cargo build respects NO_COLOR=1 but cargo test does not

[heaths]

Problem

I was adding NO_COLOR support to our code and noticed that cargo build respects NO_COLOR=1 but cargo test does not e.g.:

But:

Steps

1. cargo new repro --lib
2. NO_COLOR=1 cargo build
3. NO_COLOR=1 cargo test

Possible Solution(s)

No response

Notes

I appreciate there are lots of well-known/used environment variables that amount to the same thing e.g., FORCE_TTY, and I'm not advocating to support them all. It's just not possible. But it would be nice to support the ones cargo already does consistently.

Looking at some other bugs, I see you're using isatty for detection and that does seem to work in most cases e.g., piping to more. But if someone wanted to redirect output to a PTY e.g., in their own wrappers, they might still want a way to force color off since VT sequences may not be supported.

Version

[epage]

This appears to be a duplicate of #1983 and so I'm closing in favor of that. If there is a reason to keep this open, let us know!

[heaths]

Thanks. I searched for dups but didn't see that one. Found a couple similar but different behaviors e.g., build used to write color sequences regardless.

[epage]

Yeah, the core of the issue is that a test harnesses color display is independent of Cargo. We want to change that. As we stabilize json output from libtest, we want to shift to more of a cargo/rustc-like relationship where libtest outputs what it wants and Cargo is responsible for stripping it when needed. We'd like to go even further and put Cargo in charge of rendering all of the results.