Re: patch for mmap and friends

[Matt Wette]

On 1/14/23 7:18 AM, Maxime Devos wrote:
> \
> Port objects should be accepted too, as previously asked on 
> <https://lists.gnu.org/archive/html/guile-user/2022-06/msg00060.html>.
> As implied by later comments, using a raw fd causes problems with 
> 'move->fdes'.  For the remaining response, I'll assume that the 
> function accepts ports as well.

> To avoid this problem, you can add
>
>   scm_remember_upto_here_1 (fd);
>
> after the SCM_SYSCALL.
> \


> IIRC there is a C trick involving fields, arrays and types to check 
> this at compile-time instead.  Maybe:
>
> struct whatever {
>    /* if availability of zero-length arrays can be assumed */
>    int foo[sizeof(size_t) - sizeof(void*)];
>    /* alternatively, a weaker but portable check */
>    int foo[sizeof(size_t) - sizeof(void*) + 1];
> };
>
> Greetings,
> Maxime.

Thanks for the feedback.   I'm sorry I missed you comments on the 
previous round.
I did respond to the ones I did catch.    I'll work this and resubmit.

Matt