panic if TERM environment variable is "dumb"

[jimblandy]

If I do a compiletest test run under Emacs, Emacs sets the TERM environment variable to "dumb", and then the test run panics:

running 2 tests
test [compile-fail] asynchronous/async-fn-not-unpin.rs ... 
thread '[compile-fail] asynchronous/wrong_mutex.rs' panicked at 'called `Result::unwrap()` on an `Err` value: "SendError(..)"', /home/jimb/.cargo/registry/src/github.com-1ecc6299db9ec823/tester-0.7.0/src/lib.rs:598:5
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
test [compile-fail] enums/exclusive_range_patterns.rs ... 
...

If I change the test command to TERM=ansi cargo test, then all is well.

This is the panicking line, in the tester crate's run_test_in_process function:

    monitor_ch.send(message).unwrap();

where monitor_ch is a Sender<CompletedTest>.

The only thing I can imagine compiletest uses TERM for is the color in the output. So I'd guess that whoever is supposed to be receiving on this channel exited early for some color-related reason. Ideally, if color is not available, one should simply not emit the color escape codes.

[Munksgaard]

Hi @jimblandy, thank you for reporting this issue. I'm having trouble recreating it on my local machine. Running TERM=dumb cargo test in our test-project doesn't seem to exhibit the error you've described. What project are you running this on? Can you perhaps supply a minimal working example?

Furthermore, if the panicking line is in the tester crate, I think it would be best to report the error there?

[jimblandy]

Here's a test case: https://github.com/jimblandy/compiletest-issue-219

[jimblandy]

This is on Fedora.

[jimblandy]

Okay, I filed an issue against the tester crate: messense/rustc-test#6

[Munksgaard]

Thank you @jimblandy, that reproduces the issue for me here. I don't have time to look at the issue at the moment, unfortunately, let's hope someone else does.