#[disable_frame_system_supertrait_check] doesn't seem to work as expected

[liamaharon]

I expect to be able to write a simple pallet where the config doesn't depend on frame_system::Config like this:

#[frame_support::pallet]
pub mod pallet {
	use super::*;

	#[pallet::pallet]
	pub struct Pallet<T>(_);

	#[pallet::config]
	#[pallet::disable_frame_system_supertrait_check]
	pub trait Config: 'static {}
}

Instead, it fails to compile

See my branch to reproduce https://github.com/liamaharon/substrate-node-template/blob/liam-disable-supertrait-issue/pallets/template/src/lib.rs

Is there something wrong with my pallet, or is this a bug with the macro?

[kianenigma]

I was initially tripped by this too, but I think it is all correct.

My mistake was also to naively think that frame actually allows you to avoid frame_system. It does not. Having T: frame_system::Config is 100% mandatory. All the code generated for things like Hooks, PalletInfo and more is unavoidably reliant on the existence of frame_system.

What disable_frame_system_supertrait_check is to just syntactically hide this.

For example, you can have things like:

trait MyOpinionatedSystem: frame_system::Config {}

You can then write pub trait Config: MyOpinionatedSystem { .. } and it would be valid with this attribute.

Or, if you tightly couple to another a pallet, and the other pallet is already bounded by frame_system::Config, you don't need to name : frame_system::Config again.

Can you post this in SE as well please, if no similar previous question exists? And I will asap write paritytech/polkadot-sdk-docs#76

[liamaharon]

OK this makes a tonne of sense, thanks!

I've updated the documentation to address this in #2638. Not sure if a SE post is required if it's explained in the docs, but happy to create one if you think it would still be valuable :)

[gupnik]

We already do have a pallet in tests without frame_system::Config using that macro:

polkadot-sdk/substrate/frame/support/test/src/lib.rs

Line 40 in b0741d4

#[pallet::disable_frame_system_supertrait_check]

It exposes a prelude and adds a few items in the config. I imagine the similar changes would help here. If so, agreed that it would be good to get the docs updated.

[liamaharon]

If this is possible, could #[pallet::disable_frame_system_supertrait_check] add the necessary items from config for the dev so they don't need to do it themselves? Then it should function like we initially thought?

[bkchr]

Then it should function like we initially thought?

As you already found out, this is not the purpose of this macro. The linked "pallet" is also a test and nothing that should be repeated in the real world. FRAME expects that every runtime has the system pallet and that the config is a super trait. This is backed deeply into the entire FRAME macro machinery etc and nothing we want/can change.