KDE/kdeedu/kstars/kstars/tools

SVN commit 976195 by prakash:

Reducing the size of the AltVsTime tool to support smaller screens.

CCMAIL: kstars-devel <at> kde.org
CCBUG: 179386

 M  +136 -135  altvstime.ui  

http://websvn.kde.org/?view=rev&revision=976195

Fwd: [Kde-cvs-announce] KDE 4.3 Beta2 tagging tomorrow

FYI.

Regards
Akarsh

From: Dirk Müller <dmueller <at> suse.de>
Subject: [Kde-cvs-announce] KDE 4.3 Beta2 tagging tomorrow
Date: 2009-06-02 20:48:51 GMT

Hi, 

We have a KDE 4.3 Beta2 tagging milestone tonight at 23:59 UTC. I will, 
assuming that trunk compiles, perform tagging tomorrow morning (12 hours from 
now).

If there are urgent outstanding issues not ready until then, please tell 
release-team <at> kde.org. 

Thanks,
Dirk
_______________________________________________
Kde-cvs-announce mailing list
Kde-cvs-announce <at> kde.org

KDE/kdeedu/kstars/kstars

SVN commit 976936 by asimha:

Version microbump.

CCMAIL: kstars-devel <at> kde.org

 M  +1 -1      main.cpp  

--- trunk/KDE/kdeedu/kstars/kstars/main.cpp #976935:976936
@@ -32,7 +32,7 @@
 #include <QPixmap>
 #include <kglobal.h>

-#define KSTARS_VERSION "1.5.1"
+#define KSTARS_VERSION "1.5.2"

 static const char description[] =
     I18N_NOOP("Desktop Planetarium");

KDE/kdeedu/kstars/kstars/skyobjects

SVN commit 977139 by khudyakov:

In this changeset clone function is implemented. It allows safe
copying of objects. E.g. KSPlanet stored as SkyObject* pointer
would be copied as KSPlanet and not sliced. 

For all subclasses of SkyObject but DeepSkyObject copy
constructors are removed. In most cases they are pretty useless
and could be replaced with default implementations. 

* Changes in copy constructors
 - Trail in TrailObject wasn't copied. Now it's copied
 - const annotation added in DeepSkyObject copy constructor

* Overloaded assignment operator in EclipticPosition
  (skyobjects/ksplanetbase.h) removed. Default
  implementation would be just fine too.

Please note, despite the fact that a lot of files are changed
only conjustion tool is affected by this change. Objects of
SkyObject class family are not copied much.

CCMAIL: kstars-devel <at> kde.org

 M  +6 -1      deepskyobject.cpp  
 M  +3 -1      deepskyobject.h  
 M  +3 -7      ksasteroid.cpp  
 M  +2 -6      ksasteroid.h  
 M  +4 -12     kscomet.cpp  
 M  +2 -6      kscomet.h  

Re: KDE/kdeedu/kstars/kstars/skyobjects

On Wed, Jun 03, 2009 at 02:12:01PM +0000, Alexey Khudyakov wrote:
> SVN commit 977139 by khudyakov:
> 
> In this changeset clone function is implemented. It allows safe
> copying of objects. E.g. KSPlanet stored as SkyObject* pointer
> would be copied as KSPlanet and not sliced. 
> 
> For all subclasses of SkyObject but DeepSkyObject copy
> constructors are removed. In most cases they are pretty useless
> and could be replaced with default implementations. 
> 
> * Changes in copy constructors
>  - Trail in TrailObject wasn't copied. Now it's copied
>  - const annotation added in DeepSkyObject copy constructor
> 
> * Overloaded assignment operator in EclipticPosition
>   (skyobjects/ksplanetbase.h) removed. Default
>   implementation would be just fine too.
> 
> Please note, despite the fact that a lot of files are changed
> only conjustion tool is affected by this change. Objects of
> SkyObject class family are not copied much.

No regressions for sure?

Particularly "dangerous" areas are Conjunctions, Planets section of
AstroCalculator and the Deep Star Component, as far as I can remember.

We're about to tag KDE 4.3 Beta tomorrow, so I'd appreciate if someone
could test out for problems.

Re: KDE/kdeedu/kstars/kstars/skyobjects

On Wednesday 03 of June 2009 18:32:41 Akarsh Simha wrote:
> No regressions for sure?
>
> Particularly "dangerous" areas are Conjunctions, Planets section of
> AstroCalculator and the Deep Star Component, as far as I can remember.
>
This change only affect copy constructors. And objects are copied only in one 
place, in Conjuction tool. It didn't show any problem for now but I'll test 
more throughfully after I apply all patches in this 'Fix copy constructors 
series' 

I'm trying to test everything that I put in SVN. 

> We're about to tag KDE 4.3 Beta tomorrow, so I'd appreciate if someone
> could test out for problems.
>
Yes. I thought that if these changes do intoduce regessions it's better to 
have them in beta than intoduce them later. 

Actually I have few fixes for possible problems, so I decided to test them put 
in SVN before beta tagging. 

KDE/kdeedu/kstars/kstars/skyobjects

SVN commit 977218 by khudyakov:

Eliminate possible double free problem in SkyObject.

AuxInfo wasn't copied in the copy ctor pointer to it was copied 
instead. So if object was copied AuxInfo wuold be be deleted twice and
program will most likely crash.

This is solved via Qt mechanism for implicit data sharing. 
QSharedDataPointer & QSharedData.

CCMAIL: kstars-devel <at> kde.org

 M  +9 -4      auxinfo.h  
 M  +8 -9      skyobject.cpp  
 M  +3 -2      skyobject.h  

--- trunk/KDE/kdeedu/kstars/kstars/skyobjects/auxinfo.h #977217:977218
@@ -20,6 +20,7 @@

 #include <QString>
 #include <QStringList>
+#include <QSharedData>

 /**
  *@struct AuxInfo
@@ -30,11 +31,15 @@
  *@version 1.0
  */

KDE/kdeedu/kstars/kstars

SVN commit 977251 by khudyakov:

* Oops. Restore copy constructor for StarObject. It set updateID
  and updateNumID to 0. Compiler derived version does not. Fixed.

* In ConjunctionsTool::slotFindObject() use clone() instead of
  kind of RTTI. It shorter and safer this way.

* Fix copy constructor of DeepSkyObject so it copy QImage right
  and does not cause segfaults

CCMAIL: kstars-devel <at> kde.org

 M  +2 -2      skyobjects/deepskyobject.cpp  
 M  +2 -4      skyobjects/deepskyobject.h  
 M  +14 -0     skyobjects/starobject.cpp  
 M  +5 -1      skyobjects/starobject.h  
 M  +1 -20     tools/conjunctions.cpp  

--- trunk/KDE/kdeedu/kstars/kstars/skyobjects/deepskyobject.cpp #977250:977251
@@ -38,9 +38,9 @@
     UGC( o.UGC ),
     PGC( o.PGC ),
     MajorAxis( o.MajorAxis ),
-    MinorAxis( o.MinorAxis ),
-    Image(new QImage(*o.Image))
+    MinorAxis( o.MinorAxis )
 {
+    Image = o.Image ? new QImage(*o.Image) : 0;
     updateID = updateNumID = 0;

Re: KDE/kdeedu/kstars/kstars/skyobjects

> Particularly "dangerous" areas are Conjunctions, Planets section of
> AstroCalculator and the Deep Star Component, as far as I can remember.
>

And if  Deep Star Component and friend will bomb at some moment I wouldn't 
mind much. They are broken anyway. Mixing malloc/free and C++ object won't do 
any good unless these objects could be expressed as C structs. They don't have 
nontrivial destructors and so on... SkyObject do have nontrivial destructor. 

Stars in StarBlock do not release AuxInfo when deleted because 'free' is used 
not 'delete'.

KDE/kdeedu/kstars/kstars/tools

SVN commit 977267 by khudyakov:

Fix bug in batch mode of Planets module of Calculator. Objects were reference 
when they went out of scope. 

I suppose it wasn't noticed because batch mode feature is pretty unusable. 

CCMAIL: kstars-devel <at> kde.org

 M  +11 -18    modcalcplanets.cpp  

--- trunk/KDE/kdeedu/kstars/kstars/tools/modcalcplanets.cpp #977266:977267
@@ -466,28 +466,19 @@
         KSPlanet Earth( kd, I18N_NOOP( "Earth" ));
         Earth.findPosition( &num );

-        KSPlanetBase *kspb;
+        // FIXME: allocate new object for every iteration is probably not wisest idea.
+        KSPlanetBase *kspb = 0 ;
         if ( pn == "Pluto" ) {
-            KSPluto ksp( kd );
-            ksp.findPosition( &num, &latB, &LST, &Earth );
-            ksp.EquatorialToHorizontal( &LST, &latB );
-            kspb = (KSPlanetBase*)&ksp;
+            kspb = new KSPluto(kd);
         } else if ( pn == "Sun" ) {
-            KSSun ksp( kd );
-            ksp.findPosition( &num, &latB, &LST, &Earth );
-            ksp.EquatorialToHorizontal( &LST, &latB );
-            kspb = (KSPlanetBase*)&ksp;