LEAK: WheelEventImpl leaked when mousewheel used

• Component: New Bugs
• Status: RESOLVED FIXED

[Kurt Kohler]

A WheelEventImpl is leaked in NodeImpl::dispatchWheelEvent when a mouse wheel
event is being handled. The problem appears to be missing ref/deref's around the
call to dispatchEvent. I have a patch.

[Kurt Kohler]

Created attachment 3590 [details]
Patch to add ref/deref

This patch makes the code consistent with other event dispatchers, it *may* be
correct.

[Kurt Kohler]

I forgot to include the call stack that started the whole thing.

Leak: 0x0bde14d0  size=112
        0x019a4fa8 0x00000000 0x00000000 0x00000001     ..O.............
        0x00000001 0x00000000 0x00000000 0x00000000     ................
        0x00000000 0x0000003a 0x00000000 0x000000ae     .......:........
        0x00000000 0x41a17f6c 0xddc367a1 0x0a443030     ....A..l..g..D00
        0x00000000 0x000003ed 0x00000259 0x00000176     ...........Y...v
        0x00000230 0x00000371 0x00000230 0x00000371     ...0...q...0...q
        0x00000001 0x000000c0 0x000014cc 0x00000007     ................
        Call stack: [thread 26df]: |
        0x0 |
        0x57190 |
        0x26e0 |
        NSApplicationMain |
        -[NSApplication run] |
        0x1a684 |
        -[NSApplication sendEvent:] |
        0x1d69c |
        -[NSWindow sendEvent:] |
        -[WebHTMLView scrollWheel:] |
        -[WebCoreBridge sendScrollWheelEvent:] |
        KWQKHTMLPart::wheelEvent(NSEvent*) |
        KHTMLView::viewportWheelEvent(QWheelEvent*) |
        DOM::NodeImpl::dispatchWheelEvent(QWheelEvent*) |
        DOM::EventImpl::operator new(unsigned long) |
        khtml::main_thread_malloc(unsigned long) |
        malloc 

[Oliver Hunt]

Do you have a testcase/site that triggers the leak?

[Michael J. Cohen]

simpy scrolling up and down seems to reproduce this for me.

[Joost de Valk (AlthA)]

Reassigning to webkit-unassigned, to make sure more people see this.

[Joost de Valk (AlthA)]

Comment on attachment 3590 [details]
Patch to add ref/deref

Though this is an old patch this might yet be valid.

[Joost de Valk (AlthA)]

Confirming.

[Geoffrey Garen]

Comment on attachment 3590 [details]
Patch to add ref/deref

Should use a refptr. I'll submit a patch shortly.

[Geoffrey Garen]

Created attachment 6514 [details]
refptr patch

[Eric Seidel (no email)]

Comment on attachment 6514 [details]
refptr patch

r=me.

[Geoffrey Garen]

Landed.