WANProxy 0.5.2 release notes.

[Juli Mallett]

WANProxy 0.5.2 is now available from Subversion and on the web!

Subversion: http://wanproxy.org/svn/releases/wanproxy-0.5.2
WWW: http://wanproxy.org/releases/wanproxy-0.5.2.tar.gz

This release features:
o) Adds support for epoll(2) as a polling mechanism on Linux.
o) Adds support for port(3C) as a polling mechanism on SunOS.
o) Adds a workaround for a compiler bug in OpenBSD, making WANProxy
work on OpenBSD again.
o) More test and example code if you're thinking of extending WANProxy.
o) The configuration system will now tell you what is wrong when there
is a configuration error instead of just telling you that there is
one.
o) Hash-comments (`#') are now allowed at the end-of-line in a
configuration file, not just at the start.
o) Increased support for Debian GNU/Interix 3.5.  May work with stock
Interix 3.5 and a good version of GCC, too.  WANProxy runs on Windows!
 If you want WANProxy on a newer version of Interix, it should be
trivial.
o) A Splice (and SplicePair) mechanism has been added to join Channels
to each other with Pipes.
o) Switched to using the new Splice mechanism for proxying, which adds
some overhead in exchange for badly-needed flexibility.
o) Some performance improvements to the XCodec encoder.
o) TCPClient and UDPClient are less horrible now and no longer cause
sockets to be leaked when something weird happens.  This is mostly an
API fix but can help with socket exhaustion.  UnixClient is still
backwards.
o) The file descriptor resource limit will be increased at startup
where it is possible to do so.
o) Stub getaddrinfo(3) and getnameinfo(3) implementations have been
provided for systems that do not provide them, like Interix 3.5.
o) Work-around a bug with Mac OS X 10.6 (Snow Leopard): their
implementation of getaddrinfo(3) has regressed in its ability to
handle a service of "0" but not other numeric ports.
o) SO_REUSEADDR is now set at startup so that if WANProxy crashes or
needs restarted on a system that does not support tcpdrop(8), you can
recover quickly.
o) We now no longer rely on bsd.obj.mk.
o) The default logging level for tack has been reduced, which makes it
much less noisy.  (A -v option has been added.)
o) SOCKS5 using host names now works for IPv6 as well as IPv4.
Explicit IPv6 support for SOCKS5 is still missing.

Note: this release is mostly going out because bugfixes have been
accumulating and the key features that have been a goal of late have
not materialized due to a lack of time in suitably-large chunks.  In
the meantime, testing of the new Pipe/Splice/SplicePair infrastructure
would be valuable, especially with an eye to memory leaks, etc.