[Rohilalala]

What type of PR is this?

• bug

What this PR does / why we need it:

With wrapAt set, wrapLine measured lines and words in bytes, so help text with multi-byte characters wrapped before the configured column. At wrapAt 30, app - ändert die Größe für Überprüfung broke after "Größe"; it now breaks after "für". offset and the tabwriter that aligns help output already count runes (#2493), and this makes wrapping agree with them. Wide East Asian characters now count as one column each, as they do in the tabwriter, so text in those scripts can run past wrapAt.

Which issue(s) this PR fixes:

None.

Testing

TestWrapLineCountsRunes in help_test.go.

Release Notes

Help text wrapping counts characters instead of bytes, so non-ASCII usage text no longer wraps early.

[abitrolly]

LGTM.

[paoValle]

Counting runes matches textWidth and the tabwriter, so wrapping and alignment now agree for multi-byte text. Two things before this merges: the new test only covers offset 0, while the diff also moves the spaceLeft arithmetic in the offset path, so a case with offset 4 and two multi-byte words would pin what it changed. And runes are not columns: 日本語 is three runes and six cells, so CJK help text now under-wraps where it used to over-wrap. Consistent with the tabwriter, so not wrong, but the comment could say "runes, not display columns" and save the next reader the assumption about terminal width.

[Juneezee]

I rebased this onto current main, applied abitrolly's suggestion, and updated the CJK sentence in the description. The suggestion computes wrapAt - offset once and doesn't change behavior.

I reproduced the example from the description on main, and the fix works. Changing any of the five rune counts back to len fails the new test. Offset only sets the line width, so I don't think the test needs an offset > 0 case.

On the CJK point above: wide characters now count as one column each. Korean usage text can run past wrapAt (the first line is 47 cells wide at wrapAt 30). Counting display width would need East Asian Width data, which the standard library doesn't have. offset and the tabwriter already count runes, so I'd keep this consistent with them:

cli/help.go

Lines 577 to 581 in d854bb4

	// offset counts runes, not bytes, the same way the tabwriter that aligns
	// the help output does.
	func offset(input string, fixed int) int {
	return utf8.RuneCountInString(input) + fixed
	}