[webkit-help] segfault when destroying JS function

Dani Pinyol gmane at dpinol.com
Mon Nov 1 09:38:51 PDT 2010

Luka Napotnik <luka.napotnik at ...> writes:

> Hello.I have a C program that binds a JS function to WebKit (gtk+ r47882). Now 
when I run the program for some time it segfaults. I checked it under valgrind 
and get the following backtrace:============================
> ==6195== Invalid write of size 8==6195==    at 0x815D694: 
JSC::JIT::unlinkCall(JSC::CallLinkInfo*) (in /usr/lib/libwebkit-
1.0.so.2.9.0)==6195==    by 0x819A977: JSC::CodeBlock::unlinkCallers() (in 
> ==6195==    by 0x820723B: JSC::JSFunction::~JSFunction() (in 
/usr/lib/libwebkit-1.0.so.2.9.0)==6195==    by 0x8285386: unsigned long 
JSC::Heap::sweep<(JSC::HeapType)0>() (in /usr/lib/libwebkit-
1.0.so.2.9.0)==6195==    by 0x823A1B7: JSC::Heap::collect() (in 
> ==6195==    by 0x85CF421: WebCore::ThreadTimers::fireTimers(double, 
WTF::Vector<WebCore::TimerBase*, 0ul> const&) (in /usr/lib/libwebkit-
1.0.so.2.9.0)==6195==    by 0x85CF4CA: 
WebCore::ThreadTimers::sharedTimerFiredInternal() (in /usr/lib/libwebkit-
> ==6195==    by 0x8863611: WebCore::timeout_cb(void*) (in /usr/lib/libwebkit-
1.0.so.2.9.0)==6195==    by 0xB16C889: g_main_context_dispatch 
(gmain.c:1960)==6195==    by 0xB170217: g_main_context_iterate (gmain.c:2591)
> ==6195==    by 0xB17070C: g_main_loop_run (gmain.c:2799)==6195==    by 
0x9109BC6: gtk_main (gtkmain.c:1205)==6195==  Address 0x193b7b7b is not stack'd, 
malloc'd or (recently) free'd[Mon Nov  2 07:57:18 2009] 
========================================================It seems that the JS 
function was destroyed before the destructor for JSC::JSFunction was called. I 
should mention I re-bind the same function because sometimes the page changes 
and the binding is lost. But sometimes the page stays and I bind over the 
existing binding. Can this be the reason that leads to the segfault? Doesn't 
WebKit destroy the function if I bind it over?Greets,Luka

could you solve or workaround this problem? How?


More information about the webkit-help mailing list