FS-10167: Rewrote the ks_pool allocator, no longer uses paging or internal block allocation, but still retains reference counting and auto cleanup callbacks, should be much more efficient now on windows than the original mmap approach, and all tests now run successfully!

libs/libblade/src/blade_connection.c

@@ -312,7 +312,7 @@ KS_DECLARE(void) blade_connection_disconnect(blade_connection_t *bc)
 {
 	ks_assert(bc);
 
-	if (bc->state != BLADE_CONNECTION_STATE_DETACH && bc->state != BLADE_CONNECTION_STATE_DISCONNECT) {
+	if (bc->state != BLADE_CONNECTION_STATE_DETACH && bc->state != BLADE_CONNECTION_STATE_DISCONNECT && bc->state != BLADE_CONNECTION_STATE_CLEANUP) {
 		ks_log(KS_LOG_DEBUG, "Connection (%s) disconnecting\n", bc->id);
 		blade_connection_state_set(bc, BLADE_CONNECTION_STATE_DETACH);
 	}
@@ -453,7 +453,7 @@ ks_status_t blade_connection_state_on_attach(blade_connection_t *bc)
 	callback = blade_connection_state_callback_lookup(bc, BLADE_CONNECTION_STATE_ATTACH);
 	if (callback) hook = callback(bc, BLADE_CONNECTION_STATE_CONDITION_POST);
 
-	if (hook == BLADE_CONNECTION_STATE_HOOK_DISCONNECT)	blade_connection_disconnect(bc);
+	if (hook == BLADE_CONNECTION_STATE_HOOK_DISCONNECT) blade_connection_disconnect(bc);
 	else if (hook == BLADE_CONNECTION_STATE_HOOK_SUCCESS) {
 		// @todo this is adding a second lock, since we keep it locked in the callback to allow finishing, we don't want get locking here...
 		// or just try unlocking twice to confirm...

libs/libblade/src/blade_datastore.c

@@ -77,7 +77,7 @@ KS_DECLARE(ks_status_t) blade_datastore_create(blade_datastore_t **bdsP, ks_pool
 
 	ks_assert(bdsP);
 	ks_assert(pool);
-	ks_assert(tpool);
+	//ks_assert(tpool);
 	
 	bds = ks_pool_alloc(pool, sizeof(*bds));
 	bds->pool = pool;

libs/libblade/src/blade_module_wss.c

@@ -869,7 +869,7 @@ blade_connection_state_hook_t blade_transport_wss_on_state_disconnect(blade_conn
 	list_delete(&bt_wss->module->connected, bc);
 
 	if (bt_wss_init) blade_transport_wss_init_destroy(&bt_wss_init);
-	if (bt_wss) blade_transport_wss_destroy(&bt_wss);
+	if (bt_wss) blade_transport_wss_destroy(&bt_wss); // @TODO: Scream at this very loudly until I feel better for it wasting 2 days to track down, and then fix the issue it's causing
 
 	return BLADE_CONNECTION_STATE_HOOK_SUCCESS;
 }
@@ -1092,6 +1092,8 @@ blade_connection_state_hook_t blade_transport_wss_on_state_attach_inbound(blade_
 	// behaviour to simply go as far as assigning a session to the connection and let the system handle the rest
 	if (json_req) cJSON_Delete(json_req);
 	if (json_res) cJSON_Delete(json_res);
+
+	
 	return ret;
 }
 
@@ -1223,6 +1225,7 @@ blade_connection_state_hook_t blade_transport_wss_on_state_attach_outbound(blade
  done:
 	if (json_req) cJSON_Delete(json_req);
 	if (json_res) cJSON_Delete(json_res);
+
 	return ret;
 }
 
@@ -1261,15 +1264,15 @@ blade_connection_state_hook_t blade_transport_wss_on_state_ready_outbound(blade_
 
 	if (condition == BLADE_CONNECTION_STATE_CONDITION_PRE) {
 		blade_session_t *bs = NULL;
-		cJSON *req = NULL;
+		//cJSON *req = NULL;
 
 		ks_log(KS_LOG_DEBUG, "State Callback: %d\n", (int32_t)condition);
 
 		bs = blade_handle_sessions_get(blade_connection_handle_get(bc), blade_connection_session_get(bc));
 		ks_assert(bs);
 
-		blade_rpc_request_create(blade_connection_pool_get(bc), &req, NULL, NULL, "blade.test.echo");
-		blade_session_send(bs, req, blade_test_echo_response_handler);
+		//blade_rpc_request_create(blade_connection_pool_get(bc), &req, NULL, NULL, "blade.test.echo");
+		//blade_session_send(bs, req, blade_test_echo_response_handler);
 
 		blade_session_read_unlock(bs);
 	}

libs/libblade/src/blade_session.c

@@ -159,6 +159,7 @@ KS_DECLARE(ks_status_t) blade_session_shutdown(blade_session_t *bs)
 	if (bs->state_thread) {
 		bs->shutdown = KS_TRUE;
 		ks_thread_join(bs->state_thread);
+		printf("FREEING SESSION THREAD %p\n", (void *)bs->state_thread);
 		ks_pool_free(bs->pool, &bs->state_thread);
 		bs->shutdown = KS_FALSE;
 	}