It was possible to set the encoder to scale the video to the same size
as the output, resulting in unnecessary passes through swscale and
skipping hardware encoders. This was a common user error in Advanced
Output mode where they would tick the rescale box but leave it at the
same resolution as their main output.
FFmpeg wipes priv_data if the encoder fails to start, so we need some
way of knowing if the GPU index was non-zero. This is a bit ugly but the
best way to do it for now.
This writes the result of nvEncGetLastErrorString to the last encoder
error instead of just the log file. This makes it user-visible, which
should help users diagnose and solve the problem.
Also strips off leading colons from NVENC error strings as these can
look odd given our format string. An example of such an error:
::NV_ENC_TWO_PASS_FULL_RESOLUTION is only supported multipass encoding
on this architecture.
When adding to the queue, this triggers rowCountChanged which then
calls canClearFinished, reading undefined memory for state if this is
not set. The correct state is set immediately after by checkInputPath.
Detected by UBSan.
The gain should be positive for an upward compressor. Initially, the
gain would be zeroe'd below -60 dB to ensure noise is not amplified by
the upward compressor. This created a discontinuity at -60 dB since
just above -60 dB, the audio is boosted by +20 dB (at default
settings). This was fixed in commit 50db097 which decreased smoothly
the gain so that it's 0 dB at -60 dB. However that commit forgot to
limit the gain decrease which was negative below -60 dB.
This is fixed by the current commit.
Additionally initialization allowed -inf gain. We hard limit to positive
gains only as they should be for an upward compressor.
The bugfix was found by R1ch and put in form by pkv.
Co-authored-by: pkv <pkv@obsproject.com>
Signed-off-by: pkv <pkv@obsproject.com>
- Don't repeatedly query version at runtime in case a user installs the
SDK while OBS is running
- Restore default DLL search directory
- Don't show outdated SDK message if the SDK is not found
- Protect minimum version macro with brackets
Now uses GetIfEntry2 which supports 64-bit values for reporting speed, so
10+ gbps adapters are now reported correctly in the log. Also added an
additional log line if the interface error counters are non-zero to possibly
help identify physical faults. Finally the transmit and receive speeds are
logged independently so that asynchronous mediums such as Wi-Fi that might
have good RX but poor TX can be better diagnosed.
If the output is paused the average bitrate will be zero, resulting in
infinite time until disk is full and int overflows. Similarly, if no data
has been collected yet, the result will be NaN and undefined behavior.
If the display wasn't found, the timer was not reset, causing the code
to execute on every single tick, stalling the graphics thread and using
excessive CPU.
Another call to getenv, as well as a call to the POSIX functions
setenv(), unsetenv(), and putenv() may invalidate the pointer
returned by a previous call or modify the string obtained from a
previous call.
https://en.cppreference.com/w/c/program/getenv
With the addition of server-side release rate control, having the server
know if the update check was manually initiated can allow it to deliver
the update to the user even if they would normally not be eligible.
Windows only.
With HEVC and H264 settings being near-identical, it was impossible to
figure out which codec was being used by context alone. This applies to
both ffmpeg output and jim-nvenc.
Fixes#6976.
The toggleControlLayoutAction (change between horizontal and vertical
display mode) deletes and re-recreates the volume controls, meaning we
were writing to freed memory. This is a kind of hacky fix but it's the
only action we need to be concerned with for now.
Since this was included in things like the inject helper and game
capture DLLs, it added an additional export to those programs. Instead,
simply share the source files directly in the projects that need it.
Co-authored-by: jpark37 <jpark37@users.noreply.github.com>
If the test process freezes for whatever reason, OBS would block
indefinitely on reading the pipe. This commit adds a 2.5 second timeout
on the AMF test process after which it will self-terminate.
In the event that simply loading the AMF DLL causes initialization code
to run that might crash, don't load the DLL before the AMF test process
has succeeded. Instead, we load the DLL as data to see if it exists,
then run the AMF test, then load the DLL for real.
malloc(0) is implementation defined and almost always a mistake. I
believe we should be more like malloc and treat bmalloc(0) as an error
instead of tying to handle it. For now we can log an error in case 3rd
party plugins are broken, in the future I would suggest this becomes a
crash-worthy error.
Hopefully nothing is actually using this in the first place. As a
library libobs has alignment requirements so we should probably not
allow an application to replace our aligned mallocs with something else.
This also allows more aggressive optimizations inside libobs as the call
can be completely inlined into a libc malloc.
A race condition can occur in obs-outputs where the send_thread is in
the process of (re)connecting but the encoder was shut down in the
meantime. This causes the expected header data to be garbage, resulting
in a crash.
The silent reconnect (GOAWAY) is supposed to be used for switching the
connection to a different server. As such, cleanly shutting down the
RTMP connection can destroy state that needs to be preserved in order
for the GOAWAY-enabled server to properly resume the same stream on
reconnect.
This commit closes the TLS/TCP connection before calling RTMP_Close,
causing librtmp to skip the FCUnpublish and deleteStream messages.
This adds a circular buffer to ffmpeg-mux when writing to a file.
Output from ffmpeg is buffered so that slow disk I/O does not block
ffmpeg writes, as this causes the pipe to become full and OBS stops
sending frames with a misleading "Encoding overloaded!" warning. The
buffer may grow to 256 MB depending on the rate of data coming in and
out, if the buffer is full OBS will start waiting in ffmpeg writes.
A separate I/O thread is responsible for processing the contents of
the buffer and writing them to the output file. It tries to process 1 MB
at a time to minimize small I/O.
Complicating things considerably, some formats in ffmpeg require seeking
on the output, so we can't just treat everything as a stream of bytes.
To handle this, we record offsets of each write and try to buffer as
many contiguous writes as possible. This unfortunately makes the code
quite complicated, but hopefully well commented.
This option regularly confuses users, who accidentally select it and
then ask why they have hundreds of 2-4 second video clips in their
output folder. HLS output format to file is a pretty niche use case,
so let's keep it in advanced mode only.
In the years since this code was added, ALPN is now widely supported,
and NPN is being removed entirely in the latest version of nginx. It's
time to start using ALPN!
_dx9_simple_free already frees this pointer. I don't think this code
path is currently reachable, but it should probably still be fixed.
Detected by Coverity Scan.
Currently if a module fails its load callback, it remains loaded and OBS
continues to call additional exports such as obs_module_post_load. This
goes against the documented behavior, which states that a module that
fails its load callback is unloaded.
This commit releases locale resources and frees the module's
information, preventing further callbacks from libobs. The module itself
(the DLL) is not yet unloaded from memory as os_dlclose is commented out
for causing unspecified issues - this should be revisited in the future.
For some reason, SetWindowDisplayAffinity can make windows visible even
when it sets the same value for affinity that GetWindowDisplayAffinity
reports. Possibly an API bug as this is probably not a widely used API
yet?
Fixes a weird window with no size appearing when browsing for files on
Windows.
OBS_PROPERTY_GROUP creates neither a widget nor a label, causing
AddProperty to create a default one. Without a widget, it does not
get attached to the layout, resulting in a memory leak.
This commit also simplifies a bit of the code to avoid repeatedly
testing the same condition.
This uses the SetWindowDisplayAffinity API to hide windows from capture
applications (including OBS). This is not perfect - internal windows
such as context menus, combo box dropdowns, etc will still be displayed.
Even with these limitations, it should help people with single monitors
capture content with less interference from the OBS window.
This implementation is for Windows only but the code is generic enough
that adding other platforms should be straightforward.
These were operated on by atomic functions but were not marked as
volatile or loaded with os_atomic_load_long, potentially introducing
subtle race conditions. Detected by ThreadSanitizer.
Richard StanwayandJim
Richard StanwayandJim