@zerwiz/ymir 0.1.4 → 0.1.5

package/.agents/tests/fm-backend-autodetect-smoke.test.sh

@@ -0,0 +1,176 @@
+#!/usr/bin/env bash
+# tests/fm-backend-autodetect-smoke.test.sh - real herdr smoke test for runtime
+# backend AUTO-DETECTION (bin/fm-backend.sh's fm_backend_detect, wired into
+# fm_backend_name between config/backend and the tmux default).
+#
+# Unlike tests/fm-backend-herdr.test.sh (fake herdr CLI) and
+# tests/fm-backend-herdr-smoke.test.sh (real herdr, adapter primitives called
+# directly), this suite drives the REAL bin/fm-spawn.sh and bin/fm-teardown.sh
+# end to end, because auto-detection is a fm-spawn-TIME decision, not an
+# adapter primitive - it has to be proven where fm_backend_name is actually
+# called. The real spawn runs in a helper-provisioned, per-run named Herdr lab
+# session, with a scratch FM_HOME and scratch local-only project. Concurrent
+# copies therefore never share the default session or a workspace namespace.
+#
+# The complementary "tmux nested inside herdr resolves to tmux, silently" case
+# is covered as a fast, deterministic fake-tmux fm-spawn.sh test in
+# tests/fm-backend.test.sh (test_spawn_autodetect_nesting_resolves_tmux_silently).
+# Reproducing a genuinely nested real-tmux-inside-real-herdr pane here would
+# need a live attached tmux client, which a background test script cannot
+# manufacture; the selection LOGIC for that case is already exercised for real
+# by fm_backend_detect's own unit coverage plus that fake-tmux fm-spawn test.
+#
+# Safety (2026-07-02 incident): every test-owned Herdr operation goes through
+# bin/fm-herdr-lab.sh, which appends the named session flag and verifies the
+# default fleet session is unchanged after teardown. Never replace the helper
+# with an ambient HERDR_SESSION-only command.
+set -u
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+
+fail() { printf 'not ok - %s\n' "$1" >&2; exit 1; }
+pass() { printf 'ok - %s\n' "$1"; }
+assert_contains_local() {  # <haystack> <needle> <msg>
+  case "$1" in
+    *"$2"*) : ;;
+    *) fail "$3"$'\n'"--- got ---"$'\n'"$1" ;;
+  esac
+}
+
+command -v herdr >/dev/null 2>&1 || { echo "skip: herdr not found"; exit 0; }
+command -v jq >/dev/null 2>&1 || { echo "skip: jq not found (required by the herdr adapter)"; exit 0; }
+command -v treehouse >/dev/null 2>&1 || { echo "skip: treehouse not found (required by fm-spawn.sh)"; exit 0; }
+
+export FM_GATE_REFUSE_BYPASS=1
+
+# shellcheck source=tests/herdr-test-safety.sh
+. "$ROOT/tests/herdr-test-safety.sh"
+# This suite asserts that HERDR_ENV=1 alone selects the backend, and it runs
+# against its own isolated lab session. A Herdr pane inherited from the terminal
+# it was launched in must not follow spawn into that session as a cross-session
+# parent identity; the spawn below sets HERDR_ENV explicitly.
+herdr_forget_inherited_pane
+
+# TMP_ROOT is physically resolved (mktemp -d "$(pwd -P)"-relative) to keep this
+# real-herdr smoke fixture free of unrelated OS symlink noise.
+# The old fm-spawn bug that originally motivated this fixture shape was fixed in
+# fm-spawn-symlink-guard-s8: fm-spawn.sh now normalizes PROJ_ABS and observed
+# backend cwd reads before the worktree-discovery comparison.
+# The dedicated regression is
+# tests/fm-backend.test.sh:test_spawn_symlinked_project_prefix_avoids_false_refusal.
+TMP_ROOT=$(mktemp -d "$(cd "${TMPDIR:-/tmp}" && pwd -P)/fm-backend-autodetect-smoke.XXXXXX")
+HERDR_LAB_HELPER="$ROOT/bin/fm-herdr-lab.sh"
+HERDR_LAB_SESSION=$("$HERDR_LAB_HELPER" name fm-autodetect-smoke-concurrency-h3) || {
+  rm -rf "$TMP_ROOT"
+  fail "could not generate an isolated Herdr lab session name"
+}
+export HERDR_SESSION="$HERDR_LAB_SESSION"
+ID="autodetectsmoke1"
+WT=
+cleanup_all() {
+  local cleanup_status=0
+  [ -n "$WT" ] && command -v treehouse >/dev/null 2>&1 && treehouse return --force "$WT" >/dev/null 2>&1
+  "$HERDR_LAB_HELPER" teardown "$HERDR_LAB_SESSION" || cleanup_status=$?
+  rm -rf "$TMP_ROOT"
+  return "$cleanup_status"
+}
+on_exit() {
+  local status=$?
+  cleanup_all || status=$?
+  trap - EXIT
+  exit "$status"
+}
+trap on_exit EXIT
+"$HERDR_LAB_HELPER" provision "$HERDR_LAB_SESSION" || fail "could not provision isolated Herdr lab session"
+
+# --- scratch world: FM_HOME with NO backend config, one throwaway project ---
+
+STATE="$TMP_ROOT/state"; DATA="$TMP_ROOT/data"; CONFIG="$TMP_ROOT/config"
+mkdir -p "$STATE" "$DATA/$ID" "$CONFIG"
+# Backend auto-detection is what is under test here, so opt out of the default-on
+# presentation projection and keep the assertions on the flat per-home workspace.
+printf 'off\n' > "$CONFIG/herdr-presentation-spaces"
+printf 'trivial autodetect-smoke brief: nothing to do.\n' > "$DATA/$ID/brief.md"
+
+PROJ="$TMP_ROOT/scratch-project"
+mkdir -p "$PROJ"
+git -C "$PROJ" init -q
+printf '# scratch\n' > "$PROJ/README.md"
+git -C "$PROJ" add README.md
+git -C "$PROJ" -c user.name='Firstmate Tests' -c user.email='tests@example.invalid' commit -qm initial
+git clone --quiet --bare "$PROJ" "$PROJ.origin.git"
+git -C "$PROJ" remote add origin "file://$PROJ.origin.git"
+
+# --- spawn with NO explicit backend config; HERDR_ENV=1 is the only marker --
+
+OUT_FILE="$TMP_ROOT/spawn.out"; ERR_FILE="$TMP_ROOT/spawn.err"
+env -u TMUX -u FM_BACKEND PATH="$PATH" HERDR_ENV=1 \
+  FM_ROOT_OVERRIDE="$ROOT" FM_STATE_OVERRIDE="$STATE" FM_DATA_OVERRIDE="$DATA" \
+  FM_CONFIG_OVERRIDE="$CONFIG" FM_PROJECTS_OVERRIDE="$TMP_ROOT/unused-projects" \
+  FM_SPAWN_NO_GUARD=1 \
+  "$ROOT/bin/fm-spawn.sh" "$ID" "$PROJ" "sh -c 'echo autodetect-smoke-ok'" --mode no-mistakes --yolo off \
+  >"$OUT_FILE" 2>"$ERR_FILE"
+status=$?
+[ "$status" -eq 0 ] || fail "fm-spawn.sh did not succeed auto-detecting herdr"$'\n'"--- stdout ---"$'\n'"$(cat "$OUT_FILE")"$'\n'"--- stderr ---"$'\n'"$(cat "$ERR_FILE")"
+
+assert_contains_local "$(cat "$ERR_FILE")" "NOTICE" \
+  "fm-spawn.sh did not print the auto-detect notice to stderr when selecting herdr"
+assert_contains_local "$(cat "$ERR_FILE")" "EXPERIMENTAL herdr backend" \
+  "fm-spawn.sh's auto-detect notice did not flag herdr as experimental"
+pass "real herdr: fm-spawn.sh auto-detects herdr from HERDR_ENV=1 (no explicit config) and prints the loud notice"
+
+META="$STATE/$ID.meta"
+[ -f "$META" ] || fail "fm-spawn.sh did not write a meta file for $ID"
+assert_contains_local "$(cat "$META")" "backend=herdr" \
+  "auto-detected spawn did not record backend=herdr in meta"
+assert_contains_local "$(cat "$META")" "herdr_session=$HERDR_LAB_SESSION" \
+  "auto-detected spawn did not record the isolated herdr_session in meta"
+
+WORKSPACE=$(grep '^herdr_workspace_id=' "$META" | cut -d= -f2-)
+[ -n "$WORKSPACE" ] || fail "auto-detected spawn meta is missing herdr_workspace_id"
+
+TAB=$(grep '^herdr_tab_id=' "$META" | cut -d= -f2-)
+[ -n "$TAB" ] || fail "auto-detected spawn meta is missing herdr_tab_id"
+
+WT=$(grep '^worktree=' "$META" | cut -d= -f2-)
+if [ -z "$WT" ] || [ ! -d "$WT" ]; then
+  fail "auto-detected spawn did not report a real worktree path"
+fi
+
+PANE=$(grep '^herdr_pane_id=' "$META" | cut -d= -f2-)
+[ -n "$PANE" ] || fail "auto-detected spawn meta is missing herdr_pane_id"
+pass "real herdr: auto-detected spawn records backend=herdr and herdr_session/workspace/tab/pane fields in meta"
+
+# --- confirm the trivial launch command actually ran in the herdr pane ------
+
+sleep 1
+CAPTURED=$("$HERDR_LAB_HELPER" run "$HERDR_LAB_SESSION" pane read "$PANE" --source recent --lines 200) || \
+  fail "capture failed on the auto-detected herdr pane"
+CAPTURED=$(printf '%s\n' "$CAPTURED" | tail -n 30)
+case "$CAPTURED" in
+  *autodetect-smoke-ok*) : ;;
+  *) fail "the raw launch command did not run in the auto-detected herdr pane"$'\n'"$CAPTURED" ;;
+esac
+pass "real herdr: the auto-detected spawn's launch command actually ran in the herdr pane"
+
+# --- teardown completes the trivial spawn/teardown cycle --------------------
+
+TEARDOWN_OUT="$TMP_ROOT/teardown.out"
+FM_ROOT_OVERRIDE="$ROOT" FM_STATE_OVERRIDE="$STATE" FM_DATA_OVERRIDE="$DATA" \
+  FM_CONFIG_OVERRIDE="$CONFIG" \
+  "$ROOT/bin/fm-teardown.sh" "$ID" >"$TEARDOWN_OUT" 2>&1
+status=$?
+[ "$status" -eq 0 ] || fail "fm-teardown.sh failed for the auto-detected herdr task"$'\n'"$(cat "$TEARDOWN_OUT")"
+[ -f "$META" ] && fail "fm-teardown.sh did not remove $META"
+if "$HERDR_LAB_HELPER" run "$HERDR_LAB_SESSION" pane get "$PANE" >/dev/null 2>&1; then
+  fail "fm-teardown.sh did not close the auto-detected herdr pane"
+fi
+WT=
+pass "real herdr: teardown completes the auto-detected spawn/teardown cycle (meta cleared, pane closed)"
+
+if ! cleanup_all; then
+  trap - EXIT
+  fail "isolated Herdr lab teardown failed or the default fleet session changed"
+fi
+trap - EXIT
+pass "real herdr: isolated lab session removed and default fleet session unchanged"

package/.agents/tests/fm-backend-cmux-smoke.test.sh

@@ -0,0 +1,188 @@
+#!/usr/bin/env bash
+# tests/fm-backend-cmux-smoke.test.sh - real cmux smoke test for the cmux
+# session-provider adapter (bin/backends/cmux.sh), verified against the real
+# cmux 0.64.17 binary (docs/cmux-backend.md). Mirrors
+# tests/fm-backend-zellij-smoke.test.sh's/fm-backend-herdr-smoke.test.sh's
+# structure: every other suite fakes the CLI, this one talks to the REAL app -
+# but unlike herdr/zellij there is no isolated throwaway SESSION to spin up:
+# cmux is one shared, GUI-first, macOS-only instance (the same posture as
+# Orca). So this test creates ONLY `fm-test-`-prefixed task labels, touches and
+# closes ONLY what it created, never enumerates-and-closes, never quits or
+# relaunches the app, and cleans up every artifact via
+# tests/cmux-test-safety.sh's guarded close. The adapter turns those plain
+# labels into home-scoped cmux workspace titles internally.
+#
+# Skips cleanly when cmux (or jq) is not installed/reachable, so CI/dev
+# machines without cmux, or without the one-time password-mode setup
+# (docs/cmux-backend.md "Setup"), are unaffected.
+set -u
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+
+fail() { printf 'not ok - %s\n' "$1" >&2; cleanup_all; exit 1; }
+pass() { printf 'ok - %s\n' "$1"; }
+
+command -v jq >/dev/null 2>&1 || { echo "skip: jq not found (required by the cmux adapter)"; exit 0; }
+
+# shellcheck source=/dev/null
+. "$ROOT/bin/fm-backend.sh"
+fm_backend_source cmux || { echo "skip: could not source the cmux adapter"; exit 0; }
+
+fm_backend_cmux_tool_check >/dev/null 2>&1 || { echo "skip: cmux CLI not found on PATH or at the bundle path"; exit 0; }
+fm_backend_cmux_version_check >/dev/null 2>&1 || { echo "skip: installed cmux is older than the verified minimum"; exit 0; }
+PING_STATE=$(fm_backend_cmux_ping_state)
+[ "$PING_STATE" = ok ] || { echo "skip: cmux socket not reachable/authenticated (state=$PING_STATE) - see docs/cmux-backend.md 'Setup'"; exit 0; }
+
+# shellcheck source=tests/cmux-test-safety.sh
+. "$ROOT/tests/cmux-test-safety.sh"
+
+WS1=""
+WS2=""
+cleanup_all() {
+  [ -z "$WS1" ] || cmux_safe_close_workspace "$WS1" "fm-test-smoke1"
+  [ -z "$WS2" ] || cmux_safe_close_workspace "$WS2" "fm-test-smoke2"
+}
+trap cleanup_all EXIT
+
+# --- create_task + duplicate refusal -----------------------------------------
+
+LABEL="fm-test-smoke1"
+TASK_IDS=$(fm_backend_cmux_create_task "$LABEL" /tmp) || fail "create_task failed"
+read -r WS1 SF1 <<EOF
+$TASK_IDS
+EOF
+if [ -z "$WS1" ] || [ -z "$SF1" ]; then
+  fail "create_task did not return workspace/surface ids"
+fi
+TARGET="$WS1:$SF1"
+
+if fm_backend_cmux_create_task "$LABEL" /tmp >/dev/null 2>&1; then
+  fail "create_task should refuse a duplicate workspace title (cmux itself does not enforce uniqueness)"
+fi
+pass "real cmux: create_task creates a workspace/surface and refuses a duplicate title"
+
+fm_backend_cmux_send_key "$TARGET" Escape "$LABEL" \
+  || fail "send_key with a matching expected task label should succeed"
+if fm_backend_cmux_send_key "$TARGET" Escape "fm-test-not-$LABEL" >/dev/null 2>&1; then
+  fail "send_key with a mismatched expected task label should fail"
+fi
+pass "real cmux: expected task label verification accepts the matching workspace and rejects a mismatch"
+
+# --- send_literal + send_key(Enter), the two-step submit form ---------------
+
+fm_backend_cmux_send_literal "$TARGET" 'echo literal-then-key-captain' \
+  || fail "send_literal failed"
+sleep 0.3
+fm_backend_cmux_send_key "$TARGET" Enter || fail "send_key Enter failed"
+sleep 0.5
+out=$(fm_backend_cmux_capture "$TARGET" 20) || fail "capture failed after send_literal+send_key"
+case "$out" in
+  *literal-then-key-captain*) : ;;
+  *) fail "real cmux: send_literal + send_key(Enter) did not submit and echo the line"$'\n'"$out" ;;
+esac
+pass "real cmux: send_literal (unsubmitted) + send_key Enter submit as two steps and the output is capturable"
+
+# --- send_text_line (the composed form) --------------------------------------
+
+fm_backend_cmux_send_text_line "$TARGET" "echo captain-on-deck-line" \
+  || fail "send_text_line failed"
+sleep 0.5
+out=$(fm_backend_cmux_capture "$TARGET" 20) || fail "capture failed after send_text_line"
+case "$out" in
+  *captain-on-deck-line*) : ;;
+  *) fail "real cmux: send_text_line did not run and echo the line"$'\n'"$out" ;;
+esac
+pass "real cmux: send_text_line composes send+Enter and its output is capturable"
+
+# --- current_path: verified zellij-shape frozen cwd --------------------------
+
+fm_backend_cmux_send_text_line "$TARGET" "cd /tmp"
+sleep 0.3
+p=$(fm_backend_cmux_current_path "$TARGET") || fail "current_path failed"
+case "$p" in
+  */tmp) : ;;
+  *) fail "real cmux: current_path did not report the surface's cwd after a direct cd, got '$p'" ;;
+esac
+pass "real cmux: current_path reads the surface's live cwd after a direct cd"
+
+# The load-bearing case: a NESTED SUBSHELL's own cd (exactly what `treehouse
+# get` does). Verified real finding (docs/cmux-backend.md finding #2):
+# current_directory stays frozen at wherever the surface's shell was when it
+# launched the subshell as a foreground command - it never follows the
+# subshell's own cd. fm_backend_cmux_current_path's active pwd-probe is what
+# fm-spawn.sh's worktree-discovery poll actually depends on, so this must be
+# proven against a real subshell, not just a plain cd in the top-level shell.
+fm_backend_cmux_send_text_line "$TARGET" 'cd / && bash'
+sleep 0.5
+fm_backend_cmux_send_text_line "$TARGET" "cd /private/tmp"
+sleep 0.3
+p2=$(fm_backend_cmux_current_path "$TARGET") || fail "current_path failed inside a nested subshell"
+case "$p2" in
+  */private/tmp|*/tmp) : ;;
+  *) fail "real cmux: current_path did not track a nested subshell's own cd (the treehouse-get-shaped case), got '$p2'" ;;
+esac
+pass "real cmux: current_path tracks a NESTED SUBSHELL's own cd (the treehouse-get-shaped case a bare cwd read cannot see)"
+fm_backend_cmux_send_text_line "$TARGET" 'exit'
+sleep 0.3
+
+# --- key names: Escape and Ctrl-C, verified names --------------------------
+
+fm_backend_cmux_send_key "$TARGET" Escape || fail "send_key Escape failed"
+pass "real cmux: send_key Escape (natively supported, unlike Orca) succeeds"
+
+fm_backend_cmux_send_key "$TARGET" C-c || fail "send_key C-c (normalized to 'ctrl-c') failed"
+pass "real cmux: send_key C-c (normalized to the verified 'ctrl-c' name) succeeds"
+
+# --- busy_state: always unknown (no native agent-state primitive) -----------
+
+bs=$(fm_backend_busy_state cmux "$TARGET")
+[ "$bs" = unknown ] || fail "fm_backend_busy_state should report unknown for cmux (no native primitive), got '$bs'"
+pass "real cmux: fm_backend_busy_state reports unknown (watcher falls back to pane-regex, same as tmux/zellij/orca)"
+
+# --- window_of_workspace: real-cmux window/count detection -------------------
+# The last-workspace-in-a-window teardown fix (docs/cmux-backend.md "Closing the
+# last workspace in a window") pivots on window_of_workspace reading the live
+# `list-windows` / `workspace list --window` JSON correctly (the version-fragile
+# part the fake-CLI suite cannot prove). This task workspace shares its window
+# with at least the app's own workspace, so it reports "<window_id> <count>"
+# with a count of two or more. The last-in-window branch itself is proven end to
+# end in the fake-CLI suite and in this document's manual verification record;
+# it is not driven live here because closing the last workspace inherently
+# leaves a window cmux cannot close over the control socket.
+WININFO=$(fm_backend_cmux_window_of_workspace "$WS1")
+case "$WININFO" in
+  *' '[0-9]*) : ;;
+  *) fail "window_of_workspace did not report '<window_id> <count>' for a live task workspace, got '$WININFO'" ;;
+esac
+WCOUNT=${WININFO##* }
+[ "$WCOUNT" -ge 2 ] 2>/dev/null \
+  || fail "task workspace shares its window with the app default, so the count should be >= 2, got '$WININFO'"
+pass "real cmux: window_of_workspace locates a task workspace's window and counts its workspaces"
+
+# --- kill: whole-workspace close ----------------------------------------------
+
+fm_backend_cmux_kill "$TARGET"
+sleep 0.5
+STILL_LIVE=$(fm_backend_cmux_cli workspace list --json --id-format uuids 2>/dev/null | jq -r --arg id "$WS1" '.workspaces[]? | select(.id == $id) | .id' 2>/dev/null)
+[ -z "$STILL_LIVE" ] || fail "kill did not remove the whole task workspace"
+WS1=""
+# Best-effort contract: killing an already-gone target must not error.
+fm_backend_cmux_kill "$TARGET" || fail "kill on an already-dead target must stay best-effort (never fail)"
+pass "real cmux: kill removes the whole workspace and is idempotent/best-effort"
+
+# --- list_live (title-based recovery discovery) ------------------------------
+
+LABEL2="fm-test-smoke2"
+TASK_IDS2=$(fm_backend_cmux_create_task "$LABEL2" /tmp) || fail "second create_task failed"
+read -r WS2 _SF2 <<EOF
+$TASK_IDS2
+EOF
+live=$(fm_backend_cmux_list_live)
+case "$live" in
+  *"$LABEL2"*) : ;;
+  *) fail "list_live did not report the freshly created task workspace by title"$'\n'"--- got ---"$'\n'"$live" ;;
+esac
+pass "real cmux: list_live discovers a live task workspace by fm-<id> title"
+
+cleanup_all
+trap - EXIT