[rfc] dont use absolute paths in ldscripts if they arent needed
Daniel Jacobowitz
drow@false.org
Sun Apr 2 18:11:00 GMT 2006
On Sun, Apr 02, 2006 at 01:46:52PM -0400, Mike Frysinger wrote:
> On Sunday 02 April 2006 13:32, Daniel Jacobowitz wrote:
> > On Sun, Apr 02, 2006 at 01:25:49PM -0400, Mike Frysinger wrote:
> > > yes and no ... the way we're cross building/installing glibc now, the
> > > patch is no longer "needed" ... but the paths themselves are pointless
> > > when the files are together
> >
> > No, they aren't. They guarantee that a libc.so.6 somewhere else on
> > your path won't be used.
>
> the linker will perform a search for libc.so.6 before the current path (of the
> ldscript) ? if that's the case, then dropping the absolute paths wouldnt
> work
Yes, I'm pretty sure that's right.
> the patch is needed when using sysrooted ld with an incorrect prefix for
> glibc ... one thing we experimented with was this patch and rather than
> simply discard it since we've moved on to other methods, i thought about
> getting it integrated if it proved useful. the idea it may actually be
> useful was spurred by the fact that crosstool as well strips all paths from
> glibc linker scripts, but i dont have any history as to why it does that.
It predates sysroot support in binutils; you needed to do this if you
didn't have a linker that knew to add the prefix.
--
Daniel Jacobowitz
CodeSourcery
More information about the Libc-alpha
mailing list