feat(master): route jobs to agents and reconcile them on reconnect (#7)

internal/master/hub/hub_test.go

@@ -3,6 +3,7 @@ package hub
 import (
 	"bytes"
 	"context"
+	"fmt"
 	"io"
 	"log/slog"
 	"net/http"
@@ -35,6 +36,18 @@ type fakeStore struct {
 	// revokeFromLookup, when > 0, makes AgentByTokenHash return the agent as
 	// revoked from that (1-based) lookup onwards.
 	revokeFromLookup int
+
+	// Job calls, recorded; the *Result fields script the answers.
+	lostCalls    [][]int64
+	lostResult   []int64
+	lostErr      error
+	lostGate     chan struct{} // when set, FailLostJobs blocks until it is closed
+	progress     []string      // "agent|job|chunk"
+	progressOK   bool
+	finishes     []model.JobFinish
+	finishAgents []int64
+	finishOK     bool
+	finishErr    error
 }
 
 func newFakeStore(agents ...model.Agent) *fakeStore {
@@ -94,6 +107,31 @@ func (f *fakeStore) MarkSeen(_ context.Context, id int64, _ time.Time) error {
 	return nil
 }
 
+func (f *fakeStore) FailLostJobs(_ context.Context, _ int64, known []int64, _ time.Time) ([]int64, error) {
+	if f.lostGate != nil {
+		<-f.lostGate
+	}
+	f.mu.Lock()
+	defer f.mu.Unlock()
+	f.lostCalls = append(f.lostCalls, known)
+	return f.lostResult, f.lostErr
+}
+
+func (f *fakeStore) JobProgress(_ context.Context, agentID, jobID int64, chunk string, _ time.Time) (bool, error) {
+	f.mu.Lock()
+	defer f.mu.Unlock()
+	f.progress = append(f.progress, fmt.Sprintf("%d|%d|%s", agentID, jobID, chunk))
+	return f.progressOK, nil
+}
+
+func (f *fakeStore) FinishJob(_ context.Context, agentID, _ int64, fin model.JobFinish) (bool, error) {
+	f.mu.Lock()
+	defer f.mu.Unlock()
+	f.finishAgents = append(f.finishAgents, agentID)
+	f.finishes = append(f.finishes, fin)
+	return f.finishOK, f.finishErr
+}
+
 func (f *fakeStore) count(get func(*fakeStore) int) func() int {
 	return func() int {
 		f.mu.Lock()