Implements exponential backoff for consecutive reconnects, which is
useful to prevent too many connections from trying to reconnect back to
a service at once over a short period of time in the case of potential
service downtime. Exponential backoff causes each subsequent reconnect
attempt to double its timeout duration.
This optionally allows the front-end to know what the current timeout
value in seconds is set to for the reconnection without having to call
an extra API function to find that out.
Implement the log_processor_info function on bsd and add ifdefs to only
build the implementation specific to the platform.
Also add an ifdef around the call to that function to make sure it will
only be called on platforms where it is actually implemented.
Split the function logging the processor information on nix into two
parts. The part logging the number of logical cores is portable and
works on all systems that support POSIX.1 while the other part is
specific to linux.
Add the relevant header file needed on FreeBSD and utilize yet another
ifdef to call pthread_set_name_np as the function name differs from
those on the other platforms.
Add definition on FreeBSD to enable getline in stdio, as it is not
(yet ?) available by default. According to the manpage getline was a
GNU extension but was standardized in POSIX.1-2008.
API Changed (in struct obs_encoder_info):
----------------------------------------
bool (*get_audio_info)(void *data, struct audio_convert_info *info);
bool (*get_video_info)(void *data, struct video_scale_info *info);
To:
----------------------------------------
void (*get_audio_info)(void *data, struct audio_convert_info *info);
void (*get_video_info)(void *data, struct video_scale_info *info);
The encoder video/audio information callbacks no longer need to manually
query the libobs video/audio information, that information is now passed
via the parameter, which the callbacks can modify.
The refactor that reduces boilerplate in the encoder video/audio
information callbacks also removes the need for their return values, so
change the return types to void.
I realized that the get_video_info and get_audio_info encoder callbacks
always have to manually query the libobs audio/video information.
This fixes that problem by passing the libobs video/audio information in
the structures passed to those callbacks so they don't have to query it
each time, reducing needless boilerplate code for encoders.
Allows the ability to hint at encoders what format should be used.
This is particularly useful if libobs is currently operating in planar
4:4:4, but you want to force an encoder used for streaming to convert to
NV12 to prevent streaming issues.
Fixes a crash that could happen if any of the mutexes are used in the
create callback, or before the obs_source_init function is called.
I'm not sure how this function order slipped because it seems fairly
obvious that these mutexes should be created before the create callback.
Had this crash happen to me when creating a WASAPI output source, the
create callback of the WASAPI source creates a thread which outputs
audio, and that thread managed to call obs_source_output_audio before
the obs_source_init function was called, which in turn caused it to try
to use a null mutex.
When caching a new frame, keep a reference to the frame while copying to
ensure that the frame is not potentially destroyed for whatever reason
while that data is being copied.
The obs_source::async_reset_texture variable can cause a data race
between threads to occur because it could be set to true in one thread
then changed back to false in another thread. This could cause the
async texture to not update its size when it's supposed to, which can
cause a crash or corruption when copying data from a frame of a
differing size.
The solution to this is to:
- Delete the async_reset_texture variable, and make the
set_async_texture_size function change the texture size if the
async_width, async_height, or async_format variables differ from the
frame's width/height/format. Those variables are then only ever set
in the libobs graphics thread.
- Make the cache_video function use separate variables from other
functions to detect a change in size (due to the fact that the texture
size should only be resized in the libobs graphics thread). These
variables are async_cache_width, async_cache_height, and
async_cache_format, which are only be set in the thread that calls
obs_source_output_video.
How to replicate the data race:
- On OSX, use window capture on a textedit window, then continually
resize the textedit window.
The bilinear lowres scale effect was using 'output' for a variable,
which is apparently a reserved keyword in GLSL on macs. This slipped
by me due to the fact that this didn't occur with OpenGL on my windows
machine.
The minimum and maximum color range values were not being set by the
video_format_get_parameters function when full range was in use; I
assume it's because the expected min/max values of full range is
{0.0, 0.0, 0.0} and {1.0, 1.0, 1.0}, so I'm just making it so that it
sets those values if using full range.
Way to replicate the issue (windows):
1.) Create a win-dshow device capture source
2.) Select a device and set it to a YUV color format to enable YUV color
conversion
3.) Select "Full Range" in the color range property.
4.) Restart OBS, the device will then start up, but will display green
due to the fact that the min/max range values were never set, and are
left at their default value of 0.
Due to a bad 'if' expression, when a filter that is not last in the
chain is disabled or being bypassed, it ends up still calling the
filter's video processing function unintentionally.
This fix makes sure that it only calls the appropriate render functions
if the next filter target is the source, otherwise it will just call
obs_source_video_render to process the next filter in the chain.
How to replicate the bug:
1. Create two crop filters on the same source
2. Give each crop filter a different distinct value
3. Disable both crop filters
4. The image would still be cropped
The normal scaling methods cannot sample enough pixels to create an
accurate output image when the output size is under half the base size,
so use the bilinear low resolution scaling effect in that case instead
to ensure a more accurate low resolution image.
If you don't need to see what's displayed, then this is particularly
useful for two reasons:
1. It reduces the number of draw/present calls
2. It can prevent issues with certain hardware setups where rendering on
a monitor hooked up to a separate card can experience slowdowns
This reverts commit 99c674e41f.
These commits were originally added to allow multiple user interfaces to
use the same plugins, but I soon realized that multiple user interfaces
can use multiple libobs versions, so each user interface should have its
own set of plugins to manage. Some user interfaces may not wish to use
certain plugins anyway, so this fixes that issue as well.
This reverts commit 92d800cc18, reversing
changes made to 35a4acede0.
These commits were originally added to allow multiple user interfaces to
use the same plugins, but I soon realized that multiple user interfaces
can use multiple libobs versions, so each user interface should have its
own set of plugins to manage. Some user interfaces may not wish to use
certain plugins anyway, so this fixes that issue as well.
Instead of manually setting the blend state to the desired values, use
gs_reset_blend_state to ensure we have the default blend state (which
for color is a typical srcalpha/invsrcalpha alpha blending operation,
then the alpha channels are added together).
This fixes an issue where filtered scenes would look strange due to the
fact that alpha was not being blended properly.
This allows the ability to separate the blend states of color and alpha.
The default blend state has also changed so that alpha is always added
together to ensure that the destination image always gets an alpha value
that is actually usable after the operation (for render targets).
Old default state:
color source: GS_BLEND_SRCALPHA, color dest: GS_BLEND_INVSRCALPHA
alpha source: GS_BLEND_SRCALPHA, alpha dest: GS_BLEND_INVSRCALPHA
New default state:
color source: GS_BLEND_SRCALPHA, color dest: GS_BLEND_INVSRCALPHA
alpha source: GS_BLEND_ONE, alpha dest: GS_BLEND_ONE
This fixes an issue where cache frames would not free at all after
having been allocated with no upper limit on the cached frame size. If
cached frames go unused for a specific period of time, they are
deallocated and removed from the cache.
This is preferable to having an upper cache limit due to the potential
for async delay filtering.
Under certain circumstances the cache could be prone to growing too
large unintentionally. Setting a hard maximum limit should prevent
memory from growing if we suddenly get a lot of frames.
Async frames are only swapping when rendering, or when not visible.
This is a flawed design due to the fact that there are certain
circumstances where the source is neither visible nor currently
rendering.
This is what caused a memory leak when scene items were marked as
invisible, because if a source has an async child source and decides not
to render that source for whatever reason, the child source would not
process the async frames at all, and the cache would just grow.
To fix this, simply moving the async frame cycle to tick fixes the issue
due to the fact that tick is always called regardless of circumstance.
Filtering the video before it's output to the texture means that it
happens after all the processing on the timestamps and such of the video
data. This way, the video filter does not have to worry about what's
currently buffered, and it won't affect timing.
When OBS is shutting down, if for some reason the filter is destroyed
before the parent source is destroyed, it would try to remove itself
from the source, but it would decrement the reference and try to destroy
itself again while already in the process of destroying itself.
So, the solution was simply to make sure that if it's removing itself
from the source that it doesn't decrement its own reference.
The "last" filter that's rendered is technically at filter index 0, so
enumeration needs to be from the last index in the list to the first
index in the list.
When rendering a filter to a texture, the target is empty and unused, so
there's no reason for blending to be on when rendering the filter to a
render target.
obs_source_process_filter tried to do everything in a single function,
but the problem is that effect parameters would not properly be
accounted for due to the way it internally draws, therefore it was
necessary to split the functions in to two, you first call
obs_source_process_filter_begin, then you set your effect parameters,
then you finally call obs_source_process_filter_end. This ensures that
when the filter is drawn, that the effect parameters are set.
When the filter chain finally reaches the source and the last filter in
the chain is set to not render directly (meaning it has to render to
texture), it would not render the source with any effect due to the fact
that it expects a filter to be present.
Adds a source callback function that is used when a filter is removed
from a source. This ensures that any data that might be associated with
that source can be cleaned up if necessary.
There are a few more conditions which need to be checked to ensure
whether we can actually bypass or not; in this particular case we are
using the YUV texture shaders, plus the image can also be flipped, so we
can't really use the bypass optimization in those situations.
These functions are primarily for use with filters, filters need to be
able to get the width/height of a target source without it necessarily
getting the post-filtered dimensions.
Previously I had it set so that it would set the async_width and
async_height variables to 0,0 if a null frame occurs, but the problem
with this is that it was inadvertently cause the frame cache to clear
when it starts receiving textures again because it detects it as a size
change.
So instead of changing async_width and async_height to 0 when a null
frame occurs, change it so that if the source is set to an inactive
state, make obs_source_get_width/_get_height return 0 for their
dimensions instead.
When a frame is processed by a filter, it comes directly from the
source's video frame cache. However, if a filter is using or processing
those frames for whatever reason, there would be no guarantee that the
frames would persist during processing, and frames could eventually be
deallocated unexpected, for example when the resolution or format
changes.
So the solution is to implement simple reference counting for the frames
so that the frames will exist until they have been released by any
source or filter that's using them.
Fix a bug where when a source first starts up its async cache, it
unintentionally resets its cache, which means that the first few frames
would be lost.
This optionally specifies that the int/float value should be displayed
with a slider rather than and up/down control. UIs are not necessarily
required to implement it, it's meant to be more of a hint.
If negative was specified it would not parse the negative sign and thus
would not interpret that number as negative, and would cause
shader/effect parsing to simply fail on the file.
This is particularly important for the filter pipeline in order to
ensure that when the last filter is reached that the original blend
state is properly reset.
When render targets are used, they output to the render target inverted
due to the way that opengl works. This fixes that issue by inverting
the projection matrix so that it renders the image upside down and
inverting the front face from counterclockwise to clockwise.
The wrong function was being used to recurse through the filter chain in
obs_source_process_filter, obs_source_get_[width/height] would get the
post-effect dimensions rather than the pre-effect dimensions.
The frame is definitely meant to be modifiable if needed, so it
shouldn't be const. I originally meant for these frames to be
duplicated, but with the source video cache that's both unnecessary and
would reduce performance.
If the audio data had the same format/samplerate as the obs audio
subsystem, it would fail to simply destroy the resampler and set it to
NULL, and then any audio data going through would use the resampler that
was being used before that, causing audio to become garbage.
This bug only started appearing when I recently changed the libobs
internal audio subsystem format to non-interleaved floating point, which
is a common format, and thus caused this bug to actually occur more
often.
I when a source has both async audio/video capability, it would ignore
audio until the video has started. There's really no need to do this,
when the video starts it'll just fix up the timing automatically.
This should fix the case where things like the media source would not be
able to play audio-only files.
Core API functions changed:
-----------------------------
EXPORT bool obs_reset_audio(struct audio_output_info *aoi);
EXPORT bool obs_get_audio_info(struct audio_output_info *aoi);
To:
-----------------------------
EXPORT bool obs_reset_audio(const struct obs_audio_info *oai);
EXPORT bool obs_get_audio_info(struct obs_audio_info *oai);
Core structure added:
-----------------------------
struct obs_audio_info {
uint32_t samples_per_sec;
enum speaker_layout speakers;
uint64_t buffer_ms;
};
Non-interleaved (planar) floating point output is standard with audio
filtering, so to prevent audio filters from having to worry about
different audio format implementations and for the sake consistency
between user interfaces, make it so that audio is always set to
non-interleaved floating point output.
API changed from:
------------------------
EXPORT void obs_service_apply_encoder_settings(obs_service_t *service,
obs_encoder_t *video_encoder,
obs_encoder_t *audio_encoder);
void obs_service_info::apply_encoder_settings(void *data
obs_encoder_t *video_encoder,
obs_encoder_t *audio_encoder);
To:
------------------------
EXPORT void obs_service_apply_encoder_settings(obs_service_t *service,
obs_data_t *video_encoder_settings,
obs_data_t *audio_encoder_settings);
void obs_service_info::apply_encoder_settings(void *data
obs_data_t *video_encoder_settings,
obs_data_t *audio_encoder_settings);
These changes make it so that instead of an encoder potentially being
updated more than once with different settings, that these functions
will be called for the specific settings being used, and the settings
will be updated according to what's required by the service.
This fixes that design flaw and ensures that there's no case where
obs_encoder_update is called where the settings might not have
service-specific settings applied.
This code was originally meant to skip some checks as an optimization,
but it did not account for async video sources and would call
obs_source_default_render (which is only for synchronous video sources),
this fixes the issue by just calling obs_source_video_render.
The obs_source_get_width and obs_source_get_height functions need to
account for filters that may change their width/height such as a crop
filter or something similar.
As a side effect of this commit, because these functions need to lock
the filter mutex, these functions can no longer be used with a
const-qualified obs_source_t pointer.
This function is somewhat big and is a function called conditionally,
therefore having it be inline is somewhat of a waste if it's not always
being called.
Graphics data has to be freed inside of an active graphics context,
otherwise the resources will not be freed; in the future, we should
probably make sure that the graphics subsystem automatically
asserts/warns about this scenario.
I was iterating through the obs_source::filters array to remove filters,
but every time I removed a filter, it would remove itself from the
array, therefore it would end up skipping items in the array, therefore
leaving filters unreleased. This just removes the first filter until
all filters have been removed.
These signals introduce unnecessary complexity. Instead of emitting a
signal for a specific move direction, just signal that the scene has
been reordered, that way the target just refreshes the list.
For the show/hide and activate/deactivate callbacks, schedule these
callbacks to only be called from within the video thread rather than in
a separate thread. This ensures that any potential graphics activity
that occurs within them is kept in the same thread.
obs_source_inc_showing and obs_source_dec_showing are used to indicate
that a source is showing or no longer being shown in a particular area
of the program outside from the main render view.
One could use an obs_view, but I felt like it was unnecessary because
using an obs_view just to display a single source feels somewhat
superfluous.
The main view does not need to worry about hiding/deactivation of
sources when it's being freed here, when the obs context is shutting
down in this section of obs, all the sources are being freed, thus
there's no need to worry about deactivating/hiding sources.
When using multiple video encoders together with a single audio encoder,
the audio wouldn't be in sync.
The reason why this occurred is because the dts_usec variable of the
encoder packet (which is based on system time) would always be reset to
a value based upon the dts (which is not guaranteed to be based on
system time) in the apply_interleaved_packet_offset function. This
would then in turn cause it to miscalculate the starting audio/video
offsets, which are required to calculate sync.
So instead of calling that function unnecessarily, separate the check
for whether audio/video has been received in to a new function, and only
start applying the interleaved offsets after audio and video have
actually started up and the starting offsets have been calculated.
Instead of having services automatically apply encoder settings on
initialization (whether the output wants to or not), instead make it
something that must be explicitly called by the developer. There are
cases where the developer may not wish to apply the service-specific
settings, or may wish to override them for whatever reason.
If on windows, use the windows UTF conversion functions due to the fact
that the existing utf code is meant for 32bit wide characters, while the
windows conversion functions will properly handle 16bit wide characters.
Adds an additional search path for UI-independent and
installation-independent plugins for windows/mac.
Windows:
%appdata%/obs-plugins/
Mac:
~/Library/Application Support/obs-plugins/
Plugin directory format is [module]/bin and [module]/data.
On windows, for 32bit binaries:
[module]/bin/32bit
and 64bit binaries:
[module]/bin/64bit
Before: After:
obs_service_gettype obs_service_get_type
It seems there was an API function that was missed when we were doing
our big API consistency update. Unsquishes obs_service_gettype to
obs_service_get_type.
I didn't think it would ever need to be exported, but this function is
actually useful for applying settings to properties (to call all of
their update callbacks based upon the settings) without necessarily
having to have an object associated with it.
API changed:
--------------------------
void obs_output_set_audio_encoder(
obs_output_t *output,
obs_encoder_t *encoder);
obs_encoder_t *obs_output_get_audio_encoder(
const obs_output_t *output);
obs_encoder_t *obs_audio_encoder_create(
const char *id,
const char *name,
obs_data_t *settings);
Changed to:
--------------------------
/* 'idx' specifies the track index of the output */
void obs_output_set_audio_encoder(
obs_output_t *output,
obs_encoder_t *encoder,
size_t idx);
/* 'idx' specifies the track index of the output */
obs_encoder_t *obs_output_get_audio_encoder(
const obs_output_t *output,
size_t idx);
/* 'mixer_idx' specifies the mixer index to capture audio from */
obs_encoder_t *obs_audio_encoder_create(
const char *id,
const char *name,
obs_data_t *settings,
size_t mixer_idx);
Overview
--------------------------
This feature allows multiple audio mixers to be used at a time. This
capability was able to be added with surprisingly very little extra
overhead. Audio will not be mixed unless it's assigned to a specific
mixer, and mixers will not mix unless they have an active mix
connection.
Mostly this will be useful for being able to separate out specific audio
for recording versus streaming, but will also be useful for certain
streaming services that support multiple audio streams via RTMP.
I didn't want to use a variable amount of mixers due to the desire to
reduce heap allocations, so currently I set the limit to 4 simultaneous
mixers; this number can be increased later if needed, but honestly I
feel like it's just the right number to use.
Sources:
Sources can now specify which audio mixers their audio is mixed to; this
can be a single mixer or multiple mixers at a time. The
obs_source_set_audio_mixers function sets the audio mixer which an audio
source applies to. For example, 0xF would mean that the source applies
to all four mixers.
Audio Encoders:
Audio encoders now must specify which specific audio mixer they use when
they encode audio data.
Outputs:
Outputs that use encoders can now support multiple audio tracks at once
if they have the OBS_OUTPUT_MULTI_TRACK capability flag set. This is
mostly only useful for certain types of RTMP transmissions, though may
be useful for file formats that support multiple audio tracks as well
later on.
The comment says "these are different", but doesn't state why.
Actually, I should really rename the output flags so they're not flags,
but instead just "caps", because that's really all that they are.
obs_data_apply is used to apply the changes of a source object in to a
destination object. Problem with this however is that if sub-objects
are in use, it currently just copies the pointer of the sub-object,
meaning that the source and destination will both share the same
sub-object via reference. If anything modifies that sub-object data,
it'll modify it for both objects, which was not intended.
Instead of copying the object pointer, create a new copy and then
recursively repeat the process to ensure the data is always completely
separate.
Changed:
char *os_get_config_path(const char *name);
To:
int os_get_config_path(char *dst, size_t size, const char *name);
Also added:
char *os_get_config_path_ptr(const char *name);
I don't like this function returning an allocation by default.
Similarly to what was done with the wide character conversion functions,
this function now operates on an array argument, and if you really want
to just get a pointer for convenience, you use the *_ptr version of the
function that clearly indicates that it's returning an allocation.
The gs_enum_adapters function is an optional implementation to allow
enumeration of available graphics adapters that can be used with the
program. The ID associated with the adapter can be an index or a hash
depending on the implementation.
Direct3D textures are usually aligned to a specific pitch, so their
internal width is often not equal to the expected output width; this
means that if we want to use it on our texture output, that we must
de-align the texture while copying the texture data.
However, I unintentionally messed up the calculation at some point with
RGBA textures, so the variable size I was supposed to be using was
supposed to be multiplied by 4 (for RGBA), while I was still expecting
single channel data. So, if the texture width was something like 1332,
the source (directx) texture line size would be somewhere at or above
5328 (because it's RGBA), then destination is at 1332 (YUV luma plane),
and it would unintentionally treat 3996 (or 5328 - 1332) bytes as the
unused alignment data. So this fixes that miscalculation.
Refer to https://www.opengl.org/registry/doc/GLSLangSpec.4.10.6.clean.pdf
for a list of current (reserved) keywords.
In the future the shader compiler in libobs-opengl should probably take
care of avoiding those name conflicts (bonus points for transparently
remapping the names of effect parameters)
Instead of returning a valid string value when there are no more strings
available in the list, return NULL to indicate failure. An empty string
should really be allowed to be a valid value for the list.
The return value of os_sleepto_ns is true if it waited to the specified
time, and false if the current time is past the specified time. So it
basically returns true if it successfully waited.
I just didn't check the return value properly here, so it ended up just
setting the count of frames to 1 if overshot, ultimately causing sync
issues.
The temporary unoptimized code we were using before just completely
allocated a new copy of each frame every single time a new async frame
was output by the source plugin. This just creates a cache of frames as
needed for the current format/width/height to minimize the allocation
and deallocation. If new frames come in that are of a different
format/width/height, it'll just clear the cache. This is a fairly
important optimization.
all the async video related stuff usually started with async_*, and
there were two that didn't. So I just renamed them so they have the
same naming convention
If an async video source stops video for whatever reason, it would get
stuck on the last frame that was played. This was particularly awkward
when I wanted to give the user the ability to deactivate a source such
as a webcam because it would get stuck on the last frame.
This allows us to change the visible UI name of a property after it's
been created (particularly for a case where I want to change an
'Activate' button to 'Deactivate')
A slightly refactored version of R1CH's crash handler, allows crash
handling for windows which provides stack traces of all threads and a
list of all loaded modules. Also shows the processor, windows version,
and current libobs version.
When the encoder is set to scale to a different resolution than the obs
output resolution, make sure it uses the current video colorspace and
range by default.
I actually kind of hate how strstr returns a non-const even though it
takes a const parameter, but I can understand why they made it that way.
They really should have split it in to two functions though, one const
and one non-const or something. But alas, ultimately for a C programmer
who knows what they're doing it isn't a huge deal.
This adds support for the windows 8+ output duplicator feature which
allows the efficient capturing of a specific monitor connected to the
currently used device.
Previously, the design for the interaction between the encoder thread
and the graphics thread was that the encoder thread would signal to the
graphics thread when to start drawing each frame. The original idea
behind this was to prevent mutually cascading stalls of encoding or
graphics rendering (i.e., if rendering took too long, then encoding
would have to catch up, then rendering would have to catch up again, and
so on, cascading upon each other). The ultimate goal was to prevent
encoding from impacting graphics and vise versa.
However, eventually it was realized that there were some fundamental
flaws with this design.
1. Stray frame duplication. You could not guarantee that a frame would
render on time, so sometimes frames would unintentionally be lost if
there was any sort of minor hiccup or if the thread took too long to
be scheduled I'm guessing.
2. Frame timing in the rendering thread was less accurate. The only
place where frame timing was accurate was in the encoder thread, and
the graphics thread was at the whim of thread scheduling. On higher
end computers it was typically fine, but it was just generally not
guaranteed that a frame would be rendered when it was supposed to be
rendered.
So the solution (originally proposed by r1ch and paibox) is to instead
keep the encoding and graphics threads separate as usual, but instead of
the encoder thread controlling the graphics thread, the graphics thread
now controls the encoder thread. The encoder thread keeps a limited
cache of frames, then the graphics thread copies frames in to the cache
and increments a semaphore to schedule the encoder thread to encode that
data.
In the cache, each frame has an encode counter. If the frame cache is
full (e.g., the encoder taking too long to return frames), it will not
cache a new frame, but instead will just increment the counter on the
last frame in the cache to schedule that frame to encode again, ensuring
that frames are on time and reducing CPU usage by lowering video
complexity. If the graphics thread takes too long to render a frame,
then it will add that frame with the count value set to the total amount
of frames that were missed (actual legitimately duplicated frames).
Because the cache gives many frames of breathing room for the encoder to
encode frames, this design helps improve results especially when using
encoding presets that have higher complexity and CPU usage, minimizing
the risk of needlessly skipped or duplicated frames.
I also managed to sneak in what should be a bit of an optimization to
reduce copying of frame data, though how much of an optimization it
ultimately ends up being is debatable.
So to sum it up, this commit increases accuracy of frame timing,
completely removes stray frame duplication, gives better results for
higher complexity encoding presets, and potentially optimizes the frame
pipeline a tiny bit.