Complete freeze of stage?

[Anthony Harrison]

I'm not entirely sure where this problem is being caused, but here's  
the situation:

I've got stage configured with two robots and a connection to the  
simulation interface. After about five minutes stage freezes up  
entirely and the javaclients do as well (freezing on readAll()). (but  
my other code continues to chug along).

What makes this interesting is that after killing the javaclients,  
stage resumes, which leads me to believe a communications buffer might  
be flooded.

Has anyone experienced something similar?

Now I am doing things a tad differently. First, I'm not running  
threaded. Instead I'm using executors to parallelize a large amount of  
the processing. At the top of each iteration, readAll() is called for  
each bot, and then separate executors do the sensor processing. After  
all the sensors are processed then new commands are issued (on the  
same thread that did the reading so there is no concern about sending  
munged commands due to different threads writing to the port).

The PlayerClients are instantiated as normal but I make no calls to  
runThreaded() or setNotThreaded() (since it defaults to not threaded).




  ...
  Anthony M. Harrison, PhD
  Cognitive Robotics & Human-Robot Interaction
  Naval Research Laboratory


-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
Don't miss this year's exciting event. There's still time to save $100.
Use priority code J8TL2D2.
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Java-player-users mailing list
Java-player-users@...
https://lists.sourceforge.net/lists/listinfo/java-player-users

[John Oliver]

I am afraid I have never experienced anything like this. First question would be what version of player/stage are you using? If you read back through the mail archives people have been having issues with player/stage 2.1, so you may want to downgrade your player/stage version until this is resolved.

Otherwise if there is any more information such as error messages from P/S you could provide it would be helpful.

I wasn't entirely sure from what you said if multiple executors were reading from a given port,  since if you have the system in PULL mode multiple clients reading the same port could also cause issues.

thanks

John

2008/4/30 Anthony Harrison <anthony.harrison@...>:

I'm not entirely sure where this problem is being caused, but here's
the situation:

I've got stage configured with two robots and a connection to the
simulation interface. After about five minutes stage freezes up
entirely and the javaclients do as well (freezing on readAll()). (but
my other code continues to chug along).

What makes this interesting is that after killing the javaclients,
stage resumes, which leads me to believe a communications buffer might
be flooded.

Has anyone experienced something similar?

Now I am doing things a tad differently. First, I'm not running
threaded. Instead I'm using executors to parallelize a large amount of
the processing. At the top of each iteration, readAll() is called for
each bot, and then separate executors do the sensor processing. After
all the sensors are processed then new commands are issued (on the
same thread that did the reading so there is no concern about sending
munged commands due to different threads writing to the port).

The PlayerClients are instantiated as normal but I make no calls to
runThreaded() or setNotThreaded() (since it defaults to not threaded).




 ...
 Anthony M. Harrison, PhD
 Cognitive Robotics & Human-Robot Interaction
 Naval Research Laboratory


-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
Don't miss this year's exciting event. There's still time to save $100.
Use priority code J8TL2D2.
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Java-player-users mailing list
Java-player-users@...
https://lists.sourceforge.net/lists/listinfo/java-player-users

-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
Don't miss this year's exciting event. There's still time to save $100.
Use priority code J8TL2D2.
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Java-player-users mailing list
Java-player-users@...
https://lists.sourceforge.net/lists/listinfo/java-player-users

[Anthony M. Harrison]

Oops, forgot the details.

Yes, I'm using 2.0.4 (not going near 2.1 until the java client is  
updated). Running on OS X.. (trying to get access to a linux box to  
test it there).

I am running in PUSH mode (default), with only the main thread calling  
readAll() before delegating to the parallelized processors who just  
query the interfaces for their data (if available).

On the error front, player (-d 9) is surprisingly quiet, but the log  
output itself is locked up:

...
playertcp.cc:550 warning : skipping message from 4:0 with unsupported  
type 4:6
playertcp.cc:550 warning : skipping message from 4:0 with unsupported  
type 4:6
playertcp.cc:550 warning : skipping message from 4:0 with unsupported  
type

That warning is repeated numerous times, still tracking down what it  
is.. (interface position2D, message ACK:DATA_FREQ?) - but it's  
interesting that the lock up prevents it from even finishing the last  
line (probably just not flushing the warnings).

With javaclient debugging on:

[PlayerClient][Debug] Type = PLAYER_MSGTYPE_RESP_ACK/4 , with payload  
size = 0 for 31:0
[PlayerClient][Debug] Type = PLAYER_MSGTYPE_RESP_ACK/4 , with payload  
size = 0 for 31:0
[PlayerClient][Debug] Type = PLAYER_MSGTYPE_RESP_ACK/4 , with payload  
size = 0 for 31:0
[PlayerClient][Debug] Type = PLAYER_MSGTYPE_RESP_ACK/4 , with payload  
size = 0 for 31:0
[PlayerClient][Debug] Type = PLAYER_MSGTYPE_RESP_ACK/4 , with payload  
size = 0 for 31:0
[PlayerClient][Debug] Type = PLAYER_MSGTYPE_RESP_ACK/4 , with payload  
size = 0 for 31:0
[PlayerClient][Debug] Type = PLAYER_MSGTYPE_RESP_ACK/4 , with payload  
size = 0 for 31:0
[PlayerClient][Debug] Type = PLAYER_MSGTYPE_RESP_ACK/4 , with payload  
size = 0 for 31:0

Which are in response to the positioning of 10 objects through the  
simulation interface (but only 8 of them are acknowledged).

I'm at a loss right now. I'd rather not revert to the threading model  
of javaclient since it's so wasteful of threads (and I'm fighting the  
urge to reimplement the io in a nonblocking manner).



   ...
   Anthony M. Harrison, PhD
   http://anthonymharrison.com/
   http://jactr.org/



-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
Don't miss this year's exciting event. There's still time to save $100.
Use priority code J8TL2D2.
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Java-player-users mailing list
Java-player-users@...
https://lists.sourceforge.net/lists/listinfo/java-player-users

[Anthony M. Harrison]

After reverting to the normal threading model of the java-client, I'm  
able to push the simulation past the four minute barrier, but it then  
freezes again after around six minutes.

It looks as though this is a player issue as others have encountered  
it (http://www.nabble.com/Re%3A-request%2C-reg%3A-non-trivial-problem-tt11237170.html#a11241196 
). The fix? Go to 2.1 (argh)

I looked in the public SVN repo and saw that there has been no  
activity related to 2.1, but Leo has commented that work for 2.1  
compat is on going (presumably on local hard drives). If that code  
were to be committed, others who are stuck in a holding pattern might  
be able to assist (myself most definitely included).

Leo, would you mind taking a few moments to bring folks up to date and  
make that code available?

   ...
   Anthony M. Harrison, PhD
   http://anthonymharrison.com/
   http://jactr.org/



-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
Don't miss this year's exciting event. There's still time to save $100.
Use priority code J8TL2D2.
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Java-player-users mailing list
Java-player-users@...
https://lists.sourceforge.net/lists/listinfo/java-player-users

[Leo Nomdedeu]

I really would like to but first I need to have a talk with Radu and a few minutes to organize all the code. Right now it's not usable, neither integrable as it need a complete rewrite of the player proxies.

We talked about having a new branch where definetly anyone could help to reach the 2.1 compatibility, leaving the 2.0 branch as is for backwards compatibility.

I will try to fix some parts and organize all the stuff asap, maybe for the next week or so.

Sorry for any inconvencience ...

Bests
Leo

2008/5/5 Anthony M. Harrison <amharrison@...>:

After reverting to the normal threading model of the java-client, I'm
able to push the simulation past the four minute barrier, but it then
freezes again after around six minutes.

It looks as though this is a player issue as others have encountered
it (http://www.nabble.com/Re%3A-request%2C-reg%3A-non-trivial-problem-tt11237170.html#a11241196
). The fix? Go to 2.1 (argh)

I looked in the public SVN repo and saw that there has been no
activity related to 2.1, but Leo has commented that work for 2.1
compat is on going (presumably on local hard drives). If that code
were to be committed, others who are stuck in a holding pattern might
be able to assist (myself most definitely included).

Leo, would you mind taking a few moments to bring folks up to date and
make that code available?

  ...
  Anthony M. Harrison, PhD
  http://anthonymharrison.com/
  http://jactr.org/



-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
Don't miss this year's exciting event. There's still time to save $100.
Use priority code J8TL2D2.
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Java-player-users mailing list
Java-player-users@...
https://lists.sourceforge.net/lists/listinfo/java-player-users

--
**************************************
Leonardo Nomdedeu Calvente
leo.nomdedeu@...
-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
Don't miss this year's exciting event. There's still time to save $100.
Use priority code J8TL2D2.
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Java-player-users mailing list
Java-player-users@...
https://lists.sourceforge.net/lists/listinfo/java-player-users