Re: pg_check_dir comments and implementation mismatch

On Sun, Feb 22, 2015 at 07:57:41PM -0500, Tom Lane wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
> > On Fri, Feb 20, 2015 at 12:59 AM, Noah Misch <noah@leadboat.com> wrote:
> >> On Mon, Feb 02, 2015 at 03:48:33PM -0500, Robert Haas wrote:
> >>> If readir() fails and closedir() succeeds, the return will be -1 but
> >>> errno will be 0.
> 
> >> Out of curiosity, have you seen a closedir() implementation behave that way?
> >> It would violate C99 ("The value of errno is zero at program startup, but is
> >> never set to zero by any library function.") and POSIX.
> 
> > No.  Good point, I didn't think about that.  I think this way is safer, though.
> 
> While the spec forbids library functions from setting errno to zero, there
> is no restriction on them changing errno in other ways despite returning
> success; their exit-time value of errno is only well-defined if they fail.
> So we do need to preserve errno explicitly across closedir(), or we may
> report the wrong failure from readdir().

Yes.  I'm happy with the commit itself.