Hi,

Just out of interest, did firefox build?  My SS4 is grumpy about
one file so far, grumpy enough that the compiler segfaults.

thanks.

cheers

bruce

"Jon Buller" <jon@bullers.net> wrote:
> Date: Mon, 20 Dec 2004 07:28:19 -0800
> From: "Jon Buller" <jon@bullers.net>
> Subject: Re: GCC sparc optomizer broken?
> To: David Brownlee <abs@NetBSD.org>
> Cc: Jon Buller <jon@bullers.net>, port-sparc@NetBSD.org, jon@bullers.net
> content-length: 1428
> 
> On 00 Jan 00 00:00:00, 
> David Brownlee wrote:
> 
> > On Sun, 19 Dec 2004, Jon Buller wrote:
> > 
> > > I was building the security/gnupg package on my SS20 and I have
> > > this line in /etc/mk.conf:
> > >
> > >     .include "/usr/pkg/share/mk/cpuflags.mk"
> > >
> > > It produced the output at the end of this message, but it all works
> > > fine if I comment that line out.  I believe the only thing that
> > > line does is add -mcpu=supersparc to CFLAGS.  (The devel/cpuflags-0.78
> > > package was installed, in case anyone wants to try exactly reproducing
> > > this.)
> > 
> >  	What version of gcc are you using? 2.95 (In NetBSD 1.6.x) did
> >  	have some optimiser bugs... I'm sure 3.x has some also, but
> >  	the 2.95 ones seem worse :)
> 
> This was in my newly built 2.0 release chroot sandbox, so it was
> "gcc (GCC) 3.3.3 (NetBSD nb1 20040301)".  However, I think I've
> seen this, or very similar bugs with egcs, and 2.95 also.  As I
> remember the older versions would just spin the CPU while making
> keys, including SSL at one point I think.  At least now the gnupg
> build fails when this happens, it used to just give you a broken
> binary for you to discover later.
> 
> So I was expecting it a bit, but hoping someone had found and fixed
> it in the new version.  And this is the only thing I know of that
> breaks at the moment, I have built everything with -mcpu=supersparc,
> but I'm a little nervious about what else might be broken without
> me noticing it.
> 
> Jon