Keep nothing for a workspace given siblings, place the siblings where the tests state them, decide the hold from the listed plan keys (review D1-D3)

modules/mesh-delivery-checker/cmd/mesh-delivery-checker/forge.go

@@ -13,6 +13,7 @@ import (
 const (
 	getFileTool = "gitea.gitea_get_file"
 	compareTool = "gitea.gitea_compare"
+	commitsTool = "gitea.gitea_commits"
 )
 
 // forgeReads reads the forge through call (mesh-sdk process.Conn.Call as the checker's own account).
@@ -59,3 +60,21 @@ func (f forgeReads) OnTrunk(_ context.Context, repository, commit string) (bool,
 	}
 	return *a.Total == 0, nil
 }
+
+// FullSHA is the whole sha of a commit named by its first digits: the newest commit of the history at that ref.
+func (f forgeReads) FullSHA(_ context.Context, repository, commit string) (string, error) {
+	owner, repo := ownerRepo(repository)
+	raw, err := f.call(commitsTool, map[string]any{"owner": owner, "repo": repo, "ref": commit, "limit": 1})
+	if err != nil {
+		return "", err
+	}
+	var a struct {
+		Commits []struct {
+			SHA string `json:"sha"`
+		} `json:"commits"`
+	}
+	if err := json.Unmarshal(raw, &a); err != nil || len(a.Commits) == 0 || len(a.Commits[0].SHA) != 40 || !strings.HasPrefix(a.Commits[0].SHA, commit) {
+		return "", fmt.Errorf("the answer of %s for %s is not a commit that starts with it: %v", commitsTool, commit, err)
+	}
+	return a.Commits[0].SHA, nil
+}

modules/mesh-delivery-checker/cmd/mesh-delivery-checker/forge_test.go

@@ -53,3 +53,18 @@ func TestTheForgeReadsParseWhatTheForgeAnswered(t *testing.T) { // 1624/AC4
 		t.Error("an answer with no total was taken")
 	}
 }
+
+func TestTheForgeNamesTheWholeShaOfAShortCommit(t *testing.T) { // review D2
+	f := forgeReads{call: func(key string, body any) (json.RawMessage, error) {
+		if key != commitsTool {
+			t.Errorf("called %s", key)
+		}
+		return capturedAnswer(t, "forge-commits"), nil
+	}}
+	if sha, err := f.FullSHA(context.Background(), "novox/mesh-sdk", "cb54756c3865"); err != nil || sha != "cb54756c3865df5ac9c0dc0481ad9ebe2bf013a1" {
+		t.Errorf("sha = %q, %v", sha, err)
+	}
+	if _, err := f.FullSHA(context.Background(), "novox/mesh-sdk", "deadbeefdead"); err == nil {
+		t.Error("a commit that does not start with the pin was taken")
+	}
+}

modules/mesh-delivery-checker/cmd/mesh-delivery-checker/hold.go

@@ -11,15 +11,18 @@ import (
 	"mesh-delivery-checker/internal/check"
 )
 
-// holdWatch holds the checker back while a delivery builds (ADR 0484 clause 12): it watches the controller's bucket of the
-// delivery plans in their build phase and reads it as empty where it is absent (issue 1652 makes it) or cannot be read,
-// saying so once. The bucket may come into being after the checker started and the checker holds no timer (ADR 0437 §8), so
-// each time a check is about to start and the bucket is not watched yet, the checker looks for it.
+// holdWatch holds the checker back while a delivery builds (ADR 0484 clause 12): whether one does is the list of the keys of
+// the controller's bucket of the delivery plans in their build phase, read each time a check is about to start. Its keys age
+// out (a time to live, and no delete marker), which a watch never hears, so a watch cannot decide it (mesh-controller
+// internal/link/holds.go lists the keys too); the watch is only the wake-up, for a plan that ended by a delete. The bucket
+// is read as empty where it is absent (issue 1652 makes it) or cannot be read, saying so once. It may come into being after
+// the checker started and the checker holds no timer (ADR 0437 §8), so each time a check is about to start and the bucket is
+// not known yet, the checker looks for it.
 type holdWatch struct {
 	ctx     context.Context
 	js      bucketOpener
-	hold    check.PlanHold
-	changed func(ctx context.Context) // the last plan ended: the checks that waited may start
+	keys    func(ctx context.Context) ([]string, error)
+	changed func(ctx context.Context) // a plan key changed: the checks that waited may start
 	logger  *log.Logger
 	errs    chan error
 	mu      sync.Mutex
@@ -34,23 +37,37 @@ func newHoldWatch(ctx context.Context, js bucketOpener, changed func(context.Con
 // Held says whether a delivery builds now.
 func (w *holdWatch) Held() bool {
 	w.ensure()
-	return w.hold.Held()
+	w.mu.Lock()
+	keys := w.keys
+	w.mu.Unlock()
+	if keys == nil {
+		return false
+	}
+	list, err := keys(w.ctx)
+	if err != nil {
+		w.say("the keys of %s cannot be listed, so it is read as empty: %v", check.PlansBuildingBucket, err)
+		return false
+	}
+	return len(list) > 0
 }
 
 func (w *holdWatch) say(format string, args ...any) {
+	w.mu.Lock()
+	defer w.mu.Unlock()
 	if !w.said {
 		w.said = true
 		w.logger.Printf(format, args...)
 	}
 }
 
-// ensure starts the watch if the bucket exists and is not watched.
+// ensure finds the bucket if it is not known and exists, and starts the wake-up watch.
 func (w *holdWatch) ensure() {
 	w.mu.Lock()
-	defer w.mu.Unlock()
-	if w.watched {
+	if w.watched || w.js == nil {
+		w.mu.Unlock()
 		return
 	}
+	w.mu.Unlock()
 	kv, err := w.js.KeyValue(w.ctx, check.PlansBuildingBucket)
 	switch {
 	case errors.Is(err, jetstream.ErrBucketNotFound):
@@ -60,17 +77,25 @@ func (w *holdWatch) ensure() {
 		w.say("the bucket %s cannot be read, so it is read as empty: %v", check.PlansBuildingBucket, err)
 		return
 	}
-	watcher, err := settle(w.ctx, kv, func(e jetstream.KeyValueEntry) { w.hold.Put(e.Key(), !opOf(e)) })
-	if err != nil {
-		w.say("the bucket %s cannot be watched, so it is read as empty: %v", check.PlansBuildingBucket, err)
+	w.mu.Lock()
+	defer w.mu.Unlock()
+	if w.watched {
 		return
 	}
 	w.watched, w.said = true, false
+	w.keys = func(ctx context.Context) ([]string, error) {
+		list, err := kv.Keys(ctx)
+		if errors.Is(err, jetstream.ErrNoKeysFound) {
+			return nil, nil
+		}
+		return list, err
+	}
 	go func() {
-		w.errs <- follow(w.ctx, watcher, func(e jetstream.KeyValueEntry) {
-			if w.hold.Put(e.Key(), !opOf(e)) && !w.hold.Held() {
-				w.changed(w.ctx)
-			}
-		})
+		wake, err := kv.WatchAll(w.ctx)
+		if err != nil {
+			w.errs <- err
+			return
+		}
+		w.errs <- follow(w.ctx, wake, func(jetstream.KeyValueEntry) { w.changed(w.ctx) })
 	}()
 }

modules/mesh-delivery-checker/cmd/mesh-delivery-checker/hold_test.go

@@ -0,0 +1,35 @@
+package main
+
+import (
+	"context"
+	"errors"
+	"io"
+	"log"
+	"testing"
+)
+
+// The controller's plans-building keys age out of the bucket (a time to live, no delete marker), which a watch never hears:
+// whether a delivery builds is the list of the keys now, and the watch is only a wake-up (review D3).
+func TestADeliveryHoldsWhileItsKeyIsListedAndNotWhenItAgedOutWithNoEvent(t *testing.T) {
+	listed := []string{"plan-1"}
+	w := &holdWatch{watched: true, logger: log.New(io.Discard, "", 0), keys: func(context.Context) ([]string, error) { return listed, nil }}
+	if !w.Held() {
+		t.Fatal("a plan key is listed and nothing holds")
+	}
+	listed = nil // the key aged out: no delete reached the watch
+	if w.Held() {
+		t.Error("the key is gone from the bucket and the checker is still held")
+	}
+}
+
+func TestABucketThatCannotBeListedIsReadAsEmpty(t *testing.T) {
+	w := &holdWatch{watched: true, logger: log.New(io.Discard, "", 0), keys: func(context.Context) ([]string, error) { return []string{"plan-1"}, errors.New("the bus is down") }}
+	if w.Held() {
+		t.Error("an unreadable bucket held the checker")
+	}
+	var absent holdWatch
+	absent.logger = log.New(io.Discard, "", 0)
+	if absent.Held() {
+		t.Error("a watch with no bucket held the checker")
+	}
+}

modules/mesh-delivery-checker/cmd/mesh-delivery-checker/judge.go

@@ -31,7 +31,7 @@ type judgeDeps struct {
 
 func newJudge(d judgeDeps) check.Judge {
 	return &check.Examiner{Steps: d.runner, Toolchains: d.tools, Presence: d.presence, Health: d.have, Places: d.places,
-		Kept: check.Kept{Store: d.kept}, Previews: d.previews, Modules: d.repos, Forge: d.forge, Builds: d.builds, Announce: d.announce, Trunks: d.trunks, Logf: d.logf}
+		Kept: check.Kept{Store: d.kept}, Previews: d.previews, Modules: d.repos, Forge: d.forge, Runs: d.repos, Builds: d.builds, Announce: d.announce, Trunks: d.trunks, Logf: d.logf}
 }
 
 // manifestKV reads a manifest by its digest.
@@ -49,6 +49,7 @@ type repoModules struct {
 	mu     sync.Mutex
 	langs  map[string]map[string]bool // by manifest digest: the languages of its artifacts
 	repoOf map[string]string          // by module: its repository
+	commit map[string]string          // by module: the commit its record was computed at (the commit the mesh runs)
 	typed  map[string]string          // by module: its manifest digest
 }
 
@@ -56,11 +57,12 @@ type repoModules struct {
 func (r *repoModules) Put(ctx context.Context, key string, value []byte, deleted bool) {
 	r.mu.Lock()
 	if r.repoOf == nil {
-		r.repoOf, r.typed, r.langs = map[string]string{}, map[string]string{}, map[string]map[string]bool{}
+		r.repoOf, r.typed, r.langs, r.commit = map[string]string{}, map[string]string{}, map[string]map[string]bool{}, map[string]string{}
 	}
 	if deleted {
 		delete(r.repoOf, key)
 		delete(r.typed, key)
+		delete(r.commit, key)
 		r.mu.Unlock()
 		return
 	}
@@ -72,6 +74,7 @@ func (r *repoModules) Put(ctx context.Context, key string, value []byte, deleted
 	r.mu.Lock()
 	r.repoOf[key] = m.Repository
 	r.typed[key] = m.Manifest
+	r.commit[key] = m.Commit
 	_, known := r.langs[m.Manifest]
 	r.mu.Unlock()
 	if m.Manifest == "" || known {
@@ -128,3 +131,10 @@ func (r buildReader) Build(ctx context.Context, module, fp2 string) (delivery.Bu
 	b, err := delivery.DecodeBuilderBuild(value)
 	return b, err == nil, err
 }
+
+// CommitOf is the commit the indexer's record of the module was computed at: the commit of its repository the mesh runs.
+func (r *repoModules) CommitOf(module string) string {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	return r.commit[module]
+}

modules/mesh-delivery-checker/cmd/mesh-delivery-checker/judge_test.go

@@ -66,6 +66,9 @@ func TestARepositoryWithATypeScriptModuleIsJudgedByTheNodeBuildModuleToo(t *test
 	if got := r.BuildModules("novox/mesh-catalog"); strings.Join(got, ",") != buildrun.GoBuildModule+","+buildrun.NodeBuildModule {
 		t.Errorf("mesh-catalog: %v, want both", got)
 	}
+	if got := r.CommitOf("mesh-delivery-checker"); got != strings.Repeat("a", 40) || r.CommitOf("gone") != "" {
+		t.Errorf("the commit the mesh runs = %q", got)
+	}
 	if got := r.BuildModules("novox/mesh-sdk"); strings.Join(got, ",") != buildrun.GoBuildModule {
 		t.Errorf("a repository with no module: %v, want the Go build module", got)
 	}