GitLab

This saves a bit of space (though the message becomes a bit more
confusing). This also ensures that VLC_COMMOM_MEMBERS is actually at the
top of the struct (in C only).

- Playlist items are stored only once. No category / onelevel separation
- Services Discovery nodes always stay a tree
- The playlist and Media Library are tree or flat depending on variable "playlist-tree".
  It means that if the setting variable says flat, the items that come from playlist
  demuxers are flattened into a single level list.
- The design demanded a different playlist demuxing system. Now playlist demuxers should
  construct a tree of items under an input_item_node_t root and send that using
  input_item_AddSubItemTree. Currently, the old scheme will be retained in modules, because
  there is still some listening to the old event in code (the libvlc media system for example)

This is meant to fix the *printf format strings.
This is compatible with mingw 4.2.1 and 4.3.0
I haven't find a better place where to put this code...
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Those functions automatically abort if allocation fails (which is not
quite the same as calling assert()). Avoid these functions in new code.

Those are the same as the Linux kernel macros, and probably a bunch of
other projects. Usage:

    if (likely(condition))
        branch_likely_taken();

    if (unlikely(condition))
        branch_unlikely_taken();

Attention: those two macros convert the predicate to a boolean value.
Therefore you generally cannot use them for assignments like this:
    void *p_sys = likely (malloc (sizeof (*p_sys)));
    if (!p_sys)
        return VLC_ENOMEM;
Instead you can do this:
    void *p_sys = malloc (sizeof (*p_sys));
    if (unlikely (!p_sys))
        return VLC_ENOMEM;

when the return value is a *unique* pointer (or NULL)...

Currently, detection is done at build time. Feel free to add run-time
detect if you know how.

This is consistent with existing practice inside and outside of VLC.
We really don't need a separate error code for that.

It did not count non-ASCII characters at all (wrapping asian text must
have been fun...). Then again, to wrap even fixed-width font text, you
need to use wcwidth()/wcswidth().

David Flynn authored Apr 10, 2009 and Jean-Baptiste Kempf committed Apr 10, 2009

This reverts commit db01e438

.

Original commit was correct. -- davidf
Signed-off-by: David Flynn <davidf@rd.bbc.co.uk>
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

This reverts commit 3bceb9cf.

David Flynn authored Apr 01, 2009 and Jean-Baptiste Kempf committed Apr 03, 2009

fixup code handles this now.
Signed-off-by: David Flynn <davidf@rd.bbc.co.uk>
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

One useless thread gone (the new subsystem does not need a thread).