This is an arbitrary choice. But KiB are much more common than kB.
Also the notation is unambiguous.

Note that we keep kbps for *bit* rates.

Lower the size of the module_t structure because not any module as more than 15 shortcut (I choose 20 to be safe)

Thanks to Pierre d'Herbemont who pointed this out.

When deleting, clear cache in any case since deleting an item deletes all it's children too.

This reverts commit 58275b9d.

This is probably too late for this kind of changes. And it's probably not the best way to reduce the hudge memory footprint of the module bank.

Generally the core won't call Close() until the end of vlc's life. Hence
it's our last chance to gc collect before the end of vlc's life.

This gets important because some sd (freebox...) will open a vlc.stream
which (and that's an other issue) will eat 12MB at creation for a cache
that will never be completely filed. It also mean that if you vlc.stream
N small files, we may consume N * 12MB until finally lua decides to garabge
collect. And because we don't teach lua that vlc.stream takes 12MB this
is not likely to happen.

This reverts commit d38329b4.

This is reverted because it is a hack.

Should fix #3385

Only input jitters lower than clock-jitter will be compensated. You
may set it to 0 to avoid having the latency increased automatically
by VLC.

No functionnal changes.

Now you just need to properly set --host to build for the arch you want (ppc, i386, x86_64).

This could be mitigated one day by using a configure flag for command line builder.

There are no reason to use Git anymore, since upstream is now clear and reactive.

This reverts commit b21ac035.

Commit cc59f6552bde51010ec2fc67bfa101a44acd6967 fixes building for win32.
Signed-off-by: Jean-Paul Saman <jean-paul.saman@m2x.nl>

Conflicts:

	projects/mozilla/vlcplugin.cpp

ifdef the dependency on pthread for windows. It restricts pthread usage to XP_UNIX only. The events object should be extended to support MacOS X and Win32 too, this will be done in coming commits.
Signed-off-by: Jean-Paul Saman <jean-paul.saman@m2x.nl>

And fix logic

At least we can now see the difference between loop1 and loop-all