NimBLE-Arduino: NimBLE_Scan_Continuous.ino possible heap memory leak?

I have been noticing a potential memory leak in the continuous scan example. I have only added the following to the end of void loop():

Serial.print("HEAP: ");
Serial.println(ESP.getFreeHeap());
delay(2000);

And I see heap memory use reduce continuously:

14:15:03.247 -> HEAP: 190336
14:15:05.233 -> HEAP: 190208
14:15:07.242 -> HEAP: 190208
14:15:09.245 -> HEAP: 190208
14:15:11.230 -> HEAP: 190208
14:15:13.242 -> HEAP: 190208
14:15:15.221 -> HEAP: 190080
14:15:17.219 -> HEAP: 190080
14:15:19.245 -> HEAP: 190080
14:15:21.244 -> HEAP: 189952
14:15:23.245 -> HEAP: 189952
14:15:25.247 -> HEAP: 189952
14:15:27.215 -> HEAP: 189952
14:15:29.229 -> HEAP: 189952
14:15:31.248 -> HEAP: 189952
14:15:33.254 -> HEAP: 189952
14:15:35.086 -> Advertised Device: Name: , Address: 34:cc:2b:5e:5b:86, manufacturer data: 4c000906036a0a640627 
14:15:35.211 -> HEAP: 189952
14:15:37.245 -> HEAP: 189952
14:15:38.702 -> Advertised Device: Name: , Address: e8:fc:60:39:21:0b, manufacturer data: 4c00121900e46fc9dc373dd41da5665fc8d8293373c06494dc5b4d0100 
14:15:39.217 -> HEAP: 189824
14:15:41.251 -> HEAP: 189824
14:15:43.235 -> HEAP: 189824

Is there something that is not being cleared properly in the example?

About this issue

Original URL
State: open
Created 9 months ago
Comments: 18 (7 by maintainers)

Most upvoted comments

@h2zero I have been running an ESP32 with your new fixes to active scan for a little over 40 minutes and HEAP use has stayed constant at 229108 with the same example continuous scan with active scanning.

So I’d say this is resolved!

nravanelli on Nov 24, 2023

@scripter-co @nravanelli I have created a branch that should resolve this here: https://github.com/h2zero/NimBLE-Arduino/tree/sr-timer

Please try it out and let me know, thanks!

h2zero on Nov 24, 2023

I have looked into this and the cause is due to active scanning and how it is handled. When a device is advertising with a scan response available the device is added to the devices vector and awaits the scan response data before calling the callback and is not deleted from the vector until the scan response is received. The problem occurs when the scan response is never received, which happens quite regularly. To work around this (for now) I suggest stopping and restarting the scan occasionally to clear out the vector, or do not enable active scanning. I will contemplate options for the future to resolve this internally.

h2zero on Nov 17, 2023

Thanks, I can’t say i’ve noticed but I will test and get back

h2zero on Oct 31, 2023