Version: (using KDE 3.1.2) Installed from: (testing/unstable) Compiler: gcc version 3.3 (Debian) OS: Linux (i686) release 2.4.21-rc1 From version 3.1.1 to 3.1.2 the following patch was introduced: --- kghostview/kgv_view.cpp 2003-05-27 16:24:08.000000000 +0200 +++ kghostview.3.1.2/kgv_view.cpp 2003-05-27 16:32:54.000000000 +0200 @@ -525,6 +526,8 @@ { if( _isFileDirty ) { + slotOrientation( _selectOrientation->currentItem() ); + slotMedia( _selectMedia->currentItem() ); _docManager->goToPage( _currentPage ); _docManager->redisplay(); _isFileDirty = false; This fixes the bug that reloading a file would kill the selected orientation and media choice. Unfortunately, now the reloading (watching) of long documents does not work correctly anymore. I get different behaviors, for example upon automatic reload, I still see the old page, but when moving a page up, I'm at the last page of the document. Sometimes the document is only loaded half, sometimes gs seems to hang... My guess is that it's a timing problem, because slotOrientation, slotMedia, goToPage, and redisplay all call KGVMiniWidget::showPage( int pagenumber ) the first two possibly with a different page than the other two. Don't know what's causing the problems exactly, but I'm pretty sure that setting the orientation, media, and page variables prior to a single call to showPage would fix it.
I just tried that with a quick hack. Declaring the variables _overrideOrientation, _overridePageMedia, and _currentPage public in KGVMiniWidget and setting them directly instead of calling slotOrientation, slotMedia, and goToPage fixes all problems! GS doesn't seem to like four sequential calls to showPage()...? Anyway, the multiple calls of showPage() are redundant anyway and should go away. Not sure how to do this more elegant, though...
Ok, i see this and it is mostly my fault. I just applied a different fix to BRANCH: Changing the order in which things are done (since calling showPage in miniwidget results in resetting the _currentPage variable). In HEAD, this code has changed and it is all done with one function call so the problem is inexistent. Thanks for the report. luis