Compiled MuxTo binary too large

[thomome]

Hello everyone,

I was trying to compile and upload the MuxTO firmware for the Arduino Nano Every. I followed the guide https://www.arduino.cc/en/Guide/NANOEvery and was able to compile the sketch with arduino-cli.

arduino-cli compile -e -b MattairTech_Arduino:samd:muxto:float=default,config=enabled,clock=internal_usb,timer=timer_732Hz,bootloader=4kb,serial=two_uart,usb=cdc

After exporting the binary, I noticed that it was 280 bytes larger than the precompiled firmware binary from the repo. When I anyway tried to upload it with bossac it failed with the following error.

Does anyone have any idea what I am doing wrong?
Any help would be appreciated.

[facchinm]

Hi @thomome ,
the project grew a bit since the instructions were last written (I should have checked it, my fault).
The easiest way to make it fit again in 12KB of flash is to use a newer compiler. LTO should be a viable solution but some patches have not been ported to that core, so it's unlikely to compile out of the box.
For the records, arch linux default arm-none-eabi-gcc compiler works just fine 🙂

[per1234]

I should have checked it

I can set up CI for it if you like. The workflow I wrote for the sketch compilation "smoke test" of the Motor Carrier firmware can be a good reference:
https://github.com/arduino-libraries/ArduinoMotorCarrier/blob/master/.github/workflows/compile-firmware.yml

[facchinm]

@per1234 it would be great! The only step to add is to check that the final .bin is smaller than 12288 bytes (0x3000), since the size report is reporting a slightly smaller size.

[per1234]

Hi @facchinm. I'm afraid I don't understand the requirement. Shouldn't the compilation process fail already if the compiled sketch size exceeds the upload.maximum_size property value for the board?:

https://github.com/arduino/ArduinoCore-samd/blob/5c850ac24d9d2b5ea0b322d3fb44dbfd0b63606f/boards.txt#L1617

muxto.menu.bootloader.4kb.upload.maximum_size=12288

I can certainly add another step to the workflow to run arm-none-eabi-size on the compilation output, and a regex on the output of the command, but it doesn't seem like it should be necessary since Arduino CLI does that already.

[facchinm]

Basically arm-none-eabi-size doesn't take into account the vectors section, resulting in a smaller number than the actual size that we'll then try to upload (and fail).

[per1234]

Thanks for the explanation! Here is the PR: #105

[ganghuang]

@facchinm

As this issue shows that the latest version can not compile successfully for the size of arduino nano every.
Would you please help to point out which commit in this repo can be compiled successfully for nano every with 4k bootloader?

In the discussion
#51
mentioned the version ee0a9bd can be compiled and I tried the MuxTO in that zip file and it basic functional. But I can not reproduce the same .bin file.

And also I want to know who compiled the "factory" version of the firmware in the purchased nano every board? Which commit does it associated with?

Thanks,

[ganghuang]

I modified the MuxTO.ino as attached.
Now the project can compile to 12284, the size of nano every need <12288;
So I can upload the bin to the SAMD11D14A on the nano every.

But after upload I can not load the regular sketch and I got the error
avrdude: jtagmkII_getsync(): sign-on command: status -1
avrdude: jtagmkII_getsync(): sign-on command: status -1
avrdude: jtagmkII_getsync(): sign-on command: status -1
MuxTO.ino.txt
And also the diff to the MuxTO.ino in repo as of 4137a59 ( which is same file as e458d1d)
MuxTO.ino.diff.txt

Basically, I did the following change:

1. move global variables to the loop function except the serial_mode, baudrate and lock q;
   this doesn't work as the code need value pass between different iterations
2. we are not using the LED_BUILDIN here,
3. change the two updi_mode if to if/else structure
4. For several functions which has been called multiple times, combine them to one single call and give the result to a variable
5. Use static const for the mask and bit set for different serial new value
6. Combine the serial_mode & mask and |= new value together

For me it doesn't seems I have changed the behavior of the code, but I am not sure, especially it is not working yet.

[ganghuang]

Since my last modification is wrong, I am thinking about what can we do to reduce the size.
I think there are two possible directions:

1. Reduce the size of crc table by change it to calculation
2. Reduce the unused hardware support . Under the same building condition and start an empty sketch with empty setup() and empty loop(). I got a bin file at 8864 bytes.

[mcuee]

Unfortunately the github action never runs successfully due to this issue.

• Add "smoke test" sketch compilation CI workflow for MuxTO firmware #105