Bug 102732 – IndexedDB: Change frontend to call IDBDatabaseBackendInterface::createTransaction()

• Status: RESOLVED
• Resolution: FIXED
• Component: WebCore Misc.

[Alec Flett]

Part of the transaction id refactor

In IDBDatabase::transaction, make sure we call IDBDatabaseBackend::createTransaction() rather than transaction()

[Alec Flett]

Rather than calling through IDBTransactionBackendInterface, backend object store calls should just call through IDBDatabase with a transaction id

specifically: (leaving out the details of request/callback hookup/etc)
* IDBObjectStoreBackendInterface::openCursor() becomes 
  IDBDatabaseBackendInterface::openCursor(transaction_id, objectstore_id, -1, keyRange, direction, key=false);
* IDBObjectStoreBackendInterface::openKeyCursor() becomes 
  IDBDatabaseBackendInterface::openCursor(transaction_id, objectstore_id, -1, keyRange, direction, key=true);
* IDBObjectStoreBackendInterface::count() becomes 
  IDBDatabaseBackendInterface::count(transaction_id, objectstore_id, -1, keyRange);
* IDBObjectStoreBackendInterface::get() becomes 
  IDBDatabaseBackendInterface::get(transaction_id, objectstore_id, -1, keyRange, key=false);
* IDBObjectStoreBackendInterface::put() becomes
  IDBDatabaseBackendInterface::put(transaction_id, objectstore_id, -1, key);
* IDBObjectStoreBackendInterface::deleteFunctiuon() becomes
  IDBDatabaseBackendInterface::deleteFunction(transaction_id, objectstore_id, -1, keyRange);
* IDBObjectStoreBackendInterface::clear() becomes
  IDBDatabaseBackendInterface::clear(transaction_id, objectstore_id);
* IDBObjectStoreBackendInterface::setIndexKeys() becomes
  IDBDatabaseBackendInterface::setIndexKeys(transaction_id, objectstore_id, index_ids, index_keys, ...);
* IDBObjectStoreBackendInterface::setIndexesReady() becomes
  IDBDatabaseBackendInterface::setIndexesReady(transaction_id, objectstore_id, index_ids);


This bug does not cover:
* IDBObjectStoreBackendInterface::createIndex()
* IDBObjectStoreBackendInterface::deleteIndex()

(I'm leaving openKeyCursor in here for now, even though it isn't yet implemented in the IDL or in the spec)

[Alec Flett]

[oops, wrong bug: ignore that comment 1]

[Alec Flett]

Created attachment 176310 [details]
Patch

[Alec Flett]

jsbell@ / dgrogan@ - a quick look before passing this along to tony?

[David Grogan]

Comment on attachment 176310 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=176310&action=review

> Source/WebCore/Modules/indexeddb/IDBDatabase.cpp:80
> +    static int64_t currentTransactionId = 0;

This needs to be made thread-safe. A worker thread can race with the main thread here.

You might need AtomicallyInitializedStatic and atomicIncrement.

> Source/WebCore/Modules/indexeddb/IDBTransaction.h:73
> +    int64_t id() const;

Does anything use this or m_id? I guess they will be used when backend() and m_backend are deleted?

[Alec Flett]

Created attachment 176323 [details]
Patch

[Alec Flett]

tony@ - r? (this can't land until chrome is updated, which can't happen till the next webkit roll, that's why EWS is flailing)

[Alec Flett]

Created attachment 176585 [details]
Patch for landing

[WebKit Review Bot]

Comment on attachment 176585 [details]
Patch for landing

Clearing flags on attachment: 176585

Committed r136071: <http://trac.webkit.org/changeset/136071>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.