Keep track of my fixes

[s330chiplenovo]

*** Warning: I have not done a detail regression testing after my fixes, so it potentially could break some other CP/M functions

*** My test is on a M5Stack Cardputer which has the ESP32-S3FN8 ***

I made some fixes to let the emulator runs better on the Cardputer and would like to use this discussion thread to keep track of my changes.

(1) Accidental mask off the high nibble and missing block {...} structure in the if-construct

Sympton: It all started when I try to find a usable program languages. After failing all the C compilers I can find on the web. I switch to the CP/M comes with ASM.COM. However, I found the output .HEX file produced by ASM.COM is corrupted. I did some investigation on the bad .HEX file against a good .HEX file using the same physical ASM.COM program run on RunCPM emulator, I found all the data portion of the .HEX file is converted to the output character's actual value rather than the ASCII code. (i.e. the correct .HEX file should contain the ASCII value of '0'-'9', 'A'-'F' on the data section.)

Here is the Bad .HEX file content

M>DUMP HELLO.HEX

0000 3A 01 00 00 01 00 00 00 00 00 0E 00 09 01 01 00
0010 09 00 01 0C 0D 00 05 00 00 0C 09 04 08 04 05 04
0020 0C 04 0C 04 0F 02 00 05 07 03 07 0D 0A 3A 00 06
0030 00 01 01 00 00 00 04 0F 05 02 04 0C 04 04 02 01
0040 02 04 07 03 0D 0A 3A 00 00 00 01 00 00 00 00 0F
0050 0F 0D 0A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A
0060 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A
0070 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A

Here is the good one

M>DUMP HELLO.HEX

0000 3A 31 30 30 31 30 30 30 30 30 45 30 39 31 31 30
0010 39 30 31 43 44 30 35 30 30 43 39 34 38 34 35 34
0020 43 34 43 34 46 32 30 35 37 33 37 0D 0A 3A 30 36
0030 30 31 31 30 30 30 34 46 35 32 34 43 34 34 32 31
0040 32 34 37 33 0D 0A 3A 30 30 30 31 30 30 30 30 46
0050 46 0D 0A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A
0060 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A
0070 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A 1A

M>TYPE HELLO.HEX
:100100000E09110901CD0500C948454C4C4F205737
:060110004F524C44212473
:00010000FF

Compare the second line

Bad:
0010 09 00 01 0C 0D 00 05 00 00 0C 09 04 08 04 05 04

Good:
0010 39 30 31 43 44 30 35 30 30 43 39 34 38 34 35 34

Notice that byte 11H on the bad file is 9, on the good file is 39 (i.e. the ASCII code for 9 is 39)

So I check the cpu.h file to check for mask off the high nibble's code. (i.e. binary AND &0x0f or divided by 16)

Findings:

I am able to find the high nibble masking code. To my surprise, I found some other bugs too. A couple of if constructs missing the begin and end block i.e. {...} and one typo as well.

Here are the codes:

Fix 1: file cpu.h Line 1597-1617 LDIR emulation

Here src and dst are 16 bit integer, the code tries to split a 16 bit integer into two 8 bit integers; but the logic here is for splitting a 8 bit integer into two 4 bit (nibble) integers.

Original code:

    case 0xB0:    //** LDIR **
      uint16_t src, dst;
      V16 = (256 * B) + C;  //Byte count
      src = (256 * H) + L;  //Source
      dst = (256 * D) + E;  //Destination
      do {
        RAM[dst] = RAM[src];  //copy byte
        dst++;
        src++;
        V16--;
      } while (V16 > 0);
      H = src / 16;
      L = src & 0x0f;
      D = dst / 16;
      E = dst & 0x0f;
      B = 0;
      C = 0;
      Nf = false;
      Pf = true;
      Hf = false;
      break;

Modified code:

    case 0xB0:    //** LDIR **
      uint16_t src, dst;
      V16 = (256 * B) + C;  //Byte count
      src = (256 * H) + L;  //Source
      dst = (256 * D) + E;  //Destination
      do {
        RAM[dst] = RAM[src];  //copy byte
        dst++;
        src++;
        V16--;
      } while (V16 > 0);
      H = src / 256;     // Chip 2026-06-12 restore the upper 8 bits
      L = src & 0xff;    // Chip 2026-06-12 keep lower 8 bits
      D = dst / 256;     // Chip 2026-06-12 restore the upper 8 bits
      E = dst & 0xff;    // Chip 2026-06-12 keep lower 8 bits
      B = 0;
      C = 0;
      Nf = false;
      Pf = true;
      Hf = false;
      break;

Fix 2: file cpu.h Line 1597-1617 CPIR emulation

typo and bug on splitting the 16 bit integer into the higher 8 bit integer

Original code:

    case 0xB1:    //** CPIR **
      cfs = Cf;                 //Save Carry flag
      V32 = (256 * B) + C;      //Byte count
      V16 = (256 * H) + L;      //Source
      Zf == false;
      while (V32 > 0 && Zf == false) {
        SUB8(A, RAM[V16], 0) ;  //compare RAM to A
        V16++;
        V32--;
      }
      H = V16 / 16;             //Update HL
      L = V16 & 255;
      B = V32 / 256;            //Update BC
      C = V32 & 255;
      Cf = cfs;                 //Restore carry flag
      if (V32 != 0) Pf = true; else Pf = false;
      break;

Modified code:

    case 0xB1:    //** CPIR **
      cfs = Cf;                 //Save Carry flag
      V32 = (256 * B) + C;      //Byte count
      V16 = (256 * H) + L;      //Source
      Zf = false;               // Chip 2026-06-12 typo fix
      while (V32 > 0 && Zf == false) {
        SUB8(A, RAM[V16], 0) ;  //compare RAM to A
        V16++;
        V32--;
      }
      H = V16 / 256;            //Update HL  // Chip 2026-06-12 restore the upper 8 bits
      L = V16 & 255;
      B = V32 / 256;            //Update BC
      C = V32 & 255;
      Cf = cfs;                 //Restore carry flag
      if (V32 != 0) Pf = true; else Pf = false;
      break;

Fix 3: cpu.h Line 306-329

Missing begin and end block {...} in the if construct

Original code:

        case 0x27:                  //** DAA **
          uint8_t ua, la;
          la = A & 0x0f;          //lower nibble of A
          ua = A / 16 ;           //Upper nibble of A
          bool br;
          br = false;
          if (br == false && Cf == 0 && ua < 10 && Hf == 0 && la < 10) Cf = 0;            br = true;  //1
          if (br == false && Cf == 0 && ua <  9 && Hf == 0 && la >  9) Cf = 0; A += 0x06; br = true;  //2
          if (br == false && Cf == 0 && ua < 10 && Hf == 1 && la <  4) Cf = 0; A += 0x06; br = true;  //3
          if (br == false && Cf == 0 && ua >  9 && Hf == 0 && la < 10) Cf = 1; A += 0x60; br = true;  //4
          if (br == false && Cf == 0 && ua >  8 && Hf == 0 && la >  9) Cf = 1; A += 0x66; br = true;  //5
          if (br == false && Cf == 0 && ua >  9 && Hf == 1 && la <  4) Cf = 1; A += 0x66; br = true;  //6
          if (br == false && Cf == 1 && ua <  3 && Hf == 0 && la < 10) Cf = 1; A += 0x60; br = true;  //7
          if (br == false && Cf == 1 && ua <  3 && Hf == 0 && la >  9) Cf = 1; A += 0x66; br = true;  //8
          if (br == false && Cf == 1 && ua <  4 && Hf == 1 && la <  4) Cf = 1; A += 0x66; br = true;  //9
          if (br == false && Cf == 0 && ua < 10 && Hf == 0 && la < 10) Cf = 0;            br = true;  //10
          if (br == false && Cf == 0 && ua <  9 && Hf == 1 && la >  5) Cf = 0; A += 0xFA; br = true;  //11
          if (br == false && Cf == 1 && ua >  6 && Hf == 0 && la < 10) Cf = 1; A += 0xA0; br = true;  //12
          if (br == false && Cf == 1 && ua >  5 && ua < 8 && Hf == 1 && la >  5) Cf = 1; A += 0x9A;  //13

          if (A == 0) Zf = true; else Zf = false;
          calcP(A);                  //update parity flag
          Sf = bitRead(A, 7);       //Update S flag
          break;

Modified code:

        case 0x27:                  //** DAA **
          uint8_t ua, la;
          la = A & 0x0f;          //lower nibble of A
          ua = A / 16 ;           //Upper nibble of A
          bool br;
          br = false;
          if (br == false && Cf == 0 && ua < 10 && Hf == 0 && la < 10) { Cf = 0;            br = true; } //1 Chip 2026-06-12
          if (br == false && Cf == 0 && ua <  9 && Hf == 0 && la >  9) { Cf = 0; A += 0x06; br = true; } //2 Chip 2026-06-12
          if (br == false && Cf == 0 && ua < 10 && Hf == 1 && la <  4) { Cf = 0; A += 0x06; br = true; } //3 Chip 2026-06-12
          if (br == false && Cf == 0 && ua >  9 && Hf == 0 && la < 10) { Cf = 1; A += 0x60; br = true; } //4 Chip 2026-06-12
          if (br == false && Cf == 0 && ua >  8 && Hf == 0 && la >  9) { Cf = 1; A += 0x66; br = true; } //5 Chip 2026-06-12
          if (br == false && Cf == 0 && ua >  9 && Hf == 1 && la <  4) { Cf = 1; A += 0x66; br = true; } //6 Chip 2026-06-12
          if (br == false && Cf == 1 && ua <  3 && Hf == 0 && la < 10) { Cf = 1; A += 0x60; br = true; } //7 Chip 2026-06-12
          if (br == false && Cf == 1 && ua <  3 && Hf == 0 && la >  9) { Cf = 1; A += 0x66; br = true; } //8 Chip 2026-06-12
          if (br == false && Cf == 1 && ua <  4 && Hf == 1 && la <  4) { Cf = 1; A += 0x66; br = true; } //9 Chip 2026-06-12
          if (br == false && Cf == 0 && ua < 10 && Hf == 0 && la < 10) { Cf = 0;            br = true; } //10 Chip 2026-06-12
          if (br == false && Cf == 0 && ua <  9 && Hf == 1 && la >  5) { Cf = 0; A += 0xFA; br = true; } //11 Chip 2026-06-12
          if (br == false && Cf == 1 && ua >  6 && Hf == 0 && la < 10) { Cf = 1; A += 0xA0; br = true; } //12 Chip 2026-06-12
          if (br == false && Cf == 1 && ua >  5 && ua < 8 && Hf == 1 && la >  5) { Cf = 1; A += 0x9A; } //13 Chip 2026-06-12

          if (A == 0) Zf = true; else Zf = false;
          calcP(A);                  //update parity flag
          Sf = bitRead(A, 7);       //Update S flag
          break;

With this fix, I have ASM.COM working correctly now.

Rgds,
Chip Ling

[s330chiplenovo]

(2) Output buffer overflow

Sympton: It comes to my attention that when the output reaches a certain amount, the output to the console screen will be chopped off. It happens when your program prints out massive output. Use the TYPE command to type out a long program. Use the DIR command to list the files (if you have a lot of file in that disk)...

Findings: I found the output buffer is defined in globals.h as a 1024 bytes array. However, there is no logic to handle it's output when output is larger than 1024 bytes. (i.e. input to the buffer stop and wait for the output to screen finish before resuming input to the buffer)

globals.h

uint8_t txBuf[1024];            //Serial transmit buffer

** Fix 1 — cpu.h:2074-2079 UART_PORT: **

Original code:

    case UART_PORT:                           //UART Write
      txBuf[txInPtr] = v;                     //Write char to output buffer
      txInPtr++;
      if (txInPtr == sizeof(txBuf)) txInPtr = 0;
      bitWrite(pIn[UART_LSR], 6, 1);          //Set bit to indicate sent
      break;

Fixed code:

    case UART_PORT:                           //UART Write  // Chip 2026-06-14
      { // UART_PORT: flow-controlled write to TX circular buffer
        uint16_t next = txInPtr + 1;
        if (next == sizeof(txBuf)) next = 0;
        while (next == txOutPtr) vTaskDelay(1);  // Buffer full - yield to serial task
        txBuf[txInPtr] = v;                     //Write char to output buffer
        txInPtr = next;
        bitWrite(pIn[UART_LSR], 6, 1);          //Set bit to indicate sent
      }
      break;

** Fix 2 — serial.h:155-163 outString() **

Original code:

void outString(char buf[]) {
    int i = 0;
    while (buf[i] > 0) {
        txBuf[txInPtr] = buf[i];
        txInPtr++;
        i++;
        if (txInPtr == sizeof(txBuf)) txInPtr = 0;
    }
}

Fixed code:

void outString(char buf[]) {
    int i = 0;
    while (buf[i] > 0) {
        uint16_t next = txInPtr + 1;
        if (next == sizeof(txBuf)) next = 0;
        while (next == txOutPtr) vTaskDelay(1);  // Buffer full - yield to serial task
        txBuf[txInPtr] = buf[i];
        txInPtr = next;
        i++;
    }
}

Rgds,
Chip Ling

[s330chiplenovo]

(3) Restore BDOS and/or BIOS memory from disk when transient program overwrite those location in memory.

Sympton: DDT.COM does not exit back to CP/M via command G0 nor Ctrl+C

I cannot find any logic in the emulator to handle the case when a transient program overwrite the BDOS/BIOS memory location, and upon exiting the program, restore back the BDOS/BIOS to memory.

Fix 1: cpu.h line 23 block insert code after Line 45

Original code:

void CPUTask( void * parameter ) {
  ...
  cpu_t = true;

  while (1) {
    vTaskDelay(50);
  ...

Modified code:

void CPUTask( void * parameter ) {
  ...
  cpu_t = true;

  // Chip 2026-06-15 begin
  const uint16_t cpmBdosBase = 0xD000;
  const uint16_t cpmBdosSize = 0x1600;  // D000-E5FF
  const uint16_t cpmBiosBase = 0xE600;
  const uint16_t cpmBiosSize = 0x1A00;  // E600-FFFF
  static bool cpmImageCaptured = false;
  static uint16_t lastPC = 0xFFFF;
  static uint8_t bdosImage[cpmBdosSize];
  static uint8_t biosImage[cpmBiosSize];
  // Chip 2026-06-15 end

  while (1) {
    vTaskDelay(50);
  ...

Fix 2: cpu.h line 1491

Original code:

        default:                    //NOP or anything not recognised
          Serial.printf("Unknown OP-Code %.2X at %.4X\n\r", OC, PC - 1);
          break;
      }
      if (SingleStep == true) RUN = false;
      if (bpOn == true && PC == BP) RUN = false;
    }

Modified code:

        default:                    //NOP or anything not recognised
          Serial.printf("Unknown OP-Code %.2X at %.4X\n\r", OC, PC - 1);
          break;
      }
      // Chip 2026-06-15
      // Capture pristine BDOS/BIOS once CP/M is running.
      if (!cpmImageCaptured
          && RAM[0x0000] == 0xC3
          && RAM[0x0005] == 0xC3
          && PC >= cpmBdosBase) {
        for (uint16_t i = 0; i < cpmBdosSize; i++) bdosImage[i] = RAM[cpmBdosBase + i];
        for (uint16_t i = 0; i < cpmBiosSize; i++) biosImage[i] = RAM[cpmBiosBase + i];
        cpmImageCaptured = true;
        Serial.println("Captured CP/M BDOS/BIOS baseline image");
      }

      // On warm-boot entry (JP 0000), restore only changed BDOS/BIOS bytes.
      if (cpmImageCaptured && PC == 0x0000 && lastPC != 0x0000) {
        uint16_t bdosChanged = 0;
        uint16_t biosChanged = 0;
        for (uint16_t i = 0; i < cpmBdosSize; i++) {
          if (RAM[cpmBdosBase + i] != bdosImage[i]) {
            RAM[cpmBdosBase + i] = bdosImage[i];
            bdosChanged++;
          }
        }
        for (uint16_t i = 0; i < cpmBiosSize; i++) {
          if (RAM[cpmBiosBase + i] != biosImage[i]) {
            RAM[cpmBiosBase + i] = biosImage[i];
            biosChanged++;
          }
        }
        if (bdosChanged || biosChanged) {
          Serial.printf("Restored CP/M image: BDOS=%u BIOS=%u bytes\n\r", bdosChanged, biosChanged);
        }
      }

      lastPC = PC;

      if (SingleStep == true) RUN = false;
      if (bpOn == true && PC == BP) RUN = false;
    }

It fixes the DDT.COM exit problem. Using command G0 or Ctrl+C can return to CP/M normally and stay on the current drive.

M>DDT
DDT VERS 2.2
-G0

M>

M>DDT
DDT VERS 2.2
-^C
M>

Note 1: It also fixes the issue when exiting from MBASIC using the SYSTEM command. Before the fix, it will perform a COLD BOOT and will return to drive A instead of the current drive.

M>MBASIC
BASIC-85 Rev. 5.29
[CP/M Version]
Copyright 1985-1986  ▒  by Microsoft
Created: 28-Jul-85
29752 Bytes free
Ok
SYSTEM

M>

Note 2: the Software Toolworks C80 3.0a starts to work after this fix. I finally have a C compiler to work with.

C>type hello.c
#include "printf.c"

int main() {
  int i;
  printf("Hello from Software Toolworks C80 3.0a\n");
  for (i = 1; i <= 10; i++) {
    printf("value of i is %d\n", i);
  }
  return 0;;
}

C>c -v hello.asm=hello.c

C/80 Compiler 3.0a (12/6/83) - (c) 1983 The Software Toolworks

0 errors in compilation
9K bytes free

C>as hello.com,hello.lst=hello.asm
8080 AS 2.3 (7/15/83)
 0100   First word of program
 0F2A   Last word of program
     0  Assembly errors

C>dir hello.*
C: HELLO    ASM : HELLO    LST : HELLO    COM : HELLO    C

C>hello
Hello from Software Toolworks C80 3.0a
value of i is 1
value of i is 2
value of i is 3
value of i is 4
value of i is 5
value of i is 6
value of i is 7
value of i is 8
value of i is 9
value of i is 10

C>

Note 3: Microsoft BASIC Compiler works too

H>DIR HELLO.*
H: HELLO    BAS

H>TYPE HELLO.BAS
10100 REM HELLO PROGRAM
10200 PRINT "HELLO WORLD!"
10300 FOR I = 1 TO 5
10400   PRINT I; " ";
10500 NEXT I
10600 PRINT ""
10700 END

H>BASCOM HELLO,HELLO=HELLO

00000 Fatal Error(s)
21986 Bytes Free

H>L80 HELLO,HELLO/N/E

Link-80  3.44  09-Dec-81  Copyright (c) 1981 Microsoft

Data    4000    4209    <  521>

37997 Bytes Free
[4015   4209       66]

H>DIR HELLO.*
H: HELLO    PRN : HELLO    BAS : HELLO    REL : HELLO    COM

H>HELLO
HELLO WORLD!
 1   2   3   4   5


H>

One more compiler works...

Rgds,
Chip Ling