Upload data and migrate schema #6944
Conversation
| if (auto self = weak_self.lock()) { | ||
| util::CheckedLockGuard lock(m_mutex); | ||
| if (!m_session) | ||
| return; | ||
| m_sync_schema_migration_required = true; |
There was a problem hiding this comment.
Is there ever a case where weak_self could be locked but the this pointer is not valid? I can't think of any.
There was a problem hiding this comment.
I don't think that's really possible either
| #if REALM_ENABLE_SYNC | ||
| if (m_config.sync_config) | ||
| return; | ||
| #endif |
There was a problem hiding this comment.
What is this check for? Is it to make sure it is not a sync'ed realm?
There was a problem hiding this comment.
yeah. didn't want to break the behaviour for local realms.
| return; | ||
| } | ||
|
|
||
| std::shared_ptr<AsyncOpenTask> self(shared_from_this()); |
There was a problem hiding this comment.
Maybe add a comment that you are extending the lifetime of this object until the bootstrap is complete?
There was a problem hiding this comment.
I'll add a comment. we've already been doing that with the initial subscriptions callback.
| std::function<void()> callback; | ||
| { | ||
| util::CheckedLockGuard l(m_state_mutex); | ||
| callback = m_sync_schema_migration_callback; |
There was a problem hiding this comment.
Should you "move" or clear the callback when you grab it to call it?
There was a problem hiding this comment.
sure, let's do that.
* Main work of sync schema migrations * misc * Code review changes
* Main work of sync schema migrations * misc * Code review changes
* Main work of sync schema migrations * misc * Code review changes
* Main work of sync schema migrations * misc * Code review changes
* Add schema version to flexible sync client BIND message (#6863) * Send schema version as part of BIND messages * Add pause_async to SyncSession (#6845) * Add async shutdown of SyncSession * SyncSession::shutdown returns a Future instead of passing callbacks around * Rename shutdown() to pause_async(). Hide it from the public API. * Remove unneeded header import * Upload data and migrate schema (#6944) * Handle client resets during schema migrations + integration tests (#7106) * Add option to use draft deployments when creating a schema for baas tests * Track sync schema migration between sessions + minor refactoring * Integration tests * Bump sync protocol version to 11 * Update baas commit for evergreen tests * Allow resetting schema version to zero * Improve changelog
What, How & Why?
This is the main body of work for sync schema migrations.
A schema migrations is performed whenever the client receives
schema_version_changederror (236) from the server. In such cases, the synchronization session is left active so any unsynced changes are uploaded to the server before migrating the schema. The realm file is then deleted and reopened at the new schema version. Data is bootstrapped at the new schema version.These are the steps the client and server take in case of a schema migration:
Note: Schema migrations can only be performed when a realm is opened asynchronously.
Fixes #6842.
Integration tests will be added in a separate PR.
☑️ ToDos
[ ] 📝 Changelog update[ ] 🚦 Tests (or not relevant)[ ] C-API, if public C++ API changed.