> Now, as of yesterday, it compiled cleanly on i386. Why the compiler detects
> some things on other architectures that it happily accepts on i386 is a
> mystery to me.

Not to me :-)

	fprintf(output, s[0] == '/' && s[1] == '*' ? "\t%.*s" : "\t/* %.*s */",
		e_lab - s, s);

e_lab and s are both char *; their difference has an
implementation-defined type which is clearly "int" on x86, and "long"
on sparc.

This was not caught before because the compiler didn't know how to
grovel around in the ?: expression to find the format string..

					- Bill