Re: [Boost-users] Visual C++ 8 - Example Crashes with Access Violation

From: Ovanes Markarian (om_boost_at_[hidden])
Date: 2006-04-28 02:24:26

What I can imagine in regards of this failure is that you have in your watch
expressions some pointer dereference. And this ptr is not valid... I had
this error once in VC 6.0 and spent almost 3 hours looking for this kind of
problem... Please analyse your watch expressions carefully. May be you have
argv[1] in your watch or some other context valid name, which is not
initialized.

As I can remember this was fixed in VC 2002, but can come up in VC 8 again.

Best Regards,
Ovanes

-----Ursprüngliche Nachricht-----
Von: Rush Manbert [mailto:rush_at_[hidden]] 
Gesendet: Thursday, April 27, 2006 22:44
An: boost-users_at_[hidden]
Betreff: Re: [Boost-users] Visual C++ 8 - Example Crashes with Access
Violation

Brown Gabe wrote:
> I tried commenting out the std::cout call and it still crashes for 
> some reason.  I'm trying to track down the problem to the 
> uninitialized memory but I can't seem to locate it for some reason.  I 
> tried implementing a slight simplier version to try and track down the
fault.
> 
> Can anyone reproduce this error on a vanilla install of Visual Studio
2005?
> I'm wondering if this is a VC8 flaw for Boost or just some weird  
> libraries that I'm using with boost.
> 
> 
> Here is the sample code:
> 
> ==============[Sample Code]=====================
> 
> void ThreadTest() {
> 
>     return;
> 
> };
> 
> 
> int main(int argc, char* argv[])
> {
> 
>     boost::thread thrd(ThreadTest);
>     thrd.join();
> 
>    return 0;
> }
> 
> ==============[Sample Code]=====================
> 
I tried this code, then a variant of this code, then the code you supplied
in your first post, and they all ran without error in both Debug and Release
build configurations. I made a Win32 console app using static Boost
libraries and the static runtime. My Visual Studio 2005 installation is very
new and totally vanilla. I guess this helps point you at some other library.

I still wonder, can't you set a breakpoint to trigger when the exception is
thrown? If you break at that point, the offender should be on the stack.

- Rush