[LAU] [Jack-Devel] zombification behaviour...

Paul Davis paul at linuxaudiosystems.com
Thu Nov 18 20:04:24 UTC 2010


On Thu, Nov 18, 2010 at 2:50 PM, alex stone <compose59 at gmail.com> wrote:
> Torben, a few questions.
>
> 1. Is the uncertainty over knowing which client is the offender a
> result of the client's jack code being less than optimal? (no offence
> to coders)

not really.

there are 2 scenarios that can cause confusion:

   1) several clients, none of which is really bad by itself, but they
all contribute to excess CPU load
   2) errors by the kernel scheduler that make a particular client
look bad, when its really not

torben has done some work to reduce the impact of both of these, and
in all honest (2) is an issue that JACK can't really solve.
the first case is reduced by forcing a "bad client" to be one that on
its own takes more than an entire JACK period to execute.

> 3.) Would a more robust client ID system solve the challenge of jack
> knowing 100% which client is the offender?

no, this problem is unaffected by the type of "identifier" used for the clients.


More information about the Linux-audio-user mailing list