Used a macro for UUID generation, switched from C-style casts to reinterpret_cast, renamed callback

[Avamander]

I tried to separate these changes a bit more, but couldn't.

The PR makes some minor improvements to the music watchapp:

1. Switches to a macro for UUID generation. I think this might save some RAM+ROM, but primarily gets rid of all these hard-to-read assignments in the constructor.
2. Switches from C-style casts to reinterpret_cast
3. To avoid a name conflict, renames MSCallback to MusicCallback, also did a minor variable inlining (that the compiler will also do above -O0 very likely)
4. const-ified the value getters for more safety
5. Moved a few functions around a tiny bit to group getters

(It should remain functionally identical, at least my tests with GB didn't show any differences)

Let me know if I should strip out some of these changes.

I would also really like to know if the macro is okay.

[JF002]

Looks good to me!

2 small details:
I think MusicService::characteristicDefinition could be initialized in the initialization list instead of in the core of the ctor() like it's done in other services (ex : https://github.com/JF002/InfiniTime/blob/develop/src/components/ble/DfuService.cpp#L35).

The macro is fine. But it would be even better if we find a way to write a constexpr method to replace the macro :)

[Avamander]

I think MusicService::characteristicDefinition could be initialized in the initialization list instead of in the core of the ctor() like it's done in other services (ex : https://github.com/JF002/InfiniTime/blob/develop/src/components/ble/DfuService.cpp#L35).

I'd see if it can be done in a future PR, if you wouldn't mind?

[JF002]

I'd see if it can be done in a future PR, if you wouldn't mind?

Yes, of course :)

[jonvmey]

The ble_uuid128_t intializers can be replaced with a constexpr function that returns a ble_uuid128_t which eliminates the need for the macros. This has the additional benefit of allowing all of those members to be declared constepxr. I tried making it a member function function at one point but the compiler then complains about trying to call a method of a class before its definition is complete. It worked fine as a free function though.

Also the reinterpret_cast<ble_uuid_t*>s wouldn't be necessary if you changed (&msSomeUuid) to (&msSomeUuid.u) since ble_uuid128_t's first member u is a ble_uuid_t.

I can open a new PR with these changes if you're interested. I was going to suggest them as possibilities before the PR got merged.

[Avamander]

@jonvmey I would certainly like to see those changes. This PR was intended from the start to be a small incremental cleanup and I would certainly like to get rid of those reinterpret_casts.