[Kde-perl] Request for patches; Re: Massive PerlQt memory leakage

[Ole Christensen]

Ashley,

yes! Here is my private collection.

1. Fix for problem with overloading Qt::ToolTip::maybeTip
=========================================================

I got this patch from Richard Dale. This is his corresponding comment:

  * If a class doesn't have a virtual destructor, then no mapping was being
    kept from the C++ instance to the corresponding ruby value. If the class
    had virtual method callbacks, this meant that the ruby instance couldn't
    be found, and the callback couldn't be made.

  * Hence, the Qt::ToolTip callback in examples/qt-examples/tooltip didn't
    work, as that class doesn't have a virtual destructor.

--- snip  ---
*** old/Qt.xs	2004-10-30 10:11:42.741808984 +0200
--- new/Qt.xs	2004-10-30 10:20:17.961483688 +0200
***************
*** 1822,1830 ****
       if(!o)
           XSRETURN_EMPTY;
       SmokeClass c( o->smoke, o->classId );
!     if(!c.hasVirtual() ) {
! 	XSRETURN_EMPTY;
!    }
       mapPointer(obj, o, pointer_map, o->classId, 0);

   bool
--- 1822,1830 ----
       if(!o)
           XSRETURN_EMPTY;
       SmokeClass c( o->smoke, o->classId );
!  #   if(!c.hasVirtual() ) {
!  #	XSRETURN_EMPTY;
!  #  }
       mapPointer(obj, o, pointer_map, o->classId, 0);

   bool
--- snip  ---

2. Fix for methods returning non-scalar
=======================================

I got this patch from Germain Garand. I think this patch was intended to 
go into CVS anyway. Works fine for me.

--- snip  ---
Index: Qt.xs
===================================================================
RCS file: /cvsroot/perlqt/PerlQt-3/PerlQt/Qt.xs,v
retrieving revision 1.87
diff -u -3 -p -r1.87 Qt.xs
--- Qt.xs	29 Mar 2004 20:31:31 -0000	1.87
+++ Qt.xs	23 Nov 2004 07:50:55 -0000
@@ -993,17 +993,16 @@ XS(XS_AUTOLOAD) {
  	    sv_this = newSVsv(ST(0));
  	}

-	ENTER;
-	SAVETMPS;
  	PUSHMARK(SP - items + withObject);
  	PUTBACK;
-	int count = call_sv((SV*)GvCV(gv), G_SCALAR|G_EVAL);
+        I32 gimme = GIMME_V;
+	int count = call_sv((SV*)GvCV(gv), gimme|G_EVAL);
  	SPAGAIN;
-	SV *ret = newSVsv(TOPs);
-	SP -= count;
-	PUTBACK;
-	FREETMPS;
-	LEAVE;
+        SP -= count;
+        if (withObject) // shift the stack
+            for (int i=0; i<count; i++)
+                ST(i) = ST(i+1);
+        PUTBACK;

  	if(withObject && !isSuper) {
  	    SvREFCNT_dec(sv_this);
@@ -1014,8 +1013,10 @@ XS(XS_AUTOLOAD) {

          if(SvTRUE(ERRSV))
              croak(SvPV_nolen(ERRSV));
-	ST(0) = sv_2mortal(ret);
-	XSRETURN(1);
+        if (gimme == G_VOID)
+            XSRETURN_UNDEF;
+        else
+	    XSRETURN(count);
      }
      else if(!strcmp(method, "DESTROY")) {
          SV *old_this;
--- snip  ---

Mail traffic for both problems is on the list's archive: Oct/Nov 2004.

Thanx for your effort. PerlQt is worth the trouble!


Regards,
Ole.


Ashley Winters wrote:
> Agreed. Are there any other patches which need applying before a
> minimal bugfix release?
> 
> Ashley
> 
> --- Gary Greene <greeneg at phoenuxos.com> wrote:
> 
> 
>>On Tuesday 05 July 2005 06:00 am, Marius Kjeldahl wrote:
>>
>>>Gary Greene wrote:
>>>
>>>>I can confirm this as Germain sent me a patch for it a little
>>
>>over a year
>>
>>>>ago when I packaged it for another distribution. The patch is
>>
>>attached.
>>
>>>>
>>>>
>>>>
> ------------------------------------------------------------------------
> 
>>>>--- PerlQt/Qt.xs~       2004-06-01 10:12:35.000000000 +0100
>>>>+++ PerlQt/Qt.xs        2004-06-01 10:12:35.000000000 +0100
>>>>@@ -152,6 +152,7 @@
>>>>        SvREFCNT_dec(keysv);
>>>>        return 0;
>>>>     }
>>>>+    SvREFCNT_dec(keysv);
>>>>     return *svp;
>>>> }
>>>
>>>Patch looks golden and seems to solve memory leakage at my end. I
>>
>>would
>>
>>>suggest somebody with a clue puts out a new release with it
>>
>>included, to
>>
>>>save others from struggling with leakage as we had to go through.
>>>Hopefully, Google will also pick this up. Not sure why I never
>>
>>found any
>>
>>>references to it when I tried (and I searched a lot).
>>>
>>>Thanks,
>>>
>>>Marius K.
>>>_______________________________________________
>>>Kde-perl mailing list
>>>Kde-perl at kde.org
>>>https://mail.kde.org/mailman/listinfo/kde-perl
>>
>>Agreed, this should warrant at least a 3.0009 at least.
>>
>>-- 
>>Gary L. Greene, Jr.
>>Sent from uriel.tolharadys.net
>> 07:22:26 up 11:47,  3 users,  load average: 0.34, 0.41, 0.45
>> 
>>============================================================
>>Developer and Project Lead for PhoeNUX OS.
>> check out http://www.phoenuxos.com/ for more info.
>>EMAIL : greeneg at phoenuxos.com
>>============================================================
>> 
>>
>>>_______________________________________________
>>
>>Kde-perl mailing list
>>Kde-perl at kde.org
>>https://mail.kde.org/mailman/listinfo/kde-perl
>>
> 
> 
> 
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam protection around 
> http://mail.yahoo.com 
> _______________________________________________
> Kde-perl mailing list
> Kde-perl at kde.org
> https://mail.kde.org/mailman/listinfo/kde-perl