Migrate to rstest-bdd v0.6.0

[leynos]

Summary

Migrate Netsuke's behavioural test suite from rstest-bdd 0.5.0 to the published 0.6.0 release, and guard the correctness fix the upgrade brings.

Cargo.toml changes exactly two version strings, both caret requirements, with strict-compile-time-validation retained. Cargo.lock resolves the whole rstest-bdd family — rstest-bdd, -macros, -patterns, -policy and -harness — to 0.6.0. No harness crate is added to any manifest: rstest-bdd-harness is transitive through rstest-bdd-macros, and the opt-in adapters rstest-bdd-harness-tokio and rstest-bdd-harness-gpui are absent from the lockfile because nothing requests them. The async scenarios are synchronous, so no adapter is needed.

No MSRV bump was needed. The graph already required Rust 1.89 via pre-existing ortho_config 0.9.0 and serde-saphyr 1.2.0, above the 1.88 that gherkin 0.16 imposes. rust-toolchain.toml and its Polonius nightly pin are untouched.

Lockfile scope

The lockfile delta is 22 packages, and every one lies inside the rstest-bdd family's transitive closure — 157 packages at 0.5.0, 176 at 0.6.0. No package outside that closure changed version, and none was added or removed.

The churn is real but entirely family-owned. rstest-bdd-macros 0.6.0 swapped proc-macro-error 1.0.4 for proc-macro-error3 3.1.1, taking convert_case, heck and a windows-sys family with it; it also pulled in cargo_metadata, link-section, linktime-proc-macro, derive_more and syn 3. proc-macro-error and proc-macro-error-attr are removals, so they are absent from the post-bump closure while sitting inside the pre-bump one.

The behavioural change, and its guard

0.6.0 fixes a false green: a step whose return type is a type alias of Result<T, E> previously had its Err boxed as an unused payload, so the scenario passed despite a failed assertion. This repository's fallible steps are declared -> Result<()> where Result is anyhow::Result — exactly such an alias — so the fix lands squarely on the suite.

None of the scenarios turned red, so no latent swallowed error was exposed. A green suite is equally consistent with the propagation working and with it being silently reverted, and no existing scenario has a step that returns Err — so the fix arrived unguarded.

tests/step_error_propagation_tests.rs closes that gap. It is deliberately driven to failure and passes only when a step's Err reaches the generated step loop. Validated in both directions: green as committed, and red on its trailing assertion (trailing step ran: the earlier error was not propagated) when the failing step is neutralised to Ok(()), which reproduces the pre-0.6.0 behaviour. It cannot pass by accident.

Its feature file lives in tests/features_step_results/, not tests/features/, because scenarios! sweeps the latter directory and would otherwise collect a deliberately-failing scenario as an ordinary one expected to pass.

Breaking changes: applied or inapplicable

"Inapplicable" is evidenced by the absence of a call site, not assumed. The consumed surface is rstest_bdd::Slot plus the given, when, then and scenarios macros — nothing else, and Slot is unchanged in 0.6.0.

Breaking change	Disposition
Underscore-prefixed implicit fixture naming	inapplicable — fixtures use unprefixed world
runtime = "tokio-current-thread"	inapplicable — no scenario selector anywhere; scenarios are synchronous
HarnessAdapter::run returns HarnessResult	inapplicable — no custom harness
Step aliases of Result<T, E> propagate Err	applied, now guarded
insert_value returns InsertOutcome	inapplicable — zero direct callers
record_bypassed_steps takes BypassedScenario	inapplicable — zero callers
RustStepIndexResult from indexing entry points	inapplicable — zero references
index_feature_file removed	inapplicable — zero callers
publish_rust_diagnostics removed	inapplicable — zero callers
find_feature_files returns Result	inapplicable — zero callers
Feature paths manifest-relative in reports	inapplicable — no metadata/JSON/JUnit consumer
MSRV raised to 1.88	not binding — graph already required 1.89

Documentation

docs/rstest-bdd-users-guide.md and docs/rstest-bdd-v0-6-0-migration-guide.md were imported byte-for-byte from tag v0.6.0 (commit 72fb22635670e456545ca368805ba4c1c9d7bd69), with provenance recorded in the ExecPlan and docs/contents.md.

This PR additionally corrects the ## rstest-bdd v0.6.0 usage section of docs/developers-guide.md, which had recommended runtime = "tokio-current-thread". 0.6.0 deprecates that syntax, and it warns rather than compiles silently because this repository builds on a nightly channel — the rstest_bdd_nightly condition the macros crate's build script selects — so a -D warnings build would escalate the diagnostic to an error.

The lint failures this migration caused, and their resolution

rstest-bdd 0.6.0 declares tracing with features = ["log"] non-optionally. Because rstest-bdd is a dev-dependency while tracing is a normal dependency, Cargo unifies tracing/log onto the single node the library compiles against whenever tests are built. That feature expands every tracing::* macro into extra branches, and Clippy's cognitive_complexity is measured after macro expansion — so a macro written inline is charged to its enclosing function.

An earlier revision of this description attributed the resulting make lint failure to a pre-existing condition on main and dismissed a dev-dependency bump as a possible cause. That was wrong. The cause was proved three ways: cargo tree -e features -i tracing names rstest-bdd v0.6.0 as the sole enabler of tracing's default feature; a counter-control reverting only rstest-bdd to 0.5.0 re-runs the identical probe from exit 101 with two complexity errors to exit 0 with none; and the flagged functions each contained exactly one inline tracing macro.

Measured cost model: +7 per inline tracing macro under log, +1 without, and nothing extra for a function that merely calls an emitter holding the macro. The remedy is therefore to hoist each charged macro into a dedicated emitter function.

The scope is larger than the first count suggested. Clippy's reported number is a lower bound: it aborts the build, so clearing one diagnostic unmasks whatever the abort was hiding, and fixing the first batch kept revealing more. In the end the hoist is mechanical but wide — 76 new functions across 37 non-test source files, spanning src/stdlib, src/runner, src/manifest, src/cli and the test support crates. It is behaviour-preserving by construction: every field name, message string and emitted ordering moves verbatim into its emitter, and nothing else in those functions changes. As one illustration, record_timing_summary_sink_write keeps its counter! and histogram! calls in place and calls a new debug_timing_summary_sink_failed_from_fields() instead of writing the macro inline.

One of those sites deserves naming, because it arrived with the rebase rather than with the migration. span_fields_are_captured_by_name_and_recording_point in src/test_tracing_capture.rs is a test main added while this branch was in review; the branch had never before compiled it under unified tracing features, so it had never charged the three inline trace_span! macros. Hoisting preserves the span names, field::Empty declarations, recording order and guard lifetime, so the test's exact captured-string assertions still pin the behaviour.

Hoisting the two Kani sites then pushed tests/kani_mutation_evidence_tests/compile_guard.rs from 393 to 413 lines, past Whitaker's 400-line module_max_lines cap — a second failure invisible until now for the same cascade reason, since make lint aborts at the first failing stage and Clippy failed ahead of Whitaker. The patch-application machinery moves to a sibling apply_patch module along a real seam: one module runs git apply and decides when a reverse is owed, the other decides which patches exist and what compiling them proves. The guard is left at 294 lines.

Two further defects, both found by CI rather than locally

A tool-version skew failed the required build-test check. CI pins MDTABLEFIX_VERSION: '0.6.0'; the local install was 0.6.1, and the two have different canonical wrap points. make fmt under 0.6.1 wrote this plan into a form that 0.6.0 rejects, so CI's Format step — run: make check-fmt — failed with docs/execplans/adopt-rstest-bdd-v0-6-0.md +11 -12 and skipped Lint, Typecheck, Doc coverage, Spelling, Mermaid, Workflow contracts and Test behind it. Every locally run gate agreed with 0.6.1, which is why the divergence was invisible here. The fix is the fixpoint rather than either tool's output: converging under 0.6.0 and then checking with 0.6.1, and the reverse, both report 174 files left unchanged, and comparing the result to what 0.6.1 wrote with all whitespace removed shows the two are identical — only line breaks moved. The local binary is still 0.6.1, not the pinned 0.6.0 the earlier revision of this description claimed, so the lanes can still diverge and every Markdown edit has to be checked against the pinned binary explicitly rather than assumed safe.

The hoist cost child_exit.rs a CodeScene point. CodeScene's String Heavy Function Arguments rule is a file-level ratio, not a per-function verdict. Hoisting four tracing macros out of terminate_child and finalize_streaming took src/runner/process/child_exit.rs from 2 of 5 functions taking a string argument to 5 of 9, and the file fell from 10.00 to 9.68. Measured locally, cs check <merge-base>:./src/runner/process/child_exit.rs reports 10.00 against 9.68 on this branch's parent, which established the cause as this migration's own hoist. The four emitters move unchanged into src/runner/process/child_exit/emitters.rs, a same-stem directory beside their host, declared with an explicit #[path] — the shape stdlib/register already uses for query_helpers.rs. Both files score 10.00, and the CodeScene check passes.

Validation

Full gate set on the committed, rebased revision, with make lint reaching every stage — lint-clippy, lint-whitaker (both crates), lint-workflow-scripts, all five stages of lint-python, and github-actions-lint. The failing revision never got past lint-clippy, so lint-whitaker, lint-python and github-actions-lint had never executed on this branch at all.

make test passes both passes: 3919 nextest tests with 0 failures, 0 leaky and 6 skipped, plus 129 doctests across two separate targets. make check-fmt, make typecheck, make doc-coverage (98.88% against an 80% bar), make markdownlint and make nixie pass.

make markdownlint passing is itself worth noting: on the earliest revisions this target died at its preceding spelling stage, so markdownlint-cli2 had never actually executed.

The swept BDD inventory moves by zero: measured as sets at three revisions, the baseline (ebcedaef) had 254 scenario names, main (84447f0e) has 269, and this branch has 269. main's 15 additions arrived from work that landed while this branch was in review; the branch's own delta is nil, and all 254 baseline names survive. Two independent measurements agree on 269 — a grep of the swept feature directories, and the generated features_scenarios:: / features_unix_scenarios:: test names in the make test log.

make test-workflow-contracts also passes (1107 passed, 3 skipped). It is deliberately not a prerequisite of the make gates above, so it is run separately after a rebase.

The final sweep is the post-rebase one, and it is a different shape from the earlier runs. The branch was rebased onto fce1a746, whose rust_module_layout_test.py contract rejects sibling Rust modules sharing a first _-delimited prefix — exactly the shape all three of this branch's hoisted emitter modules had. Each moved into a same-stem directory beside its host, declared with an explicit #[path]; that removes the prefix pair outright, and since #[path] changes which file backs a module but never the module's path in the tree, every super:: and pub(super) reference resolves unchanged. The tools/kani/proof-scope.toml entries were recomputed for the two emitters the harnesses reach.

Nine targets then ran on the frozen revision 021e0524, sequentially, with the tree clean before and after: check-fmt (1s), spelling (7s), markdownlint (14s), typecheck (10s), lint (39s), test (238s), nixie (1s), doc-coverage (9s) and test-workflow-contracts (76s). All nine PIPESTATUS[0] sidecars contain 0, and each log brackets itself with HEAD_AT_START and HEAD_AT_END, both reading 021e0524, over an empty post-gate tree-status block — so the revision is pinned by the log rather than inferred from its name.

Two counts moved against the earlier sweep and neither is a regression: make test rose from 3912 to 3919 and the workflow contracts from 1082 to 1107, both because fce1a746 contributed 7 tests. make lint reached every stage with the word warning appearing exactly four times, all four being the literal -D warnings flag echoed in a command line rather than a finding. make test's one slow test, packaged_manifest_retains_build_script_sources, PASSED at 144.107s after crossing nextest's 60s and 120s SLOW thresholds; it is recorded as a pass, not a timeout.

Recording that sweep invalidated it, because this plan is itself read by a gate: tests/execplan_status_contract_tests.rs parses every plan header. The check that matters is therefore which gates can see docs/execplans/ at all — only check-fmt, spelling, markdownlint and nixie — and those four were re-run on the amended bytes. That re-run earned its keep by catching a real defect the first pass could not: prose added while recording the sweep contained a code span with an interior space, and markdownlint-cli2 reported MD038/no-space-in-code against it. Both mdtablefix 0.6.0 and 0.6.1 accepted the same text, which is the point — the fixpoint check covers the tool with the skew, not the linter. After the fix, all four gates are green on the committed revision b66ce687, with markdownlint-cli2 genuinely executing over 175 files.

make doc-coverage reads 98.88% (5044/5101) against the 80% bar; the numerator gains one item from src/runner/process/child_exit/emitters.rs.

References

• Upstream source at tag v0.6.0: leynos/rstest-bdd commit 72fb2263
• Migration guide: docs/v0-6-0-migration-guide.md
• Users' guide: docs/users-guide.md
• ExecPlan: docs/execplans/adopt-rstest-bdd-v0-6-0.md
• Working session: https://lody.ai/leynos/sessions/7bb1d019-44e1-4cc0-b860-e7ac1b312667

🤖 Generated with Claude Code

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

• Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

---

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

[leynos]

@coderabbitai Please investigate the cause of the following issue using codegraph exploration and research, identify a fix and provide an AI coding agent prompt for the fix:

error: the function has a cognitive complexity of (10/9)
Error:   --> src/cli/discovery_trace.rs:97:8
   |
97 |     fn emit(&self) {
   |        ^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity
   = note: requested on the command line with `-D clippy::cognitive-complexity`

error: the function has a cognitive complexity of (13/9)
Error:   --> src/cli/discovery_telemetry.rs:60:8
   |
60 | pub fn record_discovery_outcome<C: MonotonicClock>(
   |        ^^^^^^^^^^^^^^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (16/9)
Error:    --> src/cli/merge_observability.rs:127:4
    |
127 | fn record_default_event(event: &MergeEvent) {
    |    ^^^^^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (23/9)
Error:    --> src/cli/merge_observability.rs:140:4
    |
140 | fn record_file_event(event: &MergeEvent) {
    |    ^^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (16/9)
Error:    --> src/cli/merge_observability.rs:164:4
    |
164 | fn record_environment_event(event: &MergeEvent) {
    |    ^^^^^^^^^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (23/9)
Error:    --> src/cli/merge_observability.rs:184:4
    |
184 | fn record_cli_event(event: &MergeEvent) {
    |    ^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (22/9)
Error:    --> src/cli_localization.rs:168:8
    |
168 | pub fn build_localizer(preferred_locale: Option<&str>) -> Box<dyn Localizer> {
    |        ^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (10/9)
Error:    --> src/ir/from_manifest.rs:118:8
    |
118 |     fn process_targets(
    |        ^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (17/9)
Error:    --> src/manifest/glob/diagnostics.rs:180:4
    |
180 | fn record_template_expansion(
    |    ^^^^^^^^^^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (13/9)
Error:   --> src/manifest/jinja_macros/telemetry.rs:89:15
   |
89 | pub(super) fn instrument_macro_invocation<T>(
   |               ^^^^^^^^^^^^^^^^^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (14/9)
Error:    --> src/manifest/jinja_macros/telemetry.rs:116:15
    |
116 | pub(super) fn instrument_template_render<T>(
    |               ^^^^^^^^^^^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (16/9)
Error:   --> src/manifest/loading.rs:39:15
   |
39 | pub(super) fn trace_expansion_report(report: &ExpansionReport) {
   |               ^^^^^^^^^^^^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (16/9)
Error:    --> src/manifest/env_reader.rs:147:15
    |
147 | pub(super) fn env_var_with_default(
    |               ^^^^^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (10/9)
Error:   --> src/manifest/workspace.rs:62:15
   |
62 | pub(super) fn open_manifest_workspace(
   |               ^^^^^^^^^^^^^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

    Checking assert_cmd v2.2.2
error: the function has a cognitive complexity of (12/9)
Error:   --> src/runner/manifest_structure_telemetry.rs:46:8
   |
46 | pub fn record_manifest_structure(manifest: &NetsukeManifest) {
   |        ^^^^^^^^^^^^^^^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (10/9)
Error:   --> src/runner/graph.rs:37:15
   |
37 | pub(super) fn handle_graph(
   |               ^^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (13/9)
Error:   --> src/runner/help_telemetry.rs:19:15
   |
19 | pub(super) fn instrument_help_targets<T>(query: impl FnOnce() -> Result<T>) -> Result<T> {
   |               ^^^^^^^^^^^^^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (17/9)
Error:   --> src/runner/process/child_exit.rs:63:15
   |
63 | pub(super) fn terminate_child(child: &mut Child, context: &str) {
   |               ^^^^^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (15/9)
Error:   --> src/runner/process/command_logging.rs:97:15
   |
97 | pub(super) fn log_command_execution(
   |               ^^^^^^^^^^^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (24/9)
Error:   --> src/runner/process/ninja_program.rs:34:9
   |
34 |         |value| {
   |         ^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (17/9)
Error:    --> src/runner/process/streaming.rs:200:4
    |
200 | fn copy_with_stats<R, W>(reader: &mut R, writer: &mut W, stream_name: &'static str) -> ForwardStats
    |    ^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (12/9)
Error:   --> src/runner/recipe_shell_telemetry.rs:75:15
   |
75 | pub(super) fn instrument_recipe_shell_resolution(
   |               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (13/9)
Error:    --> src/runner/recipe_shell_telemetry.rs:115:15
    |
115 | pub(super) fn instrument_bash_preflight<T>(
    |               ^^^^^^^^^^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (13/9)
Error:    --> src/runner/recipe_shell_telemetry.rs:151:15
    |
151 | pub(super) fn instrument_legacy_recipe_operation<T>(
    |               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (10/9)
Error:    --> src/runner/mod.rs:275:4
    |
275 | fn execute_ninja_tool(
    |    ^^^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (10/9)
Error:    --> src/status_timing.rs:216:4
    |
216 | fn record_timing_summary_sink_write(elapsed: Duration, succeeded: bool) {
    |    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (13/9)
Error:    --> src/stdlib/command/execution.rs:192:4
    |
192 | fn run_child(
    |    ^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (12/9)
Error:    --> src/stdlib/command/execution.rs:260:4
    |
260 | fn run_child_inner(
    |    ^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (10/9)
Error:    --> src/stdlib/command/execution.rs:330:15
    |
330 | pub(super) fn wait_for_exit(
    |               ^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (17/9)
Error:    --> src/stdlib/command/pipes.rs:212:4
    |
212 | fn join_pipe_for_cleanup(
    |    ^^^^^^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

    Checking rust-embed v8.12.0
error: the function has a cognitive complexity of (17/9)
Error:    --> src/stdlib/network/cache.rs:112:15
    |
112 | pub(super) fn read_cached(dir: &Dir, name: &str, limit: u64) -> Result<Option<Vec<u8>>, Error> {
    |               ^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (10/9)
Error:    --> src/stdlib/network/redirect.rs:204:4
    |
204 | fn remaining_budget(deadline: Instant, url: &Url) -> Result<Duration, Error> {
    |    ^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (31/9)
Error:   --> src/stdlib/network/mod.rs:97:4
   |
97 | fn fetch_inner(
   |    ^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (10/9)
Error:    --> src/stdlib/path/path_utils.rs:181:4
    |
181 | fn resolve_home<F>(home_directory: &HomeDirectory, read_env: F) -> Result<String, Error>
    |    ^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (17/9)
Error:   --> src/stdlib/path/read_telemetry.rs:81:15
   |
81 | pub(super) fn record_file_read<T>(
   |               ^^^^^^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (15/9)
Error:    --> src/stdlib/register.rs:112:8
    |
112 | pub fn register_with_config(
    |        ^^^^^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (10/9)
Error:   --> src/stdlib/which/env.rs:51:4
   |
51 | fn capture_workspace_switch(env: &impl Env) -> WorkspaceSwitch {
   |    ^^^^^^^^^^^^^^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (15/9)
Error:   --> src/stdlib/which/lookup/workspace/mod.rs:85:15
   |
85 | pub(super) fn search_workspace(
   |               ^^^^^^^^^^^^^^^^
   |
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: the function has a cognitive complexity of (10/9)
Error:    --> src/stdlib/which/telemetry.rs:278:15
    |
278 | pub(super) fn record_resolution_error(
    |               ^^^^^^^^^^^^^^^^^^^^^^^
    |
    = help: you could split it up into multiple smaller functions
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

error: could not compile `netsuke-build` (lib) due to 39 previous errors
warning: build failed, waiting for other jobs to finish...
make: *** [Makefile:291: lint-clippy] Error 101
Error: Process completed with exit code 2.~~~

Seek a systemic fix rather than tactical. Ask yourself, can this happen again or happen elsewhere? If so, think about a long term fix of the underlying issue.

[coderabbitai]

Tip

For best results, initiate chat on the files or code changes.

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
git rev-parse HEAD
git status --short
cat Cargo.toml
cat rust-toolchain.toml
fd -i 'clippy|makefile|adopt-rstest-bdd-v0-6-0|step_error_propagation|discovery_trace|merge_observability'
ast-grep outline src/cli/discovery_trace.rs --items all
ast-grep outline src/cli/merge_observability.rs --items all
sed -n '65,135p' src/cli/discovery_trace.rs
sed -n '100,230p' src/cli/merge_observability.rs
sed -n '250,315p' Makefile
rg -n -C3 'cognitive|tracing|positive control|counter|BLOCKED|resolver' docs/execplans/adopt-rstest-bdd-v0-6-0.md
rg -n -C3 'cognitive-complexity|cognitive_complexity|cargo clippy|lint-clippy' . --glob '!Cargo.lock' --glob '!docs/rstest-bdd*'

Length of output: 84511

---

🤖 web_search executed:

Research Rust Clippy cognitive_complexity lint counting tracing macro expansion with tracing feature log. Check Clippy source whether external macro expansions are excluded, Cargo resolver 2 or 3 dev dependency feature unification for --all-targets, and whether tracing log can be disabled by downstream default-features=false. Identify systemic fixes without refactoring logging call sites.

Length of output: 4867

---

🏁 Script executed:

#!/bin/bash
sed -n '624,655p' docs/execplans/adopt-rstest-bdd-v0-6-0.md
rg -n -A24 '^name = "(rstest-bdd|tracing)"' Cargo.lock
cat clippy.toml
ast-grep outline tests/step_error_propagation_tests.rs --items all
cat tests/step_error_propagation_tests.rs
rg -n -C2 'CLIPPY_FLAGS|TEST_FLAGS' Makefile
ast-grep outline src/cli/discovery_telemetry.rs --items all
sed -n '45,125p' src/cli/discovery_telemetry.rs
rg -n -C2 'check-macro-expansion|macro expansion|cognitive' docs/whitaker-users-guide.md dylint.toml

Length of output: 13592

---

🏁 Script executed:

#!/bin/bash
python - <<'PY'
import urllib.request, tarfile, io, hashlib
url = "https://static.crates.io/crates/rstest-bdd/rstest-bdd-0.6.0.crate"
data = urllib.request.urlopen(url, timeout=30).read()
print("Published archive:", url, "sha256:", hashlib.sha256(data).hexdigest())
with tarfile.open(fileobj=io.BytesIO(data), mode="r:gz") as archive:
    for item in archive.getmembers():
        if item.name.endswith(("Cargo.toml.orig", "/Cargo.toml")):
            text = archive.extractfile(item).read().decode()
            print("\nMANIFEST:", item.name)
            print(text)
        elif item.name.endswith(".rs"):
            text = archive.extractfile(item).read().decode()
            lines = text.splitlines()
            indices = [i for i, line in enumerate(lines) if "tracing::" in line or "log::" in line or "LogTracer" in line]
            if indices:
                print("\nTRACING CALL SITES:", item.name)
                shown=set()
                for i in indices:
                    for j in range(max(0,i-3), min(len(lines),i+5)):
                        if j not in shown:
                            print(f"{j+1}: {lines[j]}")
                            shown.add(j)
url="https://raw.githubusercontent.com/rust-lang/rust-clippy/master/clippy_lints/src/cognitive_complexity.rs"
print("\nCURRENT CLIPPY SOURCE (not the pinned nightly):", url)
print(urllib.request.urlopen(url, timeout=30).read().decode())
PY

Length of output: 4374

---

🤖 web_search executed:

Inspect the published rstest-bdd 0.6.0 crate at https://docs.rs/crate/rstest-bdd/0.6.0/source/Cargo.toml and Cargo.toml.orig. Verify its tracing dependency features ["log"], and identify tracing call sites in its src, particularly whether tracing-to-log bridging is required for step or scenario correctness or is observability only. Propose gating tracing/log behind a non-default feature with compatibility considerations. Also inspect https://github.com/rust-lang/rust-clippy/blob/master/clippy_lints/src/cognitive_complexity.rs for check_macro_expansion and the per-expression visitor: can a lint-specific upstream fix exclude expansion-generated branching without skipping ordinary functions containing macros? Do not assume master equals nightly-2026-08-23.

Length of output: 3890

---

@leynos, fix the dependency feature request first, then correct the complexity measurement upstream. Do not refactor the 39 functions to accommodate generated logging branches. Thanks for the guidance to target a systemic fix.

Cause

I inspected Netsuke PR #805 at commit 27a95bbfe73dc81c8ccf633dc80468516eb3839f.

The dependency graph explains the failure:

netsuke-build
├── normal dependency: tracing 0.1.44
└── dev-dependency: rstest-bdd 0.6.0
                    └── tracing, features = ["log"]

Clippy --all-targets
  → builds the dev-dependency graph
  → unifies tracing/log into Netsuke's tracing dependency
  → expands tracing macros with additional log-bridge control flow
  → counts that generated control flow towards cognitive_complexity
  → exceeds the configured threshold of 9

The inspected code supports this diagnosis:

• Cargo.toml declares both dependency paths.
• Makefile sets CLIPPY_FLAGS to --workspace --all-targets --all-features -- -D warnings.
• clippy.toml sets cognitive-complexity-threshold = 9.
• record_default_event contains a small match and two tracing::debug! calls. The reported complexity is 16.
• record_discovery_outcome contains one application-level if/else, plus tracing and metrics macros. The reported complexity is 13.

Public source documentation confirms that published rstest-bdd 0.6.0 explicitly requests tracing/log. Its manifest describes warning delivery and log-listener detection, rather than step-result handling. Published manifest

Cargo explicitly permits this feature unification when dev-dependencies participate in an all-targets build. Changing the resolver does not remove that behaviour. Cargo resolver documentation

The ExecPlan records stronger causal evidence: a minimal reproduction, a repository-scale positive control, and a counter-control that removes only the tracing/log request. Those are recorded results, not builds I reran. This sandbox permits inspection, not repository builds or tests.

Recommended fix

1. Make the tracing-to-log bridge independently selectable

In rstest-bdd, remove features = ["log"] from the unconditional tracing dependency. Expose the bridge through a separate feature.

Use this conceptual manifest change:

[dependencies]
tracing = "0.1"

[features]
log-bridge = ["tracing/log"]

Keep native tracing instrumentation. Do not remove the direct log dependency or listener detection without inspecting their warning-delivery contracts.

Handle compatibility explicitly. Making log-bridge non-default changes diagnostics for consumers that rely on a log logger. Document and test that change. If release policy requires preserving existing defaults, keep the bridge in the upstream default set initially. Then configure Netsuke to disable upstream defaults and explicitly restore every required non-bridge feature.

Publish the upstream correction, then update Netsuke to the actual published version. Do not invent a release number or leave a local registry edit as the fix.

This addresses the feature request at its source. It avoids changes across 31 application files.

2. Correct the lint to prevent recurrence

The dependency change alone cannot prevent another dependency from enabling tracing/log.

The current Clippy cognitive_complexity source skips functions whose body span comes from expansion. However, its expression visitor does not exclude each macro-generated branch inside an ordinary function. Clippy implementation

Submit a lint-specific correction that excludes expansion-generated control flow while retaining application control flow. Preserve traversal into source-written macro arguments. Do not skip an entire function merely because it contains a macro.

Treat this as a separate upstream change. The inspected public source is master, not Netsuke's pinned nightly-2026-08-23.

Reject these alternatives

• Do not raise the threshold. That weakens the application complexity limit.
• Do not add 39 lint exemptions. That distributes a dependency problem across unrelated functions.
• Do not remove --all-targets. That removes test-target lint coverage.
• Do not disable macro checking globally. That changes unrelated lint coverage.
• Do not rely on tracing.default-features = false. Cargo features are additive; another dependency's explicit request remains enabled.

AI coding agent prompt

Fix the dependency-driven Clippy cognitive_complexity regression affecting
leynos/netsuke PR #805. Target the dependency feature request, not the 39
application functions.

Evidence:
- Netsuke was inspected at commit
  27a95bbfe73dc81c8ccf633dc80468516eb3839f.
- Netsuke declares tracing as a normal dependency and rstest-bdd 0.6.0 as
  a dev-dependency.
- Published rstest-bdd 0.6.0 explicitly enables tracing/log.
- Cargo unifies this feature when Clippy builds all targets.
- Netsuke's complexity threshold is 9.
- docs/execplans/adopt-rstest-bdd-v0-6-0.md records positive and
  counter-controls isolating the feature edge.
- The upstream repository revision was not verified. Establish the
  upstream checkout and record its exact commit before changing it.

Requirement 1 — Correct the upstream feature contract.
Inspect rstest-bdd's workspace manifests, tracing calls, direct log calls,
listener detection, and warning fallbacks.

Remove tracing/log from the unconditional tracing dependency.
Keep native tracing instrumentation.
Expose tracing/log through an independently selectable log-bridge feature.

Choose and document a compatible default-feature policy.
If the bridge remains default-enabled for compatibility, let Netsuke disable
upstream defaults and explicitly enable every required non-bridge feature.
Do not disable unrelated runtime behaviour.
Do not remove direct log usage without proving that its contracts remain intact.

Requirement 2 — Add upstream regression coverage.
Test warnings with a tracing subscriber, with a log logger, and with neither.
Test bridge-enabled and bridge-disabled configurations.
Check the resolved feature graph of a downstream consumer with both normal
and dev dependency paths.
Add a downstream Clippy reproduction for --all-targets.
Keep dependency feature checks separate from runtime behaviour checks.

Requirement 3 — Integrate the released correction into Netsuke.
Use an actual published compatible release.
Retain strict-compile-time-validation.
Keep lockfile changes targeted and explain necessary transitive changes.
Preserve tests/step_error_propagation_tests.rs and its isolated feature file.
Run the error-propagation regression and its negative control.

Preserve the complexity threshold, warning-denial policy, workspace scope,
all-targets scope, and all-features scope.
Do not add lint suppressions, global macro exclusions, or production refactors.

Requirement 4 — Address recurrence separately.
Prepare a Clippy cognitive_complexity correction that excludes generated
branching without excluding source-written branching.
Preserve traversal into source-written expressions passed to macros.
Add tests for nested macros, tracing/log on and off, source-written branches,
closures, and functions that mix macros with ordinary control flow.
Compare against the pinned nightly before proposing toolchain adoption.
Do not change Netsuke's toolchain without following its ADR.

Acceptance:
- Record the feature graph for production and all-targets builds.
- Confirm that no dependency requests tracing/log in Netsuke's intended
  bridge-disabled all-targets configuration.
- Run make lint-clippy, then make lint.
- Run make test, make typecheck, make check-fmt, and documentation gates.
- Validate the native Windows lint job.
- Record actual results. Distinguish failures from checks that did not run.
- Update the ExecPlan and migration documentation.
- Stop if upstream availability or release compatibility prevents delivery.

Verification limit: the linked upstream revision was unavailable. The archive download also failed TLS certificate verification. Public documentation supplied the manifest evidence, but it does not verify the requested Git revision. Retry that revision inspection before implementation. No fix has been implemented or build-verified here.

Linked revisions not inspected for this answer: leynos/rstest-bdd (requested revision unavailable). Their changes remain unverified.

---

✏️ Learnings added

Learnt from: leynos
URL: https://github.com/leynos/netsuke/pull/805

Timestamp: 2026-10-02T17:17:12.222Z
Learning: For leynos/netsuke investigations, leynos requests systemic fixes that address the underlying cause and recurrence across other code paths or dependency configurations, rather than tactical changes to individual diagnostic sites.

Note: Learnings are effective only in the context of similar code segments. To apply general coding guidelines, please configure review instructions. You can manage existing learnings in the app.

You are interacting with an AI system.

[codescene-access]

Gates Passed
6 Quality Gates Passed

See analysis details in CodeScene

Absence of Expected Change Pattern

• netsuke/src/stdlib/register.rs is usually changed with: netsuke/src/stdlib/config/mod.rs

Quality Gate Profile: Pay Down Tech Debt
Install CodeScene MCP: safeguard and uplift AI-generated code. Catch issues early with our IDE extension and CLI tool.