Edit directly would be fine. If I say it's completed, I'm already done and out minus any copying and pasting into new pages I'm doing. So I'll expect you to be editing or take over if I am.
#circuitpython-dev
idle owl
slender iron
any thoughts about removing all SAMD21 boards from FULL_BUILD and explicitly enabling modules for it? The msgpack PR is using FULL_BUILD and running out of space on 10 boards.
idle owl
I think it's time to do whatever we need to do to make it possible to add new features with regards to the SAMD21 boards.
@slender iron I don't entirely understand what FULL_BUILD means, but I'm in support of the idea of explicitly enabling modules for those boards. We're doing that already inconsistently, aren't we?
slender iron
we basically do it for non-express boards already
I like the idea of having the set of modules be consistent
but I'd like to freeze that set for SAMD21 essentially
idle owl
I'm fairly sure I support this entirely. I guess I'd want to know what set we go with before going totally in on it, but I think something like that is worth it for sure.
slender iron
I'm thinking whatever it supports now essentially.
idle owl
Ah so a feature freeze basically.
I suppose worst case, if something comes along that seems crucial, we can swap things around? I don't know what that might be, but I imagine it's possible something could come up.
idle owl
Alright. Yeah I support this.
slender iron
but I don't want new modules included by default
idle owl
Consistency is easier to support anyway.
slender iron
💯
tulip sleet
@slender iron I'll be afk for a while; will finish release later tonight (It just finished building and uploading), with blog post, etc.
slender iron
thanks! I'll read over the release notes soon
orchid basinBOT
ionic elk
@slender iron do you have a simple ESP32-S2 application you could recommend to just verify my internet connection is good? TCP-IP would be fine
ionic elk
perfect ty
lone sandalBOT
ionic elk
@idle owl hi are u around?
idle owl
Indeed!
ionic elk
running into an issue with the Metro S2 tutorial, AttributeError: 'module' object has no attribute 'Session', any ideas?
changed version?
idle owl
That would do it, if that's the case 😄
ionic elk
ok, cool
idle owl
@ionic elk Let me know if that's not it.
@onyx hinge For the SD Cards page in the board selector guide.... "SD Card Enabled" doesn't sit right for the page title. Do you have any suggestions on how to word it?
Or is that good?
idle owl
That's what I thought
onyx hinge
Is it really a list of boards with an SD Card Slot available?
ionic elk
@idle owl yeah my bundle was out of date, thanks!
idle owl
@onyx hinge Yeah we discussed it in the meeting yesterday.
@ionic elk Excellent!
Well it's the SD Card built in boards, and the adalogger FeatherWing.
onyx hinge
we also discussed that it was really "about" datalogging, so listing the pygamer wasn't most sensible
idle owl
She said put them at the bottom.
onyx hinge
right
idle owl
It's viable, just unlikely to be used as such.
Built In SD Card Slot is kind of long. and Built In SD Card is misleading.
onyx hinge
Spacious Storage?
onyx hinge
(subcategories: datalogging / big data files like mp3s or bitmaps)
Expandable Storage ?
onyx hinge
I like Capable better than "Enabled"
idle owl
Yeah agreed
idle owl
@onyx hinge They're all micro SD right?
idle owl
I mean they must be.... But now I'm second guessing
onyx hinge
grand central is, pygamer is, adalogger feather is ..
onyx hinge
there's also this option for adding a (micro) SD https://learn.adafruit.com/adafruit-microsd-spi-sdio
idle owl
Oh hmm yeah.
I think that's not included though because the Adalogger FeatherWing is plug-and-play on a Feather, that has to be wired to something
onyx hinge
that's fine
this is more beginner-oriented and getting those connections right can be a pain
idle owl
yeah
thorny jay
My MagTag just arrived and is working (maybe they should come with CircuitPython and UF2 preloaded). [I say working, because unfortunatly my FeatherS2 came DoA an the fix might take another 5 weeks to reach me.] Back to MagTag, @slender iron always say that we should use 6.1.0 with all ESP32S2 but when I go on S3, I only see bin and uf2 with a date and hash... I can I tell if it's a 6.0 or 6.1? https://adafruit-circuit-python.s3.amazonaws.com/index.html?prefix=bin/adafruit_magtag_2.9_grayscale/en_US/
idle owl
@thorny jay 6.1 hasn't been released yet. Use the most recent .bin file.
thorny jay
@thorny jay 6.1 hasn't been released yet. Use the most recent .bin file.
@idle owl Or maybe I need to check in the artefact of the last commit from the 6.1.x branch?
idle owl
Possibly? I'm not sure what builds beyond what ends up on S3.
thorny jay
This will be solved as soon as 6.0 is publish as 6.1 will be the bleeding edge and directly available (I assume).
idle owl
Correct.
@onyx hinge If the SD card slot is connected via SPI SERCOM, SDIO won't work. But if it's connected via SDIO (as in the STM Feather), does SPI work? No reason to include this, so I'm not even sure why I'm asking.
Also is this still valid in CircuitPython: "the SDIO SD card is not yet supported natively" ?
onyx hinge
@idle owl on the stm feather you can use the micro sd card and the spi bus simultaneously
idle owl
Ah ok
I meant does SPI work for accessing the SD Card
like on the Grand Central
or is it totally a separate thing
onyx hinge
no, you have to use sdioio, not sdcardio or adafruit_sdcard
idle owl
right on
onyx hinge
the module you use has to match the kind of bus it's attached to
idle owl
Got it
onyx hinge
(you might be able to use a bitbangio spi bus with adafruit_sdcard but it's a terrrrrible idea)
idle owl
Fair enough
onyx hinge
(so don't mention it)
tries to remember if he tested it during that guide I think it doesn't work, but can't remember the details
idle owl
@thorny jay 6.0.0 is stable. We simply haven't gotten 6.1.0-alpha/beta.0 out yet.
@onyx hinge Also is this still valid in CircuitPython: "the SDIO SD card is not yet supported natively" ? (or did you answer me and I missed it)
idle owl
Product copy.
Towards the bottom
@tidal kiln Quick question about your edits if you're around
tidal kiln
yep
idle owl
Did you remove "Adafruit" from the product names in the paragraph below the product element? Or did I inconsistently forget to include it. I didn't want to trounce your edits if that was you.
tidal kiln
you mean like "Adafruit Itsy" vs. just "Itsy"?
tidal kiln
i don't think i did. and if so wasn't intentional.
tidal kiln
sure. np.
idle owl
@onyx hinge Says you're editing a page in my guide. Can I take over?
onyx hinge
yeah oops
idle owl
no worries 🙂
onyx hinge
just looking
prime cove
I'm trying to build 6.0.0 for the grand central but I'm getting this error ```arm-none-eabi-gcc: error: unrecognized command line option '-Wimplicit-fallthrough=2'; did you mean '-Wno-fallthrough'?
../../py/mkrules.mk:55: recipe for target 'build-grandcentral_m4_express/py/mpstate.o' failed
make: *** [build-grandcentral_m4_express/py/mpstate.o] Error 1
@golden jay what is arm-none-eabi-gcc --version? You may be using an older compiler. You need 9.x
prime cove
I ran through the update procedure here with no luck: https://learn.adafruit.com/building-circuitpython/build-circuitpython
@tulip sleet arm-none-eabi-gcc (15:6.3.1+svn253039-1build1) 6.3.1 20170620
So does that mean V6.3.1?
tulip sleet
that is really old. did you download the latest version as explained in the guide?
you can't use the version that comes with your version of linux
you download and unpack the version from the ARM website, and then you need to add it to your PATH, before the system-installed version
prime cove
Ah that would explain why updating the package didn't do anything then. Thanks, I'll make sure I go to the ARM site.
hearty tapir
That is the case for almost every distribution. I also fell into that trap, adam 🙂
bronze shadow
Also watch out on Arch based distros. Their version is too new and can cause other issues.
prime cove
Good to know! I'm just running ubuntu on WSL right now, just still picking up details like these
idle owl
@tulip sleet I know you're in sleep land but do you have a few minutes to help me with this guide page? It's the Chipsets page in the Which CP Board is Right for You? guide, the page that explains why one mpu is better/different from another. I copied down everything Limor said in the meeting yesterday, but it is super sparse and I'm not sure how to fill it out more.
ionic elk
@tulip sleet @onyx hinge do you know if we turned off some aspect of the ESP-IDF native reporting to reduce the messaging over the UART bridge? I'm noticing that I don't see ESP_LOGE messages anymore. Is this intentional?
idle owl
@ionic elk Oh hmm. You can probably help. This is a guide page explaining, essentially, the pros and cons of each MPU used on CP boards. We evidently forgot STM32F405, so I don't have any notes on it. What would you say, in the context of the other microprocessors we use, are mostly the pros, and maybe a con to that proc?
ionic elk
@idle owl Oh yeah, would love to see the full guide page when you have a draft! In my opinion, the STM32 has the advantages of being a fast all-rounder board that's great for people who want to take extra steps with their project/learning process - it is much easier and cheaper to connect a debugger to a ST board compared to an Atmel one (a STLink probe is only $2 compared to a $400 Jlink), and the peripherals and memory are more powerful than other current feathers. ST F4 boards can also be programmed without an external programmer at all using DFU, and are very hard to brick.
idle owl
Thank you for helping out!
ionic elk
But I would say a lot of the biggest pros of ST might be for people who use the Feather as a starting point and reference design for their own PCBs. There are a LOT of ST chips of widely varying price and performance, and they all share a lot of code, so it's pretty easy to move between them. So you can start out with a powerful overkill chip to try things out without memory pressure, and then optimize down to something that's super cheap when you want to make a product, and you might not even need to change your PCB
idle owl
This guide is more a beginner thing so I think your first point is far more relevant, but that is good to know.
ionic elk
Yeah it's just a note for people who are trying to learn for the future - everything from the H743, which can run the original Doom, down to the F103s, which can cost under a dollar, all share the same codebase and most of the API.
idle owl
It still requires some extra steps to install CircuitPython, right?
ionic elk
So it's a nice platform to learn if you want to do lots of different kinds of projects
idle owl
As in it's not a UF2 bootloader
ionic elk
At the moment. We do have a UF2 bootloader for it though, I'm not sure why they still aren't shipping with it.
idle owl
Oh hmm ok.
User installable UF2 bootloader?
We're probably not shipping it because of existing stock, and/or it's difficult to update the testers that flash the boards in the first place.
There's always a lag on updating what ships.
ionic elk
With DFU you gotta use brew install to get a thing, connect a wire, and then run a command. So it's a little more technical in that beginners might have to use the terminal, which can be intimidating
If you're on windows there's a downloadable GUI utility, but you have to give ST your email to get it, which is annoying and silly
You can install the bootloader yourself, sure, though you'd have to get it off github, build it, etc. So better for developers and stuff.
idle owl
Oi yeah.
ionic elk
And at that point you're probably fine just using an STLinkV2 off amazon, or the DFU. STLinks are this $2 debug dongle and they're really fast and easy for development. But again, better for devs.
idle owl
I remember buying them for the STEMMA connector and then getting outright halted when I Realised there was a whole thing to install CP.
ionic elk
or people who want to become devs, which is what I'd emphasize
idle owl
Ah fair enough
I think I have enough! Thank you so much!
I'm struggling with this page 😄
Does it also have more memory than the nRF? I feel like the answer is yes
ionic elk
Good luck! These are just my opinions that kind of messaging is def more up to you and Ladyada and stuff I'd figure. I'm looking forward to seeing what you write though, I've wanted to make a guide like that for a long time on my personal site
idle owl
Thanks!
ionic elk
I'm pretty sure it does yeah
ionic elk
double checking
Yeah it has about 4x as much. 1MB vs 256. Or 2x depending on how you measure the NRF I guess, @tulip sleet would know more
idle owl
Ahhh. I was reading the thing wrong. Yeah.
Not sure which number to go with.
I'll wait until Dan's around to tweak it any more than I already have.
ionic elk
Again, if you aren't bound to the feather, the ST chips have ridiculous variability on that front, anything from like 128k to 2MB
192k of RAM as well which I also think is the highest offered
idle owl
In this case, it's explaining the different MPUs separately from their boards, but it's in the context of the boards, so it kind of means in the context of the Feather for now I suppose.
ionic elk
until we get the i.MX feather in I think it's the highest performance board in basically every category, with the exception of wireless
idle owl
Huh. Fair enough! 🙂
ionic elk
@idle owl note I'm probably biased because I deal with a lot of people who say they want to "upgrade" their skillset at my local makerspace, so I tend to slant everything I work on with that kind of language. How to move past dev boards into your own stuff, etc. Totally up to you whether it's appropriate for this case or not.
idle owl
That makes sense. Thank you so much for your help!
onyx hinge
@ionic elk I am not aware of any deliberate change, certainly not on my part. Did you forget to build with DEBUG=1 ? no recent changes in ports/esp32s2/esp-idf-config/ that look related.
ionic elk
I'm building with debug on now, we didn't used to need that but maybe we only message with it on now
ionic elk
@idle owl I dunno if this will be useful but I found it to be a fun read https://jaycarlson.net/microcontrollers/
idle owl
Cheers, I'll take a look!
orchid basinBOT
ionic elk
kind of technical but there are some little contextual notes at the start of each entry which are nice
idle owl
Nice
ionic elk
On revisiting it, I should maybe take back that it's a "fun" read. There are interesting parts of it which are less technical and more informative from a broader industry kind of viewpoint. But I'd definitely just skim for the good bits, there's a lot of technical jargon that doesn't apply to feather tier boards
idle owl
I was skimming 😄
slender iron
My MagTag just arrived and is working (maybe they should come with CircuitPython and UF2 preloaded). [I say working, because unfortunatly my FeatherS2 came DoA an the fix might take another 5 weeks to reach me.] Back to MagTag, @slender iron always say that we should use 6.1.0 with all ESP32S2 but when I go on S3, I only see bin and uf2 with a date and hash... I can I tell if it's a 6.0 or 6.1? https://adafruit-circuit-python.s3.amazonaws.com/index.html?prefix=bin/adafruit_magtag_2.9_grayscale/en_US/
@thorny jay It will come with UF2 and circuitpython eventually. It's a chicken and the egg problem. We don't want to ship it until we've had enough testing and we can't have enough testing until we have hardware available for people to test with. You are an early adopter. 🙂
@idle owl Or maybe I need to check in the artefact of the last commit from the 6.1.x branch?
@thorny jay Most recent builds will be from the main branch and will be 6.1.x. I'll do a release of 6.1 today so the tags are updated again.
thorny jay
Right now, circuitpython.org/download is a bit confused. For some board it show 6.0.0 as current and 6.0.0-rc2 as bleding edge. Where for other like MagTag it is only go to S3.
But'll survive.
slender iron
Also watch out on Arch based distros. Their version is too new and can cause other issues.
@bronze shadow @prime cove I develop on Arch so you shouldn't see any issues, just more free flash space than the older GCC version gives.
@thorny jay that seems correct to me. it'll make more sense when we have a 6.1 pre-release
idle owl
@slender iron Dan's evidently out - and I know he's busy anyway. Do you have a few minutes for a chat about the Chipsets page in the Which CP Board is Right for You? guide, the page that explains why one MPU is better/different from another? I copied down everything Limor said in the meeting yesterday, but it is super sparse and I'm not sure how to fill it out more. I also got help from Lucian on STM.
slender iron
yup!
ionic elk
@slender iron do you know about our ESP builds not logging over UART anymore?
ionic elk
Ah no Jeff was right, we just moved all messaging to debug
slender iron
@slender iron do you know about our ESP builds not logging over UART anymore?
@ionic elk It should be enabled if you build with DEBUG=1
bronze shadow
@bronze shadow @prime cove I develop on Arch so you shouldn't see any issues, just more free flash space than the older GCC version gives.
@slender iron I had lockups when I froze in adafruit_ble using Arch. It only affected some functions, but building with the same code on Ubuntu with the recomended gross compiler
idle owl
@tulip sleet Scott helped me out, FYI! Hope everything's going well with sleep 🙂
tulip sleet
it's ok 🙂 I have a lot to do, but a lot has already been done, also
idle owl
Fair enough. Progress is good regardless.
ionic elk
@slender iron what kinds of errors do you want for failed UDP transactions? I see some OS error codes and mp_raise_BrokenPipeError for TCP but that isn't appropriate for UDP, right?
slender iron
@ionic elk try and determine what CPython does. It's usually OSError with error codes from the underlying socket implementation
ionic elk
Ok, I'll check out the CPython errors
Hmm, cpython just calls Py_XDECREF(addr) and then returns NULL. I'm not sure what that means in practice, since sock_recvfrom_into isn't referenced elsewhere
@slender iron do we return NULL for other calls as an error? I don't see examples of us doing that, at least not the way CPython does.
ionic elk
nah
onyx hinge
@ionic elk I did a little slumming in Python (2.7 because it's what I had handy but I bet current is similar) and the error setting code boils down to: return PyErr_SetFromErrno(socket_error); -- it's a bit indirect so but the effect is to use the underlying error code translated to Python
hearty tapir
Congrats to the team on the CPY 6.0.0 release! 🙂
idle owl
@slender iron It's already beta-level?
I'm not up on what's going into it, so I'm unaware.
slender iron
I think so. folks are using it and we won't change too much
we may stay in beta for a bit though
idle owl
Fair enough. No objections here then.
hearty tapir
I look forward to have some time to try MQTT, sometime soon.
hearty tapir
@slender iron we could consider to update the cacert.pem in ~/circuitpython/ports/esp32s2/esp-idf/components/mbedtls/esp_crt_bundle/ to the current Mozilla bundle.
ionic elk
@onyx hinge interesting, where do I call that in practice?
slender iron
@hearty tapir I have some changes already to switch to a bundle that matches the nina-fw
that was my plan
ionic elk
I'll add you as a reviewer hang on
slender iron
any idea why the CI isn't going?
onyx hinge
@ionic elk I think the closest CP equivalent is mp_raise_OSError(int errno_);
slender iron
welp, now I see new builds
ionic elk
@onyx hinge Yeah I've seen that used elsewhere in the shared-bindings Socket.c, but I wasn't sure what in mperrno to use for a datagram send failure, so I was trying to figure out what aligned properly to CPython
slender iron
@ionic elk the linux man page can be really helpful. see https://linux.die.net/man/2/sendto
the errors section in particular
ionic elk
@slender iron ok, yeah that seems pretty thorough
onyx hinge
on a standard desktop system, recvfrom() returns a negative value for failure and sets the global errno value to a common value like -ENOSMURFS if there were insufficient smurfs available
I'm having trouble finding out what lwip_recvfrom does, it returns a negative number in case of error but beyond that ... ?
ionic elk
Wait, I guess I'm still missing something. On error, -1 is returned, and errno is set appropriately. so, what is the appropriate value for errno?
orchid basinBOT
I'd rather not merge as-is because of the duplicated images. Here is the homepage:
ionic elk
am I misunderstanding what errno is?
onyx hinge
esp-idf's code is written like there's a global errno value, shivers examples/protocols/sockets/udp_server/main/udp_server.c
// Error occurred during receiving
if (len < 0) {
ESP_LOGE(TAG, "recvfrom failed: errno %d", errno);
errno is a global (or thread-local) variable whose value is set by certain functions in the standard library. Besides "pointers can be NULL", "errors are signaled through a single, global number" is probably the 2nd worst mistake of C/Unix.
ionic elk
Lol yeah I had just found that same article and was reading it
anyway. my PR is up, I just put in some broken pipe errors which are probably wrong
I see mp_raise_OSError(0); is called in other Socket functions, I'm not 100% sure what that does?
onyx hinge
https://github.com/adafruit/circuitpython/issues/3686 it leads to this issue
(lwip source)``` err = lwip_recvfrom_udp_raw(sock, flags, &msg, &datagram_len, s);
if (err != ERR_OK) {
LWIP_DEBUGF(SOCKETS_DEBUG, ("lwip_recvfromUDP/RAW: buf == NULL, error is "%s"!\n",
s, lwip_strerr(err)));
sock_set_errno(sock, err_to_errno(err));
done_socket(sock);
return -1;
}
so it really is setting a global errno value, how quaint
slender iron
consider OSError(0) to mean I meant to figure out the appropriate error number
onyx hinge
https://github.com/adafruit/circuitpython/blob/main/extmod/modlwip.c#L213 lower level lwip APIs like lwip_recvfrom_udp_raw return ERR_xxx values and in the micropython lwip bindings this is how they're translated to CircuitPython MP_Eyyy values for OSError raising
0, /* ERR_OK 0 No error, everything OK */
MP_ENOMEM, /* ERR_MEM -1 Out of memory error */
MP_ENOBUFS, /* ERR_BUF -2 Buffer error */```
onyx hinge
but at the level you're working (lwip_recvfrom) there's already been a translation from those errors (returned as negative small integers) to newlib errno numbers and stored in a global errno value
onyx hinge
@ionic elk I expect this isn't particularly helpful to you, I'm just boggling about the layers upon layers of error handling for something "so simple", and yet it ends up difficult just to show the right error to the user/developer.
ionic elk
@onyx hinge all good, there's definitely a lot of layering in the IDF. I can't say I understand it as well as the ST HAL just yet
devout jolt
Anyone else having basic connection issues with MagTag ESP32S2 and latest CP 6.0.0-260-ga0d305042? I'm getting "Connection Error: Unknown failure"
slender iron
@devout jolt don't use 6.0.0
it's old for the ESP32-S2
you can be the first to use 6.1.0 beta 0
🙂
devout jolt
I was using the magtag latest build off S3. But I just tried 6.1.0-beta.0 and same problem
slender iron
hrm
mine was working with a build from yesterday
though it was a debug build
I'm trying a build from the 6.1.0-beta.0 tag now
devout jolt
I'm getting some really weird results. Sometimes it soft reboots, sometimes it says "no SSID" when it had just printed out the SSID in question, sometimes it just hangs on connect
and sometimes it gives that "unknown failure"
slender iron
it'll be just my luck that it broke today 🙂
at least I'll have done the release notes
one fetch on mine worked ok
are you using bin or uf2 to update?
devout jolt
always the .bin. I don't know how to get UF2 bootloader to appear. Double-tap reset doesn't work for
jaunty juniper
So I'm encountering the following with the latest main (not in 6.0.0) on the feather S2:
The following boot.py
storage.remount("/", False)```
Results in the following boot_out.txt
```Adafruit CircuitPython 6.1.0-beta.0 on 2020-11-17; FeatherS2 with ESP32S2
boot.py output:
Traceback (most recent call last):
File "boot.py", line 2, in <module>
RuntimeError: Cannot remount '/' when USB is active.```I just tried 6.0.0-rc.1-213-g61bf0e99c (the earliest one for MagTag). It does connect to my AP that runs the authoritative DHCP server, but I get "unknown failure" on the AP that's just on ethernet (and uses the authoritative DHCP server)
slender iron
@jaunty juniper please file an issue. sounds like a tinyusb bug
@devout jolt has it ever worked?
devout jolt
@slender iron with the 20201110 firmware (g61bf0e99c) the MagTag will connect on the WiFi AP that hosts the DHCP server, but "unknown failures" on the AP that gets DHCP over Ethernet. Your "firmware.bin" gives that same result too. (Also, I realize I never tested my ESP32S2 Saola1 CP scripts on my other WiFi AP that isn't a DHCP server)
devout jolt
it almost seems like an early timeout on wainting for DHCP results (like if they don't show up within 5msec)
slender iron
uh oh. the website data file is too large again
@devout jolt
the previous bin was old too
thorny jay
you can be the first to use 6.1.0 beta 0
@slender iron That solve simpleclock that was crashing on me.
slender iron
👍
devout jolt
@slender iron that firmware.bin about 50% of the time hard-reset, kicking out the CDC and MSD on wifi.radio.connect() Correction: it only seems to hard-reset on ctrl-D continue, not if I connect after it's started up
slender iron
hrm
devout jolt
my script is:
import time
import wifi
ssid = "todbot-back"
password = "maxwellcat" # hack me plz
print("scanning...")
for network in wifi.radio.start_scanning_networks():
print("\t'%s'\t\tRSSI: %d\tChannel: %d" % (str(network.ssid, "utf-8"),
network.rssi, network.channel))
wifi.radio.stop_scanning_networks()
print("connecting to:",ssid,"passwd:",password)
wifi.radio.connect(ssid=ssid, password=password)
print("connected.")
while True:
print(time.monotonic(), "loop")
time.sleep(1)
k, let me try it
yup, I'm seeing the usb crash too
oooh, and I got unknown failure
and now my usb is unhappy
I'll keep poking after I fix the website
I'm bad at doing multiple things at once
devout jolt
@slender iron doh! Another data point potentially: On Metro ESP32S2 and CP 6.1.0-beta.0, similar results of "Unknown failure" on AP w/extern DHCPd and strangely hanging on the AP w/ the DHCPd. But only on Ctrl-D or code-save restart. On hard-reset, the above script comes up fine
thorny jay
Yeah, not working very well here too. Now corrupted file system and code.py (and one font file) is broken beyond repear. Also stuck on "Connecting" in various example. Time to sleep for me.
slender iron
seems like it's gotten worse
@devout jolt mind filing an issue? I've gotta finish some website updates before I can focus on this
devout jolt
@slender iron no problem. I wish I could characterize it better than just "dudn't work fer me"
slender iron
no problem. I can reproduce it which is good
manic glacierBOT
thorny jay
Definitively something USB related. When working stand alone on battery, the clock work. When connected to USB it was stuck on connecting (from what I could see on the console).
lone axle
In a Library, would we want to keep backwards compatibility like this between Sprite and TileGrid?
# 4.0.0 Beta 2 replaces Sprite with TileGrid so use either.
self._sprite_class = getattr(displayio, "Sprite", displayio.TileGrid)
or is it okay to remove that at this point?
onyx hinge
@lone axle a rule of thumb would be, if we no longer produce a 4.x bundle then we would not need to keep workarounds or backwards compatibility for 4.x. Since that was from before stable 4.x, the case to remove it is even stronger.
(right now we make a bundle for 5.x and 6.x)
users can always get the older releases of the library via github
low sentinel
MagTag looks so good, I can’t wait for a stock alert ❤️
thorny jay
So I have been playing with MagTag and the SimpleClock code I found in the LearningGuide repo. Except for all the improvement on the core, I feel that we need a RobustClock that can serve as a template with better handling of exception, network error, ... I would even think that we need some kind of watchdog that once enabled requires your to send a keep alive message or else it reboot.
I guess that kind of remark would be true also for PyPortal and PortalMatrix.
But when you have that thing on the fridge, and the only thing it display is an error message... you start thinking about ways to "auto-resolve" them.
I took picture of the screen each time I had error, and I'll try to share them.
thorny jay
Also, I would love to see support or code for multiple SSID, trying one, or if not available, another one. Some of my error related to not beeing in reach with the main SSID. (you know my vilain lair is rather big).
So that would be multiple SSID and multiple password.
slender iron
@thorny jay Again, it's just early. I have a demo I want to work reliably as well and I'm debugging it's issues.
onyx hinge
esp32s2 question -- does anyone know if this doc (GPIO25/26) is just wrong? Other things refer to GPIO17/18 as the DAC. https://docs.espressif.com/projects/esp-idf/en/latest/esp32s2/api-reference/peripherals/i2s.html#_CPPv424I2S_DAC_CHANNEL_RIGHT_EN
I2S_DAC_CHANNEL_RIGHT_EN = 1
Enable I2S built-in DAC right channel, maps to DAC channel 1 on GPIO25I2S_DAC_CHANNEL_LEFT_EN = 2
Enable I2S built-in DAC left channel, maps to DAC channel 2 on GPIO26
vs our AnalogOut code: ``` if (pin == &pin_GPIO17) {
self->channel = DAC_CHANNEL_1;
} else if (pin == &pin_GPIO18) {
self->channel = DAC_CHANNEL_2;
analog bridge
@onyx hinge are you trying to access the built-in dac using i2s ?
onyx hinge
yeah that's what I'm up to
I've refactored the i2s code so it can be shared with AudioOut
analog bridge
I already tried that ....unfortunately built-in dac control using i2s is not supported on the s2.
I mentioned this earlier but I think you missed it
analog bridge
idf docs have errors and/or are not updated with the recent changes
onyx hinge
>>> import digitalio
>>> pin = digitalio.DigitalInOut(board.IO0)
>>> pin
<DigitalInOut>
``` this doesn't seem right -- why can I make a DigitalInOut from IO0 while it's already a part of an SPI object?bug
onyx hinge
@slender iron OK I'll check if there's an issue and file it if not
@analog bridge do you have any links about i2s / dac / esp32s2 or is it just based on your own experimentation?
components/soc/soc/esp32s2/include/soc/i2s_caps.h:#define SOC_I2S_SUPPORTS_ADC_DAC (0) // ESP32-S2 don't support ADC and DAC ah
analog bridge
I was looking for that line... I thought it was in an example
onyx hinge
@analog bridge do you think there's a different way to do it, or are we limited to i2sout?
thorny jay
@thorny jay Again, it's just early. I have a demo I want to work reliably as well and I'm debugging it's issues.
@slender iron No worries, I know this is alpha, very new hardware, ... I was just reporting the experience I had and the need for the python code to be robust (and for the clock specifically, it could continue even without being able to do the hourly check) I tried to put some exception handeling, but it is not enough. I need to check the pictureI took to see the various error I encountered. One hardware problem I had this night is the green "power" LED. I don't care if it consume power, but it is a light in a very dark room, and I also like to sleep in the dark. I am sure this is your case, so I will be interested in your solution (masking tape? unsoldering? can it be controled by software?).
analog bridge
... I would even think that we need some kind of watchdog that once enabled requires your to send a keep alive message or else it reboot.
@thorny jay my PR for watchdog support on the esp32s2 got merged recently.
thorny jay
@thorny jay my PR for watchdog support on the esp32s2 got merged recently.
@analog bridge Much needed, great. You kind of want the stuff to restart if something goes really wrong. One worry I have is how to maintain a "state" between restart, so that the software can return to the same "mode" as before the crash (example: "I was showing weather forcast"). Is there some kind of NVRAM were we can write some information, or would you do that on the flash? It need a place that survive a "reset by the watchdog" and where you can frequently write without destroying the hardware.
orchid basinBOT
Lets just use this transparent background png of the Pi 3 for that image for now. Not the same dimensions, but it should work fine. @tannewt Let me know if you had something else in mind.
onyx hinge
@analog bridge did you happen to investigate this suggestion of SPI3 -> DAC in the reference manual? (page 641)
analog bridge
nope
figuring why i2s control wasn't working exhausted me too much and I moved on...
onyx hinge
@slender iron is spi3 already "dedicated" to something? seems like it's fairly special
analog bridge
@thorny jay I have to check for whether a wdt reset will clear ram or not but if it does than for preserving state you will have to use watchdog mode RAISE which raises an exception that you can catch in your code and then set state in flash using nvm
onyx hinge
there's sure no example or code for PWM3/DAC that I can find
manic glacierBOT
onyx hinge
hmm single channel samples to i2s is missing
slender iron
@thorny jay I feel the same about the green power LED. I haven't looked into how to disable it. worse case I solder it off. 🙂
@onyx hinge I don't think spi3 is used
lapis hemlock
@onyx hinge Jeff, in the workflow file, is there any particular reason for using python 3.8? micropython tries to be 3.5-compatible. In any case, the problem is that the test scripts fail locally, while they succeed on github. The snag is that we can't add new tests, because we can't generate the expected output files for them. E.g., this https://github.com/v923z/micropython-ulab/blob/init/tests/initialisation.py spectacularly falls through: https://github.com/v923z/micropython-ulab/pull/228/checks?check_run_id=1420028333 Do you have a clue as to what is going on here?
onyx hinge
@lapis hemlock interesting, I'm not sure
onyx hinge
while it installs python3.8, the tests should be invoking python3.5 (which we're depending on being installed in some other way)
env MICROPYTHON_CPYTHON3=python3.5 MICROPY_MICROPYTHON=micropython/ports/unix/micropython micropython/tests/run-tests -d tests
lapis hemlock
@slender iron Scott, is there any way to fix the following: 253. + 238. + 248. -> 739.0000000000001? This happens on the unix port. This came up first during testing of ulab, but then I figured that it is not a ulab issue per se, it comes from micropython itself.
slender iron
I have no idea
lapis hemlock
Great. Now that makes us two.
onyx hinge
that's odd, the arithmetic should be exact
lapis hemlock
I think this is a rounding error.
The question is, why it is displayed in this particular way.
onyx hinge
bug exists in MicroPython v1.12-676-gad95c0590-dirty so consider reporting it upstream (while understanding that it's difficult for us to take fixes from them)
slender iron
I wonder if github is failing to do proper diffs for us
lapis hemlock
Even better: this is exact 253.0 + 238.0 + 248.0 + 253.0 + 238.0 + 248.0 -> 1478.0
onyx hinge
it's a display error, not an error in the value
lapis hemlock
I think you are right: 739 should fit in any float representation exactly.
onyx hinge
$ ./micropython -c 'import struct; print(struct.pack("d", 253. + 238. + 248.))'
b'\x00\x00\x00\x00\x00\x18\x87@'
$ python3 -c 'import struct; print(struct.unpack("d", b"\x00\x00\x00\x00\x00\x18\x87@"))'
(739.0,)
or more to the point, ```>>> 739.0
739.0000000000001
the floating point number to be printed gets scaled so that it has just one digit to the left of the decimal place: ```265 f *= *neg_pow;
the float-printing routine is certainly not designed to give an accurate result in all cases, but .. this is still an unfortunate consequence of the implementation.
this is py/formatfloat.c
lapis hemlock
I have just checked, this is not port-specific.
@onyx hinge Do you think I should open a ticket, or leave it at that?
onyx hinge
in circuitpython we're unlikely to address it ourselves
I don't know what mircopython will do with the report
lapis hemlock
I meant in micropython. If there is a reasonable fix, I would port it to circuitpython
onyx hinge
agreed, we can probably port a change to formatfloat.c to circuitpython
you probably are aware, printing floats accurately is a huge problem and most implementations that do it need lots of code and also perform multi precision arithmetic with allocations. those strategies won't be appealing in micropython
but maybe a special case for numbers that are exactly 32 bit integers would be feasible to add
@lapis hemlock did you try putting in a PR to install python3.5 instead of 3.8, to see what happens? I still don't know what's going on with that test. we got off on a tangent..
lapis hemlock
@lapis hemlock we got off on a tangent..
@onyx hinge That's right, I also got hung up on the float issue. I am going to open the ticket, and then see to the workflow file.
candid granite
Hi everyone, anyone know if I make a circuitpython driver for some sensor like this one for example https://github.com/adafruit/Adafruit_CircuitPython_SHTC3 is there a place where third party libraries are listed where I could have mine listed too? I also saw the CircuitPython_Community_Bundle, maybe this is where I should aim for my driver to get added?
idle owl
@candid granite The Community Bundle is exactly the place to aim.
lapis hemlock
@lapis hemlock did you try putting in a PR to install python3.5 instead of 3.8, to see what happens?
@onyx hinge That doesn't fix the issue. I still have a couple of things to try out, though...
opal crystal
Regarding the float printing – this may be relevant: https://github.com/micropython/micropython/issues/5284 (from https://forum.micropython.org/viewtopic.php?t=7177)
idle owl
@lone axle Are you around?
lapis hemlock
@opal crystal Thanks for pointing this out. So, this is a known issue.
lone axle
@idle owl yep, whats up?
idle owl
Hmm... Maybe I'm missing something. First off I feel like I shouldn't be using MagTag to test your Slideshow update. Second, should the simpletest work with the json file or is there some other thing that has to happen?
(Switching to a PyBadge)
lone axle
Ah yeah, I tested only on PyPortal so far. The simpletest can be used to test it but you would need to paste the text json file into the root of CIRCUITPY instead of images/ like it is in the repo
idle owl
Ok
lone axle
I actually just noticed that the simpletest is using / for the folder. I had assumed it was using the images. I think that is why I put the text in there.
idle owl
Simpletest isn't working.
Had to remove the BACKLIGHT line altogether to make it work on PyBadge which makes me wonder if we should have it in the simpletest, but....
It either skips your file, or it says RuntimeError: No valid images or text json files
lone axle
I can try it out on that device. perhaps there is something I have in my local one that accidentally didn't make it into the commits.
idle owl
Alright. I'll comment on the PR so this is somewhere useful.
lone axle
I will be working on it a bit more tonight. I would love to get those last two items implemented before it gets merged anyhow. I'll check it out on the magtag as well when I am working on it.
idle owl
@lone axle MagTag support in Slideshow was something I was hoping you'd sort out as well. The display requires a refresh() where others do not, so you'd have to figure out how to incorporate that cleanly. I'm not sure it's reasonable to do and may simply require another example, but having a single simpletest run on multiple things without significant changes would be ideal.
idle owl
I won't be able to test again until next week if you're getting back to it tonight, so no worries on implementing more features. You have time.
I would make a note that you are going to do that though so someone else doesn't test/merge it.
lone axle
Yep, it is in there now. Ladyada got to it before mine arrived to tinker with it. But I got it tested out over the weekend and it's already merged. So MagTag is working correctly with images. But I haven't tested it with the text slides yep.
idle owl
Oh right on. Good to know.
I think the simpletest needs some tweaking then.
Because it didn't work with either MagTag or PyBadge out of the box. 😄
lone axle
agreed. I'll try to sort that out as well I can put a fix for it in the same PR if that is okay.
idle owl
That would be fine. Just make a note of it.
lapis hemlock
@onyx hinge One indicator of the perpendicular (non-tangential) problem is that the latest version of micropython does not produce the expected files. There is actually no output for a test script that has never been run. I don't quite know what this means, though...
lone axle
@idle owl Ah I see what the deal is. The text slide support requires adafruit_display_text also, but it has a try/except to fail gracefully if it's missing.
onyx hinge
@lapis hemlock is this a change in micropython?
idle owl
@lone axle I had an old version of it temporarily and removed it and get the same error
Let me add it
lone axle
I should document that more clearly in the comments / docs for sure. Maybe even print a warning in the console.
onyx hinge
-If a test fails, run-tests produces a pair of <test>.out and <test>.exp files in the current
+If a test fails, run-tests produces a pair of <test>.out and <test>.exp files in the result
directory with the MicroPython output and the expectations, respectively.
lapis hemlock
I guess so. I just pulled the latest changes, and re-compiled the firmware, and since then I haven't seen the .out/.exp files.
idle owl
@lone axle Traceback (most recent call last): File "code.py", line 11, in <module> File "/lib/adafruit_slideshow.py", line 237, in __init__ File "/lib/adafruit_slideshow.py", line 239, in <listcomp> File "/lib/adafruit_slideshow.py", line 205, in _check_json_file OSError: [Errno 2] No such file/directory: '//.fseventsd'
onyx hinge
- cmd_parser.add_argument('-r', '--result-dir', default=base_path('results'), help='directory for test results')
is there a "results" folder/directory?
lone axle
Hmmm, not sure about that one. Will check into it though.
idle owl
@lone axle Updated the PR. Thanks!
onyx hinge
".fseventsd" is normally an empty directory, not sure what that means for the message ... supervisor/shared/filesystem.c: res = f_mkdir(&vfs_fat->fatfs, "/.fseventsd");
er no it has an empty file no_log inside it
lapis hemlock
is there a "results" folder/directory?
@onyx hinge No, there isn't.
lapis hemlock
@onyx hinge And watch this: ```cp 00smoke.py 01smoke.py
cd ..
./build.sh
....
...
pass tests/00smoke.py
Traceback (most recent call last):
File "tests/01smoke.py", line 1, in <module>
import ulab
ModuleNotFoundError: No module named 'ulab'
FAIL tests/01smoke.py
01smoke.py is a verbatim copy of 00smoke.py, and it passes only, if I also copy 00smoke.py.exp onto 01smoke.py.exp. Shouldn't the test scipt run, no matter what? The ImportError is definitely too early for anything.
onyx hinge
./micropython/tests/results/tests_cholesky.py.out
lapis hemlock
I am most probably bewitched.
onyx hinge
inside the micropython source inside ulab?
lapis hemlock
I see them now. A most unusual location.
onyx hinge
so next problem .. why it gets an error importing ulab
run-tests can work in one of two ways: first, the output of micropython compared to the output of desktop python3. second, the output of micropython compared to an ".exp" expected file
if there's no .exp file it runs the same program with python3 (or 3.5 or 3.8 or whatever...) and (A) that has to succeed and (B) then that's used as the expected output for comparison
but of course you can't import ulab
so instead you have to manually create the expected file. This USED to be possible by running micropython's run-tests in an unusual way but it's not now so I won't bother saying what that unusual way was
onyx hinge
./micropython/ports/unix/micropython tests/cholesky.py > tests/cholesky.py.exp && git add tests/cholesky.py.exp so instead you can just do this, except change the test name
in 3 places
unless your changes to output formatting mean that cpython3 numpy output and ulab output can be compared directly
we MIGHT want to just take a copy of run-tests (a python script) at some ref where we control the behavior, instead of needing to adapt to changes in the script in micropython
lapis hemlock
so instead you have to manually create the expected file. This USED to be possible by running micropython's run-tests in an unusual way
@onyx hinge Well, this is somewhat exasperating, I must say.
lapis hemlock
The tests are trivial in the sense that we have only textual output, so can't we just run micropython with the test script, catch the output, and compare it to .exp? We would have to compare the output to numpy, but we have to do that only once, at the time of creating the .exp file.
I wonder, why this has changed in micropython, but this is rather inconvenient.
And how is it in circuitpython? We could live with checks against circuitpython, if it comes to that...
idle owl
@onyx hinge Hey buddy. How's it going? I need git help.
idle owl
Text is fine.
Actually, one step back. Is there a way to verify that a submodule is up to date?
Before I try to figure out how to update submodules.
onyx hinge
by "up to date" you mean matches whatever rev is commited in the superproject?
idle owl
Um...
idle owl
By up to date I mean the submodule is even with itself.... Specifically, on circuitpython-org, is adabot up to date
with the current changes to Adabot
Wait I clicked on it in GitHub. And it shows me the latest changes to Adabot
onyx hinge
you want to know if circuitpython-org has the latest changes to adabot
onyx hinge
and now you know?
idle owl
But I think I answered my own question after rethinking it.
I believe so?
Do you know much about GitHub Actions?
onyx hinge
I pretend to, like I do with a lot of things
idle owl
Because this is a weird one, I re-ran a job and it's not seeing the updated submodule. It's installing from an old version of requirements.txt, I don't even have to wait for it to fail the whole job. I know it's wrong within 30 seconds or so.
We have to wait anyway because we hammered the API w/in the last hour with too many requests to do another full run.
manic glacierBOT
idle owl
But it installs dependencies before beginning requests, so it's possible to see it's not right before we get to that part of the job.
onyx hinge
can you point me at the run?
lapis hemlock
but of course you can't import ulab
@onyx hinge I could catch that at the very beginning like ```try:
import ulab as np
except ModuleNotFoundError:
import numpy as np
But that would imply that the workflow has got to install numpy, too.
onyx hinge
@lapis hemlock what I mean is,
[0. 0. 0.]
$ ./ports/unix/micropython -c 'import ulab as np; print(np.zeros(3))'
array([0.0, 0.0, 0.0], dtype=float)
``` the output are different@idle owl right, when you use the "re-run" button it uses the particular ref, which is before the commit that changed the adabot submodule
@lapis hemlock I'm working on scripting it a bit better
lapis hemlock
@onyx hinge This might be a bit more difficult, since, at least, in circuitpython, you have sub-modules that don't exist in numpy.
prime flower
whee, got most of minimqtt working with cpython. 📡
onyx hinge
@idle owl I think we can change the workflow so that you can trigger it from the web ui
idle owl
@onyx hinge Hmm.
onyx hinge
let me prepare a PR so you can see what I mean
onyx hinge
in a repo of mine for some random thing, I have a workflow that can run as a cron .. yup
idle owl
Excellent. Test bed.
lapis hemlock
@onyx hinge With the latest micropython, I can't just run an arbitrary script on the command line. This used to work: ./micropython /dev/shm/micropython.py , but not anymore.
onyx hinge
@lapis hemlock probably there will be some problem with it but https://github.com/v923z/micropython-ulab/pull/229 is what I'm thinking
onyx hinge
@lapis hemlock hum that is not something that I encountered yet
lapis hemlock
That line produces the garbage in the expected files, like ```micropython: ../../py/argcheck.c:111: mp_arg_parse_all: Assertion `(allowed[i].flags & MP_ARG_KIND_MASK) == MP_ARG_OBJ' failed.
CRASH
wow that's not right
if they've entered an unstable phase of development, maybe ulab needs to check out a specific revision like their last stable release
lapis hemlock
if they've entered an unstable phase of development, maybe ulab needs to check out a specific revision like their last stable release
@onyx hinge We could do that.
@lapis hemlock probably there will be some problem with it but https://github.com/v923z/micropython-ulab/pull/229 is what I'm thinking
@onyx hinge Thanks, I have merged the changes. That fixes the problem with the missing .exp files, but the issue with the crash report still persists. I will try to roll backmicropythonto a point, where it works again.
I have to take off now, but many thanks for your contributions!
This was a rather intense evening.
onyx hinge
see you!
lapis hemlock
So long.
onyx hinge
@prime flower "got most of minimqtt working with cpython" yay excellent
prime flower
@onyx hinge yep, thanks jeff! Going to convert all the recvs into recv_into with a shared buffer next, and then creep into ssl-specific code.
onyx hinge
this is different than adafruit io mqtt? or under it?
prime flower
It's the circuitpython mqtt client. Adafruit IO Circuitpython uses it.
onyx hinge
got it
I was just ruminating about how adafruit io http isn't working with esp32s2 right now
as far as I know/knew
prime flower
ideally it will after tomorrow 🙂 , i was waiting for the ssl issue to be resolved.
onyx hinge
yeah it was a chicken and egg problem I suppose
prime flower
also a lack of time on my part between school/work. I don't expect blocks, but I havent tested sending requests to IO with the ESP32S2 yet
believe me, I'm excited for the magtag on my desk to display a data feed from IO 🙂
blissful pollen
@prime flower if you need any help testing i've been working on updating a project that's reading IO data. So far its in Arduino just doing HTTP requests by hand - which is painful 🙂
prime flower
@blissful pollen that'd be great - with ESP32S2?
blissful pollen
Yup my magtag is sitting beside me right now
prime flower
oh...I updated the ESP32S2 guide to include the new HTTP stuff!
for arduino
You should be able to use WiFiClient unless you mean by hand as in writing client.printlns for requests..?
blissful pollen
Oh cool I'll take a look. Does the library allow you to retrieve the last X data points or just the most recent? I've been trying to average data out
I was using WiFi client and then parsing the JSON, just sending HTTP by client.println() is annoying and parsing headers, etc.
prime flower
They only allow retrieval of most recent since thats the most popular request from users, you could add it to the Adafruit IO Arduino/Adafruit IO CircuitPython library if you'd like
blissful pollen
Cool, I've been waiting to start playing with the CP library for it and will take a look at it for sure
prime flower
Why do you need to retrieve X data points? Are you graphing on the magtag?
blissful pollen
Averaging the last few data points - say for temperature over the last 5 minutes. Or pressure changes over say 6 hours. Graphing is a "nice to have" idea on my mind still
jovial topaz
silly question about support for esp32 and CircuitPython :
Now that support for the esp32s2 is being added, how hard it would be to support normal esp32 but without native usb ? I know that CircuitPython relies a lot on the ease of usage of having native usb, but would be cool to use it on "old" esp32 as it is more well documented and have a bigger community than micropython IMHO
prime flower
Yeah, that's currently not possible with AIO CP, but the AIO API (https://io.adafruit.com/api/docs/#get-most-recent-data) supports it. I'd looove to have a way for IO to perform the averaging instead of the device.
tulip sleet
@jovial topaz Our principle is that it's not CircuitPython without the workflow experience that USB provides. Conceivably that workflow could be provided via BLE or Wifi, but that would not be easy.
We had CircuitPython for the ESP8266, and dropped it for several reasons. One strong reason was that it was an outlier in terms of how you had to use it (no CIRCUITPY, etc.)
ionic elk
@slender iron I haven't seen much about ESP32-NOW, is support for that on the docket? or already done and I don't know about it?
jovial topaz
gotcha @tulip sleet . I think CircuitPython is awesome beyond the nice USB workflow. I would prefer to use it over micropython, as the CircuitPython libraries and tutorials are much better.
I have a LOT of esp32 here that I used to give some Arduino workshops here in Brazil, but I was planning on move to CircuitPython that seems nicer for new devs. I know that would still go thought the "hassle" of using something like ampy to transfer files, but I made some tutorials using Jupyter Notebook + esp32 + micropython (https://www.youtube.com/watch?v=PTYXsgUSoxs - it's in PT-BR) and the experience was nice even without native usb. The problem is that Micropython libs are not well documented as CircuitPython ones, so is a bit harder for newcomers
And here in Brazil is really easy to get the "old" esp32 and like I mentioned before, I already have a bunch. I ordered yesterday some esp32s2 from AliExpress to try it with CircuitPython to collab a bit more on the project. Overall I just wish that I would have better access to boards with CircuitPython support here in Brazil haha
But thanks for the feedback @tulip sleet
slender iron
@slender iron I haven't seen much about ESP32-NOW, is support for that on the docket? or already done and I don't know about it?
@ionic elk It's not on my radar to support
ionic elk
@slender iron are there other ways to connect between multiple ESP32s on the same wifi network? That was my primary interest in it, since the S2 doesn't have bluetooth
slender iron
I have no idea
ionic elk
well, just tossing it out there - device-to-device wireless communication seems like it'd be neat to have
onyx hinge
you can call the circuitpython module soonio
ionic elk
@onyx hinge I haven't even seen that one
onyx hinge
tulip sleet
@jovial topaz Adafruit_Blinka has support for pyboard, and indeed, one of the original motivations for Blinka was to run CircuitPython libraries on top of MicroPython. That has been eclipsed by supporting various Linux boards, but the support is still there. Would that help you?
onyx hinge
I'm glad you appreciated it
jovial topaz
I'll check it out @tulip sleet , I thought Blinka was just for SBC/Linux hosts
tulip sleet
I'm just looking at it now, to make sure the support is still there, but it looks like it.
onyx hinge
seriously though espnow looks interesting -- seems devices will need to be on the same wifi AP but you can send unreliable messages between them using the HW address. I assume it's unreliable "ethernet" packets from what I can see. optional encryption, not sure how you get the keys
discovery via use of the broadcast address? shivers well what else will you do
huh no it seems to be independent of an AP. interesting-er
slender iron
let's get the basics working reliably first
ionic elk
I mean I'm not the expert on wireless stuff but device-to-device communications has always seemed like a big pull for a device, so you don't have to rely on an external lora or something
onyx hinge
ionic elk
I might give it a try outside of Circuitpython in the meantime
would be nice to have something like that for my alarm clock
slender iron
you don't have an access point already?
ionic elk
If you mean me, my alarm uses NRF20L01s lol. Not the greatest. Also unsellable due to certifications
ionic elk
certainly seems to imply it
onyx hinge
implementation is in a binary blob, but it does seem to "be there" for esp32s2 components/esp_wifi/lib/esp32s2/libespnow.a
https://blogs.arubanetworks.com/industries/802-11-action-frames/ I did not know what "action frames" were. here's an overview
cobalt grail
Anyone with a working gray scale example for the MagTag willing to share the code? Mine only displays black and white... 😕
slender iron
what build are you using? does it have board.DISPLAY?
cobalt grail
@slender iron It did not. It does now. 🙂
So how would I use that board.DISPLAY stuff?
cobalt grail
I installed adafruit-circuitpython-adafruit_magtag_2.9_grayscale-sv-20201118-8ac9b17.bin and it reports 0 byte available of 963K byte for the disk. That does not sound right...
slender iron
ya, the latest has been a bit unstable
idle owl
@onyx hinge Did you end up doing up the thing with being able to trigger the job manually? I was going to say maybe don't, if you haven't, but, it occurs to me that if it's still not working, or for future use, it might still be worth it. So I'm apparently still interested?
onyx hinge
@idle owl whoops I thought I shot in a PR for it
idle owl
Maybe you did and I missed it?
onyx hinge
maybe I didn't click submit
idle owl
Tag me on it please 🙂
idle owl
@onyx hinge .... that's it? Hah!
onyx hinge
you have to KNOW what it is though
idle owl
I guess!
crimson ferry
the connect does a second scan
@slender iron does it pick the AP with the lowest RSSI from among those with the same SSID? I don't see a CircuitPython-leveldisconnect, does automatic reconnection happen in circuitpython code or espi idf? n/m on the second question, I found the event handler
idle owl
@onyx hinge So, once that's added, it shows up able to be run manually? There's so many options in the example, they're not needed?
I trust you, it just seems too easy.
onyx hinge
once that is in main, the page within actions should have a new button to click
(or master)
idle owl
Great!
onyx hinge
idle owl
It worked!
onyx hinge
yawsssss
idle owl
Whee and failed! 😄 This was the test run, so this is not a surprise.
onyx hinge
huh ERROR: Invalid requirement: 'requests-cache=0.5.2' (from line 10 of adabot/requirements.txt)Hint: = is not a valid operator. Did you mean == ?
onyx hinge
in adabot?
idle owl
yes
slender iron
idle owl
@onyx hinge ok now I need to submodule update circuitpython-org. Do I run git submodule update and then commit/merge? Or is there sync and init and all that involved?
Or is there a way to update only the adabot submodule?
Which would make me much more comfy
onyx hinge
@idle owl In the submodule, "git checkout master; git pull" (if the branch is called master)
that will change the checked out files in the submodule to be the tip of master branch, just like "git pull" always does
idle owl
there's nothing in the submodule folder :/
onyx hinge
ok better do submodule init/update then
idle owl
ah ok
onyx hinge
in the outer world
after the submodule update, go into the submodule and do the steps I said
idle owl
ok
onyx hinge
then go up to the top directory, "git add adabot" "git commit adabot" and write your commit message
onyx hinge
does it have the right content or the wrong content?
idle owl
correct
@onyx hinge On branch adabot-update-18-nov Changes to be committed: (use "git restore --staged <file>..." to unstage) modified: adabot
onyx hinge
OK try git commit
onyx hinge
@idle owl yeah looks good to me too -- did you find where you can click to verify it?
idle owl
Click to verify it?
onyx hinge
click the "1 files" link https://github.com/adafruit/adabot/compare/9c67b7f923226b628c691a48987d9eaae2940051...665d9ae7f8109d6552ff8821ee052619cba78a8e
I mean, to double check
idle owl
Yes!
onyx hinge
happy to help with gaps in your knowledge
idle owl
Now we wait another 20 minutes for this to run or fail eventually!
onyx hinge
or we fall asleep first
onyx hinge
@slender iron @ionic elk looks like maybe the implementation of calibrated ADC on esp32s2 is incomplete? uint32_t reading = adc1_get_raw(ADC1_CHANNEL_5); uint32_t voltage = esp_adc_cal_raw_to_voltage(reading, adc_chars); is suggested https://docs.espressif.com/projects/esp-idf/en/latest/esp32s2/api-reference/peripherals/adc.html but we don't actually call esp_adc_cal_raw_to_voltage that I saw
ionic elk
I'll have to check my notes, but I'm pretty sure that you literally can't yet on the ESP32-S2
The ESP32 and the S2 have different voltage curves
When I started on the S2 ADC stuff, we had to update the IDF because even the basic ADC calibration stuff had only been implemented that month. I can check in again on where things are at. If there is calibration, it's probably just a bit of smoothing at the ends
From what I recall the S2 also caps out entirely at about 2.6V and will not go higher, even at 11DB attenuation
onyx hinge
OK, it came up in the internal meeting so you may get another ping about it 🙂
ionic elk
Happy to take another look stuff might have changed
crimson ferry
@ionic elk esp32-s2 at 11dB should go to 3.3V
it's just optimal between 0.150-2.6V
crimson ferry
@onyx hinge Is this not what we need? https://github.com/adafruit/circuitpython/blob/8ac9b176c961faf7672f081953de6f78ec3005e2/ports/esp32s2/common-hal/analogio/AnalogIn.c#L92
onyx hinge
@crimson ferry mea culpa
manic glacierBOT
manic glacierBOT
lament sentinel
Making a command line tool for working with my CircuitPython boards. First thing I implemented was "backup" so that I can grab a snapshot of the current state in a zip file. https://github.com/bcr/blinka-cli
analog bridge
...are there other ways to connect between multiple ESP32s on the same wifi network?
@ionic elk ESP-MESH seems to be promising
lone axle
@lament sentinel that looks awesome. Thanks for sharing. I love the idea of CLI to backup devices. I am interested to see what other helpful utilities can gets added too.
lament sentinel
@lone axle I am adding a BOSSA updater to it now. I have some old boards that need upgrading to CircuitPython
tulip sleet
@lament sentinel be very careful with bossa. The default value of the --offset argument was changed around 1.8 or 1.9, and is now 0x0. If your bootloader is not protected, it will overwrite the bootloader with whatever you're uploading.
lone axle
💯 % true. I mostly bricked a Lora Feather that way. Wouldn't have been able to get it back to working without help from Dan 🙂 .
lament sentinel
@tulip sleet bruh tell me about it. I am super duper aware and have been reading every page to make sure I got this right. I understand the v1.7 change for offset, I understand the SAMD21 offset 0x2000 vs SAMD51 0x4000. And I am going to put dire warnings on it and explain exactly my plan for executing bossac before I do it
@lone axle yes, this is part of the motivation, to get this right. And I will make sure to ask for reviews and make it as clean as I can so we get this right
tulip sleet
it was a :🤷 forehead-slapping change, and has still not been fixed
lament sentinel
It is fully my intent to understand the BOSSA version and behave appropriately, I feel you. I once accidentally changed the default address for another update tool I made and bricked myself 😆 — part of the reason I am so careful
tidal kiln
is it possible to use a manually created displayio.Palette + displayio.Bitmap with MagTag's EPD?
onyx hinge
@tidal kiln I don't know why not
tidal kiln
do you know what the palette values should be?
trying several things and only getting black/white
onyx hinge
I could be wrong of course 🙂
is it due to this bug? https://github.com/adafruit/circuitpython/pull/3719
@tidal kiln maybe try with an artifact from that PR and see if it works?
tidal kiln
thanks. looks like it maybe related. let me try that....
ivory yew
Hey CircuitPythonistas- I haven't been active in a while but I'm still around and I wanted to thank all of you for the 6.0 release. I promise I'll be back soon to cause more trouble. 😎
onyx hinge
@ivory yew 👋
@tidal kiln fixed it for me .. before I just got two colors, now I get 4. The color values I used were 0x000000 0x555555 0xaaaaaa 0xffffff
@ivory yew as always the community will be here for you when you have the time & spoons to engage with us again!
lament sentinel
@lone axle @tulip sleet PS C:\Users\bcr00\Source\bcr\blinka-cli> python .\blinka.py --verbose bossa --port COM5 DEBUG:root:options = Namespace(verbose=True, quiet=False, locale='en_US', command='bossa', bossa_path='C:\\Program Files (x86)\\BOSSA\\bossac.exe', port='COM5', func=<function do_bossa at 0x000001EF4C0FFF70>) INFO:root:Using en_US for the locale INFO:root:Executing ['C:\\Program Files (x86)\\BOSSA\\bossac.exe', '--port=COM5', '-i'] DEBUG:root:returncode = 0 DEBUG:root:Stdout is b'Device : ATSAMD21x18\r\nVersion : v2.0 [Arduino:XYZ] Mar 5 2016 17:46:52\r\nAddress : 0x0\r\nPages : 4096\r\nPage Size : 64 bytes\r\nTotal Size : 256KB\r\nPlanes : 1\r\nLock Regions : 16\r\nLocked : none\r\nSecurity : false\r\nBOD : true\r\nBOR : true\r\n' DEBUG:root:key = 'Device' value = 'ATSAMD21x18' DEBUG:root:key = 'Version' value = 'v2.0 [Arduino:XYZ] Mar 5 2016 17:46:52' DEBUG:root:key = 'Address' value = '0x0' DEBUG:root:key = 'Pages' value = '4096' DEBUG:root:key = 'Page Size' value = '64 bytes' DEBUG:root:key = 'Total Size' value = '256KB' DEBUG:root:key = 'Planes' value = '1' DEBUG:root:key = 'Lock Regions' value = '16' DEBUG:root:key = 'Locked' value = 'none' DEBUG:root:key = 'Security' value = 'false' DEBUG:root:key = 'BOD' value = 'true' DEBUG:root:key = 'BOR' value = 'true' INFO:root:Address for device ATSAMD21x18 is 0x2000
tidal kiln
@onyx hinge thanks! that was it. works now. thanks for the tip in the issue thread about how to get the artifact also!
onyx hinge
you're welcome, and thanks for doing testing!
analog bridge
Is there an alternative to esp_log stuff ?...I tried ESP_LOGI but it just doesn't seem to work for me.
onyx hinge
mp_printf(&mp_plat_print, "i=%d\n", i);
ESP_LOGI should appear on the debug uart depending on esp-idf configuration flags though
analog bridge
there does seem to be 5 levels of log, I don't know which level we are running in cpy.
lament sentinel
@lone axle @tulip sleet Version checking
Shorter output sorry about that INFO:root:Executing ['C:\\Program Files (x86)\\BOSSA\\bossac.exe', '-h'] DEBUG:root:returncode = 1 DEBUG:root:final_version 1.9.1 DEBUG:root:Comparing BOSSA version 1.9.1 to >=1.8.0 DEBUG:root:BOSSA version is cool
tidal kiln
i'm getting a hard reset on magtag when running the bitcoin fetch example a second time
slightly modified example to just print to serial
from adafruit_magtag.magtag import MagTag
DATA_SOURCE = "https://api.coindesk.com/v1/bpi/currentprice.json"
DATA_LOCATION = ["bpi", "USD", "rate_float"]
magtag = MagTag(
url=DATA_SOURCE,
json_path=DATA_LOCATION,
)
magtag.network.connect()
resp = magtag.fetch()
print("Bitcoin =", resp)
results:
Adafruit CircuitPython 6.1.0-beta.0-13-g3e6661bc9 on 2020-11-19; MagTag with ESP32S2
>>> import fetch_test
Connecting to AP ----
Retrieving data...Reply is OK!
Bitcoin = 18023.4
>>>
soft reboot
Auto-reload is on. Simply save files over USB to run them or enter REPL to disable.
code.py output:
Hello World!
Press any key to enter the REPL. Use CTRL-D to reload.
Adafruit CircuitPython 6.1.0-beta.0-13-g3e6661bc9 on 2020-11-19; MagTag with ESP32S2
>>> import fetch_test
Connecting to AP ----
hard resets after that
slender iron
@tidal kiln want to file an issue or try and debug it?
tidal kiln
i should probably just file an issue
take it i'm not doing anything obviously wrong? (just got magtag two days ago, so only just starting to kick the tires on the thing)
jaunty juniper
I'm getting a hard reset on the Feather S2 when connecting to wifi for the second time too
slender iron
@jaunty juniper what version are you running?
tidal kiln
@slender iron is info above enough to add to issue? or need anything more?
jaunty juniper
the 6.1.0 beta, but I have had that happen for a while, but it seemed intermitent and I had other problems to look at first
slender iron
the beta is pretty broken I think
we have a fix that we're waiting on the CI for
I think the last commit before it starts with a18 iirc
tulip sleet
is it the event loop deletion PR that would fix things?
slender iron
yup
manic glacierBOT
slender iron
too much good stuff 🙂
@tidal kiln could you try with the older a18 build?
your issue might be the same that beta.0 has
tidal kiln
a185549 the one right before the beta release tag?
tidal kiln
@slender iron a18 works
slender iron
👍 thanks!
tidal kiln
np
tulip sleet
@slender iron getting my alarm branch to compile... what was your last idea about the rgb led status animation? It returns a count of how many times it's been run? Originally you were returning a bool, but parts of the code are trying to return a count. Would the caller stop calling it after <n> cycles?
jaunty juniper
a18 indeed works for me too
tulip sleet
i will just do boolean for now, based on what the caller is currently trying to do
slender iron
@tulip sleet I'm not sure. don't read too much in the code I sketched out 🙂
hearty tapir
Hi, guess who just had the idea to order a magtag and now is pretty disappointed?! ... 😄
hearty tapir
This is not even being sold for double its price on eBay ... so the community really loves the product, you can tell 😄
lapis hemlock
Could anyone, please, run the following code on bare metal? import ulab as np print(np.array(range(2**8-5, 2**8), dtype=np.int8))
This doesn't work on the linux port, while is OK in micropython. I am trying to find out, whether this is a peculiarity of the circuitpython unix port, or more general, but I haven't a circuitpython board handy at the moment.
opal crystal
@lapis hemlock ```Adafruit CircuitPython 6.1.0-beta.0-11-g8ac9b176c on 2020-11-18; FeatherS2 with ESP32S2
import ulab as np
print(np.array(range(28-5, 28), dtype=np.int8))
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
OverflowError: value must fit in 1 byte(s)```
lapis hemlock
@opal crystal Thanks! The same as on the unix port.
This appears to be related to https://github.com/adafruit/circuitpython/issues/2579, which is a closed issue.
The error is most probably caused by a call to item = mp_binary_get_val_array(...); mp_binary_set_val_array(...);
To be a bit more specific, https://github.com/v923z/micropython-ulab/blob/31d0bec9fe58c5170865a7f77609cd89296b6eed/code/ndarray.c#L868-L871
Is searching the circuitpython code on github disabled? Whatever I do, I just get a listing of matching issues, but no references to the code.
manic glacierBOT
hearty tapir
@lapis hemlock same here, I just tried to search for "esp_wifi_start" and don't get any results in "Code". This appears to be a sideeffect of being a forked project (from micropython)
lapis hemlock
@hearty tapir This is quite bad.
lapis hemlock
@slender iron So, perhaps, it would be high time to just start from scratch, and open a new repository for circuitpython.
The two projects have considerably diverged, so there is not too much to lose.
manic glacierBOT
UPDATE: PR #3719 (now merged) fixes the imageload (bottom row) thresholding error; those items can be ignored.
WRT OnDiskBitmap: looking more carefully, it appears that black is transparent in all three cases: 4-bit paletted, 8-bit paletted and 24-bit RGB. (Compare the upper & lower 8-bit images with the #3719 fix, they should be identical.) Or…here’s a photo…
Looking a...
slender iron
@lapis hemlock no way. it's important for us to connect back to it and we intend on merging mp changes in again at some point
hearty tapir
@slender iron I noticed that my featherS2 is not using the "best Wi-Fi AP in reach" but chooses the one that it finds first. (channel 1 vs. channel 6)
The last time I reviewed this in the default configurations of ESP-IDF, it should choose the best AP in reach (RSSI-wise). Unless you'd tell me this was intentional?
onyx hinge
@lapis hemlock shouldn't the code should fail, because 251 is not representable as a signed 8-bit int?
lapis hemlock
@onyx hinge No, it is simply re-casting the byte. That is also the numpy behaviour.
lapis hemlock
It is similar in python int8(250)
returns -6.
We can conditionally remove the offending lines from the test, because ulab.__version__ already tells us, whether we are running in circuitpython, but I think it should be fixed at some point. I can open an issue, if you want.
onyx hinge
@lapis hemlock looks like we made a change in circuitpython that introduces this behavior (different from micropython) -- https://github.com/adafruit/circuitpython/pull/1879
.. to align with python3 array.array: ```Python 3.7.3 (default, Jul 25 2020, 13:03:44)
import array; array.array('b', (251,))
OverflowError: signed char is greater than maximum
Oh. So what do we do now?
onyx hinge
because you can only have one "expected" text, I'm not sure how you can use an is-circuitpython block in a testcase
lapis hemlock
I have overlooked this. So we are doomed, that's what you are trying to say, aren't you?
Well, we could have two sets of test scripts in two different folders.
That should fix it, right?
onyx hinge
Maybe in ulab, if CIRCUITPY, you need to use your own copied and pasted implementation without overflow checks, instead of calling mp_binary_set_val_array
lapis hemlock
We could have a hacked copy of the mp_binary_set_val_array function, if #CIRCUITPY is set, I am OK with that.
lapis hemlock
Desperate times call for desperate measures.
@lapis hemlock no way. it's important for us to connect back to it and we intend on merging mp changes in again at some point
@slender iron You could still have a non-editable, non-forkable mirror of your repository, so that it could be searched.
slender iron
The last time I reviewed this in the default configurations of ESP-IDF, it should choose the best AP in reach (RSSI-wise). Unless you'd tell me this was intentional?
@hearty tapir This is the default behavior afaik. You can do a scan and then tell connect what channel to start with from CP.
hearty tapir
@slender iron Ok, I got it. The API defaults to "WIFI_FAST_SCAN" which means : "Do fast scan, scan will end after find SSID match AP". The other option would be "WIFI_ALL_CHANNEL_SCAN" which again would default to the strongest AP, but this time it knows it.
slender iron
ah
hearty tapir
The typical use of connect, in all examples I've seen is connect(ssid, password)
The remaining/open question is just how many have multiple APs 😉
slender iron
we could do fast only if channel is provided
hearty tapir
(or if the bssid is provided)
slender iron
sure. I couldn't remember if we allow that
crimson ferry
raises hand has multiple APs with same SSID at home
hearty tapir
This is possible. I just try to suggest that if "channel" or "bssid" are not provided, the function should do a full scan instead of fast scan. I can look at that if you want 😉
but "me against C" is still a little bit of a struggle 😄
lapis hemlock
Maybe in ulab, if CIRCUITPY, you need to use your own copied and pasted implementation without overflow checks, instead of calling mp_binary_set_val_array
@onyx hinge This won't be enough: thenumpy-compatible andcircuitpython-compatible versions will have two different call signatures for functions. C.f.np.sin(...)vs.vector.sin(...).
I would just have two separate folders for the two cases.
crimson ferry
@hearty tapir even if channel is provided, may want to do a full scan since APs often (usually?) auto-select their channel and there's no way to predict any channel:RSSI relationship
onyx hinge
it's true, if you want to switch the micropython version that you test to use numpy-compatible conventions and you don't want to use import ulab.vector as vector except NameError import numpy as vector or something like that
lapis hemlock
It is enough to change the lines MICROPY_MICROPYTHON=micropython/ports/unix/micropython ./run-tests -d tests; in the build script, and point to the test-numpy, and test-cpy folders, right?
onyx hinge
yes most likely
onyx hinge
thank you !
hearty tapir
@crimson ferry I didn't see your comment. Pull-request 3727 will do full scan if channel/BSSID are not given
If you want to give this a try (given you have a featherS2), this is the binary: https://benny.servebeer.com/nextcloud/index.php/s/fKAKWAQaRFK8WJK
tidal kiln
is this where to go for the ESP32S2 toolchain?
https://docs.espressif.com/projects/esp-idf/en/latest/esp32s2/api-guides/tools/idf-tools.html#xtensa-esp32s2-elf
also seeing python scripts in github workflow?
crimson ferry
@hearty tapir This will be really nice to connect to the best RSSI by default. I did a little more reading and WIFI_FAST_SCAN can't sort by RSSI anyway. I'll load it up on a few devices when some of the other changes get merged in too. Thanks!
hearty tapir
@tidal kiln I went through the same. I just ended up going for ./install.sh in the ESP-IDF folder that you find after the fork / git clone. This will give you all you need and you can get it back by ". ./export.sh" as well for the next shell
tidal kiln
thanks. "get it back"? it doesn't install globally via the script?
hearty tapir
It does, but when you shut down your computer / restart whatever - you have a new shell. You then do ". ./export.sh" in that ESP-IDF folder
there is probably a way to automate this, but I simply do it manually when needed.
lament sentinel
@lone axle @tulip sleet https://github.com/bcr/blinka-cli/blob/master/commands/bossa.md
orchid basinBOT
slender iron
well that language list isn't right 🙂
lament sentinel
Glad you finally released en_US
slender iron
at least the script ran this time 🙂
lone axle
@lament sentinel nice! maybe a --dryrun or similar param would be helpful too. So it will do everything up till the final flash and then just print a message and stop