26 Aug
2023
26 Aug
'23
2:33 a.m.
below...
On Fri, Aug 25, 2023 at 3:51 PM segaloco via TUHS <tuhs(a)tuhs.org> wrote:
Hello, I've been doing some research on the
history of disassembly lately,
tools available historically, today, and what sorts of developments have
been made regarding utilities and systems for taking a machine-code binary
and working it back to some semblance of source code.
So in the early days UNIX had das(I), a PDP-11 disassembler I believe
written by Ken (he's OWNER in the manual) with very little information
other than "it exists". Fast forward to the UNIX 4.1 manual in 1981 for
the 3B20S and there is dis(1), a 3B20 disassembler. Other such manuals
feature dis(1) versions for other 3B targets.
Was a disassembler ever considered part of the standard binary objects
toolkit with the assembler, linker, etc.
not to my memory - although some of the debuggers could. IIRC, the DDT
that was on the Harvard tape knew about it. I also remember on that tape
was a PDP-11 disassembler. Phil Karn wrote a table-based one for UNIX
when we were students - but it was aimed at 8-bit micros. It could do
8080/8085 and Z80; if I remember, it could also do MOS6502 and M6800. It
had a feature that it could take an external symbol table and turn out code
that was reasonable to reassemble. [ I may have a copy if it squirreled
away ].
That said, while they we not part of the core tool kit, by the time of
BSD4.2 there were a couple of disassemblers kicking around the USENET. I
remember one for the Vax and another for the 68000. You might do a grep for
dis-assembler in the USENET archives for comp.sources
or was that the sort of thing that was more niche and
therefore just kinda
cropped up when/if someone decided to write one?
exactly - need driven. Phil wrote his when we were trying to pull apart a
ROM for a tape controller. It had a funky interface on it that was not
well documented and what we did have, was wrong. So, disassembled enough
of the ROM that we could changed it.
Were there legal concerns to be grappled with when
producing a
disassembler?
Mumble ... by the mid-80s/late-90 people we disassembling code for game
controllers and PCs. So many manufacturers started adding words in the EULA
saying that was a no-no. But I don't remember worrying about it much when
I was a student 10-15 years before that.
Were such tools ever shipped or did they only appear
in the manuals as
they were technically up in the code base, just not commonly distributed or
used? Also, was there any thought given during the development of C to
producing "decompilers" as has been becoming more common lately? Or was it
a foregone conclusion that C to assembly is a "lossy" conversion and going
the other direction couldn't be fully automated.
Again - in V6/V7 with DMR's compiler, it was not always easy, but the code
generally was pretty straightforward.
Post Wulf's 'Green Book' on compiler optimization and we started to have a
generation of BLISS-style optimizers pretty much everywhere, I think those
compilers really started refactoring code plus instruction sets got more
sophisticated, so I think it started to get harder and harder to
reconstruct.
But I'll defer to someone like Paul W or Steve Johnson who loved building
those style of tools.
Thank you for any insights!
- Matt G.
ᐧ
ᐧ