Reproducible crashing software serial to tcp bridge

[celsworth]

Hi,

I can fairly reproducibly crash the following code:

#include "Arduino.h"

#include <ESP8266WiFi.h>
#include <SoftwareSerial.h>

#include <WiFiUdp.h>

#include <WiFiClient.h>
#include <ESP8266mDNS.h>
#include <ArduinoOTA.h>


WiFiServer server(2001);
WiFiClient client;

// RX on GPIO14. GPIO12 is ignored for now.
SoftwareSerial swSer(14, 12, false, 128);

void setup()
{
    Serial.begin(9600);

    WiFi.persistent(false);
    WiFi.begin("xx", "xx");
    WiFi.mode(WIFI_STA);

    ArduinoOTA.begin();

    server.begin();

    swSer.begin(9600);
}

unsigned int last_send = 0;
char buf[512];
int buflen = 0;

// this will run fine forever until you connect on the tcp socket and then
// it crashes fairly soon afterwards
void loop()
{
    ArduinoOTA.handle();

    if (server.hasClient())
        client = server.available();

    if (swSer.available())
    {
        char d = swSer.read();
        buf[buflen++] = d;

        Serial.print(d);

    }

    // if buffer is getting full, or its been 50ms since send, empty the buffer
    if (buflen > 0 && (buflen > 500 || millis() - last_send > 50))
    {
        buf[buflen] = '\0';

        if (client.connected())
            client.print(buf);

        Serial.print(buf);

        buflen = 0;
        last_send = millis();
    }   
}

SoftwareSerial is from https://github.com/plerup/espsoftwareserial

Connect the serial output of some other compatible device (I just use another ESP spitting out some debug information in a loop) to GPIO14 and wait :)

As per the comment in there, if you don't connect to the tcp socket at all then it doesn't crash. But once you connect, the crash comes reasonably quickly (can take a bit of time, depending on the serial datarate I think).

Any idea whats going on here? Some sort of interrupt clash?

There is no exception printed or anything, the ESP just hangs and then reboots. I don't even see a wdt message, oddly.

[celsworth]

Some gitter discussion reveals that this probably is because interrupts are disabled too long (or too often, or a combination of both), done by SoftwareSerial.

if I swap out SoftwareSerial for the real Serial then the problems instantly go away because I don't have to disable interrupts for such a long time.

So this is probably unfixable, just a way the ESP8266 is, unless anyone has any suggestions please feel free to close.

Many thanks @Links2004 for the help :)