3.0.0 version always consumes more Flash space than the 2.X.X versions

[sunnyguhz]

I have tried comparing multiple legacy projects, and ‌without exception‌, the 3.0.0 version always consumes more Flash space than the 2.X.X versions.
All configurations, microcontroller options, and third-party libraries have been kept identical.

Win11 25H2
Arduino 2.3.10
U8G2 2.36.19
Debug symbols and core logs:"None"
Optimize:"Smallest (-OS) with LTO"
Board part number:"Generic G030F6Px"
U(S)ART support:"Enabled (generic ‘Serial’)"
"Optimize for debug" is not checked.

For example, in this project, the Flash footprint after compilation is completely different.

#include <U8g2lib.h>
#include <SPI.h>

U8G2_ST7567_OS12864_F_4W_HW_SPI u8g2(U8G2_MIRROR, /*cs=*/PA4, /* dc=*/PA6);

byte ceshi = 0, showupdate = 1, SWJIAan = 0, SWJIANan = 0, fengshanciji = 2, fengshanshanye = 5;
short fanspeedt = 0, IRfanspeedt = 0, fanspeed = 0, IRfanspeed = 0, AVGdianliu = 0, zuidafanspeed = 0, zuidaIRfanspeed = 0, caiyangcishu = 1, qidongdianliu = 0;
long long AREFINT_CAL;
unsigned long currentMillis = 0, lastSWJIAtime = 0, lastSWJIANtime = 0, SWJIAfalltime = 0, SWJIANfalltime = 0, SWJIArisetime = 0, SWJIANrisetime = 0, lastjisutime = 0, fengshanjisutime = 0, zuidafanIRspeedtime = 0, dianliu = 0;


void setup() {
  analogReadResolution(12);  //设置ADC为12位采样
  Serial.begin(115200);
  delay(100);
  Serial.println("风扇测速器启动中……");
  pinMode(PA11, INPUT_PULLUP);                  //按钮SW-
  pinMode(PA12, INPUT_PULLUP);                  //按钮SW+
  pinMode(PA0, INPUT_ANALOG);                   //PA0读取电流
  pinMode(PA1, INPUT_PULLUP);                   //风扇测试信号线
  pinMode(PB0, INPUT_PULLUP);                   //红外测速信号
  pinMode(PB9, OUTPUT);                         //风扇电源控制
  attachInterrupt(PA11, SWJIA, CHANGE);         //按钮SW-
  attachInterrupt(PA12, SWJIAN, CHANGE);        //按钮SW+
  attachInterrupt(PA1, fanspeedtest, FALLING);  //风扇测试信号线、下降沿触发方式
  attachInterrupt(PB0, IRtestF, FALLING);       //红外测速信号、下降沿触发方式

  SPI.setSCLK(PA5);
  SPI.setMOSI(PA7);
  SPI.setMISO(PB4);            //STM32G030的硬件SPI有bug，MISO即使没有用也不能用于其他功能。默认MISO PA6如被用作其他作用要在U8g2初始化之前把MISO改为暂时不用的PB4（SPI1的MISO可以使PA6、PA11、PB4）
  u8g2.setBusClock(32000000);  //SPI默认速度是4M，G0系列全系最高到32M。
  u8g2.begin();
  u8g2.sendF("c", 0x21);  //屏幕开启升压，并设置升压目标。
  u8g2.setContrast(255);  //设置LCD屏幕对比度,3.1寸屏幕150时最清晰,1.8寸1最清楚，2.4寸45？最清楚，1.2寸建议设为20，9毛钱的1.8寸（不是M2）建议为0x24+70或者5，9毛钱的1.8寸（M2）建议为0x24+40,1.7黑色反显屏23/110,22/160/，21/255是比较合适的参数设置。
  u8g2.setFlipMode(1);    //设置LCD屏幕显示旋转180度
  u8g2.setPowerSave(0);
  u8g2.enableUTF8Print();  // enable UTF8 support for the Arduino print() function
  u8g2.setFontDirection(0);
  u8g2.setFont(u8g2_font_6x13_tr);
  u8g2.clearBuffer();
  u8g2.drawUTF8(0, 26, " FanSpeed test");
  u8g2.drawUTF8(0, 40, " S1.2");
  u8g2.drawUTF8(60, 54, "2026.08.11");
  u8g2.sendBuffer();
  u8g2.clearBuffer();
  delay(2000);
  AREFINT_CAL = *(__IO uint16_t*)(0X1FFF75AA);  //基准电压出厂校准值，校准温度30±5℃，电压3.0V±10mV，STM32G030系列通用地址，其他系列需要更换对应地址。
}

void loop() {
  currentMillis = millis();
  //SW+，短按增加风扇叶片，长按开始测试风扇转速。
  if (currentMillis - SWJIArisetime > 50 && SWJIArisetime != 0) {  //按钮弹起之后超过50毫秒无动作（50毫秒内的任何再次动作都判断为按键抖动）判断为一次短按键成立，并清空按钮弹起时间
    SWJIAan++;
    SWJIArisetime = 0;
  } else if (currentMillis - SWJIAfalltime > 200 && SWJIAfalltime != 0) {  //按钮按下200毫秒后判断单击、双击、长按，并清空按下时间
    if (SWJIAan && SWJIAan < 10) {                                         //单击
      SWJIAfalltime = 0;
      SWJIAan = 0;
      if (fengshanshanye < 50) {
        fengshanshanye++;
        showupdate = 1;
      }
    } else if (SWJIAan > 9) {  //点击超过9次判断为错误，清除
      SWJIAfalltime = 0;
      SWJIAan = 0;
      //Serial.println("多击");
    } else if (currentMillis - SWJIAfalltime > 2000) {  //按钮初次按下后超过2000毫秒没有松开按钮判断为长按成立，清空首次按下时间并执行长按功能程序。
      SWJIAfalltime = 0;
      SWJIAan = 10;
      //Serial.println("长按");
      fanspeedt = 0;
      IRfanspeedt = 0;
      fanspeed = 0;
      IRfanspeed = 0;
      dianliu = 0;
      zuidafanspeed = 0;
      zuidaIRfanspeed = 0;
      AVGdianliu = 0;
      qidongdianliu = 0;
      caiyangcishu = 1;
      showupdate = 1;
      ceshi = 1;
      digitalWrite(PB9, 1);  //开启风扇电源
      fengshanjisutime = millis();
      lastjisutime = millis();
      zuidafanIRspeedtime = millis();
    }
  }
  //SW-，短按减少风扇叶片，长按清空测试数据。
  if (currentMillis - SWJIANrisetime > 50 && SWJIANrisetime != 0) {  //按钮弹起之后超过50毫秒无动作（50毫秒内的任何再次动作都判断为按键抖动）判断为一次短按键成立，并清空按钮弹起时间
    SWJIANan++;
    SWJIANrisetime = 0;
  } else if (currentMillis - SWJIANfalltime > 200 && SWJIANfalltime != 0) {  //按钮按下200毫秒后判断单击、双击、长按，并清空按下时间
    if (SWJIANan && SWJIANan < 10) {                                         //单击
      SWJIANfalltime = 0;
      SWJIANan = 0;
      if (fengshanshanye > 2) {
        fengshanshanye--;
        showupdate = 1;
      } else if (fengshanshanye == 2) {
        fengshanshanye = 0;
        showupdate = 1;
      }
    } else if (SWJIANan > 9) {  //点击超过9次判断为错误，清除
      SWJIANfalltime = 0;
      SWJIANan = 0;
      //Serial.println("多击");
    } else if (currentMillis - SWJIANfalltime > 2000) {  //按钮初次按下后超过2000毫秒没有松开按钮判断为长按成立，清空首次按下时间并执行长按功能程序。
      SWJIANfalltime = 0;
      SWJIANan = 10;
      //Serial.println("长按");
      fengshanciji = 2;
      fengshanshanye = 5;
      fanspeedt = 0;
      IRfanspeedt = 0;
      fanspeed = 0;
      IRfanspeed = 0;
      dianliu = 0;
      zuidafanspeed = 0;
      zuidaIRfanspeed = 0;
      AVGdianliu = 0;
      qidongdianliu = 0;
      zuidafanIRspeedtime = 0;
      caiyangcishu = 1;
      showupdate = 1;
    }
  }
  if (ceshi) {
    currentMillis = millis();
    if (currentMillis - lastjisutime > 999) {
      fanspeed = fanspeedt;
      if (fengshanshanye) {
        IRfanspeed = IRfanspeedt;
      } else {  //透明扇叶风扇专用
        IRfanspeed = fanspeed;
      }
      fanspeedt = 0;
      IRfanspeedt = 0;
      lastjisutime = millis();
      if (fanspeed > zuidafanspeed) {
        zuidafanspeed = fanspeed;
      }
      if (IRfanspeed > zuidaIRfanspeed) {
        zuidaIRfanspeed = IRfanspeed;
        zuidafanIRspeedtime = millis();
      }
      if (fengshanshanye) {
        fengshanciji = fanspeed / (IRfanspeed / fengshanshanye);
      }
    }
    if (currentMillis - fengshanjisutime > 10000) {
      dianliu += analogRead(PA0);
      caiyangcishu++;
      AVGdianliu = dianliu / caiyangcishu;
      delay(10);
    } else {
      zuidafanIRspeedtime = millis();
      dianliu = analogRead(PA0);
      if (dianliu > qidongdianliu) {
        qidongdianliu = dianliu;
      }
    }
    currentMillis = millis();
    if (currentMillis - zuidafanIRspeedtime > 8000) {
      digitalWrite(PB9, 0);  //关闭风扇电源
      ceshi = 0;
    }
    showupdate = 1;
  }
  if (showupdate) {
    ALLshowupdate();
    showupdate = 0;
  }
}

void ALLshowupdate() {
  String LCDshowinfo;
  long lingshidianliu;
  if (ceshi) {
    LCDshowinfo = "Testing";
  } else {
    LCDshowinfo = "STOP";
  }
  long long Aneibujizhun = analogRead(AVREF);  //读取内部电压基准当前读数
  u8g2.drawUTF8(0, 10, LCDshowinfo.c_str());   //测试状态
  if(fengshanshanye){
    LCDshowinfo += 60 * IRfanspeed / fengshanshanye;
  }else{
    LCDshowinfo += 60 * IRfanspeed / fengshanciji;
  }
  LCDshowinfo += "RPM";
  u8g2.drawUTF8(63, 10, LCDshowinfo.c_str());  //实时转速
  if (currentMillis - fengshanjisutime > 10000) {
    lingshidianliu = 4000 * AREFINT_CAL * AVGdianliu / (12831 * Aneibujizhun);  //4个0.47欧电阻并联，INA180A1放大20倍，3.0V最大测量电流1.25A,3.3V时1.375A
  } else {
    lingshidianliu = 4000 * AREFINT_CAL * dianliu / (12831 * Aneibujizhun);  //4个0.47欧电阻并联，INA180A1放大20倍，3.0V最大测量电流1.25A,3.3V时1.375A；
  }
  LCDshowinfo += lingshidianliu / 1000;
  LCDshowinfo += ".";
  lingshidianliu = lingshidianliu % 1000;
  if (lingshidianliu < 100) {
    if (lingshidianliu < 10) {
      LCDshowinfo += "00";
    } else {
      LCDshowinfo += "0";
    }
  }
  LCDshowinfo += lingshidianliu;
  LCDshowinfo += "A ";
  u8g2.drawUTF8(0, 23, LCDshowinfo.c_str());  //当前电流
  lingshidianliu = 12 * lingshidianliu;
  LCDshowinfo += lingshidianliu / 1000;
  LCDshowinfo += ".";
  lingshidianliu = lingshidianliu % 1000;
  if (lingshidianliu < 100) {
    if (lingshidianliu < 10) {
      LCDshowinfo += "00";
    } else {
      LCDshowinfo += "0";
    }
  }
  LCDshowinfo += lingshidianliu;
  LCDshowinfo += "W";
  u8g2.drawUTF8(63, 23, LCDshowinfo.c_str());  //当前功率
  LCDshowinfo = "MAX: ";
  u8g2.drawUTF8(0, 36, LCDshowinfo.c_str());            //最大转速
  LCDshowinfo += 60 * zuidaIRfanspeed / fengshanshanye;  //最大转速
  LCDshowinfo += "RPM ";
  u8g2.drawUTF8(63, 36, LCDshowinfo.c_str());  //最大转速
  LCDshowinfo = "AVG:";
  lingshidianliu = 4000 * AREFINT_CAL * AVGdianliu / (12831 * Aneibujizhun);  //4个0.47欧电阻并联，INA180A1放大20倍，3.0V最大测量电流1.25A,3.3V时1.375A
  LCDshowinfo += lingshidianliu / 1000;
  LCDshowinfo += ".";
  lingshidianliu = lingshidianliu % 1000;
  if (lingshidianliu < 100) {
    if (lingshidianliu < 10) {
      LCDshowinfo += "00";
    } else {
      LCDshowinfo += "0";
    }
  }
  LCDshowinfo += lingshidianliu;
  LCDshowinfo += "A";
  u8g2.drawUTF8(0, 49, LCDshowinfo.c_str());  //最大电流
  LCDshowinfo = "MAX:";
  lingshidianliu = 4000 * AREFINT_CAL * qidongdianliu / (12831 * Aneibujizhun);  //4个0.47欧电阻并联，INA180A1放大20倍，3.0V最大测量电流1.25A,3.3V时1.375A
  LCDshowinfo += lingshidianliu / 1000;
  LCDshowinfo += ".";
  lingshidianliu = lingshidianliu % 1000;
  if (lingshidianliu < 100) {
    if (lingshidianliu < 10) {
      LCDshowinfo += "00";
    } else {
      LCDshowinfo += "0";
    }
  }
  LCDshowinfo += lingshidianliu;
  LCDshowinfo += "A ";
  u8g2.drawUTF8(63, 49, LCDshowinfo.c_str());  //启动电流
  LCDshowinfo = "SN:";
  LCDshowinfo += fengshanciji;
  u8g2.drawUTF8(0, 62, LCDshowinfo.c_str());  //磁极
  LCDshowinfo = "IR:";
  LCDshowinfo += fengshanshanye;
  u8g2.drawUTF8(63, 62, LCDshowinfo.c_str());  //扇叶数量
  u8g2.sendBuffer();
  u8g2.clearBuffer();
}

void SWJIA() {  //PA11，按钮SW-
  currentMillis = millis();
  if (currentMillis - lastSWJIAtime > 100) {
    if (SWJIAan == 10) {
      lastSWJIAtime = millis();
      SWJIAan = 0;
    } else if (digitalRead(PA11)) {  //按钮松开（RISING）时记录此按钮弹起时间，并更新背光灯亮起时间；
      if (SWJIAfalltime != 0) {
        SWJIArisetime = millis();
      }
    } else {                     //按钮按下（FALLING）时记录记录首次按下时间(按下时间不为0时不更新）并清空按钮弹起时间，按下时背光灯亮起并记录背光灯亮起时间
      if (SWJIAfalltime == 0) {  //如果按下时间为0则记录按下的时间
        SWJIAfalltime = millis();
      }
      SWJIArisetime = 0;  //清空按钮弹起时间
    }
  }
}

void SWJIAN() {  //PA12，按钮SW-
  currentMillis = millis();
  if (currentMillis - lastSWJIANtime > 100) {
    if (SWJIANan == 10) {
      lastSWJIANtime = millis();
      SWJIANan = 0;
    } else if (digitalRead(PA12)) {  //按钮松开（RISING）时记录此按钮弹起时间，并更新背光灯亮起时间；
      if (SWJIANfalltime != 0) {
        SWJIANrisetime = millis();
      }
    } else {                      //按钮按下（FALLING）时记录记录首次按下时间(按下时间不为0时不更新）并清空按钮弹起时间，按下时背光灯亮起并记录背光灯亮起时间
      if (SWJIANfalltime == 0) {  //如果按下时间为0则记录按下的时间
        SWJIANfalltime = millis();
      }
      SWJIANrisetime = 0;  //清空按钮弹起时间
    }
  }
}

void fanspeedtest() {  //PA1，风扇转速检测。
  fanspeedt++;
}

void IRtestF() {  //PB0，红外转速检测。
  IRfanspeedt++;
}

STM32 CORE 2.6.0
Sketch uses 31588 bytes (96%) of program storage space. Maximum is 32768 bytes.
Global variables use 3132 bytes (38%) of dynamic memory, leaving 5060 bytes for local variables. Maximum is 8192 bytes.

STM32 CORE 2.11.0
Sketch uses 31620 bytes (96%) of program storage space. Maximum is 32768 bytes.
Global variables use 3084 bytes (37%) of dynamic memory, leaving 5108 bytes for local variables. Maximum is 8192 bytes.

STM32 CORE 2.12.0
Sketch uses 31780 bytes (96%) of program storage space. Maximum is 32768 bytes.
Global variables use 3084 bytes (37%) of dynamic memory, leaving 5108 bytes for local variables. Maximum is 8192 bytes.

STM32 CORE 3.0.0
region `FLASH' overflowed by 3884 bytes

[fpistm]

Hi @sunnyguhz

Deployment of the ArduinoCore-API is the main responsible of this.
Here the analysis made using GH Copilot:

ArduinoCore-API is not merely renamed headers. It changes the concrete C++ types and method contracts that the core, Wire, SPI, Serial, and user libraries compile against.

The main cost comes from:

• New object model and virtual interface: Print, Stream, String, Uart, SPIClass, and TwoWire now conform to shared ArduinoCore-API contracts. That retains vtables, virtual dispatch entries, destructors, and compatibility methods where the old STM32-specific implementation could be simpler.

• Broader API signatures: The migration changes values such as pin identifiers and lengths to portable types like pin_size_t and size_t. On Cortex-M0+, that can require conversions, wider arithmetic, and distinct overloads. It also prevents some old call paths from folding exactly the same way under LTO.

• API compatibility surface: The new implementations need to support standard ArduinoCore-API overloads, including multiple Wire.begin, requestFrom, endTransmission, SPI.begin, SPISettings, and serial methods. Even if a sketch calls one overload, virtual/interface reachability and internal delegation can retain more code.

• Less target-specific optimization: 2.12.0 types and APIs were STM32-only, so the compiler saw more local assumptions. ArduinoCore-API requires behavior shared across Arduino platforms, preserving generic paths for pin handling, bus mode, callbacks, and configuration.

• Cascading recompile effect: With LTO, a type/signature change alters inlining, devirtualization, and constant propagation across the whole program. A small API adjustment can change generated code in U8g2, Wire, SPI, Serial, interrupts, and startup, not just in the migrated files.

Your symbols reflect this: the new image retains API-driven Wire/SPI/Uart functions totaling several KB, while the old image retained HardwareSerial, legacy String, Stream, and Print forms. The net code increase is not one large library; it is many changed call boundaries that no longer optimize into the same compact machine code.

The migration’s benefit is cross-core API consistency and easier library portability. The size cost is the price of preserving that common abstraction on a very small Cortex-M0+ target.