JsVar leak ?

[espruino-discuss]

Posted at 2017-11-28 by JumJum

I've a memory leak (JsVars) in my bluetooth implementation.
For easy testing I built a simple testfunction.
After each call of this function, process.memory.free is reduced by 2
Whats wrong in my code.

/*JSON{
	"type"	: "staticmethode",
	"class"	: "ESP32",
	"name"	: "test",
	"generate" : "jswrap_ESP32_test",
	"params":[ ["text","JsVar","test text"] ]
}
for testing only
*/
JsVar *tmp;
void jswrap_ESP32_test(JsVar *text){
	char txt[20];
	jsvGetString(text,txt,sizeof(txt)); // this is a short replacment of the data I get in Bluetooth
	tmp = jsvNewStringOfLength(sizeof(txt), txt);
	return;
}

Tried to add jsvUnLock and/or jsvUnRef on tmp, and sooner or later got HALT

[espruino-discuss]

Posted at 2017-11-28 by @gfwilliams

You're not unlocking the string created by jsvNewStringOfLength...

This should be fine?

void jswrap_ESP32_test(JsVar *text){
    char txt[20];
    jsvGetString(text,txt,sizeof(txt)); // this is a short replacment of the data I get in Bluetooth
    JsVar *tmp = jsvNewStringOfLength(sizeof(txt), txt);
    jsvUnLock(tmp);
}

I would strongly advise not keeping a global JsVar* around though - because you have to be sure to get rid of it in a _kill handler for your module (and also initialise it to 0!). If you're planning on storing data, I'd do it as a named JsVar in execInfo.hiddenRoot and grab it whenever you want it instead.

---

Posted at 2017-11-29 by JumJum

Switched to execInfo.hiddenRoot and got everything running.
This leak is history.