Delete the checker's hold, its own capacity and places, its node moves and its read of the controller's plans-building, so a check waits only in the build seats' streams (issue 1887 A5b, closes 1685; ADR 0525 §4, §7)

This commit is contained in:
2026-10-23 16:09:34 +02:00
parent e4b639c6de
commit d9d1e57e3a
24 changed files with 76 additions and 839 deletions
@@ -143,32 +143,16 @@ func Tools(r check.Reader) []stdio.Tool {
return map[string]any{"previews": ps, "unreadable": bad}, err
}},
{Name: check.StatusReadTool,
Description: "What holds the checker back and what runs: the capacity per node (the setting, the default one step per node), whether the controller's " +
"delivery plans in their build phase hold a key (no new check starts while one does), the steps in flight on the build modules by node, and the " +
"candidates without a verdict for their heads (a group of hq's pull requests alone is left out: merge-check.sh judges it), and the " +
"candidates whose check waits by kind (waiting_by_kind: new, nodes, and trunk, which start after the other two). A key a build " +
"module left running (started past the step bound, or its build module not running on its node) is listed apart under left_running " +
"with why, and not counted. The places the checker holds are in its process, not in a bucket, and the answer says so. Read-only.",
Description: "What runs: the steps in flight on the build modules by node, the candidates without a verdict for their heads, " +
"and the candidates whose check is owed by kind (waiting_by_kind: new, nodes, and trunk, a re-judge pulled after the other two). A key a " +
"build module left running (started past the step bound, or its build module not running on its node) is listed apart under " +
"left_running with why, and not counted. Nothing holds the checker and it has no capacity of its own: a check is an ask on the " +
"build seats' streams, whose backlog the delivery view shows (novox/hq ADR 0525). Read-only.",
Input: map[string]any{"type": "object", "properties": map[string]any{}},
Run: func(map[string]any) (any, error) { return r.Status() }},
}
}
func capacity() (string, error) {
path := os.Getenv("MESH_CHECKER_SETTINGS")
if path == "" {
return os.Getenv("MESH_CHECKER_CAPACITY"), nil
}
raw, err := os.ReadFile(path)
if err != nil {
return "", err
}
var s struct {
Capacity string `json:"capacity"`
}
return s.Capacity, json.Unmarshal(raw, &s)
}
// stepBound is the checker's step bound as the checker reads it (MESH_CHECKER_STEP_BOUND), its default when unset or not a
// duration.
func stepBound() time.Duration {
@@ -180,9 +164,9 @@ func stepBound() time.Duration {
func main() {
r := check.Reader{VerdictState: read(stdio.State("verdicts")), KeptState: read(stdio.State("kept")), CandidateState: read(stdio.State("mesh-delivery-indexer.candidates")),
PreviewState: read(stdio.State("previews")), PlansState: read(stdio.State("mesh-controller.plans-building")),
StepStates: map[string]check.StateRead{"mesh-build-go": read(stdio.State("mesh-build-go.steps")), "mesh-build-node": read(stdio.State("mesh-build-node.steps"))},
Capacity: capacity, HaveState: read(stdio.State("node-engine.have")), StepBound: stepBound()}
PreviewState: read(stdio.State("previews")),
StepStates: map[string]check.StateRead{"mesh-build-go": read(stdio.State("mesh-build-go.steps")), "mesh-build-node": read(stdio.State("mesh-build-node.steps"))},
HaveState: read(stdio.State("node-engine.have")), StepBound: stepBound()}
tries := stdio.State(check.TriesState)
source := stdio.State("mesh-delivery-gate.source")
trunks := func() (map[string]string, error) {
@@ -105,43 +105,6 @@ func follow(ctx context.Context, w jetstream.KeyWatcher, each func(jetstream.Key
}
}
// stepsWaiter waits for a build module's step by watching its key in the module's steps bucket (ADR 0469 clause 2): the
// watch hands over the value the key holds now, so a step that ended before the watch began is read at once, and every
// change after. No timer: the caller bounds the wait with its context.
type stepsWaiter struct{ js bucketOpener }
func (w stepsWaiter) Wait(ctx context.Context, module, key string) (delivery.Step, error) {
kv, err := w.js.KeyValue(ctx, delivery.StepsBucket(module))
if err != nil {
return delivery.Step{}, fmt.Errorf("the bucket %s: %w", delivery.StepsBucket(module), err)
}
watch, err := kv.Watch(ctx, key)
if err != nil {
return delivery.Step{}, err
}
defer watch.Stop()
for {
select {
case <-ctx.Done():
return delivery.Step{}, ctx.Err()
case e, ok := <-watch.Updates():
if !ok {
return delivery.Step{}, errors.New("the watch of the step ended")
}
if e == nil || !opOf(e) {
continue
}
step, err := delivery.DecodeStep(e.Value())
if err != nil {
return delivery.Step{}, fmt.Errorf("the step %s is not a step: %w", key, err)
}
if step.State != delivery.StepRunning {
return step, nil
}
}
}
}
// manifestReader reads manifests.<digest> of the indexer and parses the manifest it holds.
type manifestReader struct{ kv jetstream.KeyValue }
@@ -1,26 +0,0 @@
package main
import (
"context"
"encoding/json"
"errors"
"github.com/nats-io/nats.go"
"mesh-delivery-checker/internal/check"
)
// heardCaller is the build runner's caller with the calls that reached nobody marked: no responders, or none answering within
// the call's bound. The step never began there, so the check may give it to another node (issue 1820). A call that was
// answered, with a failure or not, is left as it is.
type heardCaller struct {
call func(node, key string, body any) (json.RawMessage, error)
}
func (h heardCaller) CallOn(node, key string, body any) (json.RawMessage, error) {
raw, err := h.call(node, key, body)
if errors.Is(err, nats.ErrNoResponders) || errors.Is(err, nats.ErrTimeout) || errors.Is(err, context.DeadlineExceeded) {
return nil, &check.Unheard{Err: err}
}
return raw, err
}
@@ -1,30 +0,0 @@
package main
import (
"context"
"encoding/json"
"errors"
"fmt"
"testing"
"github.com/nats-io/nats.go"
"mesh-delivery-checker/internal/check"
)
func TestACallThatReachedNobodyIsMarkedAndAnAnsweredFailureIsNot(t *testing.T) { // 1820/AC6
for _, err := range []error{nats.ErrNoResponders, nats.ErrTimeout, context.DeadlineExceeded} {
h := heardCaller{call: func(string, string, any) (json.RawMessage, error) { return nil, fmt.Errorf("call: %w", err) }}
_, got := h.CallOn("g14", "m.t", nil)
var u *check.Unheard
if !errors.As(got, &u) {
t.Errorf("%v: %v is not marked unheard", err, got)
}
}
h := heardCaller{call: func(string, string, any) (json.RawMessage, error) { return nil, errors.New("the module refused") }}
_, got := h.CallOn("g14", "m.t", nil)
var u *check.Unheard
if errors.As(got, &u) {
t.Errorf("an answered failure %v is marked unheard", got)
}
}
@@ -1,101 +0,0 @@
package main
import (
"context"
"errors"
"log"
"sync"
"github.com/nats-io/nats.go/jetstream"
"mesh-delivery-checker/internal/check"
)
// holdWatch holds the checker back while a delivery builds (ADR 0484 clause 12): whether one does is the list of the keys of
// the controller's bucket of the delivery plans in their build phase, read each time a check is about to start. Its keys age
// out (a time to live, and no delete marker), which a watch never hears, so a watch cannot decide it (mesh-controller
// internal/link/holds.go lists the keys too); the watch is only the wake-up, for a plan that ended by a delete. The bucket
// is read as empty where it is absent (issue 1652 makes it) or cannot be read, saying so once. It may come into being after
// the checker started and the checker holds no timer (ADR 0437 §8), so each time a check is about to start and the bucket is
// not known yet, the checker looks for it.
type holdWatch struct {
ctx context.Context
js bucketOpener
keys func(ctx context.Context) ([]string, error)
changed func(ctx context.Context) // a plan key changed: the checks that waited may start
logger *log.Logger
errs chan error
mu sync.Mutex
watched bool
said bool
}
func newHoldWatch(ctx context.Context, js bucketOpener, changed func(context.Context), logger *log.Logger) *holdWatch {
return &holdWatch{ctx: ctx, js: js, changed: changed, logger: logger, errs: make(chan error, 1)}
}
// Held says whether a delivery builds now.
func (w *holdWatch) Held() bool {
w.ensure()
w.mu.Lock()
keys := w.keys
w.mu.Unlock()
if keys == nil {
return false
}
list, err := keys(w.ctx)
if err != nil {
w.say("the keys of %s cannot be listed, so it is read as empty: %v", check.PlansBuildingBucket, err)
return false
}
return len(list) > 0
}
func (w *holdWatch) say(format string, args ...any) {
w.mu.Lock()
defer w.mu.Unlock()
if !w.said {
w.said = true
w.logger.Printf(format, args...)
}
}
// ensure finds the bucket if it is not known and exists, and starts the wake-up watch.
func (w *holdWatch) ensure() {
w.mu.Lock()
if w.watched || w.js == nil {
w.mu.Unlock()
return
}
w.mu.Unlock()
kv, err := w.js.KeyValue(w.ctx, check.PlansBuildingBucket)
switch {
case errors.Is(err, jetstream.ErrBucketNotFound):
w.say("the bucket %s does not exist yet: it is read as empty, so no check is held back", check.PlansBuildingBucket)
return
case err != nil:
w.say("the bucket %s cannot be read, so it is read as empty: %v", check.PlansBuildingBucket, err)
return
}
w.mu.Lock()
defer w.mu.Unlock()
if w.watched {
return
}
w.watched, w.said = true, false
w.keys = func(ctx context.Context) ([]string, error) {
list, err := kv.Keys(ctx)
if errors.Is(err, jetstream.ErrNoKeysFound) {
return nil, nil
}
return list, err
}
go func() {
wake, err := kv.WatchAll(w.ctx)
if err != nil {
w.errs <- err
return
}
w.errs <- follow(w.ctx, wake, func(jetstream.KeyValueEntry) { w.changed(w.ctx) })
}()
}
@@ -1,35 +0,0 @@
package main
import (
"context"
"errors"
"io"
"log"
"testing"
)
// The controller's plans-building keys age out of the bucket (a time to live, no delete marker), which a watch never hears:
// whether a delivery builds is the list of the keys now, and the watch is only a wake-up (review D3).
func TestADeliveryHoldsWhileItsKeyIsListedAndNotWhenItAgedOutWithNoEvent(t *testing.T) {
listed := []string{"plan-1"}
w := &holdWatch{watched: true, logger: log.New(io.Discard, "", 0), keys: func(context.Context) ([]string, error) { return listed, nil }}
if !w.Held() {
t.Fatal("a plan key is listed and nothing holds")
}
listed = nil // the key aged out: no delete reached the watch
if w.Held() {
t.Error("the key is gone from the bucket and the checker is still held")
}
}
func TestABucketThatCannotBeListedIsReadAsEmpty(t *testing.T) {
w := &holdWatch{watched: true, logger: log.New(io.Discard, "", 0), keys: func(context.Context) ([]string, error) { return []string{"plan-1"}, errors.New("the bus is down") }}
if w.Held() {
t.Error("an unreadable bucket held the checker")
}
var absent holdWatch
absent.logger = log.New(io.Discard, "", 0)
if absent.Held() {
t.Error("a watch with no bucket held the checker")
}
}
@@ -561,34 +561,6 @@ func TestACheckLeftRunningIsTakenOverWhenTheCheckerStarts(t *testing.T) { // N6/
}
}
// TestNoCheckStartsWhileADeliveryBuildsAndOneStartsWhenItEnds: the controller's bucket of the plans in their build phase
// holds a plan's key; the candidate waits, no step is called; the key deleted, the check starts and passes (ADR 0484 clause 12).
func TestNoCheckStartsWhileADeliveryBuildsAndOneStartsWhenItEnds(t *testing.T) { // N8/AC1
b := startBed(t)
plans, err := b.js.CreateKeyValue(b.ctx, jetstream.KeyValueConfig{Bucket: "mesh-controller_plans-building", History: 1})
if err != nil {
t.Fatal(err)
}
if _, err := plans.Put(b.ctx, "plan-1", []byte(`{"plan":"plan-1","since":"2026-10-21T19:00:00Z","seen":1}`)); err != nil {
t.Fatal(err)
}
key := b.candidate(t, "feat/x", pull(repo, 7, head1))
b.start(t, nil)
time.Sleep(1500 * time.Millisecond) // a wait for nothing to happen: there is no event to wait on
if _, err := b.buckets[delivery.CheckerVerdictsBucket].Get(b.ctx, key); err == nil {
t.Fatal("a check started while a delivery builds")
}
if n := b.fake.count("mesh-build-go.build_go_fetch"); n != 0 {
t.Fatalf("%d fetches while a delivery builds", n)
}
if err := plans.Delete(b.ctx, "plan-1"); err != nil {
t.Fatal(err)
}
if v := b.verdict(t, key, ended); v.State != delivery.VerdictPass {
t.Errorf("verdict = %+v", v)
}
}
// TestAMemberNamingAMeshSDKCommitOffTheTrunkFailsNamingTheCommit: the go.mod the forge answers names a commit that main does not
// hold; the verdict fails naming the member and the commit and no test runs (ADR 0484 clause 9).
func TestAMemberNamingAMeshSDKCommitOffTheTrunkFailsNamingTheCommit(t *testing.T) { // 1624/AC2
@@ -37,7 +37,8 @@ func TestTheCheckerAnnouncesEveryStateItKeepsAndOwedReadable(t *testing.T) { //
if !equal(manifest, announced) {
t.Errorf("announced %v, the manifest keeps %v", announced, manifest)
}
if want := []string{"owed", "previews", "verdicts"}; !equal(readable, want) {
// kept is read by the Go build module for a check ask (issue 1887 design limit 1).
if want := []string{"kept", "owed", "previews", "verdicts"}; !equal(readable, want) {
t.Errorf("readable %v, want %v", readable, want)
}
}
@@ -114,7 +114,9 @@ func durationEnv(getenv func(string) string, name string, def time.Duration) (ti
// keptStates is every state the checker keeps and which others may read (ADR 0495 §2-§3): owed by the forge module,
// which writes mesh/check from it (ADR 0521), and by the forge bridge; previews by the indexer; verdicts by the gate.
var keptStates = []announce.State{
{Name: "kept"},
// The kept verdicts are read by the Go build module, which takes them for a check ask's unchanged packages (issue 1887's
// design limit 1); the checker alone writes them (ADR 0444 clause 5).
{Name: "kept", Readable: true},
{Name: "owed", Readable: true},
{Name: "previews", Readable: true},
{Name: "tries"},
@@ -189,10 +191,6 @@ func run(ctx context.Context, getenv func(string) string, logOut io.Writer) erro
if err != nil {
return err
}
capacity, err := capacityFrom(strings.TrimSpace(getenv("MESH_CHECKER_SETTINGS")), getenv("MESH_CHECKER_CAPACITY"))
if err != nil {
return fmt.Errorf("the capacity setting: %w", err)
}
link := strings.TrimSpace(getenv("MESH_CHECKER_LINK"))
if link == "" {
link = defaultLink
@@ -201,7 +199,6 @@ func run(ctx context.Context, getenv func(string) string, logOut io.Writer) erro
presence := &buildrun.PresenceSet{}
have := &buildrun.HaveSet{}
trunks := &check.Trunks{}
_ = capacity
// A check is one ask per pull request and build module on the build seats' check stream, pulled by a build node with a
// free slot (novox/hq ADR 0525 §1, §3): the checker chooses no node and starts no step.
asks := buildrun.JetStreamAsks{JS: js}
@@ -226,9 +223,6 @@ func run(ctx context.Context, getenv func(string) string, logOut io.Writer) erro
}
// The tries (ADR 0501 §5) are judged by the same judge, held by the same deliveries.
tryer := &check.Tryer{Tries: kvStore{buckets[check.TriesBucket]}, Bound: checkBound, Logf: logger.Printf}
holds := newHoldWatch(ctx, js, func(ctx context.Context) { ch.Reconsider(ctx); tryer.Reconsider(ctx) }, logger)
ch.Hold = holds.Held
tryer.Hold = holds.Held
announce := func(ctx context.Context, event, key string) {
body, _ := json.Marshal(map[string]string{"key": key})
if _, err := js.Publish(ctx, eventSubject(event), body); err != nil {
@@ -379,7 +373,5 @@ func run(ctx context.Context, getenv func(string) string, logOut io.Writer) erro
return err
case err := <-watch.errs:
return err
case err := <-holds.errs:
return err
}
}
@@ -1,29 +0,0 @@
package main
import (
"encoding/json"
"fmt"
"os"
"git.novox.be/novox/mesh-sdk/go/delivery/buildrun"
)
// capacityFrom is the build capacity of each node (ADR 0478 clause 3): the module's setting `capacity`, `<node>=<steps at
// once>` pairs separated by commas, laid into the settings file the controller composes (issue 1661); empty is one step at a
// time on every node. With no settings file named, fallback is read instead, for a run outside the module's launch.
func capacityFrom(path, fallback string) (map[string]int, error) {
if path == "" {
return buildrun.ParseCapacity(fallback)
}
raw, err := os.ReadFile(path)
if err != nil {
return nil, fmt.Errorf("the settings file: %w", err)
}
var s struct {
Capacity string `json:"capacity"`
}
if err := json.Unmarshal(raw, &s); err != nil {
return nil, fmt.Errorf("the settings file %s is not JSON: %w", path, err)
}
return buildrun.ParseCapacity(s.Capacity)
}
@@ -1,68 +0,0 @@
package main
import (
"os"
"path/filepath"
"testing"
"git.novox.be/novox/mesh-sdk/go/delivery/buildrun"
)
func settingsFile(t *testing.T, body string) string {
t.Helper()
p := filepath.Join(t.TempDir(), "settings.json")
if err := os.WriteFile(p, []byte(body), 0o600); err != nil {
t.Fatal(err)
}
return p
}
// ADR 0484 clause 12: the build capacity is the module's setting `capacity`, laid into the settings file: empty is one step at a time everywhere.
func TestCapacityIsReadFromTheSettingsFile(t *testing.T) {
for body, want := range map[string]map[string]int{
`{"capacity": "shanks=3,ace=2"}`: {"shanks": 3, "ace": 2},
`{"capacity": ""}`: {},
`{}`: {},
} {
got, err := capacityFrom(settingsFile(t, body), "")
if err != nil || len(got) != len(want) || got["shanks"] != want["shanks"] || got["ace"] != want["ace"] {
t.Errorf("%s: capacity %v, %v, want %v", body, got, err, want)
}
}
p := &buildrun.Places{Capacity: mustCapacity(t, settingsFile(t, `{"capacity": ""}`))}
if p.CapacityOf("shanks") != 1 || p.CapacityOf("ace") != 1 {
t.Errorf("an empty setting gives %d and %d, want 1 everywhere", p.CapacityOf("shanks"), p.CapacityOf("ace"))
}
}
func mustCapacity(t *testing.T, path string) map[string]int {
t.Helper()
c, err := capacityFrom(path, "")
if err != nil {
t.Fatal(err)
}
return c
}
func TestACapacitySettingThatIsNotOneIsRefusedNamingIt(t *testing.T) {
if _, err := capacityFrom(settingsFile(t, `{"capacity": "shanks=-1"}`), ""); err == nil {
t.Error("shanks=-1 accepted")
}
if c, err := capacityFrom(settingsFile(t, `{"capacity": "shanks=3,ace=2,g14=0,novox=0"}`), ""); err != nil || c["g14"] != 0 || len(c) != 4 {
t.Errorf("the starting value: %v, %v", c, err)
}
if _, err := capacityFrom(settingsFile(t, `not json`), ""); err == nil {
t.Error("a file that is not JSON accepted")
}
if _, err := capacityFrom(filepath.Join(t.TempDir(), "none"), ""); err == nil {
t.Error("a settings file that is not there accepted")
}
}
// With no settings file named (a run outside the module's launch, as the tests do), the capacity is the environment's.
func TestWithNoSettingsFileTheCapacityIsTheEnvironments(t *testing.T) {
got, err := capacityFrom("", "ace=2")
if err != nil || got["ace"] != 2 {
t.Errorf("capacity %v, %v", got, err)
}
}
@@ -1,3 +0,0 @@
{
"capacity": ""
}
@@ -4,7 +4,6 @@ import (
"context"
"errors"
"fmt"
"slices"
"strings"
"sync"
"time"
@@ -69,10 +68,6 @@ type Checker struct {
Trunks interface{ Heads() map[string]string }
// Bound is how long one check may run; DefaultBound when 0.
Bound time.Duration
// Limit is how many checks run at once; two when 0. A freed place goes to the first waiting check by rank (ADR 0536 §4).
Limit int
// Hold says whether a delivery builds, so no new check starts (ADR 0484 clause 12); nil never holds.
Hold func() bool
// Link is the base of the address a status points at; the verdict key follows.
Link string
// Announce says a status was written (EventOwedWritten); nil says nothing.
@@ -81,10 +76,6 @@ type Checker struct {
Logf func(format string, args ...any)
once sync.Once
limit int
running int // checks holding a place
waiting []*waiter // checks waiting for a place, in the order they asked
seq uint64
wg sync.WaitGroup
mu sync.Mutex
known map[string]delivery.Candidate // by candidates key
@@ -97,11 +88,6 @@ type Checker struct {
func (ch *Checker) init() {
ch.once.Do(func() {
limit := ch.Limit
if limit <= 0 {
limit = 2
}
ch.limit = limit
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{}
})
@@ -190,10 +176,6 @@ func (ch *Checker) consider(ctx context.Context, key string, c delivery.Candidat
if cs == nil {
return // nothing to judge: hq's pull requests, or no trunk known
}
if ch.Hold != nil && ch.Hold() {
ch.logf("%s: not started, a delivery builds", cs.Key)
return // started when the plan bucket empties (Reconsider)
}
cs.Queued = ch.now()
ch.mu.Lock()
if ch.inflight[cs.Key] {
@@ -207,10 +189,11 @@ func (ch *Checker) consider(ctx context.Context, key string, c delivery.Candidat
go func() {
defer ch.wg.Done()
for {
cs.Kind = ch.kindOf(ctx, cs) // a re-judge after a trunk move is asked on the recheck stream (ADR 0536 §4)
ch.admit(cs, cs.Kind)
// Every check is asked at once: the build seats' streams hold the backlog, a new head's check and a check after a
// failure of the nodes on check, a re-judge after a trunk move on recheck, pulled after them (ADR 0525 §4, §9;
// ADR 0536 §4). The checker holds no place and no capacity of its own.
cs.Kind = ch.kindOf(ctx, cs)
ch.try(ctx, key, cs)
ch.leave()
ch.mu.Lock()
more := ch.again[cs.Key]
delete(ch.again, cs.Key)
@@ -223,62 +206,19 @@ func (ch *Checker) consider(ctx context.Context, key string, c delivery.Candidat
if next := CaseOf(ch.candidate(key, c), ch.heads()); next != nil {
cs = next
}
cs.Queued = ch.now() // asked again: it waits for a place by its rank like any other
cs.Queued = ch.now()
}
}()
}
// The kinds of a check waiting for a place, by the rank they start in (ADR 0536 §4): a head never judged and a head asked
// again after a failure of the nodes first, then a head asked again only because its trunk moved.
// The kinds of a check (ADR 0536 §4): a head never judged and a head asked again after a failure of the nodes are asked on
// the build seats' check stream, a head asked again only because its trunk moved on the recheck stream, pulled after.
const (
KindNew = "new"
KindNodes = "nodes"
KindTrunk = "trunk"
)
var kindRank = map[string]int{KindNew: 0, KindNodes: 0, KindTrunk: 1}
// waiter is a check waiting for a place.
type waiter struct {
group, key, kind string
queued time.Time
seq uint64
ready chan struct{}
}
// Waiting names a check waiting for a place: its group, its verdict key, its kind and when it was asked.
type Waiting struct {
Group string `json:"group"`
Key string `json:"key"`
Kind string `json:"kind"`
Queued time.Time `json:"queued"`
}
// Waiting are the checks waiting for a place, in the order they would start.
func (ch *Checker) Waiting() []Waiting {
ch.init()
ch.mu.Lock()
defer ch.mu.Unlock()
ws := slices.Clone(ch.waiting)
slices.SortStableFunc(ws, before)
out := make([]Waiting, len(ws))
for i, w := range ws {
out[i] = Waiting{Group: w.group, Key: w.key, Kind: w.kind, Queued: w.queued}
}
return out
}
// before orders waiting checks: the lower rank, then the one asked first, then the one that asked for a place first.
func before(a, b *waiter) int {
if ra, rb := kindRank[a.kind], kindRank[b.kind]; ra != rb {
return ra - rb
}
if c := a.queued.Compare(b.queued); c != 0 {
return c
}
return int(a.seq) - int(b.seq)
}
// kindOf is the kind of a case's check, read from its verdict key: none (or one left running) is a head never judged; a
// failure of the nodes is asked again after the nodes changed; a verdict on other trunk heads is a re-judge after a trunk
// move. A verdict that stands is ranked with the re-judges: its claim finds nothing to do.
@@ -297,47 +237,6 @@ func (ch *Checker) kindOf(ctx context.Context, cs *Case) string {
return KindTrunk
}
// admit returns when the check holds a place: at once when one is free and no check waits, else when a check's end hands its
// place to this one as the first by rank. No timer: a check's end wakes the next (leave).
func (ch *Checker) admit(cs *Case, kind string) {
ch.mu.Lock()
if ch.running < ch.limit && len(ch.waiting) == 0 {
ch.running++
ch.mu.Unlock()
return
}
ch.seq++
w := &waiter{group: cs.Group, key: cs.Key, kind: kind, queued: cs.Queued, seq: ch.seq, ready: make(chan struct{})}
ch.waiting = append(ch.waiting, w)
ch.handOn()
ch.mu.Unlock()
<-w.ready
}
// leave frees the check's place and hands it to the first waiting check by rank.
func (ch *Checker) leave() {
ch.mu.Lock()
ch.running--
ch.handOn()
ch.mu.Unlock()
}
// handOn gives each free place to the first waiting check by rank; it is called with mu held.
func (ch *Checker) handOn() {
for ch.running < ch.limit && len(ch.waiting) > 0 {
first := 0
for i := range ch.waiting {
if before(ch.waiting[i], ch.waiting[first]) < 0 {
first = i
}
}
w := ch.waiting[first]
ch.waiting = slices.Delete(ch.waiting, first, first+1)
ch.running++
close(w.ready)
}
}
func (ch *Checker) candidate(key string, fallback delivery.Candidate) delivery.Candidate {
ch.mu.Lock()
defer ch.mu.Unlock()
@@ -298,24 +298,6 @@ func TestATrunkMoveRewritesTheSameVerdictKeyAndTheHeadsStatus(t *testing.T) { //
}
}
func TestNoCheckStartsWhileADeliveryBuildsAndOneStartsWhenItEnds(t *testing.T) { // N8/AC1
ctx, j := context.Background(), passJudge()
r := newRig(t, j)
held := true
r.ch.Hold = func() bool { return held }
r.ch.Candidate(ctx, "feat/x", candidateOf(t, "feat/x", pullOf(repoA, 7, head1)))
r.ch.Wait()
if j.count() != 0 {
t.Fatalf("judged %d times while held", j.count())
}
held = false
r.ch.Reconsider(ctx)
r.ch.Wait()
if j.count() != 1 {
t.Fatalf("judged %d times after the hold ended, want once", j.count())
}
}
func TestTheStatusIsPendingWhileTheCheckRunsAndEndsSuccessOrFailure(t *testing.T) { // 1300/AC5 (unit part)
ctx := context.Background()
var seen []string
@@ -4,6 +4,7 @@ import (
"context"
"strings"
"testing"
"time"
"git.novox.be/novox/mesh-sdk/go/delivery"
)
@@ -71,11 +72,12 @@ func TestATrunkMoveMarksEveryOpenPullRequestIntoItPendingAtOnce(t *testing.T) {
t.Fatalf("%s#%d is %q before the move", p.repo, p.n, s)
}
}
// A delivery builds, so no check starts; mesh-delivery's trunk moves.
r.ch.Hold = func() bool { return true }
// mesh-delivery's trunk moves; the re-judges it asks are held in their judge (in the stream, since R4g) meanwhile.
release := make(chan struct{})
r.ch.Judge = &judgeFunc{f: func(context.Context, *Case, int) (Outcome, error) { <-release; return Outcome{Verdict: passing()}, nil }}
defer func() { close(release); r.ch.Wait() }()
r.setTrunks(t, map[string]string{repoA: trunkA2, repoSDK: trunkS})
r.ch.TrunkMoved(ctx)
r.ch.Wait()
if s := owedState(t, r, repoA, 7, head1); s != "pending" {
t.Errorf("mesh-delivery#7, into the moved trunk, is %q, want pending", s)
}
@@ -90,17 +92,25 @@ func TestATrunkMoveMarksEveryOpenPullRequestIntoItPendingAtOnce(t *testing.T) {
func TestACheckThatEndsAfterItsTrunkMovedOwesPendingNotItsVerdict(t *testing.T) { // gate (a): a pass on an old trunk is never shown
ctx := context.Background()
var r *rig
release, judged := make(chan struct{}), make(chan struct{})
j := &judgeFunc{f: func(ctx context.Context, c *Case, call int) (Outcome, error) {
if call == 1 {
r.setTrunks(t, map[string]string{repoA: trunkA2, repoSDK: trunkS}) // the trunk moves while it runs
r.ch.Hold = func() bool { return true } // and a delivery holds the next check
defer close(judged)
} else {
<-release // the check asked again for the new trunk waits in the stream meanwhile
}
return Outcome{Verdict: passing()}, nil
}}
r = newRig(t, j)
r.ch.Candidate(ctx, "feat%2Fx", candidateOf(t, "feat/x", pullOf(repoA, 7, head1)))
r.ch.Wait()
<-judged
for deadline := time.Now().Add(2 * time.Second); owedState(t, r, repoA, 7, head1) == "" && time.Now().Before(deadline); time.Sleep(time.Millisecond) {
}
time.Sleep(20 * time.Millisecond)
if s := owedState(t, r, repoA, 7, head1); s != "pending" {
t.Errorf("the status is %q, want pending: the pass was judged on a trunk that moved", s)
}
close(release)
r.ch.Wait()
}
@@ -1,7 +0,0 @@
package check
// PlansBuildingBucket is the controller's bucket of the delivery plans in their build phase, one key for each plan, written
// by the controller alone (issue 1652): a plan key there means a delivery builds on the build nodes, and no new check starts
// meanwhile (ADR 0484 clause 12, ADR 0483 clause 6). The keys age out with no delete marker, so the keys are listed, not watched. The bucket may be absent: the controller's issue 1652 is delivered
// before R4c-5b, and if it is not, the bucket is read as empty.
const PlansBuildingBucket = "mesh-controller_plans-building"
@@ -1,38 +0,0 @@
package check
import (
"context"
"testing"
"git.novox.be/novox/mesh-sdk/go/delivery"
)
func TestARunningCheckIsNotStoppedWhenADeliveryStartsToBuild(t *testing.T) { // ADR 0484 clause 12, ADR 0483 clause 4
ctx := context.Background()
held := false
j := &judgeFunc{f: func(_ context.Context, _ *Case, call int) (Outcome, error) {
if call == 1 {
held = true // a delivery starts while this check runs
}
return Outcome{Verdict: passing()}, nil
}}
r := newRig(t, j)
r.ch.Hold = func() bool { return held }
r.ch.Candidate(ctx, "feat/x", candidateOf(t, "feat/x", pullOf(repoA, 7, head1)))
r.ch.Wait()
if v := r.verdict(t, keyOf(t, "feat/x", candidateOf(t, "feat/x", pullOf(repoA, 7, head1)))); v.State != delivery.VerdictPass {
t.Errorf("verdict = %+v, want the running check to end as it was", v)
}
// A new candidate waits while the plan builds.
r.ch.Candidate(ctx, "feat/y", candidateOf(t, "feat/y", pullOf(repoA, 8, head2)))
r.ch.Wait()
if j.count() != 1 {
t.Errorf("judged %d times while a delivery builds, want only the one already running", j.count())
}
held = false
r.ch.Reconsider(ctx)
r.ch.Wait()
if j.count() != 2 {
t.Errorf("judged %d times after the delivery ended, want the waiting candidate too", j.count())
}
}
@@ -3,10 +3,8 @@ package check
import (
"context"
"fmt"
"runtime"
"sync"
"testing"
"time"
"git.novox.be/novox/mesh-sdk/go/delivery"
)
@@ -27,146 +25,8 @@ type rankRig struct {
errs []error // ctx errors the held check saw when released
}
func newRankRig(t *testing.T) *rankRig {
rr := &rankRig{held: make(chan struct{}), release: make(chan struct{})}
j := &judgeFunc{f: func(ctx context.Context, c *Case, _ int) (Outcome, error) {
rr.mu.Lock()
rr.started = append(rr.started, c)
hold := rr.hold
rr.hold = false
rr.mu.Unlock()
if hold {
close(rr.held)
<-rr.release
rr.mu.Lock()
rr.errs = append(rr.errs, ctx.Err())
rr.mu.Unlock()
}
return Outcome{Verdict: passing()}, nil
}}
rr.rig = newRig(t, j)
rr.ch.Limit = 1
now := time.Date(2026, 10, 23, 14, 0, 0, 0, time.UTC)
var clock sync.Mutex
rr.ch.Now = func() time.Time {
clock.Lock()
defer clock.Unlock()
now = now.Add(time.Second)
return now
}
return rr
}
func group(n int) string { return fmt.Sprintf("feat/g%d", n) }
// judged puts n candidates in and judges them on the current trunk.
func (rr *rankRig) judged(t *testing.T, n int) {
for i := 1; i <= n; i++ {
rr.ch.Candidate(context.Background(), group(i), candidateOf(t, group(i), pullOf(repoA, i, head1)))
}
rr.ch.Wait()
rr.mu.Lock()
rr.started = nil
rr.mu.Unlock()
}
// waitFor returns when n checks wait for a place.
func (rr *rankRig) waitFor(t *testing.T, n int) {
t.Helper()
deadline := time.Now().Add(5 * time.Second)
for {
if w := len(rr.ch.Waiting()); w == n {
return
} else if time.Now().After(deadline) {
t.Fatalf("%d checks wait, want %d", w, n)
}
runtime.Gosched()
}
}
// moveTrunkHoldingOne moves the trunk, so every judged candidate is asked again; the first re-judge to start is held.
func (rr *rankRig) moveTrunkHoldingOne(t *testing.T, judged int) {
rr.mu.Lock()
rr.hold = true
rr.mu.Unlock()
rr.setTrunks(t, map[string]string{repoA: trunkA2, repoSDK: trunkS})
rr.ch.TrunkMoved(context.Background())
<-rr.held
rr.waitFor(t, judged-1)
}
func TestANewHeadStartsBeforeTheReJudgesOfATrunkMove(t *testing.T) { // 1897/AC1
rr := newRankRig(t)
rr.judged(t, 4)
rr.moveTrunkHoldingOne(t, 4) // one runs, three re-judges wait
rr.ch.Candidate(context.Background(), group(9), candidateOf(t, group(9), pullOf(repoA, 9, head2)))
rr.waitFor(t, 4)
close(rr.release)
rr.ch.Wait()
if len(rr.started) != 5 {
t.Fatalf("%d checks started, want 5", len(rr.started))
}
if g := rr.started[1].Group; g != group(9) {
t.Errorf("after the running check, %s started, want the new head's %s; order %v", g, group(9), groups(rr.started))
}
}
func TestTheReJudgeQueuedFirstStartsFirst(t *testing.T) { // 1897/AC2
rr := newRankRig(t)
rr.judged(t, 3)
rr.moveTrunkHoldingOne(t, 3) // one runs, two re-judges wait
close(rr.release)
rr.ch.Wait()
if len(rr.started) != 3 {
t.Fatalf("%d checks started, want 3", len(rr.started))
}
if a, b := rr.started[1], rr.started[2]; !a.Queued.Before(b.Queued) {
t.Errorf("%s (queued %s) started before %s (queued %s)", a.Group, a.Queued.Format(time.TimeOnly), b.Group, b.Queued.Format(time.TimeOnly))
}
}
func TestANewHeadDoesNotStopARunningReJudge(t *testing.T) { // 1897/AC3
rr := newRankRig(t)
rr.judged(t, 1)
rr.moveTrunkHoldingOne(t, 1) // the re-judge runs
rr.ch.Candidate(context.Background(), group(9), candidateOf(t, group(9), pullOf(repoA, 9, head2)))
rr.waitFor(t, 1)
close(rr.release)
rr.ch.Wait()
if len(rr.errs) != 1 || rr.errs[0] != nil {
t.Fatalf("the running re-judge saw %v, want it never stopped", rr.errs)
}
if len(rr.started) != 2 || rr.started[0].Group != group(1) || rr.started[1].Group != group(9) {
t.Errorf("started %v, want the re-judge to end before the new head starts", groups(rr.started))
}
key := keyOf(t, group(1), candidateOf(t, group(1), pullOf(repoA, 1, head1)))
if v := rr.verdict(t, key); v.State != delivery.VerdictPass || v.Trunk[repoA] != trunkA2 {
t.Errorf("the re-judge's verdict = %+v, want its pass on the moved trunk", v)
}
}
func TestWaitingChecksAreNamedByKind(t *testing.T) { // 1897/AC4, the checker's own view
rr := newRankRig(t)
rr.judged(t, 2)
rr.moveTrunkHoldingOne(t, 2)
rr.ch.Candidate(context.Background(), group(9), candidateOf(t, group(9), pullOf(repoA, 9, head2)))
rr.waitFor(t, 2)
kinds := map[string]string{}
for _, w := range rr.ch.Waiting() {
kinds[w.Group] = w.Kind
}
close(rr.release)
rr.ch.Wait()
running := rr.started[0].Group
other := group(1)
if running == other {
other = group(2)
}
if kinds[group(9)] != KindNew || kinds[other] != KindTrunk || len(kinds) != 2 {
t.Errorf("waiting = %v, want %s %s and %s %s", kinds, group(9), KindNew, other, KindTrunk)
}
}
func groups(cs []*Case) []string {
out := make([]string, len(cs))
for i, c := range cs {
@@ -8,7 +8,6 @@ import (
"time"
"git.novox.be/novox/mesh-sdk/go/delivery"
"git.novox.be/novox/mesh-sdk/go/delivery/buildrun"
)
// The read tools (novox/hq ADR 0484 clauses 8 and 13, issue 1686): what the checker stored, as stored, with the line of the
@@ -34,12 +33,10 @@ type StateRead struct {
type Reader struct {
VerdictState, KeptState, CandidateState StateRead
PreviewState StateRead // the checker's merge previews
PlansState StateRead // the controller's delivery plans in their build phase; zero where the bucket is absent
StepStates map[string]StateRead // the build modules' steps states, by module
Capacity func() (string, error) // the capacity setting as written, `node=steps,...`
HaveState StateRead // the node-engine's have: whether a build module is on a node and running
StepBound time.Duration // how long the checker waits for one step; DefaultStepBound when 0
PreviewState StateRead // the checker's merge previews
StepStates map[string]StateRead // the build modules' steps states, by module
HaveState StateRead // the node-engine's have: whether a build module is on a node and running
StepBound time.Duration // how long the checker waits for one step; DefaultStepBound when 0
// Trunks are the trunk heads of every repository (the gate's source); nil names no re-judge after a trunk move.
Trunks func() (map[string]string, error)
Now func() time.Time
@@ -266,18 +263,10 @@ type LeftStep struct {
Why string `json:"why"`
}
// Status is what holds the checker back and what runs now, read from the buckets: the checker's held and running steps live in
// its process, so they are shown as what the buckets say (the steps in flight on the build modules, the candidates without a
// verdict for their heads, and the delivery plans in their build phase), not as the process's own counters.
// Status is what runs now, read from the buckets: the steps in flight on the build modules and the candidates without a
// verdict for their heads. Nothing holds a check since R4g: a check is an ask on the build seats' streams, whose backlog
// the delivery view shows (ADR 0525 §9).
type Status struct {
Capacity map[string]int `json:"capacity"`
CapacityNote string `json:"capacity_note,omitempty"`
DefaultPerNode int `json:"default_per_node"`
PlansBuilding struct {
Readable bool `json:"readable"`
Keys []string `json:"keys"`
Holds bool `json:"holds"`
} `json:"plans_building"`
Running []RunningStep `json:"running_steps"`
LeftRunning []LeftStep `json:"left_running"`
RunningByNode map[string]int `json:"running_by_node"`
@@ -288,24 +277,9 @@ type Status struct {
Notes []string `json:"notes,omitempty"`
}
// Status reads the capacity, the plan bucket, the build modules' steps and the candidates against their verdicts.
// Status reads the build modules' steps and the candidates against their verdicts.
func (r Reader) Status() (Status, error) {
st := Status{DefaultPerNode: buildrun.DefaultCapacity, Capacity: map[string]int{}, RunningByNode: map[string]int{}}
if r.Capacity == nil {
st.CapacityNote = "the capacity setting cannot be read from here"
} else if setting, err := r.Capacity(); err != nil {
st.CapacityNote = "the capacity setting cannot be read: " + err.Error()
} else if st.Capacity, err = buildrun.ParseCapacity(setting); err != nil {
st.Capacity, st.CapacityNote = map[string]int{}, "the capacity setting is not valid: "+err.Error()
}
if r.PlansState.Keys == nil {
st.Notes = append(st.Notes, PlansBuildingBucket+" is not readable here (absent), so it is read as empty: no delivery holds the checker")
} else if keys, err := r.PlansState.Keys(); err != nil {
st.Notes = append(st.Notes, PlansBuildingBucket+" cannot be listed, so it is read as empty: "+err.Error())
} else {
sort.Strings(keys)
st.PlansBuilding.Readable, st.PlansBuilding.Keys, st.PlansBuilding.Holds = true, keys, len(keys) > 0
}
st := Status{RunningByNode: map[string]int{}}
mods := make([]string, 0, len(r.StepStates))
for m := range r.StepStates {
mods = append(mods, m)
@@ -3,7 +3,6 @@ package check
import (
"context"
"encoding/json"
"os"
"strings"
"testing"
"time"
@@ -129,16 +128,14 @@ func TestThePreviewsToolShowsTheStoredPreviewOfAPullRequest(t *testing.T) {
}
}
// 1685 live acceptance: checker_status reads the capacity, whether a plan key holds the checker, the steps in flight by node and
// the candidates without a verdict.
func TestTheStatusToolReadsCapacityHoldsAndRunningSteps(t *testing.T) {
// checker_status reads the steps in flight by node and the candidates without a verdict; since R4g nothing holds the checker
// and it has no capacity of its own (ADR 0525 §4, §7; 1685 closed by the hold's deletion).
func TestTheStatusToolReadsRunningSteps(t *testing.T) {
running, _ := delivery.EncodeStep(delivery.Step{ID: "abc1", Tool: "build_go_test", Node: "shanks", State: delivery.StepRunning, Started: time.Date(2026, 10, 22, 15, 0, 0, 0, time.UTC)})
passed, _ := delivery.EncodeStep(delivery.Step{ID: "abc2", Tool: "build_go_test", Node: "ace", State: delivery.StepPassed})
cand := candidateOf(t, "feat/x", pullOf(repoA, 7, head1))
ckey, _ := delivery.CandidatesKey("feat/x")
rd := Reader{
Capacity: func() (string, error) { return "shanks=2,ace=0", nil },
PlansState: fakeState{"p1": []byte("{}")}.read(),
StepStates: map[string]StateRead{"mesh-build-go": fakeState{"shanks.abc1": running, "ace.abc2": passed}.read()},
CandidateState: fakeState{ckey: cand}.read(),
VerdictState: fakeState{}.read(),
@@ -149,21 +146,15 @@ func TestTheStatusToolReadsCapacityHoldsAndRunningSteps(t *testing.T) {
if err != nil {
t.Fatal(err)
}
if st.Capacity["shanks"] != 2 || st.Capacity["ace"] != 0 || st.DefaultPerNode != 1 {
t.Errorf("capacity = %v default %d", st.Capacity, st.DefaultPerNode)
}
if !st.PlansBuilding.Holds || len(st.PlansBuilding.Keys) != 1 {
t.Errorf("plans = %+v", st.PlansBuilding)
}
if len(st.Running) != 1 || st.Running[0].Node != "shanks" || st.RunningByNode["shanks"] != 1 || st.RunningByNode["ace"] != 0 {
t.Errorf("running = %+v %v", st.Running, st.RunningByNode)
}
if len(st.Unjudged) != 1 || st.Unjudged[0] != "feat/x" {
t.Errorf("unjudged = %v", st.Unjudged)
}
// An absent plan bucket is read as empty and said; an unreadable capacity is said, not guessed.
st, err = Reader{Capacity: func() (string, error) { return "", os.ErrNotExist }}.Status()
if err != nil || st.PlansBuilding.Holds || st.CapacityNote == "" || len(st.Notes) == 0 {
// What cannot be read is said, not guessed.
st, err = Reader{}.Status()
if err != nil || len(st.Notes) == 0 {
t.Errorf("empty status = %+v, %v", st, err)
}
}
@@ -206,13 +206,11 @@ type TryOptions struct {
Run string
}
// Tryer runs the tries: it takes a queued try, judges it through the same Judge as a pull request's check, and writes its
// end. It holds no timer: a try starts when its key is written, or when a delivery that held it ends (Reconsider).
// Tryer runs the tries: it takes a queued try, judges it through the same Judge as a pull request's check (a check ask, ADR
// 0525 §1), and writes its end. It holds no timer: a try starts when its key is written.
type Tryer struct {
Tries KV
Judge Judge
// Hold says whether a delivery builds, so no try starts (ADR 0484 clause 12); nil never holds.
Hold func() bool
Bound time.Duration
Now func() time.Time
Logf func(format string, args ...any)
@@ -250,7 +248,6 @@ func (tr *Tryer) Put(ctx context.Context, key string, value []byte) {
if t.Over() {
return
}
held := tr.Hold != nil && tr.Hold()
tr.mu.Lock()
if tr.inflight == nil {
tr.inflight, tr.seen = map[string]bool{}, map[string]bool{}
@@ -259,11 +256,6 @@ func (tr *Tryer) Put(ctx context.Context, key string, value []byte) {
tr.mu.Unlock()
return
}
if held {
tr.mu.Unlock() // not seen yet: a try left running is still taken over when the hold ends
tr.logf("%s: not started, a delivery builds", key)
return
}
tr.seen[key] = true
tr.inflight[key] = true
tr.mu.Unlock()
@@ -156,7 +156,7 @@ func TestAFailingTryNamesWhatFailedAndAStepThatCannotRunIsAnError(t *testing.T)
// No build node is no error since R4g: the try's ask waits in the check stream until a node pulls it (ADR 0525 §4).
}
func TestATryWaitsWhileADeliveryBuildsAndATryLeftRunningIsTakenOver(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())
@@ -169,14 +169,7 @@ func TestATryWaitsWhileADeliveryBuildsAndATryLeftRunningIsTakenOver(t *testing.T
t.Fatal(err)
}
b := newBench(t)
held := true
tr := &Tryer{Tries: kv, Judge: b.ex, Hold: func() bool { return held }}
tr.Reconsider(ctx)
tr.Wait()
if n := len(b.steps.tool("fetch")); n != 0 {
t.Fatalf("a try ran while a delivery builds")
}
held = false
tr := &Tryer{Tries: kv, Judge: b.ex}
tr.Reconsider(ctx)
tr.Wait()
if got := readTry(t, kv, id); got.State != TryPass {
+15 -25
View File
@@ -46,6 +46,7 @@ type manifest struct {
Consumes []string `json:"consumes"`
State []state `json:"state"`
Reads []string `json:"reads"`
Uses []string `json:"uses"`
Invokes []string `json:"invokes"`
Config map[string]struct {
Type string `json:"type"`
@@ -96,12 +97,17 @@ func TestManifestDeclarations(t *testing.T) { // 1678/AC1
t.Errorf("state = %+v, want %+v", m.State, want)
}
if want := []string{"mesh-build-go.steps", "mesh-build-go.toolchains", "mesh-build-node.steps", "mesh-build-node.toolchains",
"mesh-controller.plans-building", "mesh-delivery-builder.builds", "mesh-delivery-gate.source", "mesh-delivery-indexer.candidates",
"mesh-delivery-builder.builds", "mesh-delivery-gate.source", "mesh-delivery-indexer.candidates",
"mesh-delivery-indexer.manifests", "mesh-delivery-indexer.modules", "mesh-delivery-watchman.presence", "node-engine.have"}; !reflect.DeepEqual(sorted(m.Reads), want) {
t.Errorf("reads = %v, want %v", m.Reads, want)
}
if want := []string{"gitea.gitea_commits", "gitea.gitea_compare", "gitea.gitea_get_file", "mesh-build-go.build_go_digest", "mesh-build-go.build_go_fetch", "mesh-build-go.build_go_test", "mesh-build-node.build_node_fetch", "mesh-build-node.build_node_test"}; !reflect.DeepEqual(sorted(m.Invokes), want) {
t.Errorf("invokes = %v, want only the build modules' fetch, digest and test (no run step before issue 1309, ADR 0484 clause 5) and the forge's two reads for a mesh-sdk pin (clause 9)", m.Invokes)
// 1887/A5b: the checker reads nothing of the controller's plans-building (ADR 0483 superseded by ADR 0525 §7), calls no
// build module's tool and holds no capacity: it puts check asks on the Go and Node build seats (ADR 0525 §1, §4).
if want := []string{"gitea.gitea_commits", "gitea.gitea_compare", "gitea.gitea_get_file"}; !reflect.DeepEqual(sorted(m.Invokes), want) {
t.Errorf("invokes = %v, want only the forge's reads for a mesh-sdk pin (ADR 0484 clause 9)", m.Invokes)
}
if want := []string{"node-build-go", "node-build-node"}; !reflect.DeepEqual(sorted(m.Uses), want) {
t.Errorf("uses = %v, want the Go and Node build seats", m.Uses)
}
// The checker is woken by a build module's report of a toolchain, as the indexer is: the bucket may not exist when it starts.
if want := []string{"mesh-build-go.toolchain.reported", "mesh-build-node.toolchain.reported"}; !reflect.DeepEqual(sorted(m.Consumes), want) {
@@ -122,7 +128,7 @@ func TestManifestDeclarations(t *testing.T) { // 1678/AC1
if want := []string{"checker_kept", "checker_previews", "checker_pull", "checker_status", "checker_try", "checker_try_read", "checker_verdicts"}; !reflect.DeepEqual(sorted(m.Tools), want) {
t.Errorf("tools = %v, want %v", m.Tools, want)
}
for _, k := range []string{"logs", "emits", "claims", "provides", "requires", "uses", "defines-seats", "seat", "seats", "bus"} {
for _, k := range []string{"logs", "emits", "claims", "provides", "requires", "defines-seats", "seat", "seats", "bus", "config"} {
if _, ok := top[k]; ok {
t.Errorf("the manifest declares %q: the checker has none", k)
}
@@ -148,35 +154,19 @@ func TestManifestProcess(t *testing.T) { // 1678/AC1
t.Errorf("process = %v", p)
}
env, _ := p["env"].(map[string]any)
// The capacity of each node is the module's setting `capacity` (ADR 0484 clause 12, as the builder's, issue 1661), laid into the
// settings file the program reads; empty is one step at a time everywhere. The environment holds no capacity.
if len(env) != 2 || env["MESH_BROKER_FILE"] != "${dir:state}/broker" || env["MESH_CHECKER_SETTINGS"] != "${dir:state}/settings.json" {
t.Errorf("env = %v", p["env"])
if len(env) != 1 || env["MESH_BROKER_FILE"] != "${dir:state}/broker" {
t.Errorf("env = %v, want the broker file alone", p["env"])
}
if !reflect.DeepEqual(p["restart-on"], []any{"settings"}) {
t.Errorf("restart-on = %v, want the settings file, so a changed capacity reaches the running program", p["restart-on"])
}
c, ok := m.Config["capacity"]
if !ok || len(m.Config) != 1 || c.Type != "string" || c.Default != "" || c.Description == "" {
t.Errorf("config = %+v, want one string setting, capacity, empty (one step at a time on every node) to start", m.Config)
}
var settings map[string]any
for _, r := range m.Resources {
if r["id"] == "settings" {
settings = r
t.Errorf("a settings resource %v: the checker has no capacity of its own (ADR 0525 §4)", r)
}
}
if settings == nil || settings["type"] != "file" || settings["merge"] != "json" || !reflect.DeepEqual(settings["takes-config"], []any{"capacity"}) || settings["path"] != "${dir:state}/settings.json" {
t.Errorf("settings resource = %v", settings)
}
if raw, err := os.ReadFile("files/settings.json"); err != nil || !strings.Contains(string(raw), `"capacity": ""`) {
t.Errorf("files/settings.json = %q, %v, want the setting's starting value", raw, err)
}
if _, bus := p["bus"]; bus {
t.Errorf("the process declares a bus: it runs on its own broker credential")
}
if len(m.Build.Artifacts) != 3 || m.Build.Artifacts[2].Name != "files" || m.Build.Artifacts[2].Kind != "archive" || m.Build.Artifacts[2].From != "files" {
t.Fatalf("artifacts = %+v, want the program, the tools and the files", m.Build.Artifacts)
if len(m.Build.Artifacts) != 2 {
t.Fatalf("artifacts = %+v, want the program and the tools", m.Build.Artifacts)
}
if a := m.Build.Artifacts[1]; a.Name != "tools" || a.Kind != "bundle" || a.Language != "go" || a.Binary != "mesh-delivery-checker-tools" ||
!reflect.DeepEqual(a.Loads, []string{"mesh-delivery-checker-tools"}) {
+7 -36
View File
@@ -21,7 +21,6 @@
"mesh-build-go.toolchains",
"mesh-build-node.steps",
"mesh-build-node.toolchains",
"mesh-controller.plans-building",
"mesh-delivery-builder.builds",
"mesh-delivery-gate.source",
"mesh-delivery-indexer.candidates",
@@ -30,6 +29,10 @@
"mesh-delivery-watchman.presence",
"node-engine.have"
],
"uses": [
"node-build-go",
"node-build-node"
],
"tools": [
"checker_kept",
"checker_previews",
@@ -42,21 +45,9 @@
"invokes": [
"gitea.gitea_commits",
"gitea.gitea_compare",
"gitea.gitea_get_file",
"mesh-build-go.build_go_digest",
"mesh-build-go.build_go_fetch",
"mesh-build-go.build_go_test",
"mesh-build-node.build_node_fetch",
"mesh-build-node.build_node_test"
"gitea.gitea_get_file"
],
"config": {
"capacity": {
"type": "string",
"description": "How many build steps the checker runs at once on each node, as node=steps pairs separated by commas, such as shanks=2,ace=1 (ADR 0484 clause 12); a node set to 0 never checks, a node not named runs one at a time, and empty is one everywhere. The builder's capacity is the builder's own.",
"pattern": "|[a-z0-9_-]+=[0-9]+(,[a-z0-9_-]+=[0-9]+)*",
"default": ""
}
},
"own-secrets": {
"broker": "${dir:state}/broker"
},
@@ -77,17 +68,6 @@
"place": ".",
"unknown": "program"
},
{
"id": "settings",
"trusted": false,
"type": "file",
"path": "${dir:state}/settings.json",
"mode": "0600",
"owner": "mesh-delivery-checker",
"takes-config": ["capacity"],
"merge": "json",
"template": "files/settings.json"
},
{
"id": "checker",
"type": "process",
@@ -97,12 +77,8 @@
"./mesh-delivery-checker"
],
"env": {
"MESH_BROKER_FILE": "${dir:state}/broker",
"MESH_CHECKER_SETTINGS": "${dir:state}/settings.json"
"MESH_BROKER_FILE": "${dir:state}/broker"
},
"restart-on": [
"settings"
],
"user": "mesh-delivery-checker",
"health": {
"kind": "unit"
@@ -129,11 +105,6 @@
"loads": [
"mesh-delivery-checker-tools"
]
},
{
"name": "files",
"kind": "archive",
"from": "files"
}
]
}