Merge pull request ClickHouse#87088 from ClickHouse/backport/25.8/86563

Backport ClickHouse#86563 to 25.8: Fix Backup db engine raising exception on query with zero sized part files

Author: clickhouse-gh[bot]

src/Backups/BackupImpl.cpp

@@ -764,6 +764,16 @@ BackupImpl::readFileImpl(const String & file_name, const SizeAndChecksum & size_
     if (open_mode == OpenMode::WRITE)
         throw Exception(ErrorCodes::LOGICAL_ERROR, "The backup file should not be opened for writing. Something is wrong internally");
 
+    // Zero-sized files are not inserted into `file_infos` during metadata load,
+    // but they are present in `file_names`. Short-circuit them here and return
+    // an empty buffer without consulting `file_infos`.
+    if (size_and_checksum.first == 0)
+    {
+        std::lock_guard lock{mutex};
+        ++num_read_files;
+        return std::make_unique<ReadBufferFromOutsideMemoryFile>(file_name, std::string_view{});
+    }
+
     BackupFileInfo info;
     {
         std::lock_guard lock{mutex};
@@ -780,14 +790,6 @@ BackupImpl::readFileImpl(const String & file_name, const SizeAndChecksum & size_
         info = it->second;
     }
 
-    if (info.size == 0)
-    {
-        /// Entry's data is empty.
-        std::lock_guard lock{mutex};
-        ++num_read_files;
-        return std::make_unique<ReadBufferFromOutsideMemoryFile>(info.data_file_name, std::string_view{});
-    }
-
     if (info.encrypted_by_disk != read_encrypted)
     {
         throw Exception(

tests/queries/0_stateless/02731_zero_objects_in_metadata.reference

@@ -2,5 +2,9 @@
 1	[]
 []
 1	[]
+[]
+1	[]
+1	[]
+[]
 1	[]
 []

tests/queries/0_stateless/02731_zero_objects_in_metadata.sh

@@ -7,6 +7,9 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
 
 for DISK in s3_disk s3_cache
 do
+    BACKUP_NAME="test_s3_backup_${CLICKHOUSE_TEST_UNIQUE_NAME}_${DISK}"
+    BACKUP_DB_NAME="test_backup_db_${CLICKHOUSE_DATABASE}"
+
     ${CLICKHOUSE_CLIENT} --query "
     DROP TABLE IF EXISTS test;
     CREATE TABLE test (id Int32, empty Array(Int32))
@@ -15,16 +18,23 @@ do
 
     INSERT INTO test (id) VALUES (1);
     SELECT * FROM test;
-    "
 
-    ${CLICKHOUSE_CLIENT} --query "
-    BACKUP TABLE test TO Disk('backups', 'test_s3_backup');
+    BACKUP TABLE test TO Disk('backups', '${BACKUP_NAME}') FORMAT Null;
     DROP TABLE test;
-    RESTORE TABLE test FROM Disk('backups', 'test_s3_backup');
-    " &>/dev/null
+    RESTORE TABLE test FROM Disk('backups', '${BACKUP_NAME}') FORMAT Null;
 
-    ${CLICKHOUSE_CLIENT} --query "
     SELECT * FROM test;
     SELECT empty FROM test;
+
+    DROP DATABASE IF EXISTS "${BACKUP_DB_NAME}";
+    -- we create a backup db that points to the current test's CLICKHOUSE_DATABASE
+    -- otherwise it will have no tables.
+    CREATE DATABASE "${BACKUP_DB_NAME}"
+        ENGINE=Backup('${CLICKHOUSE_DATABASE}', Disk('backups', '${BACKUP_NAME}'));
+
+    SELECT * FROM "${BACKUP_DB_NAME}".test;
+    SELECT empty FROM "${BACKUP_DB_NAME}".test;
+
+    DROP DATABASE "${BACKUP_DB_NAME}";
     "
 done