GitLab

Basically, somewhat of a "duh" now that we have optional_arguments on OSX.
It is highly annoying however that getopt_long behaviour differs so much
between gnu and BSD implementations, and then is even BUGGY on darwin.
'--option=value' is uniformly supported  and '--option value' is a GNU getopt thing.

imperfect world etc...

We definitely don't want to use the slow mutex or risky spin locks,
unless we have to.

The exact error message is: src/.libs/libvlccore.so: undefined reference to `__sync_fetch_and_sub_4'. According to irc discussion gcc version 4.1.2 (fc8), 4.1.3, 4.2, 4,3.2 and 4.3.2 (archlinux) fail with this error message. This patch fixes it by bumping up the minor number for gcc to 4.

(Received no comments on the mailing lists)

module_list_get(): gets the list of modules
module_list_free(): releases the list

Modules have no threads, no variables (!= config items) and
no plugin instances...

1/ pgp v4 signature checks of status files with more than 506 bytes would fail
2/ pgp v4 signature checks would fail (code dupication sucks).

This code is quite a mess. We shouldn't need to implement all this lowlevel stuff. I'll remove most of it and use pgpme instead.

For the time being, this means that upcomming releases should use v3 signatures for the downloaded files and shouldn't use status files of more than 506 bytes (or maybe shouldn't use v4 signatures for status files either)

services_discovery: Partially fix service discovery. We yet need to find a way to properly get the localized name... grr.

Previously we properly initialized the localized name of the object in Open(). Now Open() also runs Run() so we are kind of stuck. Please fix.

As discussed on IRC. Yield was quite confusing.

By the way, the event management seems quite broken. Events may occur
before _Create returns (even before this commit)...

Avoid objects being detached asynchronously too early.
Side effect is that you cannot free your children in the destructor
(that would be a reference loop).