Re: mach_port_deallocate(kernel_thread) failed: (os/kern) invalid
Re: mach_port_deallocate(kernel_thread) failed: (os/kern) invalid
- Subject: Re: mach_port_deallocate(kernel_thread) failed: (os/kern) invalid
- From: "Paul Forgey" <email@hidden>
- Date: Sun, 31 Oct 2004 20:45:34 -0800
If I may butt in here, your program is not doing what you think it is. Like I
told you when we were talking about this last month, threads need to be joined
or detached. There are three ways to release these resources being taken up by
a thread, and all are mutually exclusive:
Create it detached
Detach it
Join in.
Pick one. It's important to do exactly one of these things with your thread, or
you will leak memory.
It sounds likely to me that you aren't pthread_join()'ing a valid thread handle?
That would explain both the kernel warnings and the fact a pthread_detach() is
needed despite your pthread_join() call to release the extra memory. At any
rate, if your testing indicates otherwise, then you have a bug in your program
that is causing it to not join where, when or what you think it is.
----- Original Message -----
From: "Marc Van Olmen" <email@hidden>
To: "Matt Watson" <email@hidden>
Cc: <email@hidden>
Sent: Sunday, October 31, 2004 8:03 PM
Subject: Re: mach_port_deallocate(kernel_thread) failed: (os/kern) invalid
On Oct 31, 2004, at 10:43 PM, Matt Watson wrote:
Matt,
My tests do NOT justify what you are writing here. If I remove the
pthread_detach after the pthread_join. I notice that my virtual memory
(checking with Activity monitor) is growing.... because doing this
pthread_join 2000 times in this manner my program stopped working
because no more memory available and virtual memory of my app growed to
4GB. When I added the follwing:
if (mThread) {
pthread_detach(mThread);
}
I need the join so the PTHREAD_CREATE_DETACHED wasn't an option.
I had no more crashes, no more virtual memory leaks and no errors
anymore in the console, after testing for a long time.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden