On Wed, 2 Oct 2019 at 17:48, Eli Zaretskii <
eliz@gnu.org> wrote:
> From: Andreas Schwab <schwab@linux-m68k.org>
> Date: Wed, 02 Oct 2019 13:02:43 +0200
> Cc: Richard Copley <rcopley@gmail.com>, 37576@debbugs.gnu.org
>
> On Okt 02 2019, Juanma Barranquero <lekktu@gmail.com> wrote:
>
> > When the search fails, Emacs is now reporting "permission denied" if some
> > PATH element was non-existent, or "no such file or directory" otherwise. At
> > least on Windows.
>
> And that is a bug. A PATH element pointing to a non-existing directory
> should be ignored.
I don't know what "ignore" means in this case, since call-process
should signal an error if it's unable to find the program. If you
mean that nonexistent directories should be treated as if they
existed, but didn't have the program in them, then I agree, and I've
now fixed the Windows build's behavior to match that of the Posix
builds in this case.
I agree with that too, FWIW. Thanks.
I still didn't hear from Richard confirming that his case is indeed
caused by a non-existent directory on PATH.
Yes it is. (At least, there are non-existent directories in PATH, and I get the correct message now.)
Maybe there are other
factors at work here.