On 21/4/21 4:07 pm, Sebastian Huber wrote:
> Hello,
>
> while working on the specification of some memory allocation directives I
> noticed that zero size allocations have no consistent behaviour in RTEMS. For
> example malloc( 0 ) returns NULL and posix_memalign(&p, align, 0) returns in
> p a
> unique pointer (or NULL if no memory is available). In POSIX zero size memory
> allocations are implementation-defined behaviour. The implementation has two
> options:
>
> https://pubs.opengroup.org/onlinepubs/9699919799/functions/malloc.html
>
> https://pubs.opengroup.org/onlinepubs/9699919799/functions/posix_memalign.html
>
> See also:
>
> http://devel.rtems.org/ticket/4390
>
> Linux returns a unique pointer for zero size memory allocations. I tend to use
> this approach as well and would like to remove the
>
>
> /*
> * Validate the parameters
> */
> if ( !size )
> return (void *) 0;
>
> in the RTEMS malloc().
For the record FreeBSD shows:
$ cat no-size.c
#include <stdio.h>
#include <stdlib.h>
int main(void)
{
printf("malloc size 0 = %p\n", malloc(0));
return 0;
}
$ cc -o no-size no-size.c
$ ./no-size
malloc size 0 = 0x800e16000
Are we too late to make such a change?
What if a user happens to depend on 0 returning NULL?
The FreeBSD jemalloc code has an option to assert if size is 0 which is an
interesting idea.
Chris
_______________________________________________
devel mailing list
devel@rtems.org
http://lists.rtems.org/mailman/listinfo/devel
