Xorg 6.9 & BUG #5585

Robert Connolly robert at linuxfromscratch.org
Fri Jan 13 04:08:30 PST 2006


To check for a pic compiler:
#if defined(__PIC__) || defined(__pic__)

To check for ssp:
#if defined( __SSP__)

To check for -z now they could grep from readelf. I'm not sure of a better 
way.

robert

On January 13, 2006 03:22 am, Declan Moriarty wrote:
> Recently, Somebody Somewhere wrote these words
>
> > > Right compiled again, and still get the same error. The
> > > -nonow patch went in (1 hunk), the other 2 already there. Kevin
> > > Day's Mesa-Stop_Asking went in. All appropriate seds, and bodges
> > > and it built through, but falls over on Xorg -configure
> > >
> > > The url above says
> > >
> > > dlloader will not work when LD_BIND_NOW is set, or when the
> > > modules are linked with '-z now' as in gentoo-hardened.  this is
> > > unavoidable until we abandon LoadSubModule for link-time
> > > dependencies.  you'll get an error message about LD_BIND_NOW
> > > though.
> >
> > Right, I haven't tested with these hardened flags yet.  Xorg's Modular
> > development with very little documentation had too many potential
> > problems, so I am not using (or forcing) any hardened flags of any kind.
> >
> > > Workaround: fix your scripts to clear LD_BIND_NOW before starting X
> > > Workaround (maybe): export LD_BIND_NOT=1 before starting X.
> > >
> > > Sure enough, it complains that LD_BIND_NOW is set. Using
> > >
> > > export LD_BIND_NOW="" kills that error, but not the problem
> > > export LD_BIND_NOT=1 does nothing.
> >
> > I'll make sure to look into that.
> >
> > > Is there any workaround for this before someone writes a -nonow
> > > patch for Xorg-6.9.0 or Xorg-7.0.0?
> >
> > The bug(s) fixed in the old monolithic tree  by the hlfs team probably
> > still exist in the new xorg.  Now, it just needs to be converted,
> > perhaps?
>
> Seeing as they were admitting it's broken, I filed a bug #5585 in
> bugs.freedesktop.org.
>
> They want to know how to detect a hardened compiler. I sent them
> the hardened specs script and the gcc stack-smashing test program.
> Any better ways to get this done with a configure test?
>
> --
>
> 	With best Regards,
>
>
> 	Declan Moriarty.



More information about the hlfs-dev mailing list