Adding the `UPNP_` prefix just like the other variables will avoid
potential namespaces conflicts.

libupnp 1.6 is starting to get really old and caused bugs recently with
the new winpthread rework for contribs.

Added contrib patches:
  - revert-ifaddrs.patch (revert changes breaking android)

Removed contrib patches:
  - 0001-Do-not-try[...]           -> Fixed upstream by 386b7ed7
  - dont_use_down_intf.patch       -> Fixed upstream by 1f2e3b88
  - fix_infinite_loop.patch        -> Fixed upstream by 4367cbfa
  - libpthread.patch               -> Fixed locally in rules.mak
  - missing_win32.patch            -> Fixed upstream by 02bb90c5
  - no-getifinfo.patch             -> Fixed upstream by 0296c3cd
  - pthread-w32-checks.patch       -> Fixed upstream by e0f44dd9
  - use-inet-pton.patch            -> Fixed upstream by 44cdeeb7
  - use-unicode.patch              -> Fixed upstream
  - windows-random.patch           -> Fixed upstream by 44e092a6
  - [...]-pthread-w32-checks.patch -> Fixed upstream by e0f44dd9

Edited patches (mostly to fit the new sources):
  - miniserver.patch
  - upnp-no-debugfile.patch
  - libupnp-win32.patch (Partially fixed upstream)
  - libupnp-win64.patch (Partially fixed upstream by 421d5d28)
  - windows-version-inet.patch
  - libupnp-win32-exports.patch

`UpnpInit` is deprecated since 1.8 and removed in 1.14. `UpnpInit2`
offer a better replacement, working for both ipv6 and ipv4. There is no
reason to keep using `UpnpInit` apart from keeping the module usable
with 1.6.

These functions are inline and will be removed by the compiler anyway if
unused. The #if guards add unecessary compexity.

Rémi Denis-Courmont authored Mar 24, 2022 and Jean-Baptiste Kempf committed Mar 27, 2022

The keyword is nominally from C23, but it has been supported by GCC and
Clang/LLVM for years.

Steve Lhomme authored Mar 23, 2022 and Jean-Baptiste Kempf committed Mar 27, 2022

Fixes #26738

This is what the medialib now requires

Pierre Lamot authored Mar 25, 2022 and Jean-Baptiste Kempf committed Mar 27, 2022

This remains is disabled by default, The extra cost isn't really worth it as
move operations originated from the database are very rare, move operations from
the user are usually handled "a priori".

Pierre Lamot authored Mar 23, 2022 and Jean-Baptiste Kempf committed Mar 27, 2022

this would allow to move or update elements in "a priori" before getting the
notification from the database

Pierre Lamot authored Dec 20, 2021 and Jean-Baptiste Kempf committed Mar 27, 2022

this allows to submit to the view only the changes made on the model when the
database changes instead of invalidating and resetting the whole view.

data cache is changed from a windowed view over the data, to a model where data
is loaded from the start and loaded by chunk (append to the cache), when data is
invalidated the data is reloaded in background and the difference (insertions
and deletions) are propagated to the view.

Pierre Lamot authored Feb 16, 2022 and Jean-Baptiste Kempf committed Mar 27, 2022

this allows to compare two lists and generate a list of insertions and deletion
from the old list to the new one.

This is useful to UI frontends to preform partial updates of their view when
data change on the backend.

Steve Lhomme authored Mar 24, 2022 and Jean-Baptiste Kempf committed Mar 26, 2022

Make it static in core.c, it's dirty enough not to share it.

Steve Lhomme authored Mar 24, 2022 and Jean-Baptiste Kempf committed Mar 26, 2022

No functional changes.

Alexandre Janniaux authored Mar 25, 2022 and Hugo Beauzée-Luyssen committed Mar 26, 2022

Fix memstream stream being open but never closed when zero tracks were
to be displayed to the OSD. The code was opening the vlc_memstream from
the start but was freeing the pointer only when the close of the
memstream succeeded AND the number of tracks was non-null.

Instead, this patch reorder the code so that the vlc_memstream is
allocated only when there is at least one track to display and early
exit with the N/A OSD message otherwise.

Then it tries to close the memstream and exit in case of error.

Finally, the text is sent to the OSD and the pointer is freed.

It was raising an ASAN leak report on test_src_player through the
test_teletext test function.

Fixes #26732

Alexandre Janniaux authored Mar 25, 2022 and Hugo Beauzée-Luyssen committed Mar 26, 2022

Avoid the indentation level on the whole function, allowing to easily
document the early return case as a special edge case and clarifying the
rest of the block.

Thomas Guillem authored Mar 25, 2022 and Jean-Baptiste Kempf committed Mar 26, 2022

And rename stream functions from aout_* to vlc_aout_stream_*.

Thomas Guillem authored Mar 23, 2022 and Jean-Baptiste Kempf committed Mar 26, 2022

This change is required for the gapless rework, to identify 2 different
streams, using the same aout APIS. The gapless stream, will be created,
if it has the same format as the previous one, and will wait in the
first Play to make a transition with the previous stream.

This is also a prerequisite for the aout/stream rework (in all aout
modules) planned for VLC 5.0.

Splitting the 2 API internally partition the decoder.c code: the decoder
should only handle the aout stream and not the aout devices (like sound,
mute, selection...).

For now, the aout reference and ownership is still kept in the decoder.c
because the aout and input_resource don't allow creating more than one
streams from one aout. This will be fixed in VLC 5.0.