Huge thanks to our Platinum Members Endace and LiveAction,
and our Silver Member Veeam, for supporting the Wireshark Foundation and project.

Wireshark-dev: Re: [Wireshark-dev] Systematic crash at startup when launching Wireshark GTK+ 1.

From: Gerald Combs <gerald@xxxxxxxxxxxxx>
Date: Mon, 15 Sep 2014 10:51:45 -0700
On 9/15/14 10:02 AM, Pascal Quantin wrote:
> 
> Le 15 sept. 2014 18:27, "Gerald Combs" <gerald@xxxxxxxxxxxxx
> <mailto:gerald@xxxxxxxxxxxxx>> a écrit :
>>
>> On 9/14/14 1:34 PM, Pascal Quantin wrote:
>> > Hi all,
>> >
>> > I just installed a new Windows 8.1 x64 computer and each time I launch
>> > the GTK2 GUI of master branch, I get a silent crash.
>> > After generating a memory dump and loading it in windbg, I could see
>> > that it crashes in gtk_init call:
>> > (1d90.1998): Access violation - code c0000005 (first/second chance not
>> > available)
>> > *** ERROR: Symbol file could not be found.  Defaulted to export symbols
>> > for KERNELBASE.dll -
>> > ntdll!ZwWaitForMultipleObjects+0xa:
>> > 00007ffc`67b61c2a c3              ret
>> > 0:000> .ecxr
>> > *** ERROR: Symbol file could not be found.  Defaulted to export symbols
>> > for libgobject-2.0-0.dll -
>> > rax=000000d4111b9000 rbx=00000000111c6590 rcx=000000d4111b9000
>> > rdx=0000000000000000 rsi=0000000000000020 rdi=0000000000000048
>> > rip=000000005949562b rsp=000000d40f56efe0 rbp=000000d4111c9a80
>> >  r8=0000000000000000  r9=0000000000000003 r10=0000000000000000
>> > r11=00007ffc674e1725 r12=0000000000000020 r13=0000000000000048
>> > r14=000000d40f8de318 r15=000000d4111c9c60
>> > iopl=0         nv up ei pl nz na pe nc
>> > cs=0033  ss=002b  ds=002b  es=002b  fs=0053  gs=002b
>> > efl=00010202
>> > libgobject_2_0_0!g_type_create_instance+0x15b:
>> > 00000000`5949562b 0fb67b14        movzx   edi,byte ptr [rbx+14h]
>> > ds:00000000`111c65a4=??
>> >
>> > The corresponding call stack is:
>> > libgobject_2_0_0!g_type_create_instance+0x15b
>> > libgobject_2_0_0!g_param_spec_internal+0x146
>> > libgobject_2_0_0!g_param_spec_object+0x77
>> > libgdk_win32_2_0_0!gdk_pointer_is_grabbed+0x123
>> > libgobject_2_0_0!g_type_class_ref+0x4a7
>> > libgobject_2_0_0!g_object_newv+0x26e
>> > libgobject_2_0_0!g_object_new+0x72
>> > libgdk_win32_2_0_0!gdk_display_manager_get+0x24
>> > libgtk_win32_2_0_0!gtk_misc_get_padding+0xdd3
>> > libgtk_win32_2_0_0!gtk_list_store_insert_with_valuesv+0x101c
>> > libglib_2_0_0!g_option_context_parse+0x3a7
>> > libgtk_win32_2_0_0!gtk_parse_args+0x7b
>> > libgtk_win32_2_0_0!gtk_init_check+0x9
>> > libgtk_win32_2_0_0!gtk_init+0x9
>> > Wireshark_gtk!main(int argc = 0n1, char ** argv =
> 0x000000d4`0f630540)+0x612
>> > Wireshark_gtk!WinMain(struct HINSTANCE__ * hInstance =
>> > 0x00007ff7`ac530000, struct HINSTANCE__ * hPrevInstance =
>> > 0x00000000`00000000, char * lpszCmdLine = 0x000000d4`0f612e8f "", int
>> > nCmdShow = 0n1)+0x7e
>> > Wireshark_gtk!__tmainCRTStartup(void)+0x149
>> > kernel32!BaseThreadInitThunk+0xd
>> > ntdll!RtlUserThreadStart+0x1d
>> >
>> > After doing a bisection, I noticed that it started appearing when we
>> > switched to GTK+ 2.24.23:
>> > commit 0a249087c311feacf693144372ec038938c47415
>> > Author: Gerald Combs <gerald@xxxxxxxxxxxxx
> <mailto:gerald@xxxxxxxxxxxxx> <mailto:gerald@xxxxxxxxxxxxx
> <mailto:gerald@xxxxxxxxxxxxx>>>
>> > Date:   Fri May 16 09:21:57 2014 -0700
>> >
>> >     Build with GTK+ 2.24.23.
>> >
>> >     Change-Id: Ic5c385c0fcef4d40a8cb9e7a271d14eb80905460
>> >     Reviewed-on: https://code.wireshark.org/review/1665
>> >     Reviewed-by: Gerald Combs <gerald@xxxxxxxxxxxxx
> <mailto:gerald@xxxxxxxxxxxxx>
>> > <mailto:gerald@xxxxxxxxxxxxx <mailto:gerald@xxxxxxxxxxxxx>>>
>> >     Tested-by: Gerald Combs <gerald@xxxxxxxxxxxxx
> <mailto:gerald@xxxxxxxxxxxxx>
>> > <mailto:gerald@xxxxxxxxxxxxx <mailto:gerald@xxxxxxxxxxxxx>>>
>> >
>> > But at the same time 1.12.0 release uses the same GTK2 package and works
>> > fine, so I'm puzzled. the amster x86 package works fine.
>> >
>> > I tried upgrading the GTK2 package thanks to the download-mingw-rpm.py
>> > but the latest packages from OpenSUSE seem to lack some libraries.
>> >
>> > I'm a bit out of ideas for now. Anyone familiar with this kind of
> issues?
>>
>> I can duplicate the problem here. If I create a package by hand using
>> Visual C++ 2010 Wireshark-gtk.exe runs fine.
> 
> It explains why it works fine with 1.12.0 then... And why I could not
> figure out a major difference in main.c file explaining the behavior
> while using the same libraries!

If I build a package from g0a24908 (the commit prior to the GTK+ package
upgrade) Wireshark-gtk.exe starts up OK.