Re: [Jack-Devel] jack threads

On Mon, May 2, 2011 at 7:16 AM, Tim Blechmann <[hidden]> wrote:

> atm i am doing it manually:
> if (jack_client_thread_id(self->client) == pthread_self())
>   do_something();

then i would keep on doing it manually :)

>> i think this would be a mistake too, because it would mean
>> that if thread semantics changed in the future, your code would likely
>> break (imagine a future in which libjack initiates N "realtime"
>> threads and not just 1, for example).
>
> well, if you change any part of the API, code will break, if it relies on the
> old semantic ...

my point is that the thread semantics have never been specified (which
is a mistake) but that as a result of the variety of schemes that
exist or can now be imagined, trying to tell clients "this is thread
FOO" has the effect of locking down the thread model in a way that we
currently do not. most (all) JACK clients function with both the jack1
and jack2 thread models. what precisely is the identity of the thread
in jack1 that calls the process callback? it also calls the port
registration callback ... in fact, even jack_client_thread_id() is
wrong and hopelessly jack1-centric:

"    the pthread ID of the thread running the JACK client side code. "

*what* client side code?

--p

1304335408.19547_0.ltw:2,a <BANLkTiknMOMrgj9qZvCP2b5Yob_FvekJ_Q at mail dot gmail dot com>