Fix WiFiClientRxBuffer::fillBuffer() error, likely solves issue #2212

[Jeroen88]

Fix WiFiClientRxBuffer::fillBuffer() error in handing a negative res from the call to recv()

In

arduino-esp32/libraries/WiFi/src/WiFiClient.cpp

Line 70 in 229d9b7

if(res < 0 && errno != EWOULDBLOCK) {

if res is negative and errno equals EWOULDBLOCK then the if() condition is false causing _fill to be decremented.
The condition res < 0 && errno == EWOULDBLOCK simply means that the socket has no data available (yet), so _fill should not change and, since nothing is read, the return value of fillBuffer should be zero.

This most likely solves the complete #2212 issue, at least it solves the minimum sketch reproducing the error as posted in that issue by @lbernstone. This could also be the root cause of other issues, since on slow connections it is likely that the socket not always has data available.

I also added an extra test on

arduino-esp32/libraries/WiFi/src/WiFiClient.cpp

Line 60 in 229d9b7

_buffer = (uint8_t *)malloc(_size);

by setting _failed to true if the malloc() fails. I am not sure and did not check if the _failed condition is checked in the other functions, but it should not do any harm. Moreover, this did not cause the error.

Happy New Year to everyone!

[Jeroen88]

@lbernstone, @nikuz confirmed that my PR fixes the error. Also, the if() in WiFiClientRxBuffer::fillBuffer() before my PR is obviously wrong...

[lbernstone]

Verified. Looks like you included some BLE stuff in your commit, tho.
Fixes #2212

[Jeroen88]

@lbernstone You are right, I always pull remote master before creating a branch. So I was surpised to have some BLE stuff in my local master

[me-no-dev]

I can not merge this with the BLE stuff in there ;)

[Jeroen88]

@me-no-dev can you help me get it out? I did a git pull upstream master and then it got in...????

[me-no-dev]

delete the BLE folder then call git submodule update --init --recursive in the root arduino folder

[Jeroen88]

Replaced by #2263