Skip invalid GitHub packages instead of failing the scan

Wrap package parse errors in provider.ErrInvalidPackage so only transient
asset and insert errors fail the scan; cover rpm, deb and alpine.

internal/provider/rpm/github.go

@@ -343,8 +343,8 @@ func (p *GitHubProvider) scanWithState(ctx context.Context, remote models.Remote
 		return newEtag, false, err
 	}
 
-	// A failed asset fails the scan so the ETag is not advanced and the retry
-	// re-derives it; assets that did derive are kept and served meanwhile.
+	// A transiently failed asset fails the scan so the ETag is not advanced and
+	// the retry re-derives it; an invalid package is skipped as no retry fixes it.
 	var failed []error
 	seen := map[string]bool{}
 	for _, rel := range releases {
@@ -376,7 +376,9 @@ func (p *GitHubProvider) scanWithState(ctx context.Context, remote models.Remote
 			meta, err := p.deriveAsset(ctx, remote, asset, fp)
 			if err != nil {
 				slog.Warn("github_rpm: derive asset failed", "remote", remote.Name, "asset", asset.Name, "error", err)
-				failed = append(failed, fmt.Errorf("%s: %w", asset.Name, err))
+				if !errors.Is(err, provider.ErrInvalidPackage) {
+					failed = append(failed, fmt.Errorf("%s: %w", asset.Name, err))
+				}
 				continue
 			}
 			if err := store.InsertRPMMetadata(ctx, meta); err != nil {
@@ -574,7 +576,7 @@ func (p *GitHubProvider) fetchHeader(ctx context.Context, remote models.Remote,
 			}
 			continue
 		}
-		return nil, fmt.Errorf("parse rpm header: %w", perr)
+		return nil, fmt.Errorf("%w: parse rpm header: %w", provider.ErrInvalidPackage, perr)
 	}
 }
 

internal/provider/rpm/syncer_test.go

@@ -400,42 +400,57 @@ func TestSyncerFailedScanRetriesAfterBackoff(t *testing.T) {
 	}
 }
 
-// An asset that fails to derive fails the scan: the ETag is not advanced, a
-// retry is scheduled, and the retry against the same unchanged release list
-// derives the package instead of waiting for upstream to publish again.
-func TestSyncerFailedAssetRetriesWithoutAdvancingEtag(t *testing.T) {
-	fx := newGitHubFixture(t, true)
-	fx.etag = `"v1"`
-	fx.rpmBytes["other-9-9.aarch64.rpm"] = testsupport.MinimalRPM("other", "9", "9", "aarch64")
-	fx.assetFail["other-9-9.aarch64.rpm"] = 1
-	store := newFakeSyncStore()
-	p := newTestProvider()
-	s := newSyncer(store, p, testSyncConfig())
-	remote := fx.remote()
-	bg := context.Background()
+// A transiently failed asset fails the scan: the ETag is not advanced, a retry
+// is scheduled, and the retry against the unchanged release list derives it. An
+// invalid package is skipped instead, since no retry can fix it.
+func TestSyncerFailedAssetRetry(t *testing.T) {
+	for _, tc := range []struct {
+		name      string
+		corrupt   bool
+		wantRetry bool
+	}{
+		{name: "transient", wantRetry: true},
+		{name: "invalid package", corrupt: true},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			fx := newGitHubFixture(t, true)
+			fx.etag = `"v1"`
+			if tc.corrupt {
+				fx.rpmBytes["other-9-9.aarch64.rpm"] = []byte("not a package")
+			} else {
+				fx.rpmBytes["other-9-9.aarch64.rpm"] = testsupport.MinimalRPM("other", "9", "9", "aarch64")
+				fx.assetFail["other-9-9.aarch64.rpm"] = 1
+			}
+			store := newFakeSyncStore()
+			s := newSyncer(store, newTestProvider(), testSyncConfig())
+			remote := fx.remote()
+			bg := context.Background()
 
-	s.process(bg, syncJob{remote: remote, prime: true})
-	if !store.lastResult.Failed {
-		t.Fatal("scan with a failed asset was recorded as a successful sync")
-	}
-	if _, saved := store.etags[remote.Name]; saved {
-		t.Fatalf("failed scan saved etag %q", store.etags[remote.Name])
-	}
-	if _, pending := store.retryAt[remote.Name]; !pending {
-		t.Fatal("no retry scheduled after a failed asset")
-	}
-	if rows, _ := store.ListRPMMetadataEntries(bg, remote.Name); len(rows) != 1 {
-		t.Fatalf("healthy asset not served meanwhile: %d rows", len(rows))
-	}
+			s.process(bg, syncJob{remote: remote, prime: true})
+			_, pending := store.retryAt[remote.Name]
+			if pending != tc.wantRetry {
+				t.Fatalf("retry pending=%v, want %v", pending, tc.wantRetry)
+			}
+			if _, saved := store.etags[remote.Name]; saved == tc.wantRetry {
+				t.Fatalf("etag saved=%v after first scan", saved)
+			}
+			if rows, _ := store.ListRPMMetadataEntries(bg, remote.Name); len(rows) != 1 {
+				t.Fatalf("healthy asset not served: %d rows", len(rows))
+			}
+			if !tc.wantRetry {
+				return
+			}
 
-	store.mu.Lock()
-	store.retryAt[remote.Name] = time.Now().Add(-time.Second)
-	store.mu.Unlock()
-	s.process(bg, syncJob{remote: remote})
-	if store.lastResult.Failed || store.etags[remote.Name] != `"v1"` {
-		t.Fatalf("retry not recorded: failed=%v etag=%q", store.lastResult.Failed, store.etags[remote.Name])
-	}
-	if rows, _ := store.ListRPMMetadataEntries(bg, remote.Name); len(rows) != 2 {
-		t.Fatalf("retry did not derive the failed asset: %d rows", len(rows))
+			store.mu.Lock()
+			store.retryAt[remote.Name] = time.Now().Add(-time.Second)
+			store.mu.Unlock()
+			s.process(bg, syncJob{remote: remote})
+			if _, pending := store.retryAt[remote.Name]; pending || store.etags[remote.Name] != `"v1"` {
+				t.Fatalf("retry not recorded: pending=%v etag=%q", pending, store.etags[remote.Name])
+			}
+			if rows, _ := store.ListRPMMetadataEntries(bg, remote.Name); len(rows) != 2 {
+				t.Fatalf("retry did not derive the failed asset: %d rows", len(rows))
+			}
+		})
 	}
 }