toFloat looses precision, respectively toDouble is missing

[DIRR70]

Basic Infos

• This issue complies with the issue POLICY doc.
• I have read the documentation at readthedocs and the issue is not addressed there.
• I have tested that the issue is present in current master branch (aka latest git).
• I have searched the issue tracker for a similar issue.
• If there is a stack dump, I have decoded it.
• I have filled out all fields below.

Platform

• Hardware: [ESP-12]
• Core Version: [2.5.0]
• Development Env: [Arduino IDE]
• Operating System: [Windows]

Settings in IDE

• Module: [Nodemcu]
• Flash Mode: [dio]
• Flash Size: [4MB]
• lwip Variant: [v1.4]
• Reset Method: [nodemcu]
• Flash Frequency: [40Mhz]
• CPU Frequency: [80Mhz]
• Upload Using: [SERIAL]
• Upload Speed: [921600]

Problem Description

Parsing a string into a double (using toFloat which internally uses atof) lacks of precision. It is fine for 4 Byte float, but not for 8 Byte double. In either case toFloat needs to be extended or toDouble (like in Arduino AVR) needs to be introduced!
If someone had a workaround, I would be very grateful :)

#include <Arduino.h>

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

  float f = 20.0 / 3.0;
  Serial.println (f, 8); // 6.66666651
  String fs = String (f, 8);
  Serial.println (fs); // 6.66666651
  f = fs.toFloat ();
  Serial.println (f, 8); // 6.66666651

  double d = 20.0 / 3.0;
  Serial.println (d, 16); // 6.6666666666666669
  String ds = String (d, 16);
  Serial.println (ds); // 6.6666666666666669
  d = ds.toFloat ();
  Serial.println (d, 16); // 6.6666665077209472 :(
}

void loop() {
}

[DIRR70]

Thank you earlephilhower, that was more than fast :)