A try says its end with the state it ended in, kept in the try's own state, so an end after a checker restart is still said; only the checker publishes try events (issue 1921)

Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
This commit is contained in:
2026-11-05 18:47:31 +01:00
parent bd5eb33e9e
commit 88a47be251
3 changed files with 151 additions and 59 deletions
@@ -81,6 +81,9 @@ type Try struct {
Tests string `json:"tests,omitempty"`
Level string `json:"level,omitempty"`
State string `json:"state"`
// Told is true from the moment a whole try's start is said on the checker's stream until its end is said there: kept in the
// try's own state, so an end that is written and not said (the checker stopped between) is said when the checker starts.
Told bool `json:"told,omitempty"`
// 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.
@@ -519,7 +522,7 @@ type Tryer struct {
Say func(ctx context.Context, event string, body tryevent.Event)
mu sync.Mutex
told map[string]bool // whole tries whose start was said and whose end was not
ending map[string]bool // tries whose end is being said now
wg sync.WaitGroup
inflight map[string]context.CancelFunc
seen map[string]bool // looked at since this process started: a try left running is taken over the first time
@@ -558,12 +561,8 @@ func (tr *Tryer) Put(ctx context.Context, key string, value []byte) {
if stop := tr.inflight[key]; stop != nil {
stop()
}
said := tr.told[key]
delete(tr.told, key)
tr.mu.Unlock()
if said { // ended by hand while it ran: the indexer drops its candidate
tr.say(ctx, tryevent.Ended, tryevent.Event{ID: key, State: t.State})
}
tr.sendEnd(ctx, key) // ended by hand while it ran, or its end was written and not said: the indexer drops its candidate
return
}
if tr.inflight[key] != nil || (t.State == TryRunning && tr.seen[key]) {
@@ -619,6 +618,7 @@ func (tr *Tryer) run(ctx context.Context, key string) {
return
}
t.State, t.Started = TryRunning, tr.now().UTC()
t.Told = (&TryOptions{Level: t.Level, Run: t.Tests}).Whole() // its start is said below; its end is owed from now on
data, err := encodeTry(t)
if err != nil {
return
@@ -630,17 +630,11 @@ func (tr *Tryer) run(ctx context.Context, key string) {
return
}
progress := &tryWriter{tr: tr, ctx: ctx, key: key, t: t, rev: rev}
if (&TryOptions{Level: t.Level, Run: t.Tests}).Whole() {
if t.Told {
started := tryevent.Event{ID: key}
for _, m := range t.MembersOf() {
started.Members = append(started.Members, tryevent.Member{Repository: m.Repository, Head: m.Head, Trunk: m.Trunk})
}
tr.mu.Lock()
if tr.told == nil {
tr.told = map[string]bool{}
}
tr.told[key] = true
tr.mu.Unlock()
tr.say(ctx, tryevent.Started, started)
}
cs := TryCase(key, t, progress.heard)
@@ -653,13 +647,6 @@ func (tr *Tryer) run(ctx context.Context, key string) {
return // left running: the checker takes it over when it starts; or ended by hand, which wrote its end
}
t, rev = progress.end()
tr.mu.Lock()
said := tr.told[key]
delete(tr.told, key)
tr.mu.Unlock()
if said {
defer tr.say(ctx, tryevent.Ended, tryevent.Event{ID: key})
}
v := out.Verdict
t.Ended = tr.now().UTC()
t.Took = t.Ended.Sub(t.Started).Round(time.Second).String()
@@ -707,6 +694,41 @@ func (tr *Tryer) run(ctx context.Context, key string) {
return
}
tr.logf("%s: %s", key, t.State)
tr.sendEnd(ctx, key)
}
// sendEnd says the end of an ended try whose start was said (Told) with the state it ended in, then clears Told in the try's
// own state; a checker that stopped between the two says it again at its start. Every end is said so: pass, fail, error and
// by hand.
func (tr *Tryer) sendEnd(ctx context.Context, key string) {
tr.mu.Lock()
if tr.ending[key] {
tr.mu.Unlock()
return
}
if tr.ending == nil {
tr.ending = map[string]bool{}
}
tr.ending[key] = true
tr.mu.Unlock()
defer func() { tr.mu.Lock(); delete(tr.ending, key); tr.mu.Unlock() }()
raw, rev, found, err := tr.Tries.Get(ctx, key)
if err != nil || !found {
return
}
t, err := DecodeTry(raw)
if err != nil || !t.Over() || !t.Told {
return
}
tr.say(ctx, tryevent.Ended, tryevent.Event{ID: key, State: t.State})
t.Told = false
data, err := encodeTry(t)
if err == nil {
_, err = tr.Tries.Update(ctx, key, data, rev)
}
if err != nil && !errors.Is(err, ErrConflict) {
tr.logf("%s: its end was said and Told could not be cleared: %v", key, err)
}
}
// say publishes a try event; a failure is said in the log (the indexer finds the open tries again at its start).
@@ -5,6 +5,7 @@ import (
"encoding/json"
"errors"
"fmt"
"slices"
"strings"
"testing"
"time"
@@ -181,16 +182,19 @@ func TestATryIsJudgedAsAGroupsCheckAndOwesTheForgeNothing(t *testing.T) { // ADR
written := &writtenTries{memKV: kv}
var said []string
tr := &Tryer{Tries: written, Judge: b.ex, Say: func(_ context.Context, event string, e tryevent.Event) {
said = append(said, fmt.Sprintf("%s %s %d", event, e.ID, len(e.Members)))
said = append(said, fmt.Sprintf("%s %s %d %s", event, e.ID, len(e.Members), e.State))
}}
raw, _, _, _ := kv.Get(ctx, id)
tr.Put(ctx, id, raw)
tr.Wait()
// The indexer hears the start with the members at their commits, and the end (it writes and deletes the group's candidate).
if len(said) != 2 || said[0] != "try.started "+id+" 2" || said[1] != "try.ended "+id+" 0" {
if len(said) != 2 || said[0] != "try.started "+id+" 2 " || said[1] != "try.ended "+id+" 0 "+TryPass {
t.Errorf("events = %v, want the start with 2 members and the end", said)
}
got := readTry(t, kv, id)
if got.Told { // the end was said: nothing is owed
t.Errorf("the ended try still owes its end: %+v", got)
}
if open, err := OpenTries(memTries{kv}); err != nil || len(open) != 0 {
t.Errorf("open tries after the end = %v, %v", open, err)
}
@@ -282,26 +286,24 @@ func TestATryOfOneLevelAndSomeTestsRunsThemAloneAndKeepsNothing(t *testing.T) {
if got := readTry(t, kv, id); got.State != TryPass || !strings.Contains(got.Said, "only the tests TestBus|TestStore") {
t.Errorf("try = %+v", got)
}
}
func TestATryOfOneLevelKeepsOnlyThatLevel(t *testing.T) {
ctx := context.Background()
kv := newMemKV()
id, _ := AskTry(memTries{kv}, tryTrunks, TryAsk{Repository: repoA, Head: head1, Level: delivery.LevelUnit}, time.Now())
b := newBench(t)
tr := &Tryer{Tries: kv, Judge: b.ex}
raw, _, _, _ := kv.Get(ctx, id)
tr.Put(ctx, id, raw)
tr.Wait()
keys, _ := b.kept.Keys(ctx)
if got := readTry(t, kv, id); got.State != TryPass || len(keys) == 0 {
t.Fatalf("try = %+v, kept %v", got, keys)
}
for _, k := range keys {
if !strings.HasPrefix(k, delivery.LevelUnit+".") {
t.Errorf("a try of the unit level kept %s", k)
t.Run("a try of one level keeps only that level", func(t *testing.T) {
kv := newMemKV()
id, _ := AskTry(memTries{kv}, tryTrunks, TryAsk{Repository: repoA, Head: head1, Level: delivery.LevelUnit}, time.Now())
b := newBench(t)
tr := &Tryer{Tries: kv, Judge: b.ex}
raw, _, _, _ := kv.Get(ctx, id)
tr.Put(ctx, id, raw)
tr.Wait()
keys, _ := b.kept.Keys(ctx)
if got := readTry(t, kv, id); got.State != TryPass || len(keys) == 0 {
t.Fatalf("try = %+v, kept %v", got, keys)
}
}
for _, k := range keys {
if !strings.HasPrefix(k, delivery.LevelUnit+".") {
t.Errorf("a try of the unit level kept %s", k)
}
}
})
}
func TestAFailingTryNamesWhatFailedAndAStepThatCannotRunIsAnError(t *testing.T) {
@@ -340,10 +342,18 @@ func TestAFailingTryNamesWhatFailedAndAStepThatCannotRunIsAnError(t *testing.T)
comps.v.State = delivery.CompositionPass
}
id, _ := AskTry(memTries{kv}, tryTrunks, TryAsk{Repository: repoA, Head: head1}, time.Now())
tr := &Tryer{Tries: kv, Judge: b.ex}
var ends []string
tr := &Tryer{Tries: kv, Judge: b.ex, Say: func(_ context.Context, event string, e tryevent.Event) {
if event == tryevent.Ended {
ends = append(ends, e.State)
}
}}
raw, _, _, _ := kv.Get(ctx, id)
tr.Put(ctx, id, raw)
tr.Wait()
if len(ends) != 1 || ends[0] != row.state { // every end says its state: error as well as fail
t.Errorf("%s: ends said = %v, want once %s", row.name, ends, row.state)
}
if got := readTry(t, kv, id); got.State != row.state || !strings.Contains(strings.Join(got.Failed, " | ")+got.Said, row.said) {
t.Errorf("%s: try = %s, failed %v, said %q, want %s naming %q", row.name, got.State, got.Failed, got.Said, row.state, row.said)
}
@@ -353,22 +363,56 @@ func TestAFailingTryNamesWhatFailedAndAStepThatCannotRunIsAnError(t *testing.T)
func TestATryLeftRunningIsTakenOver(t *testing.T) {
ctx := context.Background()
kv := newMemKV()
id, _ := AskTry(memTries{kv}, tryTrunks, TryAsk{Repository: repoA, Head: head1}, time.Now())
// Left running by a checker that stopped.
raw, rev, _, _ := kv.Get(ctx, id)
cur, _ := DecodeTry(raw)
cur.State = TryRunning
data, _ := encodeTry(cur)
if _, err := kv.Update(ctx, id, data, rev); err != nil {
t.Fatal(err)
}
b := newBench(t)
wholeBench(b, delivery.Candidate{})
tr := &Tryer{Tries: kv, Judge: b.ex}
tr.Reconsider(ctx)
tr.Wait()
if got := readTry(t, kv, id); got.State != TryPass {
t.Errorf("the try left running = %+v, want taken over and passed", got)
// Whether the start was said is in the try's own state (Told), so a checker that starts later says the end of a try it did
// not start, whatever way the try ended.
for _, row := range []struct {
name string
told bool
after func(t *testing.T, kv *memKV, id string)
ends []string
}{
{name: "left running after its start was said", told: true, ends: []string{TryPass}},
{name: "left running with no start said", told: false, ends: []string{TryPass}},
{name: "ended by hand after the start was said", told: true, ends: []string{TryError}, after: func(t *testing.T, kv *memKV, id string) {
if _, err := EndTry(endable{kv}, id, "stuck", time.Now()); err != nil {
t.Fatal(err)
}
}},
} {
t.Run(row.name, func(t *testing.T) {
kv := newMemKV()
id, _ := AskTry(memTries{kv}, tryTrunks, TryAsk{Repository: repoA, Head: head1}, time.Now())
// Left running by a checker that stopped.
raw, rev, _, _ := kv.Get(ctx, id)
cur, _ := DecodeTry(raw)
cur.State, cur.Told = TryRunning, row.told
data, _ := encodeTry(cur)
if _, err := kv.Update(ctx, id, data, rev); err != nil {
t.Fatal(err)
}
if row.after != nil {
row.after(t, kv, id)
}
b := newBench(t)
wholeBench(b, delivery.Candidate{})
var ends []string
tr := &Tryer{Tries: kv, Judge: b.ex, Say: func(_ context.Context, event string, e tryevent.Event) {
if event == tryevent.Ended {
ends = append(ends, e.State)
}
}}
tr.Reconsider(ctx)
tr.Wait()
tr.Reconsider(ctx) // said once: the second look finds nothing owed
tr.Wait()
if row.after == nil {
if got := readTry(t, kv, id); got.State != TryPass {
t.Errorf("the try left running = %+v, want taken over and passed", got)
}
}
if !slices.Equal(ends, row.ends) || readTry(t, kv, id).Told {
t.Errorf("ends said = %v, want %v, and nothing owed after", ends, row.ends)
}
})
}
}
@@ -134,6 +134,32 @@ func TestManifestDeclarations(t *testing.T) { // 1678/AC1
if want := []stream{{Name: "checks", Events: []string{"preview.**", "owed.**", "try.**"}}}; !reflect.DeepEqual(m.Streams, want) {
t.Errorf("streams = %+v, want %+v", m.Streams, want)
}
// Only the checker's account publishes try.* on its stream checks: the stream is declared by its owner alone, and no other
// module of the repository declares a stream carrying a try event (the indexer and the others only consume them).
siblings, err := filepath.Glob("../*/module.json")
if err != nil || len(siblings) < 2 {
t.Fatalf("the sibling manifests = %v, %v", siblings, err)
}
for _, f := range siblings {
raw, err := os.ReadFile(f)
if err != nil {
t.Fatal(err)
}
var other struct {
Module string `json:"module"`
Streams []stream `json:"streams"`
}
if err := json.Unmarshal(raw, &other); err != nil {
t.Fatalf("%s: %v", f, err)
}
for _, st := range other.Streams {
for _, e := range st.Events {
if strings.HasPrefix(e, "try.") && other.Module != m.Module {
t.Errorf("%s declares the stream %s with %s: only the checker's account publishes try events", other.Module, st.Name, e)
}
}
}
}
// The read tools (issue 1686) are served by the tools bundle, not by the checker's process.
// The try's tools (ADR 0501 §5; the end by hand, issue 734) beside them: mesh-controller's try calls them. A lost check
// is asked again by hand (checker_ask_again).