On Wed, Aug 20, 2025 at 06:04:40PM -0400, Jeff King wrote:
> I'd expect something like this:
>
> diff --git a/add-interactive.c b/add-interactive.c
> index 3e692b47ec..ad8b4907e1 100644
> --- a/add-interactive.c
> +++ b/add-interactive.c
> @@ -50,6 +50,8 @@ void init_add_i_state(struct add_i_state *s, struct repository *r,
> else
> s->use_color =
> git_config_colorbool("color.interactive", value);
> + if (s->use_color < 0 && !repo_config_get_value(r, "color.ui", &value))
> + s->use_color = git_config_colorbool("color.ui", value);
> s->use_color = want_color(s->use_color);
>
> init_color(r, s, "interactive.header", s->header_color, GIT_COLOR_BOLD);
>
> to work, but it doesn't seem to. Maybe the diff code is independently
> looking at git_use_color_default, and we really do need to set the
> variable?
Ah, indeed. There's yet another bug here. And while adding a test for
that, I found a third bug. Yikes.
So here's a series which I think addresses everything I found. These
bugs have been lurking for a while, but I guess not many people tend to
set color variables to anything exotic.
[1/4]: stash: pass --no-color to diff-tree child processes
[2/4]: add-interactive: respect color.diff for diff coloring
[3/4]: add-interactive: manually fall back color config to color.ui
[4/4]: contrib/diff-highlight: mention interactive.diffFilter
add-interactive.c | 88 ++++++++++++++++++++++-------------
add-interactive.h | 7 ++-
add-patch.c | 12 ++---
builtin/stash.c | 4 +-
contrib/diff-highlight/README | 8 ++++
t/t3701-add-interactive.sh | 51 ++++++++++++++++++++
t/t3904-stash-patch.sh | 10 ++++
7 files changed, 138 insertions(+), 42 deletions(-)
-Peff
