GitLab

If vlc_cancel() while the timer thread is between mwait() and
vlc_mutex_lock(), a dead lock would occur at vlc_join(). We need to
release the timer lock first.

This is inefficient. It could be optimized a lot with
epoll/timerfd on Linux and kqueue/kevent on BSD. Unfortunately, the
POSIX timer API is impossible to use in a thread-safe manner, other
than with signals.

Note that this can only be used safely in the core. In a plug-in, it
would introduce a race whereby dlclose() unmaps the code segment that
a detached thread is running (this was discussed over a year ago).
For that, we'd need something a bit more involved along the lines of
Win32's FreeLibraryAndExitThread().

This avoids using real-time when not asked. It should also fix a
pthread_create() permission failure on FreeBSD (Linux seems to ignore
this error silently).

Just to get compilation working. Any use of Timer API will result in an abort() for now.

Introduced in [5ead92ffd58088828ed802173d963b2571f0cd69], on PowerPC-based Macs this breaks the playback of a certain AVI and WMV files for reasons not fully understood. The commit was reverted for PPC macs in
* [49f3721c5baa88feece1d8793f08f08b2ed86a4d]
* [7786caa59f70d40794273cd741ea96f98d43621f]
but these commits were never applied to the master branch.

On Linux, these functions cannot fail. On Windows, mutexes cannot fail
and neither can real condition varaibles (but we use events instead to
support pre-Vista versions). In practice, the needed resources are
allocated as part of the mutex structure and per-thread data.

In any case, basically none of the call sites for those functions were
checking for errors. It seems best to abort in case of error than to
hit undefined behaviour. We probably do not want to clutter the tree
with untested never used error paths.