Skip to content

Native image: add Backup to the reachability metadata, and test BACKUP DATABASE on the native wheel #414

Closed

Description

@tae898

I found that the native wheel has no test for BACKUP DATABASE, and that its reachability metadata lacks the entry upstream added for it.

Upstream ArcadeData/arcadedb#12556 (commit fcc52e0f, merged 2026-11-19) registers com.arcadedb.integration.backup.Backup in the engine's own reachability-metadata.json: the (Database, String) constructor and backupDatabase(). The statement finds the class by name (Class.forName). Since ArcadeData/arcadedb#12412 (67043885, merged 2026-11-17) it looks backupDatabase up inside a lambda, which native-image cannot fold, so the nightly native build of the engine failed with NoSuchMethodException: ...Backup.backupDatabase() in RemoteDatabaseJavaApiIT.backupDatabase and JdbcQueriesIT.backupDatabase. Upstream's native-image.yml failed on its scheduled runs of 2026-11-17, 2026-11-18 and 2026-11-19, and passed on the dispatch run at fcc52e0f and on the scheduled run of 2026-11-20.

The bindings' file native/src/main/resources/META-INF/native-image/org.humemai/arcadedb-native-lib/reachability-metadata.json has no Backup entry. The only BACKUP DATABASE test, tests/test_sql_constraints_and_maintenance.py::test_backup_check_and_rebuild_index_statements, sets GlobalConfiguration.SERVER_BACKUP_DIRECTORY through JPype, so it is listed in NATIVE_NA_JVM_TESTS in tests/conftest.py and never runs on the native wheel.

Is the entry needed now? I read BackupDatabaseStatement.java at the tag 26.11.1, the engine that native/pom.xml pins. There the call is clazz.getMethod("backupDatabase").invoke(backup) in the same method as Class.forName, not in a lambda, so I expect the image to keep the method without that entry. The entry becomes needed with the first engine release that contains #12412, so with the next engine bump. I did not build a native image to confirm this, and nobody has run BACKUP DATABASE on the native wheel, so for 26.11.1 it is untested either way.

I propose two changes in one pull request:

  • Add the entry as upstream has it (same type, same two methods, no flags), in the file's own indent and alphabetical place.
  • Add a test that runs on both wheels: it starts a server (so no global setting is needed; the default backup directory is <root path>/backups), runs BACKUP DATABASE 'person.zip', finds the archive, unzips it and opens the copy. The JVM-only test stays as it is.

What I could not verify: I could not run any test on the machine where I wrote this, so the new test has not run on either wheel, and I have not built the native image. I expect the native job test-native.yml to run it (the parity suite runs the shared tests), and the JVM job test-jvm.yml to run it too. The part I am least sure of is that a server's backup lands in <root path>/backups/<database name>/; the test looks for the archive anywhere under the server root, to stay close to what I read and not to what I guess.

If you want, I will open the pull request and watch the native job until it is green or tell you what it says.

Activity

  1. self-assigned this
    on Nov 20, 2026
  2. added
    enhancementNew feature or request
    nativeThe arcadedb package (native library build)
    on Nov 20, 2026
  3. added this to the 26.12.1 milestone on Nov 20, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

enhancementNew feature or requestin progressThe team is actively working on this issuenativeThe arcadedb package (native library build)pythonPull requests that update python code

Type

No type

Projects

No projects

    Milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions