Fix the five resilience review findings

2026-09-24 21:48:18 +00:00
parent 9fc4de22b7
commit ec78b5a5db
19 changed files with 550 additions and 132 deletions
--- a/internal/framework/runner/runner.go
+++ b/internal/framework/runner/runner.go
@@ -2,6 +2,7 @@ package runner

 import (
 	"context"
+	"errors"
 	"fmt"
 	"path/filepath"
 	"sort"
@@ -86,6 +87,8 @@ type ValidatorRejectedChange struct {

 // RunInput is the deterministic runner input.
 type RunInput struct {
+	// ReportDiagnosticError may be called concurrently; it must not affect processing.
+	ReportDiagnosticError    func(error) `json:"-"`
 	Config                   *config.Config
 	Transcript               *schema.Transcript
 	Glossary                 *schema.Glossary
@@ -188,6 +191,7 @@ func (r *Runner) run(ctx context.Context, input RunInput) (RunOutput, error) {
 			ValidationClient:         input.ValidationLLMClient,
 			ValidationDiagnosticsDir: input.ValidationDiagnosticsDir,
 			Timings:                  input.timings,
+			ReportDiagnosticError:    input.ReportDiagnosticError,
 		})
 		if input.CaptureCandidates {
 			candidates = append(candidates, pipelineResult.Candidates...)
@@ -293,6 +297,8 @@ func applyWithinSections(sections []chunking.Section, approved []proposals.Enric
 }

 type collectSectionProposalsInput struct {
+	// ReportDiagnosticError may be called concurrently; it must not affect processing.
+	ReportDiagnosticError    func(error) `json:"-"`
 	Module                   contracts.TranscriptModule
 	Spec                     contracts.ModuleRunSpec
 	Policy                   proposals.ReplacementPolicy
@@ -356,11 +362,12 @@ func collectSectionProposals(ctx context.Context, input collectSectionProposalsI
 				meta := contracts.SectionMetadataFromSection(section)
 				proposalResult, err := input.Module.Propose(runCtx, contracts.ProposalRequest{
 					ExecutionContext: contracts.ExecutionContext{
-						Config:            input.Config,
-						WorkingTranscript: transcriptFromSection(section),
-						Glossary:          input.Glossary,
-						Section:           &meta,
-						DiagnosticsDir:    input.DiagnosticsDir,
+						Config:                input.Config,
+						WorkingTranscript:     transcriptFromSection(section),
+						Glossary:              input.Glossary,
+						Section:               &meta,
+						DiagnosticsDir:        input.DiagnosticsDir,
+						ReportDiagnosticError: input.ReportDiagnosticError,
 					},
 					RunSpec: contracts.ModuleRunSpec{
 						ModuleKey:         input.Spec.ModuleKey,
@@ -495,17 +502,18 @@ func runModulePipeline(ctx context.Context, input collectSectionProposalsInput)
 			go func(sectionPos int, enriched []proposals.EnrichedCorrectionProposal, sectionMetadata contracts.SectionMetadata) {
 				defer vwg.Done()
 				validated, err := validateSectionCandidates(runCtx, validateSectionCandidatesInput{
-					Spec:                input.Spec,
-					Policy:              input.Policy,
-					Glossary:            input.Glossary,
-					Config:              input.Config,
-					WorkingTranscript:   transcriptFromSection(sectionResult.section),
-					SectionIndex:        &sectionMetadata.Index,
-					Validators:          validatorsOrdered,
-					SectionEnriched:     enriched,
-					ValidationLLMClient: input.ValidationClient,
-					DiagnosticsDir:      input.ValidationDiagnosticsDir,
-					Timings:             input.Timings,
+					Spec:                  input.Spec,
+					Policy:                input.Policy,
+					Glossary:              input.Glossary,
+					Config:                input.Config,
+					WorkingTranscript:     transcriptFromSection(sectionResult.section),
+					SectionIndex:          &sectionMetadata.Index,
+					Validators:            validatorsOrdered,
+					SectionEnriched:       enriched,
+					ValidationLLMClient:   input.ValidationClient,
+					DiagnosticsDir:        input.ValidationDiagnosticsDir,
+					Timings:               input.Timings,
+					ReportDiagnosticError: input.ReportDiagnosticError,
 				})
 				setErr(err)
 				validationResults <- sectionValidationResult{
@@ -578,17 +586,19 @@ func runModulePipeline(ctx context.Context, input collectSectionProposalsInput)
 }

 type validateSectionCandidatesInput struct {
-	Spec                contracts.ModuleRunSpec
-	Policy              proposals.ReplacementPolicy
-	Glossary            *schema.Glossary
-	Config              *config.Config
-	WorkingTranscript   *schema.Transcript
-	SectionIndex        *int
-	Validators          []contracts.Validator
-	SectionEnriched     []proposals.EnrichedCorrectionProposal
-	ValidationLLMClient contracts.StructuredLLMClient
-	DiagnosticsDir      string
-	Timings             *validatorTimingCollector
+	// ReportDiagnosticError may be called concurrently; it must not affect processing.
+	ReportDiagnosticError func(error) `json:"-"`
+	Spec                  contracts.ModuleRunSpec
+	Policy                proposals.ReplacementPolicy
+	Glossary              *schema.Glossary
+	Config                *config.Config
+	WorkingTranscript     *schema.Transcript
+	SectionIndex          *int
+	Validators            []contracts.Validator
+	SectionEnriched       []proposals.EnrichedCorrectionProposal
+	ValidationLLMClient   contracts.StructuredLLMClient
+	DiagnosticsDir        string
+	Timings               *validatorTimingCollector
 }

 type validateSectionCandidatesResult struct {
@@ -622,38 +632,34 @@ func validateSectionCandidates(ctx context.Context, input validateSectionCandida

 		validatorStarted := time.Now()
 		vResult, err := validator.Validate(ctx, contracts.ValidationRequest{
-			WorkingTranscript: input.WorkingTranscript,
-			CandidateProposal: eligible,
-			ModuleKey:         input.Spec.ModuleKey,
-			ModuleInstance:    input.Spec.InstanceName,
-			SectionIndex:      input.SectionIndex,
-			ValidatorOrder:    &validatorOrder,
-			ReplacementPolicy: input.Policy,
-			Glossary:          input.Glossary,
-			Config:            input.Config,
-			LLMClient:         input.ValidationLLMClient,
-			DiagnosticsWriter: diagnosticsWriter,
+			WorkingTranscript:     input.WorkingTranscript,
+			CandidateProposal:     eligible,
+			ModuleKey:             input.Spec.ModuleKey,
+			ModuleInstance:        input.Spec.InstanceName,
+			SectionIndex:          input.SectionIndex,
+			ValidatorOrder:        &validatorOrder,
+			ReplacementPolicy:     input.Policy,
+			Glossary:              input.Glossary,
+			Config:                input.Config,
+			LLMClient:             input.ValidationLLMClient,
+			DiagnosticsWriter:     diagnosticsWriter,
+			ReportDiagnosticError: input.ReportDiagnosticError,
 		})
 		if input.Timings != nil {
 			input.Timings.record(input.Spec.ModuleKey, input.Spec.InstanceName, validator.Name(), validatormetadata.ClassOf(validator) == validatormetadata.ExecutionClassLLMBacked, time.Since(validatorStarted))
 		}
 		operations = append(operations, vResult.Operations...)

+		var decisionErr error
 		if err != nil {
-			return validateSectionCandidatesResult{
-				decisions:  decisions,
-				rejected:   rejected,
-				operations: operations,
-				filtered:   filtered,
-			}, fmt.Errorf("validator %q failed: %w", validator.Name(), err)
+			decisionErr = validators.EnforcePartialDecisionCardinality(eligible, vResult.Decisions)
+		} else {
+			decisionErr = validators.EnforceDecisionCardinality(eligible, vResult.Decisions)
 		}
-		if err := validators.EnforceDecisionCardinality(eligible, vResult.Decisions); err != nil {
+		if decisionErr != nil {
 			return validateSectionCandidatesResult{
-				decisions:  decisions,
-				rejected:   rejected,
-				operations: operations,
-				filtered:   filtered,
-			}, fmt.Errorf("validator %q cardinality failed: %w", validator.Name(), err)
+				decisions: decisions, rejected: rejected, operations: operations, filtered: filtered,
+			}, fmt.Errorf("validator %q cardinality failed: %w", validator.Name(), errors.Join(err, decisionErr))
 		}

 		nextEligible := make([]proposals.EnrichedCorrectionProposal, 0, len(eligible))
@@ -694,6 +700,13 @@ func validateSectionCandidates(ctx context.Context, input validateSectionCandida
 			})
 		}

+		if err != nil {
+			// Completed decisions are evidence only. A fatal return never grants
+			// approval to candidates that have not finished the full chain.
+			return validateSectionCandidatesResult{
+				decisions: decisions, rejected: rejected, operations: operations, filtered: filtered,
+			}, fmt.Errorf("validator %q failed: %w", validator.Name(), err)
+		}
 		eligible = nextEligible
 	}

--- a/internal/framework/runner/runner_test.go
+++ b/internal/framework/runner/runner_test.go
@@ -1280,3 +1280,20 @@ func TestRunnerGlossaryModuleUsesConfidenceThreshold(t *testing.T) {
 		t.Fatalf("expected low confidence reason, got %+v", out.ModuleResults[0].ValidatorRejected[0])
 	}
 }
+
+func TestFatalPartialDecisionsRejectUnknownAndDuplicateIdentity(t *testing.T) {
+	origin := errors.New("original fatal failure")
+	candidate := proposals.EnrichedCorrectionProposal{ProposalMetadata: proposals.ProposalMetadata{ProposalIndex: 7}}
+	for _, decisions := range [][]validators.Decision{
+		{{ProposalIndex: 8, Approved: true}},
+		{{ProposalIndex: 7, Approved: true}, {ProposalIndex: 7, Approved: true}},
+	} {
+		v := fakeValidator{name: "broken", validateF: func(contracts.ValidationRequest) (validators.Result, error) {
+			return validators.Result{Decisions: decisions}, origin
+		}}
+		result, err := validateSectionCandidates(context.Background(), validateSectionCandidatesInput{Validators: []contracts.Validator{v}, SectionEnriched: []proposals.EnrichedCorrectionProposal{candidate}})
+		if !errors.Is(err, origin) || len(result.decisions) != 0 || len(result.approved) != 0 {
+			t.Fatalf("invalid partial evidence accepted or origin lost: %+v %v", result, err)
+		}
+	}
+}