ESPNow async support: awrite concurrency

[peterhinch]

The docs are unclear whether the following is legitimate without a Lock to protect the asend calls:

async def task1():
    # code
    await espnow.asend(mac, msg)

async def task2():
    # code
    await espnow.asend(mac, msg)

If it were a UART you'd definitely need a Lock.
On my reading of the aioespnow code the writes are atomic and a Lock is not needed.

Whether or not I'm correct, the docs could do with clarifying.

[bidrohini]

I think you can post this too the esp forum too: https://www.esp32.com/search.php?search_id=unanswered

[glenn20]

Yes, this should not need a Lock (unless I have misunderstood something fundamental about uasyncio). As noted in the code, it does rely on the internal implementation in micropython/extmod/asyncio/core.py:IOQueue.queue_write().

[bulletmark]

@peterhinch, are you asking for the case where themac in task1() and task2() is the same address?

[peterhinch]

I don't believe a lock is required regardless of mac. The asend method in aioespnow calls send which is a blocking function, so there is no concurrency. The design of asend ensures that send cannot be called when a handshake is in progress. (This is on my reading of the code).

[bulletmark]

It wasn't clear whether your question was asking about concurrent messages to the same address, or concurrent messages to different addresses because I guess there could potentially be different answers. Although I agree that it probably works fine for both cases.