rofl0r

thanks for info. can you try each of those addresses for -i parameter and report back what's visible in netstat ? (also whether you can connect to v4 and v6...

> -> Working as ecpected means you can connect through both ipv4 and v6 ?

SO_BINDTODEVICE seems to be a linux-specific API. what prevents you from using the ip address of the specified interface with the existing -b option?

i can't merge this PR as-is, because it would break build on solaris, *BSD, and OSX

i dont want to litter the code with ugly ifdefs, they need to be in a single spot, if at all

> I can provide traces if needed yeah, let's start with a strace including the startup of microsocks and an outgoing connection that tries bind to v4 addr but fails....

> strace /usr/bin/microsocks this should have been `strace -f` as the interesting bit of information (what happens in client thread) is missing. do you also have netstat -lnat available ?

you seem to be using an old version of microsocks that still uses select(). the strace log is incredibly cluttered thanks to glibc's retarded getaddrinfo() implementation and it took me...

interesting, this is more conclusive. the dns result returned for ident.me seems to prefer the ipv6 address (microsocks always uses first result) and so the outgoing socket is bound to...

i won't have time to look into this today to find a good solution. i guess one thing one could do is to iterate over gai results and prefer family...