to the XDMCP specification at
http://cgit.freedesktop.org/xorg/doc/xorg-docs/plain/hardcopy/XDMCP/xdmcp.PS.gz,
it's just a junk trailer following the zero-length array of
authentication names, and that "no padding of any sort will occur within
the packets." It still correctly identifies an xdm running locally in my
testing.
The specification also says "Packets that have too little or too much
data should be ignored," but that must not be taken seiously because the
X server that comes with Mac OS X sends several junk null bytes at the
end of its XDMCP queries.
properly so our signal handing code actually activates. But our signal
handling code was a bit questionable in many respects, so I removed most
of it under the theory that it hasn't been working for years and nobody
seemed to notice/care, so we presumably don't really need it. And if we
do decide to add it again, we basically will need to do it over because
we now want to do it safely from a reentrancy perspective. In particular,
I removed sigdie() and reaper(). We still have a few signal() calls. In
particular, we ignore SIGPIPE and we set SIGINT, SIGTERM, SIGHUP, and
SIGSEGV to the defaults (SIG_DFL) explicitly. Thanks to Ithilgore, Luis,
and Solar Designer in particular for fidning or commenting on the issue.
socket have been idle for the given time. Previously it would exit
when *either* of them had been idle, meaning that the program would
quit contrary to your expectation when downloading a large file
without sending anything, for example.
or "NCAT DEBUG: " to make it clear that they are not coming from the
remote host. This only matters when output goes to a terminal, where
the standard output and standard error streams are mixed. [David]
- Cookie Support
- Pipelining Support
- New lib modularization (previously, single functions were taking care of too many tasks, we have splitted these tasks for better code reuse)
Some threads have truncated stacks after finishing with an error causing
a Lua assertion failure when we put elements on the stack without first
checking whether space is available.
is 6000-6009 *or* the service starts with "X11". In this case it didn't
matter because those ports do start with "X11" in nmap-services, but
this better matches the shortport.port_or_service model.
Unfortunately, it seems cygwin doesn't know how to expand a variable like %ProgramFiles% so that the Makefile would be locale-agnostic, so I just modified it for the most common version of "Program Files".
