Development

Whenever someone needs to modify or enhance something in libgpiod, a few things must be done to prepare the development environment for it.

Environment setup

To consume the library all it takes is to have node and libgpiod itself, just as described in the README.

To extend it, however, you need to know better the libgpiod provided by the target platform.

A tale of fragmentation

So far, there is at least 3 big releases to be aware of:

• gpiod 1.4
• gpiod 1.6
• gpiod 2.0

The 1.5 version introduced new functionality still present in 1.6, 1.4 can be considered the last in the "old" libgpiod still present on many systems out there.

I decided to give up 1.4 series for now.

The 2.0 is a brand 'new' (several years old now) API with a very distinct way to work with chips and lines and requests.

Your openSUSE Tumbleweed daily driver might not be suitable for that work

Modern linux desktop is already shipping gpiod 2.x series, but node-libgpiod is still under development. It might be a bit challenging proper setup the system to compile against an older libgpiod version.

Currently my recommendation is to setup a reasonable recent linux distro with the proper gpiod version being worked on (1.6).

I have a Fedora 38 virtual machine just for that purpose at the moment, but i am thinking on other ways to work around this version issue.

Testing

Proper test is another challenge.

You see, the examples project exists because back then i had the time and the hardware to test it, but not the experience to write proper unit and integration tests involving this.

But if the kernel module maintainer wrote 100% software tests, so can we!

I hope so.

Mocking and simulating

There are two tools, two kernel modules, to help us on that endeavor: gpio-mockup, being discontinued, and gpio-sim, the maintained one.

There is this thread and good reference material to get started with.

The fedora 38 setup offers node libgpiod 1.6.4 and gpio-sim, which is a good combo for 1.x series. But it's an EOL'ed distro, so some alternative must be worked on in near future in order to keep 1.x going on.