Fix for timeout issues on WebServer.cpp

[supersjellie]

Description of Change

Removing client temporary variable. Arduino doesn't like copying objects. And for the logisch de class variable _currentClient can be used from the start of the method

Tests scenarios

My openHAB server calls 3 ESP32 devices every minute. In the openHAB log about 5-10% fails on a timeout (3000 ms is set timeout on openHAB). With above code modification none (running for 3 days now) of the calls failed.

Related links

#8318

(same issue and resolution for ESP8266, check esp8266/Arduino#8941 )

[CLAassistant]

All committers have signed the CLA.

[SuGlider]

@supersjellie @me-no-dev

I have tested this PR and it works with no errors.
I used the WebServer HelloServer.ino example to test server.handleClient();
I couldn't notice any difference from the original code to the PR code.
I get a few errors in the WiFi Client in both cases:

[ 16123][D][WiFiClient.cpp:546] connected(): Disconnected: RES: 0, ERR: 128
[ 26460][D][WiFiClient.cpp:546] connected(): Disconnected: RES: 0, ERR: 128
[ 26463][D][WiFiClient.cpp:546] connected(): Disconnected: RES: 0, ERR: 128
[ 36087][D][WiFiClient.cpp:546] connected(): Disconnected: RES: 0, ERR: 128
[ 46242][D][WiFiClient.cpp:546] connected(): Disconnected: RES: 0, ERR: 128
[ 56364][D][WiFiClient.cpp:546] connected(): Disconnected: RES: 0, ERR: 128
[ 66167][D][WiFiClient.cpp:546] connected(): Disconnected: RES: 0, ERR: 128

[supersjellie]

I couldn't notice any difference from the original code to the PR code.

No differences in response time when your refresh 'like crazy'? In my setup (of course I use a bigger/more complicated sketch than the helloserver.ino) the difference is large. Most requests are the same but with the old code about 5-10% had large response times. And with the new code this disappeared.

@SuGlider @me-no-dev : Got curious to the why and performed an extra test and now think the ESP32 timeouts were collateral damage of the ESP8266 bug.

Setup:

1. Change refresh rate of 1 ESP32 webservice in openHAB to 10 seconds
2. New code, no timeouts in 1 hour
3. Old code, no timeouts in 1 hour
   Same as you observed. Difference with previous testsetup is that all ESP8266 devices are "patched" now. They don't timeout anymore. So ... maybe the timeouts on the ESP32 devices are collateral damage caused by the timeouts on the ESP8266. Also no network expert but thinking of some general issues in the "webservice traffic" (maybe by blocking all connections in the pool used by the openHAB server).

What would mean that for the ESP32 this code modification is more cosmetic, saving an unneeded variable and keeping it in synch with the ESP8266 code.

[SuGlider]

What would mean that for the ESP32 this code modification is more cosmetic, saving an unneeded variable and keeping it in synch with the ESP8266 code.

Yes, this PR may save a few cycles and some stack allocation in execution time.
@me-no-dev - It is an improvement, anyway, therefore, I think that we should merge it.