potentially fix deadlock issues with new queueing system

internal/jobtracker/jobtracker.go

@@ -11,6 +11,7 @@ type JobTracker struct {
 	activeWorkers int32
 	queuedJobs    int32
 	didWork       bool
+	stop          bool
 	cond          *sync.Cond
 	Queue         chan string
 	send          chan string
@@ -31,9 +32,6 @@ func NewJobTracker() *JobTracker {
 }
 
 func (jt *JobTracker) manageQueue() {
-	defer close(jt.Queue)
-	defer close(jt.send)
-
 	queue := list.New()
 	for jt.HasWork() {
 		if front := queue.Front(); front == nil {
@@ -52,6 +50,9 @@ func (jt *JobTracker) manageQueue() {
 			}
 		}
 	}
+
+	close(jt.Queue)
+	close(jt.send)
 }
 
 func (jt *JobTracker) AddJob(job string) {
@@ -71,6 +72,9 @@ func (jt *JobTracker) EndWork() {
 }
 
 func (jt *JobTracker) HasWork() bool {
+	if jt.stop {
+		return false
+	}
 	// TODO: didWork is a somewhat ugly workaround to ensure we dont exit before doing work at least once,
 	// this will however result in locking up if we create a JobTracker but never queue any jobs
 	hasWork := !jt.didWork || (atomic.LoadInt32(&jt.queuedJobs) > 0 && atomic.LoadInt32(&jt.activeWorkers) > 0)
@@ -81,11 +85,20 @@ func (jt *JobTracker) HasWork() bool {
 	return hasWork
 }
 
+func (jt *JobTracker) KillIfNoJobs() bool {
+	if atomic.LoadInt32(&jt.queuedJobs) == 0 {
+		jt.stop = true
+		return true
+	}
+	return false
+}
+
 func (jt *JobTracker) QueuedJobs() int32 {
 	return atomic.LoadInt32(&jt.queuedJobs)
 }
 
 func (jt *JobTracker) Wait() {
+	jt.KillIfNoJobs()
 	jt.cond.L.Lock()
 	for jt.HasWork() {
 		jt.cond.Wait()

pkg/goop/clone.go

@@ -169,6 +169,7 @@ func FetchGit(baseUrl, baseDir string) error {
 	}
 	jt.Wait()
 
+	jt = jobtracker.NewJobTracker()
 	log.Info().Str("base", baseUrl).Msg("finding refs")
 	for _, ref := range commonRefs {
 		jt.AddJob(ref)