Author SHA1 Message Date
jschoubben a5b763f8a2 Merge remote-tracking branch 'origin/main' into feat/try-speed 2026-10-25 03:17:48 +01:00
jschoubben 50a2e4c4e2 mesh-delivery-checker: let a try asked by a bare commit take a mesh-sdk pin that is the head of a mesh-sdk branch, naming it, as a try asked by branch does (issue 1921, issue 1624, try speed S12) 2026-10-25 02:03:26 +02:00
jschoubben 4169faf3f0 mesh-delivery-checker: report a lint finding as the lint step's failure (ADR 0542 §2); pin mesh-sdk's try speed-up head with main in it in the checker and the indexer (issue 1921) 2026-10-25 01:31:53 +02:00
jschoubben 103eec8855 Merge remote-tracking branch 'origin/main' into feat/try-speed 2026-10-25 01:29:44 +02:00
jschoubben 8a62f23afe mesh-delivery-indexer: pin mesh-sdk at the try speed-up's head, beside the checker (issue 1921) 2026-10-25 01:27:18 +02:00
jschoubben 84395420ee mesh-delivery-checker: take an ask's longest level off its time for the clone, as the levels now run side by side, and pin mesh-sdk's side-by-side levels (issue 1921, try speed S10) 2026-10-25 01:24:37 +02:00
jschoubben 9740842af7 mesh-delivery-checker: stop a check whose heads its group no longer holds and end it failed, superseded; judge a group's open members by the verdict it had once others merged, so their merge is not refused (issue 1921, issue 1791, try speed S9) 2026-10-25 01:20:29 +02:00
jschoubben 588cfd54bb mesh-delivery-indexer: keep a merged member in its group's candidate until every member merged, so the heads the Approve covers hold and no new check starts for the members left (issue 1921, issue 1791, try speed S9) 2026-10-25 01:18:20 +02:00
jschoubben 4ccd78c9ec mesh-delivery-checker: put every member's check ask of a group at once and judge the answers in member order, so a group's members run side by side (issue 1921, try speed S8) 2026-10-25 01:15:28 +02:00
jschoubben 6fe7b7dd6c Name every failed test in a check's verdict and a try's failed list (issue 1921)
The checker stopped reading a level's test results at the first that did
not pass and kept only its first failure, so a check named one fault per
round. It now reads and counts every result and writes every failed test
with its first line into the verdict's Failures and the try's failed
list (at most 50, then how many more); Failed keeps the first one's
words whole (issue 1901). Pins mesh-sdk fix/1921-every-failure.
2026-10-24 23:44:46 +02:00
42 changed files with 826 additions and 2781 deletions
@@ -281,12 +281,9 @@ func (b *bed) get(t *testing.T, bucket, key string) []byte {
return nil
}
// 1297/B4 on the bus, 1906/D5: the controller sends ace, g14 and shanks builds while the canary ace does not run F healthy yet.
// Started whole, the assigner wants F on the canary and writes no difference at all: there is no send to compare with (ADR 0539
// §8). 1945/AC19, AC21: shanks, which runs the older fingerprint (composed and granted) with no want, gets its first want at that
// fingerprint, never at F; g14, which runs a build of the old form, gets none, so web is on neither list there and the assigner
// raises the condition naming g14, web and the canary rule; once ace runs F healthy, g14's first want names F and the condition is
// cleared.
// 1297/B4 on the bus, 1906/D5: the controller sends ace and g14 builds while the canary ace does not run F healthy yet.
// Started whole, the assigner wants F on the canary, holds g14 back with no want, and writes no difference at all: there
// is no send to compare with (ADR 0539 §8).
func TestWantsAreWrittenAndNoDifferenceIs(t *testing.T) {
b := startBed(t, "")
fNew, fOld := delivery.SourcePrefix2+strings.Repeat("b", 64), delivery.SourcePrefix2+strings.Repeat("a", 64)
@@ -299,21 +296,19 @@ func TestWantsAreWrittenAndNoDifferenceIs(t *testing.T) {
}
bkey, _ := delivery.BuilderBuildKey("web", fNew)
b.putJSON(t, delivery.BuilderBuildsBucket, bkey, built)
for _, fp := range []string{fNew, fOld} {
gkey, _ := delivery.ControllerGrantKey("web", fp)
grant, _ := delivery.EncodeControllerGrant(delivery.ControllerGrant{})
b.putJSON(t, delivery.ControllerGrantsBucket, gkey, grant)
}
runs := map[string]string{"ace": fOld, "g14": "build-9", "shanks": fOld}
for _, node := range []string{"ace", "g14", "shanks"} {
b.compose(t, node, "web", fNew, 'c', delivery.JudgedNeither)
gkey, _ := delivery.ControllerGrantKey("web", fNew)
grant, _ := delivery.EncodeControllerGrant(delivery.ControllerGrant{})
b.putJSON(t, delivery.ControllerGrantsBucket, gkey, grant)
b.compose(t, "ace", "web", fNew, 'c', delivery.JudgedNeither)
b.compose(t, "g14", "web", fNew, 'c', delivery.JudgedNeither)
for _, node := range []string{"ace", "g14"} {
on, _ := delivery.EncodePresence(delivery.Presence{Online: true})
b.putJSON(t, delivery.PresenceBucket, node, on)
have, _ := delivery.EncodeHave(delivery.Have{FP: runs[node], Phase: delivery.PhaseRunning, Health: delivery.HealthOK})
have, _ := delivery.EncodeHave(delivery.Have{FP: fOld, Phase: delivery.PhaseRunning, Health: delivery.HealthOK})
b.putJSON(t, delivery.HaveBucket, node+".web", have)
b.putJSON(t, provider.AssignmentsBucket, node+".web", []byte(`{"node":"`+node+`","module":"web","build":"b-`+node+`"}`))
}
b.compose(t, "shanks", "web", fOld, 'a', delivery.JudgedNeither)
b.putJSON(t, provider.AssignmentsBucket, "ace.web", []byte(`{"node":"ace","module":"web","build":"b-8"}`))
b.putJSON(t, provider.AssignmentsBucket, "g14.web", []byte(`{"node":"g14","module":"web","build":"b-9"}`))
b.start(t)
wkey, _ := delivery.WantKey("ace", "web")
@@ -321,41 +316,12 @@ func TestWantsAreWrittenAndNoDifferenceIs(t *testing.T) {
if err != nil || want.FP != fNew {
t.Fatalf("the canary ace's want = %+v, %v", want, err)
}
if w := b.wantNow(t, "shanks", "web", "shanks' first want at what it runs", func(w delivery.Want) bool { return w.FP != "" }); w.FP != fOld || w.Composed != "sha256:"+strings.Repeat("a", 64) || w.Newest != fNew {
t.Errorf("shanks' want = %+v, want the fingerprint it runs naming its composition, held back from F", w)
}
var neither asks.Condition
for deadline := time.Now().Add(10 * time.Second); time.Now().Before(deadline) && neither.Key == ""; time.Sleep(100 * time.Millisecond) {
for k, c := range b.conditionsSaid(t) {
if strings.Contains(k, ".neither.") && c.State == asks.StateRaised {
neither = c
}
}
}
if neither.Subject.Machine != "g14" || neither.Subject.ID != "web" || neither.Severity != asks.SeverityUrgent || !strings.Contains(neither.Explanation, "canary") {
t.Errorf("the condition of web on neither list on g14 = %+v", neither)
}
time.Sleep(500 * time.Millisecond)
if keys, err := b.buckets[delivery.AssignerDifferencesBucket].Keys(b.ctx); err == nil && len(keys) != 0 {
t.Errorf("differences were written: %v", keys)
}
if e, err := b.buckets[delivery.AssignerWantsBucket].Get(b.ctx, "want.g14.web"); err == nil {
t.Errorf("g14, which runs an old-form build and is held back by the canary, was given a want: %s", e.Value())
}
// 1945/AC21: once the canary ace runs F healthy, g14's first want names F by the rules as usual, and web leaves neither list
// there by itself: the condition is cleared.
healthyF, _ := delivery.EncodeHave(delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthOK})
b.putJSON(t, delivery.HaveBucket, "ace.web", healthyF)
if w := b.wantNow(t, "g14", "web", "g14's first want at F once the canary runs it healthy", func(w delivery.Want) bool { return w.FP != "" }); w.FP != fNew || w.Composed != "sha256:"+strings.Repeat("c", 64) || w.Newest != "" {
t.Errorf("g14's want = %+v, want F with its composition", w)
}
cleared := false
for deadline := time.Now().Add(10 * time.Second); time.Now().Before(deadline) && !cleared; time.Sleep(100 * time.Millisecond) {
c, ok := b.conditionsSaid(t)[neither.Key]
cleared = ok && c.State == asks.StateCleared
}
if !cleared {
t.Errorf("the condition of web on neither list on g14 was not cleared once g14 got its want")
t.Errorf("g14, held back by the canary with no want before, was given one: %s", e.Value())
}
}
@@ -375,14 +341,12 @@ func TestAHeldFingerprintIsAskedAbout(t *testing.T) {
b.putJSON(t, provider.AssignmentsBucket, "ace.web", []byte(`{"node":"ace","module":"web","build":"b-1"}`))
b.start(t)
hkey, _ := delivery.HeldKey("web", fNew)
// 1906/H3: the held record names the machine and keeps the health, when and the node's words.
if h, err := delivery.DecodeHeld(b.get(t, delivery.AssignerWantsBucket, hkey)); err != nil || h.Node != "ace" ||
!strings.HasPrefix(h.Words, "health failed, phase running, seen at ") || !strings.HasSuffix(h.Words, ": connection refused") {
if h, err := delivery.DecodeHeld(b.get(t, delivery.AssignerWantsBucket, hkey)); err != nil || h.Node != "ace" {
t.Fatalf("held = %+v, %v", h, err)
}
akey, _ := delivery.HoldAskKey("web", fNew)
a, err := delivery.DecodeAssignerAsk(akey, b.get(t, delivery.AssignerAsksBucket, akey))
if err != nil || a.Asking <= 0 || a.Ask.ID != delivery.AskingID(akey, a.Asking) || len(a.Ask.Options) != 2 {
if err != nil || a.Ask.ID != delivery.AskingID(akey, 0) || len(a.Ask.Options) != 2 {
t.Fatalf("the ask = %+v, %v", a, err)
}
}
@@ -449,7 +413,7 @@ func TestAFailedDataModuleIsAskedToRestore(t *testing.T) {
key := ""
for deadline := time.Now().Add(10 * time.Second); time.Now().Before(deadline) && key == ""; time.Sleep(100 * time.Millisecond) {
for k, c := range b.conditionsSaid(t) {
if c.State == asks.StateRaised && c.Subject.ID == "mail" && strings.Contains(k, ".held.") {
if c.State == asks.StateRaised && c.Subject.ID == "mail" {
if c.Subject.Machine != "ace" || !strings.Contains(c.Explanation, "connection refused") {
t.Fatalf("the condition = %+v, want it to name ace and the check's words", c)
}
@@ -616,15 +580,8 @@ func (b *bed) wantNow(t *testing.T, node, module, what string, ok func(delivery.
return last
}
// answer says the teller wrote the operator's answer to the ask that stands at key (its warrant names that asking's id).
func (b *bed) answer(t *testing.T, key, option string) {
t.Helper()
a, err := delivery.DecodeAssignerAsk(key, b.get(t, delivery.AssignerAsksBucket, key))
if err != nil {
t.Fatal(err)
}
b.answerTo(t, key, option, a.Ask.ID)
}
// answer says the teller wrote the operator's answer to the ask at key.
func (b *bed) answer(t *testing.T, key, option string) { b.answerTo(t, key, option, "r1") }
// answerTo is answer to the ask with this id (the router's warrant names it).
func (b *bed) answerTo(t *testing.T, key, option, askID string) {
@@ -668,8 +625,8 @@ func TestAReleaseAnswerWritesTheWantAgain(t *testing.T) {
}
for deadline := time.Now().Add(10 * time.Second); time.Now().Before(deadline); time.Sleep(100 * time.Millisecond) {
cleared := false
for k, c := range b.conditionsSaid(t) {
cleared = cleared || (c.Subject.ID == "web" && c.State == asks.StateCleared && strings.Contains(k, ".held."))
for _, c := range b.conditionsSaid(t) {
cleared = cleared || (c.Subject.ID == "web" && c.State == asks.StateCleared)
}
if cleared {
return
@@ -816,81 +773,3 @@ func TestAUserListRevisionOfTheBusIsWrittenAtOnce(t *testing.T) {
t.Errorf("the user-list revision was asked")
}
}
// 1906/H2, whole on the bus: given a fingerprint held after one failed check, when every node that runs the module then runs
// it healthy, the assigner lifts the hold by itself: the held record and its ask are taken away, the held condition is
// cleared saying why, and the want stays at the fingerprint.
func TestAHoldIsLiftedOnceEveryNodeRunsItHealthy(t *testing.T) {
b := startBed(t, "")
fNew := delivery.SourcePrefix2 + strings.Repeat("b", 64)
b.fixture(t, "web", fNew, `{"module":"web","version":"1"}`)
b.compose(t, "ace", "web", fNew, 'b', delivery.JudgedNeither)
failed, _ := delivery.EncodeHave(delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthFailed, Detail: "connection refused"})
b.putJSON(t, delivery.HaveBucket, "ace.web", failed)
b.putWant(t, "ace", "web", delivery.Want{FP: fNew, Composed: "sha256:" + strings.Repeat("b", 64)})
b.start(t)
hkey, _ := delivery.HeldKey("web", fNew)
akey, _ := delivery.HoldAskKey("web", fNew)
b.get(t, delivery.AssignerWantsBucket, hkey)
b.get(t, delivery.AssignerAsksBucket, akey)
healthy, _ := delivery.EncodeHave(delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthOK})
b.putJSON(t, delivery.HaveBucket, "ace.web", healthy)
for deadline := time.Now().Add(15 * time.Second); time.Now().Before(deadline); time.Sleep(100 * time.Millisecond) {
if b.hasKey(delivery.AssignerWantsBucket, hkey) || b.hasKey(delivery.AssignerAsksBucket, akey) {
continue
}
for k, c := range b.conditionsSaid(t) {
if c.Subject.ID == "web" && c.State == asks.StateCleared && strings.Contains(k, ".held.") && strings.Contains(c.Resolved, "lifted its hold by itself") {
if w := b.wantNow(t, "ace", "web", "the want at the lifted fingerprint", func(w delivery.Want) bool { return w.FP == fNew }); w.Rule != 0 {
t.Errorf("the want = %+v", w)
}
return
}
}
}
t.Fatalf("the hold was not lifted: held %v, ask %v", b.hasKey(delivery.AssignerWantsBucket, hkey), b.hasKey(delivery.AssignerAsksBucket, akey))
}
// 1906/H2, whole on the bus: a hold lifted by itself and written again after the assigner was started again is asked under a
// new id, so the teller raises it; the answer to the first hold does not carry over (1906/H4).
func TestAHoldWrittenAgainAfterALiftAndARestartIsAskedAgain(t *testing.T) {
b := startBed(t, "")
fNew := delivery.SourcePrefix2 + strings.Repeat("b", 64)
b.fixture(t, "web", fNew, `{"module":"web","version":"1"}`)
b.compose(t, "ace", "web", fNew, 'b', delivery.JudgedNeither)
failed, _ := delivery.EncodeHave(delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthFailed, Detail: "connection refused"})
healthy, _ := delivery.EncodeHave(delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthOK})
b.putJSON(t, delivery.HaveBucket, "ace.web", failed)
b.putWant(t, "ace", "web", delivery.Want{FP: fNew, Composed: "sha256:" + strings.Repeat("b", 64)})
_, stop := b.startStoppable(t)
hkey, _ := delivery.HeldKey("web", fNew)
akey, _ := delivery.HoldAskKey("web", fNew)
first, err := delivery.DecodeAssignerAsk(akey, b.get(t, delivery.AssignerAsksBucket, akey))
if err != nil {
t.Fatal(err)
}
b.putJSON(t, delivery.HaveBucket, "ace.web", healthy)
for deadline := time.Now().Add(15 * time.Second); b.hasKey(delivery.AssignerWantsBucket, hkey); time.Sleep(100 * time.Millisecond) {
if time.Now().After(deadline) {
t.Fatal("the hold was not lifted")
}
}
stop()
time.Sleep(1100 * time.Millisecond) // the next hold is written in another second
b.answerTo(t, akey, "keep", first.Ask.ID) // an answer to the first hold, which the later hold must not take
b.putJSON(t, delivery.HaveBucket, "ace.web", failed)
b.start(t)
again, err := delivery.DecodeAssignerAsk(akey, b.get(t, delivery.AssignerAsksBucket, akey))
if err != nil {
t.Fatal(err)
}
if again.Ask.ID == first.Ask.ID {
t.Fatalf("the hold written again is asked under the first id %s", first.Ask.ID)
}
b.putJSON(t, delivery.HaveBucket, "ace.web", healthy)
for deadline := time.Now().Add(15 * time.Second); b.hasKey(delivery.AssignerWantsBucket, hkey); time.Sleep(100 * time.Millisecond) {
if time.Now().After(deadline) {
t.Fatal("the stale \"keep\" of the first hold kept the second one held")
}
}
}
@@ -28,18 +28,6 @@ type passer struct {
// answeredAt is, by the id of a backup ask, how many values the have had when its answer was written into the want.
answeredAt map[string]uint64
asks *composeAsks
// now is the clock; nil is time.Now.
now func() time.Time
// neither is, by node and module, each module on neither list whose condition this process raised (issue 1945). Nil until the
// first pass after a start, which clears what a previous process may have raised (startClears).
neither map[string]neitherOne
}
func (p *passer) clock() time.Time {
if p.now != nil {
return p.now()
}
return time.Now()
}
func (p *passer) pass(ctx context.Context) {
@@ -97,23 +85,11 @@ func (p *passer) wants(ctx context.Context) {
for name := range wanted {
names[name] = true
}
neither := map[string]neitherOne{}
// The first pass after a start remembers nothing a previous process said: a neither or held condition it raised would stay
// raised for ever once the module has its want or is no longer held. That pass clears, once, the neither condition of every
// assigned module not on neither list and the held condition of every module not held; a clear of a condition that is not open is
// quiet at the router.
first := p.neither == nil
stale := map[string]neitherOne{}
defer func() { p.sayNeither(neither, stale) }()
for name := range names {
fp := trunk[name].FP2
name := name
m := assign.Module{Name: name, FP: fp, Manifest: manifests[name], Nodes: nodesOf[name], Now: time.Now(),
m := assign.Module{Name: name, FP: fp, Manifest: manifests[name], Nodes: nodesOf[name],
Composed: func(node, fp string) (delivery.ComposedDigest, bool) { return p.w.composed(node, name, fp) },
GrantedAt: func(fp string) bool {
key, err := delivery.ControllerGrantKey(name, fp)
return err == nil && p.w.has(delivery.ControllerGrantsBucket, key)
},
Restored: func(node string) bool {
key, err := delivery.RestoreAskKey(node, name, fp)
return err == nil && p.answeredWith(key, assign.OptionRestore)
@@ -127,13 +103,7 @@ func (p *passer) wants(ctx context.Context) {
grant, _ := delivery.ControllerGrantKey(name, fp)
hold, _ := delivery.HoldAskKey(name, fp)
m.Released = p.answeredWith(hold, "release")
m.Kept = p.answeredWith(hold, "keep")
m.Built, m.Held, m.Granted = p.w.built(name, fp), p.w.has(delivery.AssignerWantsBucket, held) && !m.Released, p.w.has(delivery.ControllerGrantsBucket, grant)
if raw := p.w.value(delivery.ControllerGrantsBucket, grant); raw != nil {
if g, err := delivery.DecodeControllerGrant(raw); err == nil {
m.GrantedOn = func(node string) bool { return assign.GrantNamesNode(g, node, name) }
}
}
if m.Released {
p.clearHeld(name, fp)
}
@@ -143,13 +113,6 @@ func (p *passer) wants(ctx context.Context) {
last[node] = p.w.lastHealthy(node, name, fp)
}
out := assign.Wants(m, nodes, wanted[name], last)
gatherNeither(neither, name, out)
if first {
startClears(stale, m, out)
if fp != "" && !m.Held && out.Held == nil && !out.Lifted {
p.clearHeld(name, fp)
}
}
if out.Held != nil {
key, err := delivery.HeldKey(name, fp)
if err == nil {
@@ -157,11 +120,8 @@ func (p *passer) wants(ctx context.Context) {
p.put(ctx, delivery.AssignerWantsBucket, key, func() ([]byte, error) { return delivery.EncodeHeld(held) })
}
}
if out.Lifted {
p.lift(ctx, name, fp)
}
if out.Held != nil || (m.Held && !out.Lifted) {
p.holdAsk(ctx, name, fp, out.Held)
if out.Held != nil || m.Held {
p.holdAsk(ctx, name, fp)
p.heldCondition(ctx, name, fp, out.Held, len(out.Restore) > 0)
}
for _, node := range out.Restore {
@@ -182,74 +142,6 @@ func (p *passer) wants(ctx context.Context) {
}
}
// neitherOne is a module assigned to a node that holds no want for it after a pass, and the rule holding its first want back.
type neitherOne struct {
node, module string
holding assign.Holding
}
// gatherNeither adds the nodes on which module is on neither list in its outcome, keyed by node and module.
func gatherNeither(into map[string]neitherOne, module string, out assign.Outcome) {
for node, h := range out.Neither {
into[node+"."+module] = neitherOne{node: node, module: module, holding: h}
}
}
// startClears adds, for the first pass after a start, every node of m that is not on neither list after the pass: a neither
// condition a previous process raised for it is cleared. That covers a node with a want and one with none that is not on neither
// list, such as a held module (issue 1945: the hold is said on its own, so a neither raised for it before the restart would stay up
// for ever) or a move waiting for its record ask. Clearing a condition that is not open is quiet at the router.
func startClears(into map[string]neitherOne, m assign.Module, out assign.Outcome) {
for _, node := range m.Nodes {
if _, on := out.Neither[node]; !on {
into[node+"."+m.Name] = neitherOne{node: node, module: m.Name}
}
}
}
// sayNeither raises, once, the condition of every module on neither list in this pass and writes it to the journal, and clears,
// once, the condition of each one said before that has a want again (issue 1945). A say that fails is made again at the next pass.
// stale is, on the first pass after a start, the modules whose condition a previous process may have raised and that have a want:
// each is cleared once.
func (p *passer) sayNeither(now, stale map[string]neitherOne) {
if p.neither == nil {
p.neither = map[string]neitherOne{}
for key, one := range stale {
if _, on := now[key]; !on {
p.neither[key] = one
}
}
}
for key, one := range now {
if _, said := p.neither[key]; said {
continue
}
c := assign.NeitherCondition(one.node, one.module, one.holding, true)
if p.say != nil {
if err := p.say(c); err != nil {
p.logf("the condition of %s on neither list on %s could not be said, and is said at the next pass: %v", one.module, one.node, err)
continue
}
}
p.logf("%s is assigned to %s and on neither list: no want, and not in the node's own composition; rule %d holds its first want back: %s",
one.module, one.node, one.holding.Rule, one.holding.Words)
p.neither[key] = one
}
for key, one := range p.neither {
if _, still := now[key]; still {
continue
}
if p.say != nil {
if err := p.say(assign.NeitherCondition(one.node, one.module, assign.Holding{}, false)); err != nil {
p.logf("the condition of %s on neither list on %s could not be cleared, and is cleared at the next pass: %v", one.module, one.node, err)
continue
}
}
p.logf("%s on %s has a want again", one.module, one.node)
delete(p.neither, key)
}
}
// compositions writes the composition check's verdict of every group the indexer holds a candidate for, under the
// checker's key of that check: waits, with why, while the indexer holds what it composes from only in part.
func (p *passer) compositions(ctx context.Context) {
@@ -318,62 +210,15 @@ func (p *passer) put(ctx context.Context, bucket, key string, encode func() ([]b
// holdAsk writes, once, the ask about a held fingerprint (ADR 0528 clause 6, ADR 0437 clause 10): the first ask that needs
// no proof of who answered, which the view may answer and the teller fills. In release 5 an answer changes nothing on a
// node (wants are only compared, ADR 0429 §7), and the options say so.
func (p *passer) holdAsk(ctx context.Context, module, fp string, fresh *delivery.Held) {
func (p *passer) holdAsk(ctx context.Context, module, fp string) {
key, err := delivery.HoldAskKey(module, fp)
if err != nil || p.w.has(delivery.AssignerAsksBucket, key) {
return
}
// The asking is the hold's time in seconds: a hold of the same fingerprint written again after a lift is another asking, under
// an id of its own, also after a restart, since the router drops a second ask under an id it keeps (issue 1906).
now := p.clock()
asking := int(now.Unix())
h := p.heldHold(module, fp, fresh)
a := assign.HoldAsk(key, module, fp, h.Node, h.Words, asking, now)
a := assign.HoldAsk(key, module, fp, time.Now())
p.put(ctx, delivery.AssignerAsksBucket, key, func() ([]byte, error) { return delivery.EncodeAssignerAsk(key, a) })
}
// lift takes away a hold the assigner lifted by itself, since every node that runs the module runs the held fingerprint healthy
// (issue 1906): the held record and its ask (the teller then takes the ask back from the operator), and the held condition is
// cleared saying why, once. A hold of the same fingerprint written later is said and asked again.
func (p *passer) lift(ctx context.Context, module, fp string) {
held, err := delivery.HeldKey(module, fp)
if err != nil {
return
}
ask, err := delivery.HoldAskKey(module, fp)
if err != nil {
return
}
c := assign.LiftedCondition(module, fp)
if !p.said[c.Key+".lifted"] {
if p.say != nil {
if err := p.say(c); err != nil {
// The hold stays written, so the next pass lifts it again and says so then.
p.logf("the condition of the lifted hold of %s could not be cleared, and is cleared at the next pass: %v", module, err)
return
}
}
p.logf("%s", assign.LiftedWords(module, fp))
p.said[c.Key+".lifted"] = true
delete(p.said, c.Key+".raised") // a hold written again is said again
}
p.remove(ctx, delivery.AssignerWantsBucket, held)
p.remove(ctx, delivery.AssignerAsksBucket, ask)
}
// remove deletes one key of the assigner's own bucket when it holds it; a delete that fails is said and made again at the next
// pass.
func (p *passer) remove(ctx context.Context, bucket, key string) {
if !p.w.has(bucket, key) {
return
}
if err := p.kvs[bucket].Delete(ctx, key); err != nil && !errors.Is(err, context.Canceled) {
p.logf("%s %s could not be taken away: %v", bucket, key, err)
return
}
p.w.put(bucket, key, nil)
}
// heldHold is the hold of fp as written: the node whose health check failed and its words.
func (p *passer) heldHold(module, fp string, fresh *delivery.Held) delivery.Held {
if fresh != nil {
@@ -400,8 +245,6 @@ func (p *passer) heldCondition(ctx context.Context, module, fp string, fresh *de
return
}
p.said[c.Key+".raised"] = true
delete(p.said, c.Key+".lifted") // a hold raised again and lifted again is said lifted again
delete(p.said, c.Key+".cleared") // and released again is cleared again
}
// restoreAsk writes, once, the ask about a module holding data whose held fingerprint failed its health check on node.
@@ -415,21 +258,11 @@ func (p *passer) restoreAsk(ctx context.Context, module, node, fp string) {
p.put(ctx, delivery.AssignerAsksBucket, key, func() ([]byte, error) { return delivery.EncodeAssignerAsk(key, a) })
}
// answeredWith says whether the teller wrote an answer to the ask at key as it stands now and the option chosen is option. The
// answer is the operator's, given in the view or on the phone, and stays under its key (ADR 0539 §14, §16, §19); it counts only
// for the asking it answered (its warrant names that asking's id), so an answer to an ask taken away, or to an earlier hold of
// the same fingerprint, does not carry over to a later one (issue 1906).
// answeredWith says whether the teller wrote an answer to the ask at key and the option chosen is option. The answer is the
// operator's, given in the view or on the phone, and stays under its key (ADR 0539 §14, §16, §19).
func (p *passer) answeredWith(key, option string) bool {
a, ok := p.w.answer(key)
if !ok || a.Option != option {
return false
}
ask, err := delivery.DecodeAssignerAsk(key, p.w.value(delivery.AssignerAsksBucket, key))
if err != nil {
return false
}
id, _, _ := strings.Cut(a.Warrant, "@")
return id == ask.Ask.ID
return ok && a.Option == option
}
// clearHeld clears, once, the condition of a hold the operator released.
@@ -1,238 +0,0 @@
package main
import (
"context"
"errors"
"fmt"
"strings"
"testing"
"time"
"github.com/nats-io/nats.go/jetstream"
"git.novox.be/novox/mesh-sdk/go/asks"
"git.novox.be/novox/mesh-sdk/go/delivery"
"git.novox.be/novox/mesh-sdk/go/provider"
"mesh-delivery-assigner/internal/assign"
)
// memKV is the part of a bucket a pass writes: puts and deletes.
type memKV struct {
jetstream.KeyValue
values map[string][]byte
}
func (m *memKV) Put(_ context.Context, key string, value []byte) (uint64, error) {
m.values[key] = value
return uint64(len(m.values)), nil
}
func (m *memKV) Delete(_ context.Context, key string, _ ...jetstream.KVDeleteOpt) error {
delete(m.values, key)
return nil
}
// newPasser is a pass on a fresh world, as after a start, writing to the buckets given, at the clock given.
func newPasser(kvs buckets, at time.Time) *passer {
return &passer{kvs: kvs, w: newWorld(), logf: func(string, ...any) {}, said: map[string]bool{}, answeredAt: map[string]uint64{},
now: func() time.Time { return at }}
}
var fpTest = delivery.SourcePrefix2 + strings.Repeat("b", 64)
// 1906/H2: given a hold lifted by itself, when the assigner is started again and the same fingerprint is held again, then the
// hold ask is written under another asking, with an id of its own, so the teller raises it (the router drops a second ask under
// an id it keeps).
func TestAHoldAskedAgainAfterALiftAndARestartHasANewID(t *testing.T) {
asksKV, wantsKV := &memKV{values: map[string][]byte{}}, &memKV{values: map[string][]byte{}}
kvs := buckets{delivery.AssignerAsksBucket: asksKV, delivery.AssignerWantsBucket: wantsKV}
key, _ := delivery.HoldAskKey("web", fpTest)
held := &delivery.Held{Node: "ace", Words: "health failed"}
first := newPasser(kvs, time.Date(2026, 10, 25, 1, 0, 0, 0, time.UTC))
first.holdAsk(context.Background(), "web", fpTest, held)
a1, err := delivery.DecodeAssignerAsk(key, asksKV.values[key])
if err != nil {
t.Fatal(err)
}
first.lift(context.Background(), "web", fpTest)
if _, there := asksKV.values[key]; there {
t.Fatal("the lift left the hold ask")
}
again := newPasser(kvs, time.Date(2026, 10, 25, 2, 0, 0, 0, time.UTC)) // started again: it remembers nothing
again.holdAsk(context.Background(), "web", fpTest, held)
a2, err := delivery.DecodeAssignerAsk(key, asksKV.values[key])
if err != nil {
t.Fatal(err)
}
if a2.Ask.ID == a1.Ask.ID {
t.Errorf("the ask written again after a lift and a restart has the id %s of the first", a1.Ask.ID)
}
}
// 1906/H4: an answer counts only for the asking it answered: an answer to an earlier hold of the same fingerprint, or to an ask
// taken away, is not taken for the hold that stands now.
func TestAnAnswerCountsOnlyForTheAskingItAnswered(t *testing.T) {
key, _ := delivery.HoldAskKey("web", fpTest)
p := newPasser(buckets{}, time.Now())
put := func(asking int) string {
a := assign.HoldAsk(key, "web", fpTest, "ace", "health failed", asking, time.Now())
v, err := delivery.EncodeAssignerAsk(key, a)
if err != nil {
t.Fatal(err)
}
p.w.put(delivery.AssignerAsksBucket, key, v)
return a.Ask.ID
}
first := put(100)
v, _ := delivery.EncodeAnswer(delivery.Answer{Option: "release", Warrant: first + "@2026-10-25T01:00:00Z", Words: "the operator"})
p.w.put(delivery.AnswersBucket, key, v)
if !p.answeredWith(key, "release") {
t.Fatal("the answer to the ask that stands is not taken")
}
if p.answeredWith(key, "keep") {
t.Error("an answer was taken for an option not chosen")
}
put(200) // the same fingerprint held again, after a lift
if p.answeredWith(key, "release") {
t.Error("the answer to an earlier hold carried over to a later one")
}
p.w.put(delivery.AssignerAsksBucket, key, nil) // the ask taken away
if p.answeredWith(key, "release") {
t.Error("an answer to an ask taken away was taken")
}
}
// 1945/AC21: given a module on neither list on a node, when passes follow, then the assigner raises its condition once, says it
// again after a say that failed, and clears it once when the module has a want again; a module on neither list is also written to
// the journal, so it is loud with no operator channel.
func TestAModuleOnNeitherListIsRaisedOnceAndClearedOnce(t *testing.T) {
p := newPasser(buckets{}, time.Date(2026, 10, 25, 15, 0, 0, 0, time.UTC))
var said []asks.Condition
failing := true
p.say = func(c asks.Condition) error {
if failing {
failing = false
return errors.New("the channel is away")
}
said = append(said, c)
return nil
}
var logged []string
p.logf = func(f string, a ...any) { logged = append(logged, fmt.Sprintf(f, a...)) }
one := map[string]neitherOne{"g14.screen-lock": {node: "g14", module: "screen-lock",
holding: assign.Holding{Rule: delivery.RuleProviders, Words: "no provider of x11-display runs healthy on g14 or at its canary"}}}
p.sayNeither(one, nil) // the say fails
p.sayNeither(one, nil)
p.sayNeither(one, nil)
if len(said) != 1 || said[0].State != asks.StateRaised || said[0].Subject.Machine != "g14" || said[0].Subject.ID != "screen-lock" {
t.Fatalf("said = %+v, want one raised condition of screen-lock on g14", said)
}
if !strings.Contains(strings.Join(logged, "\n"), "screen-lock") {
t.Errorf("the journal does not name the module on neither list: %q", logged)
}
p.sayNeither(map[string]neitherOne{}, nil)
p.sayNeither(map[string]neitherOne{}, nil)
if len(said) != 2 || said[1].State != asks.StateCleared || said[1].Key != said[0].Key {
t.Errorf("said = %+v, want the condition cleared once under its key", said)
}
}
// 1945/AC21: the pass gathers the modules on neither list of every module's outcome, keyed by node and module.
func TestThePassGathersTheModulesOnNeitherList(t *testing.T) {
got := map[string]neitherOne{}
gatherNeither(got, "screen-lock", assign.Outcome{Neither: map[string]assign.Holding{"g14": {Rule: delivery.RuleGrant, Words: "w"}}})
gatherNeither(got, "rofi", assign.Outcome{})
if len(got) != 1 || got["g14.screen-lock"].node != "g14" || got["g14.screen-lock"].module != "screen-lock" || got["g14.screen-lock"].holding.Rule != delivery.RuleGrant {
t.Errorf("gathered = %+v", got)
}
}
// A neither or held condition a previous process raised is cleared by the first pass after a start: a fresh pass remembers
// nothing it said, so it clears, once, the neither condition of every assigned module not on neither list and the held condition of
// every module that is not held. Clearing a condition that is not open is quiet at the router.
func TestTheFirstPassAfterAStartClearsWhatAPreviousProcessRaised(t *testing.T) {
wantsKV := &memKV{values: map[string][]byte{}}
p := newPasser(buckets{delivery.AssignerWantsBucket: wantsKV, delivery.AssignerAsksBucket: &memKV{values: map[string][]byte{}}}, time.Now())
var said []asks.Condition
p.say = func(c asks.Condition) error { said = append(said, c); return nil }
mod, _ := delivery.EncodeModule(delivery.Module{FP2: fpTest, Repository: "novox/web", Commit: strings.Repeat("1", 40), Path: delivery.RootPath})
p.w.put(delivery.ModulesBucket, "web", mod)
p.w.put(provider.AssignmentsBucket, "g14.web", []byte(`{"node":"g14","module":"web","build":"b-1"}`))
on, _ := delivery.EncodePresence(delivery.Presence{Online: true})
p.w.put(delivery.PresenceBucket, "g14", on)
have, _ := delivery.EncodeHave(delivery.Have{FP: fpTest, Phase: delivery.PhaseRunning, Health: delivery.HealthOK})
p.w.put(delivery.HaveBucket, "g14.web", have)
key, _ := delivery.WantKey("g14", "web")
want, _ := delivery.EncodeWant(delivery.Want{FP: fpTest, Composed: "sha256:" + strings.Repeat("c", 64)})
p.w.put(delivery.AssignerWantsBucket, key, want)
p.wants(context.Background())
cleared := map[string]bool{}
for _, c := range said {
if c.State != asks.StateCleared {
t.Errorf("the first pass raised %+v", c)
}
cleared[c.Key] = true
}
neither := assign.NeitherCondition("g14", "web", assign.Holding{}, false).Key
held := assign.HeldCondition("web", fpTest, "", "", false, false).Key
if !cleared[neither] || !cleared[held] {
t.Errorf("cleared = %v, want %s and %s", cleared, neither, held)
}
n := len(said)
p.wants(context.Background())
if len(said) != n {
t.Errorf("a second pass said again: %+v", said[n:])
}
}
// 1945/AC24: given a module held at F under rule 2 on a node that runs it with no want, when the assigner passes, then it raises no
// neither condition for it, a neither condition this process raised for it before is cleared, and so is one a previous process
// raised (the first pass after a start), each once.
func TestAHeldModuleRaisesNoNeitherAndClearsAStandingOne(t *testing.T) {
setup := func() (*passer, *[]asks.Condition) {
p := newPasser(buckets{delivery.AssignerWantsBucket: &memKV{values: map[string][]byte{}}, delivery.AssignerAsksBucket: &memKV{values: map[string][]byte{}}}, time.Now())
said := &[]asks.Condition{}
p.say = func(c asks.Condition) error { *said = append(*said, c); return nil }
mod, _ := delivery.EncodeModule(delivery.Module{FP2: fpTest, Repository: "novox/resolved", Commit: strings.Repeat("1", 40), Path: delivery.RootPath})
p.w.put(delivery.ModulesBucket, "systemd-resolved", mod)
p.w.put(provider.AssignmentsBucket, "shanks.systemd-resolved", []byte(`{"node":"shanks","module":"systemd-resolved","build":"b-1"}`))
on, _ := delivery.EncodePresence(delivery.Presence{Online: true})
p.w.put(delivery.PresenceBucket, "shanks", on)
have, _ := delivery.EncodeHave(delivery.Have{FP: fpTest, Phase: delivery.PhaseRunning, Health: delivery.HealthFailed})
p.w.put(delivery.HaveBucket, "shanks.systemd-resolved", have)
hkey, _ := delivery.HeldKey("systemd-resolved", fpTest)
held, _ := delivery.EncodeHeld(delivery.Held{Node: "shanks", Words: "health failed"})
p.w.put(delivery.AssignerWantsBucket, hkey, held)
return p, said
}
key := assign.NeitherCondition("shanks", "systemd-resolved", assign.Holding{}, false).Key
neithers := func(said []asks.Condition) (raised, cleared int) {
for _, c := range said {
if c.Key != key {
continue
}
if c.State == asks.StateRaised {
raised++
} else {
cleared++
}
}
return
}
// a standing neither this process raised
p, said := setup()
p.neither = map[string]neitherOne{"shanks.systemd-resolved": {node: "shanks", module: "systemd-resolved", holding: assign.Holding{Rule: delivery.RuleNotHeld}}}
p.wants(context.Background())
p.wants(context.Background())
if r, c := neithers(*said); r != 0 || c != 1 {
t.Errorf("standing in this process: %d raised, %d cleared, want 0 and 1: %+v", r, c, *said)
}
// one a previous process raised: the first pass after a start clears it
p, said = setup()
p.wants(context.Background())
p.wants(context.Background())
if r, c := neithers(*said); r != 0 || c != 1 {
t.Errorf("after a start: %d raised, %d cleared, want 0 and 1: %+v", r, c, *said)
}
}
@@ -326,13 +326,6 @@ func (w *world) nodes(manifests map[string]catalogue.Manifest, assigned map[stri
return n
}
// value is what a bucket holds at a key now, nil when nothing.
func (w *world) value(bucket, key string) []byte {
w.mu.Lock()
defer w.mu.Unlock()
return w.values[bucket][key]
}
// has says whether a bucket holds a key now.
func (w *world) has(bucket, key string) bool {
w.mu.Lock()
@@ -10,17 +10,14 @@ import (
"git.novox.be/novox/mesh-sdk/go/delivery"
)
// HoldAsk is the ask about a module's fingerprint the assigner holds, at its key (delivery.HoldAskKey): keep it held or
// release it. Releasing writes the held fingerprint's want again by the rules (ADR 0539 §19). node and words are the held
// record's: the machine whose health check failed and the hold's words (issue 1906). asking numbers the writing: 0 the first, a
// larger number an ask written again after a hold lifted by itself, so the router takes it as new.
func HoldAsk(key, module, fp, node, words string, asking int, now time.Time) delivery.AssignerAsk {
return delivery.AssignerAsk{Asking: asking, Ask: asks.Ask{
ID: delivery.AskingID(key, asking),
// HoldAsk is the ask about a module's fingerprint the assigner holds, at its key (delivery.HoldAskKey), first asking: keep
// it held or release it. Releasing writes the held fingerprint's want again by the rules (ADR 0539 §19).
func HoldAsk(key, module, fp string, now time.Time) delivery.AssignerAsk {
return delivery.AssignerAsk{Ask: asks.Ask{
ID: delivery.AskingID(key, 0),
Headline: fmt.Sprintf("%s is held at %s", module, short(fp)),
Explanation: fmt.Sprintf("The health check of %s failed on %s at %s (%s), so the assigner holds that version and wants "+
"no node to move to it. It lifts the hold by itself once every node that runs %s runs it healthy; releasing it "+
"lets the wants move to it now.", module, node, short(fp), words, module),
Explanation: fmt.Sprintf("A node's health check of %s failed at %s, so the assigner holds that version and wants "+
"no node to move to it. Releasing it lets the wants move to it again.", module, short(fp)),
Options: []asks.Option{
{ID: "keep", Label: "Keep it held", Does: "the version stays held", Level: asks.Acknowledge},
{ID: "release", Label: "Release it", Does: "the version is released and the wants move to it by the rules again", Level: asks.Acknowledge},
@@ -81,35 +78,6 @@ func HeldCondition(module, fp, node, words string, raised, needsYou bool) asks.C
return c
}
// NeitherCondition is the condition of a module assigned to node that holds no want for it (issue 1945): the node's own composition
// gives up every module with a fingerprint of the new form, so nothing applies it there and its files are gone. It is raised (urgent)
// naming the module, the node and the rule holding its first want back with the rule's words, under the assigner's own name keyed by
// node and module, and cleared (raised false) when the node has a want for it again.
func NeitherCondition(node, module string, h Holding, raised bool) asks.Condition {
key := delivery.AssignerModule + ".neither." + node + "." + module
subject := asks.Scope{Scope: "module", ID: module, Machine: node}
if !raised {
return asks.Condition{Key: key, Subject: subject, State: asks.StateCleared,
Resolved: fmt.Sprintf("%s on %s has a want again, so its own composition is applied there.", module, node)}
}
headline := fmt.Sprintf("%s on %s is on no list", module, node)
if n := len([]rune(headline)); n > asks.HeadlineLength {
headline = string([]rune(headline)[:asks.HeadlineLength])
}
return asks.Condition{Key: key, Severity: asks.SeverityUrgent, Subject: subject, State: asks.StateRaised,
Summary: headline, Headline: headline,
Explanation: fmt.Sprintf("%s %s is assigned to %s, but the node holds no want for it and the node's own composition does not "+
"hold it either, so nothing applies it there. The assigner's rule %d holds its first want back: %s. The assigner writes the want once the "+
"rule holds, and clears this.", asks.NothingToDo, module, node, h.Rule, h.Words)}
}
// LiftedCondition clears the condition of a held fingerprint the assigner lifted by itself, saying why (issue 1906).
func LiftedCondition(module, fp string) asks.Condition {
c := HeldCondition(module, fp, "", "", false, false)
c.Resolved = LiftedWords(module, fp) + "."
return c
}
// The options of the record ask (ADR 0539 §14) and of the backup ask (§16).
const (
OptionYes = "yes"
@@ -17,25 +17,13 @@ func TestTheHoldAskIsValid(t *testing.T) {
t.Fatal(err)
}
now := time.Now()
a := HoldAsk(key, "mail", fNew, "ace", "health failed, phase running: connection refused", 0, now)
a := HoldAsk(key, "mail", fNew, now)
if err := a.Ask.Check(now); err != nil {
t.Errorf("the router refuses the ask: %v", err)
}
if _, err := delivery.EncodeAssignerAsk(key, a); err != nil {
t.Errorf("the ask cannot be written: %v", err)
}
// 1906/H3: the ask names the machine whose check failed and the hold's words.
if !strings.Contains(a.Ask.Explanation, "on ace") || !strings.Contains(a.Ask.Explanation, "connection refused") {
t.Errorf("the ask does not name the node and the hold's words: %q", a.Ask.Explanation)
}
// 1906/H2: an ask written again after a lift is another asking, under an id of its own.
again := HoldAsk(key, "mail", fNew, "ace", "x", 7, now)
if again.Asking != 7 || again.Ask.ID != delivery.AskingID(key, 7) || again.Ask.ID == a.Ask.ID {
t.Errorf("the asking after a lift = %d %s", again.Asking, again.Ask.ID)
}
if _, err := delivery.EncodeAssignerAsk(key, again); err != nil {
t.Errorf("the ask asked again cannot be written: %v", err)
}
if _, err := delivery.ViewAnswerKey(key); err != nil {
t.Errorf("the view may not answer it: %v", err)
}
@@ -86,17 +74,6 @@ func TestTheHeldConditionNamesModuleNodeAndWords(t *testing.T) {
}
}
// 1906/H2: a hold lifted by itself clears the held condition under its key, saying why.
func TestALiftedHoldClearsItsConditionSayingWhy(t *testing.T) {
c := LiftedCondition("mail", fNew)
if _, err := asks.EncodeCondition(c); err != nil {
t.Fatalf("the condition is refused: %v", err)
}
if c.State != asks.StateCleared || c.Key != HeldCondition("mail", fNew, "ace", "x", true, false).Key || !strings.Contains(c.Resolved, "lifted its hold by itself") {
t.Errorf("condition = %+v", c)
}
}
// 1906/D4: the record ask and the backup ask are valid asks of the router and of mesh-sdk's layout at their keys, answerable in
// the view, and name the move and the node.
func TestTheRecordAndBackupAsksAreValid(t *testing.T) {
@@ -147,19 +124,3 @@ func TestTheRecordAndBackupAsksAreValid(t *testing.T) {
t.Errorf("the backup condition is refused: %v", err)
}
}
// 1945/AC21: the condition of a module on neither list is urgent, names the module, the node and the rule holding its want back,
// under the assigner's own name keyed by node and module, and is cleared under the same key.
func TestTheNeitherConditionNamesModuleNodeAndRule(t *testing.T) {
c := NeitherCondition("g14", "i3status-rust", Holding{Rule: delivery.RuleProviders, Words: "no provider of x11-display runs healthy on g14 or at its canary"}, true)
if _, err := asks.EncodeCondition(c); err != nil {
t.Fatalf("the condition is refused: %v", err)
}
if c.State != asks.StateRaised || c.Severity != asks.SeverityUrgent || c.Subject.ID != "i3status-rust" || c.Subject.Machine != "g14" ||
!strings.Contains(c.Explanation, "x11-display") || !strings.Contains(c.Explanation, "rule 5") || c.Key != "mesh-delivery-assigner.neither.g14.i3status-rust" {
t.Errorf("condition = %+v", c)
}
if got := NeitherCondition("g14", "i3status-rust", Holding{}, false); got.State != asks.StateCleared || got.Key != c.Key {
t.Errorf("cleared = %+v", got)
}
}
@@ -4,8 +4,6 @@ import (
"fmt"
"slices"
"sort"
"strings"
"time"
"git.novox.be/novox/mesh-sdk/go/catalogue"
"git.novox.be/novox/mesh-sdk/go/delivery"
@@ -15,20 +13,13 @@ import (
// trunk (the indexer's modules), whether the builder's build of F is done, whether F is held, whether the controller
// issued its credential for F, its manifest at F, and the nodes it is assigned to.
type Module struct {
Name string
FP string
Built bool
Held bool
Granted bool
// GrantedAt says whether the controller issued the module's credential for another fingerprint than FP: the first want of a node
// that runs an older one names it (issue 1945). Nil: none.
GrantedAt func(fp string) bool
// GrantedOn says whether the controller's grant of FP names the module's accounts on a node, when the grant names any:
// a grant written for the module's other nodes says nothing of a node whose account is not issued and sealed (issue
// 1945, live on 2026-10-25 at 20:04: ace's mesh-build-go was wanted with no account on the bus). Nil: the grant alone.
GrantedOn func(node string) bool
Manifest catalogue.Manifest
Nodes []string
Name string
FP string
Built bool
Held bool
Granted bool
Manifest catalogue.Manifest
Nodes []string
// Composed answers the controller's composition of the module for a node at a fingerprint (mesh-controller_composed): its
// digest and the judgement of the move to it, and whether it exists. Nil: none exists (ADR 0539 §1, §6).
Composed func(node, fp string) (delivery.ComposedDigest, bool)
@@ -40,10 +31,6 @@ type Module struct {
// Restored says whether the operator answered "restore" to the restore ask of this data module on node: it is then set back
// as any module is, losing the writes made since (ADR 0539 §19). Nil: none did.
Restored func(node string) bool
// Kept says the operator answered "keep" to the hold ask of FP: a hold kept so is not lifted by itself.
Kept bool
// Now is when this pass reads the haves: the held record names it as when the health check was seen to fail. Zero: not said.
Now time.Time
}
func (m Module) composition(node, fp string) (delivery.ComposedDigest, bool) {
@@ -82,40 +69,27 @@ func Canary(nodes []string, online map[string]bool) string {
// with why (ADR 0429 §6). Rule 4, at most `step` others ahead, always holds: the mesh keeps no step, and step 0 is no
// limit (ADR 0528 clause 3).
func Holds(m Module, node string, n Nodes) (bool, int, string) {
return holds(m, node, n, true)
}
// holds is Holds; with ordering false it leaves out the two rules that order a move among nodes and modules, the canary (3)
// and the providers (5), and keeps those without which there is nothing to want: the build done, not held, the grant and the
// composition.
func holds(m Module, node string, n Nodes, ordering bool) (bool, int, string) {
switch {
case !m.Built:
return false, delivery.RuleBuildDone, fmt.Sprintf("the build of %s at %s is not done", m.Name, short(m.FP))
case m.Held:
return false, delivery.RuleNotHeld, fmt.Sprintf("%s is held at %s", m.Name, short(m.FP))
}
if ordering {
canary := Canary(m.Nodes, n.Online)
switch {
case canary == "":
return false, delivery.RuleCanary, fmt.Sprintf("none of the nodes of %s is online, so it has no canary", m.Name)
case node != canary && !healthy(n.Haves[canary][m.Name], m.FP):
return false, delivery.RuleCanary, fmt.Sprintf("the canary %s does not run %s healthy yet", canary, short(m.FP))
}
for _, provision := range m.Manifest.Requires {
if ok, why := providedHealthy(provision, node, n); !ok {
return false, delivery.RuleProviders, why
}
canary := Canary(m.Nodes, n.Online)
switch {
case canary == "":
return false, delivery.RuleCanary, fmt.Sprintf("none of the nodes of %s is online, so it has no canary", m.Name)
case node != canary && !healthy(n.Haves[canary][m.Name], m.FP):
return false, delivery.RuleCanary, fmt.Sprintf("the canary %s does not run %s healthy yet", canary, short(m.FP))
}
for _, provision := range m.Manifest.Requires {
if ok, why := providedHealthy(provision, node, n); !ok {
return false, delivery.RuleProviders, why
}
}
if !m.Granted {
return false, delivery.RuleGrant, fmt.Sprintf("the controller has not issued the credential of %s for %s", m.Name, short(m.FP))
}
if m.GrantedOn != nil && !m.GrantedOn(node) {
return false, delivery.RuleGrant, fmt.Sprintf("the controller has not issued the credential of %s on %s for %s", m.Name, node,
short(m.FP))
}
if _, ok := m.composition(node, m.FP); !ok {
return false, delivery.RuleComposed, fmt.Sprintf("the controller has not composed %s for %s at %s", m.Name, node, short(m.FP))
}
@@ -164,9 +138,6 @@ type Outcome struct {
Wants map[string]delivery.Want
// Held is the hold of m.FP to write, when a node's health check failed on it and it is not held yet.
Held *delivery.Held
// Lifted is the hold of m.FP lifted by itself: every node that runs the module runs m.FP healthy (issue 1906). The held
// record and its ask are taken away, and the wants move by the rules again in this pass.
Lifted bool
// Record is the moves a person must allow before the want names them: each is asked once (ADR 0539 §14).
Record []RecordNeed
// Restore is the nodes of a module holding data whose held fingerprint failed its health check there and that are not set
@@ -174,11 +145,6 @@ type Outcome struct {
Restore []string
// Rules is, by node, the rule holding the node back from m.FP and why.
Rules map[string]Holding
// Neither is, by node, the nodes the module is assigned to that run it (or whose have names it), hold no want for it after this
// pass and are not held (a held module is on the held list), with the rule holding the first want back: the node's own composition does not hold a module with a fingerprint of the new form, so such a module is on
// neither list and nothing applies it on that node (issue 1945). A module with no fingerprint of the new form goes with the node's
// own composition and is never here.
Neither map[string]Holding
}
// RecordNeed is a move of a module of policy record on a node that waits for the operator's yes: the composition's digest, the
@@ -202,15 +168,12 @@ type Holding struct {
// module holding data is not set back but asked about (Restore). A
// node that wants m but is no longer among its nodes gets its want written removed, never deleted (ADR 0408 §4).
func Wants(m Module, n Nodes, wants map[string]delivery.Want, lastHealthy map[string]string) Outcome {
out := Outcome{Wants: map[string]delivery.Want{}, Rules: map[string]Holding{}, Neither: map[string]Holding{}}
out := Outcome{Wants: map[string]delivery.Want{}, Rules: map[string]Holding{}}
held := m.Held
if held && !m.Kept && m.FP != "" && allRunHealthy(m, n) {
out.Lifted, held = true, false
}
if !held && !out.Lifted && !m.Released && m.FP != "" {
if !held && !m.Released && m.FP != "" {
for _, node := range slices.Sorted(slices.Values(m.Nodes)) {
if h, ok := n.Haves[node][m.Name]; ok && h.FP == m.FP && failed(h) {
out.Held = &delivery.Held{Node: node, Words: HeldWords(h, m.Now)}
if h, ok := n.Haves[node][m.Name]; ok && h.FP == m.FP && h.Health == delivery.HealthFailed {
out.Held = &delivery.Held{Node: node, Words: h.Detail}
held = true
break
}
@@ -233,55 +196,17 @@ func Wants(m Module, n Nodes, wants map[string]delivery.Want, lastHealthy map[st
}
continue
}
ok, rule, words := Holds(mm, node, n)
target := m.FP
if have, ran := n.Haves[node][m.Name]; !ok && (!had || before.Removed) && (rule == delivery.RuleCanary || rule == delivery.RuleProviders) && runs(have) {
// A node that runs the module and has no want for it has no want the ordering rules could hold it back to: holding its first
// want back leaves the module on neither list, since the node's own composition gives up every module with a fingerprint of
// the new form (issue 1945: on g14 the bundles of clipmenu, screen-lock and every module requiring x11-display were removed
// and never made again, because xorg's health is unknown and screen-lock's canary never ran it healthy). Its first want names
// what it runs, never a newer fingerprint, so the canary is never passed over: F when it runs F (by the rules without which
// there is nothing to want), or an older fingerprint of the new form whose composition and grant exist. Rules 3 and 5 then
// order the move from it as from any want.
switch _, composed := m.composition(node, have.FP); {
case !ran:
case have.FP == m.FP:
ok, rule, words = holds(mm, node, n, false)
case !strings.HasPrefix(have.FP, delivery.SourcePrefix2):
// An old-form build (not src2:) has no composition, so no want can ever name it: its first want names F, and only by every
// rule as usual, the canary (3) and the providers (5) included, which here hold it back. Nothing is written and the module
// stays on neither list until the canary runs F healthy; then Holds passes and the want names F (issue 1945, a design
// limit set by the lead: the canary stays in force and old-form modules leave neither list by themselves).
case composed && m.GrantedAt != nil && m.GrantedAt(have.FP):
ok, target = true, have.FP
}
}
switch {
switch ok, rule, words := Holds(mm, node, n); {
case ok:
d, _ := m.composition(node, target)
next = delivery.Want{FP: target, Composed: d.Digest}
at := m
if target != m.FP {
// the first want at the older fingerprint the node runs: held back from F by the rule, and judged as a want at it
next.Newest, next.Rule, next.Words = m.FP, rule, words
at.FP = target
}
if had && !before.Removed && before.FP == target && before.Composed == d.Digest {
d, _ := m.composition(node, m.FP)
next = delivery.Want{FP: m.FP, Composed: d.Digest}
if had && !before.Removed && before.FP == m.FP && before.Composed == d.Digest {
next.Backup, next.Words = before.Backup, before.Words // the answer to a failed backup stays while its target stays (ADR 0539 §16)
}
running, ran := n.Haves[node][m.Name]
// A removed want is no want: a first want after it goes through the record ask as a first want does.
asked, hadLive := before, had
if before.Removed {
asked, hadLive = delivery.Want{}, false
}
if need, wait := at.mustAsk(node, asked, hadLive, ran, running.FP, running.Composed == "", d); wait {
if need, wait := m.mustAsk(node, before, had, ran, running.FP, running.Composed == "", d); wait {
if need != nil {
out.Record = append(out.Record, *need)
} else if !hadLive && ran && runs(running) {
// The bus, which the assigner never moves (§15), runs here with no want: it is on neither list until the operator
// writes its want, and the pass says why (issue 1945).
out.Neither[node] = Holding{Rule: delivery.RuleComposed, Words: busMoveWords(m.Name, node, target, d)}
}
continue
}
@@ -308,33 +233,11 @@ func Wants(m Module, n Nodes, wants map[string]delivery.Want, lastHealthy map[st
default:
out.Rules[node] = Holding{Rule: rule, Words: words}
if !had || before.Removed {
// The node runs the module, or its have names it, and gets no want for it: the module is on neither list (issue 1945), and
// the pass says so. A new assignment, with no have, waits for the rules and is on no list yet by design. A held module (rule 2,
// held before or by this pass) is on the held list instead, whose hold is said and answered on its own: it is never on neither
// list (issue 1945: systemd-resolved, held on shanks by the operator, was said as urgent every hour).
if h, there := n.Haves[node][m.Name]; there && h.Phase != delivery.PhaseRemoved && !held {
out.Neither[node] = Holding{Rule: rule, Words: words}
}
continue
}
next = delivery.Want{FP: before.FP, Composed: before.Composed, Newest: m.FP, Rule: rule, Words: words}
if before.FP == m.FP {
// Held at F by the rules: a new revision of F's composition waits for them, as a new fingerprint does (ADR 0539 §5).
next = delivery.Want{FP: m.FP, Composed: before.Composed}
} else if d, there := m.composition(node, before.FP); there && d.Digest != "" && d.Digest != before.Composed {
// Held back at an earlier fingerprint whose kept composition the controller composed again with today's config
// (novox/hq ADR 0545): the want follows its new revision, so the node does not run that build with the config of
// the moment it was made (on g14 a capacity of 0 after a rollback). Asked as any move of a record module is.
at := m
at.FP = before.FP
running, ran := n.Haves[node][m.Name]
need, wait := at.mustAsk(node, before, true, ran, running.FP, running.Composed == "", d)
if need != nil {
out.Record = append(out.Record, *need)
}
if !wait {
next.Composed = d.Digest
}
}
}
if !had || !delivery.SameWant(normal(before), normal(next)) {
@@ -349,58 +252,6 @@ func Wants(m Module, n Nodes, wants map[string]delivery.Want, lastHealthy map[st
return out
}
// failed says whether a have's health check failed once the node is done moving: phase running or failed (never removed, held, or
// moving). A failed health while it
// backs up, applies or checks is a step of the move, not its outcome (on novox at 15:27:36 on 2026-10-25, "health failed, phase
// applying" held the controller's canary two seconds before the new build was healthy). Every healthy read (healthy, providedHealthy,
// the world's lastHealthy) already asks for phase running.
func failed(h delivery.Have) bool {
return h.Health == delivery.HealthFailed && (h.Phase == delivery.PhaseRunning || h.Phase == delivery.PhaseFailed)
}
// runs says whether a have is of a module the node runs: running, whatever its health.
func runs(h delivery.Have) bool {
return h.Phase == delivery.PhaseRunning
}
// allRunHealthy says whether every node of m that runs the module (its have's phase is running, or failed) runs m.FP healthy, and
// at least one does: a hold whose fingerprint runs healthy wherever the module runs holds nothing back any more (issue 1906: seven
// modules stayed held on every machine after one transient failed check). A node set back to another fingerprint does not run
// m.FP, and a failed one does not run it healthy, so the hold stays; a have that is removed, held for its backup or still moving
// (backing up, applying, checking) is not counted, or it would keep the hold up for ever.
func allRunHealthy(m Module, n Nodes) bool {
ran := 0
for _, node := range m.Nodes {
h, ok := n.Haves[node][m.Name]
if !ok || (h.Phase != delivery.PhaseRunning && h.Phase != delivery.PhaseFailed) {
continue
}
if !healthy(h, m.FP) {
return false
}
ran++
}
return ran > 0
}
// HeldWords are the words a hold keeps of its cause (issue 1906): the health and phase the node reported, when the assigner saw
// it, and the node's own words. The node is the held record's own field.
func HeldWords(h delivery.Have, at time.Time) string {
words := fmt.Sprintf("health %s, phase %s", h.Health, h.Phase)
if !at.IsZero() {
words += ", seen at " + at.UTC().Format(time.RFC3339)
}
if d := strings.TrimSpace(h.Detail); d != "" {
return words + ": " + d
}
return words + ": the node gave no words"
}
// LiftedWords say why a hold was lifted by itself.
func LiftedWords(module, fp string) string {
return fmt.Sprintf("every node that runs %s runs %s healthy, so the assigner lifted its hold by itself", module, short(fp))
}
// normal is a want without its version, as a decoded and a composed one are compared.
func normal(w delivery.Want) delivery.Want {
w.V = 0
@@ -447,17 +298,6 @@ func (m Module) mustAsk(node string, before delivery.Want, had, ran bool, runnin
return &RecordNeed{Node: node, Digest: d.Digest, Judgement: d.Judgement, Words: d.Words}, true
}
// busMoveWords say why the bus module's want at fp is not written on node: its composition differs from what the node runs in more
// than the bus's user list, so it is a move of the bus, which the operator writes (ADR 0539 §15).
func busMoveWords(module, node, fp string, d delivery.ComposedDigest) string {
why := string(d.Judgement)
if w := strings.TrimSpace(d.Words); w != "" {
why += ": " + w
}
return fmt.Sprintf("the composition of %s for %s at %s changes more than the bus's user list (judged %s), so it is a move of the bus, "+
"which the assigner never makes: the operator writes its want (ADR 0539 §15)", module, node, short(fp), why)
}
// recorded says whether the module is of policy record: as the controller holds it beside the composition's digest, the
// operator's own choice included (ADR 0539 §14); on a composition made by hand, which carries none, as the manifest derives it.
func (m Module) recorded(d delivery.ComposedDigest) bool {
@@ -471,28 +311,11 @@ func (m Module) recorded(d delivery.ComposedDigest) bool {
// userListRevision is the want that names the bus module's composition at the fingerprint the node runs when that composition
// differs from the one the node runs only in the bus's user list (ComposedDigest.UserListOnly), and whether there is one. A node
// whose want names another fingerprint than the one it runs, or a module that does not provide the bus, has none.
//
// A node whose have is of the old form (from the old node-engine: a build and no fingerprint of the new form, no composition) runs
// no fingerprint a composition is kept at, so its revision names F, the newest fingerprint, when the controller judged F's
// composition to differ from the declaration the node last applied only in the bus's user list (ADR 0539 §7, ADR 0540 §9). Any
// other difference is a move of the bus and stays the operator's (§15). Without this the bus never got its first want after the
// switch, and its user list was never written again (issue 1945: novox's nats from 11:23 on 2026-10-25).
func (m Module) userListRevision(node string, before delivery.Want, had bool, n Nodes) (delivery.Want, bool) {
if !catalogue.ProvidesBus(m.Manifest) {
return delivery.Want{}, false
}
running, ran := n.Haves[node][m.Name]
if ran && oldForm(running) && m.FP != "" && (!had || before.Removed || before.FP == m.FP) {
d, ok := m.composition(node, m.FP)
if !ok || !d.UserListOnly {
return delivery.Want{}, false
}
w := delivery.Want{FP: m.FP, Composed: d.Digest}
if had && !before.Removed && before.Composed == d.Digest {
w.Backup, w.Words = before.Backup, before.Words
}
return w, true
}
if !ran || running.FP == "" || (had && (before.Removed || before.FP != running.FP)) {
return delivery.Want{}, false
}
@@ -506,25 +329,3 @@ func (m Module) userListRevision(node string, before delivery.Want, had bool, n
}
return w, true
}
// oldForm says whether a have is of the old form: the old node-engine's, naming a build and no fingerprint of the new form, and no
// composition (issue 1945: novox's nats, build-1792674595610176744).
func oldForm(h delivery.Have) bool {
return h.Phase != delivery.PhaseRemoved && h.Composed == "" && (h.FP != "" || h.Build != "") &&
!strings.HasPrefix(h.FP, delivery.SourcePrefix2)
}
// GrantNamesNode says whether a grant of module names its accounts on node: its own `<node>.<module>` or a process's
// `<node>.<module>.<process>`. A grant that names no account (a module that does not speak on the bus) names every node.
func GrantNamesNode(g delivery.ControllerGrant, node, module string) bool {
if len(g.Accounts) == 0 {
return true
}
own := node + "." + module
for _, a := range g.Accounts {
if a == own || strings.HasPrefix(a, own+".") {
return true
}
}
return false
}
@@ -4,7 +4,6 @@ import (
"slices"
"strings"
"testing"
"time"
"git.novox.be/novox/mesh-sdk/go/catalogue"
"git.novox.be/novox/mesh-sdk/go/delivery"
@@ -86,33 +85,6 @@ func TestEachRuleHoldsTheWantBack(t *testing.T) {
t.Errorf("rule %d broken: g14's want is %+v", rule, w)
}
}
// Rule 6 per node: a grant written for the module's other nodes does not grant a node whose account it does not name
// (issue 1945, live on 2026-10-25 at 20:04: ace's mesh-build-go was wanted with no account on the bus).
m, n := ready()
grant := delivery.ControllerGrant{Accounts: []string{"ace.mail", "ace.mail.send"}}
m.GrantedOn = func(node string) bool { return GrantNamesNode(grant, node, "mail") }
out := Wants(m, n, wantedOld(), nil)
if w := out.Wants["ace"]; w.FP != fNew {
t.Errorf("ace, whose account the grant names, wants %+v", w)
}
if got := out.Rules["g14"]; got.Rule != delivery.RuleGrant || !strings.Contains(got.Words, "on g14") {
t.Errorf("g14, whose account the grant does not name, is held by %+v", got)
}
for _, c := range []struct {
accounts []string
node, module string
named bool
}{
{nil, "g14", "mail", true},
{[]string{"g14.mail.send"}, "g14", "mail", true},
{[]string{"g14.mailer"}, "g14", "mail", false},
{[]string{"ace.mail"}, "g14", "mail", false},
} {
if got := GrantNamesNode(delivery.ControllerGrant{Accounts: c.accounts}, c.node, c.module); got != c.named {
t.Errorf("%v names %s of %s: %v, want %v", c.accounts, c.module, c.node, got, c.named)
}
}
}
// 1297/B4, ADR 0435 §3: a canary that is offline is passed over: the next node by name is the canary, and moves first.
@@ -138,10 +110,8 @@ func TestAnOfflineCanaryIsPassedOver(t *testing.T) {
func TestAFailedHealthHoldsAndSetsBack(t *testing.T) {
m, n := ready()
n.Haves["ace"]["mail"] = delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthFailed, Detail: "connection refused"}
m.Now = time.Date(2026, 10, 25, 1, 0, 0, 0, time.UTC)
out := Wants(m, n, map[string]delivery.Want{"ace": {FP: fNew, Composed: dig(fNew)}, "g14": {FP: fOld, Composed: dig(fOld)}}, map[string]string{"ace": fOld})
// 1906/H3: the held record keeps its cause in words: which machine (its node), which health, when, and the node's words.
if out.Held == nil || out.Held.Node != "ace" || out.Held.Words != "health failed, phase running, seen at 2026-10-25T01:00:00Z: connection refused" {
if out.Held == nil || out.Held.Node != "ace" || out.Held.Words != "connection refused" {
t.Fatalf("held = %+v", out.Held)
}
if w := out.Wants["ace"]; w.FP != fOld || w.Composed != dig(fOld) || w.Newest != fNew || w.Rule != delivery.RuleNotHeld {
@@ -248,25 +218,6 @@ func TestARevisionAtTheSameFingerprintIsAWantByTheSameRules(t *testing.T) {
if w, ok := out.Wants["ace"]; ok && w.Composed == revised {
t.Errorf("a revision was written while a rule holds: %+v", w)
}
// A want held back at an earlier fingerprint follows the new revision of that fingerprint's kept composition, which the
// controller composed again with today's config (novox/hq ADR 0545: on g14 a capacity of 0 after a rollback), while
// the rules still hold F; for a record module whose revision recreates a container, only after the ask.
out = Wants(m, n, wantedOld(), nil)
if w := out.Wants["g14"]; w.FP != fOld || w.Composed != revised || w.Newest != fNew {
t.Errorf("g14, held back at its earlier fingerprint, wants %+v; want it at that fingerprint naming the revised digest", w)
}
m.Composed = func(node, fp string) (delivery.ComposedDigest, bool) {
return delivery.ComposedDigest{Digest: revised, Judgement: delivery.JudgedRecreates, Policy: delivery.PolicyRecord,
Against: dig(fOld), Words: "recreates"}, true
}
out = Wants(m, n, wantedOld(), nil)
if w := out.Wants["g14"]; w.Composed == revised {
t.Errorf("a recorded module's revision that recreates a container was written with no ask: %+v", w)
}
if !slices.ContainsFunc(out.Record, func(r RecordNeed) bool { return r.Node == "g14" && r.Digest == revised }) {
t.Errorf("the revision was not asked: %+v", out.Record)
}
}
// 1906/D2: a set-back names the composition of the fingerprint it sets back to, and when that is gone the want stays and says so.
@@ -350,7 +301,6 @@ func TestAMoveJudgedNeitherOrNeverRunOrNotRecordedIsWrittenAtOnce(t *testing.T)
}
m, n, _ = movingTo(delivery.JudgedRecreates)
recording(&m)
m.Nodes = []string{"ace"} // g14 runs it with no want, so its first want is asked (1945/AC19): this case is ace's alone
delete(n.Haves["ace"], "mail")
if out := Wants(m, n, nil, nil); out.Wants["ace"].FP != fNew || len(out.Record) != 0 {
t.Errorf("a node that never ran it: %+v", out)
@@ -498,53 +448,8 @@ func TestAUserListOnlyRevisionOfTheBusIsWrittenAtOnceBeforeTheRecordHold(t *test
t.Errorf("%s: the user-list revision was asked: %+v", name, out.Record)
}
}
revision := m.Composed
// Issue 1945: a have of the old form (a build, no fingerprint of the new form, no composition) runs no fingerprint a composition is
// kept at. The revision then names F when the controller judged F's composition to differ from the declaration the node last
// applied only in the user list; any other difference is a move of the bus, refused by §15 and said on neither list.
for name, c := range map[string]struct {
userListOnly bool
wanted map[string]delivery.Want
want *delivery.Want // nil: no want written
neither bool
}{
"old form, user list only, no want": {userListOnly: true, want: &delivery.Want{FP: fNew, Composed: userList}},
"old form, user list only, a removed want": {userListOnly: true, wanted: map[string]delivery.Want{"ace": {Removed: true}}, want: &delivery.Want{FP: fNew, Composed: userList}},
"old form, user list only, written before": {userListOnly: true, wanted: map[string]delivery.Want{"ace": {FP: fNew, Composed: userList}}},
"old form, a real move of the bus": {neither: true},
} {
m.Manifest.Provides = []catalogue.Offer{{Name: "mesh-bus"}}
m.Composed = func(node, fp string) (delivery.ComposedDigest, bool) {
if fp != fNew {
return delivery.ComposedDigest{}, false
}
if c.userListOnly {
return delivery.ComposedDigest{Digest: userList, Judgement: delivery.JudgedNeither, UserListOnly: true, Against: dig(fOld),
Words: "only the bus's user list changes, against the declaration ace last applied", Policy: delivery.PolicyRecord}, true
}
return delivery.ComposedDigest{Digest: dig(fNew), Judgement: delivery.JudgedNeither, Words: "the node runs no composition of mail",
Policy: delivery.PolicyRecord}, true
}
n.Haves["ace"]["mail"] = delivery.Have{Build: "build-1792674595610176744", Phase: delivery.PhaseRunning, Health: delivery.HealthOK}
out := Wants(m, n, c.wanted, nil)
w, ok := out.Wants["ace"]
switch {
case c.want == nil && ok:
t.Errorf("%s: a want was written: %+v", name, w)
case c.want != nil && (!ok || w.FP != c.want.FP || w.Composed != c.want.Composed || w.Removed):
t.Errorf("%s: ace's want = %+v (written %v), want %+v", name, w, ok, *c.want)
}
if len(out.Record) != 0 {
t.Errorf("%s: asked: %+v", name, out.Record)
}
h, on := out.Neither["ace"]
if on != c.neither || (on && (!strings.Contains(h.Words, "§15") || !strings.Contains(h.Words, "more than the bus's user list"))) {
t.Errorf("%s: on neither list %v (%+v), want %v with the words of §15", name, on, h, c.neither)
}
}
// the same flag on a module that is not the bus changes nothing
m.Manifest.Provides, m.Composed = nil, revision
n.Haves["ace"]["mail"] = delivery.Have{FP: fOld, Phase: delivery.PhaseRunning, Health: delivery.HealthOK}
m.Manifest.Provides = nil
if out := Wants(m, n, map[string]delivery.Want{"ace": {FP: fOld, Composed: dig(fOld)}}, nil); out.Wants["ace"].Composed == userList {
t.Errorf("a module that is not the bus was written at once: %+v", out.Wants["ace"])
}
@@ -615,343 +520,3 @@ func TestAHaveFromBeforeTheSwitchGetsItsWantWithNoRecordHold(t *testing.T) {
}
}
}
// 1906/H3: a failed check whose node gave no words is held saying so, never with empty words.
func TestAHoldOfAFailureWithoutWordsSaysSo(t *testing.T) {
m, n := ready()
n.Haves["ace"]["mail"] = delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthFailed}
out := Wants(m, n, map[string]delivery.Want{"ace": {FP: fNew, Composed: dig(fNew)}}, map[string]string{"ace": fOld})
if out.Held == nil || out.Held.Words != "health failed, phase running: the node gave no words" {
t.Fatalf("held = %+v", out.Held)
}
}
// 1906/H2: given a held fingerprint that every node running the module now runs healthy, when the assigner passes, then the
// hold is lifted by itself and the wants move to it by the rules again. A node set back to another fingerprint, a node whose
// check fails, or a hold the operator answered "keep" leaves it held; a node that does not run the module at all does not count.
func TestAHoldIsLiftedOnceEveryNodeRunsItHealthy(t *testing.T) {
everywhere := func() (Module, Nodes) {
m, n := ready()
m.Held = true
n.Haves["g14"]["mail"] = delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthOK}
return m, n
}
m, n := everywhere()
out := Wants(m, n, wantedOld(), map[string]string{"ace": fOld, "g14": fOld})
if !out.Lifted || out.Held != nil {
t.Fatalf("lifted = %v, held = %+v: want the hold lifted", out.Lifted, out.Held)
}
for _, node := range []string{"ace", "g14"} {
if w := out.Wants[node]; w.FP != fNew || w.Rule != 0 {
t.Errorf("%s's want = %+v, want it moved to the lifted fingerprint", node, w)
}
}
m, n = everywhere()
delete(n.Haves["g14"], "mail")
if out := Wants(m, n, wantedOld(), nil); !out.Lifted {
t.Error("a node that does not run the module kept the hold")
}
for name, change := range map[string]func(*Module, *Nodes){
"set back": func(_ *Module, n *Nodes) {
n.Haves["g14"]["mail"] = delivery.Have{FP: fOld, Phase: delivery.PhaseRunning, Health: delivery.HealthOK}
},
"failing": func(_ *Module, n *Nodes) {
n.Haves["g14"]["mail"] = delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthFailed}
},
"kept": func(m *Module, _ *Nodes) { m.Kept = true },
"no haves": func(_ *Module, n *Nodes) { delete(n.Haves["g14"], "mail"); delete(n.Haves["ace"], "mail") },
"not held": func(m *Module, _ *Nodes) { m.Held = false },
} {
m, n := everywhere()
change(&m, &n)
if out := Wants(m, n, wantedOld(), map[string]string{"ace": fOld, "g14": fOld}); out.Lifted {
t.Errorf("%s: the hold was lifted", name)
}
}
}
// x11 is screen-lock as on g14 and shanks on 2026-10-25 (issue 1945): built, granted and composed at F and at the older fOld
// on both, requiring x11-display, whose provider xorg runs on each node with its health unknown; each node runs screen-lock at fOld
// (health unknown) and holds no want for it. The canary g14 does not run F.
func x11() (Module, Nodes) {
m := Module{Name: "screen-lock", FP: fNew, Built: true, Granted: true, Composed: composedEverywhere, Nodes: []string{"g14", "shanks"},
GrantedAt: func(fp string) bool { return fp == fNew || fp == fOld },
Manifest: catalogue.Manifest{Module: "screen-lock", Requires: []string{"x11-display"}}}
n := Nodes{
Online: map[string]bool{"g14": true, "shanks": true},
Haves: map[string]map[string]delivery.Have{
"g14": {"screen-lock": {FP: fOld, Phase: delivery.PhaseRunning, Health: delivery.HealthUnknown},
"xorg": {FP: fpOf('x'), Phase: delivery.PhaseRunning, Health: delivery.HealthUnknown}},
"shanks": {"screen-lock": {FP: fOld, Phase: delivery.PhaseRunning, Health: delivery.HealthUnknown},
"xorg": {FP: fpOf('x'), Phase: delivery.PhaseRunning, Health: delivery.HealthUnknown}},
},
Providers: map[string]map[string][]string{"x11-display": {"xorg": {"g14", "shanks"}}},
}
return m, n
}
// 1945/AC19: given a node that runs an older fingerprint of the new form, composed and granted, with no want, held back from F by
// the providers or the canary rule, when the assigner passes, then its first want names the fingerprint it runs with that
// composition, never the newest, and says the newest it is held back from; the canary is never passed over.
func TestANodeRunningAModuleWithNoWantGetsItsFirstWantAtWhatItRuns(t *testing.T) {
m, n := x11()
out := Wants(m, n, nil, nil)
for _, node := range []string{"g14", "shanks"} {
if w := out.Wants[node]; w.FP != fOld || w.Composed != dig(fOld) || w.Newest != fNew || w.Rule == 0 {
t.Errorf("%s's want = %+v, want its first want at the fingerprint it runs, held back from F", node, w)
}
}
if len(out.Neither) != 0 {
t.Errorf("left on neither list: %+v", out.Neither)
}
// with the provider healthy, only the canary holds shanks back: still at what it runs
m, n = x11()
n.Haves["g14"]["xorg"] = delivery.Have{FP: fpOf('x'), Phase: delivery.PhaseRunning, Health: delivery.HealthOK}
n.Haves["shanks"]["xorg"] = n.Haves["g14"]["xorg"]
if w := Wants(m, n, nil, nil).Wants["shanks"]; w.FP != fOld || w.Rule != delivery.RuleCanary {
t.Errorf("shanks held back by the canary with no want: %+v, want at fOld by rule 3", w)
}
// a node already running F gets its want at F; a removed want is no want either
m, n = x11()
n.Haves["g14"]["screen-lock"] = delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthUnknown}
out = Wants(m, n, map[string]delivery.Want{"g14": {Removed: true}}, nil)
if w := out.Wants["g14"]; w.FP != fNew || w.Composed != dig(fNew) || w.Removed {
t.Errorf("g14 running F with a removed want = %+v, want F", w)
}
}
// 1945/AC19: a node running a fingerprint that is not of the new form, or one whose composition or grant does not exist, gets no
// want past the rules, and is named on neither list; the ordering rules still order every move from a want.
func TestNoFirstWantWithoutTheCompositionAndGrantOfWhatItRuns(t *testing.T) {
for what, change := range map[string]func(*Module, *Nodes){
"an old-form build": func(_ *Module, n *Nodes) {
n.Haves["shanks"]["screen-lock"] = delivery.Have{FP: "build-2", Phase: delivery.PhaseRunning, Health: delivery.HealthUnknown}
},
"no grant": func(m *Module, _ *Nodes) { m.GrantedAt = func(fp string) bool { return fp == fNew } },
"no composition": func(m *Module, _ *Nodes) {
m.Composed = func(_, fp string) (delivery.ComposedDigest, bool) {
return delivery.ComposedDigest{Digest: dig(fp), Judgement: delivery.JudgedNeither}, fp == fNew
}
},
} {
m, n := x11()
change(&m, &n)
out := Wants(m, n, nil, nil)
if w, ok := out.Wants["shanks"]; ok {
t.Errorf("%s: shanks got a want: %+v", what, w)
}
if h, ok := out.Neither["shanks"]; !ok || h.Rule == 0 {
t.Errorf("%s: shanks on neither list = %+v (named %v)", what, h, ok)
}
}
m, n := x11()
before := map[string]delivery.Want{"g14": {FP: fOld, Composed: dig(fOld)}, "shanks": {FP: fOld, Composed: dig(fOld)}}
out := Wants(m, n, before, nil)
if len(out.Neither) != 0 {
t.Errorf("nodes holding a want named on neither list: %+v", out.Neither)
}
for node, w := range out.Wants {
if w.FP != fOld {
t.Errorf("%s's want moved from its want past the ordering rules: %+v", node, w)
}
}
}
// 1945/AC20: a new assignment (a node with no have of the module) waiting for the canary, the build or the grant gets no want and is
// never named on neither list: only a node that runs the module, or whose have names it, is; a module with no fingerprint of the new
// form goes with the node's own composition and is never named.
func TestANewAssignmentWaitingIsNotOnNeitherList(t *testing.T) {
m, n := x11()
delete(n.Haves["shanks"], "screen-lock")
out := Wants(m, n, nil, nil)
if w, ok := out.Wants["shanks"]; ok {
t.Errorf("a new assignment waiting for the canary got a want: %+v", w)
}
if _, ok := out.Neither["shanks"]; ok {
t.Errorf("a new assignment waiting for the canary is named on neither list: %+v", out.Neither)
}
m.Built = false
if out := Wants(m, n, nil, nil); len(out.Neither) != 1 || out.Neither["g14"].Rule != delivery.RuleBuildDone {
t.Errorf("with no build: neither = %+v, want g14 alone (it runs the module), by rule 1", out.Neither)
}
m.FP = ""
if out := Wants(m, n, nil, nil); len(out.Neither) != 0 {
t.Errorf("a module of the node's own composition named on neither list: %+v", out.Neither)
}
}
// 1945/AC21: a node whose have names an old-form build (no fingerprint of the new form, so no composition of it) can never be
// named in a want at what it runs; its first want names F, under the canary rule as usual and with everything F needs and the
// providers rule. Given the canary runs F healthy, shanks's first want names F; given it does not yet, shanks gets no want and
// stays on neither list. The old-form node that is itself the canary gets F once its providers run healthy.
func TestAnOldFormNodeGetsItsFirstWantAtFOnlyUnderTheCanary(t *testing.T) {
oldForm := func(n Nodes, node string) {
n.Haves[node]["screen-lock"] = delivery.Have{FP: "build-1792260876956849831", Phase: delivery.PhaseRunning, Health: delivery.HealthUnknown}
}
providersHealthy := func(n Nodes) {
for _, node := range []string{"g14", "shanks"} {
n.Haves[node]["xorg"] = delivery.Have{FP: fpOf('x'), Phase: delivery.PhaseRunning, Health: delivery.HealthOK}
}
}
// the canary g14 runs F healthy: shanks's first want names F
m, n := x11()
providersHealthy(n)
oldForm(n, "shanks")
n.Haves["g14"]["screen-lock"] = delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthOK}
out := Wants(m, n, nil, nil)
if w := out.Wants["shanks"]; w.FP != fNew || w.Composed != dig(fNew) || w.Newest != "" || w.Rule != 0 {
t.Errorf("canary healthy on F: shanks's want = %+v, want F", w)
}
if _, ok := out.Neither["shanks"]; ok {
t.Errorf("canary healthy on F: shanks still on neither list: %+v", out.Neither)
}
// the canary does not run F healthy yet: no want, and shanks stays on neither list by rule 3
m, n = x11()
providersHealthy(n)
oldForm(n, "shanks")
n.Haves["g14"]["screen-lock"] = delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthUnknown}
out = Wants(m, n, nil, nil)
if w, ok := out.Wants["shanks"]; ok {
t.Errorf("canary not healthy on F: shanks got a want: %+v", w)
}
if h, ok := out.Neither["shanks"]; !ok || h.Rule != delivery.RuleCanary {
t.Errorf("canary not healthy on F: shanks on neither list = %+v (named %v), want rule 3", h, ok)
}
// the canary healthy on F but the providers not healthy on shanks: rule 5 still holds it back
m, n = x11()
oldForm(n, "shanks")
n.Haves["g14"]["screen-lock"] = delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthOK}
out = Wants(m, n, nil, nil)
if w, ok := out.Wants["shanks"]; ok {
t.Errorf("providers not healthy: shanks got a want: %+v", w)
}
if h := out.Neither["shanks"]; h.Rule != delivery.RuleProviders {
t.Errorf("providers not healthy: shanks on neither list = %+v, want rule 5", h)
}
// F not granted: nothing, even with the canary healthy
m, n = x11()
providersHealthy(n)
oldForm(n, "shanks")
n.Haves["g14"]["screen-lock"] = delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthOK}
m.Granted = false
if w, ok := Wants(m, n, nil, nil).Wants["shanks"]; ok {
t.Errorf("F not granted: shanks got a want: %+v", w)
}
// the old-form node is the canary itself: F once its providers run healthy
m, n = x11()
providersHealthy(n)
oldForm(n, "g14")
if w := Wants(m, n, nil, nil).Wants["g14"]; w.FP != fNew || w.Composed != dig(fNew) {
t.Errorf("old-form canary: g14's want = %+v, want F", w)
}
}
// A failed health counts only once the node is done moving: phase running or failed, never backing-up, applying or checking.
// The live row on novox at 15:27:36 on 2026-10-25 ("health failed, phase applying" at the controller's new fingerprint) held the
// canary two seconds before the new build was healthy.
func TestAFailedHealthWhileTheNodeMovesHoldsNothing(t *testing.T) {
for _, phase := range []delivery.Phase{delivery.PhaseBackingUp, delivery.PhaseApplying, delivery.PhaseChecking} {
m := Module{Name: "mesh-controller", FP: fNew, Built: true, Granted: true, Composed: composedEverywhere, Nodes: []string{"novox"},
Manifest: catalogue.Manifest{Module: "mesh-controller"}}
n := Nodes{Online: map[string]bool{"novox": true}, Haves: map[string]map[string]delivery.Have{
"novox": {"mesh-controller": {FP: fNew, Phase: phase, Health: delivery.HealthFailed}}}}
before := map[string]delivery.Want{"novox": {FP: fNew, Composed: dig(fNew)}}
if out := Wants(m, n, before, map[string]string{"novox": fOld}); out.Held != nil {
t.Errorf("phase %s, health failed: held %+v, want no hold while the node moves", phase, out.Held)
}
}
for _, phase := range []delivery.Phase{delivery.PhaseRunning, delivery.PhaseFailed} {
m := Module{Name: "mesh-controller", FP: fNew, Built: true, Granted: true, Composed: composedEverywhere, Nodes: []string{"novox"},
Manifest: catalogue.Manifest{Module: "mesh-controller"}}
n := Nodes{Online: map[string]bool{"novox": true}, Haves: map[string]map[string]delivery.Have{
"novox": {"mesh-controller": {FP: fNew, Phase: phase, Health: delivery.HealthFailed}}}}
before := map[string]delivery.Want{"novox": {FP: fNew, Composed: dig(fNew)}}
if out := Wants(m, n, before, map[string]string{"novox": fOld}); out.Held == nil || out.Held.Node != "novox" {
t.Errorf("phase %s, health failed: held %+v, want a hold on novox", phase, out.Held)
}
}
}
// A hold is lifted by itself once every node that runs the module runs F healthy: a have of a node that is not running the module
// (removed, held for its backup, or still moving) is not counted, so it does not keep the hold up for ever; a node running another
// fingerprint, or failed, still does.
func TestAHoldIsLiftedOverHavesThatAreNotRunning(t *testing.T) {
setup := func(phase delivery.Phase, fp string, health delivery.Health) (Module, Nodes) {
m := Module{Name: "web", FP: fNew, Built: true, Granted: true, Held: true, Composed: composedEverywhere, Nodes: []string{"ace", "g14"}}
n := Nodes{Online: map[string]bool{"ace": true, "g14": true}, Haves: map[string]map[string]delivery.Have{
"ace": {"web": {FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthOK}},
"g14": {"web": {FP: fp, Phase: phase, Health: health}}}}
return m, n
}
for _, phase := range []delivery.Phase{delivery.PhaseRemoved, delivery.PhaseHeld, delivery.PhaseApplying, delivery.PhaseBackingUp, delivery.PhaseChecking} {
m, n := setup(phase, fOld, delivery.HealthUnknown)
if out := Wants(m, n, nil, nil); !out.Lifted {
t.Errorf("g14's have in phase %s kept the hold up", phase)
}
}
for what, h := range map[string]delivery.Have{
"running another fingerprint": {FP: fOld, Phase: delivery.PhaseRunning, Health: delivery.HealthOK},
"failed": {FP: fNew, Phase: delivery.PhaseFailed, Health: delivery.HealthFailed},
} {
m, n := setup(h.Phase, h.FP, h.Health)
if out := Wants(m, n, nil, nil); out.Lifted {
t.Errorf("g14 %s: the hold was lifted", what)
}
}
// a failed health on a removed have holds nothing
m, n := setup(delivery.PhaseRemoved, fNew, delivery.HealthFailed)
m.Held = false
if out := Wants(m, n, nil, nil); out.Held != nil {
t.Errorf("a removed have with a failed health held F: %+v", out.Held)
}
}
// A module of policy record whose want was removed but which the node still runs gets its first want through the record ask: a
// removed want is no want, so a move the controller judges to recreate or move data waits for the operator's yes.
func TestAFirstWantAfterARemovedWantGoesThroughTheRecordAsk(t *testing.T) {
m, n := x11()
recording(&m)
m.Composed = func(_, fp string) (delivery.ComposedDigest, bool) {
return delivery.ComposedDigest{Digest: dig(fp), Judgement: delivery.JudgedRecreates, Policy: delivery.PolicyRecord}, true
}
out := Wants(m, n, map[string]delivery.Want{"shanks": {Removed: true}}, nil)
if w, ok := out.Wants["shanks"]; ok {
t.Errorf("shanks's first want after a removed want was written with no record ask: %+v", w)
}
asked := false
for _, r := range out.Record {
asked = asked || (r.Node == "shanks" && r.Digest == dig(fOld) && r.Judgement == delivery.JudgedRecreates)
}
if !asked {
t.Errorf("the record asks = %+v, want one for shanks at the composition it runs", out.Record)
}
}
// 1945/AC23: given a module held at F under rule 2 (the operator holds systemd-resolved on shanks, 2026-10-25), or held by the
// assigner in this pass after a failed health check of F, when a node that runs it holds no want for it, then the node is not on
// neither list: a held module is on the held list, whose hold is said and answered on its own.
func TestAHeldModuleWithNoWantIsNotOnNeitherList(t *testing.T) {
for what, change := range map[string]func(*Module, *Nodes){
"held at F": func(m *Module, _ *Nodes) { m.Held, m.Kept = true, true },
"held at F, running an old-form build": func(m *Module, n *Nodes) {
m.Held, m.Kept = true, true
n.Haves["shanks"]["screen-lock"] = delivery.Have{FP: "build-2", Phase: delivery.PhaseRunning, Health: delivery.HealthUnknown}
},
"held by this pass": func(_ *Module, n *Nodes) {
n.Haves["g14"]["screen-lock"] = delivery.Have{FP: fNew, Phase: delivery.PhaseRunning, Health: delivery.HealthFailed}
n.Haves["shanks"]["screen-lock"] = delivery.Have{FP: "build-2", Phase: delivery.PhaseRunning, Health: delivery.HealthUnknown}
},
} {
m, n := x11()
change(&m, &n)
out := Wants(m, n, nil, nil)
if len(out.Neither) != 0 {
t.Errorf("%s: a held module is on neither list: %+v", what, out.Neither)
}
if out.Rules["shanks"].Rule != delivery.RuleNotHeld {
t.Errorf("%s: shanks is held back by %+v, want rule 2", what, out.Rules["shanks"])
}
}
}
+1 -1
View File
@@ -3,7 +3,7 @@ module mesh-delivery-checker
go 1.26.0
require (
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261023213200-403a5d72ab93
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024232922-99407c37df95
github.com/nats-io/nats-server/v2 v2.11.17
github.com/nats-io/nats.go v1.54.0
)
+8 -2
View File
@@ -1,5 +1,11 @@
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261023213200-403a5d72ab93 h1:FAiFY2KBKEoZSH49OZTt7/zRBIQxyl5g/KhYrpggbN4=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261023213200-403a5d72ab93/go.mod h1:ouERNpZizzwZ0HLEtAypsY9FR20FyUgUYRni6J7kVCY=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024214154-bc16740feb0d h1:oJREw0ggKqJWJ5DEISSL0i2ZC5XqdXAidxo9NJ74TCc=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024214154-bc16740feb0d/go.mod h1:ouERNpZizzwZ0HLEtAypsY9FR20FyUgUYRni6J7kVCY=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024231716-571ac521f053 h1:8Kbp3aJG3jGndDOfmGdMSDgJVEcc1hyu921hVnnlnlQ=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024231716-571ac521f053/go.mod h1:ouERNpZizzwZ0HLEtAypsY9FR20FyUgUYRni6J7kVCY=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024232250-9b408db24a58 h1:4Damtdqlf6u6nJhvMA99peDy1jmjixtcwFXJ4bWEkLk=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024232250-9b408db24a58/go.mod h1:ouERNpZizzwZ0HLEtAypsY9FR20FyUgUYRni6J7kVCY=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024232922-99407c37df95 h1:I8zbgGzRjmgYydPTWo8hmaZW/P/g+qGDUfu3E32/pkU=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024232922-99407c37df95/go.mod h1:ouERNpZizzwZ0HLEtAypsY9FR20FyUgUYRni6J7kVCY=
github.com/antithesishq/antithesis-sdk-go v0.7.0-default-no-op h1:Z/MZK75wC/NSrkgqeNIa7jexam9uWzhLmFTSCPI/kn0=
github.com/antithesishq/antithesis-sdk-go v0.7.0-default-no-op/go.mod h1:FQyySiasQQM8735Ddel3MRojmy4dA1IqCeyJ5jmPMbI=
github.com/google/go-tpm v0.9.8 h1:slArAR9Ft+1ybZu0lBwpSmpwhRXaa85hWtMinMyRAWo=
@@ -0,0 +1,96 @@
package check
import (
"context"
"fmt"
"sync"
"testing"
"time"
"git.novox.be/novox/mesh-sdk/go/delivery"
)
// countingAsks puts asks through inner, and answers no wait until want asks are put (or a bound passes, when it says
// how many were): a group whose asks are put one after another, each waited for, never gets past its first.
type countingAsks struct {
inner Asks
want int
mu sync.Mutex
puts int
all chan struct{}
}
func (c *countingAsks) Put(ctx context.Context, a delivery.StepAsk) error {
if err := c.inner.Put(ctx, a); err != nil {
return err
}
c.mu.Lock()
c.puts++
if c.puts == c.want {
close(c.all)
}
c.mu.Unlock()
return nil
}
func (c *countingAsks) Wait(ctx context.Context, a delivery.StepAsk) (delivery.Step, error) {
select {
case <-c.all:
case <-time.After(time.Second):
c.mu.Lock()
defer c.mu.Unlock()
return delivery.Step{}, fmt.Errorf("%d of %d asks were put before the first was waited for", c.puts, c.want)
}
return c.inner.Wait(ctx, a)
}
// 1921/S8: a group's asks are all put before any is waited for, so its members run side by side on the build nodes;
// the verdict is judged in member order as before.
func TestAGroupPutsEveryMembersAskBeforeTheFirstAnswer(t *testing.T) {
b := newBench(t)
counting := &countingAsks{inner: b.ex.Asks, want: 3, all: make(chan struct{})}
b.ex.Asks = counting
cs := caseOf(t, pullOf(repoA, 7, head1), pullOf(repoSDK, 12, head2), pullOf("novox/mesh-controller", 3, head1))
v := b.judge(t, cs).Verdict
if v.State != delivery.VerdictPass {
t.Fatalf("verdict = %+v", v)
}
if counting.puts != 3 {
t.Errorf("%d asks were put, want 3", counting.puts)
}
}
// 1921/S10: the levels of an ask run side by side, so what of the ask's time is its fetch (Took.Clone) is the ask's time
// less its longest level, not less the sum of its levels.
func TestTheCloneTimeTakesTheLongestLevelOffNotTheSum(t *testing.T) {
b := newBench(t)
b.test = func(c stepCall) (any, error) {
r := captured[Tested](t, "go-test-slow")
r.Level, _ = c.args["level"].(string)
r.Elapsed, r.Setup, r.Lint = 1.5, 0, 0 // the ask's step takes 2 s (bench.answer)
return r, nil
}
v := b.judge(t, caseOf(t, pullOf(repoA, 7, head1))).Verdict
if v.State != delivery.VerdictPass || v.Took.Clone != 500 {
t.Errorf("clone = %d ms, want 500 (2 s less the longest level's 1.5 s): %+v", v.Took.Clone, v)
}
}
// ADR 0542 §2: a lint finding, which the Go build module marks as the lint step's, is reported as the lint step's failure,
// not as the unit level's.
func TestALintFindingIsTheLintStepsFailure(t *testing.T) {
b := newBench(t)
b.test = func(c stepCall) (any, error) {
r := captured[Tested](t, "go-test-slow")
r.Level, _ = c.args["level"].(string)
if r.Level == delivery.LevelUnit {
r.Passed = false
r.Failed = []Failure{{Package: ".", Test: "gofmt", Step: "lint", Output: "not as gofmt writes them (run gofmt -w): a.go"}}
}
return r, nil
}
v := b.judge(t, caseOf(t, pullOf(repoA, 7, head1))).Verdict
if v.State != delivery.VerdictFail || v.Failed.Test != "lint: ..gofmt" {
t.Errorf("failed = %+v, want the lint step's failure", v.Failed)
}
}
@@ -82,6 +82,9 @@ func CaseOf(c delivery.Candidate, trunk map[string]string) *Case {
}
cs := &Case{Key: key, Group: c.Group, HeadsHash: c.HeadsHash, Heads: GateHeads(c), Trunk: map[string]string{}, Progress: &Progress{}}
for _, p := range c.Pulls {
if p.Merged {
continue // merged while the group waits for its other members: judged by the verdict the group had (issue 1791)
}
switch head, known := trunk[p.Repository]; {
case !known || head == "":
cs.Left = append(cs.Left, fmt.Sprintf("%s#%d: the trunk head of %s is not known", p.Repository, p.Number, p.Repository))
@@ -119,7 +122,18 @@ func GateHeads(c delivery.Candidate) string {
}
// sameTrunk says whether two verdicts' trunk heads are the same.
func sameTrunk(a, b map[string]string) bool { return maps.Equal(a, b) }
// sameTrunk says whether a verdict judged on the trunk heads judged stands for a case on the trunk heads now: every
// repository of the case at the head it was judged on. A repository the verdict names and the case does not is a member
// that merged since (issue 1791): its own merge moved that trunk, and the members left are judged by the verdict the
// group had.
func sameTrunk(judged, now map[string]string) bool {
for repo, head := range now {
if judged[repo] != head {
return false
}
}
return len(now) > 0 || len(judged) == 0
}
// Trunks holds the trunk head of each repository, as the gate's source.main says (a watch puts it in).
type Trunks struct {
@@ -83,12 +83,24 @@ type Checker struct {
tried map[string]bool // by verdict key: looked at since this process started
epoch uint64 // changes of the nodes seen since this process started
nodesFail map[string]uint64 // by verdict key: the epoch a failure of the nodes was written at
// current is, by candidates key, the verdict key of the heads the group holds now; running holds, by candidates key,
// the stop of each check that runs, by its verdict key (issue 1791; issue 1921, try speed S9).
current map[string]string
running map[string]map[string]context.CancelCauseFunc
}
// superseded is why a check whose heads the group no longer holds was stopped: the heads it holds now.
type superseded struct{ by string }
func (s superseded) Error() string {
return "superseded: the group holds other heads now (" + s.by + "), whose check judges it"
}
func (ch *Checker) init() {
ch.once.Do(func() {
ch.known, ch.inflight, ch.again, ch.tried = map[string]delivery.Candidate{}, map[string]bool{}, map[string]bool{}, map[string]bool{}
ch.nodesFail = map[string]uint64{}
ch.current, ch.running = map[string]string{}, map[string]map[string]context.CancelCauseFunc{}
})
}
@@ -126,6 +138,15 @@ func (ch *Checker) Candidate(ctx context.Context, key string, value []byte) {
}
ch.mu.Lock()
ch.known[key] = c
if vk, err := delivery.CheckerVerdictKey(c.Group, c.HeadsHash); err == nil {
ch.current[key] = vk
// A check of heads the group no longer holds is stopped: its asks are not put, and it ends superseded.
for running, stop := range ch.running[key] {
if running != vk {
stop(superseded{by: c.HeadsHash})
}
}
}
ch.mu.Unlock()
ch.consider(ctx, key, c)
}
@@ -329,12 +350,48 @@ func (ch *Checker) try(ctx context.Context, key string, cs *Case) {
if cl == nil {
return
}
run, stop := context.WithCancelCause(ctx)
defer stop(nil)
ch.mu.Lock()
now := ch.current[key]
if ch.running[key] == nil {
ch.running[key] = map[string]context.CancelCauseFunc{}
}
ch.running[key][cs.Key] = stop
ch.mu.Unlock()
defer func() {
ch.mu.Lock()
delete(ch.running[key], cs.Key)
if len(ch.running[key]) == 0 {
delete(ch.running, key)
}
ch.mu.Unlock()
}()
if now != "" && now != cs.Key {
stop(superseded{by: now}) // the group moved on before the check began
}
began := ch.now()
ch.writeOwed(ctx, cs, delivery.Verdict{State: delivery.VerdictRunning}, Outcome{})
out := ch.judge(ctx, cs)
if run.Err() == nil {
ch.writeOwed(ctx, cs, delivery.Verdict{State: delivery.VerdictRunning}, Outcome{})
}
out := ch.judge(run, cs)
if ctx.Err() != nil {
return
}
var sup superseded
if errors.As(context.Cause(run), &sup) {
// Stopped for newer heads: the verdict ends failed, not running, and no status is put, as the newer check owns the
// status of the heads they share.
v := delivery.Verdict{Group: cs.Group, Heads: cs.Heads, Trunk: cs.Trunk, State: delivery.VerdictFail,
Failed: delivery.FailedFirst{Machine: "checker", Words: sup.Error()}}
if data, err := delivery.EncodeVerdict(v); err == nil {
if _, err := ch.Verdicts.Update(ctx, cs.Key, data, cl.rev); err != nil {
ch.logf("%s: the superseded verdict could not be written: %v", cs.Key, err)
}
}
ch.logf("%s: %s", cs.Key, sup.Error())
return
}
v := out.Verdict
v.Group, v.Heads, v.Trunk = cs.Group, cs.Heads, cs.Trunk
if v.Took.Queue == 0 && !cs.Queued.IsZero() && began.After(cs.Queued) {
@@ -0,0 +1,90 @@
package check
import (
"encoding/json"
"fmt"
"strings"
"testing"
"git.novox.be/novox/mesh-sdk/go/delivery"
)
// failingResult is a level's test result in which the named tests failed, each in its own package.
func failingResult(t *testing.T, tests int, failed ...Failure) json.RawMessage {
t.Helper()
raw, err := json.Marshal(Tested{Level: delivery.LevelIntegration, Tests: tests, Elapsed: 2, Failed: failed, Passed: len(failed) == 0})
if err != nil {
t.Fatal(err)
}
return raw
}
func judgeLevel(t *testing.T, tests ...json.RawMessage) (*delivery.Verdict, *state) {
t.Helper()
st := &state{j: &Examiner{}, c: &Case{}, levels: map[string]*acc{delivery.LevelUnit: {}, delivery.LevelIntegration: {}}, nodes: map[string]bool{}}
v, _ := st.level(t.Context(), &target{pull: Pull{Repository: "novox/x"}}, delivery.LevelAnswer{Level: delivery.LevelIntegration, Tests: tests})
return v, st
}
// Issue 1921: two failing tests in two packages are two entries of the verdict's Failures and two lines of a try's
// failed list, each with its own first line (its cause, issue 1901), whether the build module answered them in one test
// result (one go test run over both packages) or in two (the changed packages first, then the rest); every result is
// counted, the ones after a failure too; and Failed keeps the first one's words whole.
func TestTwoFailingTestsInTwoPackagesAreTwoEntries(t *testing.T) {
a := Failure{Package: "x/a", Test: "TestA", Cause: "a_test.go:9: want 1, got 2", Output: " a_test.go:3: logged\n a_test.go:9: want 1, got 2\n--- FAIL: TestA (0.00s)"}
b := Failure{Package: "x/b", Test: "TestB", Cause: "b_test.go:4: boom", Output: " b_test.go:4: boom\n--- FAIL: TestB (0.00s)"}
for name, tests := range map[string][]json.RawMessage{
"one result": {failingResult(t, 10, a, b)},
"two results": {failingResult(t, 4, a), failingResult(t, 3), failingResult(t, 3, b)},
} {
v, st := judgeLevel(t, tests...)
if v == nil {
t.Fatalf("%s: the level passed", name)
}
want := []string{"integration: x/a.TestA: a_test.go:9: want 1, got 2", "integration: x/b.TestB: b_test.go:4: boom"}
if got := failedLines(*v); strings.Join(got, "\n") != strings.Join(want, "\n") {
t.Errorf("%s: failed lines\n%s\nwant\n%s", name, strings.Join(got, "\n"), strings.Join(want, "\n"))
}
if len(v.Failures) != 2 || v.Failed.Test != "integration: x/a.TestA" || !strings.Contains(v.Failed.Words, "a_test.go:3: logged") {
t.Errorf("%s: verdict %+v", name, v)
}
if run := st.levels[delivery.LevelIntegration].run; run != 10 {
t.Errorf("%s: %d tests counted, want all 10", name, run)
}
if _, err := delivery.EncodeVerdict(delivery.Verdict{Group: "g", Heads: "0123456789ab", State: v.State, Failed: v.Failed, Failures: v.Failures}); err != nil {
t.Errorf("%s: the verdict is not written: %v", name, err)
}
}
}
// Issue 1921: past delivery.MaxFailures the list names the first fifty and how many more, the status line says how many
// failed, and each entry's line is cut to MaxFailureLine.
func TestFailuresAreCappedWithHowManyMore(t *testing.T) {
var fs []Failure
for i := range 60 {
fs = append(fs, Failure{Package: fmt.Sprintf("x/p%d", i), Test: "TestT", Cause: strings.Repeat("y", 500)})
}
v, _ := judgeLevel(t, failingResult(t, 60, fs...))
if len(v.Failures) != delivery.MaxFailures+1 || v.Failures[delivery.MaxFailures].Test != "and 10 more" {
t.Fatalf("%d entries, last %+v", len(v.Failures), v.Failures[len(v.Failures)-1])
}
if len(v.Failures[0].Words) > MaxFailureLine {
t.Errorf("an entry's line is %d bytes", len(v.Failures[0].Words))
}
if n := failedCount(*v); n != 60 {
t.Errorf("failedCount = %d, want 60", n)
}
cs := &Case{Pulls: []Pull{{Repository: "novox/x", Head: "0123456789abcdef", Trunk: "fedcba9876543210"}}}
if line := Line(*v, cs, cs.Pulls[0], nil, nil, nil); !strings.HasPrefix(line, "failed (60 tests): integration: x/p0.TestT") {
t.Errorf("line %q", line)
}
}
// A verdict that names no list (a conflict, a refused step, a verdict written before issue 1921) is one failed line, as
// before.
func TestAVerdictWithoutFailuresIsOneLine(t *testing.T) {
v := delivery.Verdict{State: delivery.VerdictFail, Failed: delivery.FailedFirst{Machine: "g14", Words: "files in conflict: go.mod\nmore"}}
if got := failedLines(v); len(got) != 1 || got[0] != "g14: files in conflict: go.mod" {
t.Fatalf("%q", got)
}
}
@@ -74,6 +74,18 @@ func (f *fakeSteps) tool(name string) []stepCall {
return out
}
// toolAt is tool's calls of one level: a check's levels run side by side (issue 1921, try speed S10), so only the calls of
// one level keep an order.
func (f *fakeSteps) toolAt(name, level string) []stepCall {
var out []stepCall
for _, c := range f.tool(name) {
if c.args["level"] == level {
out = append(out, c)
}
}
return out
}
type entries []delivery.NodeToolchain
func (e entries) Entries(language string) []delivery.NodeToolchain {
@@ -269,9 +281,9 @@ func TestOnlyThePackageWithoutAKeptVerdictRuns(t *testing.T) { // 1288/AC1
t.Fatal(err)
}
out := b.judge(t, caseOf(t, pullOf(repoA, 7, head1)))
tests := b.steps.tool("test")
if len(tests) != 2 || strings.Join(packagesOf(tests[0]), ",") != "fixture/slow" || tests[0].args["level"] != "unit" {
t.Fatalf("test calls = %+v, want the unit level of fixture/slow alone, then the integration level", tests)
tests := b.steps.toolAt("test", "unit")
if len(tests) != 1 || strings.Join(packagesOf(tests[0]), ",") != "fixture/slow" || len(b.steps.toolAt("test", "integration")) != 1 {
t.Fatalf("test calls = %+v, want the unit level of fixture/slow alone, and the integration level", b.steps.tool("test"))
}
if u := out.Verdict.Levels[delivery.LevelUnit]; u.Run != 2 || u.Kept != 40 {
t.Errorf("unit = %+v, want 2 run and 40 kept", u)
@@ -302,7 +314,7 @@ func TestAReAskRunsOnlyWhatFailedAndWhatChanged(t *testing.T) { // 1288/AC2, AC3
// Asked again with nothing changed: only the failed package runs; the good one is taken.
b.steps.calls = nil
b.judge(t, caseOf(t, pullOf(repoA, 7, head1)))
tests := b.steps.tool("test")
tests := b.steps.toolAt("test", "unit")
if len(tests) == 0 || strings.Join(packagesOf(tests[0]), ",") != "fixture/failing/bad" {
t.Fatalf("the re-ask ran %+v, want only fixture/failing/bad", tests)
}
@@ -310,7 +322,7 @@ func TestAReAskRunsOnlyWhatFailedAndWhatChanged(t *testing.T) { // 1288/AC2, AC3
b.digest[delivery.LevelUnit] = Digested{Level: "unit", Packages: map[string]string{"fixture/failing/bad": digA, "fixture/failing/good": digD}}
b.steps.calls = nil
b.judge(t, caseOf(t, pullOf(repoA, 7, head1)))
tests = b.steps.tool("test")
tests = b.steps.toolAt("test", "unit")
if got := strings.Join(packagesOf(tests[0]), ","); got != "fixture/failing/bad,fixture/failing/good" && got != "fixture/failing/good,fixture/failing/bad" {
t.Errorf("after a digest changed the run was %q, want the failed package and the changed one", got)
}
@@ -341,7 +353,9 @@ func TestANodePackageIsNeverKeptAndRunsEveryCheck(t *testing.T) { // 1288/AC4, 1
}
}
func TestTheChangesOwnFailingTestRunsFirstAndNothingElseRuns(t *testing.T) { // 1289/AC1, AC2
// 1289/AC1, AC2 as issue 1921 changed it: the change's own failing test runs first, the level's other packages still run
// after it (so one check names every failure); the later level ran beside it (S10) and is not judged.
func TestTheChangesOwnFailingTestRunsFirstThenTheRestAndNoLaterLevel(t *testing.T) {
b := newBench(t)
b.fetch.Changed = []string{"modules/m/internal/b/b_test.go"}
b.digest[delivery.LevelUnit] = Digested{Level: "unit", Packages: map[string]string{"m/internal/a": digA, "m/internal/b": digB}}
@@ -356,12 +370,12 @@ func TestTheChangesOwnFailingTestRunsFirstAndNothingElseRuns(t *testing.T) { //
if out.Verdict.State != delivery.VerdictFail || !strings.Contains(out.Verdict.Failed.Test, "TestThreeIsFour") {
t.Fatalf("verdict = %+v", out.Verdict)
}
tests := b.steps.tool("test")
if len(tests) != 1 || strings.Join(packagesOf(tests[0]), ",") != "m/internal/b" {
t.Fatalf("test calls = %+v, want only m/internal/b, once", tests)
tests := b.steps.toolAt("test", "unit")
if len(tests) != 2 || strings.Join(packagesOf(tests[0]), ",") != "m/internal/b" || strings.Join(packagesOf(tests[1]), ",") != "m/internal/a" {
t.Fatalf("test calls = %+v, want m/internal/b, then m/internal/a, each once", tests)
}
if n := len(b.steps.tool("digest")); n != 1 {
t.Errorf("%d digest calls, want the unit level's alone (the check stopped)", n)
if out.Verdict.Levels[delivery.LevelIntegration].Run != 0 {
t.Errorf("the integration level was judged after the unit level failed: %+v", out.Verdict.Levels)
}
if out.Verdict.Composition == nil || out.Verdict.Composition.State != delivery.CompositionNotRun {
t.Errorf("composition = %+v", out.Verdict.Composition)
@@ -159,6 +159,9 @@ type target struct {
runner string
// judged is what the line says of the program that judged a change to the build module; empty for any other change.
judged string
// ask is the check ask put for it, and failed why it could not be asked (issue 1921, try speed S8).
ask delivery.StepAsk
failed *delivery.Verdict
}
// judgedBy takes what a step under runner workspace answered of the program that ran it (judged_by): the candidate, or the
@@ -307,12 +310,13 @@ func (st *state) levelsOf() []string {
return []string{delivery.LevelUnit, delivery.LevelIntegration}
}
// judgeAll puts one check ask per pull request and build module the repository needs, the mesh-sdk member first, and
// judges each answer. A head that does not merge, a pin that is not allowed or a level that fails ends the check at once
// (ADR 0484 clause 13).
// judgeAll puts one check ask per pull request and build module the repository needs, every ask of the group at once so
// the members run side by side on the build nodes (issue 1921, try speed S8), and judges the answers in member order, the
// mesh-sdk member first. A head that does not merge, a pin that is not allowed or a level that fails ends the check at once
// (ADR 0484 clause 13): the verdict is the first failure in member order, as when the asks were put one after another.
func (st *state) judgeAll(ctx context.Context) *delivery.Verdict {
// The mesh-sdk member first: whether its merge changes the TypeScript package's paths decides whether the other members'
// Node fetches take the group's package (ADR 0500 §5).
// Node fetches take the group's package (ADR 0500 §5). Those Node asks alone wait for the mesh-sdk member's answer.
pulls := slices.Clone(st.c.Pulls)
slices.SortStableFunc(pulls, func(a, b Pull) int {
if (a.Repository == SDKRepository) != (b.Repository == SDKRepository) {
@@ -323,34 +327,32 @@ func (st *state) judgeAll(ctx context.Context) *delivery.Verdict {
}
return 0
})
var groupPackage map[string]any
for _, p := range pulls {
var merged []Fetched
sdkGroup := len(st.c.Pulls) > 1 && len(pulls) > 0 && pulls[0].Repository == SDKRepository
members := make([][]*target, len(pulls))
var waiting []*target // Node asks of the other members, put once the mesh-sdk member is judged
for i, p := range pulls {
for _, bm := range st.j.Modules.BuildModules(p.Repository) {
t := &target{pull: p, module: bm}
if tc, none := delivery.AgreedToolchain(st.j.Toolchains.Entries(buildrun.LanguageOf[bm])); none == "" {
t.toolchain = tc.Version
}
args := map[string]any{"repository": p.Repository, "commit": p.Head, "ref": refOf(p.Number), "base": p.Trunk}
if bm == buildrun.NodeBuildModule && groupPackage != nil && p.Repository != SDKRepository {
args["package"] = groupPackage
}
if bm == buildrun.GoBuildModule {
beside, err := st.j.besideFor(ctx, st.c, p)
if err != nil {
return fail(forgeMachine(err, bm), err.Error()+unheardWords(err))
}
if len(beside) > 0 {
args["beside"] = beside
t.beside = besideLines(beside)
}
if p.Repository == catalogueRepository {
t.runner = RunnerWorkspace // used by the build module when the merge changes it (ADR 0535 §1)
}
}
t.fetchArgs = args
t := st.target(ctx, p, bm)
members[i] = append(members[i], t)
st.targets = append(st.targets, t)
if v := st.ask(ctx, t, &merged); v != nil {
if t.failed != nil {
continue // said when the judging reaches it, so the verdict stays the first failure in member order
}
if sdkGroup && bm == buildrun.NodeBuildModule && p.Repository != SDKRepository {
waiting = append(waiting, t)
continue
}
t.failed = st.put(ctx, t)
}
}
var groupPackage map[string]any
for i, p := range pulls {
var merged []Fetched
for _, t := range members[i] {
if t.failed != nil {
return t.failed
}
if v := st.judgeAsk(ctx, t, &merged); v != nil {
return v
}
}
@@ -360,14 +362,51 @@ func (st *state) judgeAll(ctx context.Context) *delivery.Verdict {
if p.Repository == SDKRepository && len(st.c.Pulls) > 1 && st.j.changesPackage(p, merged) {
groupPackage = map[string]any{"repository": p.Repository, "commit": p.Head, "ref": refOf(p.Number)}
}
if i == 0 && sdkGroup {
for _, t := range waiting {
if groupPackage != nil {
t.fetchArgs["package"] = groupPackage
}
t.failed = st.put(ctx, t)
}
}
}
return nil
}
// ask puts the target's check ask, waits for the step of the node that pulled it and judges its answer: the fetch, the
// pins of its go.mod files, then each level's digest, kept verdicts and tests.
func (st *state) ask(ctx context.Context, t *target, merged *[]Fetched) *delivery.Verdict {
// target is the check of one pull request by one build module: its fetch's arguments, its toolchain, the siblings its
// tests read and the runner it is asked with; failed says why it cannot be asked.
func (st *state) target(ctx context.Context, p Pull, bm string) *target {
t := &target{pull: p, module: bm}
if tc, none := delivery.AgreedToolchain(st.j.Toolchains.Entries(buildrun.LanguageOf[bm])); none == "" {
t.toolchain = tc.Version
}
args := map[string]any{"repository": p.Repository, "commit": p.Head, "ref": refOf(p.Number), "base": p.Trunk}
if bm == buildrun.GoBuildModule {
beside, err := st.j.besideFor(ctx, st.c, p)
if err != nil {
t.failed = fail(forgeMachine(err, bm), err.Error()+unheardWords(err))
return t
}
if len(beside) > 0 {
args["beside"] = beside
t.beside = besideLines(beside)
}
if p.Repository == catalogueRepository {
t.runner = RunnerWorkspace // used by the build module when the merge changes it (ADR 0535 §1)
}
}
t.fetchArgs = args
return t
}
// put puts the target's check ask on its build seat's stream; a check that was stopped (its group holds other heads now,
// issue 1791) puts none.
func (st *state) put(ctx context.Context, t *target) *delivery.Verdict {
p := t.pull
if err := context.Cause(ctx); err != nil {
return fail(t.module, fmt.Sprintf("the check of %s#%d was not asked: %v", p.Repository, p.Number, err))
}
job := &delivery.CheckJob{Fetch: t.fetchArgs, Levels: st.levelsOf(), Kept: true}
if t.runner != "" {
job.Runner, job.RunnerWhen = t.runner, buildGoDir
@@ -379,13 +418,20 @@ func (st *state) ask(ctx context.Context, t *target, merged *[]Fetched) *deliver
if st.c.Kind == KindTrunk {
kind = delivery.AskRecheck // a re-judge after a trunk move goes after every new head's check (ADR 0536 §4)
}
a := delivery.StepAsk{ID: buildstep.NewID(), Kind: kind, Language: askLanguage[t.module], Asker: delivery.CheckerModule,
t.ask = delivery.StepAsk{ID: buildstep.NewID(), Kind: kind, Language: askLanguage[t.module], Asker: delivery.CheckerModule,
At: st.j.now().UTC(), Check: job}
st.c.Progress.Set(fmt.Sprintf("the %s ask %s of %s#%d", kind, a.ID, p.Repository, p.Number), "")
if err := st.j.Asks.Put(ctx, a); err != nil {
st.c.Progress.Set(fmt.Sprintf("the %s ask %s of %s#%d", kind, t.ask.ID, p.Repository, p.Number), "")
if err := st.j.Asks.Put(ctx, t.ask); err != nil {
return fail(t.module, fmt.Sprintf("the check of %s#%d could not be asked: %v", p.Repository, p.Number, err))
}
step, err := st.j.Asks.Wait(ctx, a)
return nil
}
// judgeAsk waits for the step of the node that pulled the target's ask and judges its answer: the fetch, the pins of its
// go.mod files, then each level's digest, kept verdicts and tests.
func (st *state) judgeAsk(ctx context.Context, t *target, merged *[]Fetched) *delivery.Verdict {
p := t.pull
step, err := st.j.Asks.Wait(ctx, t.ask)
if err != nil {
if ctx.Err() != nil {
return fail(t.module, ctx.Err().Error())
@@ -427,7 +473,7 @@ func (st *state) ask(ctx context.Context, t *target, merged *[]Fetched) *deliver
ran := 0.0
for _, la := range answer.Levels {
v, elapsed := st.level(ctx, t, la)
ran += elapsed
ran = max(ran, elapsed) // the levels run side by side (issue 1921, try speed S10): the longest is what they took
if v != nil {
st.clone += max(0, took-time.Duration(ran*float64(time.Second)))
return v
@@ -438,8 +484,9 @@ func (st *state) ask(ctx context.Context, t *target, merged *[]Fetched) *deliver
}
// level judges one level of a check ask's answer as the tools' answers were judged: the digest keys the kept verdicts
// of the packages that passed, the kept ones taken are counted, and the first test result that did not pass is the
// verdict. It answers the level's running time (its tests' and setup's).
// of the packages that passed, the kept ones taken are counted, and every test result is read and counted, also after
// one that did not pass, so the verdict names every failed test of the level (issue 1921). It answers the level's
// running time (its tests' and setup's).
func (st *state) level(ctx context.Context, t *target, la delivery.LevelAnswer) (*delivery.Verdict, float64) {
level := la.Level
a := st.levels[level]
@@ -461,6 +508,7 @@ func (st *state) level(ctx context.Context, t *target, la delivery.LevelAnswer)
a.kept += k.Tests
}
elapsed := 0.0
var failedResults []Tested
for _, raw := range la.Tests {
var res Tested
if err := json.Unmarshal(raw, &res); err != nil {
@@ -489,9 +537,12 @@ func (st *state) level(ctx context.Context, t *target, la delivery.LevelAnswer)
}
}
if !res.Passed {
return levelFailure(t, level, res), elapsed
failedResults = append(failedResults, res)
}
}
if len(failedResults) > 0 {
return levelFailure(t, level, failedResults...), elapsed
}
return nil, elapsed
}
@@ -531,98 +582,23 @@ const MaxWords = 4096
// failureWords is a failure's words: its output's last lines up to MaxWords, with the line the build module told apart as
// the test's error first when the output does not start with it, so the first line of the words is the cause and the
// status line, which is cut short, names it. **A failure's words are never "PASS" or "ok"** (issue 1901, its third
// meeting: a package that failed after its tests passed was worded "PASS"): go test's summary lines are dropped, and with
// no cause told apart (an older build module) the cause is the output's first "panic:", "fatal error:" or "signal:"
// line, read before the output is cut, or the words say the tests passed and the program failed after them.
// status line, which is cut short, names it.
func failureWords(f Failure) string {
out, passed := dropSummary(strings.TrimSpace(f.Output))
cause := strings.TrimSpace(f.Cause)
if cause == "" {
cause = causeIn(out)
}
switch {
case cause == "" && out == "" && passed:
cause = "the tests passed, then the test program exited with a failure: an os.Exit or a panic in TestMain or after the tests"
case cause == "" && out == "":
// The build module on main sends a package that stopped (a timeout, an exit, a kill) as its "FAIL" line alone.
cause = "the package failed with no test to name and no panic, signal or build error in its output (an older build module sent no cause)"
}
if len(cause) > MaxCause {
cause = strings.ToValidUTF8(cause[:MaxCause], "") + "..."
}
// The words keep MaxWords exactly: the cause and its newline, the marker and the kept lines (review round 1).
budget := MaxWords
if cause != "" {
budget -= len(cause) + 1
}
if len(out) > budget {
const marker = "... "
tail := out[len(out)-(budget-len(marker)):]
out := strings.TrimSpace(f.Output)
if len(out) > MaxWords {
out = out[len(out)-MaxWords:]
// Start on a whole line, and a whole character.
if i := strings.IndexByte(tail, '\n'); i >= 0 {
tail = strings.TrimLeft(tail[i:], "\n")
if i := strings.IndexByte(out, '\n'); i >= 0 {
out = "... " + strings.TrimLeft(out[i:], "\n")
}
out = marker + strings.ToValidUTF8(tail, "")
out = strings.ToValidUTF8(out, "")
}
if cause != "" && !strings.HasPrefix(firstLine(out), cause) {
out = strings.TrimRight(cause+"\n"+out, "\n")
if c := strings.TrimSpace(f.Cause); c != "" && !strings.HasPrefix(firstLine(out), c) {
out = c + "\n" + out
}
return out
}
// MaxCause is the most a cause in a failure's words holds: one long line (the 1901 review: a cause of a kilobyte could
// push the words past MaxWords).
const MaxCause = 1024
// dropSummary drops go test's summary lines from a failure's output ("PASS", "FAIL", "ok <pkg>", "FAIL <pkg>"), and says
// whether a "PASS" was among them: the package's tests passed before it failed.
func dropSummary(out string) (string, bool) {
passed := false
lines := strings.Split(out, "\n")
kept := lines[:0]
for _, l := range lines {
// go's exact forms only (review round 1): "PASS" or "FAIL" alone, "ok \t<pkg>\t<time>", "FAIL\t<pkg>\t<time>".
switch t := strings.TrimRight(l, "\r"); {
case t == "PASS":
passed = true
case t == "FAIL", strings.HasPrefix(t, "ok \t"), strings.HasPrefix(t, "FAIL\t"):
default:
kept = append(kept, l)
}
}
return strings.TrimSpace(strings.Join(kept, "\n")), passed
}
// causeIn is a race the race detector reported, as its heading, its access and the frame that made it ("WARNING: DATA
// RACE: Write at 0x... by goroutine 9 in pkg.F.func1()"); else the first "panic:", "fatal error:" or "signal:" line of a
// failure's output, in that order; or "". **Only lines at column 0**, where go prints the runtime's: a test's indented log
// line holding the same words is not the cause (review round 1).
func causeIn(out string) string {
lines := strings.Split(out, "\n")
for i, l := range lines {
if strings.TrimRight(l, "\r") != "WARNING: DATA RACE" {
continue
}
words := "WARNING: DATA RACE"
if i+1 < len(lines) && strings.TrimSpace(lines[i+1]) != "" {
words += ": " + strings.TrimSuffix(strings.TrimSpace(lines[i+1]), ":")
if i+2 < len(lines) && strings.HasPrefix(lines[i+2], " ") {
words += " in " + strings.TrimSpace(lines[i+2])
}
}
return words
}
for _, prefix := range []string{"panic: ", "fatal error: ", "signal: "} {
for _, l := range lines {
if l = strings.TrimRight(l, "\r"); strings.HasPrefix(l, prefix) {
return l
}
}
}
return ""
}
// firstLine is the first non-empty line of a failure's words.
func firstLine(s string) string {
for _, l := range strings.Split(s, "\n") {
@@ -633,23 +609,64 @@ func firstLine(s string) string {
return ""
}
// failure turns a level that did not pass into a verdict naming what failed first.
func levelFailure(t *target, level string, res Tested) *delivery.Verdict {
// MaxFailureLine is the most bytes of a failure's line in a verdict's Failures: the first line of its words, which is
// its cause where the build module told one apart (issue 1901). Failed keeps the first failure's words whole.
const MaxFailureLine = 300
// levelFailure turns a level that did not pass into a verdict: Failed names what failed first with its words whole,
// and Failures every failed test of the level's results with its first line, at most delivery.MaxFailures and an entry
// saying how many more (issue 1921).
func levelFailure(t *target, level string, results ...Tested) *delivery.Verdict {
v := delivery.Verdict{State: delivery.VerdictFail}
switch {
case len(res.Failed) > 0:
f := res.Failed[0]
name := failedName(f)
if t.module == buildrun.NodeBuildModule && f.Package == "" {
name = res.Module
var all []delivery.FailedFirst
for _, res := range results {
if len(res.Failed) == 0 {
why := res.Why
if why == "" {
why = "the " + level + " level did not pass"
}
all = append(all, delivery.FailedFirst{Test: level + ": " + t.pull.Repository, Words: why})
continue
}
v.Failed = delivery.FailedFirst{Test: level + ": " + name, Words: failureWords(f)}
default:
why := res.Why
if why == "" {
why = "the " + level + " level did not pass"
for _, f := range res.Failed {
name := failedName(f)
if t.module == buildrun.NodeBuildModule && f.Package == "" {
name = res.Module
}
step := level
if f.Step != "" {
step = f.Step // the lint step's finding, not the level's (ADR 0542 §2)
}
all = append(all, delivery.FailedFirst{Test: step + ": " + name, Words: failureWords(f)})
}
v.Failed = delivery.FailedFirst{Test: level + ": " + t.pull.Repository, Words: why}
}
if len(all) == 0 { // no result: a level said failed with nothing to name
all = append(all, delivery.FailedFirst{Test: level + ": " + t.pull.Repository, Words: "the " + level + " level did not pass"})
}
v.Failed = all[0]
v.Failures = failureLines(all)
return &v
}
// failureLines is every failure with the first line of its words, cut to MaxFailureLine, at most delivery.MaxFailures
// and then one entry saying how many more failed.
func failureLines(all []delivery.FailedFirst) []delivery.FailedFirst {
n := min(len(all), delivery.MaxFailures)
out := make([]delivery.FailedFirst, 0, n+1)
for _, f := range all[:n] {
out = append(out, delivery.FailedFirst{Test: f.Test, Machine: f.Machine, Words: clipLine(firstLine(f.Words), MaxFailureLine)})
}
if more := len(all) - n; more > 0 {
out = append(out, delivery.FailedFirst{Test: fmt.Sprintf("and %d more", more)})
}
return out
}
// clipLine cuts a line to at most n bytes on a whole character, marking the cut.
func clipLine(s string, n int) string {
if len(s) <= n {
return s
}
s = strings.ToValidUTF8(s[:n-3], "")
return s + "..."
}
@@ -177,8 +177,8 @@ func TestACandidateThatDoesNotCompileFailsTheCheckWithTheCompilersWords(t *testi
if out.Verdict.State != delivery.VerdictFail || !strings.Contains(out.Verdict.Failed.Words, "undefined: x") {
t.Fatalf("verdict = %+v, want a failure naming the compile error", out.Verdict)
}
if n := len(b.steps.tool("test")); n != 1 {
t.Errorf("%d test steps, want the first one's failure to end the check", n)
if n := len(b.steps.toolAt("test", "unit")); n != 1 {
t.Errorf("%d unit test steps, want the first one's failure to end the level", n)
}
}
@@ -49,6 +49,34 @@ func firstFailed(v delivery.Verdict) string {
return firstWordsLine(f.Words)
}
// failedLines is everything a verdict names as failed, one line each: every failed test with its first line (issue
// 1921), or what failed first when the verdict names no list (a conflict, a refused step, an older verdict).
func failedLines(v delivery.Verdict) []string {
if len(v.Failures) == 0 {
if f := firstFailed(v); f != "" {
return []string{f}
}
return nil
}
lines := make([]string, 0, len(v.Failures))
for _, f := range v.Failures {
lines = append(lines, firstFailed(delivery.Verdict{Failed: f}))
}
return lines
}
// failedCount is how many tests a verdict names as failed, the ones after delivery.MaxFailures counted too.
func failedCount(v delivery.Verdict) int {
n := len(v.Failures)
if n > delivery.MaxFailures {
var more int
if _, err := fmt.Sscanf(v.Failures[n-1].Test, "and %d more", &more); err == nil {
return delivery.MaxFailures + more
}
}
return n
}
// firstWordsLine is the first line of a failure's words: the words are several lines (issue 1901) and the status line one.
func firstWordsLine(w string) string { return firstLine(w) }
@@ -155,6 +183,9 @@ func parts(v delivery.Verdict, cs *Case, p Pull, builds []BuildNote, slowest []s
pc.head = "pass " + commits
default:
pc.head = "failed: " + firstFailed(v)
if n := failedCount(v); n > 1 {
pc.head = fmt.Sprintf("failed (%d tests): %s", n, firstFailed(v))
}
pc.commits = commits
}
var counts []string
@@ -1,92 +0,0 @@
package check
import (
"strings"
"testing"
)
// What the build module before this fix sent for two packages that failed after their tests passed (issue 1901, its third
// meeting: a try of mesh-controller kept "PASS" as its failure's words). Each is that build module's Output for the real
// `go test -json` of a throwaway package (mesh-catalog modules/mesh-build-go testdata/captured/go-test-json-pkg-*.jsonl):
// TestMain panicking in a goroutine after m.Run, TestMain exiting 2 after m.Run, and TestMain racing after m.Run.
const (
oldPassThenPanic = "PASS\npanic: the bus closed after the tests: nats: connection closed\ngoroutine 20 [running]:\n" +
"example.com/passthenfail.TestMain.func1()\n\t/tmp/fix-1901b/passthenfail/p_test.go:13 +0x25\n" +
"created by example.com/passthenfail.TestMain in goroutine 1\n\t/tmp/fix-1901b/passthenfail/p_test.go:13 +0x25\n" +
"FAIL\texample.com/passthenfail\t0.005s"
oldPassThenExit = "PASS\nFAIL\texample.com/mainexit\t0.004s"
// TestMain racing after m.Run passed, under go test -race (go-test-json-pkg-racemain.jsonl).
oldPassThenRace = "PASS\n==================\nWARNING: DATA RACE\nWrite at 0x000000836428 by goroutine 9:\n example.com/racemain.TestMain.func1()\n /tmp/fix-1901b/racemain/p_test.go:15 +0x24\nPrevious write at 0x000000836428 by main goroutine:\n example.com/racemain.TestMain()\n /tmp/fix-1901b/racemain/p_test.go:16 +0x44\n main.main()\n _testmain.go:48 +0x171\nGoroutine 9 (running) created at:\n example.com/racemain.TestMain()\n /tmp/fix-1901b/racemain/p_test.go:15 +0x37\n main.main()\n _testmain.go:48 +0x171\n==================\nFound 1 data race(s)\nFAIL\texample.com/racemain\t1.031s"
)
// A failure with no cause told apart never has "PASS" or "ok" as its words: the checker takes the panic line from the
// output, or says that the tests passed and the program failed after them (1901/AC5).
func TestAPackageFailureIsNeverWordedPass(t *testing.T) { // 1901/AC5
cases := []struct {
name, output, want string
}{
{"panic after the tests", oldPassThenPanic, "panic: the bus closed after the tests: nats: connection closed"},
{"exit after the tests", oldPassThenExit, "the tests passed, then the test program exited with a failure"},
{"race after the tests", oldPassThenRace, "WARNING: DATA RACE: Write at 0x000000836428 by goroutine 9 in example.com/racemain.TestMain.func1()"},
{"a panic deep in a long output", "PASS\npanic: the seat row vanished\n" + strings.Repeat("goroutine 7 [chan receive]:\n\tsome/frame.go:1 +0x1\n", 300) + "FAIL\tp\t2.0s", "panic: the seat row vanished"},
}
for _, c := range cases {
t.Run(c.name, func(t *testing.T) {
v := levelFailure(&target{}, "integration", Tested{Failed: []Failure{{Package: "github.com/novox/mesh-controller/cmd/mesh-controller", Output: c.output}}})
for _, l := range strings.Split(v.Failed.Words, "\n") {
if l = strings.TrimSpace(l); l == "PASS" || strings.HasPrefix(l, "ok ") || strings.HasPrefix(l, "ok\t") {
t.Errorf("1901/AC5: the words keep %q: %q", l, v.Failed.Words)
}
}
if !strings.HasPrefix(v.Failed.Words, c.want) {
t.Errorf("1901/AC5: words start %q, want %q", firstLine(v.Failed.Words), c.want)
}
if got := firstFailed(*v); !strings.Contains(got, c.want) || strings.HasSuffix(got, ": PASS") {
t.Errorf("1901/AC5: status = %q", got)
}
})
}
}
// A cause the build module told apart is cut to a line's bound, so the words keep their bound (the 1901 review's note).
func TestALongCauseIsCutToALinesBound(t *testing.T) { // 1901/AC5
f := Failure{Package: "p", Test: "TestX", Cause: strings.Repeat("c", 5000), Output: "x_test.go:1: something"}
v := levelFailure(&target{}, "unit", Tested{Failed: []Failure{f}})
if len(v.Failed.Words) > MaxWords+200 {
t.Errorf("1901/AC5: words are %d bytes, bound %d", len(v.Failed.Words), MaxWords)
}
}
// The build module on main sends a package that stopped (a timeout, an exit, a kill) as only go test's last line, with no
// cause: the words then say so, never nothing (review round 1). Each output is what that build module sends for the
// capture of the same name (go-test-json-pkg-timeout, -exiting, -killed): the package's own output, its "FAIL" line.
func TestAPackageStoppedWithOnlyItsFailLineIsStillWorded(t *testing.T) { // 1901/AC5
for _, out := range []string{"FAIL\texample.com/timeout\t2.017s", "FAIL\texample.com/exiting\t0.006s", "FAIL\texample.com/killed\t0.006s"} {
v := levelFailure(&target{}, "integration", Tested{Failed: []Failure{{Package: "example.com/x", Output: out}}})
if !strings.HasPrefix(v.Failed.Words, "the package failed with no test to name and no panic, signal or build error in its output") {
t.Errorf("1901/AC5: %q is worded %q", out, v.Failed.Words)
}
}
}
// A line of the test's own that starts as go's summary does, or that holds a runtime's words indented, is not taken for
// either (review round 1): go prints its summary in exact forms and the runtime's lines at column 0.
func TestOnlyGosExactFormsAreSummaryOrCause(t *testing.T) { // 1901/AC5
out := " x_test.go:3: ok so far\nok then: the seat is free\n panic: child retried and recovered\npanic: the real one\nFAIL\tp\t1.0s"
v := levelFailure(&target{}, "unit", Tested{Failed: []Failure{{Package: "p", Output: out}}})
if !strings.HasPrefix(v.Failed.Words, "panic: the real one") || !strings.Contains(v.Failed.Words, "ok then: the seat is free") {
t.Errorf("1901/AC5: words = %q", v.Failed.Words)
}
}
// The words keep MaxWords exactly, counting the cause, the marker and the newlines (review round 1).
func TestTheWordsKeepTheirBoundExactly(t *testing.T) { // 1901/AC5
for _, cause := range []string{"", "a cause", strings.Repeat("c", 3000)} {
for _, n := range []int{100, 170, 171, 172, 500} {
f := Failure{Package: "p", Test: "TestX", Cause: cause, Output: strings.Repeat("a log line of twenty\n", n) + "the end"}
if w := failureWords(f); len(w) > MaxWords {
t.Errorf("1901/AC5: cause %d bytes, %d lines: words are %d bytes, bound %d", len(cause), n, len(w), MaxWords)
}
}
}
}
@@ -118,8 +118,10 @@ func (st *state) checkPinsOf(ctx context.Context, t *target) verdictPtr {
return nil
}
// sameBranchPin names the branch of the tried repository that points at the tried head and that mesh-sdk also has, at the
// commit the pin names; empty when there is none.
// sameBranchPin names the mesh-sdk branch whose head the pin is, for a try: the branch of the tried repository that points
// at the tried head and that mesh-sdk also has, at the commit the pin names, first; else any branch of mesh-sdk but its
// trunk whose head the pin is, for a try asked by a bare commit that no branch of the tried repository points at (issue
// 1624; try speed S12). Empty when there is none.
func (st *state) sameBranchPin(ctx context.Context, t *target, pin string) (string, error) {
mine, err := st.j.Forge.BranchHeads(ctx, t.pull.Repository)
if err != nil {
@@ -131,9 +133,6 @@ func (st *state) sameBranchPin(ctx context.Context, t *target, pin string) (stri
same = append(same, name)
}
}
if len(same) == 0 {
return "", nil
}
sdk, err := st.j.Forge.BranchHeads(ctx, SDKRepository)
if err != nil {
return "", err
@@ -144,6 +143,16 @@ func (st *state) sameBranchPin(ctx context.Context, t *target, pin string) (stri
return name, nil
}
}
var any []string
for name, h := range sdk {
if name != "main" && strings.HasPrefix(h, pin) {
any = append(any, name)
}
}
sort.Strings(any)
if len(any) > 0 {
return any[0], nil
}
return "", nil
}
@@ -408,3 +408,22 @@ func TestAMergeCheckWithTheSameBranchPinIsStillRefusedWithoutAGroupPullRequest(t
t.Fatalf("verdict = %+v", v)
}
}
// Issue 1624 (met again on 2026-10-25; try speed S12): a try asked by a bare commit, which no branch of the tried
// repository points at, still takes a pin that is the head of a mesh-sdk branch, naming that branch; a pin that is the head
// of no mesh-sdk branch is still refused.
func TestATryByCommitTakesAPinThatIsTheHeadOfAnSDKBranch(t *testing.T) {
b, c := sameBranchBench(t, head2[:12])
b.ex.Forge.(*fakeForge).branches[repoA] = map[string]string{"main": trunkA, "feat/r6": trunkA2} // the branch moved on
if v := b.judge(t, c).Verdict; v.State != delivery.VerdictPass {
t.Fatalf("verdict = %+v", v)
}
if n := c.Notes(); len(n) != 1 || n[0] != "pin accepted: head of mesh-sdk feat/r6" {
t.Errorf("notes = %q", n)
}
b, c = sameBranchBench(t, "9aaa5c80ac5c")
b.ex.Forge.(*fakeForge).branches[repoA] = map[string]string{"main": trunkA}
if v := b.judge(t, c).Verdict; v.State != delivery.VerdictFail || !strings.Contains(v.Failed.Words, "neither on mesh-sdk's trunk") {
t.Errorf("a pin no mesh-sdk branch holds: verdict = %+v", v)
}
}
@@ -61,6 +61,9 @@ type Failure struct {
Output string `json:"output,omitempty"`
// Cause is the first line of what the test reported as its error, when the build module told it apart (issue 1901).
Cause string `json:"cause,omitempty"`
// Step is "lint" for a finding of the lint step (gofmt, go vet), which runs beside the unit level and is reported as
// its own failure, not the level's (ADR 0542 §2).
Step string `json:"step,omitempty"`
}
// TestTime is one test and the seconds it took.
@@ -0,0 +1,102 @@
package check
import (
"context"
"strings"
"testing"
"time"
"git.novox.be/novox/mesh-sdk/go/delivery"
)
const repoC = "novox/mesh-controller"
func statusOf(t *testing.T, r *rig, repo string, n int, head string) string {
t.Helper()
key, _ := delivery.CheckerOwedKey(repo, n, head)
raw, _, found, _ := r.owed.Get(context.Background(), key)
if !found {
return ""
}
o, err := delivery.DecodeCheckerOwed(raw)
if err != nil {
t.Fatal(err)
}
return o.State
}
// 1791 note 9 (issue 1921, try speed S9): two of a group's three members merge. The candidate keeps them, marked merged,
// with the same heads; their trunks moved by their own merges. No new check is asked, and the member left keeps its
// passed status, so the merge of the rest is not refused.
func TestTwoMembersMergingAskNoNewCheckAndTheThirdKeepsItsPass(t *testing.T) {
ctx, j := context.Background(), passJudge()
r := newRig(t, j)
r.setTrunks(t, map[string]string{repoA: trunkA, repoSDK: trunkS, repoC: trunkA})
a, s, c := pullOf(repoA, 7, head1), pullOf(repoSDK, 12, head2), pullOf(repoC, 3, head1)
r.ch.Candidate(ctx, "feat%2Fx", candidateOf(t, "feat/x", a, s, c))
r.ch.Wait()
if j.count() != 1 || statusOf(t, r, repoC, 3, head1) != "success" {
t.Fatalf("judged %d times, the third's status %q", j.count(), statusOf(t, r, repoC, 3, head1))
}
a.Merged, s.Merged = true, true
r.ch.Candidate(ctx, "feat%2Fx", candidateOf(t, "feat/x", a, s, c))
r.setTrunks(t, map[string]string{repoA: trunkA2, repoSDK: trunkS2, repoC: trunkA})
r.ch.TrunkMoved(ctx)
r.ch.Wait()
if j.count() != 1 {
t.Errorf("a new check was asked after two members merged: judged %d times", j.count())
}
if got := statusOf(t, r, repoC, 3, head1); got != "success" {
t.Errorf("the third member's status is %q, want success", got)
}
// The trunk of the member left moves: that is a re-judge as before.
r.setTrunks(t, map[string]string{repoA: trunkA2, repoSDK: trunkS2, repoC: trunkA2})
r.ch.TrunkMoved(ctx)
r.ch.Wait()
if j.count() != 2 {
t.Errorf("the member left's trunk moved and it was judged %d times, want again", j.count())
}
}
// 1791 note 8 (issue 1921, try speed S9): a group opened member by member writes a candidate for each partial set. When
// the group's candidate moves to new heads, the check of the older heads is stopped and ends failed, superseded, instead
// of running on or staying running; it puts no status, as the newer check owns the heads they share.
func TestACheckWhoseHeadsTheGroupNoLongerHoldsIsStoppedAndEndsSuperseded(t *testing.T) {
ctx := context.Background()
started := make(chan struct{})
j := &judgeFunc{f: func(ctx context.Context, c *Case, call int) (Outcome, error) {
if len(c.Pulls) == 1 {
close(started)
select {
case <-ctx.Done():
return Outcome{Verdict: failure("checker", ctx.Err().Error())}, nil
case <-time.After(3 * time.Second):
return Outcome{Verdict: passing()}, nil // never stopped
}
}
return Outcome{Verdict: passing()}, nil
}}
r := newRig(t, j)
partial := candidateOf(t, "feat/x", pullOf(repoA, 7, head1))
whole := candidateOf(t, "feat/x", pullOf(repoA, 7, head1), pullOf(repoSDK, 12, head2))
r.ch.Candidate(ctx, "feat%2Fx", partial)
<-started
r.ch.Candidate(ctx, "feat%2Fx", whole)
done := make(chan struct{})
go func() { r.ch.Wait(); close(done) }()
select {
case <-done:
case <-time.After(2 * time.Second):
t.Fatal("the check of the older heads was not stopped")
}
old := r.verdict(t, keyOf(t, "feat/x", partial))
if old.State != delivery.VerdictFail || !strings.Contains(old.Failed.Words, "superseded") {
t.Errorf("the older heads' verdict = %+v, want failed, superseded", old)
}
if v := r.verdict(t, keyOf(t, "feat/x", whole)); v.State != delivery.VerdictPass {
t.Errorf("the newer heads' verdict = %+v", v)
}
if got := statusOf(t, r, repoA, 7, head1); got != "success" {
t.Errorf("the shared head's status is %q, want the newer check's success", got)
}
}
@@ -67,7 +67,9 @@ type Try struct {
// Said is the try's report in words: what failed first, the tests run, kept and skipped per level, the time of each part.
Said string `json:"said,omitempty"`
// Took is the time from its start to its end, as Go writes a duration.
Took string `json:"took,omitempty"`
Took string `json:"took,omitempty"`
// Failed is every test that failed, each with its first line, at most delivery.MaxFailures and then how many more
// (issue 1921); or what else failed first (a conflict, a refused step).
Failed []string `json:"failed,omitempty"`
Levels map[string]delivery.LevelResult `json:"levels,omitempty"`
Machines []string `json:"machines,omitempty"`
@@ -339,9 +341,7 @@ func (tr *Tryer) run(ctx context.Context, key string) {
t.State = TryPass
default:
t.State = TryFail
if f := firstFailed(v); f != "" {
t.Failed = []string{f}
}
t.Failed = failedLines(v)
}
if t.Tests != "" {
t.Said += " | only the tests " + t.Tests + " (Go); nothing kept"
+1 -1
View File
@@ -3,7 +3,7 @@ module mesh-delivery-indexer
go 1.26.0
require (
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261023150321-a2c813a6b468
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024232922-99407c37df95
github.com/nats-io/nats-server/v2 v2.11.17
github.com/nats-io/nats.go v1.54.0
)
+6
View File
@@ -1,5 +1,11 @@
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261023150321-a2c813a6b468 h1:UcMee+IV2ii5E4dRZdQUaSaeh9IedTFYumuQbBiB0k0=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261023150321-a2c813a6b468/go.mod h1:ouERNpZizzwZ0HLEtAypsY9FR20FyUgUYRni6J7kVCY=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024231716-571ac521f053 h1:8Kbp3aJG3jGndDOfmGdMSDgJVEcc1hyu921hVnnlnlQ=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024231716-571ac521f053/go.mod h1:ouERNpZizzwZ0HLEtAypsY9FR20FyUgUYRni6J7kVCY=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024232250-9b408db24a58 h1:4Damtdqlf6u6nJhvMA99peDy1jmjixtcwFXJ4bWEkLk=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024232250-9b408db24a58/go.mod h1:ouERNpZizzwZ0HLEtAypsY9FR20FyUgUYRni6J7kVCY=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024232922-99407c37df95 h1:I8zbgGzRjmgYydPTWo8hmaZW/P/g+qGDUfu3E32/pkU=
git.novox.be/novox/mesh-sdk/go v0.1.14-0.20261024232922-99407c37df95/go.mod h1:ouERNpZizzwZ0HLEtAypsY9FR20FyUgUYRni6J7kVCY=
github.com/antithesishq/antithesis-sdk-go v0.7.0-default-no-op h1:Z/MZK75wC/NSrkgqeNIa7jexam9uWzhLmFTSCPI/kn0=
github.com/antithesishq/antithesis-sdk-go v0.7.0-default-no-op/go.mod h1:FQyySiasQQM8735Ddel3MRojmy4dA1IqCeyJ5jmPMbI=
github.com/google/go-tpm v0.9.8 h1:slArAR9Ft+1ybZu0lBwpSmpwhRXaa85hWtMinMyRAWo=
@@ -71,7 +71,7 @@ func (ix *Indexer) logf(format string, args ...any) {
}
// Apply takes one message of the bridge's stream FORGE into the group of its branch. A pull message adds or moves the
// pull request, a closed one or a merged message takes it out; a group left with none is deleted. A change is written
// pull request, a closed one takes it out, a merged one marks it merged; a group left with none open is deleted. A change is written
// by compare-and-set on the revision read, and read again on a conflict. A lone pull request is a group of one. A
// merged message also writes the fingerprint of each module of its repository (modules.go).
func (ix *Indexer) Apply(ctx context.Context, f delivery.Forge) error {
@@ -154,12 +154,13 @@ func (ix *Indexer) applyOpen(ctx context.Context, f delivery.Forge) error {
return nil, false, fmt.Errorf("candidates.%s holds what is not a candidate: %w", key, err)
}
kept := slices.DeleteFunc(slices.Clone(c.Pulls), func(p delivery.CandidatePull) bool {
return !named[delivery.OpenPull{Repository: p.Repository, Number: p.Number}]
// a member that merged is open no more, and stays while another member is (issue 1791)
return !p.Merged && !named[delivery.OpenPull{Repository: p.Repository, Number: p.Number}]
})
switch {
case len(kept) == len(c.Pulls):
return nil, false, nil
case len(kept) == 0:
case !anyOpen(kept):
ix.logf("%s: no pull request of the group is open, so no candidate", key)
return nil, true, nil
}
@@ -232,7 +233,16 @@ func (ix *Indexer) next(cur *delivery.Candidate, f delivery.Forge, digests map[s
if at < 0 {
return cur
}
pulls = slices.Delete(pulls, at, at+1)
if f.Kind == delivery.ForgeMerged {
// A member that merges stays, marked merged, with the head it merged at: the heads hash the group's Approve
// covers holds, so the checker starts no new check for the members still open (issue 1791; issue 1921 S9).
pulls[at].Merged = true
} else {
pulls = slices.Delete(pulls, at, at+1)
}
if !anyOpen(pulls) {
return nil // every member merged or closed: the group is done
}
} else {
entry := delivery.CandidatePull{Head: delivery.Head{Repository: f.Repository, Number: f.Number, Head: f.Head}, Modules: []string{}}
switch {
@@ -276,6 +286,11 @@ func (ix *Indexer) next(cur *delivery.Candidate, f delivery.Forge, digests map[s
return &delivery.Candidate{V: delivery.Version, Group: f.Branch, Base: base, HeadsHash: hash, Pulls: pulls, FPSet: delivery.FingerprintSet(pulls)}
}
// anyOpen says whether a group's pull requests hold one that has not merged.
func anyOpen(pulls []delivery.CandidatePull) bool {
return slices.ContainsFunc(pulls, func(p delivery.CandidatePull) bool { return !p.Merged })
}
// NoneNoPreview is why a candidate's touched module has no fingerprint of the new form: the merge preview is not
// computed in this release (novox/hq ADR 0465 clause 3; the checker comes with issue 1445).
const NoneNoPreview = "no merge preview is computed"
@@ -206,8 +206,8 @@ func TestDirsNotKnownKeepTheModulesOfTheSameHead(t *testing.T) { // 1238/AC4
}
}
func TestClosedAndMergedTakeAPullOutAndTheLastDeletesTheKey(t *testing.T) { // 1238/AC5
for _, how := range []string{"closed", "merged"} {
func TestClosedTakesAPullOutAndTheLastDeletesTheKey(t *testing.T) { // 1238/AC5 (a merge: TestAMergedMemberStays...)
for _, how := range []string{"closed"} {
ix, st, _ := newIndexer(t)
ctx := context.Background()
_ = ix.Apply(ctx, pull("novox/mesh-delivery", 12, "feat/x", h1, "modules/a"))
@@ -387,3 +387,70 @@ func TestAnOpenMessageDropsEveryCandidateItDoesNotName(t *testing.T) { // 1279/A
t.Errorf("after an empty open message the candidates hold %v", st.vals)
}
}
// 1791 (issue 1921, try speed S9): a member of a group that merges while another is open stays in the candidate, marked
// merged, with its head: the heads hash the Approve covers holds, so no new check starts for the members left. An open
// message does not drop it, a close of the last open member deletes the group, and the last merge does too.
func TestAMergedMemberStaysInItsGroupUntilEveryMemberMerged(t *testing.T) {
merged := func(repo string, n int, head string) delivery.Forge {
return delivery.Forge{V: delivery.Version, Kind: delivery.ForgeMerged, Repository: repo, Branch: "feat/x", Number: n, Head: head, Merge: h3, Root: "r", Trees: map[string]string{"a": "t"}}
}
ix, st, _ := newIndexer(t)
ctx := context.Background()
for _, f := range []delivery.Forge{pull("novox/mesh-delivery", 12, "feat/x", h1, "modules/a"), pull("novox/mesh-catalog", 300, "feat/x", h2, "modules/c"),
pull("novox/mesh-sdk", 7, "feat/x", h3)} {
if err := ix.Apply(ctx, f); err != nil {
t.Fatal(err)
}
}
before, _ := read(t, st, "feat/x")
if err := ix.Apply(ctx, merged("novox/mesh-sdk", 7, h3)); err != nil {
t.Fatal(err)
}
if err := ix.Apply(ctx, merged("novox/mesh-delivery", 12, h1)); err != nil {
t.Fatal(err)
}
c, ok := read(t, st, "feat/x")
if !ok || len(c.Pulls) != 3 || c.HeadsHash != before.HeadsHash {
t.Fatalf("after two merges the candidate = %+v (kept %v), want the three members and the same heads hash", c, ok)
}
for _, p := range c.Pulls {
if want := p.Repository != "novox/mesh-catalog"; p.Merged != want {
t.Errorf("%s#%d merged = %v, want %v", p.Repository, p.Number, p.Merged, want)
}
}
// The open message names only the member still open: the merged ones stay.
if err := ix.Apply(ctx, open(delivery.OpenPull{Repository: "novox/mesh-catalog", Number: 300})); err != nil {
t.Fatal(err)
}
if c, _ := read(t, st, "feat/x"); len(c.Pulls) != 3 || c.HeadsHash != before.HeadsHash {
t.Errorf("the open message dropped a merged member: %+v", c)
}
// The last open member merges: the group is done.
if err := ix.Apply(ctx, merged("novox/mesh-catalog", 300, h2)); err != nil {
t.Fatal(err)
}
if _, ok := read(t, st, "feat/x"); ok {
t.Error("every member merged and the group is still a candidate")
}
// A close of the last open member deletes the group too.
ix, st, _ = newIndexer(t)
_ = ix.Apply(ctx, pull("novox/mesh-delivery", 12, "feat/x", h1, "modules/a"))
_ = ix.Apply(ctx, pull("novox/mesh-catalog", 300, "feat/x", h2, "modules/c"))
_ = ix.Apply(ctx, merged("novox/mesh-delivery", 12, h1))
closed := pull("novox/mesh-catalog", 300, "feat/x", h2)
closed.Action = delivery.PullClosed
if err := ix.Apply(ctx, closed); err != nil {
t.Fatal(err)
}
if _, ok := read(t, st, "feat/x"); ok {
t.Error("the last open member closed and the group, holding only a merged one, is still a candidate")
}
// A lone pull request that merges leaves no group.
ix, st, _ = newIndexer(t)
_ = ix.Apply(ctx, pull("novox/mesh-delivery", 12, "feat/x", h1, "modules/a"))
_ = ix.Apply(ctx, merged("novox/mesh-delivery", 12, h1))
if _, ok := read(t, st, "feat/x"); ok {
t.Error("a lone merged pull request is still a candidate")
}
}
@@ -1,213 +0,0 @@
package main
import (
"encoding/json"
"fmt"
"slices"
stdio "git.novox.be/novox/mesh-sdk/go"
"git.novox.be/novox/mesh-sdk/go/asks"
"git.novox.be/novox/mesh-sdk/go/delivery"
)
// The assigner's asks reach the operator (issue 1906). The assigner writes its asks (the hold of a fingerprint, the restore
// of a data module, the backup before an apply and a move of a module of policy record) in its bucket asks; until now the
// teller only filled them from an answer given in the view, so an ask nobody opened the view for reached nobody: on
// 2026-10-25 seven modules stayed held on every machine and the operator was never asked. Now the teller raises each on the
// operator channel as it raises the gate's merge asks: as the asker mesh-delivery-teller, once per router id, through the same
// records, the same wait for room and the same warrants. The operator's answer, from the phone or from the view, is written
// under the ask's own key in the bucket answers, which the assigner reads (mesh-delivery-assigner pass.go answeredWith); none
// of these asks needs proof of who answered (mesh-sdk go/delivery AskNeedsProof).
// mostAssignerOpen is how many of the assigner's asks the teller keeps open at the router at once. The router keeps at most
// asks.MostOpen open asks of one asker, and the teller asks both the gate's merge asks and the assigner's as one asker, so a burst
// of holds (seven at once on 2026-10-25) would take all the room and a merge Approve could not be asked; one place is always
// left for a merge ask. An assigner ask beyond these waits in the wait for room and is asked when one of them ends.
const mostAssignerOpen = asks.MostOpen - 1
// assignerOpenRecord is the key of the router ids of the assigner's asks the teller asked.
const assignerOpenRecord = "assigner.open"
// assignerOpen is the router ids of the assigner's asks open now: those asked whose ask is neither answered nor ended. An id
// that ended is dropped from the record as it is read.
func (t *Teller) assignerOpen() ([]string, *stdio.StateEntry) {
e, err := t.Store.Get(assignerOpenRecord)
if err != nil || e == nil {
return nil, e
}
var ids []string
if err := json.Unmarshal(e.Value, &ids); err != nil {
t.logf("the record %s cannot be read: %v", assignerOpenRecord, err)
return nil, e
}
return slices.DeleteFunc(ids, func(id string) bool {
k, _, err := t.keptOf(id)
return err == nil && !open(k)
}), e
}
// assignerRoom says whether another of the assigner's asks may be asked now.
func (t *Teller) assignerRoom() bool {
ids, _ := t.assignerOpen()
return len(ids) < mostAssignerOpen
}
// assignerAsked keeps that the assigner's ask with this router id is asked.
func (t *Teller) assignerAsked(id string) {
ids, e := t.assignerOpen()
if !slices.Contains(ids, id) {
ids = append(ids, id)
}
var err error
if e == nil {
_, err = t.Store.Create(assignerOpenRecord, ids)
} else {
_, err = t.Store.Update(assignerOpenRecord, ids, e.Revision)
}
if err != nil {
t.logf("the record %s could not be kept: %v", assignerOpenRecord, err)
}
}
// waitForAssignerRoom puts an assigner ask in the wait for room, once.
func (t *Teller) waitForAssignerRoom(key, id string) {
ws, rec, err := t.waiting()
if err != nil {
t.logf("%v", err)
return
}
if slices.ContainsFunc(ws, func(x waitingAsk) bool { return x.ID == id }) {
return
}
if t.keepWaiting(append(ws, waitingAsk{Key: key, ID: id}), rec) == nil {
t.logf("%s: the assigner's ask %s waits: %d of the assigner's asks are open, the most the teller keeps open so a merge ask always has room", key, id, mostAssignerOpen)
}
}
// isAssignerKey says whether an ask key is one of the assigner's, not the gate's merge ask.
func isAssignerKey(key string) bool {
k, err := delivery.ParseAnyAskKey(key)
return err == nil && k.Kind != delivery.AskKindMerge
}
// assignerAsk reads the assigner's ask at key: found false when the key holds none.
func (t *Teller) assignerAsk(key string) (delivery.AssignerAsk, bool, error) {
if t.Assigner == nil {
return delivery.AssignerAsk{}, false, fmt.Errorf("the teller reads no assigner asks")
}
raw, found, err := t.Assigner.Get(key)
if err != nil || !found {
return delivery.AssignerAsk{}, false, err
}
a, err := delivery.DecodeAssignerAsk(key, raw)
if err != nil {
return delivery.AssignerAsk{}, false, fmt.Errorf("the assigner's key %s holds no ask: %w", key, err)
}
return a, true, nil
}
// raiseAssignerAsk asks the operator the assigner's ask at key, once per router id: an ask answered already (from the view or
// the phone) is not asked; an earlier asking still open is taken back first; an asking already made is only caught up with
// at the router, and one that ended is not asked again, for the router drops a second ask under an id it keeps.
func (t *Teller) raiseAssignerAsk(key string) {
a, found, err := t.assignerAsk(key)
if err != nil {
t.logf("%s: the ask is not raised: %v", key, err)
return
}
if !found {
return
}
id := a.Ask.ID
if raw, _, has, err := t.Answers.Get(delivery.AnswerKey(key)); err != nil {
t.logf("%s: the answers could not be read, so the ask is not raised: %v", key, err)
return
} else if has && answeredAsking(raw) == id {
return
}
prior, grec, err := t.seenAt(gateRecordKey(key))
if err != nil {
t.logf("%s: %v", key, err)
return
}
if grec != nil && prior.ID != id {
t.cancelOpen(key, prior.ID) // a new asking of the same key: the earlier one is taken back
}
k, _, err := t.keptOf(id)
if err != nil {
t.logf("%s: %v", key, err)
return
}
if k != nil {
if grec == nil || prior.ID != id {
_ = t.note(id, key, 0, grec)
}
if open(k) {
t.catchUp(id)
}
return
}
if err := t.note(id, key, 0, grec); err != nil {
t.logf("%s: the record of the ask %s could not be kept, so it is not raised: %v", key, id, err)
return
}
if !t.assignerRoom() {
t.waitForAssignerRoom(key, id)
return
}
if err := t.Asker.Ask(a.Ask); err != nil {
t.logf("%s: the ask %s was not asked: %v", key, id, err)
if kept, _, _ := t.keptOf(id); open(kept) {
_ = t.Store.Delete(keptKey(id)) // kept but not published: forgotten, so the next writing of the key asks it
}
return
}
t.assignerAsked(id)
t.logf("%s: the assigner's ask %s is asked of the operator (%s)", key, id, a.Ask.Headline)
}
// OnAssignerAskGone handles a key the assigner took away from its asks (a hold it lifted by itself): the open router ask of
// its latest asking is taken back, so the operator is not asked about what no longer holds.
func (t *Teller) OnAssignerAskGone(key string) {
t.mu.Lock()
defer t.mu.Unlock()
if !isAssignerKey(key) {
t.logf("the assigner's key %s was taken away and is no ask key of the assigner, and is left", key)
return
}
prior, grec, err := t.seenAt(gateRecordKey(key))
if err != nil || grec == nil {
if err != nil {
t.logf("%s: %v", key, err)
}
return
}
t.cancelOpen(key, prior.ID)
}
// answerAssigner is the act of a warrant for one of the assigner's asks: the ask must still stand at its key under the asking
// and the words the operator answered, and then the answer is written under the ask's key, once. Its options bind no act:
// the assigner reads the option chosen (asker.Client.Take has checked the option, its level and the ask's digest).
func (t *Teller) answerAssigner(key string, o asks.Option, w asks.Warrant) error {
a, found, err := t.assignerAsk(key)
if err != nil {
return err
}
if !found {
t.logf("%s: the ask %s was answered and the assigner has taken it away: the answer changes nothing", key, w.Ask)
return nil
}
if a.Ask.ID != w.Ask {
t.logf("%s: the ask %s was answered, and the assigner has asked again under %s: an answer to an earlier asking changes nothing", key, w.Ask, a.Ask.ID)
return nil
}
if a.Ask.Digest() != w.AskDigest {
return fmt.Errorf("the assigner's ask %s is no longer the ask the operator answered (its digest is %s, the warrant's %s)", key, a.Ask.Digest(), w.AskDigest)
}
body, err := delivery.EncodeAnswer(delivery.Answer{Option: o.ID, Warrant: warrantID(w), Words: w.Says()})
if err != nil {
return err
}
return t.writeAnswer(key, w.Ask, body, w.Says())
}
@@ -1,209 +0,0 @@
package main
import (
"encoding/json"
"strings"
"testing"
"git.novox.be/novox/mesh-sdk/go/asks"
"git.novox.be/novox/mesh-sdk/go/delivery"
)
// 1906/H1: given the assigner's hold ask in its bucket and no answer, when the teller hears the key, then it raises the ask on
// the operator channel as itself, unchanged and once (the key heard again, as a watch replays it, asks nothing more).
func TestAnAssignerAskIsRaisedOnTheOperatorChannelOnce(t *testing.T) {
b := newBed(t)
assigner, _ := withView(b)
key, value, ask := holdAsk(t)
assigner[key] = value
b.t.OnAssignerAsk(key)
b.t.OnAssignerAsk(key)
asked := b.askedOf(asks.AskSubject(delivery.TellerModule))
if len(asked) != 1 {
t.Fatalf("published %d asks, want one; log:\n%s", len(asked), b.logs.String())
}
var got asks.Ask
if err := json.Unmarshal(asked[0].body, &got); err != nil {
t.Fatal(err)
}
if got.ID != ask.ID || got.Digest() != ask.Digest() {
t.Errorf("the router's ask %+v differs from the assigner's", got)
}
}
// 1906/H1: without the view's bucket the assigner's asks are still raised.
func TestAnAssignerAskIsRaisedWithoutTheView(t *testing.T) {
b := newBed(t)
assigner := memGate{}
b.t.Assigner = assigner
key, value, _ := holdAsk(t)
assigner[key] = value
b.t.OnAssignerAsk(key)
if n := len(b.askedOf(asks.AskSubject(delivery.TellerModule))); n != 1 {
t.Fatalf("published %d asks, want one; log:\n%s", n, b.logs.String())
}
}
// 1906/H1: the operator's answer on the phone comes back by the same path as the view's: written under the ask's own key in
// the bucket answers, naming the asking, once; a second warrant writes nothing more and nothing is said as a fault.
func TestTheOperatorsAnswerToAnAssignerAskIsWrittenUnderItsKey(t *testing.T) {
b := newBed(t)
assigner, _ := withView(b)
key, value, ask := holdAsk(t)
assigner[key] = value
b.t.OnAssignerAsk(key)
w := ackFor(ask, "release")
b.deliver(t, w)
raw, ok := b.answers.written[delivery.AnswerKey(key)]
if !ok {
t.Fatalf("no answer under %s; log:\n%s", key, b.logs.String())
}
if a, err := delivery.DecodeAnswer(raw); err != nil || a.Option != "release" || answeredAsking(raw) != ask.ID {
t.Errorf("answer = %+v, %v", a, err)
}
b.deliver(t, w)
if b.answers.writes != 1 || len(b.conds) != 0 {
t.Errorf("%d writes, conditions %+v", b.answers.writes, b.conds)
}
}
// 1906/H1: an answer to an asking the assigner has since taken away or asked again changes nothing.
func TestAnAnswerToAnAssignerAskTakenAwayWritesNothing(t *testing.T) {
b := newBed(t)
assigner, _ := withView(b)
key, value, ask := holdAsk(t)
assigner[key] = value
b.t.OnAssignerAsk(key)
delete(assigner, key)
b.deliver(t, ackFor(ask, "release"))
if b.answers.writes != 0 {
t.Errorf("an answer to a withdrawn ask was written")
}
}
// 1906/H1, H2: an ask answered in the view is taken back from the phone, and an ask the assigner took away (a hold it lifted by
// itself) is taken back too; an ask answered before it is heard is never raised.
func TestAnAssignerAskAnsweredInTheViewOrTakenAwayIsTakenBack(t *testing.T) {
cancelled := func(b *bed, id string) bool {
for _, p := range b.askedOf(asks.CancelSubject(delivery.TellerModule)) {
if string(p.body) == `{"id":"`+id+`"}` {
return true
}
}
return false
}
b := newBed(t)
assigner, view := withView(b)
key, value, ask := holdAsk(t)
assigner[key] = value
b.t.OnAssignerAsk(key)
view[key+".view"] = viewAnswer(t, "release", "")
b.t.OnViewAnswer(key + ".view")
if !cancelled(b, ask.ID) {
t.Errorf("the ask answered in the view was not taken back; log:\n%s", b.logs.String())
}
b = newBed(t)
assigner, _ = withView(b)
assigner[key] = value
b.t.OnAssignerAsk(key)
delete(assigner, key)
b.t.OnAssignerAskGone(key)
if !cancelled(b, ask.ID) {
t.Errorf("the ask taken away was not taken back; log:\n%s", b.logs.String())
}
b = newBed(t)
assigner, view = withView(b)
assigner[key] = value
view[key+".view"] = viewAnswer(t, "keep", "")
b.t.OnAssignerAsk(key)
if n := len(b.askedOf(asks.AskSubject(delivery.TellerModule))); n != 0 {
t.Errorf("an ask answered in the view was raised %d times", n)
}
}
// 1906/H1: an assigner ask refused for want of room waits and is asked again when another of the teller's asks ends, as a merge
// ask does; a refusal for its words is said as a condition about a delivery ask, not a merge.
func TestAnAssignerAskWaitsForRoomAndARefusalIsSaidAsADeliveryAsk(t *testing.T) {
b := newBed(t)
assigner, _ := withView(b)
_, first := b.askOf(t, 1, 1)
b.askOf(t, 2, 2)
b.askOf(t, 3, 3)
key, value, ask := holdAsk(t)
assigner[key] = value
b.t.OnAssignerAsk(key)
b.refuse(t, ask.ID, roomRefusal)
b.deliver(t, warrantFor(first.Ask, "approve"))
got := b.askedIDs()
if len(got) != 5 || got[4] != ask.ID {
t.Fatalf("asks = %v, want the assigner's %s asked again", got, ask.ID)
}
b = newBed(t)
assigner, _ = withView(b)
assigner[key] = value
b.t.OnAssignerAsk(key)
b.refuse(t, ask.ID, "the ask has no headline")
if len(b.conds) != 1 || b.conds[0].Subject.Scope != "delivery" || strings.Contains(b.conds[0].Explanation, "merge") {
t.Errorf("conditions = %+v", b.conds)
}
}
// ackFor is the router's warrant for an option of level acknowledge, as the assigner's asks offer.
func ackFor(a asks.Ask, option string) asks.Warrant {
w := warrantFor(a, option)
w.Level, w.Label = asks.Acknowledge, option
return w
}
// holdAskOf is the assigner's hold ask of module plex at the n-th fingerprint, shaped as holdAsk.
func holdAskOf(t *testing.T, n int) (string, []byte, asks.Ask) {
t.Helper()
key, err := delivery.HoldAskKey("plex", "src2:"+strings.Repeat(string(rune('0'+n)), 64))
if err != nil {
t.Fatal(err)
}
_, _, ask := holdAsk(t)
ask.ID = delivery.AskID(key)
value, err := delivery.EncodeAssignerAsk(key, delivery.AssignerAsk{Ask: ask})
if err != nil {
t.Fatal(err)
}
return key, value, ask
}
// 1906/H5: given seven holds at once (as on 2026-10-25), when they are raised, then at most two are open at the router, a merge
// ask is still asked, and each hold answered lets the next waiting one be asked.
func TestABurstOfHoldsLeavesRoomForAMergeAsk(t *testing.T) {
b := newBed(t)
assigner, _ := withView(b)
var held []asks.Ask
for n := 0; n < 7; n++ {
key, value, ask := holdAskOf(t, n)
assigner[key] = value
b.t.OnAssignerAsk(key)
held = append(held, ask)
}
if got := b.askedIDs(); len(got) != mostAssignerOpen || got[0] != held[0].ID || got[1] != held[1].ID {
t.Fatalf("asked %v, want the first %d holds only; log:\n%s", got, mostAssignerOpen, b.logs.String())
}
_, merge := b.askOf(t, 1, 1)
if got := b.askedIDs(); len(got) != mostAssignerOpen+1 || got[mostAssignerOpen] != merge.Ask.ID {
t.Fatalf("asked %v, want the merge ask asked beside the holds", got)
}
if len(b.dropped) != 0 || len(b.conds) != 0 {
t.Errorf("router dropped %v, conditions %+v", b.dropped, b.conds)
}
b.deliver(t, ackFor(held[0], "release"))
got := b.askedIDs()
if len(got) != mostAssignerOpen+2 || got[len(got)-1] != held[2].ID {
t.Fatalf("asked %v after a hold was answered, want the third hold asked next; log:\n%s", got, b.logs.String())
}
// The merge ask answered frees its own place, and no hold takes it.
b.deliver(t, warrantFor(merge.Ask, "approve"))
if got := b.askedIDs(); len(got) != mostAssignerOpen+2 {
t.Errorf("asked %v after the merge ask was answered, want no further hold", got)
}
}
@@ -524,9 +524,8 @@ func TestTellerFillsAnAskFromTheViewsAnswerOnce(t *testing.T) { // 1297/B10 AC
if a, err := delivery.DecodeAnswer(e.Value()); err != nil || a.Option != "release" || answeredAsking(e.Value()) != ask.ID {
t.Fatalf("answer %+v err %v", a, err)
}
// The view's answer written again with the same option changes nothing under the ask (another option, later, is the
// operator changing their mind: TestTellerTakesAChangedHoldAnswerFromTheView, issue 1906 note 23).
if _, err := viewKV.Put(b.ctx, key+".view", viewAnswer(t, "release", "")); err != nil {
// The view's answer written again changes nothing under the ask.
if _, err := viewKV.Put(b.ctx, key+".view", viewAnswer(t, "keep", "")); err != nil {
t.Fatal(err)
}
time.Sleep(time.Second)
@@ -546,148 +545,3 @@ func TestTellerFillsAnAskFromTheViewsAnswerOnce(t *testing.T) { // 1297/B10 AC
t.Error("the merge's ask was answered from a view key")
}
}
// TestTellerRaisesAnAssignerAskAndWritesTheOperatorsAnswer starts the program against the bus: the assigner's hold ask, put in
// its bucket, is on the operator channel's queue as the teller's ask; the router's warrant for "release" is written under the
// ask's key, where the assigner reads it; a second ask the assigner takes away (a hold it lifted) is cancelled at the router
// (1906/H1, H2 on the real bus).
func TestTellerRaisesAnAssignerAskAndWritesTheOperatorsAnswer(t *testing.T) {
now = time.Now().UTC()
b := startFull(t)
assigner := b.as(t, "n1.mesh-delivery-assigner")
broker := filepath.Join(t.TempDir(), "broker")
cred, _ := json.Marshal(credential{URL: b.srv.ClientURL(), Node: "n1", Module: "mesh-delivery-teller", User: "n1.mesh-delivery-teller", Password: "pw"})
if err := os.WriteFile(broker, cred, 0o600); err != nil {
t.Fatal(err)
}
env := map[string]string{"MESH_BROKER_FILE": broker}
logs := &syncBuffer{}
rn := startRunner(b.ctx, func(c context.Context) error { return run(c, func(k string) string { return env[k] }, logs) })
t.Cleanup(func() { rn.finish(t, 10*time.Second) })
key, value, ask := holdAsk(t)
askKV, err := assigner.KeyValue(b.ctx, assignerAsks)
if err != nil {
t.Fatal(err)
}
time.Sleep(500 * time.Millisecond) // the program hears warrants and watches
if _, err := askKV.Put(b.ctx, key, value); err != nil {
t.Fatal(err)
}
seat, err := b.js.Stream(b.ctx, "SEAT_OPERATOR_CHANNEL")
if err != nil {
t.Fatal(err)
}
var raw *jetstream.RawStreamMsg
for deadline := time.Now().Add(10 * time.Second); time.Now().Before(deadline); time.Sleep(50 * time.Millisecond) {
if raw, err = seat.GetLastMsgForSubject(b.ctx, seatAsk); err == nil {
break
}
}
if raw == nil {
t.Fatalf("the assigner's ask is not on the seat's queue: %v\n%s", err, logs.String())
}
var asked asks.Ask
if err := json.Unmarshal(raw.Data, &asked); err != nil || asked.ID != ask.ID || asked.Digest() != ask.Digest() {
t.Fatalf("the ask on the queue is %+v (%v), want the assigner's", asked, err)
}
w, _ := json.Marshal(ackFor(ask, "release"))
if err := b.ctl.Publish(seatDecided, w); err != nil {
t.Fatal(err)
}
answers, err := b.js.KeyValue(b.ctx, tellerAnswers)
if err != nil {
t.Fatal(err)
}
var e jetstream.KeyValueEntry
for deadline := time.Now().Add(10 * time.Second); time.Now().Before(deadline); time.Sleep(50 * time.Millisecond) {
if e, err = answers.Get(b.ctx, delivery.AnswerKey(key)); err == nil {
break
}
}
if e == nil {
t.Fatalf("no answer under %s: %v\n%s", key, err, logs.String())
}
if a, err := delivery.DecodeAnswer(e.Value()); err != nil || a.Option != "release" || answeredAsking(e.Value()) != ask.ID {
t.Fatalf("answer %+v err %v", a, err)
}
key2, err := delivery.HoldAskKey("plex", "src2:"+strings.Repeat("3", 64))
if err != nil {
t.Fatal(err)
}
ask2 := ask
ask2.ID = delivery.AskID(key2)
value2, err := delivery.EncodeAssignerAsk(key2, delivery.AssignerAsk{Ask: ask2})
if err != nil {
t.Fatal(err)
}
if _, err := askKV.Put(b.ctx, key2, value2); err != nil {
t.Fatal(err)
}
for deadline := time.Now().Add(10 * time.Second); time.Now().Before(deadline); time.Sleep(50 * time.Millisecond) {
if strings.Contains(logs.String(), ask2.ID+" is asked") {
break
}
}
if err := askKV.Delete(b.ctx, key2); err != nil {
t.Fatal(err)
}
for deadline := time.Now().Add(10 * time.Second); time.Now().Before(deadline); time.Sleep(50 * time.Millisecond) {
if raw, err = seat.GetLastMsgForSubject(b.ctx, seatCancel); err == nil {
break
}
}
if err != nil || string(raw.Data) != `{"id":"`+ask2.ID+`"}` {
t.Errorf("the ask taken away was not cancelled: %v %v\n%s", raw, err, logs.String())
}
}
// TestTellerTakesAChangedHoldAnswerFromTheView starts the program against the bus: a hold ask answered "keep" earlier, and the
// view's answer "release" written now, replace the answer of that asking with "release" (1906/N23 on the real bus).
func TestTellerTakesAChangedHoldAnswerFromTheView(t *testing.T) {
now = time.Now().UTC()
b := startFull(t)
assigner, view := b.as(t, "n1.mesh-delivery-assigner"), b.as(t, "n1.mesh-delivery-view")
broker := filepath.Join(t.TempDir(), "broker")
cred, _ := json.Marshal(credential{URL: b.srv.ClientURL(), Node: "n1", Module: "mesh-delivery-teller", User: "n1.mesh-delivery-teller", Password: "pw"})
if err := os.WriteFile(broker, cred, 0o600); err != nil {
t.Fatal(err)
}
key, value, ask := holdAsk(t)
answers, err := b.js.KeyValue(b.ctx, tellerAnswers)
if err != nil {
t.Fatal(err)
}
kept, _ := delivery.EncodeAnswer(delivery.Answer{Option: "keep", Warrant: ask.ID + "@" + time.Now().Add(-time.Hour).UTC().Format(time.RFC3339Nano), Words: "the operator chose keep"})
if _, err := answers.Put(b.ctx, delivery.AnswerKey(key), kept); err != nil {
t.Fatal(err)
}
askKV, err := assigner.KeyValue(b.ctx, assignerAsks)
if err != nil {
t.Fatal(err)
}
if _, err := askKV.Put(b.ctx, key, value); err != nil {
t.Fatal(err)
}
env := map[string]string{"MESH_BROKER_FILE": broker}
logs := &syncBuffer{}
rn := startRunner(b.ctx, func(c context.Context) error { return run(c, func(k string) string { return env[k] }, logs) })
t.Cleanup(func() { rn.finish(t, 10*time.Second) })
viewKV, err := view.KeyValue(b.ctx, viewAnswers)
if err != nil {
t.Fatal(err)
}
time.Sleep(500 * time.Millisecond) // the program is watching
if _, err := viewKV.Put(b.ctx, key+".view", viewAnswer(t, "release", "it was a flake")); err != nil {
t.Fatal(err)
}
for deadline := time.Now().Add(10 * time.Second); time.Now().Before(deadline); time.Sleep(50 * time.Millisecond) {
if e, err := answers.Get(b.ctx, delivery.AnswerKey(key)); err == nil {
if a, err := delivery.DecodeAnswer(e.Value()); err == nil && a.Option == "release" && answeredAsking(e.Value()) == ask.ID {
return
}
}
}
t.Fatalf("the kept hold's answer was not changed to release\n%s", logs.String())
}
@@ -5,7 +5,6 @@ import (
"encoding/json"
"errors"
"fmt"
"time"
"github.com/nats-io/nats.go/jetstream"
@@ -108,15 +107,3 @@ func (g kvGate) Get(key string) ([]byte, bool, error) {
}
return e.Value(), true, nil
}
// GetAt is Get with the time the value was written.
func (g kvGate) GetAt(key string) ([]byte, time.Time, bool, error) {
e, err := g.kv.Get(g.ctx, key)
if errors.Is(err, jetstream.ErrKeyNotFound) {
return nil, time.Time{}, false, nil
}
if err != nil {
return nil, time.Time{}, false, err
}
return e.Value(), e.Created(), true, nil
}
@@ -141,20 +141,14 @@ func run(ctx context.Context, getenv func(string) string, logOut io.Writer) erro
// there is said and leaves the teller without that path rather than without the gate's asks (a wait for it needs a timer).
failed := make(chan error, 2)
assignerKV, viewKV := optionalBucket(ctx, js, delivery.AssignerAsksBucket, logger), optionalBucket(ctx, js, delivery.ViewAnswersBucket, logger)
// The assigner's asks are raised on the operator channel whether or not the view's answers can be read (issue 1906).
if viewKV != nil {
t.View = kvGate{ctx: ctx, kv: viewKV}
}
if assignerKV != nil {
t.Assigner = kvGate{ctx: ctx, kv: assignerKV}
go func() {
failed <- watchAsks(ctx, assignerKV, func(key string, _ uint64, _ []byte) { t.OnAssignerAsk(key) }, t.OnAssignerAskGone, nil)
}()
}
if assignerKV != nil && viewKV != nil {
t.Assigner, t.View = kvGate{ctx: ctx, kv: assignerKV}, kvGate{ctx: ctx, kv: viewKV}
go func() {
failed <- watchAsks(ctx, viewKV, func(key string, _ uint64, _ []byte) { t.OnViewAnswer(key) }, func(string) {}, nil)
}()
go func() {
failed <- watchAsks(ctx, assignerKV, func(key string, _ uint64, _ []byte) { t.OnAssignerAsk(key) }, func(string) {}, nil)
}()
}
go func() {
if err := watchAsks(ctx, gate, t.OnAsk, t.OnAskGone, t.OnCurrent); err != nil {
@@ -176,7 +170,7 @@ func run(ctx context.Context, getenv func(string) string, logOut io.Writer) erro
func optionalBucket(ctx context.Context, js jetstream.JetStream, name string, logger *log.Logger) jetstream.KeyValue {
kv, err := js.KeyValue(ctx, name)
if err != nil {
logger.Printf("the bucket %s cannot be read (%v): the asks and answers it holds are not carried until the teller is started again with it", name, err)
logger.Printf("the bucket %s cannot be read (%v): the view's answers fill no ask until the teller is started again with it", name, err)
return nil
}
return kv
@@ -340,9 +340,6 @@ func (t *Teller) Answer(o asks.Option, w asks.Warrant) error {
if srec == nil {
return fmt.Errorf("the teller holds no gate key for the ask %s", w.Ask)
}
if isAssignerKey(s.Key) {
return t.answerAssigner(s.Key, o, w)
}
raw, found, err := t.Gate.Get(s.Key)
if err != nil {
return fmt.Errorf("the gate's ask %s could not be read: %w", s.Key, err)
@@ -370,31 +367,26 @@ func (t *Teller) Answer(o asks.Option, w asks.Warrant) error {
if err != nil {
return err
}
return t.writeAnswer(s.Key, w.Ask, body, w.Says())
}
// writeAnswer writes the answer to the asking askID under the ask's key by compare-and-set: created when the key has no answer;
// updated at the revision read when it holds an earlier asking's answer (the reader ignores that one); left when it holds this
// asking's.
func (t *Teller) writeAnswer(key, askID string, body []byte, words string) error {
akey := delivery.AnswerKey(key)
// Compare-and-set: created when the key has no answer; updated at the revision read when it holds an earlier asking's
// answer (the gate ignores that one); left when it holds this asking's.
akey := delivery.AnswerKey(s.Key)
cur, crev, found, err := t.Answers.Get(akey)
if err != nil {
return fmt.Errorf("the answer under %s could not be read: %w", key, err)
return fmt.Errorf("the answer under %s could not be read: %w", s.Key, err)
}
if found && answeredAsking(cur) == askID {
t.logf("%s: an answer is there already, and is not written again", key)
if found && answeredAsking(cur) == w.Ask {
t.logf("%s: an answer is there already, and is not written again", s.Key)
return nil
}
written, err := t.Answers.Write(akey, body, crev, found)
if err != nil {
return fmt.Errorf("the answer under %s could not be written: %w", key, err)
return fmt.Errorf("the answer under %s could not be written: %w", s.Key, err)
}
if !written {
t.logf("%s: another writer answered first, and the answer is not written again", key)
t.logf("%s: another writer answered first, and the answer is not written again", s.Key)
return nil
}
t.logf("%s: %s", key, words)
t.logf("%s: %s", s.Key, w.Says())
return nil
}
@@ -431,10 +423,10 @@ func (t *Teller) saidOnce(h asker.Handled, w asks.Warrant, err error) {
t.clear(w.Ask)
case h == asker.Refused || (h == asker.Acted && err != nil):
t.logf("the warrant of %s does not hold, and no answer is written: %v", w.Ask, err)
t.raise(w.Ask, t.kindOf(w.Ask).answerHeadline, "its answer did not hold here: "+oneLine(err))
t.raise(w.Ask, "A merge Approve did not hold", "its answer did not hold here: "+oneLine(err))
case h == asker.Recorded && w.Outcome == asks.OutcomeRefused:
t.logf("the router refused the ask %s: %s", w.Ask, w.Words)
t.raise(w.Ask, t.kindOf(w.Ask).refusedHeadline, "the router refused it: "+w.Words)
t.raise(w.Ask, "A merge ask was refused", "the router refused it: "+w.Words)
case err != nil:
t.logf("the warrant of %s: %v", w.Ask, err)
}
@@ -454,30 +446,6 @@ func oneLine(err error) string {
// conditionKey is the key of the condition about an ask.
func conditionKey(id string) string { return delivery.TellerModule + ".merge-ask." + id }
// askWords are the words the teller's conditions use for one kind of ask: the gate's merge ask, or one of the assigner's.
type askWords struct {
scope, noun, needs, nothing, resolved string
answerHeadline, refusedHeadline, unaskedHeadline string
}
var (
mergeWords = askWords{scope: "merge", noun: "The merge ask ", needs: "to ask for the Approve again", nothing: "Nothing was merged.",
resolved: "A good answer to the merge ask was written.", answerHeadline: "A merge Approve did not hold",
refusedHeadline: "A merge ask was refused", unaskedHeadline: "A merge ask could not be asked"}
assignerWords = askWords{scope: "delivery", noun: "The delivery ask ", needs: "to answer it in the delivery view",
nothing: "Nothing was changed on a machine.", resolved: "A good answer to the delivery ask was written.",
answerHeadline: "An answer to a delivery ask did not hold", refusedHeadline: "A delivery ask was refused",
unaskedHeadline: "A delivery ask could not be asked"}
)
// kindOf is the words for the ask with this router id, by the key it was asked for.
func (t *Teller) kindOf(askID string) askWords {
if isAssignerKey(t.condID(askID)) {
return assignerWords
}
return mergeWords
}
func (t *Teller) raise(askID, headline, why string) {
if t.Say == nil {
return
@@ -489,12 +457,11 @@ func (t *Teller) raise(askID, headline, why string) {
}
t.logf("the condition of %s could not be kept: %v", id, err)
}
k := t.kindOf(askID)
needs := k.needs
c := asks.Condition{Key: conditionKey(id), Severity: asks.SeverityWarning, Subject: asks.Scope{Scope: k.scope, ID: id}, State: asks.StateRaised,
Summary: k.noun + id + ": " + why, Headline: headline, Needs: needs,
Explanation: asks.NeedsYou + needs + ". " + k.noun + id + ", " + why + ". " + k.nothing,
Resolved: k.resolved}
needs := "to ask for the Approve again"
c := asks.Condition{Key: conditionKey(id), Severity: asks.SeverityWarning, Subject: asks.Scope{Scope: "merge", ID: id}, State: asks.StateRaised,
Summary: "The merge ask " + id + ": " + why, Headline: headline, Needs: needs,
Explanation: asks.NeedsYou + needs + ". The merge ask " + id + ", " + why + ". Nothing was merged.",
Resolved: "A good answer to the merge ask was written."}
if err := t.Say(c); err != nil {
t.logf("the condition of %s could not be said: %v", id, err)
_ = t.Store.Delete(condKey(id))
@@ -510,7 +477,7 @@ func (t *Teller) clear(askID string) {
if err != nil || e == nil {
return
}
c := asks.Condition{Key: conditionKey(id), Subject: asks.Scope{Scope: t.kindOf(askID).scope, ID: id}, State: asks.StateCleared}
c := asks.Condition{Key: conditionKey(id), Subject: asks.Scope{Scope: "merge", ID: id}, State: asks.StateCleared}
if err := t.Say(c); err != nil {
t.logf("the condition of %s could not be cleared: %v", id, err)
return
@@ -611,7 +578,7 @@ func (t *Teller) roomRefused(w asks.Warrant) {
return
}
t.logf("the router refused the ask %s for want of room %d times (%s): it is asked no more, and said", w.Ask, n, w.Words)
t.raise(w.Ask, t.kindOf(w.Ask).unaskedHeadline, fmt.Sprintf("the router refused it %d times for want of room: %s", n, w.Words))
t.raise(w.Ask, "A merge ask could not be asked", fmt.Sprintf("the router refused it %d times for want of room: %s", n, w.Words))
return
}
if t.keepWaiting(ws, rec) != nil {
@@ -708,11 +675,6 @@ func (t *Teller) askWaiting(n int) {
i++
continue
}
assigner := isAssignerKey(x.Key)
if assigner && !t.assignerRoom() {
i++ // the assigner's asks hold their most open: it waits, and a merge ask after it may be asked
continue
}
a, why := t.stillAsked(x)
if why != "" {
t.logf("%s: the ask %s, waiting for room, is not asked again: %s", x.Key, x.ID, why)
@@ -725,9 +687,6 @@ func (t *Teller) askWaiting(n int) {
return
}
ws[i].Asked = true
if assigner {
t.assignerAsked(x.ID)
}
t.logf("%s: the ask %s is asked of the operator again (refused %d times for want of room)", x.Key, x.ID, x.Refusals)
if n--; n <= 0 {
return
@@ -742,9 +701,6 @@ func (t *Teller) stillAsked(x waitingAsk) (asks.Ask, string) {
if x.Key == "" {
return asks.Ask{}, "the teller holds no gate key for it"
}
if isAssignerKey(x.Key) {
return t.stillAskedByAssigner(x)
}
raw, found, err := t.Gate.Get(x.Key)
if err != nil {
return asks.Ask{}, "the gate's ask could not be read: " + err.Error()
@@ -768,26 +724,6 @@ func (t *Teller) stillAsked(x waitingAsk) (asks.Ask, string) {
return ga.Ask, ""
}
// stillAskedByAssigner is stillAsked for one of the assigner's asks.
func (t *Teller) stillAskedByAssigner(x waitingAsk) (asks.Ask, string) {
a, found, err := t.assignerAsk(x.Key)
switch {
case err != nil:
return asks.Ask{}, err.Error()
case !found:
return asks.Ask{}, "the assigner has taken it away"
case a.Ask.ID != x.ID:
return asks.Ask{}, "the assigner asks again under " + a.Ask.ID
}
if cur, _, has, err := t.Answers.Get(delivery.AnswerKey(x.Key)); err == nil && has && answeredAsking(cur) == x.ID {
return asks.Ask{}, "it is answered"
}
if err := a.Ask.Check(t.now()); err != nil {
return asks.Ask{}, "it no longer holds: " + err.Error()
}
return a.Ask, ""
}
// askAgain keeps the ask afresh as the asker would (its record of the refusal is replaced, so its warrant is taken) and
// publishes it under the bus message id given.
func (t *Teller) askAgain(a asks.Ask, msgID string) error {
@@ -2,7 +2,6 @@ package main
import (
"strings"
"time"
"git.novox.be/novox/mesh-sdk/go/delivery"
)
@@ -26,8 +25,7 @@ func (t *Teller) OnViewAnswer(key string) {
t.fillFromView(askKey)
}
// OnAssignerAsk handles a key put in the assigner's bucket asks: a view answer that came before its ask fills it now, and an
// ask not answered yet is raised on the operator channel (issue 1906; assignerasks.go).
// OnAssignerAsk handles a key put in the assigner's bucket asks: a view answer that came before its ask fills it now.
func (t *Teller) OnAssignerAsk(key string) {
t.mu.Lock()
defer t.mu.Unlock()
@@ -36,14 +34,10 @@ func (t *Teller) OnAssignerAsk(key string) {
return
}
t.fillFromView(key)
t.raiseAssignerAsk(key)
}
// fillFromView writes the answer to the assigner's ask at askKey from the view's answer, when both are there and readable, the
// ask is of a kind that needs no proof, the option is one the ask offers and the asking is not answered yet. A hold ask answered
// already takes a later answer of the view that chose another option: the operator changed their mind (issue 1906 note 23: a
// "keep" given on the phone could not be undone). Later means the view's answer was written after the answer that stands, so a
// view answer read again at a start never replaces a newer answer from the phone.
// ask is of a kind that needs no proof, the option is one the ask offers and the asking is not answered yet.
func (t *Teller) fillFromView(askKey string) {
if t.Assigner == nil || t.View == nil {
return
@@ -53,7 +47,7 @@ func (t *Teller) fillFromView(askKey string) {
t.logf("%s is not answered from the view: %v", askKey, err)
return
}
rawAnswer, at, found, err := t.viewAnswer(viewKey)
rawAnswer, found, err := t.View.Get(viewKey)
if err != nil {
t.logf("%s: the view's answer could not be read, so the ask is left: %v", askKey, err)
return
@@ -89,14 +83,14 @@ func (t *Teller) fillFromView(askKey string) {
t.logf("%s: the answer could not be read: %v", askKey, err)
return
}
if has && answeredAsking(cur) == ask.Ask.ID && !t.changesMind(askKey, cur, va, at) {
if has && answeredAsking(cur) == ask.Ask.ID {
return // this asking is answered, from the view or elsewhere
}
words := "answered in the view (mesh-delivery-view): " + va.Option
if w := strings.TrimSpace(va.Words); w != "" {
words += ": " + w
}
body, err := delivery.EncodeAnswer(delivery.Answer{Option: va.Option, Warrant: ask.Ask.ID + "@" + at.UTC().Format(time.RFC3339Nano), Words: words})
body, err := delivery.EncodeAnswer(delivery.Answer{Option: va.Option, Warrant: ask.Ask.ID + "@" + t.now().UTC().Format("2006-01-02T15:04:05.999999999Z07:00"), Words: words})
if err != nil {
t.logf("%s: the answer could not be made: %v", askKey, err)
return
@@ -109,48 +103,5 @@ func (t *Teller) fillFromView(askKey string) {
t.logf("%s: another writer answered first, and the view's answer is not written", askKey)
default:
t.logf("%s: %s", askKey, words)
t.cancelOpen(askKey, ask.Ask.ID) // answered in the view: the same ask on the phone is taken back
}
}
// timedAnswers is the view's bucket as the program reads it: each answer with the time it was written.
type timedAnswers interface {
GetAt(key string) (value []byte, at time.Time, found bool, err error)
}
// viewAnswer is the view's answer at key and when it was written; a bucket that keeps no time gives the teller's now.
func (t *Teller) viewAnswer(key string) ([]byte, time.Time, bool, error) {
if tv, ok := t.View.(timedAnswers); ok {
v, at, found, err := tv.GetAt(key)
if err == nil && !at.IsZero() {
return v, at, found, nil
}
if err != nil {
return nil, time.Time{}, false, err
}
return v, t.now(), found, nil
}
v, found, err := t.View.Get(key)
return v, t.now(), found, err
}
// changesMind says whether the view's answer va, written at at, replaces the answer cur of the same asking of the hold ask at
// askKey: it chose another option and was written after cur's warrant.
func (t *Teller) changesMind(askKey string, cur []byte, va delivery.ViewAnswer, at time.Time) bool {
k, err := delivery.ParseAnyAskKey(askKey)
if err != nil || k.Kind != delivery.AskKindHold {
return false
}
a, err := delivery.DecodeAnswer(cur)
if err != nil || a.Option == va.Option {
return false
}
_, when, _ := strings.Cut(a.Warrant, "@")
before, err := time.Parse(time.RFC3339Nano, when)
if err != nil || !at.After(before) {
t.logf("%s: the view's answer %s is not newer than the answer %s that stands (%s), and is not taken", askKey, va.Option, a.Option, when)
return false
}
t.logf("%s: the operator changed the answer from %s to %s in the view", askKey, a.Option, va.Option)
return true
}
@@ -3,7 +3,6 @@ package main
import (
"strings"
"testing"
"time"
"git.novox.be/novox/mesh-sdk/go/asks"
"git.novox.be/novox/mesh-sdk/go/delivery"
@@ -11,9 +10,8 @@ import (
const fpHeld = "src2:2222222222222222222222222222222222222222222222222222222222222222"
// holdAsk is a key of the assigner's asks: the ask about a held fingerprint, shaped as the assigner composes it
// (mesh-delivery-assigner internal/assign asks.go HoldAsk: two options of level acknowledge that bind no act, so the router
// takes it; the value is `{"v":1,"ask":<the router's ask>}`).
// holdAsk is a key of the assigner's asks: the ask about a held fingerprint, as a fixture (in release 5 the assigner writes
// none yet; the value is `{"v":1,"ask":<the router's ask>}`).
func holdAsk(t *testing.T) (key string, value []byte, ask asks.Ask) {
t.Helper()
key, err := delivery.HoldAskKey("plex", fpHeld)
@@ -22,7 +20,7 @@ func holdAsk(t *testing.T) (key string, value []byte, ask asks.Ask) {
}
ask = asks.Ask{ID: delivery.AskID(key), Headline: "plex is held after a failed health check", Explanation: "ace failed the check.", Who: asks.Operator,
Expires: now.Add(asks.ApproveLasts), OnExpiry: "the fingerprint stays held",
Options: []asks.Option{{ID: "keep", Label: "Keep it held", Does: "keeps it held", Level: asks.Acknowledge}, {ID: "release", Label: "Release it", Does: "lets the move go on", Level: asks.Acknowledge}}}
Options: []asks.Option{{ID: "release", Label: "Release", Does: "lets the move go on", Level: asks.Approve}, {ID: "keep", Label: "Keep held", Does: "keeps it held", Level: asks.Approve}}}
value, err = delivery.EncodeAssignerAsk(key, delivery.AssignerAsk{Ask: ask})
if err != nil {
t.Fatal(err)
@@ -119,11 +117,8 @@ func TestAViewAnswerOfAnOptionTheAskDoesNotOfferFillsNothing(t *testing.T) { //
}
}
// A view answer written before the answer that stands never replaces it (1906/N23: only a later one may, for a hold). The
// view's time is fixed here, not the package's clock, which the integration tests set to the wall clock.
func TestAViewAnswerNeverReplacesTheAnswerOfTheSameAsking(t *testing.T) { // 1297/B10
b := newBed(t)
b.t.Now = func() time.Time { return time.Date(2026, 10, 19, 12, 0, 0, 0, time.UTC) }
assigner, view := withView(b)
key, value, ask := holdAsk(t)
assigner[key] = value
@@ -179,53 +174,3 @@ func TestAStaleViewAnswerOfAnotherAskingFillsNothing(t *testing.T) { // 1297/B10
t.Errorf("writes %d; log:\n%s", b.answers.writes, b.logs.String())
}
}
// timedView is the view's bucket with the time each answer was written.
type timedView struct {
memGate
at map[string]time.Time
}
func (v timedView) GetAt(key string) ([]byte, time.Time, bool, error) {
b, ok := v.memGate[key]
return b, v.at[key], ok, nil
}
// 1906/N23-AC4: given a hold ask answered "keep" on the phone, when the operator answers "release" in the view later, then the
// teller replaces the answer of that asking with "release", whose warrant names the asking and the view's time; a view answer
// written before the phone's answer, the same option, or an ask of another kind replaces nothing.
func TestALaterViewAnswerChangesAHoldsAnswer(t *testing.T) {
setup := func(viewAt time.Time, option string) (*bed, string, asks.Ask) {
b := newBed(t)
assigner := memGate{}
view := timedView{memGate: memGate{}, at: map[string]time.Time{}}
b.t.Assigner, b.t.View = assigner, view
key, value, ask := holdAsk(t)
assigner[key] = value
kept, _ := delivery.EncodeAnswer(delivery.Answer{Option: "keep", Warrant: ask.ID + "@" + now.Add(time.Hour).UTC().Format(time.RFC3339Nano), Words: "the operator, via telegram, chose keep"})
b.answers.written[key], b.answers.revs[key] = kept, 3
view.memGate[key+".view"] = viewAnswer(t, option, "it was a flake")
view.at[key+".view"] = viewAt
b.t.OnViewAnswer(key + ".view")
return b, key, ask
}
later := now.Add(2 * time.Hour)
b, key, ask := setup(later, "release")
a, err := delivery.DecodeAnswer(b.answers.written[key])
if err != nil || b.answers.writes != 1 || a.Option != "release" || a.Warrant != ask.ID+"@"+later.UTC().Format(time.RFC3339Nano) {
t.Fatalf("writes %d, answer %+v, %v; log:\n%s", b.answers.writes, a, err, b.logs.String())
}
for name, b := range map[string]*bed{
"older": func() *bed { b, _, _ := setup(now, "release"); return b }(),
"same": func() *bed { b, _, _ := setup(later, "keep"); return b }(),
} {
if b.answers.writes != 0 {
t.Errorf("%s: the answer was replaced", name)
}
}
// A restore ask answered stays answered: only a hold's answer may be changed.
bb := newBed(t)
if bb.t.changesMind("restore.ace.plex.2222222222222222222222222222222222222222222222222222222222222222", b.answers.written[key], delivery.ViewAnswer{Option: "keep"}, later.Add(time.Hour)) {
t.Error("a restore ask's answer was changed")
}
}
@@ -1,8 +1,6 @@
// Command mesh-delivery-cli is the operator's command for the delivery view (novox/hq ADR 0429 §3). It asks the
// view's local socket for the rows and prints them; it holds no credential. `answer <ask key> <option> [words]` asks the view
// to write the operator's answer to an ask that needs no proof of who answered (the view refuses the merge's Approve). `holds`
// lists every fingerprint the assigner holds with its node, cause and answer; `hold-release <module> <fingerprint> [words]` and
// `hold-keep <module> <fingerprint> [words]` answer its hold ask, again when the operator changes their mind (issue 1906).
// to write the operator's answer to an ask that needs no proof of who answered (the view refuses the merge's Approve).
package main
import (
@@ -12,7 +10,6 @@ import (
"io"
"os"
"strings"
"text/tabwriter"
"mesh-delivery-view/internal/view"
)
@@ -34,26 +31,9 @@ func run(args []string, socket string, out, errOut io.Writer) int {
if err := fs.Parse(args); err != nil {
return 2
}
if fs.NArg() > 0 && fs.Arg(0) == "holds" {
return holds(socket, *asJSON, out, errOut)
}
if fs.NArg() > 0 && (fs.Arg(0) == "hold-release" || fs.Arg(0) == "hold-keep") {
if fs.NArg() < 3 {
fmt.Fprintf(errOut, "mesh-delivery-cli %s <module> <fingerprint> [words]\n", fs.Arg(0))
return 2
}
option := strings.TrimPrefix(fs.Arg(0), "hold-")
r, err := view.AnswerHoldOn(socket, fs.Arg(1), fs.Arg(2), option, strings.Join(fs.Args()[3:], " "))
if err != nil {
fmt.Fprintln(errOut, "mesh-delivery-cli:", err)
return 1
}
fmt.Fprintf(out, "answered the hold of %s at %s with %s; the teller fills the ask from it and the assigner reads it\n", r.Module, r.FP, option)
return 0
}
if fs.NArg() > 0 {
if fs.Arg(0) != "answer" || fs.NArg() < 3 {
fmt.Fprintln(errOut, "mesh-delivery-cli shows the rows, or answers: mesh-delivery-cli answer <ask key> <option> [words]; holds; hold-release|hold-keep <module> <fingerprint> [words]")
fmt.Fprintln(errOut, "mesh-delivery-cli shows the rows, or answers: mesh-delivery-cli answer <ask key> <option> [words]")
return 2
}
key, option := fs.Arg(1), fs.Arg(2)
@@ -81,53 +61,3 @@ func run(args []string, socket string, out, errOut io.Writer) int {
view.Table(out, s)
return 0
}
// holds prints every hold: module, fingerprint, node, since, answer, words.
func holds(socket string, asJSON bool, out, errOut io.Writer) int {
rows, err := view.AskHolds(socket)
if err != nil {
fmt.Fprintln(errOut, "mesh-delivery-cli:", err)
return 1
}
if asJSON {
enc := json.NewEncoder(out)
enc.SetIndent("", " ")
if err := enc.Encode(rows); err != nil {
fmt.Fprintln(errOut, "mesh-delivery-cli:", err)
return 1
}
return 0
}
if len(rows) == 0 {
fmt.Fprintln(out, "the assigner holds nothing")
return 0
}
tw := tabwriter.NewWriter(out, 0, 2, 2, ' ', 0)
fmt.Fprintln(tw, "MODULE\tFINGERPRINT\tNODE\tSINCE\tANSWER\tWORDS")
for _, r := range rows {
answer := r.Answer
switch {
case r.AskKey == "":
answer = "no ask"
case r.Pending != "":
answer = r.Pending + " (from the view, the teller fills it)"
case answer == "":
answer = "not answered"
}
fp := strings.TrimPrefix(r.FP, "src2:")
if len(fp) > 12 {
fp = fp[:12]
}
since := ""
if !r.Since.IsZero() {
since = r.Since.Local().Format("2006-01-02 15:04")
}
words := r.Words
if r.Error != "" {
words = "unreadable: " + r.Error
}
fmt.Fprintf(tw, "%s\t%s\t%s\t%s\t%s\t%s\n", r.Module, fp, r.Node, since, answer, words)
}
tw.Flush()
return 0
}
@@ -240,46 +240,3 @@ func TestTheAnswerCommandAnswersAnAskNeedingNoProof(t *testing.T) { // 1297/B9b
}
}
}
// 1906/N23-AC3: `holds` lists the holds with node, cause and answer; `hold-release <module> <fingerprint>` and `hold-keep` answer
// the hold ask through the view, also one answered already.
func TestTheHoldCommandsListAndAnswerHolds(t *testing.T) {
fp := "src2:04efd1a1aa" + strings.Repeat("2", 54)
key, err := delivery.HoldAskKey("systemd-resolved", fp)
if err != nil {
t.Fatal(err)
}
hold := asks.Ask{ID: delivery.AskID(key), Headline: "systemd-resolved is held", Explanation: "ace failed the check.", Who: asks.Operator, Expires: time.Now().Add(time.Hour), OnExpiry: "stays held",
Options: []asks.Option{{ID: "keep", Label: "Keep", Does: "stays", Level: asks.Acknowledge}, {ID: "release", Label: "Release", Does: "goes on", Level: asks.Acknowledge}}}
value, _ := delivery.EncodeAssignerAsk(key, delivery.AssignerAsk{Ask: hold})
hkey, _ := delivery.HeldKey("systemd-resolved", fp)
held, _ := delivery.EncodeHeld(delivery.Held{Node: "ace", Words: "health failed, phase running: dns timeout"})
ans, _ := delivery.EncodeAnswer(delivery.Answer{Option: "keep", Warrant: hold.ID + "@2026-10-25T04:00:00Z", Words: "the operator chose keep"})
m, w := view.New(), &putter{puts: map[string][]byte{}}
m.SetWriter(w)
m.Apply(view.Entry{Bucket: delivery.AssignerAsksBucket, Key: key, Value: value})
m.Apply(view.Entry{Bucket: delivery.AssignerWantsBucket, Key: hkey, Value: held})
m.Apply(view.Entry{Bucket: delivery.AnswersBucket, Key: key, Value: ans})
socket := serveWith(t, m)
var out, errOut bytes.Buffer
if code := run([]string{"holds"}, socket, &out, &errOut); code != 0 {
t.Fatalf("exit %d: %s", code, errOut.String())
}
for _, want := range []string{"systemd-resolved", "04efd1a1aa22", "ace", "keep", "dns timeout"} {
if !strings.Contains(out.String(), want) {
t.Errorf("holds printed %q, missing %q", out.String(), want)
}
}
out.Reset()
if code := run([]string{"hold-release", "systemd-resolved", "04efd1a1", "a", "flake"}, socket, &out, &errOut); code != 0 {
t.Fatalf("exit %d: %s", code, errOut.String())
}
if a, err := delivery.DecodeViewAnswer(w.puts[key+".view"]); err != nil || a.Option != "release" || a.Asking != hold.ID || a.Words != "a flake" {
t.Errorf("the view's answer = %+v, %v", a, err)
}
errOut.Reset()
if code := run([]string{"hold-keep", "systemd-resolved"}, socket, &out, &errOut); code != 2 {
t.Errorf("hold-keep without a fingerprint: exit %d", code)
}
}
@@ -1,188 +0,0 @@
package view
import (
"encoding/json"
"errors"
"fmt"
"net"
"slices"
"sort"
"strings"
"time"
"git.novox.be/novox/mesh-sdk/go/asks"
"git.novox.be/novox/mesh-sdk/go/delivery"
)
// The assigner's holds, listed and answered from the operator's command (issue 1906 note 23). A hold ask answered once (on the
// phone or in the view) ended with that answer, so nothing listed it any more and an answer "keep" could not be changed: the
// operator kept systemd-resolved held and then could not release it, and a kept hold is never lifted by itself. Now the view
// lists every hold with its node, its cause and its answer, and answers the hold ask that stands, again if the operator changes
// their mind, through the view's one write (`<ask key>.view`), which the teller takes into the ask's answer. Nothing else writes:
// the answer still reaches the assigner only through the teller (mesh-delivery-teller viewanswers.go).
// The options of the hold ask (mesh-delivery-assigner internal/assign asks.go HoldAsk).
const (
HoldRelease = "release"
HoldKeep = "keep"
)
type heldEntry struct {
module, fp string
value delivery.Held
err string
at time.Time
}
// applyHeld takes a change of a held key of the assigner's wants, and says whether the key was one.
func (m *Model) applyHeld(e Entry) bool {
module, fp, err := delivery.ParseHeldKey(e.Key)
if err != nil {
return false
}
if e.Deleted {
delete(m.holds, e.Key)
return true
}
h, err := delivery.DecodeHeld(e.Value)
m.holds[e.Key] = heldEntry{module: module, fp: fp, value: h, err: errText(err), at: e.At}
return true
}
// HoldRow is one fingerprint the assigner holds: the module, the fingerprint, the node whose health check failed, the hold's
// words (its cause) and since when; the ask about it (its key and the asking's id, empty when the assigner asks none), and the
// answer that stands for that asking (the option and its words, empty when none), with the view's own answer when it waits for
// the teller.
type HoldRow struct {
Module string `json:"module"`
FP string `json:"fp"`
Node string `json:"node,omitempty"`
Words string `json:"words,omitempty"`
Since time.Time `json:"since,omitempty"`
AskKey string `json:"ask-key,omitempty"`
Asking string `json:"asking,omitempty"`
Answer string `json:"answer,omitempty"`
Answered string `json:"answered,omitempty"`
Pending string `json:"pending,omitempty"`
Error string `json:"error,omitempty"`
}
// Holds is every hold, ordered by module and fingerprint.
func (m *Model) Holds() []HoldRow {
m.mu.Lock()
defer m.mu.Unlock()
out := []HoldRow{}
for _, h := range m.holds {
row := HoldRow{Module: h.module, FP: h.fp, Node: h.value.Node, Words: h.value.Words, Since: h.at, Error: h.err}
if key, err := delivery.HoldAskKey(h.module, h.fp); err == nil {
if a, ok := m.aasks[key]; ok && a.err == "" {
row.AskKey, row.Asking = key, a.ask.ID
if ans, ok := m.answerOf[key]; ok && m.answers[key] == a.ask.ID {
row.Answer, row.Answered = ans.Option, ans.Words
}
if c := m.chosenOf(key, a.ask); c != "" && c != row.Answer {
row.Pending = c
}
}
}
out = append(out, row)
}
sort.Slice(out, func(i, j int) bool {
if out[i].Module != out[j].Module {
return out[i].Module < out[j].Module
}
return out[i].FP < out[j].FP
})
return out
}
// AnswerHold writes the view's answer to the hold ask of module at the fingerprint fp: release or keep, with the operator's
// words. fp is the whole fingerprint, or the start of its hex (as the rows print it), naming one hold of the module. Unlike
// Answer it answers an ask answered already, so the operator can change their mind; an answer equal to the one that stands is
// refused, as is a hold the assigner asks nothing about.
func (m *Model) AnswerHold(module, fp, option, words string) (HoldRow, error) {
option = strings.TrimSpace(option)
if option != HoldRelease && option != HoldKeep {
return HoldRow{}, fmt.Errorf("a hold is answered %s or %s, not %q", HoldRelease, HoldKeep, option)
}
var match []HoldRow
want := strings.TrimPrefix(strings.TrimSpace(fp), delivery.SourcePrefix2)
for _, r := range m.Holds() {
if r.Module == module && want != "" && strings.HasPrefix(strings.TrimPrefix(r.FP, delivery.SourcePrefix2), want) {
match = append(match, r)
}
}
switch {
case len(match) == 0:
return HoldRow{}, fmt.Errorf("the assigner holds no fingerprint %q of %s (mesh-delivery-cli holds lists them)", fp, module)
case len(match) > 1:
return HoldRow{}, fmt.Errorf("%q names %d holds of %s: give more of the fingerprint", fp, len(match), module)
}
r := match[0]
if r.AskKey == "" {
return r, fmt.Errorf("the assigner asks nothing about the hold of %s at %s, so there is nothing to answer", module, r.FP)
}
if r.Answer == option && r.Pending == "" {
return r, fmt.Errorf("the hold of %s at %s is answered %s already", module, r.FP, option)
}
m.mu.Lock()
a, w := m.aasks[r.AskKey], m.writer
m.mu.Unlock()
if !slices.ContainsFunc(a.ask.Options, func(o asks.Option) bool { return o.ID == option }) {
return r, fmt.Errorf("the hold ask %s offers no option %q", r.AskKey, option)
}
if w == nil {
return r, fmt.Errorf("the view cannot write its answers here")
}
key, err := delivery.ViewAnswerKey(r.AskKey)
if err != nil {
return r, err
}
body, err := delivery.EncodeViewAnswer(delivery.ViewAnswer{Asking: a.ask.ID, Option: option, Words: strings.TrimSpace(words)})
if err != nil {
return r, err
}
if err := w.Put(key, body); err != nil {
return r, err
}
return r, nil
}
// askOn sends one ask to the socket at path and decodes the reply into out; the view's refusal is the error.
func askOn(path string, req ask, out any) error {
c, err := net.DialTimeout("unix", path, 5*time.Second)
if err != nil {
return fmt.Errorf("cannot reach the view at %s: %w", path, err)
}
defer c.Close()
c.SetDeadline(time.Now().Add(10 * time.Second))
raw, _ := json.Marshal(req)
if _, err := c.Write(append(raw, '\n')); err != nil {
return err
}
var reply json.RawMessage
if err := json.NewDecoder(c).Decode(&reply); err != nil {
return fmt.Errorf("the view's reply cannot be read: %w", err)
}
var refusal struct {
Error string `json:"error"`
}
if json.Unmarshal(reply, &refusal) == nil && refusal.Error != "" {
return errors.New(refusal.Error)
}
return json.Unmarshal(reply, out)
}
// AskHolds asks the socket at path for every hold.
func AskHolds(path string) ([]HoldRow, error) {
var out []HoldRow
err := askOn(path, ask{Ask: "holds"}, &out)
return out, err
}
// AnswerHoldOn asks the socket at path to answer the hold ask of module at fp with option.
func AnswerHoldOn(path, module, fp, option, words string) (HoldRow, error) {
var out HoldRow
err := askOn(path, ask{Ask: "hold", Module: module, FP: fp, Option: option, Words: words}, &out)
return out, err
}
@@ -1,91 +0,0 @@
package view
import (
"strings"
"testing"
"git.novox.be/novox/mesh-sdk/go/delivery"
)
// heldE is the assigner's hold of module at fp, as it writes it in its wants.
func heldE(t *testing.T, module, fp, node, words string) Entry {
t.Helper()
key, err := delivery.HeldKey(module, fp)
if err != nil {
t.Fatal(err)
}
v, err := delivery.EncodeHeld(delivery.Held{Node: node, Words: words})
if err != nil {
t.Fatal(err)
}
return Entry{Bucket: delivery.AssignerWantsBucket, Key: key, Value: v, At: t0}
}
func answered(t *testing.T, key, id, option string) Entry {
t.Helper()
v, err := delivery.EncodeAnswer(delivery.Answer{Option: option, Warrant: id + "@2026-10-25T04:00:00Z", Words: "the operator, via telegram, chose " + option})
if err != nil {
t.Fatal(err)
}
return Entry{Bucket: delivery.AnswersBucket, Key: key, Value: v, At: t0}
}
// 1906/N23-AC1: given holds in the assigner's wants, the view lists each with its node, its cause in words, its hold ask and
// the answer that stands for that asking; a hold without an ask says so, and a held key is never taken for a want.
func TestTheViewListsEveryHoldWithItsNodeCauseAndAnswer(t *testing.T) {
m := New()
e, a := holdAsk(t)
fill(m, e, heldE(t, "plex", fpNew, "ace", "health failed, phase running: connection refused"), answered(t, e.Key, a.ID, "keep"),
heldE(t, "web", fpNew, "g14", "health failed"))
got := m.Holds()
if len(got) != 2 {
t.Fatalf("holds = %+v", got)
}
p := got[0]
if p.Module != "plex" || p.FP != fpNew || p.Node != "ace" || !strings.Contains(p.Words, "connection refused") || p.AskKey != e.Key || p.Asking != a.ID || p.Answer != "keep" {
t.Errorf("plex = %+v", p)
}
if w := got[1]; w.Module != "web" || w.Node != "g14" || w.AskKey != "" || w.Answer != "" {
t.Errorf("web = %+v", w)
}
if len(m.Snapshot().Wants) != 0 {
t.Errorf("a hold was taken for a want: %+v", m.Snapshot().Wants)
}
}
// 1906/N23-AC2: given a hold ask answered "keep" (the ask ended with it), when the operator releases the hold, then the view
// writes its answer "release" to that asking under the ask's view key, though the ask was answered; the fingerprint may be the
// start of its hex. The same answer again, an unknown hold and another option are refused, and write nothing.
func TestAKeptHoldCanBeReleased(t *testing.T) {
m, w := New(), &memWriter{}
m.SetWriter(w)
e, a := holdAsk(t)
fill(m, e, heldE(t, "plex", fpNew, "ace", "health failed"), answered(t, e.Key, a.ID, "keep"))
short := strings.TrimPrefix(fpNew, delivery.SourcePrefix2)[:8]
if _, err := m.AnswerHold("plex", short, HoldRelease, "it was a flake"); err != nil {
t.Fatal(err)
}
va, err := delivery.DecodeViewAnswer(w.puts[e.Key+".view"])
if err != nil || va.Asking != a.ID || va.Option != HoldRelease || va.Words != "it was a flake" {
t.Fatalf("the view's answer = %+v, %v", va, err)
}
w.puts = nil
fill(m, answered(t, e.Key, a.ID, HoldRelease))
for name, try := range map[string]func() error{
"again": func() error { _, err := m.AnswerHold("plex", fpNew, HoldRelease, ""); return err },
"unknown": func() error { _, err := m.AnswerHold("plex", "ffff", HoldKeep, ""); return err },
"option": func() error { _, err := m.AnswerHold("plex", fpNew, "restore", ""); return err },
"module": func() error { _, err := m.AnswerHold("web", fpNew, HoldKeep, ""); return err },
} {
if err := try(); err == nil {
t.Errorf("%s: not refused", name)
}
}
if len(w.puts) != 0 {
t.Errorf("a refused answer wrote %v", w.puts)
}
// And kept again: the operator may change their mind both ways.
if _, err := m.AnswerHold("plex", fpNew, HoldKeep, ""); err != nil {
t.Errorf("keep after release: %v", err)
}
}
@@ -46,8 +46,6 @@ type ask struct {
Key string `json:"key,omitempty"`
Option string `json:"option,omitempty"`
Words string `json:"words,omitempty"`
Module string `json:"module,omitempty"`
FP string `json:"fp,omitempty"`
}
// ServeSocket answers {"ask":"rows"} on l, line-delimited JSON, until ctx ends, and {"ask":"answer","key":…,"option":…,"words":…}
@@ -74,21 +72,8 @@ func answer(c net.Conn, m *Model) {
return
}
var a ask
if json.Unmarshal(line, &a) != nil || (a.Ask != "rows" && a.Ask != "answer" && a.Ask != "holds" && a.Ask != "hold") {
json.NewEncoder(c).Encode(map[string]string{"error": `refused: this socket answers {"ask":"rows"}, {"ask":"answer"}, {"ask":"holds"} and {"ask":"hold"} and nothing else`})
return
}
switch a.Ask {
case "holds":
json.NewEncoder(c).Encode(m.Holds())
return
case "hold":
r, err := m.AnswerHold(a.Module, a.FP, a.Option, a.Words)
if err != nil {
json.NewEncoder(c).Encode(map[string]string{"error": err.Error()})
return
}
json.NewEncoder(c).Encode(r)
if json.Unmarshal(line, &a) != nil || (a.Ask != "rows" && a.Ask != "answer") {
json.NewEncoder(c).Encode(map[string]string{"error": `refused: this socket answers {"ask":"rows"} and {"ask":"answer"} and nothing else`})
return
}
if a.Ask == "answer" {
@@ -231,10 +231,8 @@ type Model struct {
chosen map[string]delivery.ViewAnswer // the answers the view wrote, by ask key
writer Writer
backlog *Backlog
answers map[string]string // by key: the router id of the asking the teller's answer is for ("" when unreadable)
answerOf map[string]delivery.Answer // by key: the teller's answer whole, when readable (holds.go)
holds map[string]heldEntry // the assigner's holds, by held key (holds.go)
newest map[string]string // module -> id
answers map[string]string // by key: the router id of the asking the teller's answer is for ("" when unreadable)
newest map[string]string // module -> id
haves map[string]haveEntry
nodes map[string]nodeEntry
// the watchman's buckets, by key
@@ -346,7 +344,7 @@ type nodeEntry struct {
// New is an empty model.
func New() *Model {
return &Model{assigns: map[string]assignment{}, builds: map[string]buildEntry{}, groups: map[string]groupEntry{}, modules: map[string]moduleEntry{}, keys: map[string]keyEntry{}, asks: map[string]askRecord{}, wants: map[string]wantEntry{}, told: map[string]bool{}, initial: map[string]bool{}, composed: map[string]composedEntry{}, aasks: map[string]assignerAskRecord{}, chosen: map[string]delivery.ViewAnswer{}, answers: map[string]string{}, answerOf: map[string]delivery.Answer{}, holds: map[string]heldEntry{}, newest: map[string]string{},
return &Model{assigns: map[string]assignment{}, builds: map[string]buildEntry{}, groups: map[string]groupEntry{}, modules: map[string]moduleEntry{}, keys: map[string]keyEntry{}, asks: map[string]askRecord{}, wants: map[string]wantEntry{}, told: map[string]bool{}, initial: map[string]bool{}, composed: map[string]composedEntry{}, aasks: map[string]assignerAskRecord{}, chosen: map[string]delivery.ViewAnswer{}, answers: map[string]string{}, newest: map[string]string{},
haves: map[string]haveEntry{}, nodes: map[string]nodeEntry{},
presence: map[string]presenceEntry{}, stuck: map[string]stuckEntry{}}
}
@@ -488,12 +486,9 @@ func (m *Model) apply(e Entry) {
m.chosen[askKey] = a
}
case delivery.AssignerWantsBucket:
if m.applyHeld(e) {
return
}
node, module, err := delivery.ParseWantKey(e.Key)
if err != nil {
return // the node's own key is not a module's want
return // a hold, and the node's own key, are not a module's want
}
key, kerr := delivery.HaveKey(node, module)
if kerr != nil {
@@ -525,7 +520,6 @@ func (m *Model) apply(e Entry) {
case delivery.AnswersBucket:
if e.Deleted {
delete(m.answers, e.Key)
delete(m.answerOf, e.Key)
return
}
// The answer's warrant is `<router id>@<time>`: it ends that asking, and only that one (a later asking of the key is
@@ -533,9 +527,6 @@ func (m *Model) apply(e Entry) {
id := ""
if a, err := delivery.DecodeAnswer(e.Value); err == nil {
id, _, _ = strings.Cut(a.Warrant, "@")
m.answerOf[e.Key] = a
} else {
delete(m.answerOf, e.Key)
}
m.answers[e.Key] = id
case delivery.BuilderBuildsBucket: