//go:build unix

package host

import (
	"context"
	"encoding/json"
	"testing"
	"time"

	"github.com/carlosframework/platform/internal/hostbox"
	"github.com/carlosframework/platform/internal/hostclaim"
	"github.com/carlosframework/platform/internal/hostguard"
	"github.com/carlosframework/platform/internal/registry"
)

// Carry 9 (Task 9 review, ruling 82): a relocation completes on box-1 with
// region HA on; box-1 restarts with CARLOS_REGION_HA off. One scenario,
// with no agent restart between its halves, over one control bucket, one
// guard journal, one registry and one activator:
//
//  1. the probe stop: the production probe-stop provider fences the probe
//     session under its persisted owner and root's writer-session stop
//     lands in the guard journal; the serving CAS records the completion;
//  2. the completion pass: NewRelocationRuntime with claim mode alone
//     builds the completion-only runtime, whose pass discharges through
//     main's own DischargeCompleted, releases the hold, lifts the owned
//     fence and marks the record, so the barrier opens;
//  3. the first serving wake: the wake runtime (ClaimWakeAssembly.Build)
//     wired as OpenProductionClaimAssembly wires it with moves C1 off
//     (wireProductionAdmission, N6 in front) takes the predecessor stop
//     Prerequisites takes, which re-stops nothing, then admits through
//     the runtime's own Admit, which hands root the probe's receipt as the
//     prior.
//
// The production constructor itself cannot run here in agent mode: it
// dials root at the fixed controlsock.Path, chowns its state directory to
// the carlos uid and needs a cgroup manager. Checklist item 12 is that run.
func TestCompletionPassThenFirstServingWakeWithRegionHAOff(t *testing.T) {
	t.Parallel()
	ctx := t.Context()
	f := newRecoveryFixture(t)
	const box, op = "box-1", "rel-1"
	local := claimStopLocal{Actor: hostbox.Actor{Box: box, Enrollment: "enrollment-1", Revision: 7, Binding: "7", Region: "ie"}, BootID: "boot-0", ControllerID: "controller-0", PID: 4000, ProcessStart: "start-0"}

	// The relocating claim at probing on box-1, its probe passed with
	// session 12 at install stamp 9.
	// Written whole, as the stop-probe fixture's claim is: the transitions
	// that lead here are hostrelocate's and are tested there.
	probe, _ := json.Marshal(hostclaim.Probe{Box: box, Session: 12, At: f.now})
	created := f.read()
	c := *created.Claim
	holder := *c.Holder
	holder.Stamp, holder.Session, holder.Generation, holder.WriterBox = 9, 12, "P1", box
	holder.Admissions = []hostclaim.AdmissionRecord{{Session: 12, Box: box, Generation: "P1", Mode: "cold", Region: "ie"}}
	c.Holder = &holder
	c.Gen = 30 // the history behind stamp 9 and session 12
	c.Op = hostclaim.Op{Kind: hostclaim.KindRelocating, ID: op, Phase: hostclaim.PhaseRelocatingProbing, Exec: box, Participants: []string{"src", box},
		Intent: map[string]string{"region": "ie", "source_box": "src", "target_box": box}, Journal: map[string]string{"probe": string(probe)}}
	key, _ := hostclaim.Key(recoveryHost)
	if _, err := f.claims.Put(ctx, key, mustJSON(t, &c), created.ETag); err != nil {
		t.Fatal(err)
	}
	o := f.read()
	if err := f.guard.Activate(ctx, recoveryHost, box, nil); err != nil {
		t.Fatal(err)
	}

	// 1. The probe stop, root's writer-session stop standing as Wake.Stop.
	h := o.Claim.Holder
	target := hostguard.FenceTarget{FenceID: "fence-1", Region: "ie", Host: recoveryHost, Incarnation: h.Incarnation, Stamp: 9, Session: 12, Op: op, Phase: hostclaim.PhaseRelocatingProbing,
		ObligationID: hostguard.WriterStopObligation(12), Box: box, Enrollment: local.Actor.Enrollment, BootID: local.BootID, ControllerID: local.ControllerID, ClaimETag: o.ETag, DirectoryRevision: "7", RosterRevision: 7}
	at := f.now.Add(time.Minute)
	receipt := mustMarshalStopReceipt(StopReceipt{V: 1, ID: "r1", Kind: StopReceiptStopped, Region: "ie", ClaimETag: o.ETag, ObligationID: target.ObligationID, FenceID: "fence-1", Host: recoveryHost, Incarnation: h.Incarnation,
		InstallationStamp: 9, WriterSession: 12, OperationID: op, Phase: hostclaim.PhaseRelocatingProbing, Box: box, Enrollment: local.Actor.Enrollment, BootID: local.BootID, ControllerID: local.ControllerID, DirectoryRevision: testReceiptNumeral(target), Binding: "7", EvidenceAt: at})
	var order []string
	act := newStopProbeActivator(&order)
	probeAssembly := ClaimAssembly{Claims: f.claims, Guard: f.guard, stopLocal: func(context.Context) (claimStopLocal, error) { return local, nil },
		Wake: &ClaimWakeAssembly{Stop: func(ctx context.Context, _ hostclaim.Observed) error {
			if _, err := f.guard.BeginFence(ctx, target); err != nil {
				return err
			}
			return f.guard.CompleteFence(ctx, target, receipt, at, func(context.Context, hostguard.FenceTarget, []byte, time.Time) error { return nil })
		}}}
	stop, err := probeAssembly.relocationStopProbe(RelocationAssemblyConfig{Activator: act, DataDir: f.guard.DataDir})(ctx, o)
	if err != nil {
		t.Fatal(err)
	}
	if !act.svc.Fenced(recoveryHost) {
		t.Fatal("the probe stop installed no fence")
	}
	// The serving CAS, as CompleteRelocation leaves it.
	done, _ := json.Marshal(hostclaim.CompletedRelocation{Op: op, Stop: stop, Stamp: 9, Phase: hostclaim.PhaseRelocatingProbing, Generation: "P1"})
	c.Gen = 31
	c.Op = hostclaim.Op{Kind: hostclaim.KindServing, Exec: box, Participants: []string{box}, Journal: map[string]string{hostclaim.JournalCompleted: string(done)}}
	if _, err := f.claims.Put(ctx, key, mustJSON(t, &c), o.ETag); err != nil {
		t.Fatal(err)
	}
	reg, err := registry.Open(t.TempDir() + "/registry.db")
	if err != nil {
		t.Fatal(err)
	}
	t.Cleanup(func() { reg.Close() })
	if err := reg.Upsert(registry.Route{Host: recoveryHost, Status: "relocating", HoldOp: op, Socket: "/run/shop.sock"}); err != nil {
		t.Fatal(err)
	}

	// 2. The completion pass after the restart with region HA off: claim
	// mode alone, main's own participation owner and discharge.
	identity := handoverRecoveryIdentity(t, local.Actor)
	bound := &ClaimAssembly{Claims: f.claims, Guard: f.guard, stopLocal: func(context.Context) (claimStopLocal, error) { return local, nil }}
	owner := &productionClaimLocal{legacyParticipation: true, Resources: ClaimRuntimeResources{Mode: "agent", DataDir: f.guard.DataDir}, Identity: identity, Store: f.claims,
		slots: map[string]*claimParticipation{}, assembly: bound, lifetime: ctx}
	claimOnly := func(k string) string {
		if k == "CARLOS_HOST_CLAIM" {
			return "1"
		}
		return ""
	}
	r, err := NewRelocationRuntime(ctx, claimOnly, f.claims, owner, ClaimProviders{}, reg, act, relocationRuntimeInputs{})
	if err != nil || r == nil || r.Worker != nil || r.Participant != nil || r.Convert != nil {
		t.Fatalf("not the completion-only runtime: %+v %v", r, err)
	}
	r.once(ctx, f.claims)
	prior := f.read()
	marked, ok, err := hostclaim.ReadCompletedRelocation(prior.Claim)
	if err != nil || !ok || !marked.Discharged {
		t.Fatalf("the completion was not marked: %+v %v %v", marked, ok, err)
	}
	if row, err := reg.Get(recoveryHost); err != nil || row == nil || row.HoldOp != "" {
		t.Fatalf("the hold was not released: %+v %v", row, err)
	}
	if act.svc.Fenced(recoveryHost) {
		t.Fatal("the probe's owned fence was not lifted")
	}
	if err := hostclaim.StartGuard(prior.Claim, []string{"uk"}, []string{"spa"}, ""); err != nil {
		t.Fatalf("the barrier did not open: %v", err)
	}

	// 3. The first serving wake, N6 in front as the production
	// constructor wires it with moves C1 off.
	w := f.controller("boot-1", "controller-1", 4101, "start-1")
	socket, posted := handoverRoot(t)
	locals := 0
	w.wake.AfterAdmission = func(context.Context, hostclaim.Observed, hostclaim.Observed) error { locals++; return nil }
	wireProductionAdmission(w.wake, box, false, claimCrossBoxHandover{Box: box}, nil,
		&relocationCompletedHandover{Guard: f.guard, Identity: identity, Socket: socket, Box: box})
	// Prerequisites' predecessor stop: the probe's receipt is found and
	// verified; root is asked to stop nothing.
	if err := w.wake.stopPredecessor(ctx, prior); err != nil || len(f.requests) != 0 {
		t.Fatalf("the predecessor stop: %v (root asked %d stops)", err, len(f.requests))
	}
	admitted := w.admit(t, f)
	if admitted.Claim.Holder.Session <= 12 || admitted.Claim.Holder.WriterBox != box {
		t.Fatalf("not a new writer session on %s: %+v", box, admitted.Claim.Holder)
	}
	if len(*posted) != 1 || (*posted)[0].Prior != target || (*posted)[0].Next.Session != admitted.Claim.Holder.Session || locals != 0 || len(f.requests) != 0 {
		t.Fatalf("root was handed %+v (main's local handover %d, stops %d)", *posted, locals, len(f.requests))
	}
}