and also sha256.c is in sveral places, including crypto-algorithms, and inside mbedtls
#circuitpython-dev
tulip sleet
onyx hinge
micropython has https://github.com/ARMmbed/mbedtls.git while pico-sdk has git@github.com:Mbed-TLS/mbedtls.git
tulip sleet
i thought they would be tracking pico-sdk since they were paid for the impl, but clearly not
onyx hinge
these two mbedtls versions diverged in 2019 😕
onyx hinge
I have no idea about the politics of any of this, mostly I tried to follow what MP was doing. note that they get their cyw43-driver ALSO not from pico-sdk via nested submodule but put it in <top>/cyw43-driver
tulip sleet
and that is still the case? ok, I will not do what I was doing then, and follow their lead instead
i thought we did that only
onyx hinge
CMakeLists.txt: set(PICO_CYW43_DRIVER_PATH ${MICROPY_DIR}/lib/cyw43-driver)
```I believe that they are using their own cyw43_driverthey have updated to pico-sdk 1.5.0 already or are about too. I will check out what they did. Sounds like it took a day just to track all this down.
onyx hinge
yes a few weeks ago they updated to 1.5.0
tulip sleet
thanks for your memory on this
onyx hinge
You're welcome, of course.
tulip sleet
sdk/src/rp2_common/cyw43_driver/cyw43_bus_pio_spi.pio is now .../pico_cyw43_driver/...
onyx hinge
I assume you found https://github.com/micropython/micropython/pull/10764
tulip sleet
i was going there, will do what they did for now
onyx hinge
they must have had a different & more correct way to find that pio file, I don't see any change related to that relocation
tulip sleet
i will check that
and the firmware .h include file
updates to drive repo in the past few days... https://github.com/georgerobotics/cyw43-driver/commits/main
maybe I will wait on driver repo update now until the churn is finished, and just do the minimum to get it to compile with 1.5.0.
and then test
tulip sleet
they must have had a different & more correct way to find that pio file, I don't...
it's in the cmake:
lib/pico-sdk/src/rp2_common/pico_cyw43_driver/CMakeLists.txt
50: pico_generate_pio_header(cyw43_driver_picow ${CMAKE_CURRENT_LIST_DIR}/cyw43_bus_pio_spi.pio)
I think we currently avoid cmake except for the step of building pioasm only; they have used more cmake during their build process
I appreciate you working on this, thank you.
tulip sleet
looks like I will need to update the driver at least to the 1.5.0 PR above, because of renamings. I'll investigate what the latest changes are as well.
onyx hinge
hum it doesn't seem possible to use any font but terminalio.FONT with a terminalio.Terminal. 😕
(because it must be of type BuiltinFont, but you can't create instances of it)
onyx hinge
That's interesting. Because of this line in ko.po, which is not ever going to be shown, the non-ASCII characters are included: "Last-Translator: Michal Čihař <michal@cihar.com>\n"
tulip sleet
That's interesting. Because of this line in `ko.po`, which is not ever going to ...
ha!
i think we are not building ko now anyway?
onyx hinge
I noticed that BuiltinFont explicitly lists each Unicode code point it contains; I thought maybe it would be helpful to allow ranges to be expressed directly (by bytes with the high bit off) but no translation actually seems to benefit after adding the required new logic to fontio_builtinfont_get_glyph_index
lines that are po-file comments are used too
as a consequence, the zh_Latn_pinyin translation includes the smart quote characters in its font. this only costs a few bytes though.
@tulip sleet we only disable terminalio for ja/ko/ru on some samd boards, I was looking at rpi builds
slender iron
Ya, I think it can happen on any port. ESP using FreeRTOS means that something else is checking for stack overflows in addition to CP.
manic glacierBOT
chilly jay
Hey, is there a way to debug my board using the serial connection (with the vscode addon)?
I currently am doing things that crash the board hard (playing around with FreeRTOS tasks on my unexpectedmaker feathers2) and I'd like to know why it crashes.
slender iron
I'm not sure about the vscode piece but you can definitely get the ESP_LOG output to a uart with a DEBUG=1 build
debug builds will print backtraces from the panic handler
chilly jay
I'm not sure about the vscode piece but you can definitely get the ESP_LOG outpu...
I guess I add DEBUG=1 to the make command with which I am building the firmware?
slender iron
yup, clean first too
chilly jay
Since there are also debug settings in the sdkconfig
slender iron
DEBUG=1 will change what sdkconfig is used for the CP build
chilly jay
ohhh nice
onyx hinge
how do I put back the original terminal Group in 8? setting the root_group to None isn't doing it, despite the documentation...
show has special handling for this but set_root_group doesn't .. why?
bool common_hal_displayio_display_show(displayio_display_obj_t *self, displayio_group_t *root_group) {
if (root_group == NULL) {
root_group = &circuitpython_splash;
}
return displayio_display_core_set_root_group(&self->core, root_group);
}
is it simply an oversight?
jaunty juniper
`show` has special handling for this but `set_root_group` doesn't .. why?
```
b...
display.root_group = displayio.CIRCUITPYTHON_TERMINAL
I overlooked that, thank you!
jaunty juniper
dunno if it's documented yet ? It should probably be mentioned in the root_group help ?
timber mango
Can anyone explain me what the idea of the redirects.txt is? Does it influence the build process?
timber mango
https://github.com/adafruit/circuitpython/blob/main/docs/redirects.txt --> This file i assume it is for generating documentation?
My issue is that I added a simple shared-module / binding, and I see the compiler having the compliation flag set to 1, but in the actual compile the files are not compiled. I followed the makefile changes made on this pull request, but maybe I'm missing something https://github.com/adafruit/circuitpython/pull/6757
slender iron
https://github.com/adafruit/circuitpython/blob/main/docs/redirects.txt --> This ...
Ya, thats for redirecting old urls to a new location
can you push your changes somewhere?
there are .mk files in py/ that usually need the source files added
timber mango
I will upload it to my github and send you a link in a minute
@slender iron https://github.com/adafruit/circuitpython/compare/main...binloan:circuitpython:build_test_0306
my goal was to get the compiler to compile the module in order to see if my setup is correct, but as mentioned it does not seem to regard the module at all
slender iron
@timber mango it looks right to me
timber mango
that's weird - if I run make in verbose I get the -DCIRCUITPY_EGOBUS=1 flag, but in the list of compiled files the init.c is not showing up
slender iron
did you try cleaning and rebuilding?
timber mango
I will try
It does fail
mkdir -p build-adafruit_feather_huzzah32/genhdr
GEN build-adafruit_feather_huzzah32/autogen_web_workflow_static.c
make: *** No rule to make target shared-bindings/egobus/__init__.c', needed by build-adafruit_feather_huzzah32/genhdr/qstr.split'. Stop.
so does that mean the file is compiled? if I go in the built firmware REPL trying to import egobus, it does not work tho (gives me an error the module does not exist)
slender iron
ya, I think it does get compiled
looks at your code again
MP_REGISTER_MODULE should be what you need to be in the module top
build-*/genhdr/moduledefs.h should have it
timber mango
MP_REGISTER_MODULE should be what you need to be in the module top
not sure if I understand, move it at the top of the file?
timber mango
you know any module that does it that way? the ones I saw always had it at the bottom - i will try anyhow tho
slender iron
I don't think it matters because there is a python script that looks for it in the whole file
timber mango
ok - I will try to look inside the moduledefs.h once compilation finished
slender iron
if its there then I'd expect you to be able to import it
timber mango
I don't know what happened, but I get an error in building now inside the module
seems like the make clean did not clean enough for me or some cache was still there
@slender iron thanks for your help!
timber mango
yeah seems like 😅
thorny jay
Hi, I am looking for the name (or learn guide) of a UF2 for the Pico that turn it into a kind of FT232 or CP210x and can be used as external GPIO for a PC/Mac that would run Blinka.
I am pretty sure it exist (and not a dream) but I fail to put a name on that or to have the right search keyword to find it.
Notice that I might confuse (a bit) with similar concept by Raspberry Foundation that turned a PiZero into external GPIO accessible with GPIOZero library... same concept.
ANSWER: u2if
timber mango
@slender iron looking at my previous commit I get the following build error:
In file included from ../../shared-bindings/egobus/init.c:34:
../../shared-bindings/egobus/EgoBusNodeIDs.h:7:28: error: 'tNodeId' redeclared as different kind of symbol
extern const mp_obj_type_t tNodeId;
^~~~~~~
In file included from ../../shared-bindings/egobus/EgoBusNodeIDs.h:5,
from ../../shared-bindings/egobus/init.c:34:
../../shared-module/egobus/EgoBusNodeIDs.h:61:2: note: previous declaration of 'tNodeId' was here
}tNodeId;
^~~~~~~
In file included from ../../shared-bindings/egobus/init.c:28:
../../shared-bindings/egobus/init.c:38:54: error: 'egobus_bufferedin_type' undeclared here (not in a function)
{ MP_ROM_QSTR(MP_QSTR_BufferedIn), MP_ROM_PTR(&egobus_bufferedin_type) },
^~~~~~~~~~~~~~~~~~~~~~
../../py/obj.h:312:24: note: in definition of macro 'MP_ROM_PTR'
#define MP_ROM_PTR(p) (p)
^
make: *** [build-adafruit_feather_huzzah32/shared-bindings/egobus/init.o] Error 1
I'm not sure if I can fully follow that error message. I assume based on the files I saw until now that I need to declare the type both in the bindings and modules or did I get that wrong?
tulip sleet
@onyx hinge Anne and I are debugging memory issues with OnDiskGif. MP3Decoder has a superficially similar API: you can pass it a filename or an open file. It has a deinit(), and OnDiskGif does not. OnDiskGif probably should. But where do you close an open file in MP3Decoder? I am not seeing that. I guess it's the file destructor/finalizer?
onyx hinge
@tulip sleet you're not missing anything. the file never gets closed. On standard computers, where there's a file descriptor / file number abstraction you have to close a file or that resource leaks. but, from what I could understand in CP, the file object can simply be collected by the GC and does not need to be explicitly closed. I felt it was OK inside core code to rely on an implementation detail like that.
tulip sleet
sounds ok, thanks, we'll recommend the same for gif
manic glacierBOT
onyx hinge
so this is all that is done to close the file: self->file = NULL; and when opening a fresh mp3 to the same decoder object, this is all that's done: self->file = file;
slender iron
@timber mango the type is in bindings and the obj_t goes in shared-modules
the obj_t is the instance
blissful pollen
@onyx hinge @tulip sleet Anne was asking me about the GIF memory earlier today so if you spot something I may have time to fix it up tomorrow. Hadn't had a chance to look into it yet though
tulip sleet
<@426200547024961539> <@329766224093249548> Anne was asking me about the GIF mem...
it looks like you'll need to add a .deinit(). MP3Decorder is superficially similar in a number of ways, and has a deinit() to free up stuff. Just removing all the references to the OnDiskGif object before gc-ing. is not good enough. Sometimes it works and sometimes it doesn't. del() is not a reliable way to remove references, I think.
timber mango
<@456226577798135808> the type is in bindings and the obj_t goes in shared-modul...
oh I see, thanks
blissful pollen
it looks like you'll need to add a `.deinit()`. `MP3Decorder` is superficially s...
Should be easy enough to do. I'll look at MP3Decoder for reference.
Only thought is if the bitmap is currently being displayed in a TileGrid, not sure how things will react (badly is my guess)
tulip sleet
<@426200547024961539> <@329766224093249548> Anne was asking me about the GIF mem...
Test prog:
import gifio
import gc
for i in range(1000):
print(i)
odg = None
gc.collect()
print("before odg =", gc.mem_free())
odg = gifio.OnDiskGif("/some.gif")
print("after odg =", gc.mem_free())
.next_frame() and .bitmap don't seem to cause problems. Or, they are not necessary to have the problem caused 🙂
onyx hinge
memory fragmentation will probably prevent allocating the large bitmap that is required
tulip sleet
yeah but in the above program after a hard reset the gc.collect() does not free anything. If you start the above program from the REPL (sometimes!) it works, but in other cases it does not
onyx hinge
but since the new image could have different resolution, I'm not sure you can add 'open different image' like I was able to add 'open different mp3' to the mp3decoder
tulip sleet
yeah, don't reuse
i was also seeing how the file was closed, but discussed above and that does not seem to be the issue
blissful pollen
The changing resolution wouldn't allow the bitmap to be reused sadly
tulip sleet
(it isn't closed)
onyx hinge
n = "x%d" % i
f = vfs.open(n, "w")
f.write(n)
f = None # release f without closing
[0, 1, 2, 3, 4, 5] # use up Python stack so f is really gone
gc.collect() # should finalise all N files by closing them
``` micropython/circuitpython's own test for gc things does odd gyrations to try to ensure that other stack/registers can't continue referring to an object that is intended to be freed. I don't know why a list of 6 ints makes a difference to that test (extmod/vfs_fat_finaliser.py) but it canI'm not seeing that test program lose memory myself. On a KB2040
tulip sleet
i am using a pyportal
sometimes it does and sometimes it doesn't
starting it from a hard reset, it does
blissful pollen
I'll try it on my feather M4 later. I have a ESP32-S2 set up as well (but less likely a problem there)
tulip sleet
I am using nyan.gif, prety big
onyx hinge
we have common_hal_displayio_bitmap_construct_from_buffer -- OnDiskGif could become a more generic multi-gif loader if it took a max_pixels constructor parameter, allocated a buffer that size, then when an image is actually opened it makes it a Bitmap of the appropriate WIDTH×HEIGHT, assuming it fits. The immediate 'leak' aside, my experience with mp3 playback makes me believe something ilke this will be needed to create player that can play multiple GIFs and not run into memory fragmentation.
the old Bitmap object would remain valid to refer to, but its content would reflect the new GIF data of possibly different dimensions
blissful pollen
Can someone put the nyan.gif here so I can test with it? Still can't trigger the issue. Though I'll have to resume looking at it later running out of time now
I don't own a PyPortal so can't test on device but the FeatherM4 should be close
tulip sleet
tulip sleet
I don't own a PyPortal so can't test on device but the FeatherM4 should be close
right, not necessary to display exercise the error. I was seeing very odd things, like adding an import board cuased the problem to go away, and then removing it again, came up. Seems like some field is uninitialized.
ornate breach
blissful pollen
Thanks for the gif. I'll look the code over but likely either tomorrow or the weekend. As always feel free to tag me if you do find something.
stuck elbow
and gameboy
blissful pollen
One thing just to note the nyan gif is right on the limit of what the library was set for (I think max width is 320). On the base Feather M4 there is not even enough memory to load it. It takes 153K. But this may be tomorrow's problem now.
brazen hatch
This can do proper ansi on android
This is the only one that works with ljinux
I found it today on fdroid
I hate the kb tho
rugged spindle
@brazen hatch I swear by JuiceSSH on Android - I have you tried that?
brazen hatch
Hmm no, never heard of it, but I have now, imma test it in a moment
Thanks for the suggestion
manic glacierBOT
brazen hatch
<@614096041083404300> I swear by JuiceSSH on Android - I have you tried that?
Aurora store says app not purchased even though it seems like a free app and it isn't on fdroid, soo... nope
Is there a way to use esp8266 as wifi for a CP board?
I am looking at ESP32SPI but that isn't for esp8266..
tulip sleet
Is there a way to use esp8266 as wifi for a CP board?
Deprecated library: https://github.com/adafruit/Adafruit_CircuitPython_ESP_ATcontrol esp8266 has problems including outdated certificates, unreliability, etc. Not really worth your time.
brazen hatch
Bummer, thanks for the info though!
Perhaps it would be possible to use micropy as a stepping stone
But i'm not gonna bother
I dug up a relic of the past
2019 feels like yesterday, but it really is 4 years..
So much has changed..
I honestly feel like keeping it like this, I have no use for it anyways.
tulip sleet
it becomes more historically interesting the older it gets
stuck elbow
I will just leave this here: https://github.com/lgersman/make-recipes-in-different-scripting-languages-demo/tree/main/python
midnight ember
I've run black on the file but CI keeps saying it's failing black.
stuck elbow
does it say which file?
midnight ember
not sure if that's something that can be seen by anyone other than me?
Error: black....................................................................Failed
- hook id: black
- files were modified by this hook
Error: reformatted examples/requests_api_steam.py
i ran black on that file though. and tried to run it again but black is saying it's fine
C:\Users\Devon\Documents\GitHub\Adafruit_CircuitPython_Requests\examples>python -m black requests_api_steam.py
All done! ✨ 🍰 ✨
1 file left unchanged.
did you commit that file? it doesn't seem to be included in the modified files?
midnight ember
yes, submitted the file after the changes.
stuck elbow
some other files are modified that shouldn't, though
midnight ember
github desktop automatically recognized file changes and creates a new one, all i have to do is submit it.
midnight ember
i did pip install black --upgrade before i ran it was it a mistake to update black?
hmm not sure how to understand what i'm looking at there. the github token stuff would have been with the github example.
stuck elbow
I don't think those are the changes you want to make
midnight ember
yeah those are not my changes...
maybe those are differences from the old versions that didn't fetch correctly into my local repo?
i don't even see the file i actually want to commit in those changes anyway. it's weird.
i fetched origin, pushed a commit, and i don't see the commit.
midnight ember
Figured it out I think. Do I have to close PR's on my side after they're accepted?
Seems like it was trying to push into an old PR somehow. Closed the PR. Opened a new one and everything went through fine.
blissful pollen
@tulip sleet @onyx hinge is there an issue (or maybe should there be?) about the OnDiskGif memory? I've been looking into it and the short summary seems that it only occurs on the M4 and seems to be "something" with the garbage collector vs the other ports. I've never looked at that code before so really outside my expertise.
onyx hinge
@blissful pollen I do not see that there's been a PR opened in the core about it yet.
er an issue, sorry
at my old job that I left 4 years ago, in the ticket system that was retired 15 years ago, they were called "problem reports" or PRs, so I sometimes say PR when I mean issue which is REAL CONFUSING
blissful pollen
<@228322543507865601> I do not see that there's been a PR opened in the core abo...
I'll open one up then and can put a longer post with more thoughts there.
Ha oops! Yeah I end up calling them defects because that is what I used to use. But at least that is less confusing then PR
onyx hinge
it was just fine until Pull Requests came along
tulip sleet
@analog bridge random q: sometimes people make a PR to circuitpython from their own main branch. Even if they have changed only some board-specific files, it seems to cause a full build. Example: https://github.com/adafruit/circuitpython/pull/7705 (building right now). Do you understand why?
#7695 is another example
or maybe the selective builds are accidentally broken at the moment
manic glacierBOT
analog bridge
<@747673663419187281> random q: sometimes people make a PR to circuitpython from...
its the merge commit, the diff is between the latest (head) commit in pr and an earlier commit in the pr head branch
if there is a merge commit in between those changes are counted as well
tulip sleet
ah, got it, thanks, I will just keep reminding people not to do that
analog bridge
ah, got it, thanks, I will just keep reminding people not to do that
I have thought of compiling conditional jobs based on the merge ref alongside the current workflow and then choosing the lesser of two, just haven't gotten to implementing it.
tulip sleet
I have thought of compiling conditional jobs based on the merge ref alongside th...
noticeable amount of work for an edge case. Right now I sometimes just cancel the job to save time.
analog bridge
thinking more about it, grabbing the common files looks like a better option
it will filter out changes introduced by the merge commit and will only include commit range specific changes, best of both worlds
onyx hinge
@tulip sleet @timid bolt just ran across 'anyio' which abstracts over asyncio or trio, I wonder if anything about it can help inform async work in CP or help us build the new concepts like task groups over an asyncio-ish base instead of a trio-ish base. https://pypi.org/project/anyio/
Asynchronous file I/O (using worker threads)
Well that's not going to work so hot for us
tulip sleet
<@329766224093249548> <@1065334137243836517> just ran across 'anyio' which abstr...
I looked at this at some length, but it was a lot of code. TaskGroup is being implemented in MicroPython and we can take that when it is done, or adapt the CPython one.
I also looked at porting trio, but it was -also- a lot of code, and I was not sure what to trim. But trimmed, it might not be any bigger than @timid bolt's asyncio.
I am not sure about the long-term viability of trio, given that asyncio now has TaskGroup
trio may have served its purpose by pushing asyncio in a particular direction
onyx hinge
Makes sense, just wanted to share in case it was new info
tulip sleet
i appreciate it -- this has been difficult path-finding. Lots of overgrown trails
analog bridge
noticeable amount of work for an edge case. Right now I sometimes just cancel th...
Resolved in #7708.
analog bridge
Can we turn off espcamera on adafruit_feather_esp32s3_reverse_tft? It is failing in main.
brazen hatch
C3 IS CURSED C3 IS CURSED C3 IS CURSED C3 IS CURSED C3 IS CURSED
nuked it with idf, now it doesn't boot
amazing
brazen hatch
Now it boots on it's own, but when I try to debug it, it immediately goes:
Info : [esp32c3] Halt cause (7) - (PMP Store access fault)
I'm out.
tulip sleet
Can we turn off `espcamera` on `adafruit_feather_esp32s3_reverse_tft`? It is fai...
you could make it the same as the regular TFT
slender iron
@tulip sleet have time to chat performance on imx?
tulip sleet
sure
brazen hatch
Can someone other than me try to debug esp for pystack?
I seriously am going to lose my sanity over this.
C3 is going to break me.
I have no idea how to debug S2 and rp2 is not crashing.
midnight ember
Can someone other than me try to debug esp for pystack?
I seriously am going to ...
I felt that in my soul. Can relate.
blissful pollen
If memory is allocated via m_malloc I know it can be cleared by the GC but is there another proper way to free it if you know it will be unused?
timid bolt
Can someone other than me try to debug esp for pystack?
I seriously am going to ...
I could look, but (1) I only have this board https://www.adafruit.com/product/3405, will it repro on that? and (2) won't have time until mid next week.
timid bolt
If memory is allocated via `m_malloc` I know it can be cleared by the GC but is ...
m_free
slender iron
@tulip sleet pico results look much more similar
(cp actually looks a bit faster in some cases)
brazen hatch
I could look, but (1) I only have this board https://www.adafruit.com/product/34...
Well idk if the bug is present on it, but I suppose it will, all other esp's are affected..
I cannot get it debugged anyways so yea.
tulip sleet
<@329766224093249548> pico results look much more similar
That is a relief!
tulip sleet
That will be interesting. I find its architecture so complicated.
tulip sleet
ya, but now the question is why imx is slower
maybe use svd to dump out the memory controller settings (and other relevant "peripherals") on both mpy and cpy
slender iron
totally
running mp on the 1010evk right now
I did realize I turned off caching the regular ram
I think I needed that for usb to work though
wraith crow
@brazen hatch @timid bolt I'm not seeing a crash on the Huzzah32 with a value up to 15000. setting it to 20000 resulted in a memory allocation error but no crash.
brazen hatch
def test(no):
print(no)
try:
test(no)
except RuntimeError:
print("Done.")
while True:
test(0)
let it rip for a minute
if this passes with a large pystack the board is good
this is what it takes for S2 to crash
one or two runs won't crash it
no matter how big the stack
wraith crow
did you mean test(no+1)?
wraith crow
Okay, crash reproduced
wraith crow
It doesn't look like you even need to loop the test(0) call to cause the hang, it seems to be hanging pretty reliably on the Huzzah32.
Interestingly, it hangs after printing "Done" and presenting the REPL prompt
brazen hatch
then it's the same as C3
wraith crow
The breakpoint is somewhere between 3500 and 4000
slender iron
lol
wraith crow
😆
brazen hatch
Honestly, yea. If a pc can't even do .5A it's ewaste
wraith crow
yep, wouldn't want to trade out my $25 microcontroller, need a new computer....
brazen hatch
Just wait till I do vga on ljinux
Make the microcontroller your pc
livin the 144p life
wraith crow
Have you tested on the MIMXRT boards yet?
brazen hatch
Nope, aren't those the teensy's?
brazen hatch
Expensive + microusb doesn't sit right with me.
My teacher also suggested me to get one. But I am still hesitant..
timid bolt
re the pystack issue: does the crash only happen when the stack overflows, or does it also crash at other times? Because crashing on stack overflow is not a serious bug.
wraith crow
I'll try a debug build and see if I the Huzzah indicates a stack overflow
brazen hatch
it's 1:20am so idk how much more I will sit testing
I'm just guestimating at this point
without debugging i'm absolutely incapable of actually pinpointing the fault
brazen hatch
Some random guesses for me were to try different values and multiples.
Multiples of 12 is a big nope.
I have no clue how big is a stack frame.
1536 is the stock value which works
multiples of it don't work either
so it's not that
Well I quit for today.
Goodnight.
wraith crow
The Debug build didn't output the error message over the micro usb connection. I'll take a look later tonight and see if I'm getting anything over the rx/tx pins. I did grab the latest bits for the debug build and this build no longer crashes on every run of the test script. Looping the test call does cause the crash but it now hangs in the midst of the test routine rather than after it completes.
Well maybe not in the midst, it hangs just prior to printing Done
brazen hatch
I still do not know how to have it print over uart, I will try that tomorrow since I can just read it from my pi400
tame grove
hey hey!
I've been wondering how feasible it is to write circuitpython modules in Rust… Any pre-existing projects/ideas there? I know my way around bindgen and embedded Rust, have recently picked up circuitpython again because it's just so ridiculously convenient for initial R&D, but some stuff just needs that native code speed
blissful pollen
hey hey!
I've been wondering how feasible it is to write circuitpython modules i...
I think it has been looked at but no one has done it yet. Adding a loadable module (vs compile time) is trickier as well. I experimented with that in C and if you look up the native module support you can see how it is done.
tame grove
I think it has been looked at but no one has done it yet. Adding a loadable modu...
mhm. I think I'd be fine with compile time as a first POC. didn't even know you could dynamic load native code! thought that was (m)py only. Sweet!
blissful pollen
mhm. I think I'd be fine with compile time as a first POC. didn't even know you ...
Yeah the native code does work though it is a limited subset (just like in micropython). From what I recall others saying Rust support for compile time is theoretically possible but just not done.
brazen hatch
Well I will buy a knock-off esp32s3 off of ebay.
It's not a clone of another board, but a different thing entirely.
So perhaps it could receive mainline cp support.
The thing is that all sold locally are ewaste (no psram), buying from ada directly is a unimaginably more expensive due to shipping costs, and ebay is only one left.
I'm going for N16R8. It only costs a few bucks more than N8R8
Let's just hope it's not a 20€ brique.
manic glacierBOT
Hooked up debug uart of S2 to my pi400, produced a debug build, and nothing. It doesn't crash now, even with a 12k stack.
My best guess for right now is that it's an optimisation error.
brazen hatch
In the case it is actually an optimisation error what do I do?
brazen hatch
Can I make CP expose the JTAG pins the same way UART is exposed?
I mean
MTDO / GPIO40 -> TDO
MTDI / GPIO41 -> TDI
MTCK / GPIO39 -> TCK
MTMS / GPIO42 -> TMS
esp32s2
tulip sleet
I mean
```
MTDO / GPIO40 -> TDO
MTDI / GPIO41 -> TDI
MTCK / GPIO39 -> TCK
MTMS /...
These are available as microcontroller.pin.GPIO40, etc. They are Pin objects just like board.<some-pin>
brazen hatch
I know, but how do we expose them as jtag instead?
I can't seem to find it in any board definition
tulip sleet
jtag pins are attached to things inside the chip. I'm not sure I understand -- do you want to emulate jtag? This is jtag in, not jtag out
do you want to attach these pins to the jtag pins on some other board, to use the ESP32S2 as a debugging probe?
brazen hatch
Oh no, I was just hoping to debug S2..
tulip sleet
on a Metro S2, there is an SWD connector you can hook up a debugging probe to. This works regardless of what is running on the board
brazen hatch
Well I have ordered the S3 knockoff now anyways, so in about 2 weeks I will be able to debug that.
tulip sleet
I have not successfully used openocd or J-Link to do ESP32-S2 debugging. It is flaky.
brazen hatch
on a Metro S2, there is an SWD connector you can hook up a debugging probe to. T...
Why don't more boards just do swd.. it's so much simplier than jtag..
tulip sleet
it has to be built into the chip
i don't know why espressif chose jtag over swd
jtag has a bunch of other uses as well, like debugging the hardware design
analog bridge
@brazen hatch might be helpful https://blog.wokwi.com/gdb-debugging-esp32-using-raspberry-pi/
brazen hatch
Wait
tulip sleet
@analog bridge I tried that, but not very successfully. Did it work for you?
very limited breakpoints, etc.
analog bridge
@tulip sleet same experience, it was flaky and limited in functionality
tulip sleet
they are for debugging the ESP32S2
brazen hatch
Well, I have openocd ready on my pi400 already, so I may as well test it.
brazen hatch
very limited breakpoints, etc.
rp2 is also very limited, but that didn't stop me from debugging picow-ap
I take anything over C3
brazen hatch
<@614096041083404300> might be helpful https://blog.wokwi.com/gdb-debugging-esp3...
Yea the circuitpython takes the pins, so jtag doesn't work
I set the whole thing up
I needed a different openocd
compiled it
fixed bugs
made a pr
jtag is not working with it though
[bill88t@Pi400 | ~]> sudo openocd -f interface/raspberrypi-native.cfg -f target/esp32s2.cfg -c "adapter speed 1000"
Open On-Chip Debugger v0.12.0-esp32-20230221 (2023-03-11-16:50)
Licensed under GNU GPL v2
For bug reports, read
http://openocd.org/doc/doxygen/bugs.html
Warn : TMS/SWDIO moved to GPIO 8 (pin 24). Check the wiring please!
adapter speed: 1000 kHz
Info : Listening on port 6666 for tcl connections
Info : Listening on port 4444 for telnet connections
Info : BCM2835 GPIO JTAG/SWD bitbang driver
Info : clock speed 1000 kHz
Error: JTAG scan chain interrogation failed: all zeroes
Error: Check JTAG interface, timings, target power, etc.
Error: Trying to use configured scan chain anyway...
Error: esp32s2.cpu: IR capture error; saw 0x00 not 0x01
Warn : Bypassing JTAG setup events due to errors
Warn : [esp32s2] Unexpected OCD_ID = 00000000
Warn : target esp32s2 examination failed
Info : starting gdb server for esp32s2 on 3333
Info : Listening on port 3333 for gdb connections
Tried flipping DI DO pins, nada
Imma just wait for my S3 board.
solar whale
Anyone know anything about this? Is it hapenning? https://blog.adafruit.com/2023/03/09/free-circuitpython-workshop-saturday-march-11th-circuitpython/?
guess not ...
tame grove
Yeah the native code does work though it is a limited subset (just like in micro...
limited in what way exactly? (I'd have assumed whatever gets baked in at compile time still looks like a normal python module)
blissful pollen
limited in what way exactly? (I'd have assumed whatever gets baked in at compile...
Oh sorry I meant the native submodules you load at run time are limited. If it is at compile time I don't see what limits there would be
tame grove
Oh sorry I meant the native submodules you load at run time are limited. If it i...
I meant >>> import baked_in_mod - does this have limits?
I suppose you can't e.g. do >>> baked.some_nonexisting_attr = foo ?
jaunty juniper
so I am working on adding a board with a builtin OLED display that shares the same pins as board.STEMMA_I2C, and I can't instantiate the bus since it is in use. The TFT feather, which uses board.SPI, does not have that problem, do I need to do something special to enable that with the I2C bus ?
ah nevermind I got it
lone sandalBOT
lone sandalBOT
thorny jay
ah nevermind I got it
Is it LOLIN S2 Pico by LOLIN?
jaunty juniper
I'm adding the three 01Space QT PY sized ones with oled displays (that they call LCD but...)
https://fr.aliexpress.com/store/912632218
the 0.42
blissful pollen
Anyone ever have success debugging an ItsyBitsy M4 via SWD? Using device ATSAMD51G19A.
I cannot get it to connect no matter what I try. I did hook the JTag to a Feather NRF52840 so I know the debugger works.
tulip sleet
Anyone ever have success debugging an ItsyBitsy M4 via SWD? Using device ATSAMD5...
I debug a Metro M4 via SWD often. That's an ATSAMD51J, but I'd expect it to work on ATSAMD151G as well. DO you ahve the reset line connected?
blissful pollen
I do not have the reset line connected. That goes to the enable pin?
blissful pollen
I debug a Metro M4 via SWD often. That's an ATSAMD51**J**, but I'd expect it to ...
Thanks, that fixed it. Next step get this all working from my WSL environment (but at least I know the issue there)
tulip sleet
I do not have the reset line connected. That goes to the enable pin?
it goes to RST
thorny jay
I'm adding the three 01Space QT PY sized ones with oled displays (that they call...
I love those tiny OLED: SSD1306,
tulip sleet
enable might work too if there is no-broken out RST, but I thought there was a pad or pin
have to be afk now
brazen hatch
I love those tiny OLED: SSD1306,
oled + tiny + i2c + cheap
These things are awesome.
blissful pollen
enable might work too if there is no-broken out RST, but I thought there was a p...
Thanks Dan. That all worked. I documented all I did to get it running in WSL too so I don't forget next. Now to remember how GDB works ha Been a while
devout jolt
rp2040 PIO rp2pio.StateMachine question: is there a reason why some of the state machine config settings like sm_config_set_jmp_pin() could not be made adjustable after a StateMachine is created? It's unclear to me reading the pico-sdk, rp2 micropython sdk, or the rp2040 datasheet if a state machine can be reconfigured after it's running. But it would really be handy! (I'm trying to make a multi-pin version of touchio, I have a single-pin version I call touchpio that works great but requires "jump pin")
manic glacierBOT
solar whale
@gilded cradle I used the "installer" on circuitpython.org to load CP onto a feather esp32-v2. It worked great!! That is a very nice addition.
inner jetty
```
[bill88t@Pi400 | ~]> sudo openocd -f interface/raspberrypi-native.cfg -f tar...
Hi
In the tutorial post, there is a suggestion to lower the adapter clock. You could try with 500, 200 or 100 kHz...or maybe more like 2000 Khz as I see with some esp32 boards . see the example in the link: https://learn.adafruit.com/adafruit-metro-esp32-s2/debugging-with-openocd
If you see an error, check your connections. You can also try setting a lower adapter_khz value in the openocd command. 1000 works well for me, but your mileage may vary.
inner jetty
Hi
In the tutorial post, there is a suggestion to lower the adapter clock. You ...
According to https://docs.espressif.com/projects/esp-idf/en/latest/esp32s2/api-guides/jtag-debugging/index.html and based on your logs
If you see JTAG errors (e.g., ...all ones or ...all zeroes), please check your JTAG connections, whether other signals are connected to JTAG besides ESP32-S2’s pins, and see if everything is powered on correctly.
brazen hatch
Hi
In the tutorial post, there is a suggestion to lower the adapter clock. You ...
Yea I will just wait for my S3 board. It ain't worth the trouble. Thanks though.
digital shoreBOT
blissful pollen
Quick (hopefully) question. I have GDB debugging set up and going now but I am seeing <optimized out> in a lot of places. Not sure if there is a flag I can pass to a build to remove some of these. What is confusing to me I may see a value in my current frame but the one up has 90% of the locals that seemed to generate the value optimized out so not sure where the value came from.
tulip sleet
Quick (hopefully) question. I have GDB debugging set up and going now but I am s...
you can compile -Og or -O1 or -O0 and see less of that. You will probably need to do CIRCUITPY_ULAB = 0 to make room for the larger code. You can also make local variables volatile, which will prevent them from being optimized out.
blissful pollen
you can compile -Og or -O1 or -O0 and see less of that. You will probably need t...
Thanks. Where do I add the compiler flags? Is there someplace in the makefile for it? (I do see the if DEBUG=1). I thought I had -O0 set but then I was 189Kb oversize
slender iron
optimized out can happen if you are before or after a variable is used too
tulip sleet
<@&356864093652516868> Meeting today at 2pm US Eastern time. Note that US Daylight Savings Tme started yesterday, so if you are not in the US and have not started DST yet, you may find the meeting to be one hour earlier right now in your local time zone. Add your hug reports and status reports to the notes document
https://docs.google.com/document/d/1NZeAkPhnhUytsVP1rKAScEsxPWYmZC2s7yNt4Eqj_cU/edit?usp=sharing
solar whale
1PM ?
tulip sleet
Thanks. Where do I add the compiler flags? Is there someplace in the makefile fo...
you can edit the Makefile. For CIRCUITPY_ULAB you can temporarily change mpconfigboard.mk, or you can add it to the command line: make BOARD=foo CIRCUITPY_ULAB=0
still zephyr
Question: CP Library Bundle has not been release in a while, is this release weekly thing or when a library is updated thing..?
still zephyr
mmm
tulip sleet
you could check the actions in the bundle repo to see if they are failing
still zephyr
will do thanks 🙂
cobalt grail
Seems you don't know that Discord can display events in the user's local time zone?
https://ryps.co/discord/
tulip sleet
Seems you don't know that Discord can display events in the user's local time zo...
thanks! will take a look!!
analog bridge
Seems you don't know that Discord can display events in the user's local time zo...
Neat!
still zephyr
Last Bundle update was march 7th, two PR merged after that, #422 Adding GFX library and #421 Updating- CI, then no more auto-updates..
jaunty juniper
tulip sleet
Last Bundle update was march 7th, two PR merged after that, #422 Adding GFX libr...
there has to be a release, not just a merged PR of an updated library
jaunty juniper
adabot fails
remote: error: GH006: Protected branch update failed for refs/heads/main.
looks like adabot couldn't create the bundle release for the last 5 days
still zephyr
Community bundle must be affected too then
jaunty juniper
dunno, looks like a permission issue on the bundle repo ?
tulip sleet
remote: error: GH006: Protected branch update failed for refs/heads/main.
I might have set up a branch protection rule thoughtlessly. I deleted the branch protection rule. Let me re-run the latest failed run
still zephyr
Thanks Danh I see the release now
tulip sleet
great, and sorry, sorry - github keeps bugging us about unprotected branches
vapid storm
Attempted to ask this in the MP Discord, but seems the brain trust there are stumped as well.
tulip sleet
I can have multiple files running while sharing the same state in the long run, and this isn't something that CP supports.
What do you mean by this? Threads?
This is the best place for MPy discussions: https://github.com/micropython/micropython/discussions
@vapid storm ^^
tulip sleet
I'll start the meeting in 4 minutes or so
thorny jay
Skipping the meeting without notes. Group Hug.
Also, why did I not being a barometric sensor or something that can act as an altimeter with me in vacation???
lone axle
I just got back from vacation in the mountains, and hadn't even considered that, but am now asking myself the same question 😅
onyx hinge
@tulip sleet I could take the 'libraries' section if you like
lone axle
I put in the state section stats.
tulip sleet
<@&356864093652516868> Weekly meeting is starting now (Daylight Savings Time in USA)
turbid radish
Not so great
slender iron
@tame grove I'd love to be able to use rust for native internal modules but I've never actually used Rust. I'm happy to help you on the C-side of CP.
lone axle
Nice! I will have to try some fractal images
tulip sleet
@lone axle https://www.youtube.com/watch?v=T6uR5OZ0Xtk
if you websearch the topic, you will find more
lone axle
Thank you!
slender iron
thanks for hosting dan!
still zephyr
Thank you
lone axle
Thanks for hosting Dan, have a nice day everyone!
ember iris
Thank you all! Have a great week!
onyx hinge
@slender iron terrible idea: At the end of shared-bindings/foo.c, #include "common-hal/foo.c" -- voila, calls from shared-bindings to common-hal can potentially be optimized better, even if no lto
slender iron
ya, that crossed my mind 🙂
onyx hinge
how many people have to have the same bad idea for it to actually be a good idea?
slender iron
if it works, it works 🙂
my mind has gone to doing performance guided memory placement
ti clang doesn't support m7
😦
tulip sleet
we can hope they PR it to upstream
Here is the notes document for next Monday’s CircuitPython Weekly meeting. It is at the normal time of 11am Pacific / 2pm Eastern here on Discord. Everyone is encouraged to attend! Please add your hug reports and status updates even if you’ll be attending the meeting - it’s super helpful! If you are unable to attend but would still like to include updates, feel free to include them in the notes and we’ll read them off during the meeting. Hope to see you there! <@&356864093652516868>
https://docs.google.com/document/d/1y7kZuukcZXC1eiStMaMsI_YlglAAoW_fMbaQysK_CAY/edit?usp=sharing
slender iron
@tulip sleet looks like dtcm on ocram double the perf bench scores
tulip sleet
🎉 is that instead of TCM or in addition?
tulip sleet
I was wondering if you just enable write-through that it there will not be a caching problem. Maybe that is the default
slender iron
I had to run the test over uart
last I tried over usb, tinyusb crashed
repr A had issues running out of memory 🙂
nqueens is 17.6k
up from 13.7k
chaos went from 1747 to 2735
tulip sleet
yeah, I think repr A doesn't really help, because more things are boxed, and I don't there are many savings from the slightly simpler object type checking
orchid basinBOT
orchid basinBOT
blissful pollen
Oh this is fun. Apparently if I change optimization level on my build the garbage collection issue goes away (not sure if I laugh or cry yet)
slender iron
definitely sounds like a rogue address that is keeping it alive
tulip sleet
definitely sounds like a rogue address that is keeping it alive
could there by a long-lived not-long-lived issue here? When might a short-lived allocation be moved?
tulip sleet
but we have had to make things like UART buffers be long-lived because their addresses are stored in non-CircuitPython places, and we don't want them to move out from under the non-CPy place
slender iron
right
blissful pollen
I'm seeing it on the stack just trying to figure out in which frame
tulip sleet
I'm seeing it on the stack just trying to figure out in which frame
is there some storage that you allocate for the GIF library that is passed a pointer to the large thing?
blissful pollen
is there some storage that you allocate for the GIF library that is passed a poi...
Hmm the drawing routine is passed the bitmap object to draw into
tulip sleet
if that pointer is copied into some object that might be on the stack, then it will stick around
blissful pollen
Rough order: next_frame -> GIF_playFrame(bitmap) -> GIFDraw(GIFImage struct that holds pointer)
tulip sleet
this is the SPI non-displayio version, right? Does the SPI object do antying with the buffer pointer?
I don't think so, but I don't remember
GIFImage struct, who holds that?
blissful pollen
GIFImage struct is held by OnDiskGif. This is FourWire only direct. And it does not seem to trigger if I comment out the send(odg.bitmap) line
blissful pollen
Just my ignorance here - but does the stack not clear itself when it is being reused? So therefore if something isn't used the variable is about
I'll have to look again at FourWire. It seems to always put the bitmap in the same memory address so I can watch that manually I think
tulip sleet
it doesn't clear itself, but it knows where the limits are so it won't look at leftover junk (or it should)
orchid basinBOT
lone axle
Has anyone else experienced computer freezing during builds on Ubuntu 22.04.2 LTS? This is a problem that has started cropping up for me lately. I'm not exactly sure how to troubleshoot it though. It's not consistent, some builds succeeds, but other time the whole computer just locks up at some point during the build.
secondarily, where can I look to figure out specifically what happens when I run something like:
make BOARD=adafruit_feather_esp32s2_tft
I'm guessing that it's not make directly that is causing my problem but rather something that make calls to do the build process.
If I can find out specifically what software is doing work when the freezing is occuring I'm hoping to get lucky trying to search around online about the OS version and freezing with that softwares name.
blissful pollen
Anyone know offhand where to look up all the QSTR including non-compiled ones?
slender iron
is there a specific one you are looking for?
blissful pollen
I saw 612 (decimal) and think it is a reference to a method being loaded
slender iron
ah, do you have a debugger attached
blissful pollen
Yup using GDB
slender iron
I think you can do p qstr_str(612) and it'll run the code for you
blissful pollen
Thank you! It's not a qstr because "backlight_pin" is likely not the culprit but good to know.
jaunty juniper
secondarily, where can I look to figure out specifically what happens when I run...
building ESP does bring my computer to lag at some points, it gets pretty intense
blissful pollen
(oops it was 621 not 612... 621 is bitmap which sounds more likely to be involved!)
lone axle
building ESP does bring my computer to lag at some points, it gets pretty intens...
do you know if building esspressif port use a different compiler than building something else i.e. atmel or raspberry pi pico?
slender iron
ya, espressif has their own compiler to use
its older gcc
atmel and rp2040 use arm gcc
lone axle
okay, I will try a bunch of non espressif builds. The last few times that the freeze has occured was on espressif build. If I can narrow it to that port specifically I can hopefully get closer to a solution, and atleast have a potential work around to be able to do any development until I get it fixed.
blissful pollen
I have a hunch: no clue if anyone knows this but does anything else use a property for an object. In OnDiskGif you can read the bitmap like odg.bitmap not as a function. But it is looking like something involved in that leaves the pointer on the stack unless another function happens to use that stack space before collection.
Versus changing it to odg.bitmap() as a method to return the bitmap object.
slender iron
its a different code path in the vm
you shouldn't need to worry about this
it's why the MP GC is considered imprecise
lone axle
I have a hunch: no clue if anyone knows this but does anything else use a proper...
displayio.TileGrid.bitmap exists as a property and returns and returns a Bitmap object. The rest of that is a bit over my head, not sure if that example is the same thing you're looking for.
blissful pollen
I am half way through Jim's video on the GC and he just covered that.
Partly now this is learning for me, and cause it is kinda neat
blissful pollen
`displayio.TileGrid.bitmap` exists as a property and returns and returns a Bitma...
Hmm yeah true. Though in that case you give TileGrid the bitmap.
lone axle
had an atmel port build froze in the same way 😖
blissful pollen
had an atmel port build froze in the same way 😖
More random but you don't have any issues with overheating coming up? CPU would be burning hotter during compiles
lone axle
Thank you for the tip, I will try to look into it. I'll have to figure out how to monitor it.
I haven't had any other trouble that I think would be attributed toward overheating. I bought it as a complete pre-built desktop unit not sure what the cooling is like in it tbh.
I have been running a video in a browser (and have a few browser windows with many tabs open) so the computer is definitely under a decent load when it's occuring. I'm in the process of trying to remove things that are going on to see if I can coorelate any of them with it.
So far though I've never had it occur during any operation other than building circuitpython. I can run videogames + browsers + streaming video and don't have the same problem under what I would assume is similar load.
I had been using -j 8 during builds which caused them to speed up, I assume that adds extra load. The last couple of times it's occured I've left that out though so I know it's not directly the culprit
blissful pollen
Yeah video games generally would cause the same problem (if they're using any real amount of CPU).
stuck elbow
I noticed that circuitpython's makefile doesn't always work well with -j, I suspect not all dependencies are properly defined, and some things build in the wrong order
lone axle
During a build after it prints:
GEN build-pyportal/genhdr/moduledefs.h
and before it prints
QSTR updated
What would it be doing at that time (or is there some place I can look to find this out). The last time it froze was just after that first line was printed. (though I don't think the timing is exactly consistent every time that it does)
still zephyr
I have Ubuntu, in two different laptops, no freezing here, however not building ESP, only Pico and/atmel. but I do not use anything while building. And I use all the cores avalaible aka j4
still zephyr
I have a hunch: no clue if anyone knows this but does anything else use a proper...
The only thing that comes to mind: the palette.
blissful pollen
The only thing that comes to mind: the palette.
Thanks. Yeah I'm 95% sure this is the issue now just trying to figure out the best way to deal with it. Also wondering if properties may be "bad" for objects and the garbage collector at times
lone axle
I have Ubuntu, in two different laptops, no freezing here, however not building ...
Thank you, appreciate any extra data points I can get.
I'm starting to suspect CLion or perhaps more specifically the terminal inside of it. I had been making my builds by running the command inside of that terminal in CLion with the project open and it was freezing.
I've gotten several successful builds with just the plain terminal by itself no CLion open but with browser and video running, so the only variable at this point is the IDE. 🤞 it holds up and that hopefully I can at least have it open, having to switch out to a separate terminal wouldn't be the end of the world if it can reliably not lock things up.
still zephyr
I'm starting to suspect CLion or perhaps more specifically the terminal inside o...
Let me know if you want me to test with CLion, I do not have that installed in either laptop.
lone axle
I'll let you know.
It must have heard me talking though because I did just have it freeze with a build from the terminal directly 😦
blissful pollen
I have to write it all up but I'm convinced that a property for an object can lead to confusing the GC as the pointer gets left on the stack by the load attribute function.
The VM code could blank the problematic part after but for the 99.999% of the time it isn't needed probably a performance hit not needed.
lone axle
anyone have any idea what this error is talking about?
❯ make BOARD=pyportal
Use make V=1, make V=2 or set BUILD_VERBOSE similarly in your environment to increase build verbosity.
mkdir -p build-pyportal/genhdr
GEN build-pyportal/genhdr/moduledefs.h
In file included from asf4/samd51/hri/hri_d51.h:62,
from asf4/samd51/hal/utils/include/parts.h:49,
from asf4/samd51/hal/utils/include/compiler.h:20,
from asf4/samd51/hal/include/hpl_gpio.h:56,
from asf4/samd51/hal/include/hal_gpio.h:45,
from common-hal/frequencyio/FrequencyIn.c:31:
asf4/samd51/hri/hri_mclk_d51.h:1: internal compiler error: Segmentation fault
1 | /**
|
0x7fc6c4c4251f ???
./signal/../sysdeps/unix/sysv/linux/x86_64/libc_sigaction.c:0
0x7fc6c4c29d8f __libc_start_call_main
../sysdeps/x86/libc-start.c:58
0x7fc6c4c29e3f __libc_start_main_impl
../sysdeps/nptl/libc_start_call_main.h:392
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.
Looks like the compiler crashed... does it repeat?
lone axle
it repeated a few times and then I eventually tried with -j 4 and it succeeded. But I tried a clean again after that and back to not using -j at all and it went back to failing with this error a few more times. I tried -j 2 and -j 4 and both failed (but not with this specific error output). Eventually I tried another one with no -j and it succeeded. So yes repeats some, but not reliably.
blissful pollen
Starting to sound like some sort of system error 😦
lone axle
but I wonder if the freezes are coorelated somehow. Maybe this segfault occurs and it isn't able to recover enough to print the error
It's feeling to me like somehow my compiler(s) have gotten wonky somehow. I haven't noticed any other symptoms doing anything else on the machine building circuitpython is the only activity I've had any issue with so far (and that only started within the last few weeks).
I guess maybe something within the system that is only utilized during the gcc builds? I feel in over my head in linux and with all of this tbh. I've gotten pretty good at getting things going when everything wants to cooperate. But when things go haywire I feel like I'm taking complete stabs in the dark trying to troubleshoot it 😦
currently I'm trying on kernal v6.x whereas my default is v5.x. So far I haven't frozen, but I've been getting that error I posted. Though I'm almost certain I've seen that error on the v5.x kernel before as well.
blissful pollen
yeah that's getting over my linux head as well. That's a weird error (if that makes it any better - probably not!). I'm not sure why the compiler would crash randomly
lone axle
Now I've gotten 3 or 4 successful builds in a row, of the exact same kind as the ones that had failed, without really intentionally changing anything. So definitely something involved that is not reliably failing the same way each time.
slightly different error, but still seems like compiler puking in some way:
❯ make BOARD=pyportal
Use make V=1, make V=2 or set BUILD_VERBOSE similarly in your environment to increase build verbosity.
mkdir -p build-pyportal/genhdr
GEN build-pyportal/genhdr/moduledefs.h
arm-none-eabi-gcc: fatal error: Killed signal terminated program cc1
compilation terminated.
If you can I'd at least check CPU temps, maybe RAM just to try to eliminate some things.
analog bridge
@onyx hinge should I update #7718 with the suggestions in #7722 or a separate pr is fine
onyx hinge
a separate PR is fine, the goal was to separate my .. concerns .. about the watchdog API stop blocking 7718
tulip sleet
Now I've gotten 3 or 4 successful builds in a row, of the exact same kind as the...
I saw this kind of thing a long time ago with bad RAM.
like 10-20 years ago
midnight ember
I saw this kind of thing a long time ago with bad RAM.
That checks out with the symptoms. I've had it happen a few times. Stuff starts going a little haywire, it'll crash and when it boots back up things are just not quite right with the ram.
jaunty juniper
midnight ember
@lone axle Do some ram tests.
That is cool, congrats Neradoc!
Looks kinda similar to the bangle.js stuff scott is playing with?
jaunty juniper
Lilygo T-watch (V3): boots, web workflow works, I figured out the I2C controlled power for the display, now I need to make the PSRAM work and the display init in board.c (which works in code.py but so far caused the board to hang)
it's an ESP32
midnight ember
web workflow from the watch?
jaunty juniper
yup
midnight ember
holy s***
jaunty juniper
getting all the peripherals to work will take some work, but at least the vibration is easy: just setting it as a monochrome status LED, it buzzes the status
indigo moss
hello everyone, I would like to have a step by step guide to clone, fetch submodules, update frozen libraries, create a custom branch, build, push changes to the branch, and most importantly perform updates from latest branches and merge into the custom branch...
I know how to do most of the steps, but every time I need to update and merge there is always a snag especially with submodules.
I am wondering if there is just a simple guide to follow and avoid any steps that I may be doing wrong. Thank you in advance.
proven garnet
hello everyone, I would like to have a step by step guide to clone, fetch submod...
Have you taken a look at the Learn Guide on building CircuitPython? It lays out a lot of what you mention about fetching submodules and things specific to that repo. There's another Learn Guide about contributing to CircuitPython that talks more about the Git, such as creating branches.
https://learn.adafruit.com/building-circuitpython/introduction
https://learn.adafruit.com/contribute-to-circuitpython-with-git-and-github
As far as pulling in changes to your custom branch, regular git commands like git pull should do the trick, though others more familiar with working with the core can offer more guidance (e.g., rebasing vs merging).
analog bridge
Do we need the windows job? It has become a bottle-neck.
orchid basinBOT
jaunty juniper
ok I have board.DISPLAY on the T-Watch working, and the PSRAM seems to work, but it reports 4MB, when it's supposed to have 8, which either means I got sold a V2 (??) or there's something (else) fishy
(it's not a V2, at least the brightness pin is according to the V3 pinout)
indigo moss
Have you taken a look at the Learn Guide on building CircuitPython? It lays out...
Thank you for your reply, but as far as I know those guides are out of date.
tulip sleet
Thank you for your reply, but as far as I know those guides are out of date.
If you think the guide is out of date, please let us know what is wrong with it. I try to keep it quite up to date.
indigo moss
If you think the guide is out of date, please let us know what is wrong with it....
Sure, will do. Thank you.
lone axle
There might be some opportunity to add a section or page about specifically updating / merging from main. That was part of their question which I don't think the guide currently covers.
I struggle with that process over time as well so unfortunately I don't have any advice to give. Cloning, making a branch, adding a change, commiting, pushing if done all within a relatively short window all go fine with no issues.
But if that new branch gets stale behind main, or if my local repo main gets behind remote main I have trouble getting my local instance to get updated from remote main and then not show local changes that I didn't (intentionally) make. I seem to have the most trouble with some kind of MD5 hashes stored in a file by themselves, I think they're related to the submodules, but not certain.) After some time passes my local repo gets wonkier as more things get updated in main. My IDE allows me to ignore certain files when committing so I can use that for a while to keep making new commits without having to actual solve the root problem of my instance being out of sync for some reason. But it would be great to have some repeatable steps laid out that can be used to take an out of date local repo and get it up to date such that it will get back in sync with main, submodules, hashes and all, afterward attempting to make a commit would report 'no changes detected' if the user hadn't changed anything else.
I have sometimes gotten it resolved, but never with reliable/repeatable steps, when I attempt same thing again I've run into issues. I usually end up just cloning another copy of the repo to start fresh.
indigo moss
There might be some opportunity to add a section or page about specifically upda...
Thank you for rewording my concerns more appropriately.
brazen hatch
How can I read the date modified off of a file in cp?
I can't seem to find any api for it..
The only date readily available seems to be the build date
I want to have ljinux advance the rtc in case no wifi is available to the last modified thing, so that it's not too old
os.uname()[3][os.uname()[3].rfind(" ")+1:]
onyx hinge
@brazen hatch os.stat(file)[some number]
brazen hatch
Oh alright, thanks!
onyx hinge
I don't know what index number is the modification time but you can probably guess or check the source
| This object may be accessed either as a tuple of
| (mode, ino, dev, nlink, uid, gid, size, atime, mtime, ctime)
brazen hatch
Well, I will just check source, it's easier
onyx hinge
most values are dummy on cp of course
brazen hatch
Wait, are uid gid settable though?
Or just /dev/null
I could really use them for permissioning on ljinux..
Uhh, multiuser will be such a pain..
I may end up relying on a totally virtual filesystem, relying on files to describe the real files.
Like, every file having it's .prm counterpart describing it's perms
It's gonna be such a pain.
stuck elbow
I don't think fat12 has uid and gid, I think you set them when mounting the filesysyem and they are all the same for all files
brazen hatch
Yea, prolly they are just nulled
And all the virtual filesystem stuff is going to be so easily bypassable since pexec is a command that runs direct python code.
ljinux jailbrek eta wen
kernel ring 0 direct code execution directly available.
Ljinux structure is such a meme.
And I will do nothing about it.
I may make file flags like "this file can run this".
So only specific stuff can run other specific stuff.
wraith crow
@brazen hatch This is what I'm using: fTime = localtime(max(min(2145916800,dStat[9]),946684800))
dStat = os.stat(filename)
brazen hatch
Nice thanks
Well imma implement it in ls
cuz currently in ls -l it only does the files + if they are files or folders or hidden
analog bridge
@onyx hinge The watchdog can be disabled on atmel-samd as well
tulip sleet
@slender iron are you the primary reviewer for https://github.com/pidcodes/pidcodes.github.com/pulls ? There is a backlog; a few of these are CPy-related.
onyx hinge
@analog bridge it seems like we should try to re-construct the reason it was designed NOT to be deactivated ever, then
and see if those reasons still apply
analog bridge
17.6.2.3 Enabling and Disabling
The WDT is enabled by writing a one to the Enable bit in the Control register (CTRL.ENABLE). The WDT is disabled by writing a zero to CTRL.ENABLE.
The WDT can be disabled only while the Always-On bit in the Control register (CTRL.ALWAYSON) is zero.
tulip sleet
if it can be deactivated on some chips, then maybe just raise an exception if someone tries to deactivate it on nRF or other chips where it's not possible? Or maybe I am late to the party and you are doing this already.
analog bridge
I think this should do the trick on atmel-samd: WDT->CTRLA.bit.ENABLE = 0 granted that the watchdog was initialized with WDT->CTRLA.bit.ALWAYSON = 0.
jaunty juniper
what does it mean that the pico sdk does not give a way to disable the watchdog ? does it give a reason anywhere ?
analog bridge
This leaves only nRF and from what I have read on multiple nRF forum posts, it can't be done and they explicitly state that.
Its also mentioned that the watchdog registers are disabled post watchdog initialization and subsequent writes aren't allowed.
There is a way to disable the watchdog when in sleep or debugging active, so I believe a hacky way might still exist.
analog bridge
what does it mean that the pico sdk does not give a way to disable the watchdog ...
Idk but they themselves disable the watchdog when doing some configuration internally in the sdk and then it is re-enabled.
orchid basinBOT
slender iron
<@252717193496756235> are you the primary reviewer for <https://github.com/pidco...
Yup. I'll look today. Would you mind updating your tinyusb PR to include my imx fix from yesterday?
tulip sleet
Yup. I'll look today. Would you mind updating your tinyusb PR to include my imx ...
I did, Thach mentioned it
slender iron
thanks!
timid bolt
@blissful pollen correct me if I'm wrong, but I don't see any code where the GC scans the C stack for root pointers. In which case the dest array you refer to is on the C stack and so it is never seen by the GC in anyway. Right?
tulip sleet
<@228322543507865601> correct me if I'm wrong, but I don't see any code where th...
this is via mp_state_thread_t in mp_state_ctx_t. There's a comment:
// This structure holds state that is specific to a given thread.
// Everything in this structure is scanned for root pointers.
typedef struct _mp_state_thread_t {
// Stack top at the start of program
char *stack_top;
#if MICROPY_MAX_STACK_USAGE
char *stack_bottom;
#endif
a pointer to an object might only be on the C stack for a while, so it needs to look there. ... Unless I am confused.
also note if PYSTACK is enabled, it looks on that
tulip sleet
well, there's a slightly contradictory comment:
////////////////////////////////////////////////////////////
// START ROOT POINTER SECTION
// Everything that needs GC scanning must start here, and
// is followed by state in the mp_state_vm_t structure.
//
@slender iron is more familiar with the gc code
slender iron
collect over the stack is done here: https://github.com/adafruit/circuitpython/blob/main/main.c#L1150
tulip sleet
important comment:
// This naively collects all object references from an approximate stack
// range.
@tidal kiln I'll take a look at the epd issue tomorrow when I get my epd feather and my brain into that mode
tulip sleet
@blissful pollen note approximate above. That may be the giveaway for the problem with the gif code. Even if it has return from that frame, if it goes past the end of the current frame in that collect
tidal kiln
@slender iron awesome. thanks! really curious what is going on there. super weird one.
jaunty juniper
Is it LOLIN S2 Pico by LOLIN?
do you have a Lolin S2 Pico ? I have a build to make you test with board.DISPLAY
thorny jay
do you have a Lolin S2 Pico ? I have a build to make you test with board.DISPLAY
I have, somewhere at home. But I will not be there before 19th, but I will not have much time before the 23th.
jaunty juniper
I have, somewhere at home. But I will not be there before 19th, but I will not h...
well you can bookmark this build: https://www.neradoc.me/builds/lolin_s2_pico-8.1.0-beta.0-23-gbca7d6e99-add-display-lolin-pico-dirty.uf2 with this code: https://gist.github.com/Neradoc/0d5106ac93a978d8be36ecf7b8dd9683
thorny jay
well you can bookmark this build: <https://www.neradoc.me/builds/lolin_s2_pico-...
Noted, I'll give it a try. So you also solved the I2C pin in use!
I had the plan to copycat what you were doing on the other board, but it is as great if it is already done!
jaunty juniper
well I can't test, so we might need some back-and-forth when you have time for it. I didn't know there was an I2C pin in use issue
it uses the same trick as the feather TFT, by setting up the display on the shared board.I2C(), so you can't setup the pins manually, but have to use board.I2C()
tulip sleet
@micrdev maybe I should have done a make translate in that pr
i saw your deleted message
analog bridge
@tulip sleet nevermind, false warning
tulip sleet
i think the compilation did the right thing, then, it must be weblate that doesn't like it?
slender iron
@tulip sleet @onyx hinge any thoughts about me adding an #define CP_INLINE inline __attribute__((no_instrument_function))?
tulip sleet
instead of adding it to MP_INLINE or whatever it's called?
onyx hinge
there's no MP_INLINE, oddly
slender iron
yup, just looked for that too
onyx hinge
there are MP_NOINLINE and MP_ALWAYSINLINE
tulip sleet
ALWAYS_INLINE is stronger??
onyx hinge
Is no_instrument_function something to do with the profiling work you're doing?
slender iron
yup, it exempts a function from -finstrument-functions. clang has an option to do that after inlining but not gcc
so the instrumentation prevents inlining
wraith crow
It's been awhile since I built an iMX image is there an additional tool-chain step I need?
../../shared-module/usb/core/Device.c:64:31: error: unknown type name 'tuh_xfer_t'
64 | STATIC void _transfer_done_cb(tuh_xfer_t *xfer) {
| ^~~~~~~~~~```we just updated tinyusb. you'll need to make fetch-submodules
wraith crow
Was just doing a fresh clone 😄
jaunty juniper
@tulip sleet any plan for an 8.0.4 release ? There's unreleased fixes in 8.0.x, in particular the web workflow related non-blocking sockets thing
tulip sleet
<@329766224093249548> any plan for an 8.0.4 release ? There's unreleased fixes i...
I am planning to do it tonight or tomorrow.
jaunty juniper
I am testing the support matrix PR and I get empty contents when it retrieves the past files.json even though it gets the correct size in the size field
this here: https://api.github.com/repos/adafruit/circuitpython-org/contents/_data/files.json?ref=78a1c988538d65a092c319faeec357fb0a668ac5
orchid basinBOT
jaunty juniper
Between 1-100 MB: Only the raw or object custom media types are supported. Both will work as normal, except that when using the object media type, the content field will be an empty string and the encoding field will be "none". To get the contents of these larger files, use the raw media type.
files.json is just over 1MB
lone axle
I saw this kind of thing a long time ago with bad RAM.
Thank you for this tip about potential bad RAM. I don't have a way to know for certain yet, but I swapped to dome different RAM and have not had the same issues arise yet after several builds. Anecdotally looking like a good theory so far.
tulip sleet
Thank you for this tip about potential bad RAM. I don't have a way to know for c...
sounds good. it could also have been badly-seated RAM. But it can be hard to test. I have used the Linux RAM tester that grub puts in the boot menu for hours and it didn't find the bad RAM.
I think when I really saw this was in the late 1990's, when I had to compile my own Linux kernel. Sometimes it failed, which was confusing. After some tedious swapping I found the bad stick.
tulip sleet
@jaunty juniper worked, thanks!
we'll need to PR this to 8.0.x and main. But perhaps you might go to sleep 🙂
jaunty juniper
you can add it to your PR ? It will need it anyway
tulip sleet
I can, and I will backport it to 8.0.x.
analog bridge
@tulip sleet can you check why the readthedocs CI is failing? Is it something to do with #7725?
tulip sleet
<@329766224093249548> can you check why the `readthedocs` CI is failing? Is it s...
7725 isn't merged yet. But see above for a problem that Neradoc diagnosed
analog bridge
Revision e207106d. Last updated on 07 Mar 2023.
tulip sleet
7725 just happened to be coincident with the files.json file gettign too big
I will look at this in the morning, is that ok?
trying to finish the 8.0.4 rlease and then sleep
thanks for noticing!
analog bridge
Does it run on a schedule?
tulip sleet
it runs after each merge
marble grail
Good evening - should I be able to run this example on windows 11 if I change my python version to ~3.8?
https://learn.adafruit.com/circuitpython-ble-libraries-on-any-computer/ble-uart-example
I am receiving this error:
Trying to connect...
Unhandled Exception: System.NotSupportedException: Cannot create uninitialized instances of types requiring managed activation.
tulip sleet
<@329766224093249548> can you check why the `readthedocs` CI is failing? Is it s...
re RTD builds failing: I looked at the build logs:
the issue is this:
PDF generation failed. The build log below contains information on what errors caused the failure.Our code has recently changed to fail the entire build on PDF errors, where we used to pass the build when a PDF was created.Please contact us if you need help understanding this error.
and the specific error is:
! LaTeX Error: Unknown graphics extension: .svg.
there are also a lot of warnings:
LaTeX Warning: Hyper reference `docs/index:project-structure' on page 3 undefined on input line 107.
and similar
perhaps we might turn off PDF generation for now unless .svg is easy to fix
onyx hinge
@slender iron @tulip sleet what do you think of requiring the buffers that contain PIO instructions to be of type 'H'. It's a very small potential breakage, anyone who uses adafruit_pioasm would be fine; but if you e.g., open a file and read() it you'd have a buffer with typecode 'b' instead... this is relative to https://github.com/adafruit/circuitpython/issues/7706#issuecomment-1470484705
and the fix is to memoryview(buf).cast('H') or so
tulip sleet
<@252717193496756235> <@329766224093249548> what do you think of requiring the b...
and the fix is to memoryview(buf).cast('H') or so
Are instructions always 16 bits? How hard is it to just check that the length is a multiple of 2?
onyx hinge
yes they are always 16 bits
but some strings would still be accepted, even though it would be wrong
tulip sleet
hmm... the memoryview cast seems rather obscure.
onyx hinge
like "in x" for instance, is a 4-byte string but is nonsense as a PIO instruction
maybe explicitly check for the type being 'str' and reject it, as well as checking length, so that "in x" would fail but b"in x" would be OK, just less likely to occur by accident
tulip sleet
If, say, you were reading 32 bit ints from a binary file, what would be the best way of reading those ints? Would you read bytes and use struct.unpack? I'm wondering what the usual Python idiom is
onyx hinge
if I understand this user correctly, they're saying they ran into a form of this and are wishing the problem would be diagnosed better
tulip sleet
I think checking for factor of two is easy. But is that already being done and the message is obscure?
onyx hinge
standard Python has array.Array.fromfile, CP doesn't.
some of the locations that work with SM instruction sequences check for multiple-of-2 but this one didn't (and it has the bug discussed in that issue, where it tried to execute 2x as many instructions as it should have)
brazen hatch
Progress.
I got custom logs on s2 uart.
tulip sleet
I think we could refactor the *2 check so it's small, and check everywhere. But let me stare at the api for a bit...
onyx hinge
factoring it to do the same check at all 3 (?) relevant sites would be good
there's the 2 constructor arguments + the function being discussed in that issue
tulip sleet
rp2pio.StateMachine(program: circuitpython_typing.ReadableBuffer, ...) I am somewhat reluctant to change this to an H array.array without doing it in a staged way
on the other hand, using 16-bit objects gets rid of any endianness questions
onyx hinge
what about excluding 'str' even if it's nominally a circuitpython ReadableBuffer?
'str' is always wrong, 'bytes' is might-not-be-wrong
str is not a ReadbleBuffer in real python, it's sort of an accident that it is in CP/MP
tulip sleet
yes, I think we have some similar checks elsewhere and there might be an mp_ routine or macro to check that
someday there might be an extension to PIO that has longer or shorter instructions (but maybe not)
#: shared-bindings/displayio/Palette.c
msgid "color buffer must be a bytearray or array of type 'b' or 'B'"
msgstr ""
#: shared-bindings/audiocore/RawSample.c
msgid "%q must be a bytearray or array of type 'h', 'H', 'b', or 'B'"
msgstr ""
etc.
brazen hatch
YES! IT. IS. OPTIMISATION!
I set OPTIMIZATION_FLAGS ?= -O2 for debug too and it now crashed!
Finally!
Now I just need to find which func is the one assuming the static size.
Or I can just make it volatile.
It's not the pystack funcs though.
It updates on reload just fine.
brazen hatch
Well, all the pystack logic seems to work just fine.. What is going on even..
All the functions report the same addresses and len's.
Hmm.. now what do I do?
jaunty juniper
so I just discovered this:
$ python3 -m pip install pip
error: externally-managed-environment
https://github.com/python/cpython/issues/102134#issuecomment-1445428402
As the message you're getting explains, this is actually not an issue with Python itself, but rather your Linux distribution (Kali, Debian, etc.) implementing a deliberate policy to ensure you don't break your operating system and system packages by using pip (or Poetry, Hatch, PDM or another non-OS package manager) outside the protection of a virtual environment.
This has been a long-standing recommendation from both the Python team and Linux distros, due to the severe problems that frequently occur for people as a result of using pip or other package managers directly with your system Python.
that's gonna affect Blinka
stuck elbow
other distros just make --user the default
to be honest, I can't blame the debian maintainers for getting tired of the python developers regularly messing this up
tulip sleet
Ubuntu has —user as the default. I’m surprised Debian doesn’t. Or does it?
stuck elbow
they just replaced pip with a script that shows that error
I bet you could uninstall the system pip and install it using get-pip.py and it will work
brazen hatch
python3 -m ensurepip?
brazen hatch
least decisive api
stuck elbow
ok, correction, they didn't completely replace pip
onyx hinge
so I just discovered this:
```sh
$ python3 -m pip install pip
error: externally-...
fwiw the full message in debian testing is:
error: externally-managed-environment
× This environment is externally managed
╰─> To install Python packages system-wide, try apt install
python3-xyz, where xyz is the package you are trying to
install.
If you wish to install a non-Debian-packaged Python package,
create a virtual environment using python3 -m venv path/to/venv.
Then use path/to/venv/bin/python and path/to/venv/bin/pip. Make
sure you have python3-full installed.
If you wish to install a non-Debian packaged Python application,
it may be easiest to use pipx install xyz, which will manage a
virtual environment for you. Make sure you have pipx installed.
See /usr/share/doc/python3.11/README.venv for more information.
note: If you believe this is a mistake, please contact your Python installation or OS distribution provider. You can override this, at the risk of breaking your Python installation or OS, by passing --break-system-packages.
hint: See PEP 668 for the detailed specification.
This feels .. extremely user hostile
stuck elbow
that's what you get when project politics become more important than users
onyx hinge
you can also fix it with rm /usr/lib/python3.11/EXTERNALLY-MANAGED until the python package upgrades
I am especially astounded that this applies to --user
stuck elbow
yeah, however, it's after years of https://xkcd.com/1987/
--user has its own problems, for example, it installs all binaries for different python versions together...
onyx hinge
aha dpkg-divert --divert /usr/lib/python3.11/NOT-EXTERNALLY-MANAGED --rename /usr/lib/python3.11/EXTERNALLY-MANAGED fixes it and will survive updates
stuck elbow
an interesting tool
tulip sleet
Some things to read about this pip change:
https://peps.python.org/pep-0668/
https://discuss.python.org/t/10302
Also not directly related to this, but interesting:
https://justinmayer.com/posts/homebrew-python-is-not-for-you/
orchid basinBOT
tulip sleet
in particular, they say pip install --user is still bad, because you're mixing user-installed and system-installed stuff due to sys.path, possibly causing an inconsistent mix of versions and it can cause problems for you (though not for stuff run only by the system).
onyx hinge
@slender iron there are 4 flags for increasing code alignment across the board: -falign-functions=, -labels=, -loops=, and -jumps= -- I don't know that there's one for data. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-falign-functions
slender iron
onyx hinge
I assume you don't have the .elf or .map that goes with it 😕
slender iron
offset 0x60000400
nope 😕
its not that dissimilar from something you'd compile from my branch
branch is get_perfbench_running
blissful pollen
@slender iron out of curiosity what tool are you using to get those performance/benchmark stats (just saw some of last week's show n tell). Haven't been able to follow much lately. (this isn't related to what Phil and Limor just said either on SnT ha just was interested).