When we do not explicitly catch G_IO_ERR and G_IO_HUP, GLib can go
into an infinite loop, because it won't deliver the socket error to
MPD.

GIOChannel is more portable than raw read()/write() calls.  We're
using GIOChannel anyway, because we need it for plugging the client
into the GLib main loop.

Configure the GIOChannel to the bare minimum: no character set, no
buffering.

Use g_io_channel_win32_new_socket() instead of g_io_channel_unix_new()
on WIN32.

On some platforms, g_free() must be used for memory allocated by
GLib.  This patch intends to correct a lot of occurrences, but is
probably not complete.

Simplify some code by using config_get_positive(), instead of doing
manual parsing and validation each time.

Renamed functions, types, variables.

Get rid of the non-portable Linux list library, part I.

The length of GREETING is known at compile time, don't use strlen().

We are going to migrate away from the concept of notifying the main
thread.  There should be events sent to it instead.  This patch starts
a series to implement that.

With the GLib main loop, the client manager can install its own event
in case a client is expired.  No need for main.c to call
client_manager_expire() manually.

Remove the event source from the GMainLoop object in
client_set_expired().

Other libraries may need to access the main_loop reference, to add or
remove events, or to call g_main_loop_quit().

This is a rather huge patch, which unfortunately cannot be splitted.

Instead of using our custom ioops.h library, convert everything to use
the GLib main loop.

On Windows, socket declarations reside in winsock.h and ws2tcpip.h.
The POSIX headers sys/socket.h etc. are not available.

When a response is very long (e.g. a large playlist > 100k songs),
most of it will end up in the deferred buffers.  Filling the deferred
queue is very expensive currently, because a new buffer is allocated
for every client_write() operation.  This may lead to long delays, and
the client might give up and disconnect meanwhile.  This patch makes
MPD attempt to flush the deferred queue as often as possible, to work
around this problem.  Due to the MPD 0.14 code freeze, we should not
optimize the buffering code now.

We want to remove gcc.h eventually. This takes care of all the G_GNUC_PRINTF
macros.

The client may provide the names of idle events as arguments to the
"idle" command to inform MPD that it is only interested in these
events.

Return bool instead of int.

After a partial write, chances are vanishing that another write() will
succeed.  Don't try immediately.

Another custom data structore converted to GLib.

Replace a custom data structure with a GLib one.

The list cache aims to save memory allocations, and complicates the
code a bit.  We should rather use GLib slices later, which are easy to
use.

Eliminate CamelCase in all public and static functions.

client->permission is a bit set, and should be unsigned.

Reduce two temporary variables to only one.

There is no sense in using the kernel's send buffer size (SO_SNDBUF)
for MPD's send buffer.  Convert it into a static buffer of 4 kB.

Use a literal in the struct declaration, and sizeof(client->buffer)
everywhere else.  Also shrink the buffer from 40 kB to 4 kB.  The
buffer must only be large enough to hold one line of input, and 4 kB
is still more than enough.

The buffer pointers must not exceed the buffer size.

Use ssize_t instead of int.

Commit 6eb62e47 didn't obey partial lines correctly: when a line
wasn't finished in one read, the first part was ignored when the rest
arrived.

Enable authentication over unix sockets.  Store the client's uid in
the client struct.

Use memchr() instead of manually traversing the input buffer.  Update
the client's properties after all commands have been processed.  Check
for buffer overflow once.

"idle" waits until something noteworthy happens on the server,
e.g. song change, playlist modified, database updated.  This allows
clients to keep up to date without polling.

When there are standardized headers, use these instead of the bloated
os_compat.h.