respect the NO_COLOR specification

[gasche]

Some people on the interest have decided to agree on an informal specification: if a NO_COLOR variable is defined in the user environment, then terminal tools should not enable colors by default (only if explicitly requested). The OCaml compiler shows colors by default in its error/warning messages, with various heuristics and configuration mechanisms to disable or refine it. It would be nice to respect the NO_COLOR specification.

More details: https://no-color.org/

[dbuenzli]

Is that a good specification though ?

I think the current tri-state value never, always, auto of OCAML_COLOR is what you want to be able to specify in practice (e.g. you want to be able to force colored output for cram testing). I wish they had standardized a COLOR variable that way.

If you want to proceed, what takes precedence when both variables are present ? I'd personally say the most specific i.e. OCAML_COLOR.

[gasche]

The webpage says:

User-level configuration files and per-instance command-line arguments should override $NO_COLOR. A user should be able to export $NO_COLOR in their shell configuration file as a default, but configure a specific program in its configuration file to specifically enable color.

I believe that, by the same reasoning, the more specific OCAML_COLOR should take precedence when it is specified, indeed.