textmachine/platform/internal/pgstore/sink_test.go

819 lines
35 KiB
Go

package pgstore
import (
"context"
"encoding/json"
"sync"
"testing"
"time"
"github.com/jackc/pgx/v5"
"textmachine/platform/internal/ingest"
"textmachine/platform/internal/money"
)
// sinkFixture is one live run with a chapter to fold units into.
func sinkFixture(t *testing.T) (*Store, *RunSink, string, StartedRun) {
t.Helper()
s, ctx := testDB(t)
now := fundedAccount(t, s, ctx, "u1", "10")
seedBook(t, s, ctx, "bk1", "u1", 500)
exec(t, s, ctx, `insert into chapters (id, book_id, number, units_total) values ('c1','bk1',1,50)`)
run, err := s.StartRun(ctx, StartRunInput{UserID: "u1", BookID: "bk1", CeilingChapters: 10,
Ceiling: money.MicroUSD(300_000), Now: now}, 0, nil)
if err != nil {
t.Fatal(err)
}
sink := s.NewRunSink(run.AttemptID, run.ID, "bk1")
if err := sink.Begin(t.Context(), ingest.Hello{EngineRunID: EngineStreamID(run.ID, 1)}); err != nil {
t.Fatal(err)
}
return s, sink, run.ID, run
}
func apply(t *testing.T, s *RunSink, seq int64, typ ingest.Type, data any) error {
t.Helper()
body, err := json.Marshal(data)
if err != nil {
t.Fatal(err)
}
return s.Apply(t.Context(), ingest.Envelope{Seq: seq, Type: typ, Data: body}, ingest.Cursor{Offset: seq * 10})
}
// assertBookFirst proves an operation takes the BOOK's row lock BEFORE the row `probe` asks about,
// which is the package's global order (lockBook) and the thing a concurrency stress test can only
// sample.
//
// It holds the book lock, starts the operation, waits until the operation is actually blocked, and
// then asks whether the other row is still free. Free means the operation is waiting for the book
// with nothing else held — book-first. Taken means it grabbed the other row on the way, which is the
// order that deadlocks against every other transaction in this package.
func assertBookFirst(t *testing.T, s *Store, ctx context.Context, bookID string, probe func(context.Context, pgx.Tx) error, op func() error) {
t.Helper()
holder, err := s.pool.Begin(ctx)
if err != nil {
t.Fatal(err)
}
defer func() { _ = holder.Rollback(ctx) }()
var id string
if err := holder.QueryRow(ctx, `select id from books where id = $1 for update`, bookID).Scan(&id); err != nil {
t.Fatal(err)
}
done := make(chan error, 1)
go func() { done <- op() }()
// Wait for the operation to block on something. Without this the probe below could run before the
// operation had taken any lock at all and pass for the wrong reason.
deadline := time.Now().Add(20 * time.Second)
for {
// pg_stat_activity, not pg_locks: a transaction waiting for a ROW lock waits on the HOLDER's
// transactionid, and those pg_locks rows carry a null database, so a database-scoped query over
// pg_locks never sees the wait this test is built on.
var blocked int
if err := s.pool.QueryRow(ctx, `
select count(*) from pg_stat_activity
where datname = current_database() and wait_event_type = 'Lock'`).Scan(&blocked); err != nil {
t.Fatal(err)
}
if blocked > 0 {
break
}
if time.Now().After(deadline) {
t.Fatal("the operation never blocked on the book lock")
}
time.Sleep(10 * time.Millisecond)
}
tx, err := s.pool.Begin(ctx)
if err != nil {
t.Fatal(err)
}
err = probe(ctx, tx)
_ = tx.Rollback(ctx)
if err != nil {
t.Errorf("the row was already locked while the book was not: %v — this order deadlocks", err)
}
if err := holder.Rollback(ctx); err != nil {
t.Fatal(err)
}
if err := <-done; err != nil {
t.Errorf("the operation itself failed: %v", err)
}
}
// lockedAttempt and lockedReservation are the probes assertBookFirst asks with: NOWAIT turns "someone
// else holds this row" into an error instead of a wait.
func lockedAttempt(attemptID int64) func(context.Context, pgx.Tx) error {
return func(ctx context.Context, tx pgx.Tx) error {
var id int64
return tx.QueryRow(ctx, `select id from run_attempts where id = $1 for update nowait`, attemptID).Scan(&id)
}
}
func lockedReservation(key string) func(context.Context, pgx.Tx) error {
return func(ctx context.Context, tx pgx.Tx) error {
var id string
return tx.QueryRow(ctx,
`select engine_run_id from reservations where engine_run_id = $1 for update nowait`, key).Scan(&id)
}
}
// Every path that touches a run's book and one of the rows below takes them in the package's order,
// asserted directly rather than sampled: a stress test can only fail to observe an inversion.
func TestTheMaterializerTakesTheBookBeforeTheAttempt(t *testing.T) {
s, sink, _, run := sinkFixture(t)
ctx := t.Context()
body, err := json.Marshal(ingest.Progress{Draft: ingest.Counter{Done: 1, Total: 50}})
if err != nil {
t.Fatal(err)
}
assertBookFirst(t, s, ctx, "bk1", lockedAttempt(run.AttemptID), func() error {
return sink.Apply(ctx, ingest.Envelope{Seq: 2, Type: ingest.TypeProgress, Data: body}, ingest.Cursor{Offset: 20})
})
}
func TestARestartTakesTheBookBeforeTheAttempt(t *testing.T) {
s, _, runID, run := sinkFixture(t)
ctx := t.Context()
assertBookFirst(t, s, ctx, "bk1", lockedAttempt(run.AttemptID), func() error {
_, err := s.RestartRun(ctx, RestartInput{RunID: runID, AttemptID: run.AttemptID,
UserID: "u1", BookID: "bk1", Ceiling: money.MicroUSD(1_000), Now: time.Now().UTC()})
return err
})
}
// Deleting a book takes the book before the reservations it clears. No cycle exists for this pair
// today — the delete only touches CLOSED reservations, and nothing else locks those — but an
// invariant with an undeclared exception stops being one, and the next writer of this package reads
// the invariant, not the exception.
func TestDeletingABookTakesItBeforeItsReservations(t *testing.T) {
s, ctx := testDB(t)
now := fundedAccount(t, s, ctx, "u1", "10")
seedBook(t, s, ctx, "bk1", "u1", 500)
if err := s.Hold(ctx, "u1", "bk1", "run_x#1", money.MicroUSD(300_000), now); err != nil {
t.Fatal(err)
}
if err := s.Release(ctx, "run_x#1", now); err != nil {
t.Fatal(err)
}
assertBookFirst(t, s, ctx, "bk1", lockedReservation("run_x#1"), func() error {
return s.DeleteBook(ctx, "bk1")
})
}
// The materializer and the reconciler work on one run at the same time as a matter of course — a
// live stream while a sweep ends an attempt — and they touch the same two rows. Taking them in
// opposite orders is a deadlock Postgres resolves by aborting a side, which costs a sweep or an API
// call and, worse, used to reach the reconciler as "this journal cannot be read" and quarantine the
// projection of a paying run for good.
//
// ⚠ Measured before the order was made global (lockBook): 258 of 300 concurrent pairs aborted.
func TestAMaterializerAndAReconcilerOnOneRunDoNotDeadlock(t *testing.T) {
s, sink, runID, run := sinkFixture(t)
const pairs = 60
var wg sync.WaitGroup
errs := make(chan error, 2*pairs)
body, err := json.Marshal(ingest.Progress{Draft: ingest.Counter{Done: 1, Total: 50}})
if err != nil {
t.Fatal(err)
}
ctx := t.Context()
now := time.Now().UTC()
// The restart path takes the same two rows and used to take them attempt-first as well. Its own
// LOGICAL outcome is not what this test is about — only the first of these closes the attempt and
// the rest legitimately find nothing — so its non-transient errors are tolerated and only a broken
// lock is not.
restartErrs := make(chan error, pairs)
for i := range pairs {
wg.Add(3)
go func() {
defer wg.Done()
seq := int64(i) + 2
errs <- sink.Apply(ctx, ingest.Envelope{Seq: seq, Type: ingest.TypeProgress, Data: body},
ingest.Cursor{Offset: seq * 10})
}()
go func() {
defer wg.Done()
_, err := s.FinishRun(ctx, RunEnding{RunID: runID, AttemptID: run.AttemptID,
Status: "ready", ExitResult: "exit-code", Now: now})
errs <- err
}()
go func() {
defer wg.Done()
_, err := s.RestartRun(ctx, RestartInput{RunID: runID, AttemptID: run.AttemptID,
UserID: "u1", BookID: "bk1", Ceiling: money.MicroUSD(1_000), Now: now})
restartErrs <- err
}()
}
wg.Wait()
close(errs)
close(restartErrs)
for err := range errs {
if IsTransient(err) {
t.Fatalf("the materializer and the reconciler deadlocked: %v", err)
}
if err != nil {
t.Errorf("concurrent write: %v", err)
}
}
for err := range restartErrs {
if IsTransient(err) {
t.Fatalf("the materializer and a restart deadlocked: %v", err)
}
}
}
// Each wave moves its OWN counter and no other. Which of the two then says a chapter is finished is
// a question asked at read time, of the pipeline the engine announced (pgstore.finishedUnits).
func TestADraftUnitMovesTheDraftCounterAndNotTheEditOne(t *testing.T) {
s, sink, _, _ := sinkFixture(t)
if err := apply(t, sink, 2, ingest.TypeUnitDone,
ingest.UnitDone{Chapter: 1, Unit: 1, Wave: "draft", Shipped: true}); err != nil {
t.Fatal(err)
}
var draft, edit int
if err := s.pool.QueryRow(t.Context(),
`select units_draft_done, units_edit_done from chapters where id='c1'`).
Scan(&draft, &edit); err != nil {
t.Fatal(err)
}
if draft != 1 || edit != 0 {
t.Errorf("after a draft unit: draft=%d edit=%d", draft, edit)
}
}
// The fold is an ASSIGNMENT, and this is the property that makes it one: the SAME unit announced
// again — by a re-read of the journal, or by a later process re-announcing what a crash lost before
// it reached the file — leaves the counters where they were.
//
// It is the consumer's duty and not the emitter's: delivery on this seam is at-least-once by
// ratification (D39.119 п.3), so `+ 1` per line was a counter that could only be right by luck. Two
// distinct units on the same chapter still count as two, which is the other half — an idempotent
// fold that also swallows real work is worse than the increment it replaces.
//
// Mutation caught: going back to `units_draft_done + 1`, and dropping the (chapter, unit, wave)
// primary key to (chapter, wave).
func TestAUnitAnnouncedTwiceIsCountedOnce(t *testing.T) {
s, sink, _, _ := sinkFixture(t)
counts := func() (int, int) {
t.Helper()
var draft, edit int
if err := s.pool.QueryRow(t.Context(),
`select units_draft_done, units_edit_done from chapters where id='c1'`).
Scan(&draft, &edit); err != nil {
t.Fatal(err)
}
return draft, edit
}
unit := ingest.UnitDone{Chapter: 1, Unit: 4, Wave: ingest.WaveDraft, Shipped: true}
if err := apply(t, sink, 2, ingest.TypeUnitDone, unit); err != nil {
t.Fatal(err)
}
// A NEW seq carrying a unit that was already announced. That is not a duplicate the cursor can
// absorb — the high-water mark has moved past it — which is exactly the case the identity is for.
if err := apply(t, sink, 3, ingest.TypeUnitDone, unit); err != nil {
t.Fatal(err)
}
if draft, edit := counts(); draft != 1 || edit != 0 {
t.Errorf("one unit announced twice: draft=%d edit=%d, want 1/0", draft, edit)
}
// A different unit of the same chapter is a different fact and does move the counter.
unit.Unit = 9
if err := apply(t, sink, 4, ingest.TypeUnitDone, unit); err != nil {
t.Fatal(err)
}
// …and the same unit resolved by the OTHER wave is a different fact too.
unit.Wave, unit.Flagged, unit.Reason = ingest.WaveEdit, true, "glossary_miss"
if err := apply(t, sink, 5, ingest.TypeUnitDone, unit); err != nil {
t.Fatal(err)
}
if draft, edit := counts(); draft != 2 || edit != 1 {
t.Errorf("two draft units and one edit: draft=%d edit=%d, want 2/1", draft, edit)
}
// What the engine SAID about the unit is kept and not just the fact that it happened. The stream
// delivers a disposition once and the cursor does not rewind, so a field dropped here is a field
// lost — and shipped-and-flagged together is exactly the pair the contract derives unit state
// from.
var reason string
var shipped, flagged bool
if err := s.pool.QueryRow(t.Context(), `
select reason, shipped, flagged from unit_resolutions
where book_id='bk1' and chapter=1 and unit=9 and wave='edit'`).Scan(&reason, &shipped, &flagged); err != nil {
t.Fatal(err)
}
if reason != "glossary_miss" || !shipped || !flagged {
t.Errorf("the unit was recorded as reason=%q shipped=%v flagged=%v", reason, shipped, flagged)
}
}
// A wave the engine invents in a later minor is IGNORED, not fatal. The minor rule says an unknown
// value is ignored; letting it through would put it in front of a column constraint that refuses it
// and take the whole materialization of a paying run down with it.
func TestAWaveThisBuildDoesNotKnowIsIgnoredRatherThanFatal(t *testing.T) {
s, sink, runID, _ := sinkFixture(t)
if err := apply(t, sink, 2, ingest.TypeUnitDone,
ingest.UnitDone{Chapter: 1, Unit: 1, Wave: "polish", Shipped: true}); err != nil {
t.Fatalf("an unknown wave broke the materializer: %v", err)
}
var n int
if err := s.pool.QueryRow(t.Context(),
`select count(*) from unit_resolutions where book_id='bk1'`).Scan(&n); err != nil {
t.Fatal(err)
}
if n != 0 {
t.Errorf("an unknown wave was stored (%d rows)", n)
}
// And the cursor still moved: an ignored event is applied, not skipped, or the stream stalls.
var seq int64
if err := s.pool.QueryRow(t.Context(),
`select last_seq from run_attempts where engine_run_id = $1`, EngineStreamID(runID, 1)).Scan(&seq); err != nil {
t.Fatal(err)
}
if seq != 2 {
t.Errorf("last_seq = %d after an ignored event, want 2", seq)
}
}
// WHICH ceiling stopped the run decides the reason, and the reason decides whether a resume can do
// anything about it. `day` is the engine's own limit out of a book.yaml the platform never wrote:
// calling it `credit_exhausted` would tell the user their credit ran out and would light the
// account-level halted flag, which keys on exactly that value.
//
// Mutation caught: mapping every scope onto credit_exhausted, and mapping an ABSENT scope onto
// daily_ceiling.
func TestTheCeilingScopeDecidesWhichPauseTheRunGets(t *testing.T) {
for _, tc := range []struct {
scope string
want string
}{
{ingest.ScopeBook, PausedCreditExhausted},
{ingest.ScopeDay, PausedDailyCeiling},
// Neither of the two the platform can place. Named as such rather than guessed at: the run is
// resumable under any of the three reasons, so a guess buys nothing and `credit_exhausted`
// would light the account's halted flag over somebody else's limit.
{"", PausedCeilingUnknown}, // a stream written before the field existed
{"account", PausedCeilingUnknown}, // a scope a later minor invents
} {
s, sink, runID, _ := sinkFixture(t)
if err := apply(t, sink, 2, ingest.TypeCeiling,
ingest.Ceiling{Halted: true, Scope: tc.scope}); err != nil {
t.Fatal(err)
}
var status, reason string
if err := s.pool.QueryRow(t.Context(),
`select status, coalesce(paused_reason,'') from runs where id=$1`, runID).
Scan(&status, &reason); err != nil {
t.Fatal(err)
}
// `paused`, NEVER `failed`, for every scope: the stop is resumable and the contract forbids it
// (§BookStatus, PD-113).
if status != "paused" || reason != tc.want {
t.Errorf("scope %q gave %q/%q, want paused/%s", tc.scope, status, reason, tc.want)
}
}
}
// A ceiling event that did NOT halt the run must not pause it.
func TestACeilingEventThatDidNotHaltChangesNothing(t *testing.T) {
s, sink, runID, _ := sinkFixture(t)
if err := apply(t, sink, 2, ingest.TypeCeiling, ingest.Ceiling{Halted: false}); err != nil {
t.Fatal(err)
}
var status, reason string
if err := s.pool.QueryRow(t.Context(),
`select status, coalesce(paused_reason,'') from runs where id=$1`, runID).Scan(&status, &reason); err != nil {
t.Fatal(err)
}
if status != "translating" || reason != "" {
t.Errorf("a ceiling event with halted=false gave status %q reason %q", status, reason)
}
}
// The bank stop arriving on the STREAM is the same product fact as exit code 3, and the exit path is
// pinned elsewhere; this is the other half.
func TestABankStopOnTheStreamPutsTheRunInAwaitingBank(t *testing.T) {
s, sink, runID, _ := sinkFixture(t)
if err := apply(t, sink, 2, ingest.TypeBankStop, ingest.BankStop{TermsProposed: 7}); err != nil {
t.Fatal(err)
}
var status string
if err := s.pool.QueryRow(t.Context(), `select status from runs where id=$1`, runID).Scan(&status); err != nil {
t.Fatal(err)
}
if status != "awaiting_bank" {
t.Errorf("after a bank stop the run is %q", status)
}
}
// An event with no payload is refused rather than applied as a zero struct: a `progress` line with
// empty data would otherwise write draft 0/0 over a live projection.
func TestAnEventWithNoPayloadIsRefused(t *testing.T) {
_, sink, _, _ := sinkFixture(t)
err := sink.Apply(t.Context(), ingest.Envelope{Seq: 2, Type: ingest.TypeProgress}, ingest.Cursor{})
if err == nil {
t.Fatal("an event with no data was applied")
}
}
// An ETA of zero is ABSENT, not zero: the screen renders without an estimate rather than showing
// "0 s left".
func TestAZeroEtaIsStoredAsAbsent(t *testing.T) {
s, sink, runID, _ := sinkFixture(t)
if err := apply(t, sink, 2, ingest.TypeProgress, ingest.Progress{
Draft: ingest.Counter{Done: 1, Total: 10}, ETASeconds: 0}); err != nil {
t.Fatal(err)
}
var eta *int
if err := s.pool.QueryRow(t.Context(), `select eta_seconds from runs where id=$1`, runID).Scan(&eta); err != nil {
t.Fatal(err)
}
if eta != nil {
t.Errorf("an ETA of zero was stored as %d", *eta)
}
}
// The resync materializes the engine's PER-WAVE split and never LOWERS a counter the stream has
// already moved further.
//
// ⚠ This replaces a pin whose premise the engine retired: while status carried only one end-to-end
// aggregate (backlog row 99) the resync could put it in the edit counter and nothing else, and the
// old test asserted that the draft counter stayed untouched. The split landed on 09.08 (D39.122), so
// asserting the old shape would now pin a limitation instead of a property.
// ⚠ RE-POINTED with PD-411 and the claim NARROWED to what is true, which is the finding rather than
// a concession. Its predecessor was called «the resync materializes the phase split and never lowers
// a counter» and it asserted over `runs.draft_done/draft_total/edit_done/edit_total` — four columns
// with two writers and NO reader, dropped in migration 00031. So the property it proved was real of
// those columns and meaningless of the screen: the bar is derived from `chapters`, which this channel
// does not write at all. There is nothing left for a resync to walk backwards, because there is
// nothing left for a resync to move.
//
// What that leaves is a GAP, not a passing test: the repair channel for a run whose stream is
// quarantined materialises no progress, so such a run's bar stands still however faithfully `tmctl
// status` answers. It has a register row of its own. This test pins the honest remainder — the ETA,
// the freshness stamp and the announced SHAPE — and the gap is stated here so the next reader does
// not re-derive it from a green test.
func TestTheResyncRecordsFreshnessAndShapeAndNoProgress(t *testing.T) {
s, sink, runID, _ := sinkFixture(t)
// ⚠ The stream announces a DRAFT-ONLY shape, and the resync below announces an EDITING one. The
// two must differ or the shape assertion proves nothing: this pack's adversarial pass caught the
// first edition, where the stream had already written the same shape and deleting the
// recordWaveShape call from ApplyStatus entirely left the whole eighteen-package battery green.
if err := apply(t, sink, 2, ingest.TypeProgress, ingest.Progress{
Draft: ingest.Counter{Done: 9, Total: 20}, Edit: ingest.Counter{Done: 0, Total: 0},
ETASeconds: 99}); err != nil {
t.Fatal(err)
}
if editor, _ := bookShape(t, s, "bk1"); editor == nil || *editor {
t.Fatalf("the stream announced a draft-only shape and the book reads %v: the fixture cannot"+
" observe what the RESYNC writes", editor)
}
before := bookRevision(t, s, "bk1")
rep := ingest.StatusReport{TotalUnits: 20, Done: 11,
Progress: ingest.Progress{Draft: ingest.Counter{Done: 20, Total: 20},
Edit: ingest.Counter{Done: 11, Total: 20}}, ETASeconds: 42}
at := time.Now().UTC()
if err := s.ApplyStatus(t.Context(), runID, "bk1", rep, at); err != nil {
t.Fatal(err)
}
var eta *int
var stamp *time.Time
if err := s.pool.QueryRow(t.Context(),
`select eta_seconds, last_resync_at from runs where id=$1`, runID).Scan(&eta, &stamp); err != nil {
t.Fatal(err)
}
if eta == nil || *eta != 42 {
t.Errorf("the resync's eta folded to %v, want 42", eta)
}
if stamp == nil {
t.Error("the resync left no freshness stamp: how stale its figures are is the one thing only it can say")
}
// The SHAPE the report announced reaches the book — this is the resync's other real effect, and
// the one that decides how the whole library counts this book. The stream said draft-only; only
// ApplyStatus can have written what is read here.
editor, epoch := bookShape(t, s, "bk1")
if editor == nil || !*editor {
t.Errorf("the report announced an edit total of 20 and the book's shape folded to %v: the"+
" resync's own announcement never reached the book", editor)
}
if epoch != 1 {
t.Errorf("the resync crossed a shape boundary and the epoch reads %d, want 1", epoch)
}
// And a client is told to re-read, because a resync is a materialisation like any other.
if after := bookRevision(t, s, "bk1"); after <= before {
t.Errorf("the resync left the book at revision %d: a watching client is never told to re-read", after)
}
// ⚠ THE GAP, asserted so it cannot be forgotten: the bar is derived from `chapters`, and the
// resync does not write there. A quarantined run shows what its stream last delivered.
var draftDone int
if err := s.pool.QueryRow(t.Context(),
`select units_draft_done from chapters where id='c1'`).Scan(&draftDone); err != nil {
t.Fatal(err)
}
if draftDone != 0 {
t.Errorf("the resync moved a chapter counter to %d: if this channel has learned to materialise"+
" progress, the register row that says it cannot is stale and the bar's source has changed", draftDone)
}
}
func bookShape(t *testing.T, s *Store, bookID string) (editor *bool, epoch int) {
t.Helper()
if err := s.pool.QueryRow(t.Context(),
`select epoch_editor, shape_epoch from books where id = $1`, bookID).Scan(&editor, &epoch); err != nil {
t.Fatal(err)
}
return editor, epoch
}
func bookRevision(t *testing.T, s *Store, bookID string) int64 {
t.Helper()
var rev int64
if err := s.pool.QueryRow(t.Context(), `select revision from books where id = $1`, bookID).
Scan(&rev); err != nil {
t.Fatal(err)
}
return rev
}
// What a resync produces is stale by up to the poll interval, and the ONLY thing that says how stale
// is the stamp it leaves. A run being tailed has none: the stream carries its own freshness.
func TestAResyncRecordsWhenItWasTaken(t *testing.T) {
s, _, runID, _ := sinkFixture(t)
var stamp *time.Time
if err := s.pool.QueryRow(t.Context(), `select last_resync_at from runs where id=$1`, runID).Scan(&stamp); err != nil {
t.Fatal(err)
}
if stamp != nil {
t.Fatalf("a run that was never resynced is stamped %s", stamp)
}
now := time.Now().UTC().Truncate(time.Millisecond)
if err := s.ApplyStatus(t.Context(), runID, "bk1", ingest.StatusReport{TotalUnits: 20}, now); err != nil {
t.Fatal(err)
}
if err := s.pool.QueryRow(t.Context(), `select last_resync_at from runs where id=$1`, runID).Scan(&stamp); err != nil {
t.Fatal(err)
}
if stamp == nil || !stamp.UTC().Equal(now) {
t.Fatalf("the resync stamped %v, want %s", stamp, now)
}
// And EVERY resync moves it. A stamp that only records the first one answers "when did polling
// begin", which is not the question a reader of a stale projection is asking.
later := now.Add(7 * time.Minute)
if err := s.ApplyStatus(t.Context(), runID, "bk1", ingest.StatusReport{TotalUnits: 20}, later); err != nil {
t.Fatal(err)
}
if err := s.pool.QueryRow(t.Context(), `select last_resync_at from runs where id=$1`, runID).Scan(&stamp); err != nil {
t.Fatal(err)
}
if stamp == nil || !stamp.UTC().Equal(later) {
t.Errorf("the second resync left the stamp at %v, want %s", stamp, later)
}
}
// Cross-family review of the acceptance dofix (M2): a stop that lands before its engine was spawned
// leaves an attempt that is ENDED and carries no engine run id — a shape this zone could not produce
// before the stop path existed. An old materializer still holding that attempt would then bind it to
// the handshake of the attempt that REPLACED it and apply the same journal a second time, counting
// every unit and chapter twice. Binding is refused for an attempt that is over.
func TestAMaterializerOfAnEndedAttemptDoesNotAdoptTheNextAttemptsEngine(t *testing.T) {
s, ctx := testDB(t)
now := fundedAccount(t, s, ctx, "u1", "10")
seedBook(t, s, ctx, "bk1", "u1", 500)
run, err := s.StartRun(ctx, StartRunInput{UserID: "u1", BookID: "bk1", CeilingChapters: 10,
Ceiling: money.MicroUSD(300_000), Now: now}, 0, nil)
if err != nil {
t.Fatal(err)
}
// The stop-before-spawn shape: the attempt is closed, and its name is cleared to model an attempt
// written by a build from BEFORE the platform named streams at creation — the only shape that can
// still reach Begin's adoption path at all.
exec(t, s, ctx, `update run_attempts set ended_at = now(), engine_run_id = null where id = $1`, run.AttemptID)
stale := s.NewRunSink(run.AttemptID, run.ID, "bk1")
if err := stale.Begin(ctx, ingest.Hello{EngineRunID: "eng-of-the-next-attempt"}); err == nil {
t.Fatal("an ended attempt adopted an engine run: its journal would be materialized twice")
}
var bound *string
if err := s.pool.QueryRow(ctx,
`select engine_run_id from run_attempts where id = $1`, run.AttemptID).Scan(&bound); err != nil {
t.Fatal(err)
}
if bound != nil {
t.Fatalf("the ended attempt was bound to %q", *bound)
}
}
// A resolution that arrives LATE must not drag a unit's disposition backwards.
//
// The fold assigns rather than increments, so without a comparison the upsert is "last to arrive
// wins" — and what a unit IS would then be a property of delivery order over an at-least-once seam
// (D39.119 п.3) rather than of the data. The engine itself does not emit such an inversion, and the
// acceptance was right to strike the claim that it does (see unitDone): every announcement, the
// re-announcement of a lost line included, carries the clock of the process making it. This pins the
// consumer's own guarantee, which is what a consumer of an at-least-once stream owes.
//
// The COUNTERS are unaffected either way — they count rows — which is exactly why this needs its own
// assertion: the ratified half of the fold cannot see it.
//
// Mutation caught: dropping `where excluded.at >= unit_resolutions.at` from unitDone.
func TestAReAnnouncedOlderResolutionDoesNotOverwriteANewerOne(t *testing.T) {
s, sink, _, _ := sinkFixture(t)
older := time.Date(2026, 8, 14, 10, 0, 0, 0, time.UTC)
newer := older.Add(time.Hour)
unit := ingest.UnitDone{Chapter: 1, Unit: 3, Wave: ingest.WaveEdit}
apply := func(seq int64, at time.Time, u ingest.UnitDone) {
t.Helper()
body, err := json.Marshal(u)
if err != nil {
t.Fatal(err)
}
if err := sink.Apply(t.Context(),
ingest.Envelope{Seq: seq, Type: ingest.TypeUnitDone, Time: at, Data: body},
ingest.Cursor{Offset: seq * 10}); err != nil {
t.Fatal(err)
}
}
// The redrive's verdict: the unit ships.
unit.Shipped, unit.Flagged, unit.Reason = true, false, ""
apply(2, newer, unit)
// …and then the older announcement of the same unit arrives, under a fresh seq.
unit.Shipped, unit.Flagged, unit.Reason = false, true, "glossary_miss"
apply(3, older, unit)
var shipped, flagged bool
var reason string
if err := s.pool.QueryRow(t.Context(), `
select shipped, flagged, reason from unit_resolutions
where book_id='bk1' and chapter=1 and unit=3 and wave='edit'`).Scan(&shipped, &flagged, &reason); err != nil {
t.Fatal(err)
}
if !shipped || flagged || reason != "" {
t.Errorf("the unit regressed to shipped=%v flagged=%v reason=%q on a re-announcement of an older event",
shipped, flagged, reason)
}
// And a genuinely newer disposition still wins, or the guard would have frozen the first answer.
unit.Shipped, unit.Flagged, unit.Reason = false, true, "sanitizer_stripped"
apply(4, newer.Add(time.Minute), unit)
if err := s.pool.QueryRow(t.Context(), `
select shipped, flagged, reason from unit_resolutions
where book_id='bk1' and chapter=1 and unit=3 and wave='edit'`).Scan(&shipped, &flagged, &reason); err != nil {
t.Fatal(err)
}
if shipped || !flagged || reason != "sanitizer_stripped" {
t.Errorf("a newer disposition did not win: shipped=%v flagged=%v reason=%q", shipped, flagged, reason)
}
}
// WHICH cut of the source the engine is working from reaches the book from the handshake.
//
// ⚠ It used to be recorded in `Begin`, and `Begin` is unreachable for any attempt this build creates:
// the platform names an attempt's stream when the row is written, so the tailer is never without a
// name and never adopts a handshake. Moving the effect to where the handshake arrives as an ORDINARY
// event is what kept the fact from being silently lost — a reader joining a persisted chapter tree
// cut by a different chunker would be joining on ordinals that were re-numbered.
//
// Mutation caught: removing the TypeHello case from effect.
func TestTheChunkerVersionOfTheStreamReachesTheBook(t *testing.T) {
s, sink, _, _ := sinkFixture(t)
if err := apply(t, sink, 2, ingest.TypeHello,
ingest.Hello{StreamVersion: "1.1", EngineRunID: "x", BookID: "bk1", ChunkerVersion: "chunk-2026.08"}); err != nil {
t.Fatal(err)
}
var got string
if err := s.pool.QueryRow(t.Context(),
`select coalesce(chunker_version,'') from books where id='bk1'`).Scan(&got); err != nil {
t.Fatal(err)
}
if got != "chunk-2026.08" {
t.Errorf("chunker_version = %q after the handshake", got)
}
// A handshake without one leaves the recorded value alone rather than blanking it.
if err := apply(t, sink, 3, ingest.TypeHello,
ingest.Hello{StreamVersion: "1.1", EngineRunID: "x", BookID: "bk1"}); err != nil {
t.Fatal(err)
}
if err := s.pool.QueryRow(t.Context(),
`select coalesce(chunker_version,'') from books where id='bk1'`).Scan(&got); err != nil {
t.Fatal(err)
}
if got != "chunk-2026.08" {
t.Errorf("a handshake with no chunker version overwrote the recorded one: %q", got)
}
}
// A flagged unit moves the chapter's NOTE counter as well as its wave counter, in the same statement
// and from the same rows. The card sums those counters, so a sink that stopped maintaining this one
// would leave every book on the library page reading zero notes while `GET /notes` listed them.
//
// Mutation caught: dropping note_count from the fold.
func TestAFlaggedUnitMovesTheChaptersNoteCounter(t *testing.T) {
s, sink, _, _ := sinkFixture(t)
count := func() int {
t.Helper()
var n int
if err := s.pool.QueryRow(t.Context(),
`select note_count from chapters where id='c1'`).Scan(&n); err != nil {
t.Fatal(err)
}
return n
}
if got := count(); got != 0 {
t.Fatalf("the chapter starts at %d notes: this fixture cannot observe the move", got)
}
if err := apply(t, sink, 2, ingest.TypeUnitDone,
ingest.UnitDone{Chapter: 1, Unit: 0, Wave: ingest.WaveEdit, Shipped: true, Flagged: true,
Reason: "glossary_miss"}); err != nil {
t.Fatal(err)
}
if got := count(); got != 1 {
t.Errorf("after a flagged unit the chapter counts %d notes", got)
}
// An unflagged one does not, which is what makes the counter a NOTE counter.
if err := apply(t, sink, 3, ingest.TypeUnitDone,
ingest.UnitDone{Chapter: 1, Unit: 4, Wave: ingest.WaveEdit, Shipped: true}); err != nil {
t.Fatal(err)
}
if got := count(); got != 1 {
t.Errorf("a clean unit moved the note counter to %d", got)
}
}
// The progress FRAME and the run's card answer ONE bar — same done, same total, same stage — and
// the pin rides the flipped-flag fixture on purpose: the frame is minted inside the transaction
// that flips `edit_wave` (recordWaveShape), which is exactly where a frame computed from different
// fragments than the card's would diverge first. This is the wire half of the through-bar (row
// 200): `stage` had no value pin at all, and a planted done/total swap in the payload outlived
// both batteries before it.
//
// Mutation caught: swapping done/total (or dropping stage) in emitProgress's payload; computing
// the frame from fragments other than the card's runDone/runTotal/runStage.
func TestTheProgressFrameMatchesTheCardOnAFlippedFlag(t *testing.T) {
s, ctx := testDB(t)
now := fundedAccount(t, s, ctx, "u1", "10")
seedBook(t, s, ctx, "bk1", "u1", 500)
if err := s.SaveStructure(ctx, "bk1", twoChapters("k1")); err != nil {
t.Fatal(err)
}
// The book's draft era: drafted end to end, the flag SAYS no editor, one chapter already edited
// by the time the engine speaks — so the bar is non-trivial in the very frame that flips.
exec(t, s, ctx, `update books set edit_wave = false where id = 'bk1'`)
exec(t, s, ctx, `update chapters set units_draft_done = units_total where book_id = 'bk1'`)
run, err := s.StartRun(ctx, StartRunInput{UserID: "u1", BookID: "bk1", CeilingChapters: 2,
Ceiling: money.MicroUSD(300_000), Now: now}, 0, nil)
if err != nil {
t.Fatal(err)
}
exec(t, s, ctx, `update chapters set units_edit_done = units_total
where book_id = 'bk1' and number = 1`)
sink := s.NewRunSink(run.AttemptID, run.ID, "bk1")
if err := sink.Begin(ctx, ingest.Hello{EngineRunID: EngineStreamID(run.ID, 1)}); err != nil {
t.Fatal(err)
}
// The engine's first progress event announces the editor: the flag flips INSIDE this apply, and
// the frame is minted right after it.
if err := apply(t, sink, 2, ingest.TypeProgress, ingest.Progress{
Draft: ingest.Counter{Done: 3, Total: 3}, Edit: ingest.Counter{Done: 2, Total: 3},
}); err != nil {
t.Fatal(err)
}
card, err := s.ReadRun(ctx, "u1", run.ID)
if err != nil {
t.Fatal(err)
}
if card.Progress.Done != 1 || card.Progress.Total != 2 || card.Progress.Stage != "editing" {
t.Fatalf("the fixture's card reads %d/%d %q: it cannot tell a swap from the truth",
card.Progress.Done, card.Progress.Total, card.Progress.Stage)
}
frames, err := s.ReadFrames(ctx, "bk1", 0, 100)
if err != nil {
t.Fatal(err)
}
var last json.RawMessage
for _, f := range frames {
if f.Event == FrameProgress {
last = f.Data
}
}
if last == nil {
t.Fatal("the progress event minted no frame")
}
var frame struct {
Progress struct {
Done int `json:"done"`
Total int `json:"total"`
Stage string `json:"stage"`
ETASeconds *int `json:"eta_seconds"`
} `json:"progress"`
}
if err := json.Unmarshal(last, &frame); err != nil {
t.Fatalf("the frame's payload: %v (%s)", err, last)
}
if frame.Progress.Done != card.Progress.Done || frame.Progress.Total != card.Progress.Total ||
frame.Progress.Stage != card.Progress.Stage {
t.Errorf("the frame reads %d/%d %q and the card %d/%d %q — one bar, two answers",
frame.Progress.Done, frame.Progress.Total, frame.Progress.Stage,
card.Progress.Done, card.Progress.Total, card.Progress.Stage)
}
if (frame.Progress.ETASeconds == nil) != (card.Progress.ETASeconds == nil) {
t.Errorf("eta: frame %v, card %v", frame.Progress.ETASeconds, card.Progress.ETASeconds)
}
}