fix(gc): wasn't killing zmx sessions properly

...	...		@@ -666,9 +666,6 @@ func getDomain(eventType string) string {
666	666		if eventType == "local" {
667	667		return "local"
668	668		}
669		-	if eventType == "test-ci" \|\| strings.HasPrefix(eventType, "test-") {
670		-	return eventType
671		-	}
672	669		return "ci"
673	670		}
674	671

...	...		@@ -1919,28 +1916,29 @@ func cancelJobSessions(prefix string) {
1919	1916		}
1920	1917		}
1921	1918
1922		-	// runGC kills all ci. zmx sessions older than 3 hours, regardless of status.
1923		-	func runGC(cfg *Cfg) error {
1924		-	log := cfg.Logger.With("cmd", "gc")
1925		-	log.Debug("running garbage collection")
1926		-
1927		-	listOutput, err := exec.Command("zmx", "list").CombinedOutput()
1928		-	if err != nil {
1929		-	return fmt.Errorf("zmx list: %w", err)
1930		-	}
1931		-
1932		-	sessions := parseZMXList(string(listOutput))
1933		-
1934		-	cutoff := time.Now().Add(-3 * time.Hour).Unix()
1935		-
	1919	+	// findSessionsForGC identifies finished sessions or sessions older than 3 hours matching given prefixes.
	1920	+	func findSessionsForGC(sessions []SessionInfo, prefixes []string, now time.Time) []string {
	1921	+	cutoff := now.Add(-3 * time.Hour).Unix()
1936	1922		var toKill []string
1937	1923		for _, s := range sessions {
1938		-	if !strings.HasPrefix(s.Name, "ci.") && !strings.HasPrefix(s.Name, "local.") {
	1924	+	matched := false
	1925	+	for _, prefix := range prefixes {
	1926	+	if strings.HasPrefix(s.Name, prefix) {
	1927	+	matched = true
	1928	+	break
	1929	+	}
	1930	+	}
	1931	+	if !matched {
	1932	+	continue
	1933	+	}
	1934	+
	1935	+	if s.Ended != "" {
	1936	+	toKill = append(toKill, s.Name)
1939	1937		continue
1940	1938		}
1941	1939
1942	1940		if s.Created == "" {
1943		-	continue // skip sessions with no creation time
	1941	+	continue
1944	1942		}
1945	1943
1946	1944		var created int64

...	...		@@ -1949,10 +1947,34 @@ func runGC(cfg *Cfg) error {
1949	1947		}
1950	1948
1951	1949		if created < cutoff {
1952		-	log.Debug("session expired, scheduling for gc", "session", s.Name, "created", s.Created)
1953	1950		toKill = append(toKill, s.Name)
1954	1951		}
1955	1952		}
	1953	+	return toKill
	1954	+	}
	1955	+
	1956	+	// runGC kills finished zmx sessions or sessions older than 3 hours.
	1957	+	func runGC(cfg *Cfg) error {
	1958	+	log := cfg.Logger.With("cmd", "gc")
	1959	+	log.Debug("running garbage collection")
	1960	+
	1961	+	listOutput, err := exec.Command("zmx", "list").CombinedOutput()
	1962	+	if err != nil {
	1963	+	return fmt.Errorf("zmx list: %w", err)
	1964	+	}
	1965	+
	1966	+	sessions := parseZMXList(string(listOutput))
	1967	+
	1968	+	prefixes := []string{"ci.", "local."}
	1969	+	if cfg != nil && cfg.SessionPrefix != "" {
	1970	+	p := cfg.SessionPrefix
	1971	+	if !strings.HasSuffix(p, ".") {
	1972	+	p += "."
	1973	+	}
	1974	+	prefixes = append(prefixes, p)
	1975	+	}
	1976	+
	1977	+	toKill := findSessionsForGC(sessions, prefixes, time.Now())
1956	1978
1957	1979		if len(toKill) == 0 {
1958	1980		log.Debug("no sessions to garbage collect")

...	...		@@ -48,10 +48,10 @@ zmx run step2 echo "hello from step2"
48	48		ctx, cancel := context.WithCancel(context.Background())
49	49		testJobID := fmt.Sprintf("j-%d", time.Now().UnixNano()%100000000)
50	50		t.Cleanup(func() {
51		-	_ = exec.Command("zmx", "kill", "-f", fmt.Sprintf("test-ci.test-repo.%s", testJobID)).Run()
	51	+	_ = exec.Command("zmx", "kill", "-f", fmt.Sprintf("local.test-repo.%s", testJobID)).Run()
52	52		})
53	53		event := Event{
54		-	Type: "test-ci",
	54	+	Type: "local",
55	55		Name: "test-repo",
56	56		JobID: testJobID,
57	57		Workspace: workspaceDir,

...	...		@@ -68,7 +68,7 @@ zmx run step2 echo "hello from step2"
68	68		NewWorkspace: defaultWorkspaceFactory,
69	69		StatusOutput: statusBuf,
70	70		IncludeRunning: true,
71		-	SessionPrefix: "test-ci.",
	71	+	SessionPrefix: "local.",
72	72		}
73	73
74	74		// 3. Run the runner (fire-and-forget, exits quickly)

...	...		@@ -185,8 +185,8 @@ zmx run step2 echo "hello from step2"
185	185		if finalPayload == nil {
186	186		t.Fatal("no final payload")
187	187		}
188		-	// Filter to only sessions with our job prefix to avoid picking up unrelated test-ci.* sessions
189		-	expectedPrefix := fmt.Sprintf("test-ci.%s.%s.", finalPayload.Name, finalPayload.JobID)
	188	+	// Filter to only sessions with our job prefix to avoid picking up unrelated local.* sessions
	189	+	expectedPrefix := fmt.Sprintf("local.%s.%s.", finalPayload.Name, finalPayload.JobID)
190	190		sessions := make([]SessionInfo, 0, len(finalPayload.Sessions))
191	191		for _, s := range finalPayload.Sessions {
192	192		if strings.HasPrefix(s.Name, expectedPrefix) {

...	...		@@ -559,7 +559,6 @@ func TestGetDomain(t *testing.T) {
559	559		{"ci", "ci"},
560	560		{"", "ci"},
561	561		{"local", "local"},
562		-	{"test-ci", "test-ci"},
563	562		}
564	563
565	564		for _, tt := range tests {

...	...		@@ -639,6 +638,27 @@ func findRunningJobsFromOutput(output, name string) ([]string, []SessionInfo) {
639	638		return runners, sessions
640	639		}
641	640
	641	+	func TestFindSessionsForGC(t *testing.T) {
	642	+	now := time.Unix(1786289700, 0)
	643	+	twoHoursAgo := fmt.Sprintf("%d", now.Add(-2*time.Hour).Unix())
	644	+	fourHoursAgo := fmt.Sprintf("%d", now.Add(-4*time.Hour).Unix())
	645	+
	646	+	sessions := []SessionInfo{
	647	+	{Name: "ci.repo.active.runner", Created: twoHoursAgo, Ended: ""},
	648	+	{Name: "ci.repo.finished.runner", Created: twoHoursAgo, Ended: "1786289600"},
	649	+	{Name: "ci.repo.expired.runner", Created: fourHoursAgo, Ended: ""},
	650	+	{Name: "other.session", Created: fourHoursAgo, Ended: ""},
	651	+	}
	652	+
	653	+	toKill := findSessionsForGC(sessions, []string{"ci.", "local."}, now)
	654	+	if len(toKill) != 2 {
	655	+	t.Fatalf("expected 2 sessions for GC, got %d: %v", len(toKill), toKill)
	656	+	}
	657	+	if toKill[0] != "ci.repo.finished.runner" \|\| toKill[1] != "ci.repo.expired.runner" {
	658	+	t.Errorf("unexpected sessions for GC: %v", toKill)
	659	+	}
	660	+	}
	661	+
642	662		func TestKillSessionsEmpty(t *testing.T) {
643	663		// Should not error with empty list
644	664		if err := killSessions(nil); err != nil {

...	...		@@ -795,7 +815,7 @@ func TestWaitAndReport_Cancellation(t *testing.T) {
795	815		}
796	816
797	817		jobID := fmt.Sprintf("canceltest-%d", time.Now().UnixNano())
798		-	prefix := "test-local.testrepo." + jobID + "."
	818	+	prefix := "local.testrepo." + jobID + "."
799	819		runnerSession := prefix + "runner"
800	820
801	821		t.Cleanup(func() {

...	...		@@ -812,7 +832,7 @@ func TestWaitAndReport_Cancellation(t *testing.T) {
812	832		cancel()
813	833		}()
814	834
815		-	err := waitAndReport(cfg, nil, "testrepo", jobID, "test-local")
	835	+	err := waitAndReport(cfg, nil, "testrepo", jobID, "local")
816	836		if err == nil {
817	837		t.Error("expected error when waitAndReport is cancelled, got nil")
818	838		}

Commit fb2b232