RFC: syz-cluster: consider git diffs during series triage

syz-cluster/pkg/triage/diff.go

@@ -0,0 +1,49 @@
+// Copyright 2025 syzkaller project authors. All rights reserved.
+// Use of this source code is governed by Apache 2 LICENSE that can be found in the LICENSE file.
+
+package triage
+
+import (
+	"path/filepath"
+
+	"github.com/google/syzkaller/pkg/vcs"
+	"github.com/google/syzkaller/syz-cluster/pkg/api"
+)
+
+// For now, follow the simplest strategy - just verify whether the series modifies only the files
+// irrelevant for the kernel builds.
+
+// There are many opportunities to make it smarter:
+// 1. Some source files are not compiled.
+// 2. Some source files are compiled but we have never been able to cover them during fuzzing.
+
+const reasonNotAffectsBuild = "the series only modifies irrelevant files"
+
+// NeedFuzzing returns an empty string if the series is worth of fuzzing,
+// and otherwise it returns an explanation why it is not.
+func NeedFuzzing(series *api.Series) string {
+	for _, patch := range series.Patches {
+		files := vcs.ParseGitDiff(patch.Body)
+		for _, file := range files {
+			if irrelevantPath[file] {
+				continue
+			}
+			if irrelevantExt[filepath.Ext(file)] {
+				continue
+			}
+			return ""
+		}
+	}
+	return reasonNotAffectsBuild
+}
+
+var irrelevantExt = map[string]bool{
+	".txt":  true,
+	".rst":  true,
+	".json": true,
+	".py":   true,
+}
+
+var irrelevantPath = map[string]bool{
+	"MAINTAINERS": true,
+}

syz-cluster/pkg/triage/diff_test.go

@@ -0,0 +1,66 @@
+// Copyright 2025 syzkaller project authors. All rights reserved.
+// Use of this source code is governed by Apache 2 LICENSE that can be found in the LICENSE file.
+
+package triage
+
+import (
+	"testing"
+
+	"github.com/google/syzkaller/syz-cluster/pkg/api"
+	"github.com/stretchr/testify/assert"
+)
+
+func TestNeedFuzzing(t *testing.T) {
+	tests := []struct {
+		Name    string
+		Patch   string
+		Verdict string
+	}{
+		{
+			Name: "source code changes",
+			Patch: `diff --git a/a.c b/a.c
+index c4bfa5f..7b82a4b 100644
+--- a/a.c
++++ b/a.c
+@@ -1,2 +1,2 @@
+-Old text
++new text`,
+			Verdict: "",
+		},
+		{
+			Name: "unrelated changes",
+			Patch: `diff --git a/documentation.txt b/documentation.txt
+new file mode 100644
+index 0000000..e15621b
+--- /dev/null
++++ b/documentation.txt
+@@ -0,0 +1 @@
++new text file`,
+			Verdict: reasonNotAffectsBuild,
+		},
+		{
+			Name: "irrelevant file",
+			Patch: `diff --git a/MAINTAINERS b/MAINTAINERS
+new file mode 100644
+index 0000000..e15621b
+--- /dev/null
++++ b/MAINTAINERS
+@@ -0,0 +1 @@
++new text file`,
+			Verdict: reasonNotAffectsBuild,
+		},
+	}
+
+	for _, test := range tests {
+		t.Run(test.Name, func(t *testing.T) {
+			reason := NeedFuzzing(&api.Series{
+				Patches: []api.SeriesPatch{
+					{
+						Body: []byte(test.Patch),
+					},
+				},
+			})
+			assert.Equal(t, test.Verdict, reason)
+		})
+	}
+}

syz-cluster/workflow/triage-step/main.go

@@ -57,6 +57,13 @@ func getVerdict(ctx context.Context, client *api.Client, ops triage.TreeOps) (*a
 		// TODO: the workflow step must be retried.
 		return nil, fmt.Errorf("failed to query series: %w", err)
 	}
+	if skipReason := triage.NeedFuzzing(series); skipReason != "" {
+		return &api.TriageResult{
+			Skip: &api.SkipRequest{
+				Reason: skipReason,
+			},
+		}, nil
+	}
 	trees, err := client.GetTrees(ctx)
 	if err != nil {
 		return nil, fmt.Errorf("failed to query trees: %w", err)