[minicom-Bugs][311024] segmentation fault in a corner case

[minicom-bugs at alioth.debian.org]

Bugs item #311024, was opened at 2008-08-26 16:40
Status: Open
Priority: 3
Submitted By: Paul Bolle (pebolle-guest)
Assigned to: Nobody (None)
Summary: segmentation fault in a corner case 
Category: None
Group: None
Resolution: None


Initial Comment:
0) minicom segfaults in this corner case:
- start minicom on a device (here ttyS0);
- insert a PCMCIA modem (using cs_serial);
- watch one error appear in the main window, watch the main window disappear and notice more errors appear on the clear terminal screen;
- suspend minicom on the terminal (e.g. by using ^z);
- bring minicom to the foreground.

1) screenscraping:

Welcome to minicom 2.3

OPTIONS: I18n 
Compiled on Mar 13 2008, 00:58:14.
Port /dev/ttyS0

               Press CTRL-A Z for help on special keys                      
                                                                            
Device /dev/ttyS0 access failed: No such file or directory.                 
                                                                            
[...]

CTRL-A Z for help |115200 8N1 | NOR | Minicom 2.3    | VT102 |      Offline                                                 

Device /dev/ttyS0 access failed: No such file or directory.
Device /dev/ttyS0 access failed: No such file or directory.
Device /dev/ttyS0 access failed: No such file or directory.
Device /dev/ttyS0 access failed: No such file or directory.
Device /dev/ttyS0 access failed: No such file or directory.
Device /dev/ttyS0 access failed: No such file or directory.
Device /dev/ttyS0 access failed: No such file or directory.
Device /dev/ttyS0 access failed: No such file or directory.

Suspended. Type "fg" to resume.

[1]+  Stopped                 sudo minicom ttyS0
$ fg 
sudo minicom ttyS0
Segmentation fault (core dumped)

2) backtrace and related info:

$ gdb minicom core.12991 
[...]
Core was generated by `minicom ttyS0'.
Program terminated with signal 11, Segmentation fault.
[New process 12991]
#0  mc_wreturn () at window.c:728
728           _write(e->value, -1, x, y, e->attr, e->color);
(gdb) bt
#0  mc_wreturn () at window.c:728
#1  0x0804a360 in shjump (sig=20) at minicom.c:125
#2  <signal handler called>
#3  0x00110416 in __kernel_vsyscall ()
#4  0x0081a63d in ___newselect_nocancel () from /lib/libc-2.8.so
#5  0x08061c7d in check_io (fd1=28, fd2=0, tmout=<value optimized out>, buf=0xbfa515ac "", buflen=0xbfa515a8) at ipc.c:56
#6  0x08064fc0 in do_terminal () at main.c:753
#7  0x0804b9ce in main (argc=2, argv=0xbfa52344) at minicom.c:1373
gdb) f 0
#0  mc_wreturn () at window.c:728
728	      _write(e->value, -1, x, y, e->attr, e->color);
(gdb) info locals
x = 0
y = 0
e = (ELM *) 0x0

Analysis: e == NULL because gmap == NULL; gmap was set to NULL in src/window.c/win_end(). win_end() was called by src/window.c/mc_wclose(), which was called (in this case) by src/main.c/open_term(). 

3) version etc.:
minicom version 2.3 (compiled Mar 13 2008) is minicom-2.3-2.fc9.i386 (current Fedora minicom rpm)

4) This is sort of related to https://bugzilla.redhat.com/show_bug.cgi?id=246465 Also of interest are the recent changes regarding disappearing devices: https://alioth.debian.org/plugins/scmcvs/cvsweb.php/minicom/src/main.c.diff?r1=1.21;r2=1.22;cvsroot=minicom (though here a device sort of appears during a session).

5) I've been trying to find a solution. Not sure what should be done.

My best _guess_ right now would be to remove this call to mc_wclose() in open_term():
[...]
  lockfile[0] = 0;
  if (doinit >= 0) {
       int rc = ttylock(dial_tty);
       if (rc < 0) {
               if (access(dial_tty, W_OK) == -1)
                       fprintf(stderr, _("Device %s access failed: %s.\n"),
                               dial_tty, strerror(errno));
               else
                       fprintf(stderr, _("Device %s lock failed: %s.\n"),
                               dial_tty, strerror(-rc));
       } else if (rc > 0) {
               fprintf(stderr, _("Device %s is locked.\n"), dial_tty);
       } else if (rc == 0) {
               snprintf(lockfile, sizeof(lockfile), "%s", dial_tty);
       }
       if (rc) {
               if (stdwin) mc_wclose(stdwin, 1);
               drop_privs();
               return(-1);
       }
  }
[...]

I _assume_ this would mean that:
- the main window will remain on device ;
- all errors will be printed to the main window;
- suspending (^A J) and bringing to the foreground do work as expected.

If this makes senses, I'm happy to write and test a patch.

----------------------------------------------------------------------

You can respond by visiting: 
http://alioth.debian.org/tracker/?func=detail&atid=100031&aid=311024&group_id=30018