agent v0.85.0 WIP: F12/F11/F10/F9/F2/F1 boot-recovery plane + appliance self-heal (pre-build)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017CDMFpFx84pfviCTVuGGhf
2026-07-12 07:48:49 +02:00
parent bec4bac076
commit bc4eda926b
18 changed files with 1249 additions and 104 deletions
@@ -0,0 +1,96 @@
+package storage
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"runtime"
+	"strings"
+	"testing"
+)
+
+// F2/F1 (CAMPAIGN-3): a remove (and, via the same path, a rolled-back add) must leave ZERO residue —
+// no failed-state units, no leftover mountpoint dir. RemoveNetworkMount must reset-failed the stuck
+// unit BEFORE removing the files (or systemd keeps it as not-found/failed) and rmdir the mountpoint.
+func TestRemoveNetworkMount_ZeroResidue(t *testing.T) {
+	if runtime.GOOS == "windows" {
+		t.Skip("systemd-escaped unit filename contains a backslash; exercised on the Linux build server")
+	}
+	unitDir := t.TempDir()
+	stageDir := t.TempDir()
+	spec := NetworkMountSpec{Name: "media", Protocol: ProtocolNFS, Server: "10.0.0.5", Export: "/srv/media", MappedUID: 1000, MappedGID: 1000}
+	where := spec.Where()
+	mountName, err := UnitNameForMount(where)
+	if err != nil {
+		t.Fatalf("unit name: %v", err)
+	}
+	autoName := strings.TrimSuffix(mountName, ".mount") + ".automount"
+	// Both unit files present on disk (rm is recorded, so they stay for the assertion).
+	if err := os.WriteFile(filepath.Join(unitDir, mountName), []byte(renderNetworkMountUnit(spec)), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.WriteFile(filepath.Join(unitDir, autoName), []byte(renderNetworkAutomountUnit(spec)), 0o644); err != nil {
+		t.Fatal(err)
+	}
+
+	rr := &recordingRunner{}
+	ops := NewSudoHostOps(SudoHostOpsConfig{Runner: rr, Bins: Binaries{}.withDefaults(), UnitDir: unitDir, StageDir: stageDir, Host: &fakeHostReader{}, Logger: quietLogger()})
+	// The automount is in the failed state (start-limit residue) — must be reset-failed.
+	ops.unitFailed = func(_ context.Context, unit string) bool { return unit == autoName }
+
+	if err := ops.RemoveNetworkMount(context.Background(), "media"); err != nil {
+		t.Fatalf("remove: %v", err)
+	}
+
+	var sawReset, sawRmUnit, sawRmdir, sawReload bool
+	for _, c := range rr.calls {
+		j := strings.Join(c, " ")
+		switch {
+		case strings.Contains(j, "reset-failed -- "+autoName):
+			sawReset = true
+		case strings.Contains(j, "rm -f") && strings.Contains(j, autoName):
+			sawRmUnit = true
+		case strings.Contains(j, "rmdir") && strings.Contains(j, where):
+			sawRmdir = true
+		case strings.Contains(j, "daemon-reload"):
+			sawReload = true
+		}
+	}
+	if !sawReset {
+		t.Errorf("F2: a failed unit must be reset-failed on remove; calls: %v", rr.calls)
+	}
+	if !sawRmUnit {
+		t.Errorf("the unit files must be removed; calls: %v", rr.calls)
+	}
+	if !sawRmdir {
+		t.Errorf("F1: the empty mountpoint dir must be rmdir'd; calls: %v", rr.calls)
+	}
+	if !sawReload {
+		t.Errorf("daemon-reload must run after removal; calls: %v", rr.calls)
+	}
+}
+
+// rmdir is used (never rm -rf) — the fail-safe: a non-empty dir is left in place, not force-removed.
+func TestRemoveNetworkMount_NeverForceRemovesMountpoint(t *testing.T) {
+	if runtime.GOOS == "windows" {
+		t.Skip("systemd-escaped unit filename contains a backslash; exercised on the Linux build server")
+	}
+	unitDir := t.TempDir()
+	spec := NetworkMountSpec{Name: "media", Protocol: ProtocolNFS, Server: "10.0.0.5", Export: "/srv/media", MappedUID: 1000, MappedGID: 1000}
+	mountName, _ := UnitNameForMount(spec.Where())
+	autoName := strings.TrimSuffix(mountName, ".mount") + ".automount"
+	_ = os.WriteFile(filepath.Join(unitDir, mountName), []byte(renderNetworkMountUnit(spec)), 0o644)
+	_ = os.WriteFile(filepath.Join(unitDir, autoName), []byte(renderNetworkAutomountUnit(spec)), 0o644)
+
+	rr := &recordingRunner{}
+	ops := NewSudoHostOps(SudoHostOpsConfig{Runner: rr, Bins: Binaries{}.withDefaults(), UnitDir: unitDir, StageDir: t.TempDir(), Host: &fakeHostReader{}, Logger: quietLogger()})
+	if err := ops.RemoveNetworkMount(context.Background(), "media"); err != nil {
+		t.Fatalf("remove: %v", err)
+	}
+	for _, c := range rr.calls {
+		j := strings.Join(c, " ")
+		if strings.Contains(j, "rm -rf") || (strings.Contains(j, "rm ") && strings.Contains(j, "/mnt/felhom-drives/media") && !strings.Contains(j, "rmdir")) {
+			t.Errorf("mountpoint cleanup must be rmdir-only (never rm -rf); offending call: %s", j)
+		}
+	}
+}