I don’t want to bikeshed this too hard, but I’m a little worried that the choice of test-runner is a crate-attribute, since that implies different crates can and will use different test-runners. I expect there’ll wind up being half a dozen or so competing test-runners, each with different feature sets, and there’ll be one particular one I like best - perhaps it summarizes results in a way I find easy to comprehend at a glance, maybe it produces machine-readable output in a format I can use with other tools, maybe it uses debug info to provide richer diagnostics than the standard test-runner interface would allow.
At any rate, once I have a favourite test-runner, I’d want to use that with every crate I work on, regardless of what test-runner the crate’s author prefers. If the choice of test-runner is set by a crate attribute, I’d have to tinker with the code before running any tests (running the risk of accidentally breaking the code), and it would be difficut to automate (without a full Rust parser and AST to manipulate).
If possible, I think a better solution would be some kind of
Cargo.toml option, since it’s safer and easier to edit. Best of all would be an option in
~/.cargo/config, or an environment variable, so I can set it once and have everything always use the test-runner I chose.