`bazel_test`

Resolver-generated Bazel test rule materialized as a Once graph target.

Description#

bazel_test is created by the bazel_workspace resolver for every rule whose class ends in _test (for example rust_test, cc_test, py_test). Direct authoring in once.toml is rejected: the resolver must materialize it so the bazel_label and bazel_rule_kind attributes stay tied to real Bazel state.

Both capabilities read the action graph for the label through bazel aquery and run every action themselves from <workspace>/.once/bazel-shadow/<target>/. Bazel handles analysis and external-repository download via bazel fetch; Once handles execution. When the action graph contains a Bazel-internal action (with no argv in aquery output), the capability falls back to bazel test <label> --test_output=errors (or bazel build <label> for the build capability) and records the unsupported mnemonics on the target's provider.

Test rules are not roots of the resolver output: once build or once bazel build //... never picks them up implicitly. Invoke the test explicitly with once bazel test or once test <once-id>.

Attributes#

Attribute	Type	Required	Default	Description
`bazel_label`	string	yes		Fully qualified Bazel label of the test rule, for example `//src:kura_lib_test`. Set by the resolver.
`bazel_rule_kind`	string	yes		Bazel rule class, for example `rust_test`. Set by the resolver.
`bazel`	string	no	`bazel`	Bazel executable forwarded from the workspace seed.

Providers and capabilities#

The target emits bazel_test and bazel_target and exposes the build and test capabilities. It does not expose run.

Direct use#

bash

once bazel test //src:kura_lib_test

See the bazel_workspace reference for the workspace-level attributes that control graph enumeration.

Sources#

Bazel command-line reference documents bazel test, which this target kind wraps.