Re: [RFC PATCH] color: respect the $NO_COLOR convention

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 1, 2018, 17:10 UTC
Message-ID: <xmqqefl3iuvx.fsf@gitster-ct.c.googlers.com>
In-Reply-To: <87zi3reoez.fsf@gmail.com>

Leah Neukirchen <leah@vuxu.org> writes:

> NO_COLOR (http://no-color.org/) is a comprehensive approach to disable
> colors by default for all tools:

The list of software that supports that "convention" is, eh, respectable. Is it really a "convention" yet, or yet another thing the user needs to worry about?

Show 10 quoted lines

> diff --git a/color.c b/color.c
> index d48dd947c..59e9c2459 100644
> --- a/color.c
> +++ b/color.c
> @@ -326,6 +326,8 @@ int git_config_colorbool(const char *var, const char *value)
>  
>  static int check_auto_color(void)
>  {
> +	if (getenv("NO_COLOR"))
> +		return 0;

Our convention often calls for CONFIG_VAR=false to mean "I do not want to see what CONFIG_VAR wants to do done", i.e.

	NO_COLOR=false git show

would show colored output if there is no other settings. But this code contradicts the convention, deliberately because that is what no-color.org wants. Makes me wonder if that convention is worth following in the first place.

>  	if (color_stdout_is_tty < 0)
>  		color_stdout_is_tty = isatty(1);
>  	if (color_stdout_is_tty || (pager_in_use() && pager_use_color)) {

According to no-color.org's FAQ #2, NO_COLOR should affect only the "default" behaviour, and should stay back if there is an explicit end-user configuration (or command line override). And this helper function is called only from want_color() when their is no such higher precedence setting, which is in line with the recommendation.

Which is good.