Velocity Reviews > Evaluation Order in Function Declaration

# Evaluation Order in Function Declaration

Shao Miller
Guest
Posts: n/a

 02-28-2012
Is the evaluation order guaranteed to be left-to-right for the
declaration of 'baz' in the following code? Or are the array size
expressions unsequenced relative to one another, as in 'a + b'? Or
something else?

#include <stdio.h>

int foo(void) { puts("foo"); return 1; }
int bar(void) { puts("bar"); return 1; }

int baz(int (* x)[foo()], int (* y)[bar()]) { return 0; }

int main(void) { return baz(0, 0); }

Jens Gustedt
Guest
Posts: n/a

 02-28-2012
Hello,

Am 02/28/2012 06:22 AM, schrieb Shao Miller:
> Is the evaluation order guaranteed to be left-to-right for the
> declaration of 'baz' in the following code? Or are the array size
> expressions unsequenced relative to one another, as in 'a + b'? Or
> something else?
>
> #include <stdio.h>
>
> int foo(void) { puts("foo"); return 1; }
> int bar(void) { puts("bar"); return 1; }
>
> int baz(int (* x)[foo()], int (* y)[bar()]) { return 0; }
>
> int main(void) { return baz(0, 0); }

Yes, not only that it is left to right, you are even allowed to use each
of the preceding identifiers in the expressions. Otherwise things like

double sum(size_t n, size_t m, double A[n][m]) { ... }

could never work reliably.

Jens

Kaz Kylheku
Guest
Posts: n/a

 02-28-2012
On 2012-02-28, Jens Gustedt <(E-Mail Removed)> wrote:
> Hello,
>
> Am 02/28/2012 06:22 AM, schrieb Shao Miller:
>> Is the evaluation order guaranteed to be left-to-right for the
>> declaration of 'baz' in the following code? Or are the array size
>> expressions unsequenced relative to one another, as in 'a + b'? Or
>> something else?
>>
>> #include <stdio.h>
>>
>> int foo(void) { puts("foo"); return 1; }
>> int bar(void) { puts("bar"); return 1; }
>>
>> int baz(int (* x)[foo()], int (* y)[bar()]) { return 0; }
>>
>> int main(void) { return baz(0, 0); }

>
> Yes, not only that it is left to right, you are even allowed to use each
> of the preceding identifiers in the expressions. Otherwise things like
>
> double sum(size_t n, size_t m, double A[n][m]) { ... }
>
> could never work reliably.

Sure they could. The above could mean that A[n][m], n and m independently
receive values from the same sources (the argument expressions). The [n][m]
doesn't have to refer to the values stored in the parameters n, and m
but rather the [n][m] can indicate additional sinks for the data flows
emanating from argument 1 and 2.

Such semantics can be hammered out. There is no end to how you can lay weasely
traps for the programmer (as the sordid history of this language shows.)

Jens Gustedt
Guest
Posts: n/a

 02-28-2012
Am 02/28/2012 10:31 AM, schrieb Kaz Kylheku:
> On 2012-02-28, Jens Gustedt <(E-Mail Removed)> wrote:
>> Hello,
>>
>> Am 02/28/2012 06:22 AM, schrieb Shao Miller:
>>> Is the evaluation order guaranteed to be left-to-right for the
>>> declaration of 'baz' in the following code? Or are the array size
>>> expressions unsequenced relative to one another, as in 'a + b'? Or
>>> something else?
>>>
>>> #include <stdio.h>
>>>
>>> int foo(void) { puts("foo"); return 1; }
>>> int bar(void) { puts("bar"); return 1; }
>>>
>>> int baz(int (* x)[foo()], int (* y)[bar()]) { return 0; }
>>>
>>> int main(void) { return baz(0, 0); }

>>
>> Yes, not only that it is left to right, you are even allowed to use each
>> of the preceding identifiers in the expressions. Otherwise things like
>>
>> double sum(size_t n, size_t m, double A[n][m]) { ... }
>>
>> could never work reliably.

>
> Sure they could. The above could mean that A[n][m], n and m independently
> receive values from the same sources (the argument expressions). The [n][m]
> doesn't have to refer to the values stored in the parameters n, and m
> but rather the [n][m] can indicate additional sinks for the data flows
> emanating from argument 1 and 2.

theoretically they could. pratically the standard says that the
identifier n for example is visible for the declaration of A

> Such semantics can be hammered out. There is no end to how you can lay weasely
> traps for the programmer (as the sordid history of this language shows.)

sure, I just don't see how your comment is related to the question
Shao had and how with any respect it would be helpful. You seem to
find it necessary to rant against VLA in any form you see them. Not
very helpful for anybody, probably not even for yourself.

Jens