Should rustc-dep-of-std feature really depend on c?

[RalfJung]

It used to be possible to almost entirely build libstd for some target without having a C toolchain (the "almost" being due to rust-lang/rust#56443: in the final step, a dylib gets built). To do this, we have to be careful to not enable the c feature of compiler-builtins. This could be achieved by disabling the default features of libstd.
However, that regressed when compiler-builtins got moved out of the workspace; now liballoc depends on the rustc-dep-of-std feature (curiously, libstd does not), and that feature depends on the c feature.

Is it really necessary that the rustc-dep-of-std feature depends on c?

[alexcrichton]

It isn't, just needs to be enabled somehow and that can be done with a feature of std instead of an implicit feature enabled by rustc-dep-of-std

[RalfJung]

libstd enables it per default.

So, does that mean we can remove it from rustc-dep-of-std?

[alexcrichton]

So long as libstd builds and uses the c feature for released builds I don't have a personal preference how this is all architected.