Photon 2 and Edge-Impulse local compile

Firmware
1

Hello, I am trying to build an Edge Impulse audio inference project for Particle Photon 2 using Particle Workbench on Windows.

I am working on IOT and using your fantastic products.

I am looking to see if introducing AI in IOT and with pleasure found a way to develop with Photon 2.

But get frustrated using the Particle Work Bench and the compile.

I already made the Ingestion using the Photon 2 and Microphone.

I manually inserted my numerous records and noise and unknow in Edge Impulse online.

I deployed the result using the Particle Library Compile and imported the project in Particle Workbench.

The compile result: Argument list too long…

Setup

  • OS: Windows 10/11

  • Particle Workbench (latest)

  • Device OS: 6.3.4

  • Device: Photon 2 (RTL8721DM, Cortex-M33)

  • Edge Impulse project: audio keyword detection

  • EI export: Particle / C++ SDK

Problem 1 – Windows build fails

Local compilation on Windows fails with:

/bin/bash: arm-none-eabi-gcc-ar: Argument list too long

This happens even with:

  • very short project paths

  • no libraries

  • clean project

  • fresh Workbench install

The same project builds successfully on Linux, proving the source is valid.

Problem 2 – Cloud compile unavailable

Particle cloud compile is not available for Photon 2 + Edge Impulse projects, so there is no Windows fallback.

Summary

  • :cross_mark: Native Windows local compile fails

  • :cross_mark: Cloud compile not available

  • :white_check_mark: Linux build works

Expected

At least one of:

  • Some help for Windows users

  • Cloud compile support for Photon 2 EI projects

  • Confirmation that Windows local builds are currently unsupported

At the moment, as Windows users I am blocked from local development with Edge Impulse on Photon 2 unless I use Linux. (I cann’t install WSL)

Thanks for any help.

2

it's difficult to build custom Edge Impulse projects on Windows because you will run into the argument list being too long because there are too many files in the Windows build. And cloud build does not work because the project is too large as well.

In the examples that we pre-built (Doorbell and You're Muted) we ran a script that stripped out the unnecessary files so it can be built on Windows or a cloud build, but this is a chicken-and-egg problem because you need to have a successful build to figure out which files are removed by the linker.

The only viable option for Windows is to use the Docker build, if you can install Docker.

1 Like
3

@rickkas7 thank you for your prompt answer.
I think this is bad news.

I will check if I can investigate in some other hardware, because first I have to understand AI in IOT and not spending too much time in compile understanding, plus… I never used Docker yet.

BTW: I had also problems with Particle Workbench in a Windows 10 computer. Every time I was compiling a Particle code the system32\version.dll was corrupted and my PC crashed.

After numerous Windows fix I could get rid of the problem by a new VSCode and Particlebench fresh install. This is scary.