[PATCH] gdiplus: Handle NULL elementdata.combine.left/right in delete_element

[Konstantin Romanov]

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=52423

Signed-off-by: Konstantin Romanov 
---
 dlls/gdiplus/gdiplus.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

diff --git a/dlls/gdiplus/gdiplus.c b/dlls/gdiplus/gdiplus.c
index 7c4c68f162f..7b0592c184c 100644
--- a/dlls/gdiplus/gdiplus.c
+++ b/dlls/gdiplus/gdiplus.c
@@ -473,10 +473,14 @@ void delete_element(region_element* element)
         case RegionDataInfiniteRect:
             break;
         default:
-            delete_element(element->elementdata.combine.left);
-            delete_element(element->elementdata.combine.right);
-            heap_free(element->elementdata.combine.left);
-            heap_free(element->elementdata.combine.right);
+            if(element->elementdata.combine.left){
+                delete_element(element->elementdata.combine.left);
+                heap_free(element->elementdata.combine.left);
+            }
+            if(element->elementdata.combine.right){
+                delete_element(element->elementdata.combine.right);
+                heap_free(element->elementdata.combine.right);
+            }
             break;
     }
 }

[Nikolay Sivov]

Please add a test case for this, that replicates crashing call sequence.

[Esme Povirk (she/they)]

It looks to me like the error path of clone_element could cause this,
but that's not something we can really test.

On Thu, Jan 20, 2022 at 5:05 AM Nikolay Sivov  wrote:
> Please add a test case for this, that replicates crashing call sequence.

[Esme Povirk (she/they)]

> It looks to me like the error path of clone_element could cause this,
> but that's not something we can really test.

That wouldn't cause a crash in GdipDeleteRegion like the log on the
bug shows, so I guess it's not that.