printf: panics on a numeric field width above 65535 (%65536d)

[sylvestre]

printf panics instead of printing when a numeric conversion is given a field width above 65535:

$ printf '%65536d' 5
thread 'main' panicked at src/uucore/src/lib/features/format/num_format.rs:781:17:
Formatting argument out of range
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
$ echo $?
101

GNU prints the padded field and exits 0:

$ /usr/bin/printf '%65536d' 5 | wc -c
65536

The boundary is exactly u16::MAX: %65535d works, %65536d panics. It affects numeric
conversions (%d, %f, ...) but not %s, so %65536s is fine.

The cause is that the width is passed straight to Rust's formatting machinery, which only
accepts a width that fits in a u16. The fix is to stop relying on format! for the padding
once the width exceeds that, and emit the padding directly.

Where

src/uucore/src/lib/features/format/num_format.rs, in the NumberAlignment handling — the
write!(writer, "{s:>width$}", ...) / write!(writer, "{sign_indicator}{s:>remaining_width$}")
calls (around line 781 on current main).

Related, but not the same code

The same "Formatting argument out of range" message is reported in #12593 (%111111c) and
#12900 (%-101727s), which both panic in src/uucore/src/lib/features/format/spec.rs. This one
is a separate site in num_format.rs and still reproduces with those unfixed. They share a root
cause, so whoever fixes one may want to look at all three.

Why this is a good first issue

The reproducer is a single command, the failing line is known, and the fix is local to one
function. A regression test belongs in tests/by-util/test_printf.rs.

Found with coreutils 0.10.0 (multi-call binary) at 282b8b4, compared against GNU coreutils
9.11.64.

[MannXo]

I'd like to give this a shot if it's unassigned.

[Devel08]

#12708 is an issue that covers that too