Alaric Senat authored 2 months ago and Steve Lhomme committed 2 months ago

Emscripten is known to deal badly with function casts to incompatible
types [^1]. This patch prevent us to accidentally merge harmful UB casts
to wasm enabled code. The warning is really strict but is compliant to the
standard. Code that wants to use function cast should be disabled from
the wasm platform.

[^1]: https://emscripten.org/docs/porting/guidelines/function_pointer_issues.html

Alaric Senat authored 2 months ago and Steve Lhomme committed 2 months ago

The function signature does not match the libplacebo callback (gl_t*
instead of void*). While this is supported by many platforms and
compilers, it's undefined behavior strictly by the standard [1] and it's
known to often fail with emscripten [^2].

[1]:
Section 6.3.2.3, paragraph 8:

> A pointer to a function of one type may be converted to a pointer to a
> function of another type and back again; the result shall compare
> equal to the original pointer. If a converted pointer is used to call
> a function whose type is not compatible with the pointed-to type, the
> behavior is undefined.

Quoting section 6.7.5.1, paragraph 2 for pointer compatibility:

> For two pointer types to be compatible, both shall be identically
> qualified and both shall be pointers to compatible types.


[^2]: https://emscripten.org/docs/porting/guidelines/function_pointer_issues.html

Alaric Senat authored 2 months ago and Steve Lhomme committed 2 months ago

The function signature does not match the dvbpsi callback (demux_t*
instead of void*). While this is supported by many platforms and
compilers, it's undefined behavior strictly by the standard [1] and it's
known to often fail with emscripten [^2].

[1]: 
Section 6.3.2.3, paragraph 8:

> A pointer to a function of one type may be converted to a pointer to a
> function of another type and back again; the result shall compare
> equal to the original pointer. If a converted pointer is used to call
> a function whose type is not compatible with the pointed-to type, the
> behavior is undefined.

Quoting section 6.7.5.1, paragraph 2 for pointer compatibility:

> For two pointer types to be compatible, both shall be identically
> qualified and both shall be pointers to compatible types.


[^2]: https://emscripten.org/docs/porting/guidelines/function_pointer_issues.html

Especially when mixing LPCWSTR and LPWSTR*.

Maxime Chapelet authored 2 months ago and Steve Lhomme committed 2 months ago

Given contribs are bootstrapped with `--disable-optimizations` option,
the CMAKE flag `-DCMAKE_BUILD_TYPE` is set to `Debug`, and when
build configuration is set to `Debug` in libpng contrib, it CMAKE
configuration append a `d` char to the static library name,
resulting in a lib named `libpng16d.a` instead of `libpng16.a`.
Then some vlc module depending on libpng aren't able to link
properly as they can't find their lpng16 dependency.

Alaric Senat authored 1 year ago and Steve Lhomme committed 2 months ago

Fixes a regression in the SAP descriptor generation.

Pierre Lamot authored 2 months ago and François Cartegnie committed 2 months ago

we can't rely on index to refer to an item across views as the ordering of the
model is not stable across the views.

This MR uses the artist MLId to set which subview should be loaded

fix: #28556

Pierre Lamot authored 2 months ago and François Cartegnie committed 2 months ago

this allows to request more data from the cache using the standart growth policy
rather than requiring the caller to `refer` and item after what is currently
loaded. This function ensure that the cache tries to load at least the current
count + chunk size

Pierre Lamot authored 2 months ago and François Cartegnie committed 2 months ago

This allows getting the position of an item in the model from its MLItemID

Pierre Lamot authored 2 months ago and François Cartegnie committed 2 months ago

this allows to retrieve the content of and item given its id

It avoids the warning in the core that uses whatever default it expects.
A SPU with absolute coordinates should always tell what dimensions it's
referring to.

For DVD SPU we know it has to correspond to the visible video area.

Even in 3.0 it's never set to true.

Fixes #28952

Fatih Uzunoğlu authored 3 months ago and Steve Lhomme committed 2 months ago

We can use the transparent area to restrict the client
size and use that remaining extra size for resizing.

This size needs to be adjusted with regard to the dpi.

Note that on Windows 10 and Windows 7, this brings
borders.

Fatih Uzunoğlu authored 2 months ago and Steve Lhomme committed 2 months ago

Not all `QOpenGL` stuff are in the Qt OpenGL
module. It appears that `QOpenGLFrameBufferObject`
was the only class provided by Qt OpenGL and
the rest are provided within Qt GUI.

Fatih Uzunoğlu authored 2 months ago and Steve Lhomme committed 2 months ago

Attachment does not need to be done each time the UI is rendered.

Fatih Uzunoğlu authored 3 months ago and François Cartegnie committed 2 months ago

Here, Qt wants to create a layer because when source is `QQuickImage`,
`QGfxSourceProxyME` checks its fill mode and source size.

In this case, we need to tell Qt that we don't need a layer here, as
we are doing the mapping as we want but Qt currently does not offer
a way to tell that and simply tries to decide itself: "If needed,
MultiEffect will internally generate a ShaderEffectSource as the
texture source"

It was added by mistake.

lbdvbcsa is GPLv2 [^1].

[^1] https://code.videolan.org/videolan/libdvbcsa/-/blob/1.1.0/COPYING

It's only picked as a dependency from the dvdread contrib which also requires GPL.
But it's cleaner to make it explicit.

Fixes this odd warning:
```
vlc/modules/text_renderer/sapi.cpp:64:12: warning: unused function 'RenderTextMTA' [-Wunused-function]
   64 | static int RenderTextMTA(filter_t *, subpicture_region_t *);
      |            ^~~~~~~~~~~~~
```

We don't use the content of the VIDEOINFOHEADER in the loop.
There is another loop later doing the same thing.

We can copy the known size without pointer casting.