The __func__ macro is part of the C99 standard whereas __FUNCTION__ is a
legacy gcc specific alias for it:
http://gcc.gnu.org/onlinedocs/gcc/Function-Names.html
Additionally, checkpatch.pl that's commonly used to verify coding style
also recommends to use __func__ instead of __FUNCTION__.
g_atomic_* end up using G_STATIC_ASSERT, causing gcc 4.8 to yell due to
-Wunused-local-typedefs.
/usr/include/glib-2.0/glib/gmacros.h:162:53: error: typedef ‘_GStaticAssertCompileTimeAssertion_2’ locally defined but not used [-Werror=unused-local-typedefs]
#define G_STATIC_ASSERT(expr) typedef char G_PASTE (_GStaticAssertCompileTimeAssertion_, __COUNTER__)[(expr) ? 1 : -1]
Most of the uses of atomic operations were wrong. They were fixed as
well. If we are using atomic operations, reading the variable again
later for logging is not an option, we should use the return of the
atomic function used to fetch the variable.
If attrib is freed in cmd->func(), then it will be used if either
request or response queue has some data to send.
This patch moves calling wake_up_sender() which increases the ref
count of attrib so that it wont get freed in cmd->func().
We want only the profile that implements a service to be notified of
changes on that service. Before this patch, all the registered event
notifiers are being called.
size_t/ssize_t/off_t/etc are more appropriate for variables denoting
some kind of size than simply using int.
This patch includes a couple of other related changes to avoid gcc
signedness errors resulting from it treating (for whatever reason) const
variables and integer literals as signed.
The attrib server code relies on these id's to be unique globally and
not just per GAttrib instance. As an easy fix make them global by adding
a static guint to g_attrib_register.
The callbacks could result with the reference count dropping to 0 and
the object being freed. This patch fixes the issue by adding one extra
reference for the duration of the timeout function.
23 octets is the default (and minimum) ATT_MTU value. If someone tries
to set ATT_MTU less than 23 octets g_attrib_set_mtu should fail (return
FALSE). Additionally, there is no constraint regarding the maximum value
of ATT_MTU, so we should not check for it.
Also, we should not change the L2CAP ATT fixed channel MTU. bt_io_set
call will always fail since we are not supposed to change L2CAP MTU
after connection is established.
GAttrib buffer should be allocated according to ATT_MTU value. Over
BR/EDR, ATT_MTU should be set to the L2CAP imtu negotiated during
L2CAP configuration phase. Over LE, ATT_MTU should be 23 octets.
A timer is set when a response is expected. The timer is removed when
data is received, regardless of whether or not the data is a response.
As a result, the timer may be cleared even though a response was not
received and there would be no way to detect a command timeout.
Fix this by clearing the timer only after verifying a response was
received.
This patch fixes command timeout handling. Previously attrib_destroy was
explicitly called which ignored any reference holders. This patch fixes
the issue by first passing errors to command callbacks and after that
marking the GAttrib object as stale so no further operations can be
done.
New requests and responses are never sent if a request was sent and the
response for it hasn't been received yet. As a result, if both end
points send requests at the same time, a deadlock could occur. This
could happen, for instance, if the client sends a read request and the
server sends an indication before responding to the read request.
Fix this by introducing an additional queue for responses. Responses may
be sent while there's still a pending request/indication.
The new buffer is allocated in g_attrib_new() and it will be used to
send/receive PDUs. The buffer size is the MTU read from L2CAP channel
limited to ATT_MAX_MTU. Functions to handle the buffer size were also
created.
This patch puts the new UUID functions into use for GATT-related
code, and adds some convenience functions to ATT API (att.h).
Example GATT server is also changed.
This patch adds support to destroying the GATT connection
when a GATT server doesn't respond for more than 30 seconds.
A function to destroy the GAttrib is introduced and it is used
in the timeout case and when the last GAttrib reference is
dropped.
Overall purpose of change is to enable a GATT procedure to be
executed atomically, even if it requires multiple ATT
request/response transactions.
Fix g_attrib_send() to include an ID parameter, if the pkt to
be sent should be added to the Head of the pkt queue. If the
ID is Zero, legacy functionality is maintained, and the pkt will
be added at the tail of the queuer, and a new ID will be generated,
and returned to the caller. If ID is non-zero, the pkt will be
added to the head of the queue, with the ID value requested, which
will also be returned to the caller.
Fix received_data() to not service the send queue until after the
received data has been processed by calling the cmd->func()
callback, to allow the callback to insert another pkt on the head
of the queue.
Fix all callers of g_attrib_send() to include new parameter.
If a characteristic requires a higher security level, change it on
demand and re-send the GATT Charateristic Value Read. Request will not
be sent until the SMP negotiation finishes. This change doesn't affect
GATT over BR/EDR, since encryption is mandatory for BR/EDR.
If the GIOChannel is in the buffered state (the default) the watch
function is called without receiving a POLLOUT from the socket. GLib
adds a G_IO_OUT condition just because there is space in the GIOChannel
internal buffer.
The solution is disabling the internal buffer, which in turn, makes the
call of g_io_channel_flush() useless.
In the client, after local or remote initiated disconnection the GAttrib
reference shall be set to NULL to allow a proper control of references
and further connections.
Add a new function to notify the GAttrib "users" when the GAttrib has
beed destroyed. The disconnect function is called only for remote initiated
disconnections or external events(not unref).
