indexeddb: fix getall crash with null receiver

When transactions are aborted, any get all operations should still send
an error with a valid receiver.  This avoids a dcheck (and probable
crash later) from mojo due to passing a null receiver.

Bug: 1110781
Change-Id: If142072cae93f29dc3d4c1ac388bc27a3f3939f3
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2324525
Commit-Queue: Daniel Murphy <dmurph@chromium.org>
Reviewed-by: Daniel Murphy <dmurph@chromium.org>
Auto-Submit: enne <enne@chromium.org>
Cr-Commit-Position: refs/heads/master@{#792837}

content/browser/indexed_db/indexed_db_callback_helpers.cc

@@ -12,7 +12,14 @@ mojo::PendingReceiver<blink::mojom::IDBDatabaseGetAllResultSink> AbortCallback(
     base::WeakPtr<IndexedDBTransaction> transaction) {
   if (transaction)
     transaction->IncrementNumErrorsSent();
-  return mojo::NullReceiver();
+
+  mojo::Remote<blink::mojom::IDBDatabaseGetAllResultSink> remote;
+  auto receiver = remote.BindNewPipeAndPassReceiver();
+
+  IndexedDBDatabaseError error(blink::mojom::IDBException::kIgnorableAbortError,
+                               "Backend aborted error");
+  remote->OnError(blink::mojom::IDBError::New(error.code(), error.message()));
+  return receiver;
 }
 
 }  // namespace indexed_db_callback_helpers_internal