Re: emacsclient on WXP does not work (as it used to)

[dhruva]

The following patch fixes this issue since tty and TERM does not make
sense on windows (a wild guess though) ...

diff --git a/lib-src/emacsclient.c b/lib-src/emacsclient.c
index 0bbc385..0452cb3 100644
--- a/lib-src/emacsclient.c
+++ b/lib-src/emacsclient.c
@@ -1446,7 +1446,6 @@ main (argc, argv)
       char *tty_name = NULL;
 #ifndef WINDOWSNT
       tty_name = ttyname (fileno (stdout));
-#endif

       if (! tty_name)
         {
@@ -1468,15 +1467,19 @@ main (argc, argv)
                    " is not supported\n", progname);
           fail ();
         }
+#endif
+
 #if !defined (NO_SOCKETS_IN_FILE_SYSTEM)
       init_signals ();
 #endif

+#ifndef WINDOWSNT
       send_to_emacs (emacs_socket, "-tty ");
       quote_argument (emacs_socket, tty_name);
       send_to_emacs (emacs_socket, " ");
       quote_argument (emacs_socket, type);
       send_to_emacs (emacs_socket, " ");
+#endif
     }

   if (window_system)

On Fri, Oct 31, 2008 at 5:57 PM, Juanma Barranquero <address@hidden> wrote:
> On Fri, Oct 31, 2008 at 12:15, Eli Zaretskii <address@hidden> wrote:
>
>> Can you please explain why is this a problem on Windows?  I understand
>> that the addition of current_frame to the `if' clause causes it to be
>> entered on Windows, whereas it wasn't before, but how does that, and
>> the fact that tty_name is NULL, cause emacsclient failure reported by
>> dhruva?
>
> I don't know why dhruva got the "connect" error message. My guess is
> that they are two different problems:
>
>  1.- emacsclient opened the socket (so no INVALID_SOCKET), but couldn't 
> connect
>  2.- emacsclient couldn't get the terminal name, and it didn't have
> alternate-editor
>
> though I don't know how or why 2) would happen, unless there were two
> different emacsclients involved, called through alternate-editor (and
> even so, it doesn't make much sense).
>
> But anyway, my point was just that the very next code after the if() is
>
>      char *type = egetenv ("TERM");
>      char *tty_name = NULL;
> #ifndef WINDOWSNT
>      tty_name = ttyname (fileno (stdout));
> #endif
>
>      if (! tty_name)
>        {
>          message (TRUE, "%s: could not get terminal name\n", progname);
>          fail ();
>        }
>
> so entering the if() is *always* going to fail(). I'm unable to see
> how that be useful on Windows.
>
>> Agreed (to the ``ugly'' part).
>
> Glad we agree on this :)
>
> Have you any comment about the other part? ("Are there legitimate uses
> of tty == 1 on Windows?")
>
>> I'd like to try to find a less ugly fix.
>
> It's Chong's patch, perhaps he can shed a bit of light.
>
>  Juanma
>



-- 
Contents reflect my personal views only!