Command line

Here is how you can choose which tests run and how they are reported, without editing any file.

Passing arguments

Arguments after -- are passed by dub to the test runner:

dub test -- -s "my.module" -r spec,result

You can also call the built test executable directly, for example ./my-project-test-unittest -t "parses an empty list".

Flags

Flag	Value	What it does
`-t`, `--testName`	filter	Runs only the tests whose name matches the filter
`-s`, `--suiteName`	filter	Runs only the tests whose suite name matches the filter
`-f`, `--filter`	filter	Runs only the tests whose `<suite> <test name>` matches the filter, so one value can select both the suite and the test
`--at`	`file:line`	Runs the test that contains that line, for example `--at source/my/module.d:42`
`-r`, `--reporters`	names	Replaces the reporter list. Separate names with commas: `-r spec,result,xunit`
`-e`, `--executor`	name	Replaces the executor: `default`, `parallel` or `process`

When several filters are given, a test must match all of them.

When the filters match no test, the run fails instead of passing with nothing executed, so a typo can’t turn CI green:

No tests matched the filters: -s "my.module" -t "retruns 404"

A run without filters still passes when the project has no tests.

An unknown reporter or executor name stops the run with an error like There is no `nope` reporter.

Filter values

Value	Matches	Example
plain text	names containing the text	`-t "returns 404"`
`=text`	the exact name only	`-t "=returns 404"`
`/regex/`	names matching the regular expression	`-t "/^returns (404\|403)$/"`

A plain filter can select more than you expect: -t "returns 404" also runs returns 404 when the map is private. Use the = form to run exactly one test.

Selecting a test by line

--at file:line runs the test whose body contains that line, which is handy when a stack trace or your editor points at a line. The file only needs to end with the given path, so --at my/module.d:42 is enough. Trial knows only the line where each test starts, so it picks the test in that file that starts closest before the line.

Settings precedence

The settings are resolved in this order, the last one wins:

the defaults
the trial.json file from the current directory, when it exists
the command line flags

The trial.json file can set reporters, executor, maxThreads, artifactsLocation, warningTestDuration and dangerTestDuration. Missing keys keep their defaults and unknown keys are ignored.

When there is no reporters value, these reporters are used: spec, result, stats, html, allure and xunit.

Agent mode

When the tests are run by an AI coding agent, the spec and result console reporters are replaced by the agent reporter, which prints only the failed tests and one summary line. The file reporters, like xunit and html, keep running.

An agent is detected when one of these environment variables is set: AI_AGENT, CLAUDECODE, CURSOR_AGENT, CODEX_SANDBOX, GEMINI_CLI, OPENCODE or CLINE_ACTIVE.

TRIAL_AGENT=1 turns agent mode on and TRIAL_AGENT=0 turns it off, whatever environment variables are set. Passing -r always wins over agent mode, since you chose the reporters yourself.