From 8d41190df0b6184bac7d8b34765fc84395f27cf4 Mon Sep 17 00:00:00 2001
From: Aleksandr Nogikh <nogikh@google.com>
Date: Wed, 18 Jan 2023 16:22:07 +0100
Subject: pkg/subsystem/linux: keep one list for list-merged groups

Otherwise we can get too many mailing lists at the same time.
---
 pkg/subsystem/linux/subsystems.go      | 19 +++++++++++--------
 pkg/subsystem/linux/subsystems_test.go |  3 ++-
 2 files changed, 13 insertions(+), 9 deletions(-)

(limited to 'pkg/subsystem')

diff --git a/pkg/subsystem/linux/subsystems.go b/pkg/subsystem/linux/subsystems.go
index 9d9688758..8d8716bf9 100644
--- a/pkg/subsystem/linux/subsystems.go
+++ b/pkg/subsystem/linux/subsystems.go
@@ -79,7 +79,7 @@ func (ctx *linuxCtx) getSubsystems() ([]*entity.Subsystem, error) {
 	ret := []*entity.Subsystem{}
 	for _, raw := range ctx.groupByList() {
 		s := &entity.Subsystem{}
-		mergeRawRecords(s, raw.records)
+		raw.mergeRawRecords(s)
 		ret = append(ret, s)
 	}
 	if err := setSubsystemNames(ret); err != nil {
@@ -98,7 +98,7 @@ func (ctx *linuxCtx) applyExtraRules(list []*entity.Subsystem) {
 	}
 }
 
-func mergeRawRecords(subsystem *entity.Subsystem, records []*maintainersRecord) {
+func (candidate *subsystemCandidate) mergeRawRecords(subsystem *entity.Subsystem) {
 	unique := func(list []string) []string {
 		m := make(map[string]struct{})
 		for _, s := range list {
@@ -111,19 +111,22 @@ func mergeRawRecords(subsystem *entity.Subsystem, records []*maintainersRecord)
 		sort.Strings(ret)
 		return ret
 	}
-	var lists, maintainers []string
-	for _, record := range records {
+	var maintainers []string
+	for _, record := range candidate.records {
 		rule := record.ToPathRule()
 		if !rule.IsEmpty() {
 			subsystem.PathRules = append(subsystem.PathRules, rule)
 		}
-		lists = append(lists, record.lists...)
 		maintainers = append(maintainers, record.maintainers...)
 	}
-	subsystem.Lists = unique(lists)
-	// But there's a risk that we collect too many unrelated maintainers, so
+	if candidate.commonEmail != "" {
+		// For list-grouped subsystems, we risk merging just too many lists.
+		// Keep the list short in this case.
+		subsystem.Lists = []string{candidate.commonEmail}
+	}
+	// There's a risk that we collect too many unrelated maintainers, so
 	// let's only merge them if there are no lists.
-	if len(records) <= 1 {
+	if len(candidate.records) <= 1 {
 		subsystem.Maintainers = unique(maintainers)
 	}
 }
diff --git a/pkg/subsystem/linux/subsystems_test.go b/pkg/subsystem/linux/subsystems_test.go
index b4f0818ff..09d835e3f 100644
--- a/pkg/subsystem/linux/subsystems_test.go
+++ b/pkg/subsystem/linux/subsystems_test.go
@@ -75,7 +75,7 @@ func TestLinuxSubsystemPaths(t *testing.T) {
 	// For the list of subsystems, see TestLinuxSubsystemsList.
 	// Here we rely on the same ones.
 	repo := prepareTestLinuxRepo(t, []byte(testMaintainers))
-	subsystems, err := listFromRepoInner(repo, testRules)
+	subsystems, err := listFromRepoInner(repo, nil)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -106,6 +106,7 @@ func TestLinuxSubsystemPaths(t *testing.T) {
 			list: []string{"kernel", "mm"},
 		},
 		{
+			// Shmem has no mailing list.
 			path: `mm/shmem.c`,
 			list: []string{"kernel", "mm"},
 		},
-- 
cgit mrf-deployment