On Sat, Dec 10, 2022 at 05:24:24PM -0500, Jan Schaumann wrote:
> Martin Husemann <mar...@duskware.de> wrote:
> How do you feel about adding a separate manual page
> for the sizeof operator? (See attached proposal,
> though my mandoc is not very good and likely needs
> clean up.)
some kibitzing on the page itself, since I'm not by any means
convinced this would be a bad page to have in the system:
(first, since it's an identifier in the C function namespace, even
though it's not actually a function, the page belongs in section 3)
> .Sh NAME
> .Nm sizeof
> operator
Pretty sure having "operator" there is not correct mdoc.
> .Nd yield the storage size of the given operand
"find", not "yield"
> The unary
> .Nm
> operator yields the storage size of an expression or
Just say that .Nm finds the size. It being an operator rather than a
function is a secondary point and belongs as an aside a later section.
> data type in
> .Em char sized units .
> As a result, 'sizeof(char)' is always guaranteed to be 1.
It is _defined_ to be 1.
> (The number of bits per
> .Dv char
> is given by the
> .Vt CHAR_BITS
that's CHAR_BIT
> .Sh EXAMPLES
> The following example illustrate the possible results
^s
> For initialized data or uninitialized arrays of a
> fixed size known at compile time,
> .Nm
> will return the correct storage size:
"correct" does not seem like the right word here.
> When applied to a struct or union,
> .Nm
> returns the total number of units in the object,
"char-sized units", otherwise it sounds like it returns the number of
members in the struct.
> When applied to a struct containing a flexible array
> member,
^ at the end
since most people, even those with plenty of experience reading the
standard, have no idea precisely what widgetry "flexible array member"
means without looking it up.
> if ((buf = malloc(BUFSIZ)) == NULL) {
> perror("malloc");
> }
> /* Warning: wrong! */
> (void)strncat(buf, input, sizeof(buf) - 1);
Don't encourage people to use strncat. Or even strncpy. Make it
snprintf if you don't want to use strlcpy. And calling strncat on an
uninitialized array is itself wrong :-)
> Since it is evaluated by the compiler and not the
> preprocessor, the
> .Nm
> operator cannot be used in a preprocessor expression.
There have been compilers where sizeof() works in #if, e.g.
Borland's. FWIW. But it's not standard, nor generally implemented.
(Though TBH, I'm surprised not, given the large number of #defines
about type sizes that a modern compiler's cpp has to know anyway.)
--
David A. Holland
dholl...@netbsd.org
