Age | Commit message (Collapse) | Author |
|
[Nexus] Bump version 20.1
|
|
|
|
fuzzard/backport_CrystalP_quickfix-window-pos-size
[BP][Windows] shrink window to fit the screen and avoid a crash
|
|
[BP] ExifParser: Fix several out of bounds accesses while parsing exif information
|
|
[BP][Windows][dxva] fix processor height
|
|
[BP] VideoPlayerCodec: Stop dividing by zero
|
|
[BP] tools/depends: make the hash file an order only prerequisite
|
|
[BP] [jenkins][Apple] Some fixes for apple buildsteps
|
|
|
|
|
|
Partially via upstream commit:
https://android.googlesource.com/platform/external/jhead/+/2a4c12f5e5808e309b9ba04fe8b1539debf466d1
|
|
When parsing exif information the number of entries could be tainted. Make
sure to not run behind the exif data by properly checking out of bounds.
Upstream? fix: https://android.googlesource.com/platform/external/jhead/+/34a2564d3268a5ca1472c5076675782fbaf724d6
|
|
this fixes an issue when rebuilding the dependencies that if a target is already built it doesn't get rebuilt
ref: https://www.gnu.org/software/make/manual/html_node/Prerequisite-Types.html
|
|
With the new Build system in xcode, it now has a changed behaviour with a new setting
"Based on dependency analysis". For us, essentially our dependency
targets (eg fmt, spdlog, etc) are added as Run scripts by the cmake xcode generator.
This means that any dep added as a run script will always be exectuted.
Currently with master buildsteps make-xbmc builds the build target, and then the package step
building the deb target, dependencies are built twice regardless of state. We will have to
look into the cmake generation side of things but the easiest solution for now is to just
generate the deb target as part of the make-xbmc buildstep, and then just handle upload
of artifact in the package buildstep.
Shaves off ~3 minutes of build time on jenkins.
|
|
|
|
When parsing a file failed so that needed fields, in our case frameSize
are not available a guess work like with m_channels or m_samplerate does
not help. Therefore fail early.
|
|
[upnp] fix regressions for folder definitions
|
|
[Windows] Fix SMB read chunk size when is not used FileCache
|
|
[Backport] Sync controller add-ons with controller-topology-project
|
|
ActiveAESink: workaround for TrueHD and DD+ broken after pause or seek in AudioTrack RAW (Android only)
|
|
64K is the dafault in CFileStreamBuffer and can be used when FileCache is
not enabled or for file types which is not used FileCache (Blu-Ray folders)
Otherwise defaults to only 6144 bytes causing performance issues over SMB
|
|
AudioTrack RAW (Android only)
|
|
Revert "AESinkAudioTrack: Use most simple pause() logic"
|
|
[Backport] Fixes and improvements from "show actual controllers" PR
|
|
[Backport] Controller Dialog: Fix "Ignore Input" buttons not being saved
|
|
|
|
[backport] NFSv3 and NFSv4 fixes and improvements
|
|
[Addons] Having no repos is no error
|
|
Be able to install addons from ZIP without having valid repos because
of no internet connection.
|
|
[Windows][Installer] fix the missing .html files of Chorus2 with proper exclusion of .h
|
|
NFSv4 not has maximun chunk size limitation at server side (used
TCP vs UDP in NFSv3, etc.) then 'nfs_get_readmax' returns zero or not
implemented.
Clients must set the chunk size they want to use.
|
|
Retrying with other error codes and force deinit connection may cause
crash, specifically when a file does not exist.
Anyway, NFS4ERR_EXPIRED should not occur now that contexts expiration and keep alives are
handled correctly.
|
|
In case an NFSv4 server are configured with low "lese_time" i.e: 30s
the session will expire before receiving the first keep alive and we
will realize it by nfs_read errors here
|
|
|
|
That means the nfsv4 context has expired and needs to be recreated
Signed-off-by: Lukas Rusak <lorusak@gmail.com>
|
|
exclusion of .h
Fix displayed path of the generated installer.
|
|
This reverts commit 3f69f160e6bdcfae51e0c8448ad0196907615920.
|
|
AESinkAudioTrack: Revert sanity checking
|
|
This technically reverts:
0917ef9d7a00b04eb69961dbde6e93c2b15248f4
and
1e3715ce15259fe630a3774044b833f64579d8e3
When I check forum posts I see that on this platform several
devices really work awefully bad. They open large buffers and
start running very late ... we will hit more and more of those
devices, therefore I want to revert this for our stable
release and continue testing and finding such devices in master.
|
|
[backport][Skin][Estuary] Fix player settings adjustments if autoclose video osd is active
|
|
[video] Fix watched status not taken into account when queueing items.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Before:
PERIPHERALS::CPeripherals::LoadMappings - loaded node ""
After:
PERIPHERALS::CPeripherals::LoadMappings - loaded node "joystick"
|