X-Authentication-Warning:	delorie.com: mail set sender to djgpp-bounces using -f
From:	Rugxulo <rugxulo AT gmail DOT com>
Newsgroups:	comp.os.msdos.djgpp
Subject:	GAS ".code16" and DJASM quirks ...
Date:	Mon, 9 Nov 2009 20:03:14 -0800 (PST)
Organization:	http://groups.google.com
Lines:	69
Message-ID:	<0008cd2e-63c9-43cb-82d4-11aece7c209f@c3g2000yqd.googlegroups.com>
NNTP-Posting-Host:	68.220.162.125
Mime-Version:	1.0
X-Trace:	posting.google.com 1257825794 29226 127.0.0.1 (10 Nov 2009 04:03:14 GMT)
X-Complaints-To:	groups-abuse AT google DOT com
NNTP-Posting-Date:	Tue, 10 Nov 2009 04:03:14 +0000 (UTC)
Complaints-To:	groups-abuse AT google DOT com
Injection-Info:	c3g2000yqd.googlegroups.com; posting-host=68.220.162.125;
	posting-account=p5rsXQoAAAB8KPnVlgg9E_vlm2dvVhfO
User-Agent:	G2/1.0
X-HTTP-UserAgent:	Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US)
	AppleWebKit/532.0 (KHTML, like Gecko) Chrome/3.0.195.32 Safari/532.0,gzip(gfe),gzip(gfe)
To:	djgpp AT delorie DOT com
DJ-Gateway:	from newsgroup comp.os.msdos.djgpp
Reply-To:	djgpp AT delorie DOT com

I've been messing with a small Befunge93 interpreter in assembly. I
didn't write it, just tweaked it a bit, converted it to something
besides TASM, fixed a few bugs (though some obscure ones still exist),
and shrank its size a lot. I know it's silly, but ...

http://board.flatassembler.net/topic.php?t=10810

Yesterday I made it work with DJASM (via MAKE.BAT which uses sed
script). Unfortunately, I couldn't manually get GAS working earlier
today. Anyways, here's what I discovered:

DJASM:

1). no "salc" (0xD6) support (though other assemblers have it except
GAS)
2). no hex numbers a la "30h" (only "0x30" ??), that's very annoying
and not Intel-friendly, IMHO, can't be hard to add support for that
can it?
3). '.type "com"' even when using ".org 0x100" seems useless, "djasm
befi.dj" still makes an .EXE unless explicitly saying "djasm befi.dj
befi.com"
4). "rep movsb" must be split to use two lines (vaguely annoying but
not the only assembler to require this, e.g. GAS or WolfASM)
5). it's also picky on "byte" or "word" overrides, which is kinda
clunky
6). jumps that are > 128 bytes away aren't manually converted to 3-
byte jumps, so they have to be manually tweaked ("jmp" -> "jmpl"),
which is annoying

Okay, so I know it's not perfect and never was meant for too much
initially. I also know it's probably moot mentioning this, but I tried
anyways. At least it works!   ;-)

While GAS now supports ".code16" and everything mostly seems to work
(almost), I have a few doubts that anybody really uses it much for 16-
bit (or maybe I did it wrong, who knows).

GAS:

1). no "salc" support (argh, oh well)
2). "0x300" required instead of "300h" (seriously?? bah)
3). ".org 0x100" doesn't seem to do anything useful (adds a bunch of
zero bytes at the beginning), nor does ".data" (adds bunch of crud to
end)
4). seems to align the code with no way to turn that off (is there?),
so it's "too big" a result (e.g. smallest .COM I got was 1536 bytes)
5). can't even directly output a flat binary (.COM) file (can it?)
except via external tools like OBJCOPY (yes) or LD (doubt it), can
it???
6). it seems to assume 0x1000 as "entry" and using "-e0x100" (or
anything else I tried) doesn't seem to change the following (of what I
assume is a) "bug":

(normal assembler)
dw offset lnop

(GAS)
.short lnop

For some reason, while most ".intel_syntax noprefix" code requires
"offset" and "word ptr" etc., using "offset" like ".short offset lnop"
isn't accepted. So the closest I can get is as mentioned above, BUT it
doesn't generate a correct address/offset (instead of being 0x262 it's
like 0x1162).  I'm blindly guessing entry is accidentally hard-coded
or maybe not handled correctly or whatever. (This was 2.19, I think.)

So based upon that and not knowing a solution, I gave up on GAS (for
now). I almost wouldn't even post this, but there's no much other
traffic (die, spam!) and it is at least *almost* on topic.   ;-)

- Raw text -