Tobias Hildebrandt authored 2 years ago and Adrien Béraud committed 2 years ago

* rename EchoCanceller to AudioProcessor
* enable echo cancellation and voice activity detection
* add minimum active duration and trigger time to voice activation

Gitlab: #741
Change-Id: I98662462c17539fca1d042482e97fdb3eff86130

This will facilitate the work of the client in order to order messages
as linearized parents for all commits until the merge base can be
updated.

Change-Id: I95266174d58913fbfb2ca665da18b08475aa9c93

This is useful for the conversation_module if any channel is pending
while shutting down, as the previous fetch status was not updated.
More generally, every connectDevice() should call the associated
callback.

Add a unit test to replicate this scenario.

Change-Id: I72f2975dc15dd4bac3f55c2f899ebb1ae5a7a7f3

As we only need one git notification, we can just refresh the
last sending token to avoid to populate the cache with lot of
notifications that will be unused.

Change-Id: Icb8252c8a4b4068e0bd8475a8cc70d77224858da

Adrien Béraud authored 2 years ago and Sébastien Blin committed 2 years ago

Change-Id: I2cb56b8320a2a0f8fa8cdef0c163bf4f662bdf0d

Adrien Béraud authored 2 years ago and Sébastien Blin committed 2 years ago

Change-Id: I7c2e6b95fad95049565f9298e2690c3ef8b0b5d1

Adrien Béraud authored 2 years ago and Sébastien Blin committed 2 years ago

Change-Id: I1df11887d2c1e9a3e18333e7eb42c297d5ef14f5

Adrien Béraud authored 2 years ago and Sébastien Blin committed 2 years ago

Change-Id: If659bfb7f1140fed5159b193f9e6757ab8f8b103

Change-Id: Ie810524fa960832154be0319d1c4e2383e87eb90

If an account is disabled during the initialization of the ICE,
the callacks must be called to correctly clean conversation_module.
This was not correctly done as we only check the onShutdown() after
that the TLSSocketEndpoint is initialized.
It was a problem when the account was disabled, because in
connectDeviceStartIce() isDestroying_ was true at this point, so
we didn't erase the pending callback.

Change-Id: I21cdd30057d9879b6d6fef189fc43bf916d4c310

Change-Id: I58028cad9b28b694ed271ebd1e9ebad7cd72f9fa

* fixes black bar on the left of conference videos

Change-Id: Ie3725ad144035529288e0a97353a977abffb1133

Change-Id: Ic2f1cf4151e55f5f6a1fc18c857c90143656ad35

Change-Id: Ic8576fbb9633d768cdb6ac5e895e954ee460aa05

Change-Id: I3a1c6b59083426ceb28c8fb2fe4b3c2d7866e491

Change-Id: I434255ea4caf66c630927bdfea572c49e7e1c658
GitLab: #748

This reverts commit 80abd8b7.

Reason for revert: should now work with {fmt} 6

Change-Id: I9e7f287a8fb79e33d343eadbceb84c4290629699

Change-Id: I561e60a14a363bb8a1658eb7f740bd67a85c2d99

Change-Id: Ib8fcedb805085add521d27eaef521b2c7faecea8

Change-Id: I6d7fe3f7593764bbb99570ba02b987943ba35a18

Project name is Jami

Change-Id: I844d2933c49249806b89c2530194e468935ad3cf

Olivier Dion authored 2 years ago and Adrien Béraud committed 2 years ago

Also, extend padding of function name to 24 characters.

Change-Id: I98b39048152fc65bd9c04112393e47e481498170

On a lot of iOS devices, a lot of weird IPv6 candidates seems to
be added from generally unused interfaces (tun0->2, ipsec_0-9, pdp_ip1->4)
Those interfaces seems to be up and running but instead a classic
link-local address (fe80:: like on macOS), some un-assigned ipv6
seems to be generated (2605:8d80:: or  2001:56f::) and seems un-documented.

Except ignoring point to point network for now, I don't see any good
solution. And this problem seems to only appears on iOS (macOS got tun0->2
but with a correct link local address).

Change-Id: I7724150ea2fdd93fbccac83cfa0c0c691452e820

Change-Id: I5f427f924fc34f7f5218de815ba4077e61b45ea5

Change-Id: I9560cb0bd8bfab5a07039e93d4db8827906982ef

Change-Id: I3638c05a8bccad9326fb74dff0c0cc4acbaf1e65

Change-Id: Ic7a89119c5db5023e91529402377bdc591afdf3a

Unloading a plugin without erasing the exit function might cause Jami
to crash when closing.

Change-Id: I63d5c481036c4718507396f76df3f132e543c48c

bad CPPUNIT_ASSERT and timing issue

Change-Id: I16267d00325ebc1875cb4a92f73eaefb50614f99

Change-Id: I50593324cef00af15f5a77d73e2b60eae7d8f63b

Change-Id: I2d8d29077e15967f3869ffc341eda07852aa8fe8

Change-Id: Ib2ba3429d0bc7b63a4608d34b321e4ae2759b48d

Change-Id: I3ee8a5051e8aace1737d40364bd5defae044636c

Andreas Traczyk authored 2 years ago and Adrien Béraud committed 2 years ago

The natpmp CMake build has no installation specified, so currently
we use the root source directory which contains the header files as
an include path in the daemon build. This may cause issues as case
insensative header includes for the standard library <version> will
incorrectly load natpmp's 'VERSION' file.

This patch copies the project's headers to an include directory
post-build, and adjusts the daemon's include path.

Change-Id: I4b318fb0309c72b94f4ea4e757035780ee2a8a63

Change-Id: I0fe99a8d44f8bc5271bea0b7b6a0e377f051f0ab

Change-Id: I55664ee12cb3b7ca372ec9efcb654cb9f68242fb

Sébastien Blin authored 2 years ago and Adrien Béraud committed 2 years ago

Because an incorrect TURN can cause long delay for connections,
we cache the resolved address in a file which is refresh at the
next status changes.
However if the resolution fails it can take a long time (moreover
on servers) to retry the resolution if it fails.
Here, if turn is enabled but not resolved, we do the resolution
on incoming requests to avoid to wait too long before re-using
the TURN.

Change-Id: I39a8e83b43c7342a6c296e57316bd85497360011

Kateryna Kostiuk authored 2 years ago and Adrien Béraud committed 2 years ago

Change-Id: I97867faadfcdbe3896464c6ca05c5857f042f1e2

Sébastien Blin authored 2 years ago and Adrien Béraud committed 2 years ago

Change-Id: Ib3bbc21e513e0dd22ce28d099356f3286b9dffd1

Sébastien Blin authored 2 years ago and Adrien Béraud committed 2 years ago

Now that it's been months since all clients supports swarm, we
can assume that everybody is using a version compatible with
swarm now.
So, downgrading conversations to a non swarm if a DHT message
comes is not necessary and it's just unwanted code.

Change-Id: I83b83d592ea43219cbdecb31fa8b5a71b897e487
GitLab: #312