--color=always doesn't flow through to diffs

[max-sixty]

What happened?

Currently --color=always doesn't affect the colorization of diffs

Reproduction steps

Take a failing test:

cargo insta test --color=always -p prql-compiler --lib --check -- test_upper

gives color in the diffs:

But when piped to a file, it only has color in the cargo parts, not the diffs:

�[0m�[0m�[1m�[32m    Finished�[0m test [unoptimized + debuginfo] target(s) in 0.63s
�[0m�[0m�[1m�[32m     Running�[0m unittests src/lib.rs (target/debug/deps/prql_compiler-454bbecdaff4c0a9)

running 1 test
test tests::test::test_upper ... FAILED

failures:

---- tests::test::test_upper stdout ----
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Snapshot Summary ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Snapshot: upper
Source: prql-compiler/src/tests/test.rs:3436
────────────────────────────────────────────────────────────────────────────────
Expression: compile(
    r#"
    from test_tbles
    derive [upper_name = upper name]
    select [upper_name]
    "#,
)
.unwrap()
────────────────────────────────────────────────────────────────────────────────
-old snapshot
+new results
────────────┬───────────────────────────────────────────────────────────────────
    0     0 │ SELECT␊
    1     1 │   UPPER(name) AS upper_name␊
    2     2 │ FROM␊
    3       │-  test_tables
          3 │+  test_tbles␊
────────────┴───────────────────────────────────────────────────────────────────
Stopped on the first failure. Run `cargo insta test` to run all snapshots.
thread 'tests::test::test_upper' panicked at 'snapshot assertion for 'upper' failed in line 3436', /Users/maximilian/.cargo/registry/src/github.com-1ecc6299db9ec823/insta-1.29.0/src/runtime.rs:569:9
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace


failures:
    tests::test::test_upper

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 148 filtered out; finished in 0.10s

�[0m�[0m�[1m�[31merror�[0m�[1m:�[0m test failed, to rerun pass `-p prql-compiler --lib`
�[0m�[0m�[1m�[31merror�[0m�[1m:�[0m 1 target failed:
    `-p prql-compiler --lib`

This also occurs for CARGO_TERM_COLOR

Insta Version

cargo-insta 1.29.0

rustc Version

1.65

What did you expect?

As above

[mitsuhiko]

Very surprised about this. Both insta (which renders the diff) and cargo-insta use console so they should share that setting.

[xxchan]

Actually this part by cargo test runner is also not colored.

running 1 test
test tests::test::test_upper ... FAILED

But by cargo test --color=always -- --color=always, according to (rust-lang/cargo#1983), you can have that part colored, but the diff is still not colored. 😄

[max-sixty]

If an example is helpful:

• Here's an insta diff: https://github.com/PRQL/prql/actions/runs/5049856404/jobs/9059847037?pr=2601#step:11:225

• Here's a diff from: pretty_assertions::assert_eq! of almost the same diff: `https://github.com/PRQL/prql/actions/runs/5048986899/jobs/9057951221?pr=2601#step:11:220

[max-sixty]

I still don't have a great mental model of this — but at least some of the reason is that libtest uses a different option from cargo, as @xxchan points out:

But by cargo test --color=always -- --color=always, according to (rust-lang/cargo#1983)

...hence the --color=always -- --color=always.

The good news is that the outputs might become consolidated under cargo (rust-lang/cargo#1983 (comment)), which might solve then solve this, or at least make it more tractable

[max-sixty]

Closed by #375 (I think). Passing CLICOLOR_FORCE in CI seems to work; it's still not clean given --color= doesn't always work.