2.1 KiB
Changes to the microsoft/surface top-level
Overview
When I (@mexisme) created the microsoft/surface profile, there weren't that many differences between the various models of Surface.
I had just acquired a Surface Go 1, and it was mostly safe to enable all the options for all the models, and they would fail gracefully enough that we could mostly ignore warnings or errors.
Now, however --- as-of 2023-01-10 --- we have a much wider variety of chipsets, incl. models with some of the newer AMD CPUs, and this is breaking small things in annoying ways for more people.
Changes
Model Specialisations
In keeping with the broader structure of "nixos-hardware", I've also changed the structure of the microsoft/surface profile to make it easier for people to specialise for their hardware.
Any code or modules that are specialised for a Surface model now have their own directory under this top-level.
E.g. I am moving the Surface Go specialisations into surface-go/, and there may be further
specialisations like surface-go-3/.
"Common" modules
All the "common" modules that were once in the top-level of the microsoft/surface profile have moved
under the common/ directory.
Tools / services that are shared among several models are now extracted to their own module under
common/ and imported by common/default.nix.
These new "common" modules now have an enable option, which is false by default.
"Old Behaviour" module
The original ./default.nix module has been replaced by a new old/ module, which replicates the
original behaviour.
The new ./default.nix will load the new old/ module, but will also pop-up a warning asking users
to update how they use the microsoft/surface profile from now on.
This warning will probably change to a fail assertion at some point in the future.
Adding a new Model Specialisation
This hasn't been finalised, partly as I now only have access to a Surface Go 1, these days, so I'm maybe not the best custodian of this code any longer.
However, hopefully the (imminent) surface-go/ module is a reasonable exmample, and we should be
able to gather more examples for more model specialisations over time.