Bug 168292 – REGRESSION(r212239): Crash in DragImage::operator=(WebCore::DragImage&&) when DragImageRef is the same

• Component: Platform
• Status: NEW

[Carlos Garcia Campos]

This happens at least in the GTK+ port where DragImageRef is a pointer (we should definitely change that). It caused several crashes in the bots:

Thread 1 (Thread 0x2b4ba8e96940 (LWP 11637)):
#0  0x00002b4ba28b3067 in __GI_raise (sig=sig@entry=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:56
#1  0x00002b4ba28b4448 in __GI_abort () at abort.c:89
#2  0x00002b4ba28ac266 in __assert_fail_base (fmt=0x2b4ba29e5238 "%s%s%s:%u: %s%sAssertion `%s' failed.\n%n", assertion=assertion@entry=0x2b4b9ba8ce08 "((*&(&surface->ref_count)->ref_count) > 0)", file=file@entry=0x2b4b9ba8cd38 "/home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/cairo-1.14.2/src/cairo-surface.c", line=line@entry=953, function=function@entry=0x2b4b9ba9d9f0 <__PRETTY_FUNCTION__.11168> "cairo_surface_destroy") at assert.c:92
#3  0x00002b4ba28ac312 in __GI___assert_fail (assertion=0x2b4b9ba8ce08 "((*&(&surface->ref_count)->ref_count) > 0)", file=0x2b4b9ba8cd38 "/home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/cairo-1.14.2/src/cairo-surface.c", line=953, function=0x2b4b9ba9d9f0 <__PRETTY_FUNCTION__.11168> "cairo_surface_destroy") at assert.c:101
#4  0x00002b4b9ba1a7c2 in cairo_surface_destroy () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/cairo-1.14.2/src/cairo-surface.c:953
#5  0x00002b4b9968b7f9 in WebCore::DragImage::operator=(WebCore::DragImage&&) () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libwebkit2gtk-4.0.so.37
#6  0x00002b4b995d7e8b in WebCore::DragController::doImageDrag(WebCore::Element&, WebCore::IntPoint const&, WebCore::IntRect const&, WebCore::DataTransfer&, WebCore::Frame&, WebCore::IntPoint&) () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libwebkit2gtk-4.0.so.37
#7  0x00002b4b995dbcc2 in WebCore::DragController::startDrag(WebCore::Frame&, WebCore::DragState const&, WebCore::DragOperation, WebCore::PlatformMouseEvent const&, WebCore::IntPoint const&) () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libwebkit2gtk-4.0.so.37
#8  0x00002b4b995e3994 in WebCore::EventHandler::handleDrag(WebCore::MouseEventWithHitTestResults const&, WebCore::CheckDragHysteresis) () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libwebkit2gtk-4.0.so.37
#9  0x00002b4b995e3eb8 in WebCore::EventHandler::handleMouseDraggedEvent(WebCore::MouseEventWithHitTestResults const&, WebCore::CheckDragHysteresis) () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libwebkit2gtk-4.0.so.37
#10 0x00002b4b995eab67 in WebCore::EventHandler::handleMouseMoveEvent(WebCore::PlatformMouseEvent const&, WebCore::HitTestResult*, bool) () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libwebkit2gtk-4.0.so.37

When m_dragImageRef is the same as other.m_dragImageRef we end up deleting twice.

[Carlos Garcia Campos]

Created attachment 301472 [details]
Patch

[Anders Carlsson]

Comment on attachment 301472 [details]
Patch

This isn't right. Just self-move a no-op.

[Carlos Garcia Campos]

(In reply to comment #2)
> Comment on attachment 301472 [details]
> Patch
> 
> This isn't right. Just self-move a no-op.

I'm not sure this is actually a self-move, this and other are not the same but both wrap the same pointer. This is no longer a problem for us, though, because I changed our DragImageRef to be a RefPtr too, and I don't think there are any other ports using raw pointers now.