bug#2403: And the guilty is: Garbage Collecting leak

[Pierre Poissinger <pierre.poissinger@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1416 bytes --]

> Than you very much, installed.
You are welcome

> Other arches don't use GCPROs at all, instead they conservatively scan
> the whole stack.  They do that by setting
>
>   #define GC_MARK_STACK GC_MAKE_GCPROS_NOOPS
>
> in the corresponding emacs/src/s/*.h.  This is the preferred way to
> function, so if you can test it and confirm that it works as well under
> AIX, we could set it for that OS as well.
Oups - I actually didn't look so far - readed that same bug hit
GNU/Linux on Super7 so didn't look further...
Since I was in a good mood (Emacs 23 working on my boxes) I took the
liberty to also define GC_SETJMP_WORKS since I saw it next to
GC_MARK_STACK [and ok, I felt lucky...]

So far so good - full build/bootstrapping went fine from latest git
master - basic testing don't show regressions - will see tomorrow if
something fishy pops during work, otherwise looks fine.

In case of need: the compiler used is a gcc-2.9-aix51 on a good old
aix 5.2 - didn't have the courage to try the outdated xlc...

Notes:
* To be able to build emacs with X support (I was really in a good
mood): Had to correct 2 files with "old compilers are grumpy" errors
[0001-gcc-2.9-aix51-020209-compile-fix.patch]
* The actual change I made to src/s/aix-4-2.h
[0002-GC-test-for-AIX.patch]

Regs,
Pierre
-- 
>>> horsemen = ['war', 'pestilence', 'famine']
>>> horsemen.append('Powerbuilder')

[-- Attachment #2: 0001-gcc-2.9-aix51-020209-compile-fix.patch --]
[-- Type: text/x-patch, Size: 1488 bytes --]

From b543f5c64a2456722b91c29016027dc958d39677 Mon Sep 17 00:00:00 2001
From: Pierre Poissinger <pierre.poissinger@gmail.com>
Date: Mon, 30 Mar 2009 17:10:35 -0400
Subject: [PATCH 1/2] gcc 2.9-aix51-020209 compile fix

---
 src/frame.c |    3 ++-
 src/xterm.c |    3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/src/frame.c b/src/frame.c
index 1a11021..6408571 100644
--- a/src/frame.c
+++ b/src/frame.c
@@ -3370,6 +3370,7 @@ x_set_font (f, arg, oldval)
   Lisp_Object frame;
   int fontset = -1;
   Lisp_Object font_object;
+  Lisp_Object lval;
 
   /* Set the frame parameter back to the old value because we may
      fail to use ARG as the new parameter value.  */
@@ -3427,7 +3428,7 @@ x_set_font (f, arg, oldval)
     return;
 
   
-  Lisp_Object lval = Fassq (Qfullscreen, f->param_alist);
+  lval = Fassq (Qfullscreen, f->param_alist);
   if (CONSP (lval)) lval = CDR (lval);
 
   x_new_font (f, font_object, fontset);
diff --git a/src/xterm.c b/src/xterm.c
index 76beb62..8444061 100644
--- a/src/xterm.c
+++ b/src/xterm.c
@@ -8675,6 +8675,7 @@ x_handle_net_wm_state (f, event)
   Display *dpy = FRAME_X_DISPLAY (f);
   unsigned char *tmp_data = NULL;
   Atom target_type = XA_ATOM;
+  Lisp_Object lval;
 
   BLOCK_INPUT;
   x_catch_errors (dpy);
@@ -8704,7 +8705,7 @@ x_handle_net_wm_state (f, event)
         value |= FULLSCREEN_BOTH;
     }
 
-  Lisp_Object lval = Qnil;
+  lval = Qnil;
   switch (value) 
     {
     case FULLSCREEN_WIDTH:
-- 
1.6.2.1


[-- Attachment #3: 0002-GC-test-for-AIX.patch --]
[-- Type: text/x-patch, Size: 722 bytes --]

From 8293f6bec069a4c00aa20202173990597ab84ef4 Mon Sep 17 00:00:00 2001
From: Pierre Poissinger <pierre.poissinger@gmail.com>
Date: Mon, 30 Mar 2009 17:12:12 -0400
Subject: [PATCH 2/2] GC test for AIX

---
 src/s/aix4-2.h |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/src/s/aix4-2.h b/src/s/aix4-2.h
index 89154ee..c6e806c 100644
--- a/src/s/aix4-2.h
+++ b/src/s/aix4-2.h
@@ -168,5 +168,9 @@ along with GNU Emacs.  If not, see <http://www.gnu.org/licenses/>.  */
 */
 #define BROKEN_GET_CURRENT_DIR_NAME 1
 
+/* test */
+#define GC_SETJMP_WORKS 1
+#define GC_MARK_STACK GC_MAKE_GCPROS_NOOPS
+
 /* arch-tag: 38fe75ea-6aef-42bd-8449-bc34d921a562
    (do not change this comment) */
-- 
1.6.2.1