This is growing severely off-topic, but...
Andy Isaacson wrote:
>
> (4)
> dpy = XOpenDisplay('\0');
>
> is pretty questionable. I was going to say that the compiler should
> warn you in this case, but apparently this gets away with being legal
> due to the "sizeof('c')==sizeof(int)" clause.
>
> It seems likely that there could be systems where this example would
> fail _if sizeof('c') were 1_, but since sizeof('c')==sizeof(int), it's
> unlikely that there could be a system perverse enough to make it fail.
>
Not at all. Standard C allows implicit conversion between any numeric
types in an assignment. There is nothing wrong with saying 'foobar(c)'
where c is of type 'char' and foobar() is declared to have an argument
of type 'int'. Compilers usually provides a warning when an implicit
integer-to-integer conversion can cause loss of information, but not
otherwise.
However, that discussion is irreleveant since we're talking about a
conversion from 'int' to 'char *' here (standard C specifies the type of
a character constant as 'int'), or possibly a conversion from 'int' to
'int' if the arguments to XOpenDisplay() have not been declared.
A conversion from a numeric type to a pointer type is unspecified in
ANSI C, and this is where there might be a problem. However, the case of
NULL == 0 == '\0' is so common that this would probably work just fine
on just about any platform. In fact, C++ and possibly also C89 and C99
_requires_ that NULL == 0 == '\0', and allows implicit conversion
between any numeric and pointer type in an assignment where the
initializer is a constant 0, NULL, or '\0'.
Not that I would ever write XOpenDisplay('\0'), which is utterly
confusing :-)
--
Christer Palm
_______________________________________________
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert
