- 06 Jun, 2017 2 commits
-
-
Olivier SOLDANO authored
higten the colors, and reduce the size of the file paths to fit in the fixed size window of win32 debug Change-Id: I9b0d167b5f6b8ebdc74937129c31c7a828320bcf Reviewed-by:
Andreas Traczyk <andreas.traczyk@savoirfairelinux.com>
-
Adrien Béraud authored
Use the account configuration to store the registered name, allowing to know the account registered name for disabled accounts, before account registration happens, or if the name server is not reachable and the name cache was cleared. This allows to show the account registered name instead of the RingID more consistently in clients. The field is not authoritative as information from the name server always take precedence. Change-Id: Ief8725da57d609ce9bbaeda6b2d564638ea3b191
-
- 05 Jun, 2017 1 commit
-
-
Philippe Gorley authored
Adds VideoToolbox and VDA hardware accelerations. VideoToolbox supports H.264, H.263 and MPEG4, while VDA only supports H.264. VDA is implemented in case libav is used instead of FFmpeg, as only the latter implements VideoToolbox. This being said, Ring will prefer VideoToolbox. VideoToolbox is OSX 10.8+ and iOS 8+. VDA is OSX 10.6.3+. Both have their respective configure switches. Change-Id: I588fcbb92809a9d6a56bb9b6a7ac3a59874c0186 Tuleap: #1090 Reviewed-by:
Anthony Léonard <anthony.leonard@savoirfairelinux.com>
-
- 02 Jun, 2017 8 commits
-
-
Philippe Gorley authored
With the bumps to FFmpeg 3.3.1 and libav 12.1, these code paths will never be executed. Change-Id: Ia8d7602a91896620dd6b3848dd6ed63d79fcc809 Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
Philippe Gorley authored
This only activates hardware decoding if the setting wasn't found in dring.yml or if there was an error while reading the file. Change-Id: Ib256e327a4136a72df11873e9b0c662e0e00afce Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
Philippe Gorley authored
If Ring was compiled with acceleration, but is disabled in dring.yml, the call would crash. This checks if the accel_ object isn't null before checking if it has failed. Change-Id: I098f573bd17271824fcfbdb60d783af085ef19be Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
Adrien Béraud authored
Use latest libvpx (was already the case on Android), to profit from latest optimizations. This should mostly profit to low-end Intel CPUs, according to libvpx changelog: Devices based on Celeron processors have seen substantial decoding improvements (...) While Celeron benefited most, Core-i5 also improved 5% and 10%. https://github.com/webmproject/libvpx/blob/master/CHANGELOG Change-Id: I471408c2b09d580fd564b3a968eb4e0f22599541 Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
Philippe Gorley authored
Bump minimum required ffmpeg version in contrib to avoid using outdated system version. Change-Id: I50aa8ecea8be355a4fb60eaf35ca3f6224f2a32e Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
Adrien Béraud authored
Add patch to fix restbed DNS resolution error, that was ignored, causing indefinite hangup. Change-Id: I40fff6a8662e8d204e0d4b813c4a94adadb99b65 Reviewed-by:
Anthony Léonard <anthony.leonard@savoirfairelinux.com>
-
Philippe Gorley authored
Fixes certain cases where the fallback to software decoding was not done. Change-Id: Ie0276a79f3d7434795876a7f417e7b0c6a0b8851 Reviewed-by : Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
Philippe Gorley authored
Bumps up libav to match the FFmpeg bump. Includes many bug fixes. Full list of changes here: https://git.libav.org/?p=libav.git;a=blob;f=Changelog;hb=refs/tags/v12.1 Change-Id: Ieca119af8d56eed45184378080c710285fb60b1c Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
- 31 May, 2017 2 commits
-
-
Anthony Léonard authored
The count argument passed to std::fill_n was in bytes instead of element's count (which are of Float32 type) so zeroes were written outside the scope of the audio buffer. This was causing crashes not on this instruction but anywhere in the application when a malloc call was checking for memory sanity (which isn't done on every allocation). Change-Id: Iec6fa9d3b00b828c6bf798c77419b8bad970d6e1 Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
Anthony Léonard authored
It was previously set to 1s which is too slow for other components such as OpenDHT which kept connecting/deconnecting. Moreover, this change was already made in the GUI client but the CLI one was forgotten at that time. Change-Id: Ifed007b90e9c2f9ccc01e11d90bc2d619bebd40c
-
- 30 May, 2017 4 commits
-
-
Guillaume Roguez authored
Fix Manager::addParticipant() to always accept (return true) to add a call already in given conference (no-op). Without this patch, doing so causes a daemon crash. Change-Id: I7b03a1349e7cac148ba3991b8e0bc2a9f1521c85 Reviewed-by:
Olivier Soldano <olivier.soldano@savoirfairelinux.com>
-
Guillaume Roguez authored
Add some inclusion to <functional> Std C++ header to support build on GCC 7.x. Change-Id: I82ae10a300807cb6b3cab172bf5d7399264cd08e Reviewed-by:
Nicolas Jäger <nicolas.jager@savoirfairelinux.com>
-
Guillaume Roguez authored
Fix regression found in commit [33447cc6: certstore: re-factor peer certificate authorization routine] An unknown dht call is not accepted even if the configuration flags "dhtPublicInCall" is true. Now, accept the gnutls certificate error "INVALID + SIGNER_NOT_FOUND" when this flags is true. Change-Id: I90575ab82296d3d29140a7394ae178c196aa98c4 Reviewed-by:
Andreas Traczyk <andreas.traczyk@savoirfairelinux.com> Reviewed-by:
Adrien Béraud <adrien.beraud@savoirfairelinux.com>
-
Adrien Béraud authored
Currently the dring API provides no way to simply remove a contact without banning it. This adds a ban flag to removeContact. If set to true, the behavior stays the same and the contact will be banned. If set to false, the contact will be removed from the contact list. After the call, it won't be neither added or banned. In both cases the ContactRemoved signal is emitted if the contact status changes. Change-Id: I95510f6db767b7b0e7fe15e8405b39b0bc3dd594 Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
- 29 May, 2017 5 commits
-
-
Adrien Béraud authored
* update ffmpeg to latest stable 3.3.1, update patches and remove deprecated patches and flags, * enable ffmpeg ASM optimizations for all platforms but Android x86 (see https://trac.ffmpeg.org/ticket/4928 ) * enable Android hardware accelerated decoding (MediaCodec), for now only with VP8, MPEG4 Change-Id: Ifac82ec7b9b71bd92a65a3575ac63e7ed80b510b
-
Adrien Béraud authored
If the certificate status can't be set it means the contact is already allowed (or banned) and it's not an error preventing Ring to work properly. Still print a debug message to annoy client developers so they stop allowing everybody on startup. Change-Id: I334b2d1f0140c7019f07dd11f8ce64e61510f038 Reviewed-by:
Nicolas Jäger <nicolas.jager@savoirfairelinux.com>
-
Adrien Béraud authored
Version of x264 currently used in contrib is from 2013, doesn't include any ARM64 optimization, and misses many ARMv7, x86, x86_64 optimizations introduced since 2013. This patch updates x264 to the latest commit in the official x264 "stable" branch, except on iOS which doesn't build. Change-Id: Ifedbe80b0c4d7c628adc701018603cf01ed12dfa Reviewed-by:
Nicolas Jäger <nicolas.jager@savoirfairelinux.com>
-
Adrien Béraud authored
Change-Id: I5a83272244891166103538aee07507476628543a Reviewed-by:
Nicolas Jäger <nicolas.jager@savoirfairelinux.com>
-
Adrien Béraud authored
* cleanup CertStore to remove unused features * ensure that at least one certificate in the chain is allowed if public calls are disabled and that no certificate in the chain is banned, additionally to the chain match. This prevent allowing any certificate signed by the CA of a contact, while still supporting authorization through a CA (with setCertificateStatus). * use dht::crypto::TrustList when relevant (requires OpenDHT bump) * rebuild trust list when appropriate as GnuTLS' trust list doesn't allow "unbanning". This fix issues when banning then allowing a certificate during the same session Change-Id: I8ea28f3a673eebf2174e8fab3a413c20630b28ca Reviewed-by:
Nicolas Jäger <nicolas.jager@savoirfairelinux.com>
-
- 26 May, 2017 2 commits
-
-
Alexandre Lision authored
Switch to ffmpeg for video processing on OSX. FFmpeg provides hardware acceleration for video decoding Change-Id: I6f4a318e79871c8c9e798d213f1c9deeb80d7398 Reviewed-by:
Philippe Gorley <philippe.gorley@savoirfairelinux.com>
-
Adrien Béraud authored
SipIceTransport was a first implementation of a SIP transport working over an ICE transport. It was insecure and is now unused and unmaintained (last actual change is from February 2015), and deprecated by SipsIceTransport. Remove it from the code base to speed up compilation and avoid accidental usage by confusion with SipsIceTransport. Also fix minor warnings in SipsIceTransport (std::move of an rvalue). Change-Id: I7a8ce2c0570d92bfe8bd191f2f6c8c6f2b191980 Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
- 24 May, 2017 2 commits
-
-
Guillaume Roguez authored
Remove a race-condition on packet reorder std::map during its flush. This caused application crashes. Change-Id: I01ebf56f8cca141585f11c1090262d3b44f34544 Reviewed-by:
Andreas Traczyk <andreas.traczyk@savoirfairelinux.com>
-
Anthony Léonard authored
This CoreAudio function was called every time the read callback was triggered which is inefficient and causing deadlock if AudioOutputUnitStop is called at the same time. This problem was occuring often during call stop. We now keep necessary properties in the CoreLayer object during initialization of the AudioUnit and use them instead of querying the API. Change-Id: I38e685e3dcbf00d95253f27f99ae25a708f31b19 Reviewed-by:
Kateryna Kostiuk <kateryna.kostiuk@savoirfairelinux.com>
-
- 19 May, 2017 2 commits
-
-
Adrien Béraud authored
Avoid to build unused argon2 binary and to perform unnecessary clean before building, which can cause build failure when building in parallel. Change-Id: I5fad09d8c8c9a22678e6a673a288efc40cc165b9
-
atraczyk authored
This patch adds libopus to the UWP contrib and configures the ffmpeg build script to use it. Change-Id: I20e3b547eb7228be0a293a3514eb267afac1c068 Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
- 18 May, 2017 1 commit
-
-
Philippe Gorley authored
Renames the HardwareAccel class's check method for two reasons: - Better describe the purpose of the method - Mac sometimes defines a "check" macro Sources: http://stackoverflow.com/questions/16766843/using-boost-from-xcode-4-6-1-what-are-the-proper-build-settings https://github.com/USCiLab/cereal/issues/104 http://stackoverflow.com/questions/31665095/xcode-error-compiling-c-expected-member-name-or-after-declaration-specifie Change-Id: I5f67fe1a9d2b262942eb56f05d99073587208d9e Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
- 17 May, 2017 2 commits
-
-
Philippe Gorley authored
Adds the VDPAU acceleration for GNU/Linux systems and enable it by devault. Can be disabled using ./configure --disable-vdpau Support for H.264, H.263 and MPEG4, as long as the hardware supports them. Requires libvdpau and ffmpeg 3.3 (contrib bump included by this patch) [guillaume roguez: re-word ci msg to be more explicit] Change-Id: I7479f4d7e3d51caf702be8c85284a4fca01605b3 Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
Guillaume Roguez authored
give a chance to rx queue to immediatly push packets to application when packet is received. This reduces a bit the latency of packet handling. Change-Id: I7c09d91ab505e9e65ebbb990f1d155b3d639e024
-
- 16 May, 2017 2 commits
-
-
atraczyk authored
The argon2 static projects are named "static" and output with a .lib extension, but are not configured correctly as static libraries. This patch corrects that configuration setting for the project used by opendht/ring-daemon. Change-Id: I8e8d4f30bcf5e60c5f22c42f4b68b7f1f6e42d30
-
Adrien Béraud authored
Despite setting LIB_SH="", a (broken) symbolic is still created, and sometimes attempted to be linked with. Make sure to remove any potential shared library. Change-Id: Iba8242dd952b536b261cfd7d30c8ce72cf5af3d0
-
- 15 May, 2017 7 commits
-
-
Philippe Gorley authored
There was a missing override of shared object target in the argon2 build rules. Change-Id: I6e47ff55fd684357cc0ac674cb27b9b39068f5a0 Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
Guillaume Roguez authored
setCallMediaLocal is used to setup transport related options. We also don't use anymore the local IP argument. So this patch remove the argument, move it into ctor and make setCallMediaLocal private (all calls removed). This has the good side effect to remove the non-permited ICE API call causing an error in logs. Change-Id: I69220c253164777e8e88b6a4deef451325458c2c Reviewed-by:
Olivier Soldano <olivier.soldano@savoirfairelinux.com>
-
Adrien Béraud authored
adapt daemon to support Android NDK Unified Headers Change-Id: I36f7225c2d235b4301a33ad59a18d4a7c4749e25
-
Philippe Gorley authored
This is because argon2's shared object target always makes the executable, causing some undefined references to 'main'. This nulls that target, to only build in static. Change-Id: Ic6fbd6a99c5df24866bd07bbbf11803e19d39af1
-
Adrien Béraud authored
* bump fixes a crash with null certificates * use a release tag as a best practice to help packagers Change-Id: Ief90ea9d833419eb10a9250d7e5d956f95a2a5bc Reviewed-by:
Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>
-
Houmin authored
Change-Id: I3621c804bff191ff877d8681864a40ee56de8444 Reviewed-by:
Stepan Salenikovich <stepan.salenikovich@savoirfairelinux.com>
-
Guillaume Roguez authored
There are various issues in the PMTUD code: - OOO handler wasn't applied to the first packet due to unseen code duplication in PMTU code. - first packet sequence has to be known in case of OOO on it - bug in losts detection. - decrease the lost threshold time. - temporary packet allocation is not efficient. - code duplication and functional flow not well designed. - comments needed This patch fixes all of that. Change-Id: I93ec71e22f6cb7a66ad9ab0f927d31044966f1e3 Reviewed-by:
Anthony Léonard <anthony.leonard@savoirfairelinux.com>
-