MAMEWorld Forums - EmuChat - Debugger command 'symlist' bugs

Tested version : 0.140u1

1) can't print global symbols when input this command without any parameter.
debughlp says "symlist : Displays the global symbol table." But printed data is always CPU's symbols even without parameter.

See debugcmd.c:execute_symlist()


Code:


	/* validate parameters */
	if (!debug_command_parameter_cpu(machine, param[0], &cpu))
		return;

	if (cpu != NULL)
	{
		symtable = &cpu->debug()->symtable();
		debug_console_printf(machine, "CPU '%s' symbols:\n", cpu->tag());
	}
	else
	{
		symtable = debug_cpu_get_global_symtable(machine);
		debug_console_printf(machine, "Global symbols:\n");
	}

debug_command_parameter_cpu() attempts to return "visible CPU" in case that parameter strings is NULL so that "cpu != NULL" is always TRUE.
Simple solution is that change "cpu != NULL" to "params != 0" (or "params != 0 || cpu != NULL") though.

2) MAME crashes when try to get symbols for Z80 CPU.
I find this problem in bygone. Inputting "symlist maincpu" (or "symlist audiocpu") causes the crash.
This problem happens in the following sequence in execute_symlist()


Code:


	/* iterate over symbols and print out relevant ones */
	for (symnum = 0; symnum < count; symnum++)
	{
		const symbol_entry *entry = symtable->find(namelist[symnum]);
		assert(entry != NULL);
		UINT64 value = entry->value();

		/* only display "register" type symbols */
		debug_console_printf(machine, "%s = %s", namelist[symnum], core_i64_hex_format(value, 0));
		if (!entry->is_lval())
			debug_console_printf(machine, "  (read-only)");
		debug_console_printf(machine, "\n");
	}

"value = entry->value()" causes NULL (entry) access after 8-bit register entry ("A", "B", "C", "D", "E", "H", "L", "I", "R").
I guess this is very similar to the problem for Testers Bug ID:4121 - Debugger cannot set 8bit registers with 'do' command.

find() returns NULL when matched hash is not found though. So I doubt the reliability of tagmap in case that a tag is a single character.
See tagmap.h


Code:

INLINE UINT32 tagmap_hash(const char *string)
{
	UINT32 hash = (string[0] << 5) + string[1];
	char c;

	string += 2;
	while ((c = *string++) != 0)
		hash = ((hash << 5) | (hash >> 27)) + c;
	return hash;
}

For example : "a" register, string[0] = 'a', string[1] = '\0' (stopper) but it attemps to calculate a character at "string +=2" (= string[2]) or later.
I don't know C/C++ guarantees a content after stopper character (I think it is "unknown"). But I guess this will be the reason of wrong hash calculation.

[ATTACHED IMAGE]

Attachment


INLINE UINT32 tagmap_hash(const char *string)
{
	UINT32 hash = (string[0] << 5) + string[1];
	char c;

	if (strlen(string) > 1)
		string += 2;

	while ((c = *string++) != 0)
		hash = ((hash << 5) | (hash >> 27)) + c;
	return hash;
}


860D: C9            ret
860E: 4F            ld   c,a
860F: 3A 82 F0      ld   a,($F082)
8612: CB 5F         bit  3,a
8614: C0            ret  nz
8615: 21 FF E7      ld   hl,$E7FF
8618: 7E            ld   a,(hl)
8619: 71            ld   (hl),c
861A: BE            cp   (hl)
861B: 32 A3 F0      ld   ($F0A3),a
861E: 28 FA         jr   z,$861A
8620: C9            ret