[Bug 59609] New: comctl32: LVM_GETITEMPOSITION returns client coordinates instead of view coordinates in ICON/SMALLICON views (breaks FastStone Image Viewer)

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=59609

            Bug ID: 59609
           Summary: comctl32: LVM_GETITEMPOSITION returns client
                    coordinates instead of view coordinates in
                    ICON/SMALLICON views (breaks FastStone Image Viewer)
           Product: Wine
           Version: 11.5
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: major
          Priority: P2
         Component: comctl32
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

Bug overview:
In `dlls/comctl32/listview.c`, the function `LISTVIEW_GetItemPosition`
unconditionally adds `Origin.x` and `Origin.y` to the item's coordinates. This
converts the coordinates to Client coordinates.
However, according to MSDN, for `LV_VIEW_ICON` and `LV_VIEW_SMALLICON` modes,
`LVM_GETITEMPOSITION` must return View coordinates (absolute coordinates on the
canvas), not Client coordinates.

MSDN Documentation reference for LVM_GETITEMPOSITION:
"In icon or small icon view, the LVM_GETITEMPOSITION message returns the item's
position in view coordinates. To get the client coordinates of an item,
subtract the origin from the view coordinates."

Symptoms & Application Behavior:
This bug breaks the top thumbnail strip in full screen mode of FastStone Image
Viewer (FSViewer.exe) when scrolling horizontally. The thumbnail strip is a
ListView in Icon mode (height: 150px). 

When the user scrolls to the right by approximately the width of the screen
(e.g., ~1900px on a 1080p monitor), the thumbnails suddenly disappear, leaving
blank white spaces and cropped thumbnails.

The Mathematical Root Cause:
Because Wine incorrectly returns Client coordinates instead of View
coordinates, a "double subtraction" occurs in the application's logic:
1. Wine calculates internally: `Returned_X = Absolute_X - ScrollPos` (by adding
the negative Origin).
2. FSViewer (following MSDN) attempts to calculate the client position to
determine if the item is visible: `App_Client_X = Returned_X - ScrollPos`.
3. The resulting math is: `App_Client_X = Absolute_X - 2 * ScrollPos`.

When `ScrollPos` exceeds the width of the window (e.g., scrolling past 1920px),
the calculated right edge of the visible items becomes negative. FSViewer
incorrectly concludes that all items have moved off-screen to the left. 
To save resources, FSViewer returns `CDRF_SKIPDEFAULT` (0x4) during the
`CDDS_PREPAINT` custom draw notification. Wine respects this flag, skips
drawing the items, and `BitBlt` copies a blank/white double-buffer background
to the screen.

Proposed Fix / Patch:
The addition of `Origin` should be restricted to `LV_VIEW_DETAILS` and
`LV_VIEW_LIST` views, where Client coordinates are actually expected.

In `dlls/comctl32/listview.c`, modify `LISTVIEW_GetItemPosition`:

--- a/dlls/comctl32/listview.c
+++ b/dlls/comctl32/listview.c
@@ -XXXX,X +XXXX,X @@ static BOOL LISTVIEW_GetItemPosition(const LISTVIEW_INFO
*infoPtr, INT nItem, LP
     if (infoPtr->uView == LV_VIEW_ICON)
     {
         lpptPosition->x += (infoPtr->nItemWidth - infoPtr->iconSize.cx) / 2;
         lpptPosition->y += ICON_TOP_PADDING;
     }
-    lpptPosition->x += Origin.x;
-    lpptPosition->y += Origin.y;
+    
+    /* In Icon/SmallIcon views, position is in view coordinates (absolute).
+     * In Details/List views, position is in client coordinates. */
+    if (infoPtr->uView == LV_VIEW_DETAILS || infoPtr->uView == LV_VIEW_LIST)
+    {
+        lpptPosition->x += Origin.x;
+        lpptPosition->y += Origin.y;
+    }

     TRACE ("  lpptPosition=%s\n", wine_dbgstr_point(lpptPosition));
     return TRUE;

Testing:
Applying this exact logic fully resolves the white-space rendering bug in
FastStone Image Viewer without breaking standard ListView internal mechanics
(since Wine internally relies on `LISTVIEW_GetItemOrigin` for drawing, which
remains unaffected).

-- 
Do not reply to this email, post in Bugzilla using the
above URL to reply.
You are receiving this mail because:
You are watching all bug changes.