[Scummvm-cvs-logs] SF.net SVN: scummvm:[54975] scummvm/trunk/engines/sci/graphics/paint16.cpp

[thebluegr at users.sourceforge.net]

Revision: 54975
          http://scummvm.svn.sourceforge.net/scummvm/?rev=54975&view=rev
Author:   thebluegr
Date:     2010-12-20 12:21:18 +0000 (Mon, 20 Dec 2010)

Log Message:
-----------
SCI: Fixed bug #3037945 - "PQ3 : Graphical Glitch with White Mercedes on Freeway"

Modified Paths:
--------------
    scummvm/trunk/engines/sci/graphics/paint16.cpp

Modified: scummvm/trunk/engines/sci/graphics/paint16.cpp
===================================================================
--- scummvm/trunk/engines/sci/graphics/paint16.cpp	2010-12-20 12:19:28 UTC (rev 54974)
+++ scummvm/trunk/engines/sci/graphics/paint16.cpp	2010-12-20 12:21:18 UTC (rev 54975)
@@ -534,7 +534,20 @@
 		case SCI_DISPLAY_RESTOREUNDER:
 			bitsGetRect(argv[0], &rect);
 			rect.translate(-_ports->getPort()->left, -_ports->getPort()->top);
-			bitsRestore(argv[0]);
+			if (g_sci->getGameId() == GID_PQ3 && g_sci->getEngineState()->currentRoomNumber() == 29) {
+				// WORKAROUND: PQ3 calls this without calling the associated
+				// kDisplay(SCI_DISPLAY_SAVEUNDER) call before. Theoretically,
+				// this would result in no rect getting restored. However, we
+				// still maintain a pointer from the previous room, resulting
+				// in invalidated content being restored on screen, and causing
+				// graphics glitches. Thus, we simply don't restore a rect in
+				// that room. The correct fix for this would be to erase hunk
+				// pointers when changing rooms, but this will suffice for now,
+				// as restoring from a totally invalid pointer is very rare.
+				// Fixes bug #3037945.
+			} else {
+				bitsRestore(argv[0]);
+			}
 			kernelGraphRedrawBox(rect);
 			// finishing loop
 			argc = 0;


This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.