Justin Tobler [off-list ref] writes:

On 25/11/06 09:39AM, Karthik Nayak wrote:

quoted

@@ -1858,33 +1902,8 @@ static int do_fetch(struct transport *transport,
 	if (retcode)
 		goto cleanup;

-	retcode = ref_transaction_commit(transaction, &err);
-	if (retcode) {
-		/*
-		 * Explicitly handle transaction cleanup to avoid
-		 * aborting an already closed transaction.
-		 */
-		ref_transaction_free(transaction);
-		transaction = NULL;
-		goto cleanup;
-	}
-
-	if (!atomic_fetch) {
-		struct ref_rejection_data data = {
-			.retcode = &retcode,
-			.conflict_msg_shown = 0,
-			.remote_name = transport->remote->name,
-		};
-
-		ref_transaction_for_each_rejected_update(transaction,
-							 ref_transaction_rejection_handler,
-							 &data);
-		if (retcode) {
-			ref_transaction_free(transaction);
-			transaction = NULL;
-			goto cleanup;
-		}
-	}
+	retcode = commit_ref_transaction(&transaction, atomic_fetch,
+					 transport->remote->name, &err);

It looks like previously, whenever `ref_transaction_commit()` or
`ref_transaction_rejection_handler()` returned a non-zero value, we
would "goto cleanup". Now that these operations are handled via
`commit_ref_transaction()` though, it looks like we no longer handle the
"retcode" return value and just continue. I think we still need to check
the return value here.

-Justin

Good catch, will add this in. Thanks