Simplify some error handling

[faern]

This is in preparation of ditching error-chain. We had pretty tangled errors. Lots of our own errors are just stringly typed errors (ensure! kind of stuff).

This PR just massage away some errors.

• Types that implemented TryCopyTo but actually had no code path leading to an error was changed into implementations of the infallible CopyTo
• The internal compare_cstr_safe was changed to panic instead of returning an error. This because it only deals with parsing C strings returned by the system. Since the system should return C strings, it's something really bad going on with the system if they don't null-terminate their strings. I regard this as a case for panic rather than user error handling.

---

This change is 

[dlon]

Reviewed 7 of 7 files at r1, all commit messages.
Reviewable status: complete! all files reviewed, all discussions resolved

---

src/lib.rs line 162 at r1 (raw file):

    // SAFETY: We point to the same memory region as `c_chars`,
    // which is a valid slice, so it's guaranteed to be safe
    let c_chars_u8 = unsafe { slice::from_raw_parts(c_chars_ptr, c_chars.len()) };

I'm curious how you feel about simply casting this from *const [i8] to *const [u8]:

let c_chars_u8 = unsafe { &*(c_chars as *const [i8] as *const [u8]) };

I'm sure their bits will be identical.

[faern]

Reviewable status: complete! all files reviewed, all discussions resolved

---

src/lib.rs line 162 at r1 (raw file):

Previously, dlon (David Lönnhager) wrote…

I'm curious how you feel about simply casting this from *const [i8] to *const [u8]:

let c_chars_ut = unsafe { &*(c_chars as *const [i8] as *const [u8]) };

I'm sure their bits will be identical.

As discussed on slack: That is not sound. because the memory layout of &[T] can't be assumed to be the same as &[U]. https://users.rust-lang.org/t/how-to-convert-i8-to-u8/16308/11