Serial Corruption Isolated (Tiny 87, maybe others)

[ahshah]

I've been struggling with this odd problem. Here is the sample code:

#include <EEPROM.h>
#define CAL_ADDR 0x0C
#ifdef __AVR_ATtiny841__
#define CAL_REG OSCCAL0
#else
#define CAL_REG OSCCAL
#endif
void performCalibration() {
  int8_t valEnabled = EEPROM.read(CAL_ADDR);
  int8_t valCalib = EEPROM.read(CAL_ADDR + 1);
  if (valEnabled == 1) {
    CAL_REG = valCalib;
    __asm__("nop\n\t");
    __asm__("nop\n\t");
    __asm__("nop\n\t");
    __asm__("nop\n\t");
  }
}


// the setup routine runs once when you press reset:
void setup() {
  // initialize serial communication at 9600 bits per second:
  performCalibration();
  Serial.begin(9600);
}

// the loop routine runs over and over again forever:
void loop() {
  
  uint8_t sensorValue = 13;
  Serial.print("XXX 1 Received: ");
  Serial.println(sensorValue);
  delay(500);  // delay in between reads for stability
  Serial.printf("YYY 2 Received: %d\n", sensorValue);
  delay(500);  // delay in between reads for stability
}

This code prints like so in the 2.0.0 codeline (on thetip of master):

XXX 1 Received: 13
YReceived: 13
YYY 2 Received: 13
X Received: 13
XXX 1 Received: 13
YReceived: 13
YYY 2 Received: 13
X Received: 13
XXX 1 Received: 13
YReceived: 13
YYY 2 Received: 13

On 1.5.2 it prints like so:

XXX 1 Received: 13
YYY 2 Received: 13
XXX 1 Received: 13
YYY 2 Received: 13
XXX 1 Received: 13
YYY 2 Received: 13
XXX 1 Received: 13
YYY 2 Received: 13

The breaking change occurred here:

commit f083ebe10d36031822428b8c99e2639d0ce42156
Author: Spence Konde <[email protected]>
Date:   Wed Feb 2 07:36:41 2022 -0500

    Refactor serial.

    Split off separate files to lose overhead when only one of a pair of UARTs is used on 1624, 841, and 441. Remove cpi/sbi abuse as noted in the comments which was truly rampant in this file

Checkout to the previous commit (cc0a052) confirms that everything works fine.

I've narrowed down the breakage in f083ebe to how we reference the tx buffer. I've created a commit that 'fixes' this issue:
v2.0.0-devThis-is-the-head-submit-PRs-against-this...ahshah:ATTinyCore:87_Serial_issue

However I cannot explain it besides "compiler optimization woes" or "compiler bug" . I've tried throwing 'volatile' everywhere in Serial0/Hardware CPP + H files, I cannot fix the issue. This has been the only way to get it behaving normally.

Would love to learn what in the world is going on here. I believe this should affect all sorts of chips besides the 87.

[ahshah]

Okay tested some chips really quick, but the issue only seems to occur on the 87 and not the x41 nor the x4

[ahshah]

I've figured out the issue and have created a pull request:
#844