GNU bug report logs - #1836
emacs -Q -fn "nonexistent" hangs

Previous Next

To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 1836 in the body.
You can then email your comments to 1836 AT debbugs.gnu.org in the normal way.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> emacsbugs.donarmstrong.com (full text, mbox):

From: "Juanma Barranquero" <lekktu <at> gmail.com>
To: "Emacs Bug Tracker" <submit <at> debbugs.gnu.org>
Subject: emacs -Q -fn "nonexistent" hangs
Date: Sat, 10 Jan 2009 03:42:07 +0100

Package: emacs
Version: 23.0.60
X-Debbugs-CC: monnier <at> iro.umontreal.ca

[Note: This is *not* bug#1548, which apparently was just fixed by Jason.]

Passing a nonexistent font/fontset in the command line makes Emacs hang.

emacs -Q -fn "Courier Old"

(where Courier Old does not exist, of course), and Emacs hangs. It
does not consume CPU, just sits idle and does not respond to C-c.

I don't know whether the bug is w32-specific. It disappears when the
attached commit (by Stefan) is removed.

Attaching gdb to the hungup Emacs I get this backtrace:

#0  0x7c91e4f4 in ntdll!LdrAccessResource () from C:\WINDOWS\system32\ntdll.dll
#1  0x7c91df2c in ntdll!ZwWaitForMultipleObjects () from
C:\WINDOWS\system32\ntdll.dll
#2  0x7c809574 in KERNEL32!CreateFileMappingA () from
C:\WINDOWS\system32\kernel32.dll
#3  0x7e3995f9 in USER32!GetLastInputInfo () from C:\WINDOWS\system32\user32.dll
#4  0x7e3996a8 in USER32!MsgWaitForMultipleObjects () from
C:\WINDOWS\system32\user32.dll
#5  0x010e2979 in sys_select (nfds=1, rfds=0x82f558, wfds=0x0,
efds=0x0, timeout=0x82f550) at w32proc.c:1271
#6  0x010d985a in wait_reading_process_output (time_limit=0,
microsecs=0, read_kbd=-1, do_display=1, wait_for_cell=49018881,
    wait_proc=0x0, just_wait_proc=0) at process.c:4818
#7  0x0101094f in kbd_buffer_get_event (kbp=0x82f6e0,
used_mouse_menu=0x82fa24, end_time=0x0) at keyboard.c:4052
#8  0x0100d499 in read_char (commandflag=1, nmaps=2, maps=0x82f860,
prev_event=49018881, used_mouse_menu=0x82fa24, end_time=0x0)
    at keyboard.c:3012
#9  0x0101ebac in read_key_sequence (keybuf=0x82fc48, bufsize=30,
prompt=49018881, dont_downcase_last=0, can_return_switch_frame=1,
    fix_current_buffer=1) at keyboard.c:9359
#10 0x01007a53 in command_loop_1 () at keyboard.c:1632
#11 0x0103b2ba in internal_condition_case (bfun=0x1007228
<command_loop_1>, handlers=49082561, hfun=0x10069b4 <cmd_error>) at
eval.c:1511
#12 0x01006e37 in command_loop_2 () at keyboard.c:1349
#13 0x0103ad13 in internal_catch (tag=49078681, func=0x1006e17
<command_loop_2>, arg=49018881) at eval.c:1247
#14 0x01006dee in command_loop () at keyboard.c:1328
#15 0x0100610b in recursive_edit_1 () at keyboard.c:942
#16 0x010065e0 in Frecursive_edit () at keyboard.c:1004
#17 0x01002a71 in main (argc=4, argv=0xa926d0) at emacs.c:1786

    Juanma


commit 4c7b4c352abdd735268f9c876bd298fe2eb0cdf8
Author: Stefan Monnier <monnier <at> cs.yale.edu>
Date:   Sun Dec 21 04:13:46 2008 +0000

    (cmd_error_internal): Don't exit in daemon mode, bug#1310.

--- a/src/keyboard.c
+++ b/src/keyboard.c
@@ -1265,7 +1265,18 @@ cmd_error_internal (data, context)
   /* If the window system or terminal frame hasn't been initialized
      yet, or we're not interactive, write the message to stderr and exit.  */
   else if (!sf->glyphs_initialized_p
-	   || FRAME_INITIAL_P (sf)
+	   /* We used to check if "This is the case of the frame dumped with
+              Emacs, when we're running under a window system" with
+	        || (!NILP (Vwindow_system) && !inhibit_window_system
+	            && FRAME_TERMCAP_P (sf))
+	      then the multi-tty code generalized this check to
+	        || FRAME_INITIAL_P (sf)
+	      but this leads to undesirable behavior in daemon mode where
+	      we don't want to exit just because we got an error without
+	      having a frame (bug#1310).
+	      So I just removed the check, and rely instead on the `message_*'
+	      functions properly using FRAME_INITIAL_P.  In the worst case
+	      this should just make Emacs not exit when it should.  */
 	   || noninteractive)
     {
       print_error_message (data, Qexternal_debugging_output,

Message #10 received at 1836 <at> emacsbugs.donarmstrong.com (full text, mbox):

From: Jason Rumney <jasonr <at> gnu.org>
To: Juanma Barranquero <lekktu <at> gmail.com>, 1836 <at> debbugs.gnu.org
Subject: Re: bug#1836: emacs -Q -fn "nonexistent" hangs
Date: Sat, 10 Jan 2009 11:19:52 +0800

Juanma Barranquero wrote:
> Package: emacs
> Version: 23.0.60
> X-Debbugs-CC: monnier <at> iro.umontreal.ca
>
> [Note: This is *not* bug#1548, which apparently was just fixed by Jason.]
>   

Yes, I haven't yet marked 1548 as fixed, because I came across this, but 
this one seems to exist on GNU/Linux too so I was going to open a 
separate bug report about it before closing 1548.

What seems to be happening is that an error is thrown during frame 
creation ("No fonts match `nonexistent'" - the same error that Emacs 
22.3 produces), but the top-level lisp loop catches this error and 
continues executing the main lisp loop without ever having created a frame.

My guess was either the multi-tty or daemon changes had introduced this, 
your investigations seem to point to the latter.

> +	   /* We used to check if "This is the case of the frame dumped with
> +              Emacs, when we're running under a window system" with
> +	        || (!NILP (Vwindow_system) && !inhibit_window_system
> +	            && FRAME_TERMCAP_P (sf))
> +	      then the multi-tty code generalized this check to
> +	        || FRAME_INITIAL_P (sf)
> +	      but this leads to undesirable behavior in daemon mode where
> +	      we don't want to exit just because we got an error without
> +	      having a frame (bug#1310).
> +	      So I just removed the check, and rely instead on the `message_*'
> +	      functions properly using FRAME_INITIAL_P.  In the worst case
> +	      this should just make Emacs not exit when it should.  *


Do we have a global flag to indicate that Emacs was started in daemon 
mode that we could use here? How do we decide whether to exit when the 
last frame is closed for example?

Message #15 received at 1836 <at> emacsbugs.donarmstrong.com (full text, mbox):

From: "Juanma Barranquero" <lekktu <at> gmail.com>
To: "Jason Rumney" <jasonr <at> gnu.org>
Cc: 1836 <at> debbugs.gnu.org
Subject: Re: bug#1836: emacs -Q -fn "nonexistent" hangs
Date: Sat, 10 Jan 2009 04:26:20 +0100

On Sat, Jan 10, 2009 at 04:19, Jason Rumney <jasonr <at> gnu.org> wrote:

> Do we have a global flag to indicate that Emacs was started in daemon mode
> that we could use here?

There's a C function Fdaemonp, which uses IS_DAEMON, from lisp.h:

/* Pipe used to send exit notification to the daemon parent at
   startup.  */
extern int daemon_pipe[2];
#define IS_DAEMON (daemon_pipe[1] != 0)

    Juanma

Message #20 received at 1836 <at> emacsbugs.donarmstrong.com (full text, mbox):

From: Jason Rumney <jasonr <at> gnu.org>
To: Juanma Barranquero <lekktu <at> gmail.com>
Cc: 1836 <at> debbugs.gnu.org
Subject: Re: bug#1836: emacs -Q -fn "nonexistent" hangs
Date: Sat, 10 Jan 2009 11:53:56 +0800

The following change seems to fix the problem for me, but I haven't 
tested daemon mode, since I don't know the circumstances in which there 
was a problem in that case (surely a non-existent font should still 
cause an error in daemon mode, since it will prevent frames from being 
created later).


*** keyboard.c.~1.988.~    2009-01-10 00:01:32.703125000 +0800
--- keyboard.c    2009-01-10 11:50:39.984375000 +0800
***************
*** 1265,1282 ****
   /* If the window system or terminal frame hasn't been initialized
      yet, or we're not interactive, write the message to stderr and 
exit.  */
   else if (!sf->glyphs_initialized_p
!        /* We used to check if "This is the case of the frame dumped with
!               Emacs, when we're running under a window system" with
!             || (!NILP (Vwindow_system) && !inhibit_window_system
!                 && FRAME_TERMCAP_P (sf))
!           then the multi-tty code generalized this check to
!             || FRAME_INITIAL_P (sf)
!           but this leads to undesirable behavior in daemon mode where
!           we don't want to exit just because we got an error without
!           having a frame (bug#1310).
!           So I just removed the check, and rely instead on the `message_*'
!           functions properly using FRAME_INITIAL_P.  In the worst case
!           this should just make Emacs not exit when it should.  */
        || noninteractive)
     {
       print_error_message (data, Qexternal_debugging_output,
--- 1265,1271 ----
   /* If the window system or terminal frame hasn't been initialized
      yet, or we're not interactive, write the message to stderr and 
exit.  */
   else if (!sf->glyphs_initialized_p
!        || (!IS_DAEMON && FRAME_INITIAL_P (sf))
        || noninteractive)
     {
       print_error_message (data, Qexternal_debugging_output,

Message #25 received at 1836 <at> emacsbugs.donarmstrong.com (full text, mbox):

From: "Juanma Barranquero" <lekktu <at> gmail.com>
To: "Jason Rumney" <jasonr <at> gnu.org>
Cc: 1836 <at> debbugs.gnu.org
Subject: Re: bug#1836: emacs -Q -fn "nonexistent" hangs
Date: Sat, 10 Jan 2009 05:03:14 +0100

On Sat, Jan 10, 2009 at 04:53, Jason Rumney <jasonr <at> gnu.org> wrote:

> The following change seems to fix the problem for me

It works for me, too.

> but I haven't tested
> daemon mode, since I don't know the circumstances in which there was a
> problem in that case

??  Stefan's change was a fix for bug#1310, so the circumstances are
described in the bug report.

> (surely a non-existent font should still cause an error
> in daemon mode, since it will prevent frames from being created later).

Yes, I think so.

    Juanma

Message #30 received at 1836 <at> emacsbugs.donarmstrong.com (full text, mbox):

From: Jason Rumney <jasonr <at> gnu.org>
To: Juanma Barranquero <lekktu <at> gmail.com>
Cc: 1836 <at> debbugs.gnu.org, Stefan Monnier <monnier <at> iro.umontreal.ca>,
        Dan Nicolaescu <dann <at> ics.uci.edu>
Subject: Re: bug#1836: emacs -Q -fn "nonexistent" hangs
Date: Sat, 10 Jan 2009 12:26:01 +0800

Juanma Barranquero wrote:
> ??  Stefan's change was a fix for bug#1310, so the circumstances are
> described in the bug report.
>   

So the same change I just proposed to fix this bug was first proposed by 
Dan for bug#1310, but Stefan responded thus:

   > So, any reason why above patch cannot be checked in?

   I've installed a slightly more bold patch which just removes the
   FRAME_INITIAL_P check altogether.
     


(there seem to be a lot of intermediate mails missing from that bug 
report, so it is not clear if there was a problem pointed out with Dan's 
proposed patch, or Stefan just did not foresee this type of problem and 
thought it was safe to go further - Cc'ed both for their recollection of 
events).

Message #35 received at 1836 <at> emacsbugs.donarmstrong.com (full text, mbox):

From: Stefan Monnier <monnier <at> iro.umontreal.ca>
To: Jason Rumney <jasonr <at> gnu.org>
Cc: Juanma Barranquero <lekktu <at> gmail.com>, 1836 <at> debbugs.gnu.org,
        Dan Nicolaescu <dann <at> ics.uci.edu>
Subject: Re: bug#1836: emacs -Q -fn "nonexistent" hangs
Date: Sat, 10 Jan 2009 18:13:50 -0500

>    I've installed a slightly more bold patch which just removes the
>    FRAME_INITIAL_P check altogether.
> (there seem to be a lot of intermediate mails missing from that bug report,
> so it is not clear if there was a problem pointed out with Dan's proposed
> patch, or Stefan just did not foresee this type of problem and thought it
> was safe to go further - Cc'ed both for their recollection of events).

No, I just thought I'd try my luck.  Also I figured that if the extra
check was needed, we'd then learn why.  And indeed, now we learned why.
So when you re-add it, please make sure you add a comment that explains
the circumstance in which it's needed.


        Stefan

Message #40 received at 1836-done <at> emacsbugs.donarmstrong.com (full text, mbox):

From: Jason Rumney <jasonr <at> gnu.org>
To: Stefan Monnier <monnier <at> iro.umontreal.ca>
Cc: Juanma Barranquero <lekktu <at> gmail.com>,
        1836-done <at> debbugs.gnu.org,
        Dan Nicolaescu <dann <at> ics.uci.edu>
Subject: Re: bug#1836: emacs -Q -fn "nonexistent" hangs
Date: Sun, 11 Jan 2009 21:30:22 +0800

Stefan Monnier wrote:
> No, I just thought I'd try my luck.  Also I figured that if the extra
> check was needed, we'd then learn why.  And indeed, now we learned why.
> So when you re-add it, please make sure you add a comment that explains
> the circumstance in which it's needed.
>   

OK, I've written a rather lengthy comment to go with the change, 
explaining both why we want to exit when not in daemon mode, and why we 
don't want to in daemon mode.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.