detached eed2010

fix(harness): scope observer gap span suppression

commit
eed2010c969651de211378b9cafbe4addd8e76ab
author
brice · · 2026-08-02 10:57 UTC
parent
7b0921e
signature
GPG VerifiedCD88F9E564E0102646FB6077A1A07C7C7C3E349D · arch-gpg Verified for the committer email
2 files · +59 -23
Unified
UnifiedSplit
M cmd/vev-perf-harness/evidence.go +8 -5
@@ -200,21 +200,23 @@ func mergeProcessTraces(mappings []processMapping) ([]span, error) {
200200 // and exact correlation domain may pair, and each pair's ticks must retain
201201 // its own start-before-end ordering.
202202 starts := map[string]traceRecord{}
203- observerGap := false
203+ discarded := map[string]struct{}{}
204204 for _, r := range records {
205205 if r.Component == "observer" && r.Kind == "transport_diagnostic" && !r.Valid {
206206 // A bounded producer queue reports dropped marks explicitly. Any
207207 // in-flight span may have lost either endpoint, so discard those
208- // partial samples; an end after the gap is not independently
209- // attributable to a missing start.
208+ // partial samples; only their exact keys may be missing an end.
209+ for key := range starts {
210+ discarded[key] = struct{}{}
211+ }
210212 starts = map[string]traceRecord{}
211- observerGap = true
212213 continue
213214 }
214215 for _, pair := range spanPairs {
215216 key := fmt.Sprintf("%s/%s/%s/%d/%d/%d/%d", r.ProcessID, r.Component, m.Scenario, m.Run, r.Sequence, r.RequestID, r.Epoch)
216217 if r.Kind == pair.start {
217218 k := pair.name + "/" + key
219+ delete(discarded, k)
218220 if _, exists := starts[k]; exists {
219221 return nil, errors.New("duplicate process-local span start")
220222 }
@@ -224,7 +226,8 @@ func mergeProcessTraces(mappings []processMapping) ([]span, error) {
224226 k := pair.name + "/" + key
225227 start, ok := starts[k]
226228 if !ok {
227- if observerGap {
229+ if _, ok := discarded[k]; ok {
230+ delete(discarded, k)
228231 continue
229232 }
230233 return nil, errors.New("span end without same-process start")
@@ -200,21 +200,23 @@ func mergeProcessTraces(mappings []processMapping) ([]span, error) {
200 // and exact correlation domain may pair, and each pair's ticks must retain200 // and exact correlation domain may pair, and each pair's ticks must retain
201 // its own start-before-end ordering.201 // its own start-before-end ordering.
202 starts := map[string]traceRecord{}202 starts := map[string]traceRecord{}
203- observerGap := false203+ discarded := map[string]struct{}{}
204 for _, r := range records {204 for _, r := range records {
205 if r.Component == "observer" && r.Kind == "transport_diagnostic" && !r.Valid {205 if r.Component == "observer" && r.Kind == "transport_diagnostic" && !r.Valid {
206 // A bounded producer queue reports dropped marks explicitly. Any206 // A bounded producer queue reports dropped marks explicitly. Any
207 // in-flight span may have lost either endpoint, so discard those207 // in-flight span may have lost either endpoint, so discard those
208- // partial samples; an end after the gap is not independently208+ // partial samples; only their exact keys may be missing an end.
209- // attributable to a missing start.209+ for key := range starts {
210+ discarded[key] = struct{}{}
211+ }
210 starts = map[string]traceRecord{}212 starts = map[string]traceRecord{}
211- observerGap = true
212 continue213 continue
213 }214 }
214 for _, pair := range spanPairs {215 for _, pair := range spanPairs {
215 key := fmt.Sprintf("%s/%s/%s/%d/%d/%d/%d", r.ProcessID, r.Component, m.Scenario, m.Run, r.Sequence, r.RequestID, r.Epoch)216 key := fmt.Sprintf("%s/%s/%s/%d/%d/%d/%d", r.ProcessID, r.Component, m.Scenario, m.Run, r.Sequence, r.RequestID, r.Epoch)
216 if r.Kind == pair.start {217 if r.Kind == pair.start {
217 k := pair.name + "/" + key218 k := pair.name + "/" + key
219+ delete(discarded, k)
218 if _, exists := starts[k]; exists {220 if _, exists := starts[k]; exists {
219 return nil, errors.New("duplicate process-local span start")221 return nil, errors.New("duplicate process-local span start")
220 }222 }
@@ -224,7 +226,8 @@ func mergeProcessTraces(mappings []processMapping) ([]span, error) {
224 k := pair.name + "/" + key226 k := pair.name + "/" + key
225 start, ok := starts[k]227 start, ok := starts[k]
226 if !ok {228 if !ok {
227- if observerGap {229+ if _, ok := discarded[k]; ok {
230+ delete(discarded, k)
228 continue231 continue
229 }232 }
230 return nil, errors.New("span end without same-process start")233 return nil, errors.New("span end without same-process start")
M cmd/vev-perf-harness/evidence_test.go +51 -18
@@ -364,25 +364,58 @@ func TestHarnessRejectsCrossProcessAndBadTraceSpans(t *testing.T) {
364364 }
365365 }
366366
367-func TestHarnessSkipsUnpairedEndAfterObserverGap(t *testing.T) {
368- path := filepath.Join(t.TempDir(), "trace.jsonl")
369- records := []traceRecord{
370- {ProcessID: "one", Component: "observer", Scenario: "s", Run: 1, Sequence: 1, RequestID: 1, Epoch: 1, Kind: "transport_diagnostic", Valid: false},
371- {ProcessID: "one", Component: "ipc", Scenario: "s", Run: 1, Sequence: 2, RequestID: 2, Epoch: 2, Kind: "adapter_receive_end", Tick: 20, Valid: true},
367+func TestHarnessObserverGapSuppressesOnlyDiscardedSpanKeys(t *testing.T) {
368+ base := func(kind string, sequence, requestID, epoch uint64, tick int64) traceRecord {
369+ return traceRecord{ProcessID: "one", Component: "ipc", Scenario: "s", Run: 1, Sequence: sequence, RequestID: requestID, Epoch: epoch, Kind: kind, Tick: tick, Valid: true}
372370 }
373- var lines []string
374- for _, record := range records {
375- lines = append(lines, mustJSON(record))
376- }
377- if err := os.WriteFile(path, []byte(strings.Join(lines, "\n")+"\n"), 0o600); err != nil {
378- t.Fatal(err)
379- }
380- spans, err := mergeProcessTraces([]processMapping{{ProcessID: "one", ClockDomain: "one", TracePath: path, Scenario: "s", Run: 1}})
381- if err != nil {
382- t.Fatalf("observer gap did not invalidate the dropped span: %v", err)
383- }
384- if len(spans) != 0 {
385- t.Fatalf("spans=%+v, want no sample after observer gap", spans)
371+ gap := traceRecord{ProcessID: "one", Component: "observer", Scenario: "s", Run: 1, Sequence: 1, RequestID: 1, Epoch: 1, Kind: "transport_diagnostic", Valid: false}
372+ for _, tc := range []struct {
373+ name string
374+ records []traceRecord
375+ wantErr bool
376+ }{
377+ {
378+ name: "discarded span end is suppressed",
379+ records: []traceRecord{
380+ base("adapter_receive_start", 2, 2, 2, 10),
381+ gap,
382+ base("adapter_receive_end", 2, 2, 2, 20),
383+ },
384+ },
385+ {
386+ name: "unrelated unmatched end remains an error",
387+ records: []traceRecord{
388+ base("adapter_receive_start", 2, 2, 2, 10),
389+ gap,
390+ base("adapter_receive_end", 2, 2, 2, 20),
391+ base("adapter_receive_end", 3, 3, 3, 30),
392+ },
393+ wantErr: true,
394+ },
395+ } {
396+ t.Run(tc.name, func(t *testing.T) {
397+ path := filepath.Join(t.TempDir(), "trace.jsonl")
398+ var lines []string
399+ for _, record := range tc.records {
400+ lines = append(lines, mustJSON(record))
401+ }
402+ if err := os.WriteFile(path, []byte(strings.Join(lines, "\n")+"\n"), 0o600); err != nil {
403+ t.Fatal(err)
404+ }
405+ spans, err := mergeProcessTraces([]processMapping{{ProcessID: "one", ClockDomain: "one", TracePath: path, Scenario: "s", Run: 1}})
406+ if tc.wantErr {
407+ if err == nil {
408+ t.Fatalf("unrelated unmatched end accepted: spans=%+v", spans)
409+ }
410+ return
411+ }
412+ if err != nil {
413+ t.Fatalf("discarded span end was rejected: %v", err)
414+ }
415+ if len(spans) != 0 {
416+ t.Fatalf("spans=%+v, want no sample after observer gap", spans)
417+ }
418+ })
386419 }
387420 }
388421
@@ -364,25 +364,58 @@ func TestHarnessRejectsCrossProcessAndBadTraceSpans(t *testing.T) {
364 }364 }
365 }365 }
366 366
367-func TestHarnessSkipsUnpairedEndAfterObserverGap(t *testing.T) {367+func TestHarnessObserverGapSuppressesOnlyDiscardedSpanKeys(t *testing.T) {
368- path := filepath.Join(t.TempDir(), "trace.jsonl")368+ base := func(kind string, sequence, requestID, epoch uint64, tick int64) traceRecord {
369- records := []traceRecord{369+ return traceRecord{ProcessID: "one", Component: "ipc", Scenario: "s", Run: 1, Sequence: sequence, RequestID: requestID, Epoch: epoch, Kind: kind, Tick: tick, Valid: true}
370- {ProcessID: "one", Component: "observer", Scenario: "s", Run: 1, Sequence: 1, RequestID: 1, Epoch: 1, Kind: "transport_diagnostic", Valid: false},
371- {ProcessID: "one", Component: "ipc", Scenario: "s", Run: 1, Sequence: 2, RequestID: 2, Epoch: 2, Kind: "adapter_receive_end", Tick: 20, Valid: true},
372 }370 }
373- var lines []string371+ gap := traceRecord{ProcessID: "one", Component: "observer", Scenario: "s", Run: 1, Sequence: 1, RequestID: 1, Epoch: 1, Kind: "transport_diagnostic", Valid: false}
374- for _, record := range records {372+ for _, tc := range []struct {
375- lines = append(lines, mustJSON(record))373+ name string
376- }374+ records []traceRecord
377- if err := os.WriteFile(path, []byte(strings.Join(lines, "\n")+"\n"), 0o600); err != nil {375+ wantErr bool
378- t.Fatal(err)376+ }{
379- }377+ {
380- spans, err := mergeProcessTraces([]processMapping{{ProcessID: "one", ClockDomain: "one", TracePath: path, Scenario: "s", Run: 1}})378+ name: "discarded span end is suppressed",
381- if err != nil {379+ records: []traceRecord{
382- t.Fatalf("observer gap did not invalidate the dropped span: %v", err)380+ base("adapter_receive_start", 2, 2, 2, 10),
383- }381+ gap,
384- if len(spans) != 0 {382+ base("adapter_receive_end", 2, 2, 2, 20),
385- t.Fatalf("spans=%+v, want no sample after observer gap", spans)383+ },
384+ },
385+ {
386+ name: "unrelated unmatched end remains an error",
387+ records: []traceRecord{
388+ base("adapter_receive_start", 2, 2, 2, 10),
389+ gap,
390+ base("adapter_receive_end", 2, 2, 2, 20),
391+ base("adapter_receive_end", 3, 3, 3, 30),
392+ },
393+ wantErr: true,
394+ },
395+ } {
396+ t.Run(tc.name, func(t *testing.T) {
397+ path := filepath.Join(t.TempDir(), "trace.jsonl")
398+ var lines []string
399+ for _, record := range tc.records {
400+ lines = append(lines, mustJSON(record))
401+ }
402+ if err := os.WriteFile(path, []byte(strings.Join(lines, "\n")+"\n"), 0o600); err != nil {
403+ t.Fatal(err)
404+ }
405+ spans, err := mergeProcessTraces([]processMapping{{ProcessID: "one", ClockDomain: "one", TracePath: path, Scenario: "s", Run: 1}})
406+ if tc.wantErr {
407+ if err == nil {
408+ t.Fatalf("unrelated unmatched end accepted: spans=%+v", spans)
409+ }
410+ return
411+ }
412+ if err != nil {
413+ t.Fatalf("discarded span end was rejected: %v", err)
414+ }
415+ if len(spans) != 0 {
416+ t.Fatalf("spans=%+v, want no sample after observer gap", spans)
417+ }
418+ })
386 }419 }
387 }420 }
388 421

Keyboard shortcuts

Single-key shortcuts are on. ctrl k and esc work whatever this says.

ctrl k :
go to a section, branch, tag, file or repository
?
this list
j k
move the row cursor down / up
enter
open the row under the cursor; on Files, a directory opens or closes in place
↑ ↓ ← → home end
in the file tree (focus on a row): previous / next row, close / open a directory or go to its parent / first entry, first / last row. They never open a file
l h
on Files: open a directory or a file / close a directory or go up to its parent, then to the parent page. Elsewhere: into the side panel (on a merge request, its files) / back to the content; there j k move between its controls
g g G
first / last row
/
filter the list; on Files, find a file anywhere in the repository (esc clears, enter opens the best match)
b
switch branch or tag
g s g f g c g b g m g i g r g a
tabs: summary, files, commits, branches & tags, merge requests, issues, releases, actions (when available)
g h
home
w
wide mode: use the whole width of the screen, or not (remembered)
y
copy a permalink pinned to the commit
d u x
diff: split / unified / collapse or expand all
e E
diff: the whole file in view / every file, with the lines the hunks hide
n
merge request: next note in the diff
esc
close a dialog, leave a field

Branches 12

Tags 11

↑↓ moveenter openesc close