Do you have a real (non-VMWare) OpenBSD machine you can test on?  
Another thing to try is to upgrade to GCC 3.3.

Is anyone else using Crypto++ on OpenBSD/x86 without disabling the 
optimized assembly code?

On Thu, Sep 11, 2003 at 02:20:08PM -0400, Guy Smith wrote:
> For now we've "solved" this problem by forcing crypto++ to use the
> "Portable" base class instead of the derived "PentiumOptimized" class in
> integer.cpp, but obviously we're not thrilled by this workaround.
> 
> --Guy
> 
> ----- Original Message ----- 
> From: "Guy Smith" <[EMAIL PROTECTED]>
> To: <[EMAIL PROTECTED]>
> Sent: Thursday, September 11, 2003 10:43 AM
> Subject: BlumBlumShub fails on OpenBSD
> 
> 
> > Hi there.  We've been using Crypto++ for about eight months now on Windows
> > and have recently begun porting over to OpenBSD.  One problem we had
> involed
> > osrng.cpp, which wanted to use /dev/random.  On OpenBSD, it should be
> > /dev/srandom.  The second problem is more difficult for us.  When running
> > the validation suite (cryptest v) we get the following assertion during
> > BlumBlubShub.  We also get a similar assertion at line 2012 when running
> our
> > own project.  The situation is complicated by the fact that we're running
> > OpenBSD in VMWare on Windows XP.  Any help on this problem would be vastly
> > appreciated.  Thanks.
> >
> > BlumBlumShub validation suite running...
> > assertion "carry >= 0 && carry <= 2" failed: file "integer.cpp", line 1864
> >
> > Versions:
> >     Crypto++ 5.1
> >     OpenBSD 3.3
> >     gcc 2.95.3
> >     VMWare 4.0.2 build 5592
> >     WinXP 2002 SP1
> >
> >
> > Guy L. Smith
> > [EMAIL PROTECTED]
> > [EMAIL PROTECTED]
> >
> > I feel like I'm debugging a zen koan.
> >

Reply via email to