[TW#15753] ESP-IDF GATT CLIENT demo

[pbrito55]

Hi I'm having this error:

CORRUPT HEAP: Bad tail at 0x3fffee40. Expected 0xbaad5678 got 0xbaad2902 assertion "head != NULL" failed: file "/Users/ficeto/Espressif/ESP32/esp-idf/components/heap/./multi_heap_poisoning.c", line 195, function: multi_heap_free abort() was called at PC 0x40147ebf on core 0

Backtrace: 0x40089db0:0x3fff14a0 0x40089eaf:0x3fff14c0 0x40147ebf:0x3fff14e0 0x40089af1:0x3fff1510 0x4008512a:0x3fff1530 0x400855a1:0x3fff1550 0x4000bec7:0x3fff1570 0x400d3708:0x3fff1590 0x400d3c19:0x3fff1610 0x400dfa45:0x3fff1640 0x400dfdf6:0x3fff1680 0x400dd18d:0x3fff16b0

Rebooting...

While running the the example GATT CLIENT demo.(esp-idf/examples/bluetooth/gatt_client/main/gattc_demo.c)

This happen in code line 195 & 249 while doing the free(descr_elem_result); and free(char_elem_result);

If I comment this lines the example just starts working. This is strange to me because if we have a malloc we should have then a free but he looks like he is complaining like we already had done the free.

Has I'm using the example just to interface a ble server I'm afraid if I do not do the free then I will start having memory leak.

My SW verssion Sep 29, 2017 e6afe28 Sorry about my English...

Thanks //PB

[gerbert]

I'd suggest to replace dynamic allocation in favor of static declaration of two arrays - one for 'char_elem_result' of size, say, 10 entries, the same for 'descr_elem_result'. In such limited environments dynamic memory allocation is not good. That's why, e.g., in automotive it's forbidden to use dynamic memory allocation :).

[pbrito55]

I'm already doing that.... but i looks that something strange is there I do not like when I do not understand... why it is doing strange things special with dynamic allocation.

[gerbert]

I'll tell you what, I had a trouble with one simple expression that looked like this (on master branch of ESP-IDF), comparison was done in a standalone task that was reading the data from the UART interface: if (a + b < c) { ... } (i even tried ((a + b) < c) with the same result)

it crashed (system hang and reboot)! Variable types where correct 100%, as well as values. The solution, in my case was :): uint16_t k = a + b; if (k < c) { ... } Which worked :). Later I switched to a branch 2.1 which is stable enough... in compare to master branch. Even now I don't what have happened. Maybe some stack issues or else...

[Yulong-espressif]

@pbrito55 Can you show me the code, I didn't have the same trouble with you in my environment. So I do not know how to be able to reproduce your problem. Thanks.

[pbrito55]

I was using this example in esp-arduino and got the reboot. Then decided to try it in esp-idf.

Arduino esp has heap corruption detection active. so I activated also in esp-idf by make menuconfig
(if the heap corruption is not active the there is no error in esp-idf but there is no detection also...)

The esp complains about corruption when we do the free of the memory.

Heap corruption detection (Comprehensive) ---> │ │ │ │ [*] Enable heap tracing │ │ │ │ (2) Heap tracing stack depth

and made 4 changes in the example code.

define REMOTE_SERVICE_UUID 0x1816

define REMOTE_NOTIFY_CHAR_UUID 0x2a5b

static const char remote_device_name[] = "Bike Combo Sensor";

and in line 274 / esp_ble_gattc_write_char( gattc_if, gl_profile_tab[PROFILE_A_APP_ID].conn_id, gl_profile_tab[PROFILE_A_APP_ID].char_handle, sizeof(write_char_data), write_char_data, ESP_GATT_WRITE_TYPE_RSP, ESP_GATT_AUTH_REQ_NONE); /

Thanks for your support. And if I did not explain myself correctly please ask again as my English ... Also the heap corruption functionality is not my speciality and I could be doing something wrong.

But the fact is that this is activated in the arduino and I do not know how to disable it if there is no problem in the code. And form my point of view the code of the example is correct if we have a problem must be in the GATTC API. I gess...

Once more thanks for the help.

Here is the full code:

gatt_client.zip

[Weijian-Espressif]

@pbrito55 Our client demo has a malloc error, maybe it is the reason. the line 173 "char_elem_result = (esp_gattc_char_elem_t )malloc(sizeof(char_elem_result) count);" changed to "char_elem_result = (esp_gattc_char_elem_t )malloc(sizeof(esp_gattc_char_elem_t) count);". please have a try.

[pbrito55]

Changed lines : line173: char_elem_result = (esp_gattc_char_elem_t )malloc(sizeof(esp_gattc_char_elem_t) count); line 219: descr_elem_result = (esp_gattc_descr_elem_t )malloc(sizeof(esp_gattc_descr_elem_t) count);

No more CORRUPT HEAP :-) thanks :+1: