bug#13319: 24.2.91; emacsclient doesn't find server with nonstandard name

bug#13319: 24.2.91; emacsclient doesn't find server with nonstandard name

[Michael Heerdegen]

Hi,

In emacs -Q, do this:

(progn
  (require 'server)
  (let ((server-name (concat "server_" (format-time-string  "%H:%M:%S"))))
    (server-start)))

Then I try to connect to this server, but it fails:

| bash-4.2:mh:~$ /usr/bin/emacsclient ~/.bashrc 
| /usr/bin/emacsclient: can't find socket; have you started the server?
| To start the server in Emacs, type "M-x server-start".
| /usr/bin/emacsclient: No socket or alternate editor.  Please use:
| 
|         --socket-name
|         --server-file      (or environment variable EMACS_SERVER_FILE)
|         --alternate-editor (or environment variable ALTERNATE_EDITOR)


When I explicitly specify the server name with the -s option,
emacsclient succeeds.  But due to the doc (man page), "emacsclient"
should connect to "the first server it finds":

| `-s SERVER-NAME'
| `--socket-name=SERVER-NAME'
|      Connect to the Emacs server named SERVER-NAME.  The server name is
|      given by the variable `server-name' on the Emacs server.  If this
|      option is omitted, `emacsclient' connects to the first server it
|      finds.  (This option is not supported on MS-Windows.)

Obviously this is not fulfilled here.  And it is no useful behavior.  If
only one server is running (and the socket is located under the
canonical path), emacsclient should find and use it.  It is arguable
which server should be used if more than one is running (and what the
"first server it finds" should be).

BTW, here is an explanation for why I want to start the server like
above (not substantially important for this report):

I often make changes to my config file/ elpa installations or rebuild
Emacs.  To guarantee that I always have a running Emacs (as a fallback
when something went wrong and a newly started Emacs fails to start), I
keep my old Emacs open and start a new one.  If I had just

  (server-start)

in my config, the new Emacs would not start a server (since the old one
is still running).  After having closed the old Emacs, I had no server
running, so I had to manually start the server every time.  This is
annoying.  So I want to start a server with a unique name.


Thanks,

Michael.


In GNU Emacs 24.2.91.2 (x86_64-unknown-linux-gnu, GTK+ Version 2.24.10)
 of 2012-12-30 on drachen
Bzr revision: rgm@gnu.org-20121230122326-m6gw29pkwjgwqf9d
Windowing system distributor `The X.Org Foundation', version 11.0.10707000
System Description:	Debian GNU/Linux testing (wheezy)

bug#13319: 24.2.91; emacsclient doesn't find server with nonstandard name

[Andreas Schwab]

Michael Heerdegen <michael_heerdegen@web.de> writes:

> Obviously this is not fulfilled here.  And it is no useful behavior.  If
> only one server is running (and the socket is located under the
> canonical path), emacsclient should find and use it.

This is questionable.  How does it know that a random socket is a emacs
server socket?

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#13319: 24.2.91; emacsclient doesn't find server with nonstandard name

[Michael Heerdegen]

Andreas Schwab <schwab@linux-m68k.org> writes:

> Michael Heerdegen <michael_heerdegen@web.de> writes:
>
> > Obviously this is not fulfilled here.  And it is no useful behavior.  If
> > only one server is running (and the socket is located under the
> > canonical path), emacsclient should find and use it.
>
> This is questionable.  How does it know that a random socket is a emacs
> server socket?

Here on Debian, it looks under /tmp/emacs$UID.  Can there be any other
socket?


Regards,

Michael.

bug#13319: 24.2.91; emacsclient doesn't find server with nonstandard name

[Lars Ingebrigtsen]

Michael Heerdegen <michael_heerdegen@web.de> writes:

>> > Obviously this is not fulfilled here.  And it is no useful behavior.  If
>> > only one server is running (and the socket is located under the
>> > canonical path), emacsclient should find and use it.
>>
>> This is questionable.  How does it know that a random socket is a emacs
>> server socket?
>
> Here on Debian, it looks under /tmp/emacs$UID.  Can there be any other
> socket?

This was eight years ago, and some things have changed in this area
(like the location of the server socket), but the issue still remains, I
think?  That is, the client won't look around for a socket with a
non-standard name.

But I tend to think that this is actually the correct behaviour --
having emacsclient guess at what socket name to use sounds a bit too
DWIM to me:  I may deliberately have set up a server to do one very
specific thing under a non-standard name, and I would not want any
random emacsclient invocation to use that special server just because it
can't find the standard socket.  (Which I may just have forgotten to
start the server for.)

So I think this behaves as it should.  Anybody else got another opinion?

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#13319: 24.2.91; emacsclient doesn't find server with nonstandard name

[Michael Heerdegen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> >> This is questionable.  How does it know that a random socket is a emacs
> >> server socket?
> >
> > Here on Debian, it looks under /tmp/emacs$UID.  Can there be any other
> > socket?
>
> This was eight years ago, and some things have changed in this area
> (like the location of the server socket), but the issue still remains, I
> think?

Seems so.

> That is, the client won't look around for a socket with a
> non-standard name.
>
> But I tend to think that this is actually the correct behaviour --
> having emacsclient guess at what socket name to use sounds a bit too
> DWIM to me:  I may deliberately have set up a server to do one very
> specific thing under a non-standard name, and I would not want any
> random emacsclient invocation to use that special server just because it
> can't find the standard socket.  (Which I may just have forgotten to
> start the server for.)

What would happen in this case?  Can't you use another place to store
the socket then?

> So I think this behaves as it should.  Anybody else got another opinion?

I don't know if something like you describe could be prevented.  If you
are sure that my suggestion would cause trouble, then I'm ok with
closing.  I'm too ignorant about how this socket stuff is working to
have an opinion.

Michael.

bug#13319: 24.2.91; emacsclient doesn't find server with nonstandard name

[Lars Ingebrigtsen]

Michael Heerdegen <michael_heerdegen@web.de> writes:

> What would happen in this case?  Can't you use another place to store
> the socket then?

I can, but I don't see how that's a better design.

>> So I think this behaves as it should.  Anybody else got another opinion?
>
> I don't know if something like you describe could be prevented.  If you
> are sure that my suggestion would cause trouble, then I'm ok with
> closing.

I think it would, but I'd like to hear whether anybody else has an
opinion here...

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#13319: 24.2.91; emacsclient doesn't find server with nonstandard name

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Date: Fri, 16 Jul 2021 10:49:24 +0200
> Cc: Andreas Schwab <schwab@linux-m68k.org>, 13319@debbugs.gnu.org
> 
> >> So I think this behaves as it should.  Anybody else got another opinion?
> >
> > I don't know if something like you describe could be prevented.  If you
> > are sure that my suggestion would cause trouble, then I'm ok with
> > closing.
> 
> I think it would, but I'd like to hear whether anybody else has an
> opinion here...

I think we should fix the documentation to not say anything about "the
first available socket" there.  If the socket name is not the default
one, the user should tell emacsclient which socket to use.

bug#13319: 24.2.91; emacsclient doesn't find server with nonstandard name

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> I think we should fix the documentation to not say anything about "the
> first available socket" there.  If the socket name is not the default
> one, the user should tell emacsclient which socket to use.

Yup.  Done, and closing.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no