Alexandre Janniaux authored 1 month ago and Felix Paul Kühne committed 1 month ago

libtool doesn't handle frameworks as dependencies, and only support them
as linker argument (ie. -Wl,-framework arguments). Remove them from the
link to prevent the linkage failure. This also means that they must be
provided manually in the makefile afterwards...

No need to set it on each iteration.

realloc works fine on NULL.
The pointer is set in p_subtitle->psz_text which is free'd later.
It's OK to free a NULL pointer if there wasn't any line to use.

No need to do a lengthy strlen() call each time the concatenated string grows.

We're just concatenating the lines in psz_text.

We already use flags to fill some arrays like AOUT_MIX_MODE_xxx
or VOUT_ALIGN_xxx.

The downside is that we cannot change the defines without breaking
the libvlc API. But SUBPICTURE_ALIGN_xxx defines are also used by
many module as source for parameters, so these would break as well.

It's clearer when it's not weird.

Fatih Uzunoğlu authored 1 month ago and Steve Lhomme committed 1 month ago

- `GetSystemMetrics()` function already scales most of the metrics for the
primary monitor [1], evidently because Qt makes the application DPI aware.
- Currently by multiplying the `GetSystemMetrics()` result with
`windowDpi / screenDpi`, we apply the window's DPI which is what is wanted
here.
- If the window is in the primary monitor, there is no problem. If there
are multiple monitors but they share the same DPI, there should also be no
problem. But if there are multiple monitors that have different DPIs, in
that case dragging would be problematic.
- The solution that Windows recommends [2] is using `GetSystemMetricsForDpi`
and `GetDpiForWindow`. Unfortunately these are only available on Windows 10,
so I dynamically load the necessary library and resolve the symbols.
- Handling `WM_DPICHANGED` and calling `SetWindowPos()` should not be
necessary, because Qt already calls it to resize the window according to
the new dpi.

[1] https://learn.microsoft.com/en-us/windows/win32/gdi/multiple-monitor-system-metrics

[2] https://learn.microsoft.com/en-us/windows/win32/hidpi/high-dpi-desktop-application-development-on-windows#updating-existing-applications

Alexandre Janniaux authored 3 months ago and Steve Lhomme committed 1 month ago

If the display is opened on a window that is not matching with a
DirectComposition backend or HWND backend, the module should not even
try to open.

Alexandre Janniaux authored 1 month ago and Steve Lhomme committed 1 month ago

The decoder fifo level metric allows interpreting the situation that led
to the fifo from the decoder to be flushed (ie. the fifo reached 400MiB
of undecoded data stored). The metrics generates a lot of points though,
so be sure to aggregate when displaying it.

Example of query in flux language:

    from(bucket: v.defaultBucket)
    |> range(start: v.timeRangeStart, stop:v.timeRangeStop)
    |> filter(fn: (r) =>
      r._measurement == "measurement" and
      r._field == "fifo_size"
    )
    |> aggregateWindow(every: 1s, fn: max)

The size metrics are casted into uint64_t, but are always below the
uint64_t max value anyway given they represent lesser values than
400MiB.

Alexandre Janniaux authored 1 month ago and Steve Lhomme committed 1 month ago

Unsigned support allows tracing sizes and counters.

Now that we can position subpictures outside of the video.

If subtitles in black bar is disabled, it's equivalent to the video area.

SPUs using substext can change it if they need to be in the video.

This does the same as using the window dimensions when SPU in black
bars was forced by the core.

...rather than in the video.
It's off by default for now.

This effectively disables "subtitles in black bars" as no SPU source is
setting this flag yet.

"subtitle in black bars" is not opt-in per SPU codec, and per region
for each codec. It is off by default.

Each subpicture needs to decide for itself if it's using the whole display area.

We can't tell before the region is created/updated if it's going to be in video coordinates
or display coordinates.

The core doesn't need to know anymore if a picture is a "subtitle"
(contains timestamps) in "absolute" (video or screen?) coordinates.
It just needs to provide both dimensions.

Since we keep the alignment as well.

They are always "absolute", no need to set it on internal regions.

This is how the IOCallback abstraction works [^1].
Either you read the amount requested, or you return 0.

vlc_stream_Read() reads the whole amount of data, unless it reaches the end of file.
In that case it returns the amount it could read.

[^1]: https://github.com/Matroska-Org/libebml/blob/ba5707e3268be330516c678c7de14af734f0f3b9/ebml/IOCallback.h#L30

Since a878e9d6.

Also remove duplicate include of <climits>.

Alexandre Janniaux authored 1 year ago and Felix Paul Kühne committed 1 month ago

This commit signal the mostlyclean and other dependant targets that it
should clean the targets generated by cargo.

We use a separate target name and add the dependency to
mostlyclean-local so that we can still extend the known automake target.

We also don't use CLEANFILES so that dependencies files are also removed
when using `make mostlyclean-local-rust` to avoid a corrupted tree where
the dependencies would not be removed but the target files they
reference would not exist anymore.

Alexandre Janniaux authored 1 year ago and Felix Paul Kühne committed 1 month ago

It's easier to setup CLEANFILES to $(NULL) at the top makefile and then
always append to it whenever needed.

Steve Lhomme authored 1 month ago and Felix Paul Kühne committed 1 month ago

Either when the video is in portrait mode or when the subtitles
are shown in the black bars and the video area is in portrait mode,
the text can become very large just because we apply a percentage base on
the output height.

We switch mode when the height becomes bigger than the width
so there's a continuity in the ratio when growing/shrinking the
video in one direction or the other.

The static libraries are used in multiple modules for the android
codebase.