The requirement for iot_uart_ioctl() with xUartRequest = eGetTxNoOfbytes or eGetRxNoOfbytes is not reasonable

Start with the eGetTxNoOfbytes and eGetRxNoOfbytes definitions in iot_uart.h, then read test/test_iot_uart.c to understand the assumed concurrent async operations. Determine the intended behavior when reads and writes overlap, and document the agreed API semantics in iot_uart.h with tests covering the decision.

Hello,

We are thinking of developing our own device driver as compliant to common-io-basic interface.
Among all, we are facing an issue with iot_uart_ioctl() definition as below.

Description

The requirement for iot_uart_ioctl() with xUartRequest = eGetTxNoOfbytes or eGetRxNoOfbytes is not reasonable.

Test Steps

N/A.
This is an issue with API definitions.

Target Revision

ddfb538 (or any version on and after 8151c98).

Details

iot_uart.h says eGetTxNoOfbytes (eGetRxNoOfbytes) requires "If the last operation was read, this returns 0" (If the last operation was write, this returns 0"). See below.

 * @note eGetTxNoOfbytes returns the number of written bytes in last operation.
 * This is supposed to be called in the caller task or application callback, right after last operation completes.
 * This request expects 2 bytes buffer (uint16_t).
 *
 * - If the last operation was write, this returns the actual number of written bytes which might be smaller than the requested number (partial write).
 * - If the last operation was read, this returns 0.
 *
 * @note eGetRxNoOfbytes returns the number of read bytes in last operation.
 * This is supposed to be called in the caller task or application callback, right after last operation completes.
 * This request expects 2 bytes buffer (uint16_t).
 *
 * - If the last operation was read, this returns the actual number of read bytes which might be smaller than the requested number (partial read).
 * - If the last operation was write, this returns 0.

As test/test_iot_uart.c implies, we assume here you can call iot_uart_read_async() and iot_uart_write_async() concurrently.

In such concurrent call case, "the last operation" could change in an unexpected manner, and therefore the return value for eGetTxNoOfbytes/eGetRxNoOfbytes could be unreliable.

Imagine the events occur in the following order.

• (1) Call iot_uart_write_async()
• (2) Call iot_uart_read_async()
• (3) Async write operation is complete and the callback function is called with xOpStatus = eUartWriteCompleted
• (4) Call iot_uart_ioctl() with xUartRequest = eGetTxNoOfbytes

In (4), the guy who codes this does want to get the number of successfully written bytes in (1) and (3).

But according to iot_uart.h, this guy could get 0 in this case because "the last operation was(or could be) read".
Probably, this result is not what most people expect.

I suggest we remove "If the last operation was read, this returns 0" and "If the last operation was write, this returns 0" out of this definition.
What do you think of this idea ?