For some reason, this old Coverity issue was already marked as solved.
Also it is slightly more effective to store the end pointer than
evaluate in every pass.
cfg cannot be NULL so do not handle it in any way
In the historic API, empty options was passed as NULL but it is not the
case for some time so remove the NULL handling entirely not to be
confusing.
In addition to error code, the OSStatus usually represent a FourCC that
can be presented to user.
Some generic error may be interpreted by:
NSError *error = [NSError errorWithDomain:NSOSStatusErrorDomain code:err userInfo:nil];
const char *err_str = [[error localizedDescription] UTF8String];
But it looks like it almost always returns 'The operation couldn’t be
completed.' for unknown errors. Also there doesn't seem to be any domain
specific for CoreAudio.
This option is now mostly useless. From the prespective of UltraGrid
build system (CI), more defining are the binary dependencies by
Homebrow, which doesn't allow setting older Mac version.
Also the problem is, that the macros defined in Availability.h, like
__MAC_10_7 relate to the build system, not taking into account build
target. __ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ should perhaps be
used but it seems to be quite poorly defined.
Anyways, users can now provide the environment variable
$COMMON_OSX_FLAGS if really needed.
Refer also to GH-318 - now testcard may be used with group of devicese
all using analog or aesebu.
Also until that, although analog/aesebu were not accepted to play audio,
no error was prineted and the options was just unused (no sound
generated and no error).
strstr("help") has accidentally catched also the above case
+ do not check fmt to NULL, which cannot happen (also was incorrect - if
it could have (== no opts), we no longer display help but use def. vals)
The reported problem was not actually a bug, since feof() and ferror()
were both checked but both gcc and coverity complained about that and
changing it accordingly actually simplifies the code.
support 29.98, because 48 kHz is divisible by 24000/1001 (it's 2002
samples)
Note that 29.97 (or 59.94) is still not supported because 48 kHz is not
evenly divisible by that value (it is 1601.6 audio samples per 29.97
video frame).
On incorrect device index, the further call succeeds but rate returned
is 0, resulting in wrong assert in ring_buffer_init().
+ s/MODULE_NAME/MOD_NAME for convenience (usual in other code)
if no configuration is set, empty string ("") is passed, not NULL
This resulted in wrong device set 0 set because the atoi doesn't check
its arguments. Example of fixed behavior:
uv -s testcard -r coreaudio
`AudioUnitSetProperty(s->auHALComponentInstance, kAudioOutputUnitProperty_CurrentDevice,
kAudioUnitScope_Global, 1, &device, sizeof(device));` succeeded with 0
for some reason (but not with other values).
Since in last commit, the printed index has changed, print device name
to ensure user that it is the correct one.
+ fixed not freed device state array on cleanup
- describe 'fps' option - although the semantics is obvious, it was not
mentioned that it accepts 'i' (interlaced) suffix
- change FPS in examples to '59.94i' to illustrate the syntax
refer to GH-317
Limits frame rate to specified value.
This can be used if source doesn't keep its nominal frame rate, as some
Magewells do and use higher.
Another use case is just to limit the framerate from eg. 60 to 25 (which
are coprimes, for 60 and 30, already existing capture filter 'every')
can be used.
In Wayland, when resizing windows programmatically glfwSetWindowSize(),
if swap buffers was not triggered, event returning back to the original
resolution is triggered when swap buffer is triggered just after
drawing. It looks like that no drawing should occur before window size
change and swap buffers, seems there has been reports for SDL, eg. [1].
Also do not call gl_resize() after glfw_resize_window() (needless, it
will be triggered via callback after swapping the buffers).
To reproduce the wrong behavior - use Wayland with GLFW native Wayland
build, then: `uv -t testcard -d gl` (size may be any other than the
splashscreen size 512x512).
[1] https://github.com/libsdl-org/SDL/pull/4821
Window width is not taken directly from video desc but rather computed
according to height (to reflect optional aspect ratio). So the window
size needs to be rather rounded to respect odd sizes like 187x77, which
rounded down 186x77, eg.:
uv -t testcard:size=187x77 -d gl
Instead of crashing on assert, print user-friendly message that the Y4M
is not supported. Also print the properties of the file and what is
supported, which allows easy debug and user can eg. request the feature.
