Support renaming of IndexedDB indexes and object stores.

third_party/WebKit/Source/modules/indexeddb/IDBIndex.cpp

Index: third_party/WebKit/Source/modules/indexeddb/IDBIndex.cpp
diff --git a/third_party/WebKit/Source/modules/indexeddb/IDBIndex.cpp b/third_party/WebKit/Source/modules/indexeddb/IDBIndex.cpp
index 86e2166751b60c43799b018e011b477ce57e10a7..1599c7f233ae0a68d8ac6491df5dfad7fab1cb39 100644
--- a/third_party/WebKit/Source/modules/indexeddb/IDBIndex.cpp
+++ b/third_party/WebKit/Source/modules/indexeddb/IDBIndex.cpp
@@ -65,6 +65,46 @@ DEFINE_TRACE(IDBIndex)
     visitor->trace(m_transaction);
 }
 
+void IDBIndex::setName(const String& name, ExceptionState& exceptionState)
+{
+    if (!RuntimeEnabledFeatures::indexedDBExperimentalEnabled())

[foolip, 2016/09/09 08:12:17]

With 'use strict', trying to set a readonly property should actually throw
TypeError. I don't think trying to emulate that here makes sense, so the choice
is between silent return or throwing. Whichever you pick, can you add another a
test that fails on stable but will start passing once this is enabled?

[jsbell, 2016/09/09 18:17:22]

Did you want something different than what's captured in
third_party/WebKit/LayoutTests/webexposed/indexeddb-renames-should-not-be-exposed.html
? (that has the reversed pass/fail semantics than what you suggested)

[foolip, 2016/09/09 20:04:34]

I was thinking about something like this:

assert_throws(new TypeError, function() {
  'use strict';
  store.name = 'foo';
});

That would pass before this change and once this workaround is removed again.
But it's a bit pedantic, OK with me to not test it.

[pwnall, 2016/09/09 20:50:35]

I wrote the test and jsbell helped me figure out what it'd take for it to pass.
It seems like I can only determine if the setter was called from strict mode is
to write a custom binding. I think that's overkill, because we plan to ship this
pretty soon, together with all the other IndexedDB v2 changes.

WDYT? Are you ok with the strict mode behavior being incorrect until we ship?

[foolip, 2016/09/09 22:31:17]

Yeah, that seems OK.

+        return;
+
+    IDB_TRACE("IDBIndex::setName");
+    if (!m_transaction->isVersionChange()) {
+        exceptionState.throwDOMException(InvalidStateError, IDBDatabase::notVersionChangeTransactionErrorMessage);
+        return;
+    }
+    if (isDeleted()) {
+        exceptionState.throwDOMException(InvalidStateError, IDBDatabase::indexDeletedErrorMessage);
+        return;
+    }
+    if (m_transaction->isFinished() || m_transaction->isFinishing()) {
+        exceptionState.throwDOMException(TransactionInactiveError, IDBDatabase::transactionFinishedErrorMessage);
+        return;
+    }
+    if (!m_transaction->isActive()) {
+        exceptionState.throwDOMException(TransactionInactiveError, IDBDatabase::transactionInactiveErrorMessage);
+        return;
+    }
+
+    if (m_metadata.name == name)
+        return;
+    if (m_objectStore->containsIndex(name)) {
+        exceptionState.throwDOMException(ConstraintError, IDBDatabase::indexNameTakenErrorMessage);
+        return;
+    }
+    if (!backendDB()) {
+        exceptionState.throwDOMException(InvalidStateError, IDBDatabase::databaseClosedErrorMessage);
+        return;
+    }
+
+    backendDB()->renameIndex(m_transaction->id(), m_objectStore->id(), id(), name);
+    m_metadata.name = name;
+    m_objectStore->indexRenamed(m_metadata.id, name);
+    m_transaction->db()->indexRenamed(m_objectStore->id(), id(), name);
+}
+
 ScriptValue IDBIndex::keyPath(ScriptState* scriptState) const
 {
     return ScriptValue::from(scriptState, m_metadata.keyPath);