Re: delete-process bug

[Michaël Cadilhac]

Stefan Monnier <address@hidden> writes:

>>>> Error in post-command-hook: (wrong-type-argument listp (nil DEAD . 
>>>> 18764377))
>>> 
>>> Oh yes, the live_deleted_processes variable needs to be gcpro'd.
>
>> I'm not  familiar with Emacs C code,  so I don't know  how to proceed.
>> Can you send a patch, or at least, the lines that have to be modified,
>> and how ?
>
> Add something like
>
>    staticpro (&live_deleted_processes);
>
> in syms_of_process

Works fine!

I used two tests:

  1) (original test) Quick switching between two flyspellized buffers
     that have different dictionaries,

  2) 

(let (processes)
  (dotimes (i 100)
    (push (start-process "bug" nil "/tmp/sleep") processes))
  (dotimes (i 50)
    (if (zerop (mod i 2))
        (delete-process (pop processes))
      (kill-process (pop processes))))
  (message "%S" (call-process "/tmp/sleep"))
  (dotimes (i 50)
    (delete-process (pop processes)))
  (message "%S" (call-process "/tmp/sleep")))

None of them succeeded before, but now it's OK.

(Final ?) patch attached.

Agustin, can you check that you can't trigger the bug anymore?

Index: src/process.c
===================================================================
RCS file: /sources/emacs/emacs/src/process.c,v
retrieving revision 1.481
diff -c -r1.481 process.c
*** src/process.c       8 May 2006 05:19:42 -0000       1.481
--- src/process.c       28 May 2006 20:24:11 -0000
***************
*** 778,783 ****
--- 778,792 ----
    return proc;
  }
  
+ 
+ /* Fdelete_process promises to immediately forget about the process, but in
+    reality, Emacs needs to remember those processes until they have been
+    treated by sigchld_handler; otherwise this handler would consider the
+    process as being synchronous and say that the synchronous process is
+    dead.  */
+ static Lisp_Object deleted_pid_list;
+ 
+ 
  DEFUN ("delete-process", Fdelete_process, Sdelete_process, 1, 1, 0,
         doc: /* Delete PROCESS: kill it and forget about it immediately.
  PROCESS may be a process, a buffer, the name of a process or buffer, or
***************
*** 800,805 ****
--- 809,818 ----
    else if (XINT (p->infd) >= 0)
      {
        Fkill_process (process, Qnil);
+       /* No problem storing the pid here, as it is still in Vprocess_alist.  
*/
+       deleted_pid_list = Fcons (make_fixnum_or_float (p->pid),
+                               /* GC treated elements set to nil.  */
+                               Fdelq (Qnil, deleted_pid_list));
        /* Do this now, since remove_process will make sigchld_handler do 
nothing.  */
        p->status
        = Fcons (Qsignal, Fcons (make_number (SIGKILL), Qnil));
***************
*** 6373,6378 ****
--- 6386,6400 ----
  
        /* Find the process that signaled us, and record its status.  */
  
+       /* The process can have been deleted by Fdelete_process.  */
+       tail = Fmember (make_fixnum_or_float (pid), deleted_pid_list);
+       if (!NILP (tail))
+       {
+         Fsetcar (tail, Qnil);
+         goto sigchld_end_of_loop;
+       }
+ 
+       /* Otherwise, if it is asynchronous, it is in Vprocess_alist.  */
        p = 0;
        for (tail = Vprocess_alist; GC_CONSP (tail); tail = XCDR (tail))
        {
***************
*** 6424,6431 ****
            EMACS_SET_SECS_USECS (*input_available_clear_time, 0, 0);
        }
  
!       /* There was no asynchronous process found for that id.  Check
!          if we have a synchronous process.  */
        else
        {
          synch_process_alive = 0;
--- 6446,6453 ----
            EMACS_SET_SECS_USECS (*input_available_clear_time, 0, 0);
        }
  
!       /* There was no asynchronous process found for that pid: we have
!        a synchronous process.  */
        else
        {
          synch_process_alive = 0;
***************
*** 6442,6447 ****
--- 6464,6473 ----
            EMACS_SET_SECS_USECS (*input_available_clear_time, 0, 0);
        }
  
+ 
+     sigchld_end_of_loop:
+       ;
+ 
        /* On some systems, we must return right away.
         If any more processes want to signal us, we will
         get another signal.
***************
*** 6843,6848 ****
--- 6869,6875 ----
    FD_SET (0, &input_wait_mask);
  
    Vprocess_alist = Qnil;
+   deleted_pid_list = Qnil;
    for (i = 0; i < MAXDESC; i++)
      {
        chan_process[i] = Qnil;
***************
*** 6981,6986 ****
--- 7008,7014 ----
    staticpro (&Qlast_nonmenu_event);
  
    staticpro (&Vprocess_alist);
+   staticpro (&deleted_pid_list);
  
    DEFVAR_BOOL ("delete-exited-processes", &delete_exited_processes,
               doc: /* *Non-nil means delete processes immediately when they 
exit.

-- 
 |      Michaël `Micha' Cadilhac   |   Would someone please DTRT with this  |
 |         Epita/LRDE Promo 2007   |         then ACK?                      |
 | http://www.lrde.org/~cadilh_m   |           -- Richard Stallman          |
 `--  -   JID: address@hidden --'                                   -  --'

pgp2ohWqx_HPp.pgp
Description: PGP signature