gcc3 vs 176.gcc

[dewar@gnat.com]

 > In C, we cannot divide all user code into "right" and "wrong" in this
 > kind of simple way, and certainly not based on the ISO standard.  That
 > standard is just the decisions of a certain committee (which I was a
 > member of) about what cases conforming compilers should commit to
 > support.  We must not let ourselves start thinking that C code is
 > "wrong", just because it is not conforming ISO C code.

Sure, that is reasonable with respect to GNU-C and our world. But for
a standard benchmark, that is to be processed by arbitrary compilers, it
is only reasonable to restrict the benchmark to either standard C or
some very well defined extension of it. The issue is not whether the
code is right or wrong, but whether a conforming ANSI C compiler is
required to make sense of it.

The GNU project can of course simply declare that "if you want to 
compile gcc, your C compiler had better do sensible things" -- this
attitude being a little more supportable if we clearly define what
we mean by sensible things.

But for an official benchmark to take this viewpoint is a little different
I think. In other words, I am not saying that the original code is wrong.
Indeed for me I think that at this stage it is reasonable to simply require
that gcc be compilable by GNU-C period. If however you want the criterion
to be that gcc can be compiled by *any* ISO-conforming C compiler, then
you can no longer take the above quoted position. 

An interesting situation ...