#circuitpython-dev
manic glacierBOT
slender iron
@lucid solar Interesting! I'd love to see your draft of it. I don't know cmake so it'd be tougher for me to do.
lucid solar
I'm also making sure you can build it directly without having to use their ./install.sh scripts etc.
slender iron
@tulip sleet do you have cycles today to do further lower_power tests for me? I'm curious if master has a problem for and and if you can repro on a debugger
tidal kiln
@slender iron i've lost track of things - what's the desire for pixel_order? should it be a string, a tuple, or support both?
slender iron
looks
I imagine it should use the pixelbuf variables and not care what the value is
tidal kiln
where are those defined?
tulip sleet
@slender iron I can try master, within an hour or so.
slender iron
@tulip sleet ok, that would be awesome. poking at it now. chatting with helen soon I think
idle owl
@onyx hinge No rush. When you get a chance, can you reply to this? I am assuming you may know. https://forums.adafruit.com/viewtopic.php?f=60&t=164898
tulip sleet
@idle owl when I'm creating an alias for a readthedocs subproject of circuitpython, should the alias use dashes or underscores? I seem to be finding both
tidal kiln
@slender iron nevermind, moved question / discussion to PR.
https://github.com/adafruit/Adafruit_CircuitPython_NeoPixel_SPI/pull/15
tulip sleet
idle owl
@tulip sleet Look at the RTD link in the README of the lib. And make it exactly what it is in the README.
tulip sleet
got it; there is also a link in the release notes, so maybe one is wrong
onyx hinge
@idle owl thanks, done!
idle owl
@onyx hinge thank you!
tulip sleet
@slender iron I am not seeing the same Metro M0 trouble on the tip of master that I am seeing on lower_power.
also I still hate sphinx:
/home/halbert/repos/Adafruit_CircuitPython_BLE_BerryMed_Pulse_Oximeter/adafruit_ble_berrymed_pulse_oximeter/__init__.py:docstring of adafruit_ble_berrymed_pulse_oximeter.PulseOximeterValues.finger_present:1: WARNING: duplicate object description of adafruit_ble_berrymed_pulse_oximeter.PulseOximeterValues.finger_present, other instance in api, use :noindex: for one of them
sphinx is generating useless "alias for field number <n>" for namedtuples that conflict with the real docstring ... actually this appears to be an artifact of CPython, which generates useless docstrings, and sphinx doesn't override them properly
(That's why I didn't respond to you so promptly)
slender iron
@tulip sleet I don't think I've ever reproed the partial enum you see. can you send me a uf2?
one thing I've seen is the DFLL not coming up
tulip sleet
did you see the extraneous reset in the trace?
I thought the xobs comment might be helpful: https://github.com/adafruit/circuitpython/pull/2685#issuecomment-620386943
This is built from a completely updated clone of your repo, for Metro M0, so it should be the same as yours
slender iron
I think my tinyusb asserts are on but I don't hit any
when I see a usb failure it looks like this:
tulip sleet
did you look at the trace i put in the issue?
tulip sleet
what's the hardware you have that's running Arch?
tulip sleet
what motherboard, and is it an Intel CPU?
slender iron
no, a ryzen
idle owl
@tulip sleet What did you end up doing for the namedtuple Sphinx issue? The last time I ran into it was another one of your libs, and we tweaked it until it passed, but the "alias for" lines are still in the docs.
tulip sleet
@idle owl not solved yet; it seems to have reappeared maybe with 3.0.3 of sphinx, or maybe it was always there
total pain
idle owl
We found one workaround but it was complicated and extraneous and I'm not even sure sommersoft ever got it to work.
tulip sleet
@slender iron do you have an vanilla intel box or laptop you could try?
slender iron
my thinkpad might be an intel
tulip sleet
could boot off an ubuntu dvd; that might replicate it
tulip sleet
i haven't tried yet
idle owl
@tulip sleet I think this is the one we updated but left the alias lines in the docs - it wasn't worth the fight: https://github.com/adafruit/Adafruit_CircuitPython_BLE_Heart_Rate/blob/master/adafruit_ble_heart_rate.py
tulip sleet
@idle owl yes, I was doing the same as that one; I'm going to try that one again and see if it fails anew
idle owl
Ok.
tulip sleet
@slender iron I can certainly try it in the debugger but I wasn't sure what to look for
slender iron
once it stops you can see if it's in the sleep_until function
that's mostly what I'd expect
or it could be some other timing loop that is wrong
tulip sleet
What I was thinking of doing was triggering a pin on each call to TinyUSB task and see whether it's late
compared with the same thing in master
slender iron
do we actually need to call it regularly or just after any interrupt?
tulip sleet
i thought we had to call it every msec; that's what Thach told me
in the lastest master, do you do that still?
I think we should ask him about it when he gets up
slender iron
in the lower power we only call it when awake
the frustrating part is that it works for me
tulip sleet
right, it's clearly some timing thing that's different on your USB hardware or the drvier
what does your thinkpad run?
tulip sleet
i mean, is it linux or some old windows?
tulip sleet
if you can bring up the thinkpad and bring it up to date with whatever, just try it, but mainly let's wait for Thach and ask him if we WFI and don't call tud_task is that fatal?
it's possible he's scheduling something that needs to be done not on an interrupt but sometime after an interrupt
tulip sleet
is there a timer interrupt every msec
i.e. does the WFI return every 1/1024 of a sec?
slender iron
only if ticks are enabled
which you could turn on to try
but we also disabled wfi and had the issue
tulip sleet
definitely; how?
it still might not be calling the background task every tick
for some reason
slender iron
it should over-call it if wfi isn't called
tulip sleet
so maybe I should instrument that call to the tud task and look at it on the Saleaea
slender iron
ya, and see if you can see where it is with a debugger
does the neopixel ramp up and down?
tulip sleet
or maybe you turned off some interrupts? did you redo the priorities
yes, the board seems fine; I just can'talk via USB
slender iron
I think I replaced systick with rtc but nothing else
tulip sleet
anyway, how do I enable the ticks? I'll try that just in case
tulip sleet
I'll try that; I'll turn it on at the beginning of main or something
tulip sleet
@idle owl it's actually still broken in heart-rate: it doesn't throw errors but the documentation is missing:
idle owl
That's odd, it was included previously - I remember the documentation existing with the aliases below it.
Maybe when built locally or something. I have no idea.
Be back in a bit. Recording.
slender iron
@tulip sleet looks like my nuc may have tickled it
tulip sleet
@idle owl mine has py:attribute, heart-rate has just :attribute::, which I think might be a bug: the leading : is wrong. If I remove that, I start to get
/home/halbert/repos/Adafruit_CircuitPython_BLE_Heart_Rate/adafruit_ble_heart_rate.py:docstring of adafruit_ble_heart_rate.HeartRateMeasurementValues.heart_rate:1: WARNING: duplicate object description of adafruit_ble_heart_rate.HeartRateMeasurementValues.heart_rate, other instance in api, use :noindex: for one of them
@slender iron excellent (I guess!). I'm just trying to get the pulse ox library into the bundle because the guide went live
slender iron
I wonder if I'm calling it too fast
tulip sleet
@idle owl I'm going to do this for now, just to get it to sort of work. I'll file a bug on sphinx eventually.
raven canopy
@tulip sleet catching up on the scroll-back; i took most of the day away from the PC.
the namedtuple alias thing is certainly annoying. like Kattni said, we dug into it, and the 3.x "fix" was the error getting thrown. there is an issue about the alias fields though: https://github.com/sphinx-doc/sphinx/issues/7353
i did try building with the workaround mentioned in the issue, but my notes (DMs w/Kattni) said it failed; i just don't remember why.
tulip sleet
@raven canopy thanks, I found that workaround, but it's from 2015. There may be some other way. I might submit a new issue because of the index warnings, which are more trouble (since we treat them as errors). I made it into a bulleted list for now instead of using attribute. Apparently CPython may generate those useless docstrings. I would like any supplied docstring to override those for namedtuples.
raven canopy
huh. it does appear that CPython creates the aliases...
>>> bar = namedtuple("bar_tuple", ("foo", "bar", "baz"))
>>> help(bar)
Help on class bar_tuple in module __main__:
class bar_tuple(builtins.tuple)
| bar_tuple(foo, bar, baz)
|
| bar_tuple(foo, bar, baz)
|
| Method resolution order:
| bar_tuple
| builtins.tuple
| builtins.object
... methods removed ...
|
| ----------------------------------------------------------------------
| Data descriptors defined here:
|
| foo
| Alias for field number 0
|
| bar
| Alias for field number 1
|
| baz
| Alias for field number 2
@tulip sleet I think the host controller is changing behavior because we're responding faster
and tinyusb is assuming we can handle the tx complete interrupt before the next setup packet arrives
tulip sleet
do you want to bring this up with Thach?
We could have him debug this with lower_power
slender iron
ya I'll ask him about it
tulip sleet
is it also dependent on the chip periph? Why don't we see this on SAMD51 or nrf, for instance?
slender iron
I think xobs is seeing the same on '833
tulip sleet
aha, ok
slender iron
the SAMD51 may be fast enough to handle the tx complete before the next setup packet
tulip sleet
i can try on an 840 also
slender iron
I think we're responding faster to the string descriptor fetches which is making the host assume we're fast
tulip sleet
'840 seems OK, so the SAMD21 is really the test
slender iron
are you trying with a debugger?
there may be tinyusb asserts that fail with asm("bkpt")
tulip sleet
no, I can do that
simple pulsar
For displayio has there been any discussion on the Labels for being able to choose different gravity/justification? I'd like to be able to do top left positioning.
tulip sleet
do i need to compile DEBUG to enable the asserts?
slender iron
that I don't know
raven canopy
@tulip sleet there is an older CPython issue about this as well. not much movement on it though. seems the most effective workaround (applies to Sphinx and CPython help) is to subclass namedtuple, and supply the desired docstring. but then...might as well just write it as a class. :D
https://bugs.python.org/issue16669
not trying to interrupt debug; can be dealt with later
raven canopy
hehe. ISRs everywhere!
tulip sleet
@slender iron I see no bkpt asserts either way on '840, but that doesn't mean xobs isn't seeing them. But I'm glad it's reproducible for you. I think Thach can help us.
taking a break now
marble hornet
what is the difference between py/runtime.h and py/runtime0.h ?
tulip sleet
@marble hornet that's a good question; I haven't figured out why both exist
marble hornet
maybe it's like coke vs coke0 π
raven canopy
a quick diff was not very helpful. but runtime0 is included with 4 files in py/.
tulip sleet
if you grep for either's inclusion in various files, there are a lot of examples, and I don't detect an obvious rhyme or reason. I looked at the micropython issues and PR's and also didn't come up with an explanation
the redundancy bothers me
marble hornet
i'm attempting to write a module based on the extending circuitpython guide. in the shared bindings the bindings include an __exit__ method but i don't see the __enter__ method defined . the comments say it is made via a "Provided by context manager helper." but i don't see where it is used
//| .. method:: deinit()
//|
//| Deinitializes the Meaning and releases any hardware resources for reuse.
//|
STATIC mp_obj_t tinydisplay_tinytft_deinit(mp_obj_t self_in) {
shared_module_tinydisplay_tinytft_deinit(self_in);
return mp_const_none;
}
STATIC MP_DEFINE_CONST_FUN_OBJ_1(tinydisplay_tinytft_deinit_obj, tinydisplay_tinytft_deinit);
//| .. method:: __enter__()
//|
//| No-op used by Context Managers.
//|
// Provided by context manager helper.
//| .. method:: __exit__()
//|
//| Automatically deinitializes the hardware when exiting a context. See
//| :ref:`lifetime-and-contextmanagers` for more info.
//|
STATIC mp_obj_t tinydisplay_tinytft_obj___exit__(size_t n_args, const mp_obj_t *args) {
shared_module_tinydisplay_tinytft_deinit(args[0]);
return mp_const_none;
}
STATIC MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(tinydisplay_tinytft___exit___obj, 4, 4, tinydisplay_tinytft_obj___exit__);
it doesn't subclass anything
tulip sleet
or there's a default __enter__() which doesn't do anything
raven canopy
{ MP_ROM_QSTR(MP_QSTR___enter__), MP_ROM_PTR(&default___enter___obj) }
trying to find where the default __enter__ is.
marble hornet
ah
tulip sleet
do you have a PR #?
raven canopy
trying. only commit surfing now...
this is the original commit for runtime0: https://github.com/micropython/micropython/commit/d99b05282d14ceb0163cbcd059aa37bdb415af43#diff-7f2d7c73676c7b412b695ef33348571a
this is the original commit for runtime0: https://github.com/micropython/micropython/commit/d99b05282d14ceb0163cbcd059aa37bdb415af43#diff-7f2d7c73676c7b412b695ef33348571a
this is the original commit for runtime0: https://github.com/micropython/micropython/commit/d99b05282d14ceb0163cbcd059aa37bdb415af43#diff-7f2d7c73676c7b412b695ef33348571a
this is the original commit for runtime0: https://github.com/micropython/micropython/commit/d99b05282d14ceb0163cbcd059aa37bdb415af43#diff-7f2d7c73676c7b412b695ef33348571a
marble hornet
okay, i think i finished the c but would now like to hook it in
if i remember properly the guide is out of date for that
the first steps of this link seem to be wrong https://learn.adafruit.com/extending-circuitpython/inside-the-virtual-machine#ports-slash-atmel-samd-3-29
tulip sleet
it's from 2018, so it may be outdated
raven canopy
a bit, since most of that is built in the .mk now. but, still mostly doable: https://github.com/adafruit/circuitpython/blob/master/ports/atmel-samd/Makefile#L287
marble hornet
okay:
here is what i tried SRC_SHARED_MODULE += tinydisplay/__init__.c tinydisplay/TinyTFT.c
and i am getting this error
Use make V=1, make V=2 or set BUILD_VERBOSE similarly in your environment to increase build verbosity.
make: *** No rule to make target `shared-module/tinydisplay/__init__.c', needed by `build-pyportal/genhdr/qstr.i.last'. Stop.```rule?
tulip sleet
I would say mostly ignore that guide, and just look at an existing module, and model what it does
makefile rule
what existing module is similar to this?
marble hornet
oh, noted. should it be worth taking the guide down? or a warning at the top? and it will be a procedural display library, like the python rgb_display but in c
so i'm not totally sure. i'm just trying to get methods and properties working on an object
tulip sleet
which native module is this most simlar to? maybe the new RGBMatrix? (I don't mean in the details, but the purpose).
i'll add a red warning to that guide
marble hornet
looking to see which it would be most similar to...
i'm drawing a blank, displayio is kind of close in that it lets a user control the screen
tulip sleet
do you have port-specific code, or can it all be in shared-module?
marble hornet
it should all be in shared module
tulip sleet
does it need to live across VM instantiations? E.g., are you showing the REPL or something like that?
marble hornet
no, it won't need to
tulip sleet
_pixelbuf is straightforward (even if it is an _ module). See where that is added, turned on and off, etc.
it is only shared-module, no common-hal
emulate that. Search for where it's mentioned in the makefiles, etc.
tulip sleet
i refactored all that a while ago. I didn't really know about that guide. I've put a warning on it.
ok, π€, good luck!
marble hornet
sleep well. thank you (i think it is working! )
slender iron
@gentle bronze you around?
gentle bronze
@slender iron yup
slender iron
We found a weird issue on samd21 in my lower power branch. Let me go up to my computer
gentle bronze
ok
slender iron
kk, so I think the issue is that we're calling tusb_task more frequently than before
enumeration is failing here: https://github.com/hathach/tinyusb/blob/master/src/device/usbd_control.c#L169
its weird because trcpt and setup received bits are set
I can only repro it with an intel system too
my macbook and ryzen pcs are ok
but my intel nuc has issues
(my changes are here: https://github.com/tannewt/circuitpython/tree/lower_power )
this seems similar: https://github.com/hathach/tinyusb/issues/296
gentle bronze
it is different, the nuc120v is more difficult to work with, I couldn't get it running on my board at all. majbthrd could get it running on his dev board (different one) though.
slender iron
right, different nuc π
dan sees it have after set configuration
mine stops after get string descriptor
gentle bronze
hold on, I am checking out your tree to test with.
slender iron
k, I'll try with a tinyusb example
gentle bronze
if you could reproduce it with tinyusb example repo, try to enable LOG=2 LOGGER=rtt https://github.com/hathach/tinyusb/tree/master/examples#debug-log
and use jlink rtt viewer for the log, that hopefully would not intefere with the timing.
gentle bronze
building for metro m0
@slender iron I didn't follow low power issue, could you sum up the major modification it does π
slender iron
we switch timekeeping from systick to the RTC
in samd21 that meant we run gclk2 faster than before
we also always call tusb_task instead of at a 1 ms-ish interval
sleeps should actually sleep by calling wfi
gentle bronze
π π
slender iron
let me know if you can reproduce it. I was happy I could earlier
gentle bronze
I am able to reproduce it, it stall every request after the first device descriptor, very consistent
slender iron
great! that's the first step
gentle bronze
It must be late for you, please take the rest, I will do a bit of bisect to catch up with the low power changes first
then try to narrow down the issue π
gentle bronze
great, though it is still late. I will get you updated if I could figure out something.
slender iron
π thanks!
gentle bronze
no problems π
manic glacierBOT
gentle bronze
@slender iron btw, do you know from which commit the issue occurs
slender iron
no I don't unfortunately
gentle bronze
no problems, I do a binary search π
slender iron
it's probably worth looking at which commits are in the samd directory
since many of them will be for others
gentle bronze
it is OK, I could narrow it down for several trials:)
slender iron
I get a bit further if I do: ```
// Handle SETUP packet
if (USB->DEVICE.DeviceEndpoint[0].EPINTFLAG.bit.RXSTP)
{
// This copies the data elsewhere so we can reuse the buffer.
dcd_event_setup_received(0, _setup_packet, true);
USB->DEVICE.DeviceEndpoint[0].EPINTFLAG.reg = USB_DEVICE_EPINTFLAG_RXSTP | USB_DEVICE_EPINTFLAG_TRCPT0;
return;
}
yup, that seems to fix it for me
based on one trial π
ok tv time. thanks @gentle bronze !
gentle bronze
great finding, I will double check it with manual, and update tinyusb if needed
gentle bronze
I tried but it doesn't fix the issue, samd doesn't reponse correctly, I will do more check
basically, samd response with device descriptors contents for all the GET_DESCRIPTOR request (string/configuration etc ...)
plucky flint
Morning @here - today is MicroPython's 7th birthday. I can't think of a nicer way to celebrate than to publicly thank Damien for all his friendliness, extraordinary effort (lots of that) and technical excellence. e.g. I just tweeted this: https://twitter.com/ntoll/status/1255412859435126785 Thanks, and apologies for the loud "PING" in this channel.
Twitter
onyx hinge
@ionic elk stm32 and timer/counters question. I'm working on rgbmatrix/protomatter for stm32f405 feather. Is statically allocating TIM6, the "other" timer with no associated PWM pins, reasonable? Any hints for debugging my register setup of the timers? I feel like I am more in the "cargo cult" stage of trying to get this to work, than the "understanding it" phase, and it's been frustrating me for a bit of time now. I can share my WIP code with you if it's helpful.
between not being familiar with stm timers and not quite grasping how protomatter needs to use them it's been vexing me and I'm not making progress.
ionic elk
@onyx hinge You'll be competing with RotaryIO, which will also need an unallocated timer and I had using TIM6, though I haven't progressed very far on it. It seems we're outstripping these "unallocated" timers - what we might need is some kind of intermediary that has a list of low-priority timers it can check and dynamically allocate.
STM32 basic timers like TIM6 and TIM5 are further complicated by not being fully represented by the stm32 HAL (presumably because they're not really intended for "timer" use but more for DAC stuff?). You have to manually go in and implement on the register level via the reference manual. And if you ever switch to a general purpose everything changes and you have to rewrite.
onyx hinge
I have a whole raft of pins, I could later change to use a timer that is related to one of those pins I guess
I actually do have my interrupt executing now, yay.
slender iron
It'd be good to add a generic way to grab a free timer. Many chips don't have basic ones
@gentle bronze any insights on my usb issues?
gentle bronze
@slender iron yeah, it is a race condition with setup packets, I have a solution, only need to check with manual a bit more
slender iron
ya, seemed like a setup packet was catching up to OUT completion
gentle bronze
yeah, it is right after the out, causing missing setup, though the event still registerred and dcd duplicate the old setup packet as new. Should be a quick fix, just a bit more reading
onyx hinge
$68 = {ptr = 0x2001f3a0, length = 3168}
``` Is the stm stack smaller than the "usual" stack? I am hitting `reset_into_safe_mode (reason=HEAP_OVERWRITTEN)` a lot. The stack at the time this occurs looks "sane, but full".ports/nrf/mpconfigport.h:#define CIRCUITPY_DEFAULT_STACK_SIZE 0x6000
ports/atmel-samd/mpconfigport.h:#define CIRCUITPY_DEFAULT_STACK_SIZE (24*1024)
ports/stm/mpconfigport.h:#define CIRCUITPY_DEFAULT_STACK_SIZE ((uint32_t) &_ld_default_stack_size)
``` I can sort of tell that the stm approach is trying to do something allowing the default stack size to be in the linker file .. but it's a wildly smaller stack than nrf and samd...what board are you on?
onyx hinge
ports/stm/boards/STM32F405_default.ld:_minimum_stack_size = 2K;
the stm feather (f405)
stack when I hit the soft reset
slender iron
I thought the stack was dynamically sized
the .ld stuff is just to make sure we have a minimum size
onyx hinge
(gdb) p next_stack_size
$72 = 2048
static uint32_t next_stack_size = CIRCUITPY_DEFAULT_STACK_SIZE;
right now it's also being used as the default "next stack size"
slender iron
that's pretty small π
onyx hinge
I can chuck an issue in github, add a workaround in my boot.py, and probably get on with things. Mind if I do?
slender iron
you can just increase it π
even samd21 is 4096
samd51 is 24*1024
it'd be cool to move it to CCMRAM which is 64k
onyx hinge
I think I need to avoid getting side-tracked, I'm at 1 week with nothing to show for my rgbmatrix port to stm π¦
simple pulsar
@onyx hinge I raise you, two days spent just drawing plots of a rather simple passive filter.
BTW, I put a question about displayio in the forums earlier: https://forums.adafruit.com/viewtopic.php?f=60&t=164925
onyx hinge
86 if (current_safe_mode > BROWNOUT && reason > BROWNOUT) {
(gdb) where
#0 reset_into_safe_mode (reason=HARD_CRASH) at ../../supervisor/shared/safe_mode.c:86
#1 0x0802f2a8 in HardFault_Handler () at supervisor/port.c:244
#2 <signal handler called>
#3 0x0803088e in allocate_stack () at ../../supervisor/shared/stack.c:61
#4 0x0bff133a in ?? ()
``` well that didn't go so wellOK, moving on/back
manic glacierBOT
lucid solar
@slender iron esp32s2 port is going to be quite hard, there's a lot of linking config we need to sort out. I believe I know why it was crashing the linker now, we were not including some .ld files that it needs...so I think it was confused how big some sections were and that caused some crash.
manic glacierBOT
lucid solar
-L/Users/mark/src/esp-idf/examples/get-started/blink/build/esp-idf/esp32s2
-Tesp32s2_out.ld
-uapp_main
-L/Users/mark/src/esp-idf/examples/get-started/blink/build/esp-idf/esp32s2/ld
-Tesp32s2.project.ld
-L/Users/mark/src/esp-idf/components/esp32s2/ld
-Tesp32s2.peripherals.ld
@ionic elk I removed the ld-defined minimum stack size in the atmel and nrf ports when I parameterized .ld files. There may be an inconsistency in the stm port.
onyx hinge
I like montage for creating sprite sheets for displayio. https://imagemagick.org/Usage/montage/ really handy
manic glacierBOT
ionic elk
@slender iron by the way, I wrote a bunch of implementations for rotated rectangles and images when I was working on ExpressionEngine. Is that something you might be interested in having in DisplayIO someday?
gentle bronze
@slender iron after lots of testing and logging, I think we finally fixed it. PR to tinyusb is on the way https://github.com/hathach/tinyusb/pull/388
slender iron
nice! I was thinking it might be due to delaying the interrupt handling
gentle bronze
it is hardware race condition, samd doesn't have an way to NAK setup, it will just always acked. We need to ready the buffer at anytime.
slender iron
I didn't think the usb spec allowed NAKing a setup
gentle bronze
I actually make the race worse when doing the last clean up, by moving the setup prepare after the status complete. But it must be before queueing the last status (both IN and OUT status)
slender iron
right, if the assert didn't catch it first
gentle bronze
ah, I mean a way to not acked and discard setup. normally other mcus has special sram/pointer to store setup
anyway, the samd controller is very fast in this regard π
slender iron
right, that makes more sense
ionic elk
@slender iron you ok with me grabbing a Toasty? https://www.tindie.com/products/webtronics/toasty-480mhz-stm32-usb-development-board/
slender iron
@ionic elk sure
gentle bronze
@slender iron do you want to review the pr https://github.com/hathach/tinyusb/pull/388
slender iron
@gentle bronze ya, looking now. is it always a ZLP?
gentle bronze
the status is always ZLP and has opposite direction to the setup[0]'s bit7
slender iron
Yup! will do. Thanks again!
slender iron
good night
@tulip sleet USB should be working in lower_power now. I think I have two boards I need to find space for
tulip sleet
excellent! I'm still working on PacketBuffer and its vagaries with connection handles. It has been painful.
the connection handle is not easily found. I saw how you cached it and thought I could get it from the service, but that doesn't work for local services
slender iron
what do you need it for?
tulip sleet
and there is no "read" evt, so we couldn't get it if the client is just reading
I needed the connection handle to get the mtu for packet length
slender iron
there is a way to inject yourself in a read
tulip sleet
yes, I know about that (authorize stuff), but I hoped to avoid that
tulip sleet
ble_midi_simplein, for example
slender iron
midi should only be using notify and write
it does do a read at the beginning but it's not critical I believe
tulip sleet
the characteristic assigned to packetbuffer is notify/read (remember my "pathological" remark?)
it just has to do with the buffer setups in the MIDI service
slender iron
right, the read is just for a single read at the start
not for the actual midi packets
tulip sleet
so the packet size for that read is always below MTU?
even if MTU is 20?
i thought it was using the same characteristic bidirectionally, either for sending midi notes (et al) or receiving them
tulip sleet
def __init__(self):
super().__init__(
properties=Characteristic.WRITE_NO_RESPONSE
| Characteristic.READ
| Characteristic.NOTIFY,
for _MidiCharacteristic
slender iron
right. the read is only for the initial read
tulip sleet
so "a packet that has no payload" is going to be <= 20 bytes?
slender iron
it is empty
slender iron
packets usually have a header byte but this packet is empty
I'm not sure any implementations actually check it's contents
tulip sleet
the current MIDI code does this in __init()__:
self._in_buffer = bytearray(self._raw.packet_size)
self._out_buffer = None
but we don't know packet_size without knowing the MTU, so I rewrote it to delay creating this buffer. I see that you adjusted the size by re-checking .packet_size, so maybe you were assuming it might be too small and then grew it as necessary. There were other services (e.g. heart rate, iBBQ), where I just called packet_size once in __init()__, and I now defer creating the buffer until connected
I will have some draft PR's for those
tulip sleet
anyway, I got waylaid because I thought the service had the connection handle, which it doesn't unless the service is remote, because there's no way to find out the connection handle for a local service with a connection. You have to spy on it like you did in PacketBuffer. I wondered why I didn't run into this in CharacteristicBuffer, but it's because I never implemented writing.
service->connection is null for local services even if connected
slender iron
CharacteristicBuffer doesn't try to be exclusive to a single connection
slender iron
ah right, doesn't characteristicbuffer just notify on every connection?
tulip sleet
it only implements read, not write or notify, so it never ran into this problem
all it does is get notified
it doesn't send out notifications
slender iron
kk. that must be how the general set value for notify works
tulip sleet
but PacketBuffer does
slender iron
right
tulip sleet
sd_ble_gatts_hvx() needs a real connection handle, but sd_ble_gatts_value_set() doesn't care about the connection handle unless it's writing a CCCD
timber mango
@slender iron you ok with me grabbing a Toasty? https://www.tindie.com/products/webtronics/toasty-480mhz-stm32-usb-development-board/
@ionic elk I just ordered two Toastys. π π I can hardly wait to get them!
ionic elk
it looks like a pretty neat board
It'll be cool to try running some really crazy displayio stuff and see how it goes.
Though... that means I'll have to get displayio enabled for the H7
slender iron
@tulip sleet this is what I'm thinking of: https://github.com/adafruit/circuitpython/blob/master/ports/nrf/common-hal/_bleio/Characteristic.c#L156-L179
tulip sleet
i forgot about that. so packet buffer could do the same thing, instead of caching the conn_handle
tulip sleet
did you consider doing that, instead of grabbing the conn_handle on certain events in PacketBuffer?
slender iron
I didn't want to do that because it feels like the wrong model because midi is one to one in my mind.
tulip sleet
i'll leave it alone for now. I don't know if BLE MIDI is meant to support fanout to write-only devices, but from the spec, it sounds like not (like the single read packet)
slender iron
kk, time for a run
thorny jay
I'll contact the maker/vendor (8086.net) but I have a strange behaviour with the Commander board. Mu does not find the serial interface. However, if I use device manager to find the serial port, then use Putty to access COM26, I have the REPL.
Any idea?
That sound like a Mu problem. But it works on this computer with every other board I have (all Adafruit).
tulip sleet
Mu looks at hte USB PID/VID to see if it's a board it can talk to
thorny jay
So when new board appear...
tulip sleet
for Adafruit boards, it just checks the VID (Vendor ID).
What is the VID/PID for the Commander board?
thorny jay
I might need to boot in Linux for that... dunno how to do it in Windows.
tulip sleet
the check for only the Adafruit VID (as opposed to particular PID's) is relatively new
you can look at the properties in Device Manager
what version of Mu are you using?
tulip sleet
that already has the adafruit VID-only check
thorny jay
I upgraded the bootloader and CP.
tulip sleet
but the Adafruit VID is 0x239a
here the VID is 0x3171, and Mu doesn't know it
you can hack mu to add it
thorny jay
The VID/PID is in the bootloader?
tulip sleet
no, when CIRCUITPY is running (actually both, but Mu doesn't care about the one for the bootloader)
thorny jay
There are only a handfull of Commander (there was 5 in stock and it is brand new). I already aquired great stuff from 8086.net, so I felt like I should support and spend on this "unique" device.
I have a working solution with Putty.
I'll share this conversation with the maker, he might want to ask Adafruit for a VID/PID or check with Mu?
tulip sleet
@thorny jay you can add the board to mu/modes/base.py
i'm a little confused about the list there; there's also adafruit.py, which has the wildcard entry
orchid basinBOT
onyx hinge
@tulip sleet @slender iron is there something that needs to be done for board sorting due to me releasing 5.3.0? I am still fuzzy on that.
tulip sleet
the download counts are all zero to start, and we have to update them by hand. with zero counts, they show up in reverse alphabetical order, I think
I haven't started owrking on that yet. You could add fake download counts to sort them alphabetically or something. "Adafruit" comes conveniently at the beginning of the alphabet
maybe it should be fixed so that they show up in alphaetical order if the counts are zero
for now
ionic elk
@slender iron can I get an openMV-H7 PID?
thorny jay
@tulip sleet Yes, you made it... it works now. I will try to make a PR for Mu. Thank you so much.
tulip sleet
@ionic elk limor maintains the file of the PIDs. She pushes to an internal repo periodically, but it's probably best to ask her
I periodically refresh that to add a bunch of empty slots
ionic elk
Ok, sure. I don't know the workflow there, I just know it's usually Scott who posts the VID/PID collection for people.
@meager fog could I get an openMV-H7 PID?
tulip sleet
maybe send her an email or open an issue so she can queue the work. She could pass it to me too but she's pushing to that repo all the time
open a circitpython issue
and assign her
onyx hinge
typical PID request PR, if you want a template @ionic elk https://github.com/adafruit/circuitpython/issues/2829
tulip sleet
I;m afk for an hour or so
idle owl
@tidal kiln This issue: https://github.com/adafruit/Adafruit_CircuitPython_Display_Text/issues/11 doesn't make sense on this library anymore, as the solution appears to be changing the Group full message, which isn't generated by the library. Dan created another issue referencing it on CircuitPython - am I ok to close this one?
Or should I leave it until the issue is resolved?
idle owl
@slender iron Where are we at with pixelbuf? Asking as I'm going through lib issues and some are related or potentially resolved by pixelbuf, and I'm not sure of the status.
onyx hinge
@slender iron "build-mac" failed 5.3.0 but that doesn't inhibit all boards from building (we know why and it's fixed in master, but 5.3.0 is affected). Is it still okay to complete the 5.3.0 release in this state? https://github.com/adafruit/circuitpython/runs/630740349?check_suite_focus=true
slender iron
@idle owl @tidal kiln should be reviewing NeoPixel_SPI which has an open PR to update to newer pypixelbuf
@onyx hinge ya, fine with me
idle owl
Ok thanks
manic glacierBOT
tidal kiln
@idle owl yah, i'd say leave it open until the other issue is resolved. if that one results in changes that will work for also fixing the issue in Display_Text then can close it
idle owl
@tidal kiln Ok.
tidal kiln
looks like the issues have been cross referenced?
idle owl
@tidal kiln They have.
timber mango
Though... that means I'll have to get displayio enabled for the H7
@ionic elk This is a good thing, right? π I am looking forward to doing some crazy stuff with this too, especially for robotics. π
tidal kiln
cool. is there a pressing need to close it?
idle owl
@tidal kiln No, I'm cleaning library house basically. It's welcome to stay open, but if it could be closed, I wanted to close it. Trying to clear out some of the longstanding issues, but there are many that cannot be closed still. So it is what it is.
slender iron
any objections to me turning off bitbangio in the specialty cpx builds? we need to turn pixelbuf on to make the upcoming frozen neopixel useful
idle owl
No. Do it.
slender iron
π also means my lower power stuff will fit
idle owl
Also good.
tidal kiln
@slender iron looks good to go:
https://github.com/adafruit/Adafruit_CircuitPython_NeoPixel_SPI/pull/15#issuecomment-621428914
@idle owl ^^ fyi
slender iron
ok, awesome!
idle owl
@tidal kiln cheers.
timber mango
I am also getting a few display featherwings. I want to see if I can add multi-display support to the HT16K33 library for the alphanumeric and 7-segment displays. I could not quite fit the matrix displays into my Digi-Key order though because priorities.
ionic elk
@timber mango it shouldn't be tough to get displayio into the H7. The only reason it isn't already in is because DisplayIO by default requires PWM, which requires timers, which have some significant differences on the H7 series. But I think I can just turn it off with a flag, it's been a little while since I worked with it.
I agree it'd be a nice feature to have since powerful LCD capability is probably one of the top use cases for a chip that overpowered alongside camera work and DSP.
timber mango
I agree it'd be a nice feature to have since powerful LCD capability is probably one of the top use cases for a chip that overpowered alongside camera work and DSP.
@ionic elk Overpowered is good! π I am very excited about the Toasty board! I see a lot of possibilities for this board in my projects. It is breadboard friendly too! π
I wish I had been able to fit a couple of nice TFT displays into my order.
bright aspen
I'm a little confused about crypto. I can see things are happening, perhaps with the cryptography library. Is all of cryptography being added, or just parts. Will it be broken into pieces?
slender iron
@bright aspen definitely pieces so that we can enable/disable based on the board
xobs has a PR for AES currently
copper crescent
@thorny jay I added a PR a month ago to add commander support to Mu but not heard anything yet.
thorny jay
Oups. I did not check. So I made a duplicate...
@copper crescent Any plan in enhancing your documentation. I have tested most part using Adafruit documentation. I could write some stuff, like a bigger demo.
My commander run the lattest CP and lattest bootloader. So far, so good.
copper crescent
@thorny jay yes, I'm planning a week off the dayjob to try and catch up with things like that. :)
ionic elk
@slender iron what is the official name of the driver software for the i.MX series? Does it have a manual somewhere?
slender iron
not that I know of. I usually just look through the source and data sheet
marble hornet
okay: I am still trying add a new module in c. there continues to be build errors (see below) b/c the module does not have a rule to make it. from what i can see there are three places new module (w/out port specific code) need to be tied in:
py/circuitpy_defns.mkin theSRC_SHARED_MODULE_ALLlist andSRC_PATTERNSpy/circuitpy_mpconfig.hfor defining theextern _mp_obj_module_tpy/circuitpy_mpconfig.mkfor defining the contstant used to turn the module on or off
am i correct in my assessment that these are all the place? where else does it?
if so i am still getting build errors.
@tulip sleet gave me some good pointers to help find this information, these are all the places that pixelbuf is defined that i could find through a search tool but i;m not sure where else to add them.
marble hornet
the error is: the display is named tinydisplay and has been tied into the above three mentioned locations```
Use make V=1, make V=2 or set BUILD_VERBOSE similarly in your environment to increase build verbosity.
make: *** No rule to make target shared-module/tinydisplay/__init__.c', needed by build-pyportal/genhdr/qstr.i.last'. Stop.
@marble hornet can you push your code to a branch on your fork? would be easier to spot any disparities for folks that are up-to-speed on it. (i haven't in a while, but willing to take a look)
marble hornet
@raven canopy sure! sorry for delay, there was an unexpected dinner. https://github.com/TG-Techie/circuitpython/tree/add_tinytft
onyx hinge
@marble hornet you should create that above-named file (relative to the top source directory), as an empty file if necessary.
marble hornet
thank you! i'm getting a difference error now! (schrodinger's "yay")
marble hornet
where can i find the docs for mp_arg_check_num ? i'm curious what the laster argument is, and why it is often true EX:mp_arg_check_num(n_args, kw_args, 0, 0, true);
if there are docs for the inner working ?
raven canopy
no docs that i've ever seen. just the source code...
for which, there is a lot. π
marble hornet
IT"S WORKING !
raven canopy
π
marble hornet
and of much code that is sure
i have a question about how MP_QSTR work?
it seems like sometimes they are only used in one place and not defined as variables
is there source code manipulation to turn QSTRs into c char arrays in the background?
marble hornet
I am trying to write a getter for a new class (TinyTFT working title). the attribute i want to return is an instance of displayio's displayio_display_obj_t when given as an input it works fine, however the compiler says that is in an incompatible pointer type with an mp_obj_t.
does it need to be wrapped in something?
../../shared-module/tinydisplay/TinyTFT.c:11:16: error: incompatible types when returning type 'displayio_display_obj_t {aka struct <anonymous>}' but 'mp_obj_t {aka void *}' was expected
return self->display;
~~~~^~~~~~~~~
../../shared-module/tinydisplay/TinyTFT.c:12:1: error: control reaches end of non-void function [-Werror=return-type]
}```the branch is https://github.com/TG-Techie/circuitpython/tree/add_tinytft and the module name is tinydisplay
when it is passed in you heave to get the display itself by getting .u_obj from the argument
args[display_arg_index].u_obj) what is the wrapper that contains the .u_obj?
manic glacierBOT
manic glacierBOT
tulip sleet
@marble hornet
i have a question about how MP_QSTR work?
it seems like sometimes they are only used in one place and not defined as variables
is there source code manipulation to turn QSTRs into c char arrays in the background?
Yes there is a source code preprocessing step that finds all the QSTR's and generates the appropriate code for them. If you change a QSTR, do a clean make because the dependency checking for QSTR's won't find the changes (or at least it didn't used to)
marble hornet
@tulip sleet i was just curious if it was pre-processed or some combination of c features (but i could;t find any from googleing).
solar whale
FYI -- I built 5.3.0 after the release yesterday -- the build went fine, but they do not have the 5.3.0 tag -- is that expected?
Adafruit CircuitPython 5.3.0-54-g2d7cf4b79 on 2020-04-29; Adafruit PyPortal with samd51j20
I pulled from master -- It looks like the released version is the latest commit to the. 5.3.x branch. and I should have just checked out the 5.3.0 tag.... Someday I'll get this straight π
tulip sleet
@marble hornet QSTR stuff is specific to micropython, I think (it's not a general tool). The ideas may have been copied from elesewhere.
tulip sleet
π€·
raven canopy
public reminder: verify that you're using good USB cables. been fighting a sneaky USB bug on RosiePi for over a week when attempting to upload firmware. cable works on multiple computers; assumed it wasn't the issue. came to the end of my research abilities, threw a hail-mary on a different cable...issue solved.
ionic elk
I exile haunted cables to the ghost-box
solar whale
Life is too short for bad cables ...
ionic elk
it has markings of warding to keep the spirits contained
raven canopy
i mean, the RPi is the only one that has issue with it. it still has capability. this is an old motorola cable, which are quality in my experience. (granted, all of that belief is what got me here in the first place. π)
tulip sleet
@onyx hinge could we just remove it from the commited files? does weblate get upset if it's missing, or does it just restore it?
onyx hinge
it'll put it back every time, I think
tulip sleet
that's ok, if we just remove it before the commit
i think that would save useless diffs
onyx hinge
it's weblate that creates the PR, and I think if a human adds commits on top outside of the weblate service they get stomped on
tulip sleet
I just did this once : sed -i '/^"PO-Revision-Date:"/d'
oh, i didn't realize it created its own commits
onyx hinge
right, the idea is to use this web service to make it easier for community to contribute localizations
slender iron
@onyx hinge do you want to merge 5.3.0 for the website or should I update the download count for it?
onyx hinge
@slender iron oh I thought I had merged it βΉοΈ but I most certainly didn't do anything about download counts, because I couldn't figure anything smarter than hand-editing a huge json file and just "nope"'d out of it
if you can do it, that would be awesome and I'll owe you for it
slender iron
ya, nope is fair
dan has the ball rolling for it so I don't mind doing it manually
will do it into your PR
onyx hinge
you rock, thanks!
slender iron
np
onyx hinge
oh no wonder I'm hungry, it's lunchtime in the central US
lucid solar
I'm hungry as it's evening meal time in CET. But my lovely wife is making fish & mash π π₯
Can someone explain safemode for me? what causes it to go into it?
tulip sleet
reasons include: fatal error (e.g. hard fault), power brownout
typedef enum {
NO_SAFE_MODE = 0,
BROWNOUT,
HARD_CRASH,
USER_SAFE_MODE,
HEAP_OVERWRITTEN,
MANUAL_SAFE_MODE,
MICROPY_NLR_JUMP_FAIL,
MICROPY_FATAL_ERROR,
GC_ALLOC_OUTSIDE_VM,
PROGRAMMATIC_SAFE_MODE,
NORDIC_SOFT_DEVICE_ASSERT,
FLASH_WRITE_FAIL,
MEM_MANAGE,
} safe_mode_t;
I've always been curious what you can actually do from safe mode. What was the original purpose of it? Can you get it to reveal information about the crash?
Or is it just the above enum?
lucid solar
@tulip sleet thanks, will look into it after food
slender iron
the goal is to give you drive access if your user code gets you into deep trouble
tulip sleet
examples: bad user code, a bug in CircuitPython, power sag (we see this on the Trellis M4, for instance: user code turns on all 32 neopixels full blast, which draws too much current from a USB port)
there has been discussion of a setting to NOT go into safe mode when there's a power sag, for (example:) proper operation of solar-powered devices
slender iron
@onyx hinge website PR has download counts now
slender iron
@tulip sleet do we have bootloader updater for nrf?
tulip sleet
not yet, thach is working on it
timber mango
I exile haunted cables to the ghost-box
@ionic elk If I prove a cable (any cable) is bad, I cut the connectors off and scrap them. I check the wires for continuity, and If the wire is good, I keep it for future use. Otherwise I scrap that too. I can not have bad or even flaky cables around.
lucid solar
@slender iron I'm getting tired, so I've commited and did a draft PR of what I've got so far. I'll come back to it either later tonight or tomorrow.
tulip sleet
@slender iron could i have yet another video chat with you re packet_size?
slender iron
yup, got a bit of time now
simple pulsar
@lucid solar For safe mode, here's an old discussion mainly on reboots but one that touched on safe mode: https://forums.adafruit.com/viewtopic.php?f=60&t=151877
onyx hinge
It'll keep for In the Weeds on Monday, but I made good progress with weblate today, and I wonder what should definitely be done before inviting people to use it. "Soft launch" would mean actually asking people on Discord to use it, with the understanding that it might not work out in the end. "Launch" meaning stuff like newslettering and blogging it and being relatively committed to running the service as long as it's bringing in quality translations. Here's my internal list, with the (DONE) tasks being stuff I accomplished today: ``` * enable github auth (DONE)
- enable pull request creation (DONE, but re-disabled temporarily)
- enable google Cloud Translation (MT) (DONE)
- some sort of backups (DONE; digitalocean weekly "system level" backups)
- make sure we can accept the PRs it generates - (danh's metadata problem)
---- soft launch here? - Move the github authentication API keys to adafruit organization
- Move the github PR creation to adafruit organization (adabot?)
- enable weblate.circuitpython.org via DNS
---- launch here? - Move the Google Clould Translation (MT) to adafruit organization
cost: $20 / 1 million characters translated (probably not a lot of $$) - Move the digitalocean droplet to adafruit organization
hey folks. I'll need a quick review on this: https://github.com/adafruit/Adafruit_CircuitPython_NeoPixel/pull/83
it adds pypixelbuf as a pypi dep for neopixel
onyx hinge
@slender iron (I can review that PR but not merge it apparently)
(the neopixel one, that is)
slender iron
Ok! Lower power is merged and NeoPixel on PixelBuf is released!
let's see what bugs we find π
onyx hinge
:rubs hands together gleefully:
now I can do the testing I should have done before the PR was merged π π
slender iron
and then fix the bugs you find π
lone sandal
lucid solar
Does anyone know if there is a way to determine if you're inside the REPL, a script or waiting for "press any key to enter repl"?
lone axle
@lucid solar it looks like accessing __name__ from inside of code.py or another file will return the name of the file. But in the REPL accessing __name__ will raise a NameError
lucid solar
@lone axle Thanks, but not quite what I was needing, I want to know this inside of C.
lone axle
Ah, I imagine that functionality is backed by C code somewhere, perhaps you could get access to the same info at that layer. I'm not sure how to do that though.
lucid solar
Thanks anyway
onyx hinge
@lucid solar when at "press any key", the GC heap is deinitialized so MP_STATE_MEM(gc_pool_start) is 0
lucid solar
ok, that's probably good enough for what I want. Thanks @onyx hinge
lucid solar
I have to figure out if I need that...
onyx hinge
but you could add code in main.c:run_repl() to set a global and check it
ionic elk
@indigo wedge @slender iron I'm trying to help out with the pin claiming issues in the i.MX port but I'm kind of confused by the GPIO bank system. What's the relationship between the actual GPIO banks (GPIO1, GPIO2, etc) and the pad names? GPIO, GPIO_EMC, GPIO_AD, GPIO_SD... I can't find anything in the datasheet or reference manual that explains what the labels actually mean
indigo wedge
good question, i don't think there is much relation there, i think they reuse HW blocks from their CPUs and it's some legacy stuff
ionic elk
Hmmm gross. I dunno why they use that labelling when it seems like they have perfectly good GPIO blocks right inside
indigo wedge
yea it just adds to the confusion
their defines are also a bit off IOMUXC_GPIO_AD_01_GPIOMUX_IO15 vs IOMUXC_GPIO_SD_13_GPIO2_IO13
slender iron
@ionic elk did you find the mapping between pin and gpio?
idle owl
@slender iron Did this ever get addressed? https://github.com/adafruit/Adafruit_CircuitPython_MIDI/issues/7
orchid basinBOT
slender iron
@idle owl I thought so but I don't think I've actually tried it
idle owl
Ok I'll dig a little deeper then. Thanks.
manic glacierBOT
ionic elk
@slender iron wasn't really a specific issue, more a documentation confusion. Sounds like the answer is "they're using legacy names that don't make sense" which is what it is
slender iron
I think the pin naming might be power domain
ionic elk
Hmm, I'll look into that
slender iron
Β―_(γ)_/Β―
ionic elk
I was asking because I'm looking at implementing pin claiming so I just have to pick some way of sorting GPIO banks
But the internal ones are probably fine
slender iron
does the gpio handle muxing?
ionic elk
I'll do that. Real quick can I ask for your help on a git rebase thing? I screwed up the base of my last PR
slender iron
totally
ionic elk
I made my branch off another branch I was working on instead of master (which I could have SWORN I specifically checked out to not do but hey) and now I need to get these commits onto master
but all the rebase tutorials only talk about moving a base forward in history, not backward, so I'm afraid of messing up
slender iron
git rebase -i will give you a list of what commits to move over
ionic elk
My confusion is what constitutes moving over? I'm in my branch, do I need to make a new one?
Basically I'm trying to remove two commits from the start of my branch
that are from another branch
slender iron
yup, you can edit the list of what to do with -i
so you can just delete the first two commits
it lists back to the common commit between your current branch and your target I think
ionic elk
I'm just getting a rebase noop?
slender iron
what command are you running?
lucid solar
git rebase -i HEAD~4
try that
then assuming it opens in vi/nano...just delete the lines that are the commits you want to delete...then save...
@slender iron I've got a very weird issue, when in the REPL, it's reading lots and lots of \b (0x08) and I have no idea why
slender iron
from the uart?
lucid solar
yeah
slender iron
what is 0x08 in ascii?
lucid solar
\b (backspace)
slender iron
is the stlink sending it?
ionic elk
@lucid solar that did the trick! Thanks, I'm using that forever now
lucid solar
@ionic elk when I learnt it about 2 weeks ago...mind blown!
slender iron
@ionic elk do you understand why it worked?
ionic elk
Yeah it makes sense to just manually move back by X commits
lucid solar
I knew about rebase interactive...just didn't know about deleting commits
ionic elk
can't you just use drop? That's what I just did
lucid solar
@slender iron I'll get out my logic analyzer...
@ionic elk yeah, I forgot about that...
slender iron
to move back four commits you can just check out the older commit and make a new branch
ionic elk
true but my concern with making new branches is they can't fix already-submitted PRs
slender iron
they can be used in a new pr though
and if you are fixing something commited you'll want to start with the latest version of master
ionic elk
Right, I'd have to close the PR and open a new one. I've done that with git stuff that's gone really wrong but rebasing out a couple commits was all that was needed for this one
slender iron
to change an existing PR you can change the branch completely and force push it
beware that it makes existing comments hard to find though because they are associated with commits
lucid solar
@slender iron weird...the analyzer says it's 0xC9
slender iron
does seem like the repl should handle backspace
lucid solar
But I shouldn't really be sending it
slender iron
are you doing tab completion? maybe the internal autocomplete is trying to use it
manic glacierBOT
lucid solar
I'm not doing anything...I don't type anything...
lucid solar
Hmm...I was reading the TX not RX I think...that explains one part of it...
ionic elk
@slender iron I think github has some intermediary comment saving. I force push quite a lot, as does @lucid solar, and it preserves previous PR comments despite the history changing. The Zephyr folks actually require all PRs to be a single commit.
slender iron
I see the comments in the overview but not in the file diff
lucid solar
@ionic elk I always clean my commits....but they don't require a single commit...just clean ones. My last PR had 4 commits for 4 parts of the thing
slender iron
my preference is to rebase until making a PR and then merge from then on
ionic elk
makes sense
lucid solar
@slender iron Something really weird was happening...so I've upgraded stlink on that board and it's working again...not tested the \b thing yet though.
It's not sending the \b over the UART itself.
simple pulsar
@tulip sleet Would you like me to ticket that namedtuple thing? 3rd/4th posts in https://forums.adafruit.com/viewtopic.php?f=60&t=164835 for anyone else interesting in os.uname()
lucid solar
@slender iron the \b is because the HAL_UART_Receive is exiting with HAL_BUSY...it must be some default value or something weird...
@slender iron are you streaming in ~2 hours
slender iron
@lucid solar yup streaming at the normal time
lucid solar
@slender iron will you do esp32 stuff? it could be good for me to switch my brain off from uart for a bit π I've made a lot of progress, the only thing to fix is why it's reporting it's busy when it shouldn't be.
slender iron
Yup! Went through your changes yesterday and will talk through them today
Then going to do pin info
lucid solar
ok, cool. The biggest thing I learnt is that some linker files were missing, which I think caused the linker to crash. Just to sumerize π
slender iron
ooooooh! the other way to get it to not crash was to remove gc-sections
I should try adding it back
lucid solar
Yes...gc-sections stopped it crashing, because it didn't reach as far as it couldn't find the right linker files...
Can't explain it right...but gc-sections is good to have enabled
slender iron
kk
lucid solar
@ionic elk do you know why, after I receive one character with HAL_UART_Receive(&huart, &data, 1, 500) the next time it's called I get HAL_BUSY returned?
lucid solar
@slender iron IT WORKS! I can enter the REPL, exit it, reload the script...it all works!!
crimson ferry
@simple pulsar I use os.uname().machine quite a bit to adapt to whatever hardware the code is running on. Have you figured out a way to auto-ID SPI devices? I was using input pins but that was very limiting, so now using a manually-edited file to itemize SPI displays, etc.
idle owl
@gilded cradle Is this something simple and would you be interested in adding it? I recall you working with Pillow. https://github.com/adafruit/Adafruit_CircuitPython_MLX90640/issues/3
tidal kiln
@idle owl i think that's a one liner
idle owl
@tidal kiln Do you want to pick it up?
tidal kiln
sure. got the stuff right here. i can check and PR real quick.
idle owl
@tidal kiln Perfect thank you!
gilded cradle
Sorry @idle owl, just saw this. Thanks @tidal kiln.
tidal kiln
no worries
idle owl
@gilded cradle No worries! Carter's got it
gilded cradle
π
tidal kiln
@idle owl do you know why the examples are named the way they are? noticing the one named pygamer has nothing to do with pygame - looks like it uses displayio
idle owl
@tidal kiln No idea. Feel free to rename as needed. Do a quick search for guides, though they throw an error when code links are broken.
timber mango
@slender iron Are you deep diving today?? π
slender iron
@timber mango just started!
tidal kiln
@idle owl meh. punting. i have question. responded in issue thread.
idle owl
@tidal kiln Ok.
tidal kiln
@idle owl and figured out the filename - it's "pygamer" as in the PyGamer, not the Python pygame module π€¦ββοΈ
idle owl
Fair enough π
tidal kiln
@idle owl actually, back on, found something existing to just PR:
https://github.com/adafruit/Adafruit_CircuitPython_MLX90640/pull/17
tidal kiln
@idle owl done. i'd say release and close issue.
simple pulsar
@crimson ferry No, nothing clever I'm afraid, I was just after a basic way to distinguish between CLUE and CPB+Gizmo. I'm currently just checking machine and if I don't find Circuit Playground assuming it's a CLUE.
@crimson ferry In my particular case it's reasonable to assume the Gizmo is present if it's a CPB.
timber mango
@simple pulsar Does CLUE show up for the CLUE board?
simple pulsar
@timber mango Yep, ```>>> import os
os.uname()
(sysname='nrf52', nodename='nrf52', release='5.2.0', version='5.2.0 on 2020-04-09', machine='Adafruit CLUE nRF52840 Express with nRF52840')
As far as I've seen, those .machine strings are descriptive enough to uniquely identify each variant
(suspect that's by design)
timber mango
@gilded cradle I ordered the displays I need to start working on multi-display support for the HT16K33 based displays. Three each of the 14-Segment and 7-Segment displays. π
This is going to be a very interesting project, which will probably touch several sections of the library.
I will also be able to add animation support for the 7-Segment displays. π
marble hornet
animation on 7-segment?
timber mango
animation on 7-segment?
@marble hornet Yes! If you happen to have any of the HT16K33 14-Segment alphanumeric displays Adafruit sells, you could run the animation script I wrote. It allows one to do simple to medium difficult animations on those displays.
marble hornet
what kind of animation? like fading ? unfortunately i don't have any 7-segment displays
queen mason
Hello people
I'm new to hardware programming and I have some questions about Circuitpython/Micropython
I recently got some boards to start fiddling and learning, including an Esp32
I have been having fun with it using an ultrasonic sensor, I'm using micropython
but I am a bit confused with something. As far as I know circuitpython is a version of micropython for Adafruit boards
if my esp32 is not from Adafruit, can I use circuitpython?
solar whale
@queen mason CircuitPython does not support the ESP32 because it does not have native USB. There is work in progress to support the ESP32-s2
queen mason
@solar whale oh ok, thanks!
tulip sleet
@slender iron I rewrote .packet_size and related code, and discovered that ... at least first write (and maybe more of the early writes) when doing ble_midi_simpletest.py is/are not actually happening. PacketBuffer.write() used to just return quietly if the conn_handle was INVALID. Now I have it return an error value and shared-bindings raises and error, and it got raised. So the write is happening before intercepting the events that contain the conn_handle.
I'm thinking that maybe PacketBuffer should have a routine that lets you pass in the current connection for it to remember. That would make this easier.
slender iron
hrm, interesting
tulip sleet
it might just be a race: the client hasn't gotten around to setting the CCCD yet or writing that empty packet
tulip sleet
could have midi.start(connection) as something required before you start writing. Anyway, can wait until monday if you'd like; I just saw you were still on
manic glacierBOT
Shape uses a bitmap to represent blocked regions. This is fine
for some shapes, but others may be more compactly represented.
This lets users make large, possibly overlapping areas of their screen be colored without running out of memory due to bitmap arrays in adafruit_display_shapes.Rect.
Used like:
palette = displayio.Palette(1)
palette[0] = 0x001100
bitmap_gri...@crimson ferry I don't know what it's doing but I just ran some code that uses displayio and a gizmo and it runs ok without the gizmo. The only difference is a ~5 second pause somewhere in the initialisation. After that all the code seems to run without any ill effect to my surprise! I'd imagine some of the shapes are doing pointless drawing, just about to see if I can detect the absence.
simple pulsar
@crimson ferry I may be wrong about pause, that may just be my application. Possible and probably the gizmo library is just firing off signals "hoping" the gizmo is there.
vast kettle
I have been playing a bit with a 2.9" eink screen that uses the IL0373 driver (specifically the GDEW029T5) and it seems to work pretty ok with only minor changes to the examples (swap_rams=True and black_bits_inverted=True), but there is a black border all the way around the active pixels, and it's pretty thick on the 2.9" screen.
When I look closer on the smaller 2.13" tri-color (as on the e-ink featherwing) I also see a black border, but it a bit thinner and kind of hidden behind the edges of the screen, so not as noticeable.
After a bit of digging, I have found the VBD bits in the CDI command and changing the last byte in the start sequence from x37 to xf7 makes the border white (the light background color the display)
crimson ferry
@simple pulsar Yeah, if I'm reading the docs right, that range of st7789 displays don't even have a MISO connection, so they're write-only
low sentinel
Hi friends! I'm trying to get this pull request green and the hallowing_m0_express build is failing on GitHub.
253888 bytes used, -192 bytes free in flash firmware space out of 253696 bytes (247.75kB).
is what github says. When I build it locally to try and repro I see
251748 bytes used, 1948 bytes free in flash firmware space out of 253696 bytes (247.75kB).
I don't think I added 1948 bytes of flash demand in this pull request & I'm not entirely sure how to get to the bottom of this. Any ideas?
simple pulsar
@crimson ferry There might be some hacks to work out if the TFT Gizmo is attached: https://forums.adafruit.com/viewtopic.php?f=19&t=165042
@crimson ferry BTW, i think you need to power cycle to get rid of the library screen code once its active that will be holding loads of the pins in use even if there's no attached Gizmo. It lasts even when code.py has been interrupted, presumably to keep the screen alive for showing the serial console.
crimson ferry
@simple pulsar That's a useful find. I'll look into whether the backlight could be used on standalone displays. Yes, from what I understand, even imports can't be un-imported. I usually do conditional import and setup based on hardware config, but some cases are tough if you have to initialize to detect... at least the runtime code can be skipped, and device objects can possibly be manually de-inited.
Looks like there's a resistor-transistor arrangement on the standalone IPS display Backlight pin that may give a useful level for a presence test. I'll wire one up and see.
onyx hinge
@low sentinel Without any extra flags, you get an English build of CircuitPython. The CI builds all languages. The first failing language was korean (language code ko): "Build hallowing_m0_express for ko took 14.96s and failed"
the CI's build of english had 1948 bytes left just like you saw locally
low sentinel
Gotcha. So I guess I'm curious how I bloated the Korean build that much π
manic glacierBOT
onyx hinge
Other languages just take up more space in flash
low sentinel
You are in safe mode: something unanticipated happened.
CircuitPython core code crashed hard. Whoops!
Crash into the HardFault_Handler.
Who did this? You need lots of socially-distant hugs. Thank you! (I wrote a bug and this lovely thing rescued me gracefully)
indigo wedge
Twitter
manic glacierBOT
manic glacierBOT
ivory yew
Hmm if you do spi.write and then a spi.readinto is it normal to see what you wrote in the response buffer?
like i'm seeing bytearray(b'*Hello World'), and * is what I'm sending to spi.write and Hello World is what the device is sending back.
hrm if I toggle the CS pin between writing and reading it works as expected
slender iron
You are in safe mode: something unanticipated happened. CircuitPython core code crashed hard. Whoops! Crash into the HardFault_Handler.Who did this? You need lots of socially-distant hugs. Thank you! (I wrote a bug and this lovely thing rescued me gracefully)
@low sentinel That is exactly why we have it. π
low sentinel
Hey @slender iron thanks for the note on Github! :-)
I'm doing just that today, and while I'm at it I'm trying to put together a non-TileGrid binding for vector shapes.
I got full vector polygons working last night π¬. I was using displayio_vector as the module name; shapes felt squatty and it clashed in my mind with displayio.Shape. Any module naming input you have before the PR (whenever I get it ready) would be welcome!
slender iron
@low sentinel no problem! it's something I'd love to see included. We generally add io to the end of our modules so they aren't too generic.
we can brainstorm naming ideas π
low sentinel
shapesio, vectorio, display_shapes, ...
Lots of ways to skin it. I don't know if I have all the seds figured out to rename in place but I'll try π
low sentinel
It lets me think of factorio every time I say it.
K, I like it. I'll make it vectorio. Thanks!
onyx hinge
Me too (factorio)
low sentinel
>>> import vectorio
>>> help(vectorio)
object <module 'vectorio'> is of type module
__name__ -- vectorio
Polygon -- <class 'Polygon'>
Rectangle -- <class 'Rectangle'>
VectorShape -- <class 'VectorShape'>
looking reasonable. Time to exercise that Safe Mode...
ivory yew
ah so weird, I'm seeing that behavior with SPIDevice as well
with spi as io:
io.write(bytearray([42]))
with spi as io:
spi_result = bytearray(13)
io.readinto(spi_result)
print(spi_result)
# Prints bytearray(b'*Hello World!')
# the '*' is the 42 sent above, which shouldn't be in the read buffer?
Is that .. normal?
tulip sleet
@ivory yew what device is it?
ivory yew
I'm running the circuitpython code on a Metro M4
it's talking to a barebones SAMD21 board running custom firmware.
The firmware on that device just waits for a SPI command that equal 42 and responds with "Hello World"
void spi_respond() {
struct io_descriptor *io;
spi_s_sync_get_io_descriptor(&SPI_0, &io);
int32_t recv_count;
do {
recv_count = io_read(io, spi_in_buf, 1);
} while (recv_count == 0);
if(spi_in_buf[0] == 42) {
io_write(io, example_SPI_0, 13);
}
}
i think it may have to do with how the SPI peripheral works; looking...
so you are transmitting the last byte you read on the SAMD21 side
tulip sleet
in the SAMD51 datasheet
ivory yew
but it applies to the SAMD21 as well?
tulip sleet
yes, their peripherals are very similar
the exact same paragraph is 27.6.2.4 in the SAMD21 datasheet
i think you may not be ending the transaction or something. Did you take that code from an existing example? Maybe look at the Arduino SPI libraries for both sides
ivory yew
the Atmel example for SPI slave literally just had reading, not writing.
gonna do some more research
thanks, it helps just to know this isn't "normal" behavior.
tulip sleet
it's common to do a write_readinto, where you send a command and then start getting back data at the same time
ivory yew
I was looking at the ESP32SPI library for the CircuitPython side.
they send the command, then use a function to way for the "start response" byte then use readinto
which I think makes more sense here, as the device might not be ready to respond immediately.
tulip sleet
same problem: https://forum.arduino.cc/index.php?topic=578373.0 (on a '328!)
response in the thread is not so clear
i think you're late on the write
that last one is the clearest
raven canopy
@tulip sleet have you seen any complaints about BOOT drive issues on RPis/Linux? i've been fighting this for two weeks now, and the closest i can get is USB/FS errors; just not sure at what level since it just immediately locks up and logging isn't captured. new (to me) Grand Central is on a 1.x bootloader and doesn't have issues. Metro M4 was on 3.9, now 3.6, and immediately locks up. I'm walking back bootloader versions in hopes to find a diff...
ivory yew
interesting
"When transmitting several characters in one SPI transaction, the data has to be written into DATA register with at least three SCK clock cycles left in the current character transmission. If this criteria is not met, the previously received character will be transmitted."
tulip sleet
With SPI, as with so many other things in life, it's much harder being a slave than a master. from the last link
tulip sleet
i have not had to write an SPI slave driver yet, but I thought that ASF4 might provide an abstraction for you
tulip sleet
i've written code at that level, but only on the master side
sigh
@raven canopy @tidal kiln and I have both seen this, but we've been seeing it for months, well before the newer bootloaders. I used to get random Linux hangs, but then I installed kdump and it's really crashes. I even opened an issue: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1871143
I thought it was past the bootloader, in the TinyUSB in CircuitPython, but I think @tidal kiln said he saw it on Arduino too. It's quite frustrating to press reset on a board and have the host crash. It seems to be timing-related.
i've done a lot of websearching on this and haven't come up with similar reports
except maybe from years ago
i think it's some kind of storage-management issue in the driver, based on the backtraces from the crashes
raven canopy
that second dmesg is similar from what i've seen. i haven't seen any of it on my laptop (Ubuntu 18.04), so i was leaning toward it being related to the RPi's dwc_otg USB driver. nice (relative) to see it happening on ehci as well. i've been to narrow in my launchpad searches, since i haven't come across that report. i've been digging down the udev route... π
Seeing it plenty on the RPi with both 19.10 & 20.04.
tulip sleet
i have a very vanilla Dell Optiplex about 8 years old. I think it's happening more often, but I don't know whether to attribute it to more recent Linux kernels or more recent TinyUSB commits
I was running 18.04, now running 20.04. I upgraded to 20.04 early in the hopes it might have been fixed in the newer kernel series
@tidal kiln has an older HP machine
ivory yew
@tulip sleet thanks for your help, I think I understand this now, and I can make it work as-is. π
tulip sleet
I haven't seen it on my laptop either. But if it's ehci, it would be the same driver, though maybe that doesn't mean much if it's hw-specific in the driver
@ivory yew great! you probably understand it better than me already
raven canopy
yeah, it could definitely be hub/driver specific...