21.1 on Solaris 2.5.1 (sparc), on terminal, core dump

[Koji Hino]

Hi,

# I am not on this list, so please CC to me if you want to contact me.

I found that Emacs 21.1 on terminal (not w/ X) dumps core when using
subprocess...

Emacs source: plain emacs-21.1.tar.gz with leim-21.1.tar.gz
Platform: Solaris 2.5.1 on sparc (Ultra5) with recent Sun's recommended patches
          NAME: Solaris 2.5.1 Recommended Patch Cluster
          DATE: Nov/15/01
Configure: env CFLAGS="-O -g" ../emacs-21.1/configure  --prefix=/usr/Local 
--without-x
CC: gcc version 2.95.3 20010315 (release)

Note that same bug arise with X compiled-in binary (no --without-x
switch to configure), if it runs under terminal with 'emacs -q -nw'.

How to coredump:

% ./src/emacs -q [ret]

C-g
 -> Bell rings, 'Quit' printed on mini buffer. good.

M-x shell [ret]
C-g
Fatal error (11).Segmentation fault

gdb backtrace:
Program received signal SIGSEGV, Segmentation fault.
0x00000000 in ?? ()
(gdb) bt
#0  0x00000000 in ?? ()
#1  <signal handler called>
#2  0xef5bce0c in poll () from /usr/lib/libc.so.1
#3  0xef5da404 in select () from /usr/lib/libc.so.1
#4  0x0010bd04 in wait_reading_process_input (time_limit=30, microsecs=0,
    read_kbd=268435455, do_display=1)
    at /hirosige/src/emacs/emacs-21.1/src/process.c:2596
#5  0x0001b9bc in sit_for (sec=30, usec=0, reading=1, display=1,
    initial_display=0) at /hirosige/src/emacs/emacs-21.1/src/dispnew.c:6243
#6  0x0006f2b8 in read_char (commandflag=1, nmaps=2, maps=0xefffee20,
    prev_event=270769156, used_mouse_menu=0xefffee48)
    at /hirosige/src/emacs/emacs-21.1/src/keyboard.c:2491
#7  0x00075cd8 in read_key_sequence (keybuf=0xefffef80, bufsize=30,
    prompt=270769156, dont_downcase_last=0, can_return_switch_frame=1,
    fix_current_buffer=1) at /hirosige/src/emacs/emacs-21.1/src/keyboard.c:8180
#8  0x0006d174 in command_loop_1 ()
    at /hirosige/src/emacs/emacs-21.1/src/keyboard.c:1440
#9  0x000d2044 in internal_condition_case (bfun=0x6cdc0 <command_loop_1>,
    handlers=270897228, hfun=0x6c880 <cmd_error>)
    at /hirosige/src/emacs/emacs-21.1/src/eval.c:1267
#10 0x0006cbe4 in command_loop_2 ()
    at /hirosige/src/emacs/emacs-21.1/src/keyboard.c:1245
#11 0x000d1b74 in internal_catch (tag=270843412,
    func=0x6cbc0 <command_loop_2>, arg=270769156)
    at /hirosige/src/emacs/emacs-21.1/src/eval.c:1030
#12 0x0006cb68 in command_loop ()
    at /hirosige/src/emacs/emacs-21.1/src/keyboard.c:1224
#13 0x0006c590 in recursive_edit_1 ()
    at /hirosige/src/emacs/emacs-21.1/src/keyboard.c:950
#14 0x0006c70c in Frecursive_edit ()
    at /hirosige/src/emacs/emacs-21.1/src/keyboard.c:1006
#15 0x0006b16c in main (argc=2, argv=0xeffff4b4, envp=0x1)
    at /hirosige/src/emacs/emacs-21.1/src/emacs.c:1547
...

It seems that there are no signal handler for SIGINT, in this case..

Workaround:
add
#define HAVE_VFORK 1
to src/config.h or src/s/sol2-5.h.
With HAVE_VFORK enabled, emacs works fine.

Consideration:
configure script says this machine has working vfork, but there are no
#define enabled from this judgement.

Here is what ./configure said:
....
checking for pid_t... yes
checking for vfork.h... no
checking for working vfork... yes
checking for size_t... yes
....

But, even without vfork, emacs should set signal handler correctly. Is
this problem come from Solaris 2.5.1's bad behavior?

Are there anyone who love terminal? :-)
# I am 'screen' lover, so I have to use Emacs with terminal..

Koji

====================================================================
Koji HINO(HINO is my family name)
C&C Research Laboratories, NEC USA, Inc.
110 Rio Robles, San Jose, CA 95134, USA
E-mail: hino@ccrl.sj.nec.com
----------
DISCLAIMER: this message is the author's personal opinion and does not
constitute the support, opinion, or policy of NEC USA, Inc.