Re: [PATCH 11/11] getdtablesize: Fix RLIMIT_NOFILE fallback case

[Paul Eggert]

Kevin Cernekee wrote:

> @@ -99,8 +99,8 @@ rpl_getdtablesize(void)
>        a smaller soft limit, the smaller limit is not enforced, so
>        we might as well just report the hard limit.  */
>     struct rlimit lim;
> -  if (!getrlimit (RLIMIT_NOFILE, &lim) && lim.rlim_max != RLIM_INFINITY)
> -    return lim.rlim_max;
> +  if (!getrlimit (RLIMIT_NOFILE, &lim) && lim.rlim_cur != RLIM_INFINITY)
> +    return lim.rlim_cur;
>     return getdtablesize ();
>   }

This one doesn't look right, as it causes the code to disagree with the 
preceding comment.  Doesn't the patch lose a Cygwin fix?  That is, shouldn't the 
code do one thing on Cygwin, and another on other platforms where rlim_cur 
presumably works?