From 2441d038e528242dde19547406018d0f81b2b7eb Mon Sep 17 00:00:00 2001
From: Sandy Xu <sandy@juicedata.io>
Date: Thu, 12 Sep 2024 14:36:33 +0800
Subject: [PATCH] meta: in compaction don't skip the first slice if it's
 duplicated (#5159)

---
 pkg/meta/base_test.go | 15 +++++++++++++++
 pkg/meta/slice.go     |  6 ++++++
 2 files changed, 21 insertions(+)

diff --git a/pkg/meta/base_test.go b/pkg/meta/base_test.go
index 4cd96fc3630d..4bc4fdda5e27 100644
--- a/pkg/meta/base_test.go
+++ b/pkg/meta/base_test.go
@@ -1324,6 +1324,21 @@ func testCompaction(t *testing.T, m Meta, trash bool) {
 		t.Fatalf("inode %d should be compacted, but have %d slices, size %d,%d,%d",
 			inode, len(slices), slices[0].Len, slices[1].Len, slices[2].Len)
 	}
+
+	m.NewSlice(ctx, &sliceId)
+	_ = m.Write(ctx, inode, 3, 0, Slice{Id: sliceId, Size: 2338508, Len: 2338508}, time.Now())
+	_ = m.CopyFileRange(ctx, inode, 3*ChunkSize, inode, 4*ChunkSize, 2338508, 0, nil, nil)
+	_ = m.Fallocate(ctx, inode, fallocZeroRange, 4*ChunkSize, ChunkSize, nil)
+	_ = m.CopyFileRange(ctx, inode, 3*ChunkSize, inode, 4*ChunkSize, 2338508, 0, nil, nil)
+	if c, ok := m.(compactor); ok {
+		c.compactChunk(inode, 4, false, true)
+	}
+	if st := m.Read(ctx, inode, 4, &slices); st != 0 {
+		t.Fatalf("read inode %d chunk 4: %s", inode, st)
+	}
+	if len(slices) != 1 || slices[0].Len != 2338508 {
+		t.Fatalf("inode %d should be compacted, but have %d slices, size %d", inode, len(slices), slices[0].Len)
+	}
 }
 
 func testConcurrentWrite(t *testing.T, m Meta) {
diff --git a/pkg/meta/slice.go b/pkg/meta/slice.go
index d036954dba80..08bdda4c63a6 100644
--- a/pkg/meta/slice.go
+++ b/pkg/meta/slice.go
@@ -174,6 +174,7 @@ func compactChunk(ss []*slice) (uint32, uint32, []Slice) {
 func skipSome(chunk []*slice) int {
 	var skipped int
 	var total = len(chunk)
+OUT:
 	for skipped < total {
 		ss := chunk[skipped:]
 		pos, size, c := compactChunk(ss)
@@ -189,6 +190,11 @@ func skipSome(chunk []*slice) int {
 			// it's not the first slice, compact it
 			break
 		}
+		for _, s := range ss[1:] {
+			if *s == *first {
+				break OUT
+			}
+		}
 		skipped++
 	}
 	return skipped