AI junk

[sean-kim05]

Summary

click.style() (and therefore secho()) silently drops the 256-color index 0 (black) when passed as fg or bg, producing uncolored text.

The color arguments are guarded with a truthiness check:

if fg:
    bits.append(f"\033[{_interpret_color(fg)}m")
if bg:
    bits.append(f"\033[{_interpret_color(bg, 10)}m")

0 is a valid 8-bit color index (documented range "[0, 255]") but is falsy, so the guard skips it and no escape is emitted. Every other input works — nonzero indices, named colors, and even RGB black (0, 0, 0) (a truthy tuple) — which makes index 0 an inconsistency, not a design choice.

Reproduction

>>> import click
>>> click.style("x", fg=0)      # 256-color black
'x\x1b[0m'                       # color dropped
>>> click.style("x", fg=1)      # works
'\x1b[38;5;1mx\x1b[0m'
>>> click.style("x", fg=(0, 0, 0))  # RGB black works
'\x1b[38;2;0;0;0mx\x1b[0m'

Fix

Guard on is not None instead of truthiness, matching how the other style attributes (bold, dim, …) in the same function are already handled. None (the default) stays suppressed, unknown color names still raise TypeError, and index 0 is now emitted correctly.

The truthiness guards date to the original 2014 implementation when only named colors (all truthy) existed; integer/RGB support was later added inside the same blocks without updating them.

Tests

Added {"fg": 0} and {"bg": 0} cases to the parametrized test_styling table in tests/test_utils.py. They fail on main and pass with this change; the full suite is unaffected (1682 passed).

[kdeldycke]

This is right on time as I am myself digging into the ANSI-side of Click (see: #3653 and #3420). I confirm the issue, which also made me aware of a new class of edge-cases the style() method is not processing. This affects both my Click Extra project and Cloup (cc @janluke).

So I just updated your PR with an additional commit to illustrate all the edge-cases not covered in unittests, and also fix the behavior by validating all colors passed to style().

Now I need a quick pass from other maintainers to check that I did not miss anything but to me this is good to merged in the upcoming 8.5.0 release.

[kdeldycke]

@davidism do you want me to redo it?