Re: [Gnustep-cvs] gnustep/core/gui ChangeLog Source/NSTableView.m

[Fred Kiefer]

Enrico Sersale wrote:
> On 2005-05-28 06:34:22 +0300 Matt Rice <address@hidden> wrote:
>
> > --- Enrico Sersale <address@hidden> wrote:
> >
> > > On 2005-05-26 16:38:12 +0300 Fred Kiefer
> > > <address@hidden> wrote:
> > >
> > > > CVSROOT:    /cvsroot/gnustep
> > > > Module name:    gnustep
> > > > Branch:    
> > > > Changes by:    Fred Kiefer
> > >
> > > <address@hidden>    05/05/26 13:38:11
> > >
> > > > > Modified files:
> > > >
> > > >     core/gui       : ChangeLog     core/gui/Source:
> > >
> > > NSTableView.m
> > >
> > > > Log message:
> > > >     Improved mouseDown call handling for table view.
> > > >
> > > > > CVSWeb URLs:
> > > I think that it would be a good habit to try to run
> > > one or two of the (few) GNUstep applications before
> > > committing changes that can break things...
> >
> > just to elaborate.. this breaks (at least)
> > gworkspace's
> > view->list segfaults when selecting a row.
>
>
> Just to be more precise :) ... I'm not referring to this; I've fixed it 
> implementing -copyWithZone: in my cell class when I've seen that 
> gworkspace segfaults trying to release an ivar of this class.
> The real problem is that it is not possible anymore to start a drag from 
> a cell (this is visible in GNUMail, too). Making 
> -startTrackingAt:inView: to return NO in the cell class fixes this 
> problem too, but I think that there is something wrong in -mouseDown: 
> because, before theese changes, both the apps worked well.

Enrico,
It is not always bad to break things. If in this case my change showed 
some memory problem in GWorkspace and GNUMail this is a good thing. 
Provided we are able to fix it before the next offical release of 
course. So there was at least one benefit from applying the patch. :-)

We should work together to track down the actual problems and fix them. 
GNUstep is stil full of hacks to get partial functionality working. We 
need to try to get it correct completely.

> > anyhow the culprit being cell copy, which was added
> > because NSComboBoxCell caches the cellFrame so it can
> > pop up the window through a NSButtonCell's action...
> >
> > though it was causing multiple drawWithFrame: calls
> > for the combobox cells with its setNeedsDisplay: then
> > popping up in the wrong place...
> >
> > anyhow this seems to work in both cases..

Matt,
You are surely correct about the fix for NSComboBoxCell, it should only 
mark its onl´e frame for a redraw. I am not totally convinced about the 
copy of the active cell in NSTableView. Copying the cell also gets done 
when editing is started for a cell. So cells that don't handle it 
correctly will also pose problems when edited.

I will test this myself and propably apply it as well, to get GWorkspace 
and GNUMail working again. But my feeling is still that there is another 
problem hidden here, which we need to resolve.