BGLIB SiliconLabs code bug? - c

Question

I'm using the SiliconLabs's BGLIB static library for comunicate via SerialPort, with BLED112.
For use this library it's necessary to define two function: one for send and another for receive bytes to and from SerialPort. All data received from serial can be divided in two part: header and data.
Following the "main.c" code of "thermometer-demo" example the function that receive data from serial, receive first of all header and than data. When the header and data has been received, the code parse the header by follow code and initialize a *msg variable:
unsigned char data[256];
struct ble_header hdr;
int r;
r = uart_rx(sizeof(hdr), (unsigned char *)&hdr, UART_TIMEOUT);
.
.
.
if (hdr.lolen) {
r = uart_rx(hdr.lolen, data, UART_TIMEOUT);
.
.
.
}
const struct ble_msg *msg = ble_get_msg_hdr(hdr);
The *msg variable is a ble_msg struct pointer and this struct is so defined:
struct ble_msg
{
struct ble_header hdr;
uint32 params;
ble_cmd_handler handler;
};
handler is a function pointer that is initialized when *msg is inizialized. After parse header the code call this function pointer:
The functions used to initialize this function pointer are defined empty from the library and if you want to use one of this, it's necessary to delete it and rewrite.
In the "thermometer-demo" example is used void ble_evt_gap_scan_response(const struct ble_msg_gap_scan_response_evt_t *msg) function to receive all visibled devices and the ble_msg_gap_scan_response_evt_t are so defined:
PACKSTRUCT(struct ble_msg_gap_scan_response_evt_t
{
int8 rssi;
uint8 packet_type;
bd_addr sender;
uint8 address_type;
uint8 bond;
uint8array data;
});
with uint8array so defined:
typedef struct
{
uint8 len;
uint8 *data;
}uint8array;
Here for me there is the problem: when the msg->handler(data) code is executed and than the ble_evt_gap_scan_response function is called, a cast is executed between data buffer and ble_msg_gap_scan_response_evt_t struct and the content of uint8 *data (in uint8array struct) is initialized with data received from SerialPort and than, will point to a non correct RAM location.
I think the intention is to provide a way to have a directly access to the data but for me is not the correct way.
It's my wrong evaluation or is a bglib bug?

Answer 1

I have contacted SiliconLabs support and that the original libraries are correct and it seem I have a incorrect version of the libraries.
Like the SiliconLabs support suggest me (and like explained very good in this link: http://www.drdobbs.com/questions-answers-creating-variable-siz/184403480), there are 3 possible kind of declarations:
1° declaration:
typedef struct
{
uint8 len;
uint8 data[0];
}uint8array;
and I access to data with:
msg->data->data[0]
msg->data->data[1]
msg->data->data[2]
...
msg->data->data[n]
In this case could happen compiler warns that the buffer index exceeds the limits.
2° declaration:
typedef struct
{
uint8 len;
uint8 data[1000];
}uint8array;
In this case it's possible to access to data in the same way as previous one but or more space has been allocaded than necessary (in embedded sometime it's important optimize because the resources are limited) or, like previous, you could exceed the buffer limits.
3° declaration:
typedef struct
{
uint8 len;
uint8 data[];
}uint8array;
The name of this declaration is Flexible array member (https://en.wikipedia.org/wiki/Flexible_array_member) and are introduced in the C99 standard. In this case it's possible to access to data in the same way as previouses one but the compiler can warns that isn't a standard declaration.
This declaration is the same adopted by SiliconLabs for the library.
In my case I'm using VisualStudio and I have adopted the third solution but compiler return me the warning C4200 (non standard declaration). I tried the first solution and compiler doesn't return me no warns but for me, the clarity of code is missing and could be a problem for future uses.