We have the most common ones and we won't be able to convert them
otherwise.

We pass the mask verbatim so that's how the mask were interpreted so far.

It's never used as such. Only 2 masks existed and they have been replaced with
VLC_CODEC_RGB24 and VLC_CODEC_BGR24.

So that code passing a chroma as a string ends up on the on the same chroma
layout.

Using BI_BITFIELDS is only allowed for 16/32 bits. VLC always writes BI_RGB
and flips the bits to match it if necessary.

We have the most common ones and we won't be able to convert them
otherwise.

We pass the mask verbatim so that's how the mask were interpreted so far.

This is the format once the mask will be "fixed".

The RGB32/RGB24 masks are passed verbatim to VLC. So the same assumptions
we made on the masks were correct, regardless of endianness.

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

texture() should be used after GLSL 130 on OpenGL, and must be used
after GLSL 300 es on OpenGL ES.

Fixes #28369

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

The GLSL version will be used to generate the correct function names for
non-forward compatible functions.

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

Getting the shader source enable better debugging when an error happens.

Alexandre Janniaux authored 1 year ago and Jean-Baptiste Kempf committed 1 year ago

The es_id string is not used in the RTP module, leading to:

    ../../modules/stream_out/rtp.c: In function ‘Add’:
    ../../modules/stream_out/rtp.c:920:83: warning: unused parameter ‘es_id’ [-Wunused-parameter]
      920 | static void *Add( sout_stream_t *p_stream, const es_format_t *p_fmt, const char * es_id )
          |                                                                      ~~~~~~~~~~~~~^~~~~

Alexandre Janniaux authored 1 year ago and Jean-Baptiste Kempf committed 1 year ago

The alpha is not initialized for some chroma which were not using it.

    ../../modules/video_filter/blend.cpp: In function ‘void Blend(const {anonymous}::CPicture&, const {anonymous}::CPicture&, unsigned int, unsigned int, int) [with TDst = {anonymous}::CPictureYUVPlanar<short unsigned int, 1, 1, false, false>; TSrc = {anonymous}::CPictureRGBX<4>; TConvert = {anonymous}::compose<{anonymous}::convertBits<10, 8>, {anonymous}::convertRgbToYuv8>]’:
    ../../modules/video_filter/blend.cpp:542:32: warning: ‘spx.{anonymous}::CPixel::a’ may be used uninitialized [-Wmaybe-uninitialized]
      542 |             unsigned a = div255(alpha * spx.a);
          |                          ~~~~~~^~~~~~~~~~~~~~~
    ../../modules/video_filter/blend.cpp:537:20: note: ‘spx.{anonymous}::CPixel::a’ was declared here
      537 |             CPixel spx;
          |                    ^~~
    ../../modules/video_filter/blend.cpp: In function ‘void Blend(const {anonymous}::CPicture&, const {anonymous}::CPicture&, unsigned int, unsigned int, int) [with TDst = {anonymous}::CPictureYUVPlanar<short unsigned int, 1, 1, false, false>; TSrc = {anonymous}::CPictureRGBX<4>; TConvert = {anonymous}::compose<{anonymous}::convertBits<16, 8>, {anonymous}::convertRgbToYuv8>]’:
    ../../modules/video_filter/blend.cpp:542:32: warning: ‘spx.{anonymous}::CPixel::a’ may be used uninitialized [-Wmaybe-uninitialized]
      542 |             unsigned a = div255(alpha * spx.a);
          |                          ~~~~~~^~~~~~~~~~~~~~~
    ../../modules/video_filter/blend.cpp:537:20: note: ‘spx.{anonymous}::CPixel::a’ was declared here
      537 |             CPixel spx;
          |                    ^~~

Alexandre Janniaux authored 1 year ago and Jean-Baptiste Kempf committed 1 year ago

Fix warning:

    ../../modules/gui/ncurses.c: In function ‘DrawStats’:
    ../../modules/gui/ncurses.c:691:23: warning: comparison of integer expressions of different signedness: ‘int’ and ‘size_t’ {aka ‘long unsigned int’} [-Wsign-compare]
      691 |     for (int i = 0; i < item->es_vec.size ; i++) {
          |                       ^

Thomas Guillem authored 1 year ago and Steve Lhomme committed 1 year ago

Don't try to copy the whole frame to the jarray buffer but copy it chunk
by chunk (50ms max). Therefore, a same frame can now be used for more
than one java write.

This fixes the following error and abort when allocating a too big
jarray (likely when sending a lot of silence to catch up with a playback
way to early).

E VLC   : [9b14b710/138c] libvlc generic: jarray allocation failed
E VLC   : [9b14b710/138c] libvlc generic: Write failed: ERROR
F DEBUG : Abort message: 'art/runtime/java_vm_ext.cc:470] JNI DETECTED ERROR IN APPLICATION: JNI GetJavaVM called with pending exception java.lang.OutOfMemoryError: Failed to allocate a 99591132 byte allocation with 3792244 free bytes and 90MB until OOM'

Thomas Guillem authored 1 year ago and Steve Lhomme committed 1 year ago

It was only used with Android API 21 and 22 (Android 5.x).
Remove this code that behave differently than all other write methods
and that is not as tested as the others.

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

... otherwise match() does not work properly
when previous() is called.

All build scripts just call the build that way and end up installing
each contrib.

It should build fine.

The forced nasm builds nothing and we may need other tools. But
the logs only mention meson is needed.

After 9ca790a6 the DWORD turned into
an unsigned, but it should be an unsigned long.

It should be using an unsigned long, not an unsigned.

DRM_FORMAT_RGBA8888 haa a VLC mapping since c6a3c197

They return the same thing.

So we can remove vlc_video_format_drm().

We don't have any way to convert them back to anything useful.

They won't be used from rgb_fourcc_list in that case