🚀 go-pugleaf

In article <7360@boring.mcvax.cwi.nl}, steven@mcvax.cwi.nl (Steven Pemberton) writes:
} In article <1403@uwmacc.UUCP> edwards@unix.macc.wisc.edu.UUCP (mark edwards) writes:
} >   Yes indeed C can describe itself.
} 
} Actually, I'm not completely convinced of this. The question of whether C
} can be used to describe itself or not seems to be the same as whether the
} sentence "This sentence is true" is true or false: it depends on your
} assumptions.
} 
} Evidence of this is in Dennis Ritchie's Turing award lecture: he had built a
} trojan horse into the C compiler. The compiler looked identical to its
} previous self, but actually compiled a (slightly) different language. This
} suggests to me that you can't look at the compiler source to determine what
} a certain construct in C means, even if you know the meaning of the language
} it compiles to.
} 
} Steven Pemberton, CWI, Amsterdam; steven@cwi.nl or try steven@mcvax.uucp

If I recall correctly, he first wrote a compiler that on a special
string input, output a binary that did NOT correctly match the
source code.  Then he fed it a "clean" C compiler, and it output
that C compiler with the additional property that it also contained
the trojan horse.  Thus, as long as you kept compiling the C
compiler with the C compiler, you would get a trojan horse C,
which would do a couple of interesting things when fed the right
input file.

But, the point is, that the original compiler and all its
descendants were NOT (!!!!!) actual C compilers, since the
did NOT correctly transform all proper inputs into equivalent
outputs.  So, this example is not relevant.

---------------------------------------------------------------------------
Steve Chapin                    |      Chapin's Law of Motion:  
ARPA:  sjc@gwen.cs.purdue.edu   |      You can get anywhere in 10 minutes
UUCP:  ...!purdue!sjc           |      if you drive fast enough.
---------------------------------------------------------------------------