🚀 go-pugleaf

The following reply was made to PR ports/6930; it has been noted by GNATS.

From: Doug Ambrisko <ambrisko@whistle.com>
To: kkennawa@physics.adelaide.edu.au
Cc: freebsd-gnats-submit@freebsd.org
Subject: Re: ports/6930: socks5 port broken by /usr/lib/aout changes?
Date: Fri, 12 Jun 1998 13:55:35 -0700 (PDT)

 kkennawa@physics.adelaide.edu.au writes:
 | Recent /usr/lib/aout changes seem to have broken at least part of the socks5
 | port (/usr/ports/net/socks5). Specifically, the 'runsocks' command no
 | longer seems to work - this command performs a runtime re-linking of executables
 | to use the socks5 libraries and thus communicate through a socks5 proxy.
 | >How-To-Repeat:
 | Install port on a post-/usr/lib/aout machine, then attempt 'runsocks <cmd>' where <cmd>
 | is a non-socks aware command, such as telnet (yeah, theres a rtelnet installed by the port,
 | thats not the point :-)
 |
 | runsocks telnet immediately exits to the command line with no errors
 | runsocks ftp hangs, refuses to respond to ^C or most other keystrokes
 |
 | Pre-aout changes, runsocks worked fine for me :)

 Here's some help.

 The problem you will run into is when the libsocks.so shim does a dlopen
 of libc etc (the same problem as with ppp -alias if I recall).  Sym-links
 of the a.out version to /usr/lib should fix it.  Alternatively rebuild
 libsocks.so with /usr/lib/aout for the path.  This will require a patch
 to the "configure" script.

 What runsocks does is to setup LD_PRELOAD to install the shim.  The for
 example:

 	Application
 		-> call sendto
 			-> libsocks version of sendto intercepts this
 			   call and socks'ifies it and does a dlopen
 			   of libc (since it is pre-loaded before libc)
 				-> the socks'ified network call go direct
 				   into libc.
 					-> kernel does its stuff


 This may all blow over and just work again when elf is done.

 Doug A.

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-ports" in the body of the message