The events table had no attempt column, so two attempts of the same task were indistinguishable: a retried task's stage events interleaved into one timeline and nothing could say which run produced a given result. - events gains an `attempt` column through the existing presence-gated migrateTasks loop rather than a version bump, so user_version stays 3 and an existing archie.db is migrated in place. A legacy-database test builds the pre-change schema and proves both that existing rows survive unattributed (attempt = 0) and that a new value round-trips. - Event.attempt = 11 and ReadTaskLogRequest.stage = 9 are field additions to existing messages: no new RPC, no new table, and no change to any canonical error message that crosses the state-store wire. - Every task event the workflow engine emits now carries the attempt. Producers that have no attempt yet (task_queued, binding_dispatch_failure) stay unattributed rather than guessing one. - Changed files are captured as a durable changes_captured event at commit and push time, reusing the existing three-dot diff primitive: paths from FilePatches(), counts from Stats(), base_sha recorded as the merge base actually diffed rather than the base tip, a 200-file cap with totals kept complete. Consumption goes through an unexported optional interface so workflow.Trees -- the yaegi capability surface for interpreted stage plugins -- is not widened. A capture is also taken once the pull request exists, so the capture can link it. - The per-attempt configuration is captured as a durable config_captured event carrying the non-secret config.TaskConfig, instead of a new table plus two new state-store RPCs: events are already this system's per-attempt provenance stream, so a table would have given one feature two provenance mechanisms. Redaction is proven against canary secrets. Measured payload: 539 bytes typical, 2333 bytes at 20 models. - Captures carry a schema marker and are refused unless their key set matches the declared payload shape, so drift surfaces as "recorded but could not be read" rather than as a capture of zeroes. - Three read endpoints serve the attempt rail, the changed-file captures and a raw debug view, mapping to UI-owned response types rather than dumping store types. - The task-log read gains a stage filter, bound to the stage logger for that stage's own lines; agent and tool output records no stage. A task-scoped grant may already insert events on its own task, so this provenance is exactly as trustworthy as the rest of that task's event stream. That limit is documented rather than papered over.
131 lines
4.5 KiB
Go
131 lines
4.5 KiB
Go
package agentexec
|
|
|
|
import (
|
|
"encoding/json"
|
|
"errors"
|
|
"log/slog"
|
|
"testing"
|
|
"time"
|
|
|
|
"github.com/samcharles93/archie-core/internal/events"
|
|
)
|
|
|
|
var errPublishBoom = errors.New("publish boom")
|
|
|
|
// eventPublisherStub is an EventPublisher that records every call and can be
|
|
// told to fail, mirroring fakePublisher in systemlog_test.go for the same
|
|
// reason: assert both the happy path and "one failure must not stop the
|
|
// rest."
|
|
type eventPublisherStub struct {
|
|
subjects []string
|
|
payloads [][]byte
|
|
failWith error
|
|
}
|
|
|
|
func (p *eventPublisherStub) Publish(subject string, data []byte) error {
|
|
p.subjects = append(p.subjects, subject)
|
|
p.payloads = append(p.payloads, append([]byte(nil), data...))
|
|
return p.failWith
|
|
}
|
|
|
|
// TestForwardTaskEventsPublishesEachEventOnItsTaskSubject is the core
|
|
// contract: everything published on the bus while ForwardTaskEvents is
|
|
// running arrives at pub, JSON-decodable back into the same Event, on
|
|
// SubjectForEvents(taskID) -- regardless of which task's events happen to be
|
|
// flowing, since taskID is a ForwardTaskEvents parameter, not read from the
|
|
// event itself.
|
|
func TestForwardTaskEventsPublishesEachEventOnItsTaskSubject(t *testing.T) {
|
|
bus := events.NewBus()
|
|
sub := bus.Subscribe(8)
|
|
pub := &eventPublisherStub{}
|
|
|
|
done := make(chan struct{})
|
|
go func() {
|
|
ForwardTaskEvents(sub, pub, 42, slog.New(slog.DiscardHandler))
|
|
close(done)
|
|
}()
|
|
|
|
want := []events.Event{
|
|
{Kind: events.KindStageStart, TaskID: 42, Stage: "baseline", Attempt: 1},
|
|
{Kind: events.KindStageFinish, TaskID: 42, Stage: "baseline", Attempt: 1, Data: map[string]any{"duration_ms": float64(120)}},
|
|
{Kind: events.KindParked, TaskID: 42, Attempt: 1, Detail: "baseline red"},
|
|
// R4's document rides this bridge too: a worker captures its own
|
|
// effective configuration, and the daemon is a different process, so
|
|
// the payload has to survive the hop or the Config tab reads nothing.
|
|
// The daemon's own key for it is asserted below by name.
|
|
{Kind: events.KindConfigCaptured, TaskID: 42, Attempt: 1, Data: map[string]any{
|
|
"schema": events.ConfigCapturedSchema,
|
|
"document": map[string]any{"bot_user": "archie"},
|
|
}},
|
|
}
|
|
for _, e := range want {
|
|
bus.Publish(e)
|
|
}
|
|
sub.Close()
|
|
<-done
|
|
|
|
if len(pub.payloads) != len(want) {
|
|
t.Fatalf("published %d events, want %d", len(pub.payloads), len(want))
|
|
}
|
|
wantSubject := SubjectForEvents(42)
|
|
for i, payload := range pub.payloads {
|
|
if pub.subjects[i] != wantSubject {
|
|
t.Errorf("event %d subject = %q, want %q", i, pub.subjects[i], wantSubject)
|
|
}
|
|
var got events.Event
|
|
if err := json.Unmarshal(payload, &got); err != nil {
|
|
t.Fatalf("event %d undecodable: %v", i, err)
|
|
}
|
|
if got.Kind != want[i].Kind || got.TaskID != want[i].TaskID || got.Stage != want[i].Stage || got.Detail != want[i].Detail {
|
|
t.Errorf("event %d = %+v, want %+v", i, got, want[i])
|
|
}
|
|
if got.Attempt != want[i].Attempt {
|
|
t.Errorf("event %d attempt = %d, want %d (provenance must not be dropped on the hop)", i, got.Attempt, want[i].Attempt)
|
|
}
|
|
if got.Kind == events.KindConfigCaptured {
|
|
if got.Data["schema"] != events.ConfigCapturedSchema {
|
|
t.Errorf("event %d schema = %v, want %q", i, got.Data["schema"], events.ConfigCapturedSchema)
|
|
}
|
|
if _, renamed := got.Data["config"]; renamed {
|
|
t.Errorf(`event %d crossed under "config"; the producer writes the document under "document"`, i)
|
|
}
|
|
doc, ok := got.Data["document"].(map[string]any)
|
|
if !ok || doc["bot_user"] != "archie" {
|
|
t.Errorf("event %d document payload = %#v, want the decoded document under its own key", i, got.Data["document"])
|
|
}
|
|
}
|
|
if got.At.IsZero() {
|
|
t.Errorf("event %d has zero At, want the bus-stamped publish time", i)
|
|
}
|
|
}
|
|
}
|
|
|
|
// TestForwardTaskEventsSurvivesAPublishFailure pins "one missing event must
|
|
// never abort the ones after it": event shipping must never block or fail
|
|
// the workflow run it is reporting on.
|
|
func TestForwardTaskEventsSurvivesAPublishFailure(t *testing.T) {
|
|
bus := events.NewBus()
|
|
sub := bus.Subscribe(8)
|
|
pub := &eventPublisherStub{failWith: errPublishBoom}
|
|
|
|
done := make(chan struct{})
|
|
go func() {
|
|
ForwardTaskEvents(sub, pub, 7, slog.New(slog.DiscardHandler))
|
|
close(done)
|
|
}()
|
|
|
|
bus.Publish(events.Event{Kind: events.KindStageStart, TaskID: 7})
|
|
bus.Publish(events.Event{Kind: events.KindOutcome, TaskID: 7})
|
|
sub.Close()
|
|
|
|
select {
|
|
case <-done:
|
|
case <-time.After(2 * time.Second):
|
|
t.Fatal("ForwardTaskEvents did not return after sub.Close(); a publish failure blocked the loop")
|
|
}
|
|
|
|
if len(pub.payloads) != 2 {
|
|
t.Fatalf("published %d events despite failures, want 2 (both attempted)", len(pub.payloads))
|
|
}
|
|
}
|