print contents of sentinel-terminated pointers in std.fmt

[daurnimator]

Not sure if we need to make error messages more explicit? Would be a good place to use #1669 (comment)

[andrewrk]

The way the test was before is testing what it intends to test - calculating the length of a null terminated pointer. e.g. the zig equivalent of strlen.

[daurnimator]

Should it perhaps be .ptr instead?

[daurnimator]

The type of a string literal is *const [9:0]u8; IMO asking for the length of a pointer seems weird, and we need to .* to make sense...

[LemonBoy]

ping, @daurnimator any chance you could rebase this and address Andrew's comments?

[daurnimator]

Rebased and addressed some comments.

I went with a new function mem.pointerToSlice where you pass the result type that you want: that result type is used to know the sentinel as well as the const-ness (thereby avoiding the need for two functions: one for const, one for non-const as we currently have).

[andrewrk]

I'm not ready for any of these changes to std.mem. Can you just make pointerToSlice a private function in fmt.zig, or, better yet, inline the logic?

Here's a new set of functions to std.mem I would be willing to add (and adding deprecation notices to toSlice, toSliceConst, and len):

/// Takes a pointer to an array, a sentinel-terminated pointer, or a slice, and
/// returns a slice. If there is a sentinel on the input type, there will be a
/// sentinel on the output type. The constness of the output type matches
/// the constness of the input type.
pub fn Slice(comptime T: type) type {
    // TODO
}

/// Takes a pointer to an array, a sentinel-terminated pointer, or a slice, and
/// returns a slice. If there is a sentinel on the input type, there will be a
/// sentinel on the output type. The constness of the output type matches
/// the constness of the input type.
pub fn slice(ptr: var) Slice(@TypeOf(ptr)) {
    // TODO
}

/// Takes a pointer to an array, an array, a sentinel-terminated pointer,
/// or a slice, and returns the length.
/// TODO deprecate `len` and make this become the new `len`.
pub fn length(ptr: var) usize {
    // TODO
}

I believe then std.mem.slice would take the place of pointerToSlice in the fmt.zig code.

The rest of the PR looks good. The changes to std.mem are my only concern.

[daurnimator]

Can you just make pointerToSlice a private function in fmt.zig, or, better yet, inline the logic?

How would this then help with #4304 (comment)