Rust "cargo nextest" Failures - Leaked/Timed-Out Tests in CI

A cargo nextest run had at least one test that failed, leaked a child process, or exceeded its slow/leak timeout and was aborted. nextest runs each test in its own process, so its failure modes (LEAK, slow, timeout) differ from the built-in harness.

What this error means

nextest prints a summary with FAIL, LEAK, or TIMEOUT/ABORT lines and exits non-zero. A LEAK means the test passed but left a child process or open handle; a slow/timeout abort means it ran past the configured limit.

nextest output

        FAIL [   0.012s] app::math tests::divides
        LEAK [   5.030s] app::net tests::spawns_server
------------
     Summary [   5.1s] 24 tests run: 22 passed, 1 failed, 1 leaked
error: test run failed

Common causes

A test assertion failed

Like the built-in harness, a FAIL line is a genuine failing test - an assertion didn’t hold or the test panicked. It reproduces on the same code.

A leaked process or a slow/timeout abort

nextest flags a LEAK when a test leaves a child process or handle open after returning, and aborts tests that exceed the slow-timeout/leak-timeout. A test spawning a server without cleaning it up is the classic leak.

How to fix it

Read the summary and reproduce the failure

Identify each FAIL/LEAK/TIMEOUT line and the test it names.
Re-run just that test with a filter expression, e.g. cargo nextest run -E "test(divides)".
For a LEAK, ensure the test joins/kills any process or task it spawns before returning.

Tune timeouts deliberately, not to mask bugs

Raise the slow/leak timeout only for genuinely long tests; don’t use it to hide a hang or a leak.

.config/nextest.toml

# .config/nextest.toml
[profile.ci]
slow-timeout = { period = "60s", terminate-after = 2 }
leak-timeout = "1s"

How to prevent it

Clean up spawned processes/tasks in tests so they don’t leak.
Keep per-test timeouts tight enough to catch hangs but generous for real long tests.
Run cargo nextest run locally before pushing.