platform/upstream/libnice.git
10 years agotcp-bsd: Don't clear error that hasn't been set
Olivier Crête [Wed, 2 Apr 2014 15:25:27 +0000 (11:25 -0400)]
tcp-bsd: Don't clear error that hasn't been set

10 years agoagent: Don't return the socket to the app in any reliable case
Olivier Crête [Wed, 9 Apr 2014 03:10:34 +0000 (23:10 -0400)]
agent: Don't return the socket to the app in any reliable case

10 years agoNo need to strdup foundation in new_selected_pair since we queue signal
Youness Alaoui [Tue, 8 Apr 2014 01:30:49 +0000 (21:30 -0400)]
No need to strdup foundation in new_selected_pair since we queue signal

10 years agopseudotcp: Also cause Writable callback to be emitted if can_send() says no
Olivier Crête [Wed, 9 Apr 2014 02:43:30 +0000 (22:43 -0400)]
pseudotcp: Also cause Writable callback to be emitted if can_send() says no

The GOutputStream based systems can sometimes call this, decide the stream is
not writable and then poll for it. We have to make sure that poll is woken up.

10 years agoFix TCP TURN for Google compatibility. Message length was sent uninitialized
Youness Alaoui [Fri, 4 Apr 2014 06:43:20 +0000 (02:43 -0400)]
Fix TCP TURN for Google compatibility. Message length was sent uninitialized

10 years agoSmall cleaning of tcp-bsd.c
Youness Alaoui [Tue, 1 Apr 2014 00:36:58 +0000 (20:36 -0400)]
Small cleaning of tcp-bsd.c

10 years agoFix possible segfault in stun_message_validate_buffer_length()
Jakub Adam [Sat, 4 Jun 2011 21:25:33 +0000 (23:25 +0200)]
Fix possible segfault in stun_message_validate_buffer_length()

10 years agobuild: Check for [s]size_t before redefining them on MinGW
Philip Withnall [Thu, 3 Apr 2014 11:29:04 +0000 (12:29 +0100)]
build: Check for [s]size_t before redefining them on MinGW

MinGW defines size_t and ssize_t for us, so we should not
unconditionally redefine them in stun/win32_common.h. Add an
AC_CHECK_TYPES configure check to avoid this.

10 years agoagent: Define IN6_ARE_ADDR_EQUAL macro when building under MinGW
Philip Withnall [Thu, 3 Apr 2014 11:22:27 +0000 (12:22 +0100)]
agent: Define IN6_ARE_ADDR_EQUAL macro when building under MinGW

Older versions of MinGW don’t define IN6_ARE_ADDR_EQUAL, as normally
defined in netinet/in.h. Conditionally define it in address.c to fix the
build on MinGW.

MinGW has added support for the macro here:
https://github.com/mirror/mingw-w64/commit/0f4899473c4ba2e34fa447b1931a04e38c1f105e

Based on a patch originally by Luciana Fujii Pontello
<luciana@fujii.eti.br>.

10 years agopseudotcp: Validate option lengths
Olivier Crête [Wed, 2 Apr 2014 01:41:04 +0000 (21:41 -0400)]
pseudotcp: Validate option lengths

10 years agoRemove all traces of compile time marshallers
Olivier Crête [Tue, 1 Apr 2014 15:00:55 +0000 (11:00 -0400)]
Remove all traces of compile time marshallers

10 years agoagent: Use generic marshaller
Olivier Crête [Mon, 31 Mar 2014 23:28:24 +0000 (19:28 -0400)]
agent: Use generic marshaller

10 years agoInitialize variables to reduce unexpected behavior in case something goes wrong
Olivier Crête [Mon, 31 Mar 2014 23:17:28 +0000 (19:17 -0400)]
Initialize variables to reduce unexpected behavior in case something goes wrong

10 years agoRename variables that shadow global definitions
Olivier Crête [Mon, 31 Mar 2014 23:15:07 +0000 (19:15 -0400)]
Rename variables that shadow global definitions

10 years agostun: Use struct sockaddr_storage * for any pointer that we write to
Olivier Crête [Mon, 31 Mar 2014 22:32:40 +0000 (18:32 -0400)]
stun: Use struct sockaddr_storage * for any pointer that we write to

Some platforms have a larger alignment requirement for struct sockaddr_* than
for plain struct sockaddr, in that case, we need to make sure that the incoming pointer
has the right alignement before taking it in.

10 years agostun tests: Fix a bunch of issues found by more aggressive GCC warnings
Olivier Crête [Mon, 31 Mar 2014 21:46:34 +0000 (17:46 -0400)]
stun tests: Fix a bunch of issues found by more aggressive GCC warnings

10 years agoDon't mix gsize vs size_t, seems like they're not always the same size
Olivier Crête [Mon, 31 Mar 2014 21:46:15 +0000 (17:46 -0400)]
Don't mix gsize vs size_t, seems like they're not always the same size

10 years agocandidate: Use explicit 64-bit constants to make 32-bit gcc happy
Olivier Crête [Mon, 31 Mar 2014 21:45:36 +0000 (17:45 -0400)]
candidate: Use explicit 64-bit constants to make 32-bit gcc happy

10 years agopseudotcp: Don't mix "long" and guint, one is signed, the other is unsigned
Olivier Crête [Mon, 31 Mar 2014 21:44:45 +0000 (17:44 -0400)]
pseudotcp: Don't mix "long" and guint, one is signed, the other is unsigned

So use guint64 everywhere, this is never supposed to be negative.

10 years agotcp-turn: Don't access output message after freeing it
Olivier Crête [Sun, 30 Mar 2014 01:28:02 +0000 (21:28 -0400)]
tcp-turn: Don't access output message after freeing it

10 years agoturn: Don't overwrite source, destroy it first
Olivier Crête [Sun, 30 Mar 2014 01:23:40 +0000 (21:23 -0400)]
turn: Don't overwrite source, destroy it first

10 years agoturn: Don't access list node after freeing it
Olivier Crête [Sun, 30 Mar 2014 01:23:11 +0000 (21:23 -0400)]
turn: Don't access list node after freeing it

10 years agocomponent: Set dummy callback from cancellable GSource
Olivier Crête [Fri, 28 Mar 2014 21:38:01 +0000 (17:38 -0400)]
component: Set dummy callback from cancellable GSource

10 years agoagent: Make it possible to set the relays after gathering
Olivier Crête [Fri, 28 Mar 2014 01:10:33 +0000 (21:10 -0400)]
agent: Make it possible to set the relays after gathering

This allows finding the relays asynchronously while trying to connect
with other methods.

10 years agoagent: Prevent nice_agent_set_port_range() from being called after a gather
Olivier Crête [Fri, 28 Mar 2014 01:10:02 +0000 (21:10 -0400)]
agent: Prevent nice_agent_set_port_range() from being called after a gather

The port range is only used when gathering, print a critical if it is
called afterwards as it will have no effect.

10 years agoagent: Prevent gather from being called multiple times
Olivier Crête [Fri, 28 Mar 2014 01:09:15 +0000 (21:09 -0400)]
agent: Prevent gather from being called multiple times

This can cause odd behaviors as it is not a idempotent function.

10 years agoagent: Refactor nice_agent_set_relay_info()
Olivier Crête [Fri, 28 Mar 2014 01:06:55 +0000 (21:06 -0400)]
agent: Refactor nice_agent_set_relay_info()

No behavior change is introduced

10 years agoturn: Put timeouts on the specific main context
Olivier Crête [Fri, 28 Mar 2014 00:13:55 +0000 (20:13 -0400)]
turn: Put timeouts on the specific main context

10 years agoturn: Don't try to process empty messages
Olivier Crête [Thu, 27 Mar 2014 21:55:31 +0000 (17:55 -0400)]
turn: Don't try to process empty messages

10 years agoagent: Bail out early on empty message
Olivier Crête [Thu, 27 Mar 2014 21:55:01 +0000 (17:55 -0400)]
agent: Bail out early on empty message

This can happen if a lower layer has consumed the message, bail out earlier in
that case.

10 years agoagent: Don't overwrite the size of the GInputBuffer, only set the global size
Olivier Crête [Thu, 27 Mar 2014 21:54:00 +0000 (17:54 -0400)]
agent: Don't overwrite the size of the GInputBuffer, only set the global size

10 years agopseudotcp: Also accept NiceInputMessage array with a single message
Olivier Crête [Thu, 27 Mar 2014 21:51:03 +0000 (17:51 -0400)]
pseudotcp: Also accept NiceInputMessage array with a single message

10 years agotest-fullmode: Don't use gmainloop when not required
Olivier Crête [Thu, 27 Mar 2014 21:50:31 +0000 (17:50 -0400)]
test-fullmode: Don't use gmainloop when not required

There are more than one callback that could stop it

10 years agotest-fullmode: Do #if instead of #ifdef as we set it to 0
Olivier Crête [Thu, 27 Mar 2014 21:50:08 +0000 (17:50 -0400)]
test-fullmode: Do #if instead of #ifdef as we set it to 0

10 years agotest-fullmode: Only wait for data on component 1
Olivier Crête [Thu, 27 Mar 2014 21:49:43 +0000 (17:49 -0400)]
test-fullmode: Only wait for data on component 1

10 years agotcp-bsd: Use default main context if none specified
Olivier Crête [Thu, 27 Mar 2014 16:33:23 +0000 (12:33 -0400)]
tcp-bsd: Use default main context if none specified

10 years agoturn-tcp: Don't corrupt messages
Olivier Crête [Thu, 27 Mar 2014 07:54:32 +0000 (03:54 -0400)]
turn-tcp: Don't corrupt messages

10 years agoagent: Free every parameter
Olivier Crête [Sat, 15 Mar 2014 03:53:06 +0000 (23:53 -0400)]
agent: Free every parameter

Fixes off-by-one error

10 years agoagent: Create a custom pointer GType to recognize the stream ids
Olivier Crête [Sat, 15 Mar 2014 03:52:45 +0000 (23:52 -0400)]
agent: Create a custom pointer GType to recognize the stream ids

10 years agoagent: Improve documentation for NiceAgent::new-candidate
Philip Withnall [Fri, 7 Mar 2014 09:18:04 +0000 (09:18 +0000)]
agent: Improve documentation for NiceAgent::new-candidate

Note that it’s a new _local_ candidate.

10 years agoNEWS: Remove superfluous empty lines
Olivier Crête [Fri, 7 Mar 2014 02:38:07 +0000 (21:38 -0500)]
NEWS: Remove superfluous empty lines

10 years agoVersion 0.1.5.1
Olivier Crête [Fri, 7 Mar 2014 02:30:26 +0000 (21:30 -0500)]
Version 0.1.5.1

10 years agoVersion 0.1.5 0.1.5
Olivier Crête [Fri, 7 Mar 2014 02:21:37 +0000 (21:21 -0500)]
Version 0.1.5

10 years agoconfigure: Remove -Wunsafe-loop-optimizations, it causes false warnings
Olivier Crête [Fri, 7 Mar 2014 01:51:23 +0000 (20:51 -0500)]
configure: Remove -Wunsafe-loop-optimizations, it causes false warnings

10 years agotests: Add private header to tarball
Olivier Crête [Fri, 7 Mar 2014 01:45:43 +0000 (20:45 -0500)]
tests: Add private header to tarball

10 years agotests: Don't verify the message content with not in reliable mode, because its unreliable
Olivier Crête [Fri, 7 Mar 2014 01:00:38 +0000 (20:00 -0500)]
tests: Don't verify the message content with not in reliable mode, because its unreliable

10 years agoagent: Enforce limits on the size buffers, because the retval is signed
Olivier Crête [Fri, 7 Mar 2014 00:59:45 +0000 (19:59 -0500)]
agent: Enforce limits on the size buffers, because the retval is signed

10 years agodoc: Update doc generation
Olivier Crête [Thu, 6 Mar 2014 22:06:16 +0000 (17:06 -0500)]
doc: Update doc generation

10 years agoiostream: De-gtkdoc the comments, as it's not public API
Olivier Crête [Thu, 6 Mar 2014 22:06:06 +0000 (17:06 -0500)]
iostream: De-gtkdoc the comments, as it's not public API

10 years agoagent: Remove unused #include
Philip Withnall [Thu, 6 Mar 2014 16:00:07 +0000 (16:00 +0000)]
agent: Remove unused #include

net/if_arp.h doesn’t exist on OS X, and hence causes cross-compilation
to fail. It doesn’t appear to be needed for any of the functions
currently used in interfaces.c anyway.

10 years agoagent: Ensure the pending signal queue is cleared on dispose
Philip Withnall [Wed, 5 Mar 2014 10:45:00 +0000 (10:45 +0000)]
agent: Ensure the pending signal queue is cleared on dispose

10 years agoagent: Fix a potential re-entrancy situation
Philip Withnall [Tue, 4 Mar 2014 13:48:03 +0000 (13:48 +0000)]
agent: Fix a potential re-entrancy situation

If we emit signals from component_io_cb() in the middle of a read, it’s
possible that one of those signals will be picked up by the client to
perform another read. A likely candidate (and what was triggering the
re-entrancy here) would be component-state-changed. Since signals are
emitted synchronously, the second read would being inside the first, and
trigger the anti-re-entrancy assertion.

10 years agotests: Use dynamically allocated port for test-new-dribble
Olivier Crête [Tue, 25 Feb 2014 05:48:56 +0000 (00:48 -0500)]
tests: Use dynamically allocated port for test-new-dribble

10 years agotests: Fix horribly raciness in test-new-dribble
Olivier Crête [Tue, 25 Feb 2014 05:23:47 +0000 (00:23 -0500)]
tests: Fix horribly raciness in test-new-dribble

10 years agotests: Split racy "closing" into a test for stopping while reading or writing
Olivier Crête [Tue, 25 Feb 2014 04:05:44 +0000 (23:05 -0500)]
tests: Split racy "closing" into a test for stopping while reading or writing

We wan't to make sure not to have a race where a thread blocks in a blocking
receive or send while nice_agent_remove_stream() is called

10 years agotest: Use a less racy way to know if the test is ready to start
Olivier Crête [Tue, 25 Feb 2014 04:03:42 +0000 (23:03 -0500)]
test: Use a less racy way to know if the test is ready to start

Also, keep a pointer to the other structure for tests to use

10 years agooutputstream: Add a cancellable to get out of blocking write if the stream is removed
Olivier Crête [Tue, 25 Feb 2014 04:02:39 +0000 (23:02 -0500)]
outputstream: Add a cancellable to get out of blocking write if the stream is removed

10 years agocomponent: Add a GCancellable to get of out blocking nice_agent_recv()
Olivier Crête [Tue, 25 Feb 2014 04:02:04 +0000 (23:02 -0500)]
component: Add a GCancellable to get of out blocking nice_agent_recv()

Otherwise, it may be stuck iterating the context forever if all of the
other sources are gone.

10 years agocomponent: Keep its own context at all times
Olivier Crête [Tue, 25 Feb 2014 04:00:33 +0000 (23:00 -0500)]
component: Keep its own context at all times

Don't re-create it.

10 years agoIgnore .trs and .log file from newer automake
Olivier Crête [Mon, 24 Feb 2014 23:59:39 +0000 (18:59 -0500)]
Ignore .trs and .log file from newer automake

10 years agoagent: Delay signal emission after the lock has been released
Olivier Crête [Mon, 24 Feb 2014 23:50:59 +0000 (18:50 -0500)]
agent: Delay signal emission after the lock has been released

This way, there can be no annoying re-entrancy in our code.

10 years agoagent: Replace recursive mutex with non-recursive
Olivier Crête [Mon, 24 Feb 2014 23:51:31 +0000 (18:51 -0500)]
agent: Replace recursive mutex with non-recursive

This should prevent us from re-adding re-entrancy in the future

10 years agoagent: Delay signal emission after the lock has been released
Olivier Crête [Mon, 24 Feb 2014 23:50:59 +0000 (18:50 -0500)]
agent: Delay signal emission after the lock has been released

This way, there can be no annoying re-entrancy in our code.

10 years agoiostream: remove re-entrant call into NiceAgent
Olivier Crête [Mon, 24 Feb 2014 22:25:50 +0000 (17:25 -0500)]
iostream: remove re-entrant call into NiceAgent

10 years agooutputstream: If tcp, only check for space in the PseudoTCP queue
Olivier Crête [Mon, 24 Feb 2014 23:39:21 +0000 (18:39 -0500)]
outputstream: If tcp, only check for space in the PseudoTCP queue

10 years agoagent: Expand documentation for a couple of NiceAgent functions
Philip Withnall [Tue, 11 Feb 2014 10:01:18 +0000 (10:01 +0000)]
agent: Expand documentation for a couple of NiceAgent functions

Expand documentation for nice_agent_gather_candidates() and
nice_agent_set_remote_credentials().

10 years agoagent: Fix inlining of nice_debug() if NDEBUG is defined
Philip Withnall [Fri, 31 Jan 2014 11:52:15 +0000 (11:52 +0000)]
agent: Fix inlining of nice_debug() if NDEBUG is defined

10 years agoagent: Assert on re-entrant reads
Philip Withnall [Fri, 31 Jan 2014 11:43:19 +0000 (11:43 +0000)]
agent: Assert on re-entrant reads

Ensure the agent’s context doesn’t get iterated while in the middle of
reading a message, as that will corrupt the component->recv_messages
state.

10 years agoagent: Improve debug output
Philip Withnall [Fri, 31 Jan 2014 11:43:03 +0000 (11:43 +0000)]
agent: Improve debug output

10 years agoagent: Add nice_input_message_iter_compare()
Philip Withnall [Fri, 31 Jan 2014 11:42:16 +0000 (11:42 +0000)]
agent: Add nice_input_message_iter_compare()

Use it instead of memcmp() to avoid comparing padding bytes.

10 years agodocs: Remove NiceIOStream, nothing specific to document
Olivier Crête [Fri, 21 Feb 2014 20:05:59 +0000 (15:05 -0500)]
docs: Remove NiceIOStream, nothing specific to document

10 years agoagent: returning 0 valid messages is valid, it means EOF
Olivier Crête [Fri, 21 Feb 2014 19:34:28 +0000 (14:34 -0500)]
agent: returning 0 valid messages is valid, it means EOF

10 years agoiostream: iostreams are only for reliable streams, no need to check
Olivier Crête [Thu, 20 Feb 2014 06:41:22 +0000 (01:41 -0500)]
iostream: iostreams are only for reliable streams, no need to check

10 years agoagent: Use locally allocated memory if the application buffer is not big enough for...
Olivier Crête [Thu, 20 Feb 2014 06:40:30 +0000 (01:40 -0500)]
agent: Use locally allocated memory if the application buffer is not big enough for one STUN packet

10 years agooutputstream: Trigger the pollable source on error or removal
Olivier Crête [Wed, 19 Feb 2014 04:52:37 +0000 (23:52 -0500)]
outputstream: Trigger the pollable source on error or removal

10 years agoagent: Fix compiler warnings
Olivier Crête [Thu, 13 Feb 2014 20:14:23 +0000 (15:14 -0500)]
agent: Fix compiler warnings

10 years agoconfigure: Add newline ignoring
Olivier Crête [Wed, 5 Feb 2014 23:16:02 +0000 (18:16 -0500)]
configure: Add newline ignoring

Otherwise it ran the "yes" command

10 years agoconfigure: Move -Werror=unsafe-loop-optimizations to the maximum level
Olivier Crête [Wed, 5 Feb 2014 02:29:04 +0000 (21:29 -0500)]
configure: Move -Werror=unsafe-loop-optimizations to the maximum level

It tends to cause false warnings

10 years agopseudotcp: Remove one duplicated call to g_get_monotonic_time()
Olivier Crête [Fri, 31 Jan 2014 06:25:57 +0000 (01:25 -0500)]
pseudotcp: Remove one duplicated call to g_get_monotonic_time()

10 years agoagent: Remove weak pointers, they aren't thread safe anyway
Olivier Crête [Fri, 31 Jan 2014 06:06:08 +0000 (01:06 -0500)]
agent: Remove weak pointers, they aren't thread safe anyway

And we get close to 10% perf boost

10 years agopseudotcp: Add list of unsent segment
Olivier Crête [Fri, 31 Jan 2014 05:18:15 +0000 (00:18 -0500)]
pseudotcp: Add list of unsent segment

Looping through the list of send segment was the only thing from libnice
showing up in my profiler. Keeping a list of segment that were never sent
has made it disappear.

10 years agoagent: Avoid calling nice_address_to_string() when debugging is disabled
Olivier Crête [Fri, 31 Jan 2014 05:11:04 +0000 (00:11 -0500)]
agent: Avoid calling nice_address_to_string() when debugging is disabled

10 years agopseudotcp: Use GQueue for list that has stuff inserted in the end
Olivier Crête [Fri, 31 Jan 2014 04:33:16 +0000 (23:33 -0500)]
pseudotcp: Use GQueue for list that has stuff inserted in the end

10 years agopseudotcp: Don't change the receive buffer size if it hasnt changed
Olivier Crête [Fri, 31 Jan 2014 04:08:36 +0000 (23:08 -0500)]
pseudotcp: Don't change the receive buffer size if it hasnt changed

10 years agoagent: Restore the ability nice_agent_send() to send partial buffers
Olivier Crête [Fri, 31 Jan 2014 03:06:32 +0000 (22:06 -0500)]
agent: Restore the ability nice_agent_send() to send partial buffers

This is very important for reliable mode.

Also use it in the GOutputStream so as to not get into the case where
there is still some space in the TCP buffer, but not enough for one message.
Also warn against this problem.

10 years agoagent: Replace nice_agent_build_io_stream() with nice_agent_get_io_stream()
Olivier Crête [Fri, 31 Jan 2014 02:59:33 +0000 (21:59 -0500)]
agent: Replace nice_agent_build_io_stream() with nice_agent_get_io_stream()

Also made the GIOStream into a singleton, it always returns the same one.
Also make it impossible to create a GIOStream for a non-existing stream/component

10 years agoRemove the "length" parameter from NiceOutputMessage
Olivier Crête [Fri, 31 Jan 2014 02:21:13 +0000 (21:21 -0500)]
Remove the "length" parameter from NiceOutputMessage

It was used correctly only half the time anyway

10 years agoRemove the "to" parameter from NiceOutputMessage
Olivier Crête [Fri, 31 Jan 2014 01:38:15 +0000 (20:38 -0500)]
Remove the "to" parameter from NiceOutputMessage

We never send multiple messages to separate targets in practice, so
this will simplify the code

10 years agoagent: Attaching to a NULL context should attach to the default one
Olivier Crête [Thu, 30 Jan 2014 23:37:25 +0000 (18:37 -0500)]
agent: Attaching to a NULL context should attach to the default one

Otherwise it would have attached to a newly created context

10 years agosocket: Add note that the packet dropping in tcp-bsd is broken behavior
Olivier Crête [Fri, 24 Jan 2014 08:02:38 +0000 (03:02 -0500)]
socket: Add note that the packet dropping in tcp-bsd is broken behavior

10 years agotests: Fix test-restart to only quit on state changes
Philip Withnall [Wed, 22 Jan 2014 12:36:33 +0000 (12:36 +0000)]
tests: Fix test-restart to only quit on state changes

Previously, there was a race between receiving the data packets and
changing state to READY, which would cause one of the final assertions
to fail.

10 years agoagent: Add support for vectored I/O for sends
Philip Withnall [Tue, 21 Jan 2014 15:56:18 +0000 (15:56 +0000)]
agent: Add support for vectored I/O for sends

Add one new public function, nice_agent_send_messages_nonblocking(),
which replaces nice_agent_send_full(). This isn’t an API break, because
nice_agent_send_full() hasn’t been in a release yet. The new API allows
sending multiple messages in a single call, and supports vectors of
buffers to transmit the messages from.

The existing nice_agent_send() API has been left untouched, although
it’s a bit of a bugbear because it’s non-blocking and doesn’t fit with
the new *_nonblocking() naming scheme. Oh well.

This doesn’t bring any notable changes to the number of memcpy()s on the
critical path: it remains at zero for the common cases and common socket
types. It introduces the possibility for future work to eliminate some
memcpy()s in more complex socket types, like tcp-turn and tcp-bsd, but
these optimisations have not been made yet. FIXME comments have been
added.

This includes modifications to the test-send-recv unit test to cover the
new API.

10 years agosocket: Add vectored I/O support for sending on sockets
Philip Withnall [Mon, 20 Jan 2014 07:56:47 +0000 (07:56 +0000)]
socket: Add vectored I/O support for sending on sockets

Replace the send() API with a send_messages() API, which supports
sending multiple messages, each with multiple buffers rather than a
single monolithic buffer.

This doesn’t break API, as the socket API is not exposed outside
libnice. It does introduce a new struct: NiceOutputMessage, which is
analogous to struct mmsghdr and NiceInputMessage.

This includes updates to the test-bsd test to cover the changed API.

The existing nice_socket_send() API has been retained as a thin wrapper
around nice_socket_send_messages(), for convenience only. It’s hoped
that internal usage of this API will decline to the point where it can
be removed.

10 years agoagent: Move gtk-doc comments from agent.c to agent.h for new API
Philip Withnall [Tue, 21 Jan 2014 10:50:19 +0000 (10:50 +0000)]
agent: Move gtk-doc comments from agent.c to agent.h for new API

I completely disagree with this, and believe the C file is a much better
place for them, as then they’re:
 • easier to read while hacking on the functions, and
 • easier to modify once finished hacking on the functions.

I think the argument for putting them in the header files (so that the
documentation is available by the function declarations) is weak, as the
generated gtk-doc manual should be installed on the system alongside the
header files in any case.

10 years agotests: Update test-send-recv to support vectored I/O
Philip Withnall [Thu, 16 Jan 2014 16:11:25 +0000 (16:11 +0000)]
tests: Update test-send-recv to support vectored I/O

10 years agoagent: Add support for vectored I/O for receives
Philip Withnall [Thu, 16 Jan 2014 16:06:30 +0000 (16:06 +0000)]
agent: Add support for vectored I/O for receives

Add two new public functions:
 • nice_agent_recv_messages()
 • nice_agent_recv_messages_nonblocking()
which allow receiving multiple messages in a single call, and support
vectors of buffers to receive the messages into.

The existing nice_agent_recv[_nonblocking]() APIs have been left
untouched.

This tidies up a lot of the message handling code internally, and
eliminates a couple of memcpy()s. There are still a few more memcpy()s
on the critical path, which could be eliminated with further work.

In the reliable agent case, every message is memcpy()ed twice: once into
the pseudo-TCP receive buffer, and once out of it. The copy on input
could be eliminated (in the case of in-order delivery of packets) by
receiving directly into the receive buffer. The copy on output can’t be
eliminated except in the I/O callback case (when
nice_agent_attach_recv() has been used), in which case the callback
could be invoked with a pointer directly into the pseudo-TCP receive
buffer.

In the non-reliable agent case, zero memcpy()s are used.

A couple of the more complex socket implementations (TURN and HTTP) have
slow paths during setup, and partially also during normal use. These
could be optimised further, and FIXME comments have been added.

10 years agotests: Fix a shadowed function name in test-pseudotcp
Philip Withnall [Thu, 16 Jan 2014 16:05:12 +0000 (16:05 +0000)]
tests: Fix a shadowed function name in test-pseudotcp

The write() function already exists; we can’t re-define it as a
callback.

10 years agostun: Add a fast version of stun_message_validate_buffer_length()
Philip Withnall [Thu, 16 Jan 2014 16:02:07 +0000 (16:02 +0000)]
stun: Add a fast version of stun_message_validate_buffer_length()

stun_message_validate_buffer_length() is already fast, but requires the
entire message to be in a single monolithic buffer. For introducing
vectored I/O, this becomes impossible to guarantee.

Rather than rewriting the STUN code to natively support vectors of
buffers (which would be a huge undertaking, and would probably slow
the code down considerably), implement a fast check of whether a message
is likely to be a STUN packet which *does* support vectored I/O. This
can then be used to determine whether to compact the vector of buffers
to a single monolithic one in order to validate the message more
thoroughly.

10 years agosocket: Fix the formal parameter types of nice_turn_socket_parse_recv()
Philip Withnall [Thu, 16 Jan 2014 16:00:42 +0000 (16:00 +0000)]
socket: Fix the formal parameter types of nice_turn_socket_parse_recv()

gsize and guint8 are more obviously for binary data than gint and gchar.

Note the signedness of the return type has changed: the function never
returns negative values.

10 years agosocket: Add vectored I/O support for receiving on sockets
Philip Withnall [Thu, 16 Jan 2014 15:45:12 +0000 (15:45 +0000)]
socket: Add vectored I/O support for receiving on sockets

Replace the recv() API with a recv_messages() API, which supports
receiving multiple messages, each with multiple buffers rather than a
single monolithic buffer.

This doesn’t break API, as the socket API is not exposed outside
libnice. It does introduce a new struct: NiceInputMessage, which is
analogous to struct mmsghdr.

This includes updates to the test-bsd test to cover the changed API.