The DWM1001 struct can't be created with RTFM (safely) about rust-dwm1001 HOT 4 CLOSED

Oh man, that really sucks! This means that RTFM is incompatible with any crate taking ownership of CorePeriphals/Peripherals, like my lpc82x-hal crate.

This is bad, in my opinion, because I think the whole extension trait pattern the other HALs have going on is deeply flawed. I wrote about that earlier. I've been planning to write a more detailed article about this issue. I need to prioritize this.

Unless there's some easy way out that I can't see right now, either someone needs to prove me wrong, or RTFM needs to be fixed. Otherwise we'll get to choose between sup-par static guarantees in HAL APIs and a split ecosystem (although, to be honest, I seem to be the only one who's split from the rest of the ecosystem right now :-) ).

from rust-dwm1001.

I'm mostly through your stream now. I think your particular problem can be solved, or at least alleviated, by making the constructors of Led, DW_RST and other such structs public. Then you could construct your own versions of those from what RTFM gives you.

The more general problem I talked about in my previous comment remains. Pretty sure I couldn't do this in lpc82x-hal without significantly weakening its static guarantees.

from rust-dwm1001.

@jamesmunns, @japaric (don't want to ping him needlessly), and I talked about this topic on IRC yesterday. Since RTFM only actually requires CorePeripherals, and handles Peripherals just as a convenience to the user, the broader problem of RTFM clashing with all libraries that take Peripherals can be solved in RTFM. See https://github.com/japaric/cortex-m-rtfm/issues/163.

The more specific problem in this library can be solved by the following action items:

1. Make constructors of Led, DW_RST, and similiar structs public, as previously mentioned.
2. Remove anything related to CorePeripherals from DWM1001, or somehow make it optional. I'm not sure right now whether having CorePeripherals is just a convenience or actually required. If it's a convenience, it would be nice to make it optional, if that wouldn't be too complex, but removing it would be fine. If it's required, a solution needs to be found.
3. Make constructor of DWM1001 public too. I'm have this as a separate item, since the constructor may need to change, depending on how item 2. shakes out.

from rust-dwm1001.

I'm going to close this. The immediate problem was fixed in #82. The more general problem is being discussed in https://github.com/japaric/cortex-m-rtfm/issues/163.

from rust-dwm1001.