This was broken in cd222f8ce0 which had a
horrible commit message that makes replicating the issue impossible if
there weren't others who reported similar visual studio issues when
using a Japanese locale
This improves logging for when audio data insertion is way out of bounds
or is getting cut off in the front due to a bad negative sync offset.
Instead of throwing out a log message for every time this happens with
each piece of data, it now states when the out of bounds or cutoff has
started and stopped only.
This fixes a case where an insertion of audio data would pass
valid_timestamp_range yet the insert position would cause a negative
integer position and thus an unsigned integer overflow.
obs_data_create_from_json_file_safe: Attempts to create an obs_data
object from a file, and if that fails and a backup file exists, deletes
the old file and tries to open it again.
obs_data_save_json_safe: Saves json data to a temporary file first,
optionally backs up the target file if the file exists and backup_ext is
valid (otherwise deletes it), and then renames the temporary file to the
target file. This helps reduce the chance of json corruption on save.
This helper function saves to a temporary file first, (optionally) backs
up the original file, then renames the temporary file to the actual file
name. This helps reduce the chance of file corruption under various
circumstances (such as shutdown or crash while the file is being written
to disk).
API Changed:
---------------------------
From:
- bool obs_startup(const char *locale, profiler_name_store_t *store);
To:
- bool obs_startup(const char *locale, const char *module_config_path,
profiler_name_store_t *store);
Summary:
---------------------------
This allows plugin modules to store plugin-specific configuration data
(rather than only allowing objects to store configuration data). This
will be useful for things like caching data, for example looking up and
storing ingests from remote (rather than storing locally), or caching
font data (so it doesn't have to build a font cache each time), among
other things.
Also adds a module-specific directory for the UI
Manually specifies the UTF-8 character codes used by the file rather
than relying on the compiler to determine what the codes are manually.
I was getting compile errors due to the fact that my current code page
is not at the default code page; so visual c++ tried to use my current
code page rather than UTF-8 and it would cause an error on the file.
When getting a blank module data file (indicating you want to get the
path to the module data directory itself), on certain operating systems
(windows) it will fail if you end the path with '/' or '\'. So simply
make sure that if a blank module sub-path is used, don't try to append a
slash.
Just a little helper function that allows you to create an obs_data_t
object from a json file (rather than having to manually open it each
time and then call obs_data_create_from_json on the file data)
The rationale for rejecting these register calls is that these functions
may be required for the plugin to work properly, which can't be
guaranteed when libobs doesn't know about them.
This behavior may be revisited once the plugin manager is implemented,
to e.g. make it configurable (potentially per plugin) to allow loading
newer plugins in case they are known to work with the older libobs
The id and parent_id fields should now allow better recovery of the
actual call trees, though they aren't compatible between different data
dumps in a single profiler session anymore; for that reason the new
fields name_id and parent_name_id are introduced, they hold the old id
and parent_id values respectively
Due to all the threads in libobs it wouldn't be safe to make that
parameter reconfigurable after libobs is initialized without adding
even more synchronization. On the other hand, adding a function to set
the name store before calling obs_startup would solve the problem of
passing a name store into libobs, but it can lead to more complicated
semantics for obs_get_profiler_name_store (e.g., should it always return
the current name store even if libobs isn't initialized until someone
calls set_name_store(NULL)? should obs_shutdown call
set_name_store(NULL)? Passing it as obs_startup parameter avoids
these (and hopefully other) potential misunderstandings
(Non-compiling commit: windowless-context branch)
API Changed:
---------------------
Removed functions:
- obs_add_draw_callback
- obs_remove_draw_callback
- obs_resize
- obs_preview_set_enabled
- obs_preview_enabled
Removed member variables from struct obs_video_info:
- window_width
- window_height
- window
Summary:
---------------------
Changes the core libobs API to not be dependent upon a main window/view.
If you wish to draw to a window/view, use an obs_display object to
handle it.
This allows the use of libobs without requiring a window to be present
on the system. This is also prunes code that had to be needlessly
duplicated to handle the "main" window.
(Non-compiling commit: windowless-context branch)
Changes API from:
---------------------
EXPORT int gs_create(graphics_t **graphics, const char *module,
const struct gs_init_data *data);
To:
---------------------
EXPORT int gs_create(graphics_t **graphics, const char *module,
uint32_t adapter);
Summary:
---------------------
Changes the gs_create function to use an adapter parameter instead of
requiring a gs_init_data with window/color/etc information.
Intentionally breaks compilation when trying to compile the specific
merged commits within the windowless-context branch. This is meant to
be used in conjunction with a merge commit so that bisecting will never
see any non-compiling commits.
Microsoft basically deprecated GetVersion/GetVersionEx, so now you have
to query the file version of kernel32.dll in order to get the actual
windows version. Because of the steps involved in getting the windows
version are fairly complicated, this is an exported libobs helper
function that can be used to get the windows version if needed.
(Microsoft does have its own set of helper functions for this but the
API feels a bit.. awkward, and you can't actually get the real windows
version with them)
A minor optimization: in copy_rgbx_frame (used when libobs is set to
output RGBA frames instead of YUV frames), if the line sizes for the
source and destination match, just use a single memcpy call for all of
the data instead of multiple memcpy calls.
Added cast to unsigned and the assert because microsoft's compiler
doesn't support "%zu"
Actual warning:
libobs/graphics/effect-parser.c:1387:4: warning: format specifies type
'unsigned int' but the argument has type 'size_t' (aka 'unsigned long')
[-Wformat]
Certain plugins/modules that are loaded on windows may be dependent on
libraries that are located in the same directory as the module in
question. This makes is so that LoadLibrary will also search for
dependent libraries for that module in the module's directory.
When globbing for modules, the intent was to ignore the . and ..
directories that might also be included in the glob search. It was
mistakenly doing a string compare on the 'path' variable which contains
the full path, rather than the 'file' variable which just contains the
found file itself, so the string compare always failed.
In case the encoder has to use a different sample rate (due to the
sample rate being unsupported), we need an API function for the encoder
to get the sample rate that the encoder is actually running at.
When an audio filter is applied to a video source that also has
accompanying audio, it would cause the video from the source to stop
rendering.
The original code this was from was to prevent audio-only sources from
rendering video, but I neglected to make sure that this would not apply
to filters, and thus when an audio filter is on a source with video, the
code would kill the video.
The size parameter is the size of the elements, not the size of the
data. The size parameter should be 1, and the elements should be the
number of bytes.
The reason why I'm making this change is because the fread/fwrite would
fail when the parameters were swapped.
In the (unlikely) event of multiple concurrent calls to
input_method_changed it was possible that the log messages would appear
out of order with respect to which layout would actually be active
after the last log message
Due to the fact that async timestamps themselves can be susceptible to
minor jitter from certain types of inputs, increase the allowable jitter
compensation value to ensure that the rendered frame timing from async
video sources is always as close as possible to the compositor.
When the framerate of the source is the same as the framerate as the
compositor, this (combined with the fact that clamped video timing now
being used with async video frames) helps ensure that buffered async
video sources will sync up their rendering to the compositor as
accurately as possible despite jitter from the source's timestamps.
If there is no jitter in the source's timestamps then it'll always sync
up perfectly with the compositor, thanks to clamped video timing.
When playing back buffered async frames, this reduces the probability
that new frames will be missed/skipped due to jitter in the system
timestamps.
If a buffered async source is playing at the same framerate as the
compositor and there is no jitter in the async source's timestamps, then
the async source will play back perfectly in sync with the compositor
thanks to this change, ensuring that there's no skipped or missed frames
in video playback.
The "clamped" video time is the system time per video frame that is
closest to the current system time, but always divisible by the frame
interval. For example, if the last frame system timestamp was 1600 and
the new frame is 2500, but the frame interval is 800, then the
"clamped" video time is 2400.
This clamped value is useful to get the relative system time without any
jitter.
When buffering is enabled for an async video source, sometimes minor
drift in timestamps or unexpected delays to frames can cause frames to
slowly buffer more and more in memory, in some cases eventually causing
the system to run out of memory.
The circumstances in which this can happen seems to depend on both the
computer and the devices in use. So far, the only known circumstances
in which this happens are with heavily buffered devices, such as
hauppauge, where decoding can sometimes take too long and cause
continual frame playback delay, and thus continual buffering until
memory runs out. I've never been able to replicate it on any of my
machines however, even after hours of testing.
This patch is a precautionary measure that puts a hard limit on the
number of async frames that can be currently queued to prevent any case
where memory might continually build for whatever reason. If it goes
over the limit, it clears the cache to reset the buffering.
I had a user with this problem test this patch with success and positive
feedback, and the intervals between buffering resets were long to where
it wasn't even noticeable while streaming/recording.
Ideally when decoding frames (such as from those devices), frame
dropping should be used to ensure playback doesn't incur extra delay,
although this sort of hard limit on the frame cache should still be
implemented regardless just as a safety precaution. For DirectShow
encoded devices I should just switch to faruton's libff for decoding and
enable the frame dropping options. It would probably explain why no
one's ever reported it for the media source, and pretty much only from
DirectShow device usage.
Found via UBSan, actual errors (addresses not pruned for illustrative purposes):
"runtime error: store to misaligned address 0x7f9a9178e84c for type
'size_t' (aka 'unsigned long'), which requires 8 byte alignment"
"runtime error: load of misaligned address 0x7f9a9140f2cf for type
'size_t' (aka 'unsigned long'), which requires 8 byte alignment"
The screen index returned from XDefaultScreen is 0-based, and we were
decrementing it before the check to see if it had reached 0 rather than
after, so in the default_screen function it would always end up getting
either the wrong screen or no screen.
When xcb_query_pointer and xcb_query_pointer_reply was called with no
valid screen, it would fail with an error, thus making it so that the
mouse buttons could not be properly captured as hotkeys.
jp9000
Palana
kc5nra
Anthony Catel
Carl Fürstenberg
lemmi
Richard Stanway