WiFiClient::flush() can get stuck indefinitely

[tko]

Hardware:

Board: DOIT ESP32 DEVKIT V1
Core Installation version: 1.0.4
IDE name: Platform.io
Flash Frequency: Can't tell (Crystal is 40MHz?)
PSRAM enabled: Can't tell
Upload Speed: 921600
Computer OS: Mac OSX

Description:

After reading the http response line with WiFiClient::readBytesUntil() calling WiFiClient::flush() hangs indefinitely. Doesn't always happen, best guess is it depends on response length. First noticed with 204 No Content response from an influxdb server.

Sketch:

#include <WiFi.h>
#include <WiFiClient.h>

const char ssid[] = "FIXME";
const char pass[] = "FIXME";

void setup()
{
  Serial.begin(115200);

  WiFiClient client;
  WiFi.begin(ssid, pass);
  WiFi.waitForConnectResult();
  client.connect("httpbin.org", 80);
  client.printf("GET /status/200 HTTP/1.0\r\nHost: httpbin.org\r\n\r\n");

  auto res = client.readStringUntil('\n');
  Serial.println(res);

  client.flush();
  Serial.println("Done.");
}

void loop()
{
}

Debug Messages:

[D][WiFiGeneric.cpp:337] _eventCallback(): Event: 0 - WIFI_READY
[D][WiFiGeneric.cpp:337] _eventCallback(): Event: 2 - STA_START
[D][WiFiGeneric.cpp:337] _eventCallback(): Event: 4 - STA_CONNECTED
[D][WiFiGeneric.cpp:337] _eventCallback(): Event: 7 - STA_GOT_IP
[D][WiFiGeneric.cpp:381] _eventCallback(): STA IP: 192.168.1.2, MASK: 255.255.255.0, GW: 192.168.1.1
HTTP/1.1 204 NO CONTENT

[enwi]

Having the same issue. The while loop inside WiFiClient::flush() hangs indefinitely due to recv always returning 0, but the previous call to available() returning a value larger 0. Adding some sort of timeout would definitely help.

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

Line 483 in 3236358

res = recv(fd(), buf, toRead, MSG_DONTWAIT);

[enwi]

Ahh I think I found the culprit. The call to WiFiClient::available() inside WiFiClient::flush() refers to the available data inside WiFiClientRxBuffer which could contain data already received from the socket. In that case the socket can't receive any more data, since it was already received by the RXBuffer thus looping forever waiting for data.

I fixed the code by changing the implementation of WiFiclient::flush() to

void WiFiClient::flush() {
    if(available()){
       while(read() >= 0) {};
    }
}

This way the WiFiClientRxBuffer is in charge of handling the data and receiving from the socket.

[stale]

[STALE_SET] This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 14 days if no further activity occurs. Thank you for your contributions.

[enwi]

I dare you to close the issue 🤣

[stale]

[STALE_CLR] This issue has been removed from the stale queue. Please ensure activity to keep it openin the future.

[Darktemp]

I had the same issue, that would hand PubSubClient::connected() if the connection was lost and it lead to the same spot in WiFiClient::flush(), where available() reported non-zero byte but recv always returned 0. Checking errno anyways shows this:
[E][WiFiClient.cpp:491] flush(): fail on fd 58, errno: 128, "Socket is not connected"
For now, I just changed if(res < 0) { to if(res <= 0) { but I'm not sure if there are valid scenarios, where res==0 is fine.

[VojtechBartoska]

Hello @tko, sorry for late reply. Are you able to test your issue on development version 2.0.3-RC1 to check if this is still valid? You can take a look on Docs where is explained how to choose development release version in Arduino IDE.

[tko]

Now I get

...
[  2617][D][WiFiGeneric.cpp:852] _eventCallback(): Arduino Event: 7 - STA_GOT_IP
[  2620][D][WiFiGeneric.cpp:914] _eventCallback(): STA IP: 192.168.1.2, MASK: 255.255.255.0, GW: 192.168.1.1
HTTP/1.1 200 OK
[  3127][E][WiFiClient.cpp:499] flush(): fail on fd 48, errno: 128, "Socket is not connected"
Done.

I suppose it's better for not hanging anymore though the error seems little out of place.

Adding bit more logging I see before flush() calling available() returns 214, and checking connected() returns 0 and logs additionally

[  3164][D][WiFiClient.cpp:528] connected(): Disconnected: RES: 0, ERR: 128