Worktrunk is a CLI for Git worktree management, designed for parallel AI agent workflows
ci: run the full test suite across the feature powerset, not just lib (#3068)
Maximilian Roos465b8fd0229f
diff --git a/.github/workflows/nightly.yaml b/.github/workflows/nightly.yamlindex aca4b0fc7..47ac99f19 100644--- a/.github/workflows/nightly.yaml+++ b/.github/workflows/nightly.yaml@@ -1,7 +1,7 @@name: nightly# Slower checks that aren't worth running on every PR but should run before a# release. Jobs:-# - feature-powerset: feature-flag unification, check + lib tests (motivated by #2442)+# - feature-powerset: feature-flag unification, check + test suite (motivated by #2442)# - release-target: PTY+shell suite on the release triples ci.yaml doesn't cover# - benchmarks: full criterion suite + time-series gist append (cron only)# - check-unused-dependencies: cargo-udeps on nightly toolchain@@ -90,14 +90,16 @@ jobs:# worktrunk, or feature unification will mask gating bugs by silently# enabling `cli` in the lib build (see tests/helpers/wt-perf/Cargo.toml).#- # The test step runs the unit tests per combination — the check step+ # The test step runs the test suite per combination — the check step# strips dev-deps, so `#[cfg(test)]` code is invisible to it, and a test# importing a feature-gated symbol without its own gate, or a snapshot# baking feature-dependent output, only surfaces when tests are built and- # run per combo. Bounded to `--lib`: the integration suite snapshots `wt`- # output, which bakes in syntax-highlighted command rendering, so it only- # passes with syntax-highlighting on; ci.yaml's `test` matrix already- # runs it (with shell-integration-tests) on every PR.+ # run per combo. The integration suite declares+ # `required-features = ["cli", "syntax-highlighting"]` (Cargo.toml) because+ # it execs the `wt` binary and snapshots its highlighted output, so cargo+ # runs it only on combinations that satisfy both and runs lib + doc tests+ # everywhere. No target filter, so default-members builds the mock-stub+ # helper bin the lib suite shells out to.needs: gateif: needs.gate.outputs.run == 'true'runs-on: ubuntu-24.04@@ -113,15 +115,23 @@ jobs:with:tool: cargo-hack- - run: cargo hack check --feature-powerset --no-dev-deps+ # The `shell-integration-tests` combinations run PTY tests that spawn real+ # zsh/fish and probe nushell; match the test matrix's shell setup+ # (.github/actions/test-setup) so those combinations run, not just compile.+ - name: Install shells (zsh, fish)+ uses: awalsh128/[email protected]+ with:+ packages: zsh fish+ version: 1.0- # The lib suite shells out to the mock-stub helper bin, which `--lib`- # doesn't build (tests/CLAUDE.md documents the fresh-target panic).- # mock-stub doesn't depend on worktrunk features, so one build serves- # every combo.- - run: cargo build -p mock-stub+ - name: Install nushell+ uses: hustcer/setup-nu@v3+ with:+ version: '0.113.1'++ - run: cargo hack check --feature-powerset --no-dev-deps- - run: cargo hack test --feature-powerset --lib+ - run: cargo hack test --feature-powersetrelease-target:# Build and run the test suite on the release triples that ci.yaml'sdiff --git a/Cargo.toml b/Cargo.tomlindex cdcf31bec..32adcddd3 100644--- a/Cargo.toml+++ b/Cargo.toml@@ -71,6 +71,22 @@ required-features = ["cli"]name = "git-wt"path = "src/git_wt.rs"required-features = ["cli", "git-wt"]+# `git_wt.rs` is `include!("main.rs")`, so its unit tests are byte-identical+# duplicates of the `wt` binary's. Running them adds no coverage and breaks+# under `cargo hack test --feature-powerset`: insta derives snapshot filenames+# from the crate name, so the duplicates look for `git_wt__*.snap` files that+# don't exist (only `wt__*`). The `check` powerset still compiles this target.+test = false++# The integration suite execs the `wt` binary (needs `cli`) and snapshots its+# rendered output, which bakes in syntax highlighting (needs+# `syntax-highlighting`). Declaring both lets `cargo hack test+# --feature-powerset` skip this target on combinations that can't satisfy it+# rather than fail; lib and doc tests still run on every combination.+[[test]]+name = "integration"+path = "tests/integration.rs"+required-features = ["cli", "syntax-highlighting"][dependencies]anyhow = "1.0"diff --git a/tests/CLAUDE.md b/tests/CLAUDE.mdindex 4b30e00a7..8c1243e25 100644--- a/tests/CLAUDE.md+++ b/tests/CLAUDE.md@@ -10,7 +10,7 @@ cargo test --test integration # integrationcargo test --test integration --features shell-integration-tests # + shell tests```-A filtered `--test integration` run on a fresh `target/` panics with "mock-stub binary not found" (a target filter skips the helper-bin build). Fix: `cargo build -p mock-stub`, or use `cargo nextest run` / `cargo llvm-cov nextest`.+A target-filtered run (`--lib`, `--test integration`, …) on a fresh `target/` panics with "mock-stub binary not found" (a target filter skips the helper-bin build). Fix: `cargo build -p mock-stub`, or use `cargo nextest run` / `cargo llvm-cov nextest`.**Claude Code web:** `task setup-web` installs zsh/fish/nushell, `gh`, and dev tools. Install `task` first if needed: `sh -c "$(curl --location https://taskfile.dev/install.sh)" -- -d -b ~/bin` then `export PATH="$HOME/bin:$PATH"`. The permission tests (`test_permission_error_prevents_save`, `test_approval_prompt_permission_error`) skip automatically when running as root.