The view says a lagging node's build by fingerprint, what moved, and a failed newest build (issue 1966)

The hold named the same commit twice when only a mesh package moved. It now names both builds
by fingerprint with their commits, the input that moved when the commit is the same, and says a
newest build that failed as failed with its first words.

Claude-Session: https://claude.ai/code/session_01JG73KEq4zoDfM9WtYpcQgZ
This commit is contained in:
2026-11-09 22:41:22 +01:00
parent 29e491e715
commit 47026d5ebb
3 changed files with 163 additions and 2 deletions
@@ -10,6 +10,7 @@ import (
"git.novox.be/novox/mesh-delivery/contracts/go/clients/meshcatalogclient"
"git.novox.be/novox/mesh-delivery/contracts/go/clients/meshdeliveryclient"
"git.novox.be/novox/mesh-delivery/contracts/go/seats/meshcatalog"
"git.novox.be/novox/mesh-delivery/modules/mesh-delivery-builder/builds"
"git.novox.be/novox/mesh-delivery/modules/mesh-delivery-view/viewstate"
"git.novox.be/novox/mesh-delivery/sdk/trunk"
)
@@ -195,6 +196,7 @@ func (m *Model) moduleReach(name string, mod meshcatalog.ModulesByModuleValue, h
if n := len(hist); n == 0 || hist[n-1].Fp2 != mod.Fp2 || hist[n-1].Commit != mod.Commit {
hist = append(slices.Clone(hist), mod) // the value watched now is the newest, whatever the history watch has given
}
var values []meshcatalog.ModulesByModuleValue // the catalogue's value of each change, in step with mr.Changes
for i := len(hist) - 1; i >= 0 && len(mr.Changes) < viewstate.MostChanges; i-- {
c := viewstate.Change{Commit: hist[i].Commit, FP: hist[i].Fp2, Src: uint64(hist[i].Src)}
if c.Src == 0 {
@@ -202,9 +204,11 @@ func (m *Model) moduleReach(name string, mod meshcatalog.ModulesByModuleValue, h
}
if n := len(mr.Changes); n > 0 && mr.Changes[n-1].FP == c.FP {
mr.Changes[n-1] = c // the same fingerprint again, back to back: its first appearance is the older commit
values[n-1] = hist[i]
continue
}
mr.Changes = append(mr.Changes, c)
values = append(values, hist[i])
}
var waiting []string
away := m.awayNodes()
@@ -264,7 +268,7 @@ func (m *Model) moduleReach(name string, mod meshcatalog.ModulesByModuleValue, h
return mr, noBound, ""
}
next := mr.Changes[floor-1]
words := fmt.Sprintf("%s on %s runs the build of %.12s, not yet that of %.12s", name, floorNode, mr.Floor, next.Commit)
words := m.laggingWords(name, floorNode, mr.Changes[floor], next, values[floor], values[floor-1])
if next.Src == 0 {
// The next change's order is not known: the head is held at the floor's own revision, or before every head when that is
// not known either.
@@ -276,6 +280,86 @@ func (m *Model) moduleReach(name string, mod meshcatalog.ModulesByModuleValue, h
return mr, next.Src, words
}
// laggingWords says what a node's running build lacks of the module's next change: both builds by fingerprint with their
// commits, which input moved when the commit is the same (the catalogue holds the module's inputs, not the version a mesh
// package resolved to), and a next build that failed, with the first words of its error, instead of "not yet".
func (m *Model) laggingWords(name, node string, run, next viewstate.Change, runV, nextV meshcatalog.ModulesByModuleValue) string {
words := fmt.Sprintf("%s on %s runs build %s (%.8s)", name, node, shortFP(run.FP), run.Commit)
nextIs := fmt.Sprintf("%s (%.8s)", shortFP(next.FP), next.Commit)
if run.Commit == next.Commit {
if moved := inputsMoved(runV, nextV); moved != "" {
nextIs = fmt.Sprintf("%s (%.8s, %s)", shortFP(next.FP), next.Commit, moved)
}
}
if key, err := builds.BuilderBuildKey(name, next.FP); err == nil {
if k, ok := m.keys[key]; ok && k.err == "" && k.value.State == builds.BuildFailed {
return fmt.Sprintf("%s, its newest build %s failed: %s", words, nextIs, firstWords(k.value.Error))
}
}
return fmt.Sprintf("%s, not yet %s", words, nextIs)
}
// inputsMoved names what of a module's inputs differs between two values with the same commit, or "".
func inputsMoved(a, b meshcatalog.ModulesByModuleValue) string {
if a.Inputs2 == nil || b.Inputs2 == nil {
return ""
}
x, y := a.Inputs2, b.Inputs2
var moved []string
if x.Tree != y.Tree {
moved = append(moved, "tree")
}
for _, name := range slices.Sorted(maps.Keys(mergedKeys(x.Locks, y.Locks))) {
if x.Locks[name] != y.Locks[name] {
moved = append(moved, "lock "+name)
}
}
if !slices.Equal(x.Bases, y.Bases) {
moved = append(moved, "a base module")
}
if len(moved) == 0 && x.Builder != y.Builder {
moved = append(moved, "builder")
}
if len(moved) == 0 && len(y.Packages) > 0 {
// Nothing of the module's own moved: a mesh package it resolves at its latest published version did.
moved = append(moved, "a mesh package of "+strings.Join(y.Packages, ", "))
}
return strings.Join(moved, ", ")
}
func mergedKeys(a, b map[string]string) map[string]bool {
out := map[string]bool{}
for k := range a {
out[k] = true
}
for k := range b {
out[k] = true
}
return out
}
// firstWords is the start of an error, up to eighty characters cut at a word.
func firstWords(s string) string {
s = strings.Join(strings.Fields(s), " ")
if len(s) <= 80 {
return s
}
s = s[:80]
if i := strings.LastIndexByte(s, ' '); i > 0 {
s = s[:i]
}
return s + "..."
}
// shortFP is a fingerprint's first eight characters, after its form.
func shortFP(fp string) string {
fp = strings.TrimPrefix(fp, meshcatalogclient.SourcePrefix2)
if len(fp) > 8 {
fp = fp[:8]
}
return fp
}
// runs is whether a have in this phase runs the fingerprint it names: running, being checked after its apply, or held after a
// failed backup, which keeps running what it has (ADR 0539 clause 16). A have backing up or applying, failed or removed runs
// nothing the derivation can count.
@@ -82,7 +82,7 @@ func TestTheViewDerivesHowFarEachRepositoryReached(t *testing.T) {
then: []Entry{haveFP(t, "g14", "web", fpOld, nodeengineclient.HealthOK), presenceE(t, "g14", false, t0)}},
{name: "a node returned lagging holds the head again", reached: commitB, holds: "web on g14", waiting: "not yet the",
then: []Entry{haveFP(t, "g14", "web", fpOld, nodeengineclient.HealthOK), presenceE(t, "g14", false, t0), presenceE(t, "g14", true, t0)}},
{name: "a want held back at an older build holds the head as a lag does", reached: commitB, holds: "web on g14 runs the build of",
{name: "a want held back at an older build holds the head as a lag does", reached: commitB, holds: "web on g14 runs build",
then: []Entry{wantE(t, "g14", "web", meshwantsclient.Want{Fp: fpOld, Newest: fpNew, Rule: 2}), haveFP(t, "g14", "web", fpOld, nodeengineclient.HealthOK)}},
{name: "a failing health check is said and holds nothing", reached: commitC, waiting: "health check failed",
then: []Entry{haveFP(t, "g14", "web", fpNew, nodeengineclient.HealthFailed)}},
@@ -0,0 +1,77 @@
package view
import (
"strings"
"testing"
"git.novox.be/novox/mesh-contracts/go/clients/nodeengineclient"
"git.novox.be/novox/mesh-delivery/contracts/go/clients/meshcatalogclient"
"git.novox.be/novox/mesh-delivery/contracts/go/clients/meshdeliveryclient"
"git.novox.be/novox/mesh-delivery/contracts/go/clients/meshwantsclient"
"git.novox.be/novox/mesh-delivery/contracts/go/seats/meshcatalog"
"git.novox.be/novox/mesh-delivery/contracts/go/seats/meshdelivery"
"git.novox.be/novox/mesh-delivery/modules/mesh-delivery-builder/builds"
"git.novox.be/novox/mesh-delivery/modules/mesh-delivery-indexer/takein"
"git.novox.be/novox/mesh-delivery/sdk/trunk"
)
// Issue 1966: a node that runs an older build of a module than its newest names both builds by fingerprint with their commits,
// says which input moved when the commit is the same, and says a newest build that failed as failed, with its first words.
func TestALaggingNodeNamesTheBuildsAndWhatDiffers(t *testing.T) {
const web = "novox/web"
commitA, commitB, commitC := strings.Repeat("a", 40), strings.Repeat("b", 40), strings.Repeat("c", 40)
inputs := func(tree string, pkgs ...string) *meshcatalog.ModulesByModuleValueInputs2 {
return &meshcatalog.ModulesByModuleValueInputs2{Tree: tree, Packages: pkgs}
}
for _, tc := range []struct {
name string
oldIn, newIn *meshcatalog.ModulesByModuleValueInputs2
newCommit string
failed string
want, not string
}{
{name: "same commit, a mesh package moved", oldIn: inputs("t", "mesh-sdk/go"), newIn: inputs("t", "mesh-sdk/go"), newCommit: commitA,
want: "web on g14 runs build 11111111 (aaaaaaaa), not yet 22222222 (aaaaaaaa, a mesh package of mesh-sdk/go)", not: "the build of"},
{name: "same commit, its tree moved", oldIn: inputs("t1"), newIn: inputs("t2"), newCommit: commitA,
want: "runs build 11111111 (aaaaaaaa), not yet 22222222 (aaaaaaaa, tree)"},
{name: "another commit names no input", oldIn: inputs("t1"), newIn: inputs("t2"), newCommit: commitB,
want: "runs build 11111111 (aaaaaaaa), not yet 22222222 (bbbbbbbb)", not: "tree"},
{name: "a newest build that failed is said failed", oldIn: inputs("t1"), newIn: inputs("t2"), newCommit: commitB, failed: "go build ./...: undefined: Foo",
want: "its newest build 22222222 (bbbbbbbb) failed: go build ./...: undefined: Foo", not: "not yet"},
} {
t.Run(tc.name, func(t *testing.T) {
mod := func(fp, commit string, src int64, in *meshcatalog.ModulesByModuleValueInputs2, history bool) Entry {
v, err := meshcatalogclient.EncodeModule(meshcatalog.ModulesByModuleValue{Fp2: fp, Manifest: strings.Repeat("a", 64), Repository: web, Commit: commit, Src: src, Path: "web", Inputs2: in})
if err != nil {
t.Fatal(err)
}
return Entry{Bucket: ModulesBucket, Key: "web", Value: v, At: t0, History: history}
}
src := func(rev uint64, head string, history bool) Entry {
v, _ := meshdeliveryclient.EncodeSource(meshdelivery.SourceByMainValue{Group: "g", Heads: map[string]string{web: head}})
return Entry{Bucket: SourceBucket, Key: meshdeliveryclient.SourceMainKey, Value: v, At: t0, Revision: rev, History: history}
}
key, _ := trunk.RepositoryKey(web)
v, _ := takein.EncodeTaken(takein.Taken{Repository: web, Merge: commitC, At: t0})
m := New()
m.SetReachWriter(func(ReachWrite) error { return nil })
m.SetDeliveredEvent(func(DeliveredEvent) error { return nil })
entries := []Entry{src(10, commitA, true), src(11, tc.newCommit, true), src(12, commitC, true), src(12, commitC, false),
mod(fpOld, commitA, 10, tc.oldIn, true), mod(fpNew, tc.newCommit, 11, tc.newIn, true), mod(fpNew, tc.newCommit, 11, tc.newIn, false),
{Bucket: TakenBucket, Key: key, Value: v, At: t0},
wantE(t, "g14", "web", meshwantsclient.Want{Fp: fpOld}), haveFP(t, "g14", "web", fpOld, nodeengineclient.HealthOK)}
if tc.failed != "" {
entries = append(entries, builderKey(t, "web", fpNew, builds.BuilderBuild{State: builds.BuildFailed, Error: tc.failed}))
}
fill(m, entries...)
for _, b := range reachBuckets {
m.Initial(b)
}
fill(m, haveFP(t, "g14", "web", fpOld, nodeengineclient.HealthOK))
got := strings.Join(m.Reach()[key].Holds, "; ")
if !strings.Contains(got, tc.want) || (tc.not != "" && strings.Contains(got, tc.not)) {
t.Errorf("holds = %q, want words with %q and none with %q", got, tc.want, tc.not)
}
})
}
}