[Qgis-developer] Re: [QGIS Commit] [g_j_m] r5068 - trunk/qgis/src/gui

[Brendan Morley]

Gavin,

An idea I had for getting around the nested paint event is this:

Somehow we "triple-buffer" the canvas, where Qt provides the first 2
buffers and canvas operations write to the third.

Then a paint event is "simply" a matter of copying the third buffer to
the first pair. - a quick event.  Before and after, the canvas can
continue to be composed.

Thoughts?


Brendan

On Mon, 2006-03-20 at 02:42 +0000, noreply at qgis.org wrote:
> Author: g_j_m
> Date: 2006-03-20 02:42:21 +0000 (Mon, 20 Mar 2006)
> New Revision: 5068
> 
> Modified:
>    trunk/qgis/src/gui/qgsmaprender.cpp
> Log:
> Add a little debug output to make clear the source of the QPaintEngine
> error
> 
> 
> Modified: trunk/qgis/src/gui/qgsmaprender.cpp
> ===================================================================
> --- trunk/qgis/src/gui/qgsmaprender.cpp	2006-03-19 17:58:13 UTC (rev 5067)
> +++ trunk/qgis/src/gui/qgsmaprender.cpp	2006-03-20 02:42:21 UTC (rev 5068)
> @@ -205,6 +205,15 @@
>      std::cout << "QgsMapRender::render: at layer item '" << (*li).toLocal8Bit().data() << "'." << std::endl;
>  #endif
>  
> +    // This call is supposed to cause the progress bar to
> +    // advance. However, it seems that updating the progress bar is
> +    // incompatible with having a QPainter active (the one that is
> +    // passed into this function), as Qt produces a number of errors
> +    // when try to do so. I'm (Gavin) not sure how to fix this, but
> +    // added these comments and debug statement to help others...
> +    std::cerr << "If there is a QPaintEngine here, it is caused by an emit "
> +      "call just after line " << __LINE__ << " in file " << __FILE__ << ".\n";
> +
>      emit setProgress(myRenderCounter++,layers.size());
>      QgsMapLayer *ml = QgsMapLayerRegistry::instance()->mapLayer(*li);
>  
-- 
Brendan Morley <morb at ozemail.com.au>