unofficial mirror of guile-devel@gnu.org 
 help / color / mirror / Atom feed
From: ludovic.courtes@laas.fr (Ludovic Courtès)
Cc: Han-Wen Nienhuys <hanwen@xs4all.nl>, guile-devel@gnu.org
Subject: Re: [PATCH] Reference leak in `iprin1 ()'
Date: Mon, 14 Nov 2005 10:58:46 +0100	[thread overview]
Message-ID: <874q6fsg89.fsf@laas.fr> (raw)
In-Reply-To: <87zmoa2pad.fsf@ossau.uklinux.net> (Neil Jerram's message of "Sat, 12 Nov 2005 09:23:06 +0000")

Hello,

Neil Jerram <neil@ossau.uklinux.net> writes:

> Nice work, but it looks to me that PUSH_REF sets the value of the
> (pstate->top)th element _before_ incrementing pstate->top.  So
> shouldn't your fix do the decrement first and then set the slot to
> undefined?

Yes, you're perfectly right.

> (Note that your fix will still prevent most reference leaks, just not
> the outermost one.  So that's why it may appear to be working
> already.)

Exactly.

> Also, is the () -> (void) change strictly related?

In fact, it's strictly unrelated, but while we're at it (really `(void)'
was meant here, not `()')...

Below is an updated patch.  I modified `PUSH_REF ()' as well so that it
does PSTATE->TOP++ _after_ using the `PSTATE_STACK_SET ()' macro: this
is safer.  And, well, I couldn't resist the desire to "beautifully
backslashify" `ENTER_NESTED_DATA ()' as well.  :-)  I hope this is not
too much pollution for you.

Thanks,
Ludovic.


2005-11-14  Ludovic Courtès  <ludovic.courtes@laas.fr>

	* print.c (EXIT_NESTED_DATA): Before popping from the stack,
	reset the value at its top.  This fixes a reference leak.
	(PUSH_REF): Perform `pstate->top++' after calling
	`PSTATE_STACK_SET ()' in order to avoid undesired side effects.
	(scm_current_pstate): Replaced `()' by `(void)'.


--- orig/libguile/print.c
+++ mod/libguile/print.c
@@ -112,31 +112,40 @@
  * time complexity (O (depth * N)), The printer code can be
  * rewritten to be O(N).
  */
-#define PUSH_REF(pstate, obj) \
-do { \
-  PSTATE_STACK_SET (pstate, pstate->top++, obj); \
-  if (pstate->top == pstate->ceiling) \
-    grow_ref_stack (pstate); \
+#define PUSH_REF(pstate, obj)			\
+do						\
+{						\
+  PSTATE_STACK_SET (pstate, pstate->top, obj);	\
+  pstate->top++;				\
+  if (pstate->top == pstate->ceiling)		\
+    grow_ref_stack (pstate);			\
 } while(0)
 
-#define ENTER_NESTED_DATA(pstate, obj, label) \
-do { \
-  register unsigned long i; \
-  for (i = 0; i < pstate->top; ++i) \
-    if (scm_is_eq (PSTATE_STACK_REF (pstate, i), (obj))) \
-      goto label; \
-  if (pstate->fancyp) \
-    { \
-      if (pstate->top - pstate->list_offset >= pstate->level) \
-	{ \
-	  scm_putc ('#', port); \
-	  return; \
-	} \
-    } \
-  PUSH_REF(pstate, obj); \
+#define ENTER_NESTED_DATA(pstate, obj, label)			\
+do								\
+{								\
+  register unsigned long i;					\
+  for (i = 0; i < pstate->top; ++i)				\
+    if (scm_is_eq (PSTATE_STACK_REF (pstate, i), (obj)))	\
+      goto label;						\
+  if (pstate->fancyp)						\
+    {								\
+      if (pstate->top - pstate->list_offset >= pstate->level)	\
+	{							\
+	  scm_putc ('#', port);					\
+	  return;						\
+	}							\
+    }								\
+  PUSH_REF(pstate, obj);					\
 } while(0)
 
-#define EXIT_NESTED_DATA(pstate) { --pstate->top; }
+#define EXIT_NESTED_DATA(pstate)				\
+do								\
+{								\
+  --pstate->top;						\
+  PSTATE_STACK_SET (pstate, pstate->top, SCM_UNDEFINED);	\
+}								\
+while (0)
 
 SCM scm_print_state_vtable = SCM_BOOL_F;
 static SCM print_state_pool = SCM_EOL;
@@ -144,8 +153,8 @@
 
 #ifdef GUILE_DEBUG /* Used for debugging purposes */
 
-SCM_DEFINE (scm_current_pstate, "current-pstate", 0, 0, 0, 
-           (),
+SCM_DEFINE (scm_current_pstate, "current-pstate", 0, 0, 0,
+	    (void),
 	    "Return the current-pstate -- the car of the\n"
 	    "@code{print_state_pool}.  @code{current-pstate} is only\n"
 	    "included in @code{--enable-guile-debug} builds.")




_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel


  reply	other threads:[~2005-11-14  9:58 UTC|newest]

Thread overview: 35+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2005-11-08 17:22 [PATCH] Marking weak alist vectors Ludovic Courtès
2005-11-08 23:51 ` Marius Vollmer
2005-11-09  9:03   ` Ludovic Courtès
2005-12-06 23:55     ` Marius Vollmer
2005-12-07 10:33       ` Ludovic Courtès
2005-12-13 23:45         ` Marius Vollmer
2005-12-14  9:30           ` Ludovic Courtès
2005-11-09 10:28 ` Han-Wen Nienhuys
2005-11-09 16:28   ` Ludovic Courtès
2005-11-09 18:36     ` Han-Wen Nienhuys
2005-11-09 21:11     ` Kevin Ryde
2005-11-09 22:45       ` Marius Vollmer
2005-11-10 12:11         ` Han-Wen Nienhuys
2005-11-10  9:47       ` [PATCH] Reference leak in `iprin1 ()' Ludovic Courtès
2005-11-12  9:23         ` Neil Jerram
2005-11-14  9:58           ` Ludovic Courtès [this message]
2005-11-16 21:18             ` Neil Jerram
2005-11-17  9:54               ` Ludovic Courtès
2005-11-17 18:52                 ` Neil Jerram
2005-11-23 10:19     ` [PATCH] Marking weak alist vectors, #2 Ludovic Courtès
2005-11-24  0:59       ` Han-Wen Nienhuys
2005-11-24  9:01         ` Ludovic Courtès
2005-11-26  0:49       ` Kevin Ryde
2006-01-09 14:51       ` [PATCH] Marking weak alist vectors, epilogue Ludovic Courtès
2006-01-09 19:29         ` Neil Jerram
2006-01-10  8:21           ` Ludovic Courtès
2006-01-10  9:33             ` Neil Jerram
2006-01-10 15:43               ` Ludovic Courtès
2005-11-17 13:21 ` [PATCH] Fixing `gc-live-object-stats' Ludovic Courtès
2005-11-17 14:12   ` Han-Wen Nienhuys
2005-11-30  8:54   ` Ludovic Courtès
2005-11-30 23:45     ` Han-Wen Nienhuys
2005-12-03 19:31       ` Neil Jerram
2005-12-05  8:50         ` Ludovic Courtès
2005-12-06 19:14           ` Neil Jerram

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: https://www.gnu.org/software/guile/

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=874q6fsg89.fsf@laas.fr \
    --to=ludovic.courtes@laas.fr \
    --cc=guile-devel@gnu.org \
    --cc=hanwen@xs4all.nl \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).