Converter for IDA Pro .lst file to beebasm

[tricky]

I wrote this last weekend (and this desc!) but forgot to post it.
I know that this won't have wide appeal, but here goes.
For the ports that I have done to the beeb: Asteroids, Canyon Bomber, Circus, Rip Cord, Sprint and Warlords, I use IDA Pro (from a previous employer) to disassemble the game ROM and usually do some debugging in MAME.
Since I have done all the games that I think can easily be ported without modifying the ROM other than patching on load, I now need to be able to remove chunks of unused code/data to make room or need to relocate the code as if uses &C000..&FFFF.
This that can be removed include tilt/slam sensor code, language data and low-level code to deal with the games original hardware.

Anyway, this leaves me with 6502 assembler, but not in a format that I am going to use as I don't have a suitable assembler and anyway, I really only use beebasm unless it is a very small BBC BASIC sample or loader program.

So this tool's job is to take output like:

Code: Select all

RAM:0000 byte_0:	 .BYTE 0 ; (uninited)	 ; DATA	XREF: sub_64D3+38w
RAM:0000					 ; sub_657C+35w ...
RAM:0001 byte_1:	 .BYTE 0 ; (uninited)	 ; DATA	XREF: sub_64D3+32w
RAM:0001					 ; sub_64D3+77w ...
RAM:0002 tmp_02:	 .BYTE 0 ; (uninited)	 ; DATA	XREF: sub_6686+33w
RAM:0002					 ; sub_6686+5Dr ...
RAM:0002					 ; general use

and turn it into this:

Code: Select all

.byte_0
	EQUB 0
.byte_1
	EQUB 0
.tmp_02
	EQUB 0
	; general use

or:

Code: Select all

IRQ_ACK:4D00 ; ===========================================================================
IRQ_ACK:4D00
IRQ_ACK:4D00 ; Segment type: Regular
IRQ_ACK:4D00		     ;.segment IRQ_ACK
IRQ_ACK:4D00		     * =  $4D00
IRQ_ACK:4D00 irq_ack:	     ; 0 .BYTE uninited	& unexplored ; IRQ Ack
IRQ_ACK:4D01		     ; 0 .BYTE uninited	& unexplored
IRQ_ACK:4D02		     ; 0 .BYTE uninited	& unexplored
IRQ_ACK:4D03		     ; 0 .BYTE uninited	& unexplored
IRQ_ACK:4D04		     ; 0 .BYTE uninited	& unexplored
. . .
IRQ_ACK:4DFC		     ; 0 .BYTE uninited	& unexplored
IRQ_ACK:4DFD		     ; 0 .BYTE uninited	& unexplored
IRQ_ACK:4DFE		     ; 0 .BYTE uninited	& unexplored
IRQ_ACK:4DFF		     ; 0 .BYTE uninited	& unexplored
IRQ_ACK:4DFF ; end of 'IRQ_ACK'
IRQ_ACK:4DFF

to:

Code: Select all

ORG &4D00 ; segment IRQ_ACK
	; ===========================================================================

ORG &4D00 ; segment IRQ_ACK

.irq_ack

	; end of 'IRQ_ACK'
	SKIP &100

ORG &5000 ; segment ROM

It has several options:

Code: Select all

usage: input.lst [-disable|+enable] {output.asm}
+/-  a +/-  check_addr + Adds a CHECK_ADDR addr between every line
+/-  m +/-   add_macro + Adds the CHECK_ADDR addr macro
+/-  s +/-  subroutine + includes " == subroutine ==" comments
+/-  g +/-     segment + includes "; Segment" and ";.segment" comments
+/-  f +/-  end_of_fun + includes "; End of function" comments
+/-  c +/-    comments - strips all comments not on code
+/-  e +/-   empty_com + includes empty comments in output
+/-  t +/-  tab_to_spc + changes a tab character to a space
+/-  x +/-   data_xref + include DATA XREF comments
+/-  u +/-    uninited + include (uninited) comments

but the one that I wrote first was -a to add a macro to line of data or code to error if it wasn't where it was supposed to be.

Code: Select all

MACRO CHECK_ADDR ad
	IF ad <> P%
		PRINT CALLSTACK$
		PRINT "ERROR: P% missmatch is",~P%,"should be",~ad
	ENDIF
ENDMACRO
. . .
CHECK_ADDR &7FDF
	BPL  loc_7FBD
CHECK_ADDR &7FE1
	RTS
	; ---------------------------------------------------------------------------
CHECK_ADDR &7FE2
.test_grid_horz
	EQUB 0,&33,&66,&99,&CC,&FF
	; Vertical and horizontal positions for grid lines, stored as pointer values.
CHECK_ADDR &7FE8
.test_grid_vert

I've just noticed a bug, in that if you turn on the macro, it puts one on each byte that would normally be skipped as a block!
It probably doesn't matter that much as it is only really to use once.

When using it on Missile Command, it found quite a few places where there was an instruction like LDA &F0 which was in the original ROM as LDA &00F0 taking 3 bytes and an extra cycle. I added NOPs in the .LST to fix this and when everything aligned, reconverted without the MACROs and did a binary diff out the ROMs produced. They were the same apart from the NOPs and the instruction before.

An improvement would be to add a way to include a "changes" file so that the process could start from IDA Pro without having to redo changes to the .lst file, but for now, Perforce will help with that!

The project is in C++, but should be fine in a modern C compiler.
It is for Visual Studio and as such, there are a few bits to change if not using that:

#include <windows.h>

int error(const char * err)
{
fprintf(stderr, "%s(%i): error [%i]: \"%s\"+'%c' %s\n", argv1, line, colm, token, ch, err);
if (IsDebuggerPresent())
{
char error[1024];
sprintf(error, "%s(%i): error [%i]: \"%s\"+'%c' %s\n", argv1, line, colm, token, ch, err);
OutputDebugString(error);
DebugBreak();
}
return 4;
}

[markusher]

That's really interesting. IDA is one of my favourite tools and I used it quite extensively when looking at the various FileServer versions and FileStore ROMs.
Fortunately I had some source code to work with but I had considered doing something similar to output code for TurboMASM. Clearly though, beebasm is far more useful in conjunction with other available tools.

Really interesting post and glad to see I'm not alone in the use of IDA.

[tricky]

You might find 6502Bench useful.