www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - OwnerTerminated graceful handling

reply Eitan Frachtenberg <etc facebook.com> writes:
N00b question: Based on the TPDL, explicitly receive()ing OwnerTerminated p=
revents an exception from throwing when the owner thread terminates. Yet th=
is is exactly what happens in the following code:

import std.concurrency;

void thr() {
    receive(
      (OwnerTerminated) { return; }
   );
}

void main() {
  spawn(&thr);
}

Try to wrap the receive with a try/catch pair only causes a segfault, but t=
hat's not the point anyway. What am I missing something here?

Thanks,
--Eitan.
Oct 06 2010
next sibling parent reply Sean Kelly <sean invisibleduck.org> writes:
Eitan Frachtenberg Wrote:

 N00b question: Based on the TPDL, explicitly receive()ing OwnerTerminated
prevents an exception from throwing when the owner thread terminates. Yet this
is exactly what happens in the following code:
 
 import std.concurrency;
 
 void thr() {
     receive(
       (OwnerTerminated) { return; }
    );
 }
 
 void main() {
   spawn(&thr);
 }
 
 Try to wrap the receive with a try/catch pair only causes a segfault, but
that's not the point anyway. What am I missing something here?

You're missing the fact that I forgot to implement this feature :-) Fixed in SVN.
Oct 07 2010
parent Sean Kelly <sean invisibleduck.org> writes:
I'm not sure if you've read TDPL, but the basic idea is that if thread A spawns
threads B and C, B and C should typically not outlive thread A.  Manual cleanup
of spawned threads tends to be messy, so this functionality was built into the
messaging system as the OwnerTerminated exception.  So currently, if a spawned
thread should outlive its creator it should receive OwnerTerminated and ignore
it.

I've added the beginnings of bidirectional linking as well via spawnLinked and
the LinkTerminated exception, so monitor threads can be created to track the
status of worker threads.  There is still a lot of functionality missing
however.

Eitan Frachtenberg Wrote:

 I still don't understand how it's supposed to work. Replace the line
'(OwnerTerminated) { return; }' with '(MsgToClient) { return; }', where
MsgToClient is an enum, still crashes gloriously with an OwnerTerminated
exception. Is there any way I can cleanly stop a thread without crashing?
  
 On Oct 7, 2010, at 10:38 AM, Sean Kelly wrote:
 
 Eitan Frachtenberg Wrote:
 
 N00b question: Based on the TPDL, explicitly receive()ing OwnerTerminated
prevents an exception from throwing when the owner thread terminates. Yet this
is exactly what happens in the following code:
 
 import std.concurrency;
 
 void thr() {
    receive(
      (OwnerTerminated) { return; }
   );
 }
 
 void main() {
  spawn(&thr);
 }
 
 Try to wrap the receive with a try/catch pair only causes a segfault, but
that's not the point anyway. What am I missing something here?

You're missing the fact that I forgot to implement this feature :-) Fixed in SVN.


Oct 13 2010
prev sibling next sibling parent sybrandy <sybrandy gmail.com> writes:
On 10/06/2010 07:48 PM, Eitan Frachtenberg wrote:
 N00b question: Based on the TPDL, explicitly receive()ing OwnerTerminated
prevents an exception from throwing when the owner thread terminates. Yet this
is exactly what happens in the following code:

 import std.concurrency;

 void thr() {
      receive(
        (OwnerTerminated) { return; }
     );
 }

 void main() {
    spawn(&thr);
 }

 Try to wrap the receive with a try/catch pair only causes a segfault, but
that's not the point anyway. What am I missing something here?

 Thanks,
 --Eitan.

I've seen the same issue. From what I've seen, std.concurrency is still a work in progress, so hopefully the bugs will get worked out soon. Casey
Oct 07 2010
prev sibling next sibling parent Eitan Frachtenberg <etc facebook.com> writes:
I still don't understand how it's supposed to work. Replace the line '(Owne=
rTerminated) { return; }' with '(MsgToClient) { return; }', where MsgToClie=
nt is an enum, still crashes gloriously with an OwnerTerminated exception. =
Is there any way I can cleanly stop a thread without crashing?
=20
On Oct 7, 2010, at 10:38 AM, Sean Kelly wrote:

 Eitan Frachtenberg Wrote:
=20
 N00b question: Based on the TPDL, explicitly receive()ing OwnerTerminate=


this is exactly what happens in the following code:
=20
 import std.concurrency;
=20
 void thr() {
    receive(
      (OwnerTerminated) { return; }
   );
 }
=20
 void main() {
  spawn(&thr);
 }
=20
 Try to wrap the receive with a try/catch pair only causes a segfault, bu=


=20
 You're missing the fact that I forgot to implement this feature :-)  Fixe=

Oct 11 2010
prev sibling parent Eitan Frachtenberg <etc facebook.com> writes:
On Oct 13, 2010, at 10:09 AM, Sean Kelly wrote:

 I'm not sure if you've read TDPL, but the basic idea is that if thread A =

ual cleanup of spawned threads tends to be messy, so this functionality was= built into the messaging system as the OwnerTerminated exception. So curr= ently, if a spawned thread should outlive its creator it should receive Own= erTerminated and ignore it.
=20
 I've added the beginnings of bidirectional linking as well via spawnLinke=

ack the status of worker threads. There is still a lot of functionality mi= ssing however.
=20

Yes, I have read TDPL. I suppose I don't really care about spawned worker t= hreads outliving owner thread, as long as the owner has a clean way to sign= al to the worker that it should stop now.
Oct 13 2010