int x[static 10]

[Serve Laurijssen]

A question about this C99 feature.

suppose I have a function

void f(int x[static 10]);

and I call it with

int *x = malloc(20 * sizeof(int));

f(x);

Should a compiler emit a warning then or fail?



What if it's called with int x[20]; ? Is there a difference?

 

[lawrence.jones]

suppose I have a function

void f(int x[static 10]);

and I call it with

int *x = malloc(20 * sizeof(int));

f(x);

Should a compiler emit a warning then or fail?

No. In this context, "static" indicates that the argument will contain
*at least* 10 elements, not exactly 10 elements.

What if it's called with int x[20]; ? Is there a difference?

No.

-Larry Jones

This game lends itself to certain abuses. -- Calvin

 

[Micah Cowan]

A question about this C99 feature.

suppose I have a function

void f(int x[static 10]);

and I call it with

int *x = malloc(20 * sizeof(int));

f(x);

Should a compiler emit a warning then or fail?

No. Nore is there UB invoked, nor any other problem. the [static 10]
means that the parameter must have at *least* 10 elements. No problems
there. However, if you attempted to pass in *fewer* than 10, you would
be invoking UB. No diagnostic is required. And there is no difference
in behavior between dynamic allocation using malloc(), and static or
automatic allocation.

HTH,
-Micah

 

[Dan Pop]

A question about this C99 feature.

suppose I have a function

void f(int x[static 10]);

and I call it with

int *x = malloc(20 * sizeof(int));

f(x);

Should a compiler emit a warning then or fail?

What if it's called with int x[20]; ? Is there a difference?

You've already gotten the correct replies. I only want to point out that
this is probably the most useless feature introduced by C99: even if the
function prototype is in scope and a function call is in obvious violation
(e.g. an array of 5 int is passed to your function), no diagnostic is
required. So, why bother with this feature? A helpful compiler could
produce the expected diagnostic even in the absence of the static
keyword...

Dan

 

[lawrence.jones]

You've already gotten the correct replies. I only want to point out that
this is probably the most useless feature introduced by C99: even if the
function prototype is in scope and a function call is in obvious violation
(e.g. an array of 5 int is passed to your function), no diagnostic is
required. So, why bother with this feature?

Because you've completely missed the point. The point is not to provide
better diagnostics (although a high-quality implementation can, and
should, use the information to do so) but to provide better performance
by allowing the compiler to generate code to pre-fetch some of the
values without having to worry about handling invalid addresses.

-Larry Jones

Aw Mom, you act like I'm not even wearing a bungee cord! -- Calvin