bun test: nest results under describe scopes instead of repeating the scope path

[jakeboone02]

Branch: nest-describe-blocks-streaming · Commit: 70323bea
Closes #3342

This is one of two alternative implementations of #3342. The sibling PR, #35794,
solves the same problem by buffering each file's results. Pick one.
See Streaming vs. buffered below.

What does this PR do?

bun test repeats the full describe path on every single result line. In a suite with
any real nesting that means the same 40 characters of scope are re-printed hundreds of
times, and the actual test names are pushed off to the right.

This PR prints each describe scope once, as its own dim header line, and indents
its tests beneath it — two spaces per level.

Before

demo.test.ts:
(pass) math > arithmetic > add
(fail) math > arithmetic > subtract [1.84ms]
(pass) math > is a module
(skip) all skipped > x
(todo) all todo > y
(pass) top-level test

After

demo.test.ts:
math
  arithmetic
    (pass) add
    (fail) subtract [0.09ms]
  (pass) is a module
all skipped
  (skip) x
all todo
  (todo) y
(pass) top-level test

Headers are emitted lazily: a scope line is written the first time a result inside it is
about to print, and re-announced if results from a different scope interleave in between.
Nothing is buffered — every line still goes out the instant the test finishes.

Design notes & edge cases

• Scope path is tracked, not the tree. The reporter keeps a single
  printed_scope_path: Vec<u8> (\x1f-separated). Before printing a result it diffs the
  test's scope path against the last-printed one, emits headers for the segments that
  differ, and stores the new path. This is O(depth) per test with one allocation reused
  for the whole run.
• Unnamed describes are transparent. describe(() => {...}) contributes no header
  and no indent level, so the anonymous scope doesn't shift its children right.
• Depth is capped at 64 scopes (BoundedArray), matching the existing limit in
  collect_scopes. Beyond that, indentation stops growing rather than running away.
• Recap buffers keep the flat format. The skip/todo/fail lists printed after the run
  still read math > arithmetic > subtract — they're a flat index, and nesting them out of
  context would be worse. This means each line is formatted twice, so the side effects of
  formatting (assertion-count errors, GitHub Actions timeout annotations) were extracted
  into print_status_side_effects so they fire exactly once.
• --reporter=dots is untouched. It uses Layout::Flat and never nests.
• --only-failures nests the same way; scopes with no failures never print a header.
• JUnit / LCOV output is unchanged — this is purely the terminal reporter.

How did you verify your code works?

Test runs

• bun bd test test/cli/test/ — 220 pass, 6 todo, 0 fail
• bun bd test test/js/bun/test/ — no regressions vs. the parent commit
  (df6c7eed6b has 6 pre-existing failures, 4 of them missing-node_modules load
  errors; this branch has the same set)
• All test/regression/issue/*.test.ts files that spawn bun test — green
  (5 inline snapshots regenerated for the new shape)

New tests in test/cli/test/bun-test.test.ts cover indentation depth, unnamed-describe
transparency, re-announcing a scope after interleaving, and --only-failures. They fail
under USE_SYSTEM_BUN=1 and pass under bun bd test.

Manually verified: default reporter, FORCE_COLOR=1, NO_COLOR=1, --parallel=2,
--reporter=dots, --only-failures, --rerun-each=2, test.concurrent,
--bail, process.exit() mid-file, GitHub Actions ::group:: output, and JUnit output.

Performance

Release builds, macOS arm64, best-of-5, stdout discarded, stderr piped.
Workloads: wide = 200 files x 200 tests x 3 levels (40k tests); deep = 20 files x
5,000 tests x 8 levels (100k tests).

workload	mode	build	wall (ms)	peak RSS (MB)	first output (ms)
wide	default	baseline	89	26.2	6
wide	default	streaming	92	26.5	6
deep	default	baseline	143	59.4	6
deep	default	streaming	159	60.4	7
wide	parallel	baseline	152	53.5	11
wide	parallel	streaming	154	53.3	11
deep	dots	baseline	136	57.7	13
deep	dots	streaming	131	59.1	13

Full benchmark matrix (all 3 builds x 2 workloads x 4 modes)

workload	mode	build	wall (ms)	peak RSS (MB)	first output (ms)
wide	default	baseline	89	26.2	6
wide	default	streaming	92	26.5	6
wide	default	buffered	98	26.4	7
wide	parallel	baseline	152	53.5	11
wide	parallel	streaming	154	53.3	11
wide	parallel	buffered	181	53.2	12
wide	dots	baseline	86	25.8	7
wide	dots	streaming	87	26.3	8
wide	dots	buffered	86	26.0	8
wide	only-failures	baseline	71	26.0	70
wide	only-failures	streaming	70	25.9	69
wide	only-failures	buffered	70	26.1	69
deep	default	baseline	143	59.4	6
deep	default	streaming	159	60.4	7
deep	default	buffered	163	58.2	18
deep	parallel	baseline	95	35.3	20
deep	parallel	streaming	129	35.8	20
deep	parallel	buffered	113	35.4	30
deep	dots	baseline	136	57.7	13
deep	dots	streaming	131	59.1	13
deep	dots	buffered	134	59.5	14
deep	only-failures	baseline	92	52.4	91
deep	only-failures	streaming	92	51.8	91
deep	only-failures	buffered	91	51.8	90

baseline = df6c7eed6b (main), streaming = this branch, buffered = the sibling PR.

Summary: memory is unchanged (single reused Vec<u8> for the scope path, ~1 KB).
Wall time is +3% on wide and +11% on deep for the default reporter; --dots and
--only-failures are within noise. Time-to-first-output is identical to today.

Streaming vs. buffered

	this PR (streaming)	sibling PR (buffered)
Groups split when results interleave	yes, scope is re-announced	never
test.concurrent / --parallel grouping	best-effort	guaranteed contiguous
Aggregate status on file/describe lines (#12378)	no	yes
Time to first output	unchanged	+12 ms on a 5,000-test file
Peak RSS	unchanged	unchanged (bounded ~190 KB/file)
Wall time vs. main (default reporter)	+3% / +11%	+10% / +14%
Failure diagnostics position	inline, under their test	above the file's block
Lines changed in src/	~370	~770

Streaming is the conservative option: it changes only how the scope prefix is rendered and
keeps every other timing property of the reporter identical. The cost is that under
test.concurrent or --parallel a describe can appear more than once, because the
reporter can't know what's still coming.

Known limitations

• A scope header is re-printed whenever results from another scope interleave. With
  test.concurrent this is common; with sequential tests it never happens.
• The end-of-run recap lists keep the flat a > b > name form (intentional, see design
  notes).

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: 915ea137-2649-43b2-9173-aca65905bcbd

📥 Commits

Reviewing files that changed from the base of the PR and between 70323be and b839efa.

📒 Files selected for processing (1)

• test/cli/test/bun-test.test.ts

---

Walkthrough

Bun’s test console reporter now renders nested describe scopes as hierarchical headers with indented test results. Scope paths are serialized through parallel worker IPC, while recap and flat modes retain path-oriented formatting. Documentation and snapshots reflect the new output.

Changes

Nested test reporting

Layer / File(s)	Summary
Scope-aware reporter formatting src/runtime/cli/test_command.rs, src/runtime/test_runner/bun_test.rs	Describe scopes are serialized and emitted once per scope, with nested and flat layouts selected by reporter mode.
Parallel scope-path transport src/runtime/cli/test/parallel/*	Workers send scope paths with test results, and the coordinator resets and emits scope headers per file.
Reporter behavior validation test/cli/test/*, test/js/bun/test/*, test/regression/issue/*	Snapshots and CLI tests cover nested indentation, filtering, recap output, failures, and parallel execution.
Reporter output documentation docs/guides/test/coverage*.mdx, docs/test/reporters.mdx	Examples show grouped coverage and hierarchical console output.

Possibly related PRs

• oven-sh/bun#35794: Both PRs implement nested describe scope output and related scope-path transport and rendering changes.

Suggested reviewers: cirospaciari, jarred-sumner

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly describes the main change: nesting bun test results under describe scopes instead of repeating scope paths.
Description check	✅ Passed	The description matches the required template and includes both what changed and how it was verified.
Linked Issues check	✅ Passed	The changes satisfy #3342 by rendering describe labels once with nested test output and preserving the requested terminal-report behavior.
Out of Scope Changes check	✅ Passed	The diff stays focused on reporter behavior, matching docs, and test snapshots, with no obvious unrelated code changes.

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@test/cli/test/bun-test.test.ts`:
- Around line 1526-1537: Update the run helper to drain proc.stdout alongside
proc.stderr, such as by reading both streams in the existing Promise.all, while
preserving the current stderr normalization and exitCode return behavior.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: 51767a38-8693-4d0e-b345-c55fd0232c3f

📥 Commits

Reviewing files that changed from the base of the PR and between 04bb5c4 and 70323be.

📒 Files selected for processing (17)

• docs/guides/test/coverage-threshold.mdx
• docs/guides/test/coverage.mdx
• docs/test/reporters.mdx
• src/runtime/cli/test/parallel/Coordinator.rs
• src/runtime/cli/test/parallel/Frame.rs
• src/runtime/cli/test/parallel/runner.rs
• src/runtime/cli/test_command.rs
• src/runtime/test_runner/bun_test.rs
• test/cli/test/bun-test.test.ts
• test/cli/test/test-filter-lifecycle-snapshot.test.ts
• test/js/bun/test/bun_test.test.ts
• test/js/bun/test/concurrent.test.ts
• test/js/bun/test/describe.test.ts
• test/js/bun/test/nested-describes.test.ts
• test/regression/issue/12782.test.ts
• test/regression/issue/19875.test.ts
• test/regression/issue/20092.test.ts