Fix fmt UTF-8 ─ character as fill

[vinnichase]

fixes #17025

I created this PR as a zig learning opportunity.

My changes allowed me to exchange fmt.zig in my project, compile and run it on two platforms with the fill character being correctly formatted.

Of course my changes involve replacing u8s with u21 in various positions which allocates more memory. I clearly don't have the foresight to know if that is against the goals for fmt or causes other problems. I wasn't even sure if I should report a bug until I found the existing one since It is a very narrow edge case only covering one character. Maybe it was even intended to keep it ASCII letters.

However as of now I find it more intuitive to allow all unicode characters.

I'm also happy to discuss implementation details and revise my solution many times 😄

[nektro]

nice work! please add a test though to be sure and so it doesn't regress :)

[vinnichase]

nice work! please add a test though to be sure and so it doesn't regress :)

Test added. Thanks for feedback!

3 pipelines are green on my fork https://github.com/vinnichase/zig/actions/runs/7510739735
the rest aren't picked up by the runners, however I assume the others to succeed either.

[vinnichase]

To clarify, the issue I was referring to in my previous comment is that the following incorrect program:

OK gotcha! I misunderstood your first suggestion.

This issue needs to be fixed in three places in fmt.Placeholder.parse.

Added your suggested utf8EncodeComptime and invoked it in the three places. Works as you anticipated. Great!

Do we need a test for utf8EncodeComptime?

[jacobly0]

Implementation LGTM

[vinnichase]

@jacobly0

std.debug.print("wow thank you so much for the babysitting! {s:💙<3}\n", .{ "" });

I very much appreciate that. learned some good lessons today.