[uClibc] [RFC] RUNPATH config option for ld.so

Joakim Tjernlund Joakim.Tjernlund at lumentis.se
Sun Jun 26 19:02:27 UTC 2005


> On Sunday 26 June 2005 07:08 am, Joakim Tjernlund wrote:
> > > On Sunday 26 June 2005 05:26 am, Joakim Tjernlund wrote:
> > > > > On Saturday 25 June 2005 03:28 am, Erik Andersen wrote:
> > > > > > On Sat Jun 25, 2005 at 02:22:08AM -0400, Mike Frysinger wrote:
> > > > > > > +config LDSO_RUNPATH_FULL
> > > > > > > +	bool "FULL"
> > > > > > > +	help
> > > > > > > +	  Respect all RUNPATH's regardless (same behavior as glibc).
> > > > > > > +config LDSO_RUNPATH_SAFE
> > > > > > > +	bool "SAFE"
> > > > > > > +	help
> > > > > > > +	  Only respect RUNPATH's which contain absolute paths.
> > > > > > > +config LDSO_RUNPATH_NONE
> > > > > > > +	bool "NONE"
> > > > > > > +	help
> > > > > > > +	  Disable RUNPATH support all together.
> > > > > > > +endchoice
> > > > > >
> > > > > > BTW, I think this is too many choices for no real benefit....
> > > > > > IMHO it should be all or nothing.
> > > > >
> > > > > i offered one more option ('safe') than you guys wanted ... the back
> > > > > end is trivial to support this, it'd require one ifdef around a path
> > > > > check ...
> > > >
> > > > show us please.
> > >
> > > attached
> >
> > This also does the same to LD_LIBRARY_PATH.
> 
> inttended side effect ... wrapper scripts often do things like 'export 
> LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:blahblah' which, much more often than not, 
> evaluate to ":blahblah" ...
> -mike

hmm, you also want to disable the possibility to set LD_LIBRARY_PATH to "." as well?

While we are at it, why not remove RPATH compleatly and just have RUNPATH?

 Jocke



More information about the uClibc mailing list