Re: fg/bg on FreeBSD.

[Tanaka Akira <akr@m17n.org>]

In article <hvozoq3mux1.fsf@serein.m17n.org>,
  Tanaka Akira <akr@m17n.org> writes:

> Possibly.  But NetBSD has no problem...

I found the difference between FreeBSD and NetBSD.

FreeBSD fails to kill(-PGID,0) for zombie process but NetBSD (and
Linux) succeeds.

FreeBSD:

> | Z:akr@dhcp21% kdump|egrep 'kill|fork|exit|setpgid'
> |  29256 zsh      CALL  fork                    # zsh: fork for echo
> |  29256 zsh      RET   fork 29257/0x7249
> |  29257 zsh      RET   fork 0
> |  29257 zsh      CALL  setpgid(0,0x7249)       # echo: setpgid(0,29257)
> |  29257 zsh      RET   setpgid 0
> |  29256 zsh      CALL  fork                    # zsh: fork for sleep
> |  29256 zsh      RET   fork 29258/0x724a
> |  29257 zsh      CALL  exit(0)                 # echo: exit(0)

29257(echo) is zombie now.

> |  29258 zsh      RET   fork 0

> |  29258 zsh      CALL  kill(0xffff8db7,0)
> |  29258 zsh      RET   kill -1 errno 3 No such process

The process to be sleep command tries to check a leader is exists...
0xffff8db7 is -29257.  kill(-29257,0) is failed.

> |  29258 zsh      CALL  setpgid(0,0x724a)       # sleep: setpgid(0,29258)
> |  29258 zsh      RET   setpgid 0

So, it decides to be a leader itself.

NetBSD:

> | Z:akr@netbsd% kdump|egrep 'kill|fork|exit|setpgid'
> |  16519 zsh      CALL  fork                    # zsh: fork for echo
> |  16519 zsh      RET   fork 16520/0x4088
> |  16520 zsh      RET   fork 0
> |  16520 zsh      CALL  setpgid(0,0x4088)
> |  16520 zsh      RET   setpgid 0
> |  16520 zsh      CALL  exit(0)                 # echo: exit(0)

16520(echo) is zombie now.

> |  16519 zsh      CALL  fork                    # zsh: fork for sleep
> |  16519 zsh      RET   fork 16521/0x4089
> |  16521 zsh      RET   fork 0
> |  16521 zsh      CALL  kill(0xffffbf78,0)
> |  16521 zsh      RET   kill 0

The process to be sleep command tries to check a leader is exists...
0xffffbf78 is -16520.  kill(-16520,0) is succeed.

> |  16521 zsh      CALL  setpgid(0,0x4088)       # sleep: setpgid(0,16520)
> |  16521 zsh      RET   setpgid 0

So, it comes into the process group of echo.

I confirmed following modification prevents the problem.  But
definitely this is too radical...

Index: Src/exec.c
===================================================================
RCS file: /cvsroot/zsh/zsh/Src/exec.c,v
retrieving revision 1.7
diff -u -r1.7 exec.c
--- Src/exec.c  2000/05/04 13:40:05     1.7
+++ Src/exec.c  2000/05/06 19:53:20
@@ -2466,7 +2466,7 @@
        }
     } else if (thisjob != -1 && cl) {
        if (jobtab[list_pipe_job].gleader && (list_pipe || list_pipe_child)) {
-           if (killpg(jobtab[list_pipe_job].gleader, 0) == -1 ||
+           if (/* killpg(jobtab[list_pipe_job].gleader, 0) == -1 || */
                setpgrp(0L, jobtab[list_pipe_job].gleader) == -1) {
                jobtab[list_pipe_job].gleader =
                    jobtab[thisjob].gleader = (list_pipe_child ? mypgrp : getpid());
-- 
Tanaka Akira