Ok, so following a bit on this. I'll summarize some of our findings, some of them maybe obvious for some people in this list.
I saw that there are actually two different kind of VMs for *nix [1] : - threaded heartbeat - itimer + signal heartbeat
Specially, I'd like to cite the following paragraph for the lazy:
A distinction on linux is between VMs with an itimer hearbeat or a threaded heartbeat. VMs with an itimer hearbeat use setitimer to deliver a SIGALRM signal at regular intervals to interrupt the VM to check for events. These signals can be troublesome, interrupting foreign code that cannot cope with such signals. VMs with a threaded heartbeat use a high-priority thread that loops, blocking on nanosleep and then interrupting the VM, performing the same function as the itimer heartbeat but without using signals. These VMs are to be preferred but suport for multiple thread priorities in user-level processes has only been available on linux in kernels later than 2.6.12.
So, I downloaded the heartbeat squeak VM from bintray [2]. This VM requires so deploy some configuration files in /etc/security [2].
Under this configuration OSSubprocess worked like a charm (or I did not found the issue again so far).
Now, this heartbeat threaded VM is the recommended in the README file, and we see that OSSubprocess generates the exact issue stated. The main problem remains for the moment since Pharo's default download includes not this VM but the itimer one. I talked with Esteban about it and he was aware of these two VM flavours, and the reason why we are using the itimer one is the need to deploy those permission files in /etc, which makes installation a bit less automatic.
Cheers, Guille
[1] https://github.com/OpenSmalltalk/opensmalltalk-vm/blob/e17db79411cfec767e04f... [2] https://bintray.com/opensmalltalk/vm/download_file?file_path=cog_linux32x86_... [3] https://github.com/OpenSmalltalk/opensmalltalk-vm/releases/tag/r3732#linux
-------- Original Message --------
I asked Santa an OSSubprocess that won't hang :)
I really appreciate to see such not so small polar elfves helping santa :)
Stef
Thanks,
Guille
-------- Original Message --------
On 19 Dec 2016, at 14:41, Mariano Martinez Peck marianopeck@gmail.com wrote:
Hi guys,
Guille Polito kept one of these images if someone can give us a hand. He also proposed the great idea of using `strace` to see what was going on. He (together with Pable Tesone) suspected that the heartbeat could be interrupting the `clone()` function which is (I think) called internally by the 'posix_spawn()' which is the one used by OSSubprocess.
When these images are "hung" they found at a infinitive loop like this:
Okay, how many child processes do you have at that point? How many processes does the system have?
[pid 17477] --- SIGALRM {si_signo=SIGALRM, si_code=SI_KERNEL} ---
[pid 17477] gettimeofday({1482152630, 593498}, NULL) = 0
[pid 17477] sigreturn() (mask []) = 120
[pid 17477] clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0xf7578768) = ? ERESTARTNOINTR (To be restarted)
[pid 17477] --- SIGALRM {si_signo=SIGALRM, si_code=SI_KERNEL} ---
[pid 17477] gettimeofday({1482152630, 600126}, NULL) = 0
[pid 17477] sigreturn() (mask []) = 120
[pid 17477] clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0xf7578768) = ? ERESTARTNOINTR (To be restarted)
so above.. 7ms between the two gettimeofday calls? Nothing else? Set a breakpoint on clone/fork in gdb and look at the c-stack at this point? Could you strace with timestamps to see how much time is spent? Is the process suspicious in other ways?
So yes.. sounds like clone doesn't complete.. the question is why? Is it out of resources? Is something in the VM blocking longer than the heartbeat, is the heartbeat more frequent than expected?
As you can see, there is a SIGALARM involved. It also looks like the `gettimeofday` is used by the heartbeat ? Could it be that someone the heartbeat is interrupting the `clone()` ?
Guille also showed me the `strace` output with a regular / working image:
[pid 18647] --- SIGALRM {si_signo=SIGALRM, si_code=SI_KERNEL} ---
[pid 18647] gettimeofday({1482152829, 481014}, NULL) = 0
[pid 18647] sigreturn() (mask []) = -1 EINTR (Interrupted system call)
[pid 18647] getitimer(ITIMER_REAL, {it_interval={0, 2000}, it_value={0, 1917}}) = 0
[pid 18647] recvmsg(3, 0xff7b0734, 0) = -1 EAGAIN (Resource temporarily unavailable)
[pid 18647] select(4, [3], [], [3], {0, 1000}) = 0 (Timeout)
[pid 18647] getitimer(ITIMER_REAL, {it_interval={0, 2000}, it_value={0, 797}}) = 0
[pid 18647] recvmsg(3, 0xff7b0734, 0) = -1 EAGAIN (Resource temporarily unavailable)
[pid 18647] select(4, [3], [], [3], {0, 1000}) = ? ERESTARTNOHAND (To be restarted if no handler)
[pid 18647] --- SIGALRM {si_signo=SIGALRM, si_code=SI_KERNEL} ---
Does anyone have any hint here?
Get timestamps in there. How long does it take to fail/end in this situation?
holger