test: run the suite in parallel (bats --jobs), 48 min -> ~4.5 min

.github/workflows/test.yml

@@ -174,16 +174,17 @@ jobs:
 
       - name: Install bats-core + jq
         run: |
+          # GNU parallel lets bats run tests side by side (--jobs, below).
           if [[ "$RUNNER_OS" == "macOS" ]]; then
-            brew install bats-core jq
+            brew install bats-core jq parallel
           elif [[ "$RUNNER_OS" == "Windows" ]]; then
             # bats-core ships an npm package that drops a Windows-compatible
             # bats wrapper on PATH. jq is preinstalled on windows-latest.
             npm install -g bats
             jq --version
           else
             sudo apt-get update
-            sudo apt-get install -y bats jq
+            sudo apt-get install -y bats jq parallel
           fi
 
       - name: Verify dependencies
@@ -198,8 +199,20 @@ jobs:
           git config --global user.email "ci@example.com"
           git config --global user.name "CI"
 
+      # One job per CPU (serial on the Mac took 48 min for 2055 tests,
+      # 2026-10-29; 14 jobs took 4 min). Serial when GNU parallel is absent.
       - name: Run test suite
-        run: bats tests/
+        run: |
+          jobs=1
+          if command -v parallel >/dev/null 2>&1; then
+            jobs="$(getconf _NPROCESSORS_ONLN 2>/dev/null || echo 2)"
+          fi
+          echo "bats jobs: $jobs"
+          if (( jobs > 1 )); then
+            bats --jobs "$jobs" tests/
+          else
+            bats tests/
+          fi
 
   shellcheck:
     needs: changes

CONTRIBUTING.md

@@ -8,7 +8,7 @@ most things can ship quickly if they fit the design.
 1. **Fork** the repo and clone locally
 2. **Branch** off `main`: `git checkout -b your-branch-name`
 3. **Make changes** — see "What kinds of contributions are welcome" below
-4. **Run tests**: `bats tests/` (install via `brew install bats-core` if needed)
+4. **Run tests**: `bats --jobs 8 tests/` (install via `brew install bats-core parallel` if needed; plain `bats tests/` runs the same suite serially, about 10x slower)
 5. **Open a PR** against `main` with a description of the change
 
 ## What kinds of contributions are welcome
@@ -89,10 +89,15 @@ tests/             bats-core test suite
 The bats test suite is the primary safety net:
 
 ```bash
-bats tests/                    # run all
+bats --jobs 8 tests/           # run all, 8 at a time (needs GNU parallel)
+bats tests/                    # run all, one at a time
 bats tests/canonicalize.bats   # run one file
 ```
 
+Tests run side by side under `--jobs`, so a test keeps its state in its own
+temp dirs (`fi_setup_tmp`, `$BATS_TEST_TMPDIR`) and never kills processes by a
+pattern another test could also match.
+
 When adding code, add tests. Existing files in `tests/` are good
 reference — they cover lib functions, CLI subcommands, and hooks.
 

docs/found-issues.md

@@ -59,4 +59,5 @@ Maintained automatically by Claude. See <https://github.com/AltDoug/found-issues
 - [fixed] 2026-10-27 lib/autofix-ship.sh:658 — the 3.10.7 squash rebuild runs only after GitHub's AutomaticBaseChangeSucceededEvent, which needs the landing branch deleted; in a repo with delete_branch_on_merge off the landing branch survives its squash merge, so a fix PR still merges into that dead branch and never reaches the default branch. Seen on ByteTechSoftwares/sayciao#320 (base gsd/phase-24-resume squash-merged into main as #318 on 2026-10-25, branch kept, 49 ahead / 6 behind main) (suggested: before merging, if the base has a merged PR into the default branch whose squash is not an ancestor-carrier of the base tip, retarget with gh pr edit --base and reuse _fi_af_mwg_squashed with that PR as #M) (decide: When a fix PR's landing branch was squash-merged into the default branch but not deleted [no GitHub retarget], should merge-when-green retarget the PR to the default branch and rebuild it there, or stop and report?) (PR: AltDoug/found-issues#292) (fixed: 2026-10-27)
 - [fixed] 2026-10-27 tests/autofix-sweep.bats:1067 — 'a continuation's PR leaves the PR branch ledger alone' flaked on the Mac 2026-10-27 with status 141: git log ... | grep -q under pipefail, grep -q exits on the first match and git log dies of SIGPIPE (suggested: capture the log into a variable first, then grep it, as other tests do) (fix: small) (PR: AltDoug/found-issues#292) (fixed: 2026-10-27)
 - [open] 2026-10-29 lib/sync.sh:770 — SessionStart sync still writes the ledger on a pull-only clone through auto-archive (and merged-PR closures): measured 2026-10-29 on a main-only agent-config clone, sync moved 5 [fixed] entries to found-issues-archive.md; git refuses a later pull --ff-only even when upstream made the byte-identical edit, so the clone's update path breaks whenever upstream commits the same archive (suggested: read-only sync mode for consumer clones) (decided: build the read-only sync mode: a per-clone git config key plus an env override; picked over keeping the deploy workaround and over doing both [operator, 2026-10-29, via agent-config: Mods Dougstation]) (PR: AltDoug/found-issues#295)
-- [open] 2026-10-29 tests/autofix-sweep.bats — full bats suite runs serially and keeps growing: 24m08s on 2026-10-07, ~40 min on 2026-10-11, 48 min for 2048 tests on 2026-10-29 (this Mac); each autofix-* test builds a real repo + bare remote + worktree and runs the pipeline (measured 3.5-7.2 s per sweep test), and the fix-ship gate is 60 min (FOUND_ISSUES_FIX_SHIP_TIMEOUT_SECS), so headroom fell from ~20 to ~12 min (suggested: run bats --jobs N with GNU parallel, each test already uses its own mktemp dir; then cache per-file fixtures if still slow) (decided: fix it after 3.11.0 ships: run the suite in parallel with bats --jobs [GNU parallel as dev + CI dependency], then per-file fixture caching if still slow [operator, 2026-10-29])
+- [open] 2026-10-29 tests/autofix-sweep.bats — full bats suite runs serially and keeps growing: 24m08s on 2026-10-07, ~40 min on 2026-10-11, 48 min for 2048 tests on 2026-10-29 (this Mac); each autofix-* test builds a real repo + bare remote + worktree and runs the pipeline (measured 3.5-7.2 s per sweep test), and the fix-ship gate is 60 min (FOUND_ISSUES_FIX_SHIP_TIMEOUT_SECS), so headroom fell from ~20 to ~12 min (suggested: run bats --jobs N with GNU parallel, each test already uses its own mktemp dir; then cache per-file fixtures if still slow) (decided: fix it after 3.11.0 ships: run the suite in parallel with bats --jobs [GNU parallel as dev + CI dependency], then per-file fixture caching if still slow [operator, 2026-10-29]) (PR: AltDoug/found-issues#296)
+- [open] 2026-10-29 tests/helpers.bash:25 — tests that keep the real HOME write the operator's real ~/.cache/found-issues: measured 2026-10-29, ~/.cache/found-issues/mode_foo_bar was rewritten at 23:11 by a suite run (foo/bar is only a test fixture slug); one shared file across tests that run side by side under bats --jobs can also hand one test another's cached mode (suggested: per-test HOME in helpers, with git identity via GIT_AUTHOR_*/GIT_COMMITTER_* env so commits still work, or point XDG_CACHE_HOME at $BATS_TEST_TMPDIR in the files that do not set HOME) (fix: medium)

tests/autofix-cancel.bats

@@ -8,9 +8,12 @@ setup() {
   fi_setup_tmp; fi_af_fixture; fi_use_standins; fi_af_queue_fixture
   ST="$FI_AF_ST"
   export GH_MOCK_TRACE="$TMP/gh.trace"
+  # Sleep lengths unique to this test process, so the teardown below kills
+  # only this test's sleeps, never those of a test running beside it.
+  NAP_ENGINE="9${$}3" NAP_FOREIGN="9${$}4"
 }
 # A failing cancel must not leave the stand-in sleeping for an hour.
-teardown() { pkill -f 'sleep 471[34]' 2>/dev/null || true; fi_teardown_tmp; }
+teardown() { pkill -f "^sleep (${NAP_ENGINE}|${NAP_FOREIGN})\$" 2>/dev/null || true; fi_teardown_tmp; }
 
 # The engine child's group, once it runs: claim's base test run (3.2.1) has
 # its own, already finished, child group first.
@@ -77,7 +80,7 @@ wait_engine() {
 }
 
 @test "autofix cancel: an A run and its engine child are stopped" {
-  export FI_STANDIN_SLEEP=4713
+  export FI_STANDIN_SLEEP="$NAP_ENGINE"
   "$FI_BIN" autofix run "$ID" --engine claude >/dev/null 2>&1 &
   rpid=$!
   wait_engine
@@ -101,7 +104,7 @@ wait_engine() {
 }
 
 @test "autofix cancel: a pid that is not an autofix run is never signalled" {
-  sleep 4714 & spid=$!
+  sleep "$NAP_FOREIGN" & spid=$!
   wt="$("$FI_BIN" autofix claim "$ID")"
   printf 'pid=%s\n' "$spid" >>"$ST/running/$ID"
   run "$FI_BIN" autofix cancel "$ID"
@@ -111,7 +114,7 @@ wait_engine() {
 }
 
 @test "autofix cancel: a stale running item never signals the run that moved on" {
-  export FI_STANDIN_SLEEP=4713
+  export FI_STANDIN_SLEEP="$NAP_ENGINE"
   "$FI_BIN" autofix run "$ID" --engine claude >/dev/null 2>&1 &
   rpid=$!
   wait_engine

tests/cli-sync-default-branch.bats

@@ -5,12 +5,14 @@ setup() {
   fi_setup_tmp
   fi_init_git
   fi_use_gh_shim
+  # The default-branch cache lives under XDG_CACHE_HOME: give each test its
+  # own, inside $TMP. A shared ~/.cache plus a teardown that emptied it let
+  # one test delete another's cache when they ran side by side (bats --jobs).
+  export XDG_CACHE_HOME="$TMP/xdg-cache"
 }
 
 teardown() {
   fi_teardown_tmp
-  # Clean up any cache files this test created (they live globally in ~/.cache)
-  rm -rf ~/.cache/found-issues/default-branch-* 2>/dev/null || true
 }
 
 @test "sync: uses gh repo view fallback when origin/HEAD is unset and gh available" {

tests/fix-batch8-long-root.bats

@@ -10,7 +10,7 @@ setup() {
   fi_setup_tmp
   TMP_TOP="$TMP"
 }
-teardown() { pkill -f 'sleep 471[34]' 2>/dev/null || true; TMP="$TMP_TOP"; fi_teardown_tmp; }
+teardown() { TMP="$TMP_TOP"; fi_teardown_tmp; }
 
 long_dir() { printf "$1%.0s" $(seq 1 100); }
 

tests/helpers.bash

@@ -22,7 +22,10 @@ export FOUND_ISSUES_SEGMENT_AUTOSYNC=off
 
 # Segment renders write a cache (lib/segment-cache.sh) under the cache dir;
 # keep it out of the real ~/.cache. Tests that inspect it set their own.
-export FOUND_ISSUES_CACHE_DIR="${BATS_RUN_TMPDIR:-${TMPDIR:-/tmp}}/fi-test-cache"
+# One per test (BATS_TEST_TMPDIR exists when this file loads): auto-fix keeps
+# its run logs there (FI_AF_RUNS), and tests running in parallel (bats
+# --jobs) shared and mixed one another's logs.
+export FOUND_ISSUES_CACHE_DIR="${BATS_TEST_TMPDIR:-${BATS_RUN_TMPDIR:-${TMPDIR:-/tmp}}}/fi-test-cache"
 
 # A suite started inside an auto-fix run inherits FOUND_ISSUES_AUTOFIX_CHILD=1,
 # which makes session-start take its headless path and fails every test of