[bjj@sequent.com: memleak in tasks.c]

------- Start of forwarded message -------
Return-Path: nop@ccs.neu.edu
X-Authentication-Warning: eng4.sequent.com: Host localhost didn't use HELO protocol
To: moo-cows@the-b.org
Subject: memleak in tasks.c
Date: Thu, 22 May 97 16:39:16 PDT
From: Ben Jackson <bjj@sequent.com>
Sender: owner-moo-cows@the-b.org
Precedence: bulk
Resent-From: clue-cows <nop@nop.com>
Errors-To: clue-cows <nop@nop.com>

Jay found a particular structure element that seemed to be leaking and
despite my disbelief I ran across it while reading tasks.c the next day.

The leak happens when you kill_task a forked task which has not yet
started to run.  One way to do this would be:

	;;fork t (0) endfork ; kill_task(t);

If this happens in a verb (rather than an eval) the strings that would
be "leaked" would not actually be freed anyway (they would still be
referenced by the object or the verb code) so the actual leak wouldn't
happen until you decided to @rmverb or @program the verb, which may
never happen in the life of the server.  It's pretty hard to make the
server grow by any significant amount with this bug.

Note that this is *not* "r2" specific, but the patch below is against
r2 and may not apply cleanly to p6.  I trust if that's the case you can
add the two lines by hand.  :-)

- --Ben

*** tasks.c	1997/03/08 06:25:43	1.3
- --- tasks.c	1997/05/21 03:41:34	1.3.2.1
***************
*** 319,327 ****
  	free_str(t->t.input.string);
  	break;
      case TASK_FORKED:
! 	if (strong)
  	    free_rt_env(t->t.forked.rt_env,
  			t->t.forked.program->num_var_names);
  	free_program(t->t.forked.program);
  	break;
      case TASK_SUSPENDED:
- --- 319,330 ----
  	free_str(t->t.input.string);
  	break;
      case TASK_FORKED:
! 	if (strong) {
  	    free_rt_env(t->t.forked.rt_env,
  			t->t.forked.program->num_var_names);
+ 	    free_str(t->t.forked.a.verb);
+ 	    free_str(t->t.forked.a.verbname);
+ 	}
  	free_program(t->t.forked.program);
  	break;
      case TASK_SUSPENDED:
***************
*** 1934,1942 ****
      register_function("flush_input", 1, 2, bf_flush_input, TYPE_OBJ, TYPE_ANY);
  }
  
! char rcsid_tasks[] = "$Id: tasks.c,v 1.3 1997/03/08 06:25:43 nop Exp $";
  
  /* $Log: tasks.c,v $
   * Revision 1.3  1997/03/08 06:25:43  nop
   * 1.8.0p6 merge by hand.
   *
- --- 1937,1948 ----
      register_function("flush_input", 1, 2, bf_flush_input, TYPE_OBJ, TYPE_ANY);
  }
  
! char rcsid_tasks[] = "$Id: tasks.c,v 1.3.2.1 1997/05/21 03:41:34 bjj Exp $";
  
  /* $Log: tasks.c,v $
+  * Revision 1.3.2.1  1997/05/21 03:41:34  bjj
+  * Fix a memleak when a forked task was killed before it ever started.
+  *
   * Revision 1.3  1997/03/08 06:25:43  nop
   * 1.8.0p6 merge by hand.
   *
------- End of forwarded message -------

---

• Prev by Subject: [asb@media.mit.edu: signal 10 crash with 1.8.0p6]
• Next by Subject: [bmgustav@bioinformatics.weizmann.ac.il: Re: Memory leak?]
• Prev by thread: [rfc@MICROSOFT.com: RE: #define LOG COMMANDS]
• Next by thread: [egoff@combase.com: #define LOG_COMMANDS]
• Index(es):
  • Subject
  • Thread

Home |
Subject Index |
Thread Index