Release 1.3.4

Make the 1.3 branch build with --disable-debug again. Fixes LM-113. committer: Mikael Hallendal <micke@imendio.com>

Merge branch 'master' of git@github.com:hallski/loudmouth

Close the socket before setting it to -1. This hopefully solves LM-95. Added a socket_close_io_channel call that unrefs the IOChannel and closes the file descriptor.

Some code style fixes

Return FALSE from lm_connection_open in the synchronous case instead of open callback This patch fixes things up so that synchronous failures don't call the open callback but instead return the correct FALSE result from lm_connection_open(). Otherwise this means that we will get the callback before the lm_connection_open call has returned. committer: Mikael Hallendal <micke@imendio.com>

Fixed reentrancy problem for async connections. Fixes first bit of LM-117. Patch adding reference to sockets in places where otherwise we can run into trouble where the user callback causes the socket to be freed. (Note that we actually double ref the socket in some cases ...both inside lm_socket_fail_with_error() and in the caller. This is, of course, harmless, and needed because the inner ref is necessary to handle the calls to _lm_socket_fail_with_error() from lm-proxy.c, the outer ref necessary because we reference socket after calling _lm_socket_fail_with_error()) committer: Mikael Hallendal <micke@imendio.com>

Protect EAI_NODATA with #ifdef. Fixes LM-120. A recent change to GNU libc changes the EAI_NODATA constant in netdb.h to only be defined when _GNU_SOURCE is defined. A google for "loudmouth EAI_NODATA" indicates that EAI_NODATA is also problematical on FreeBSD. I'm attaching a patch that protects the usage of EAI_NODATA with an ifdef ... I think this is preferable to the sledgehammer of _GNU_SOURCE; while GNU libc can apparently still return this error value, using the default message in this rare corner case should not be a problem. committer: Mikael Hallendal <micke@imendio.com>

Fixed the keep alive rate when using TCP for handling keep alive. Fixes LM-121. The keep alive rate was multiplied with 1000 to fit into g_timeout_add. Moved it so that the multiplication is done when calling g_time_add instead.

Added new line for g_log printouts