pkg/subsystem: extract names after all preprocessing

In the previous steps we eliminate some of the extracted subsystems. It
helps to have fewer contention while assigning the names.

As a result, we need to only rely on emails during parents
trasnformations.

3 files changed, 15 insertions, 11 deletions

diff --git a/pkg/subsystem/linux/parents.go b/pkg/subsystem/linux/parents.go
index 739514c68..6d27565bf 100644
--- a/pkg/subsystem/linux/parents.go
+++ b/pkg/subsystem/linux/parents.go
@@ -68,12 +68,15 @@ func dropSmallSubsystems(matrix *match.CoincidenceMatrix, list []*entity.Subsyst
 // we drop A, since it brings little value.
 func dropDuplicateSubsystems(matrix *match.CoincidenceMatrix, list []*entity.Subsystem) []*entity.Subsystem {
 	drop := map[*entity.Subsystem]struct{}{}
-	nameIsBetter := func(first, second string) bool {
-		// Let's prefer shorter names first.
-		if len(first) < len(second) {
-			return true
+	firstIsBetter := func(first, second *entity.Subsystem) bool {
+		firstEmail, secondEmail := "", ""
+		if len(first.Lists) > 0 {
+			firstEmail = first.Lists[0]
+		}
+		if len(second.Lists) > 0 {
+			secondEmail = second.Lists[0]
 		}
-		return first < second
+		return firstEmail < secondEmail
 	}
 	matrix.NonEmptyPairs(func(a, b *entity.Subsystem, count int) {
 		// Only consider cases when A is fully enclosed in B, i.e. M[A][B] == M[A][A].
@@ -82,7 +85,7 @@ func dropDuplicateSubsystems(matrix *match.CoincidenceMatrix, list []*entity.Sub
 		}
 		// If A and B 100% coincide, eliminate A and keep B if A > B.
 		if count == matrix.Count(b) {
-			if nameIsBetter(a.Name, b.Name) {
+			if firstIsBetter(a, b) {
 				return
 			}
 			drop[a] = struct{}{}
diff --git a/pkg/subsystem/linux/parents_test.go b/pkg/subsystem/linux/parents_test.go
index 06f7e1a3c..f1942671d 100644
--- a/pkg/subsystem/linux/parents_test.go
+++ b/pkg/subsystem/linux/parents_test.go
@@ -39,7 +39,8 @@ func TestDropDuplicateSubsystems(t *testing.T) {
 	expected = append(expected, kernel)
 
 	// Fully overlap.
-	sameA, sameB := &entity.Subsystem{Name: "SameA"}, &entity.Subsystem{Name: "SameB"}
+	sameA := &entity.Subsystem{Lists: []string{"SameA@gmail.com"}}
+	sameB := &entity.Subsystem{Lists: []string{"SameB@gmail.com"}}
 	matrix.Record(kernel, sameA, sameB)
 	matrix.Record(kernel, sameA, sameB)
 	matrix.Record(kernel, sameA, sameB)
diff --git a/pkg/subsystem/linux/subsystems.go b/pkg/subsystem/linux/subsystems.go
index 8c7d1a673..5ca57a1be 100644
--- a/pkg/subsystem/linux/subsystems.go
+++ b/pkg/subsystem/linux/subsystems.go
@@ -31,10 +31,6 @@ func listFromRepoInner(root fs.FS, rules *customRules) ([]*entity.Subsystem, err
 		extraRules: rules,
 	}
 	list := ctx.groupByList()
-	if err := setSubsystemNames(list); err != nil {
-		return nil, fmt.Errorf("failed to set names: %w", err)
-	}
-	ctx.applyExtraRules(list)
 	matrix, err := match.BuildCoincidenceMatrix(root, list, dropPatterns)
 	if err != nil {
 		return nil, err
@@ -43,6 +39,10 @@ func listFromRepoInner(root fs.FS, rules *customRules) ([]*entity.Subsystem, err
 	if err != nil {
 		return nil, err
 	}
+	if err := setSubsystemNames(list); err != nil {
+		return nil, fmt.Errorf("failed to set names: %w", err)
+	}
+	ctx.applyExtraRules(list)
 
 	// Sort subsystems by name to keep output consistent.
 	sort.Slice(list, func(i, j int) bool { return list[i].Name < list[j].Name })