(bug) isDarkColor returns incorrect value for some blues

isDarkColor (color.go:225) takes its lightness from Hsl(), which is (max+min)/2 of the channels and is not perceptual. it returns exactly 0.500 for all three saturated primaries, so pure blue -- relative luminance 0.072 -- is reported light.

reachable from public api, not internal. HasDarkBackground calls it, and compat/color.go:13 evaluates

HasDarkBackground = lipgloss.HasDarkBackground(os.Stdin, os.Stdout)

as a package-level variable at init. the v2 upgrade guide's quick start, item one, tells every v1 user migrating AdaptiveColor to use compat. on a saturated blue background that resolves the wrong branch once, for the whole program.

contrast, measured: a developer's light-background choice is typically dark text, which against #0000FF is 1.47:1 to 2.03:1 where WCAG AA asks 4.5:1. the dark-background choice would have been 8.59:1. included image shows both

the fix is Hsl() -> Lab(). go-colorful is already imported and normalises L to 0..1, so the threshold does not move and no dependency changes. every real terminal background gives the same answer as before -- solarized dark and light, dracula, DOS blue, black, white -- and the only value that moves is saturated blue.

Setup Please complete the following information along with version numbers, if applicable.

• OS macos
• Shell zsh
• Terminal Emulator ghostty
• Terminal Multiplexer tmux
• Locale en_US.UTF-8

Source Code

• see difff

Screenshots

• see above