Re: multilib dirs and ld.so

[Mike Frysinger]

On Wednesday 22 August 2007, Peter O'Gorman wrote:
> On Wed, 2007-08-22 at 06:40 -0400, Mike Frysinger wrote:
> > On Wednesday 22 August 2007, Peter O'Gorman wrote:
> > > libsuff=`ldd conftest 2>/dev/null | awk '/libc\.so/
> > > {n=split([$]3,x,"/"); for (i=0; i < n; i++) { if (x[[i]] == "lib64")
> > > {print "64"}}}'`
> >
> > this test would still be subject to the environment of ldd ... in
> > otherwords, it is still possible to have ldd return /lib/libc.so.6 when
> > /lib is a symlink to /lib64 ...
> >
> > since the dynamic loader is the only hardcoded path, wouldnt it be better
> > to check that ?  so drop the /libc\.so/ and change the $3 to $1 ...
>
> Makes sense to me. However, I think we'd need to match the first
> absolute path output from ldd because:
>
> address@hidden ]$ ldd fool
>         linux-gate.so.1 =>  (0x009be000)
>         libc.so.6 => /lib/libc.so.6 (0x00b82000)
>         /lib/ld-linux.so.2 (0x00b65000)

i'm not sure how this sample output explains the comment that we need to match 
the first absolute path ?  afaik, the only absolute path ive ever seen in $1 
is the ld.so as ldd shows:
(1) [DT_NEEDED form] => [resolve DT_NEEDED entry to full path] (load addr)
(2) the interpreter on its own line
(3) the linux-gate/vdso stuff

> So, how about the attached?

seems to work on my Gentoo/amd64

> Ralph, the only linux that I have access to outside of TWW are debian 32
> bit, and Fedora 32 bit, glad that your list is so much longer :)

if you're interested, i have no problem giving out access to some exoctic 
Linux/Gentoo setups ... i know Ralph has a login on one of our Gentoo/parisc 
boxes ...
-mike

signature.asc
Description: This is a digitally signed message part.