PRG send causes BS Monitor's X command to fail

[dansanderson]

Test Environment (required)

• Tools Platform: MacOS
• Tools Version: 20220519.21-HEAD_n-d9a2452
• MEGA65 Platform: MEGA65
• ROM Release: 920376
• Core Commit: f555316

Describe the bug
Something about the way m65 sends PRG files confuses BS Monitor's X command. The X command is supposed to exit to BASIC, but if the program was sent via m65/M65Connect, X actually goes somewhere else then re-breaks back into the monitor with PC 002005 (which is a weird PC because that's in BASIC memory).

To Reproduce

1. Transfer any PRG from a PC to the MEGA65. Don't run it or switch modes. m65 -l /dev/cu... somefile.prg
2. On the MEGA65 at the READY. prompt, type MONITOR. BS Monitor opens.
3. Type X, hit Return.

Expected behavior
The X command is supposed to exit the monitor and return to BASIC. It works if no program is loaded, or if a program is loaded from disk/SD card. It doesn't matter if the monitor was invoked by the MONITOR command or a brk instruction from the program.

Instead, the monitor attempts to exit, but then re-breaks with the PC at 002005. This appears to be true only if the PRG was transferred by m65/M65Connect. The content of the program does not appear to matter, and you don't have to run the program to reproduce the issue.

Additional context
Here's the beginning of monexit, which is where X goes:

cd9b          monexit
cd9b 20 80 03  jsr restore_system ;Restore system configuration
cd9e ad 05 11  lda save_status  ;Restore BASIC's init status
cda1 8d 04 11  sta init_status
cda4 ea        nop  
cda5 6c 00 11  jmp (system_vector) ;Return to system

(Issue relocated from MEGA65/m65connect#11.)

[dansanderson]

[dansanderson]

Trying with the refactor-ftp branch, 20220729.00-refact-00e36ae. Can still repro.

% ./bin/m65.osx --debug -l /dev/cu.usbserial-251633006E0EA ~/websites/site-dansanderson.com/content/lab-notes/2022/07/26/mega65-monitor/golml.prg
2022-07-29T07:36:06.785Z DEBG parameter parsing done
2022-07-29T07:36:06.786Z NOTE MEGA65 Cross-Development Tool 20220729.00-refact-00e36ae
2022-07-29T07:36:06.786Z INFO opening serial port /dev/cu.usbserial-251633006E0EA
2022-07-29T07:36:06.788Z DEBG set_serial_speed: 2000000 bps (OSX)
2022-07-29T07:36:06.935Z INFO checking if MEGA65 has RX buffer
2022-07-29T07:36:06.959Z INFO RX buffer detected. Latency will be reduced.
2022-07-29T07:36:06.979Z INFO detecting C64/C65 mode status
2022-07-29T07:36:07.075Z DEBG in C65 mode
2022-07-29T07:36:07.075Z NOTE loading file '/Users/dan/websites/site-dansanderson.com/content/lab-notes/2022/07/26/mega65-monitor/golml.prg'
2022-07-29T07:36:07.091Z INFO LOAD vector from ROM is $f4c4
2022-07-29T07:36:07.139Z DEBG Injecting string into key buffer at $02B0 : DLo"![$0d]
2022-07-29T07:36:07.139Z INFO waiting for breakpoint at $F4C4 to trigger
2022-07-29T07:36:07.155Z INFO breakpoint @ $F4C4 triggered
2022-07-29T07:36:07.235Z INFO requested file is '0:!' (len=3)
2022-07-29T07:36:07.235Z INFO forcing load address to $2001
2022-07-29T07:36:07.235Z DEBG read block for $2001 -- $2290 (656 bytes)
2022-07-29T07:36:07.363Z INFO storing end of program pointer at $82
2022-07-29T07:36:07.363Z INFO returning top of load address = $2291
2022-07-29T07:36:07.411Z NOTE loaded

[dansanderson]

It appears some bytes at $000380 are getting corrupted by the transfer, specifically:

 . 0380  A2 91    LDX  #$91
 . 0382  A0 22    LDY  #$22
 . 0384  18       CLC
 . 0385  60       RTS

These locations need to be preserved for a clean exit from the monitor. Original code before it is overwritten:

 . 0380  AD 10 01 LDA  $0110
 . 0383  AE 11 01 LDX  $0111

[lydon42]

should be fixed in refactor-m65

[lydon42]

also fixed in merge2-m65dbg

[dansanderson]

The latest in merge2-m65dbg resolves this issue. Per @lydon42 I'll close this ticket. Thanks very much!