT
onyx hinge
@slender iron https://github.com/adafruit/circuitpython/pull/10028 updated again.
supervisor.runtime.displaywill be set to a display set by Cboard_init()but not automatically set to a display set by a python script- but the property is settable, so boot or code files can init a display and then assign it to
supervisor.runtime.displayfor use in future runs - the stage is set for
reset_displaysto only reset other displays than thesupervisor.runtime.displaybut this isn't turned on yet.
Hopefully it builds OK.
If this is still not passing muster maybe we should get on a voice or video call to iron it out, because this is somewhat blocking both of us right now and I'd like to get it sorted.
onyx hinge
@slender iron once again the history there has become a mess. but it might be better to merge the messy history than have me spend time rebasing it. If you don't feel strongly.
slender iron
either way is fine with me
onyx hinge
@slender iron let me see how it looks after rebasing
onyx hinge
@slender iron the rebased version is now at https://github.com/adafruit/circuitpython/pull/10062 so that old reviews are still useful.
onyx hinge
CI still waiting to build #10062 😢
onyx hinge
@slender iron it might be interesting to set the max display count to 2 on the fruit jam, since it has both a HSTX connector and an EyeSpi connector.
wraith crow
I'm getting an error when circup attempts to get the 9.x bundle:
There was a problem downloading that platform bundle. Skipping and using existing download if available.
Not really a critical issue, as I assume I can just go out and manually get the files I need....
devout jolt
I'm getting an error when circup attempts to get the 9.x bundle:
There was a pro...
with --verbose, it looks like circup is trying to get the 8.x bundles and getting a 404.
specifically:
02/13/2025 21:58:38 WARNING: Unable to connect to https://github.com/adafruit/Adafruit_CircuitPython_Bundle/releases/download/20250214/adafruit-circuitpython-bundle-8.x-mpy-20250214.zip
There was a problem downloading that platform bundle. Skipping and using existing download if available.
02/13/2025 21:58:38 ERROR: 404 Client Error: Not Found for url: https://github.com/adafruit/Adafruit_CircuitPython_Bundle/releases/download/20250214/adafruit-circuitpython-bundle-8.x-mpy-20250214.zip
Traceback (most recent call last):
File "/opt/homebrew/lib/python3.13/site-packages/circup/command_utils.py", line 137, in ensure_latest_bundle
Ah, I think there was talk of discontinuing the 8.x bundles so it's probably related....
manic glacierBOT
jaunty juniper
Ah, I think there was talk of discontinuing the 8.x bundles so it's probably rel...
and there hasn't been a release of circup yet, the change was made yesterday (in circup)
lone axle
Ah, I think there was talk of discontinuing the 8.x bundles so it's probably rel...
I've made a new release for circup just now that should resolve that issue. Thanks for mentioning it here, and thanks to Neradoc and todbot for checking into it.
zinc walrus
I've been excitedly watching the CP Zephyr port advance, and I'm eager to test what's working on my nRF53 Audio DK, but not sure how. @slender iron previously mentioned adding support, but where do I look?
Do I need to build the entire project in NCS/Zephyr and flash it that way, or is there a prebuilt .hex available that I can flash via J-Link? Any suggestions?
slender iron
I've been excitedly watching the CP Zephyr port advance, and I'm eager to test w...
I can build you a hex
zinc walrus
I can build you a hex
I don't want to distract you from FruitJam!
zinc walrus
Well, if you need a momentary distraction, I'd love to experiment. Let me know...
slender iron
building it now while I eat breakfast. no promises it works
zinc walrus
building it now while I eat breakfast. no promises it works
Thanks very much. Looking forward to this afternoon's Deep Dive.
slender iron
for audio dk
zinc walrus
for audio dk
Awesome. TYVM.
slender iron
doesn't look like it has any external flash
slender iron
(so no circuitpy drive. just repl)
zinc walrus
(so no circuitpy drive. just repl)
Understood.
I'm still working my way up to dev, but maybe the SD card is a possibility.
slender iron
west build --sysbuild -b nrf5340_audio_dk/nrf5340/cpuapp
have you done zephyr before?
zinc walrus
`west build --sysbuild -b nrf5340_audio_dk/nrf5340/cpuapp`
Yes, though by force as it's the only way to program the ADK.
slender iron
You may know better than me how to change the internal flash partition. If you add a circuitpy_partition then it should use it.
zinc walrus
You may know better than me how to change the internal flash partition. If you a...
I'm still intermediate, but looking forward to playing with this all weekend!
slender iron
great! it should be familiar to you if you've done zephyr
do west init within ports/zephyr-cp and then west update
zinc walrus
great! it should be familiar to you if you've done zephyr
Got it. Thanks again for your help, and especially for CP, which is a joy.
slender iron
np!
manic glacierBOT
short tendon
Is it possible for circup to error gracefully if it can't find a bundle and say it's using the last one of got? I can imagine people struggling to find out why it's not working and not automatically upgrading...
onyx hinge
it's also worth considering that different bundles might support different versions rather than all changing on the same day...
manic glacierBOT
Do you think you could give this logic-analyzer a try: https://github.com/gusmanb/logicanalyzer
It is very simple to use, you basically need a Pico for it (the logic-analyzer is on top, the DUT with the AHT20 below):
You will see something like this in the analysis software:
As you can see, this is with version...
manic glacierBOT
manic glacierBOT
onyx hinge
heya <@&356864093652516868> just a quick reminder that due to a US holiday the meeting is on TUESDAY this week. See you tomorrow! As always you can add your notes in advance to the notes doc: https://docs.google.com/document/d/1oM_x1t2iD_iStf4riPa5j_naS-yqZAR-Y8_qKyVpa7s/edit?usp=sharing
onyx hinge
by now cooper knows how much of an amateur I am at coding 🤣 🙊
manic glacierBOT
fleet hollow
by now cooper knows how much of an amateur I am at coding 🤣 🙊
Hey dude, we all miss stuff here and there. And I am by no means an expert!
lone axle
Would it be possible / desirable to have a board.SD_SPI() on something like the Metro RP2350 where it has a built-in SDCard reader which is connected to one of the two available SPI busses? It has board.SPI() for the other available bus, and it has the pins which can be used with busio, but it could be convenient to have a shorthand for the SDCard bus perhaps.
candid sun
personally i think that would be really useful
stuck elbow
The tutorials for how to use the sd card still couldn't use it, because not all boards have it, and you don't want to share the spi connected to the sd card with other devices, because the card doesn't start in spi mode and would interfere with communication, so making it a singleton doesn't help in this particular case much. So the only benefit would be not specifying the pins explicitly. But explicit is better than implicit...
slender iron
Would it be possible / desirable to have a `board.SD_SPI()` on something like th...
That's fine with me! My hope is to add SD card automounting though
That may remove the need for it at all
onyx hinge
howdy folks and especially <@&356864093652516868>, I'll step on host @lone axle's toes a tiny bit to announce that our weekly meeting is happening in about 70 minutes in the CircuitPython voice channel, meaning it's a good time to add your notes if you haven't already. The google doc is at https://docs.google.com/document/d/1oM_x1t2iD_iStf4riPa5j_naS-yqZAR-Y8_qKyVpa7s/edit?usp=sharing and I'm going to go add mine right now!
tulip sleet
@lone axle if you select the pasted in report stuff in the notes doc, copy it, and then right-click "Paste from markdown" on top of the selection, it will reformat as styled Google Docs text. Then, when you "Download as Markdown", it will get converted back to Markdown.
slender iron
Paste from markdown does require a chrome extension to work I believe
tulip sleet
It's built in to Google docs now, but you might have to enable markdown support.
"Paste from Markdown" is also in the Edit menu
@lone axle I am especially thinking of the bullet lists
lone axle
I do see it in my context menu after enabling the checkbox in preferences. Thank you
🙁
slender iron
I switched to chrome to do it
lone axle
they allow you to enable it, but not actually use it.
onyx hinge
oh google, you've come so far from the scrappy little company that wanted to do no evil. </sarcasm>
lone axle
Sorry to gank the last ones, chrome works for me as well. I'll try to remember to copy them that way next time.
tulip sleet
and I also set the headings to the right levels with the paragraph style menu in the toolbar at the top
onyx hinge
@lone axle I can read core or blinka if you like. just let me know.
lone axle
<@382939733107408897> I can read core or blinka if you like. just let me know.
Thank you, I'll pass to you for Blinka.
lone axle
@slender iron will you be around able to read for the core section?
slender iron
yup
onyx hinge
the link is https://adafruit-playground.com/playground/pages.atom but I may have spoken too soon, my feed reader (newsblur) doesn't seem to like the content there
I sent an email internally to the team in charge of playground so with any luck that'll get resolved soon.
720x400 is an aspect ratio of 1.8 (9/5), vs 1.7778 (16/9), so just a few percent un-square
dvhstx has a dedicated text mode with "RGB111" color (foreground colors only, always on black background, I think?). It'd be much higher performance than doing everything via bitmaps.... but it wouldn't "be displayio".
[I was writing in the notes about the history of why CAMERA is disabled on a bunch of espressif 4MB boards in the notes doc...]
slender iron
thanks all! 🦷 dentist time for me
onyx hinge
thanks Tim!
lone axle
Thanks everyone. Have a great week!
solar whale
Thanks!
thorny jay
This is the most usefull and frustrating feature of Adafruit shop... now I know I have one, but I need to find it:
stuck elbow
This is the most usefull and frustrating feature of Adafruit shop... now I know ...
hold an empty box in your hand, pretend you just received it, and go put it away — it will be there
thorny jay
hold an empty box in your hand, pretend you just received it, and go put it away...
I am checking the other item in the same order... my expectation is that once I see one of those, the other should not be too far. 😉
stuck elbow
I've been searching for a bag of connectors I'm sure I have for the last three months, yesterday I gave up and reordered them, I expect to find them any minute now
lone axle
Here is the notes document for next Monday’s CircuitPython Weekly Meeting. It is at the normal time of 11am Pacific / 2pm US Eastern here on Discord. Add your hug reports and status updates to the document before the meeting. If you are unable to attend but would still like to contribute, feel free to add your notes and we’ll read them off during the meeting. Hope to see you there! <@&356864093652516868> https://docs.google.com/document/d/1nJs-ZrRGrey76S8DlqTuS4wzoIXLvEyXdewi0lbNx3o/edit?usp=sharing
tiny peak
Is there a good place to look to understand, for a given port and board, how specifically are the bootloader, CircuitPython firmware image, NVM, and CIRCUITPY drive assigned to addresses in flash. Like, what goes where in internal flash and external flash (not where things get loaded into RAM)? I see that ports/$FOO/mpconfigport.h has stuff like CIRCUITPY_FIRMWARE_SIZE, CIRCUITPY_INTERNAL_NVM_SIZE, and CIRCUITPY_CIRCUITPY_DRIVE_START_ADDR which look relevant. I also see address space layout info in .ld files, but I'm not sure how that relates to addresses in RAM vs. where things get stored in flash.
Context for my question is, to define flash partition layouts for Zephyr board definitions, it would be nice to avoid stomping on the CircuitPython bootloader, CIRCUITPY drive contents, etc.
short tendon
Is there a current plan in motion to kill all the secrets in guides and libs and examples?
If not and wanted, I can start down this path...
slender iron
Is there a current plan in motion to kill all the secrets in guides and libs and...
I don't know of one
slender iron
Is there a good place to look to understand, for a given port and board, how spe...
It varies port to port unfortunately. Very related to @devout jolt's question on issue #10074
circuitpython doesn't vary the CIRCUITPY location build to build except some major version changes. we want to preserve the CIRCUITPY contents on upgrades.
tiny peak
so maybe it would make sense to just try matching whatever address range is in the firmware .uf2 file for a given port/board?
slender iron
the linker scripts are probably the "ground truth" since linking fails when the firmware is too large
slender iron
so maybe it would make sense to just try matching whatever address range is in t...
I don't think that'd be reliable because the firmware may be much smaller than the space we've set aside
tiny peak
oh, right, I follow
hmm... guess I'm gonna have to spend some more time reading code and pondering this
thanks
tiny peak
My current quest is writing guides on display/sensors/CLUE-nrf52840-BLE
I was using the Feather RP2350 as an intro-to-writing-a-board-def practice
but now my order showed up with the CLUE board, so I'd been thinking of switching more of my attention to that
tiny peak
yeah. lots of partition excitement there
Also, if you see the potential for me to plug into something with rp2 in a way that would help things along for you, and pt likes it, that works too
slender iron
nrf uses the C preprocessor to get info about the size of things unfortunately
nrf is more interesting than rp2 I think
for my zephyr stuff because the newer nordic boards will be some of the first on zephyr
we don't need to be partition compatible for those though
tiny peak
yeah, I'm expecting it to be tricky, but should make for some good writeups
slender iron
(we don't need to be partition compatible because they are new)