Hi all,

what causes this? I will greatly appreciate your help:

Traceback (most recent call last):
  File "C:\Users\tim\code_base\svn\octopus\parallel_python\pp.py", line 762, in
__run
    sresult = worker.t.receive()
  File "C:\Users\tim\code_base\svn\octopus\parallel_python\pptransport.py", line
 133, in receive
    msg = self.r.read(msg_len)
OverflowError: long int too large to convert to int
Unhandled exception in thread started by <bound method Server.__run of <parallel
_python.pp.Server object at 0x017EEC70>>
Traceback (most recent call last):
  File "C:\Users\tim\code_base\svn\octopus\parallel_python\pp.py", line 771, in
__run
    job.finalize(sresult)
UnboundLocalError: local variable 'sresult' referenced before assignment

Hi everyone

I'm getting similar issues. In my case, the message is:


Traceback (most recent call last):
  File "/home/wynand/.local/lib/python/pp.py", line 760, in __run
    worker.t.csend(sfunc)
  File "/home/wynand/.local/lib/python/pptransport.py", line 96, in csend
    self.send("H" + hash1)
  File "/home/wynand/.local/lib/python/pptransport.py", line 125, in send
    self.w.write(struct.pack("!Q", len(msg)))
IOError: [Errno 32] Broken pipe
Unhandled exception in thread started by <bound method Server.__run of <pp.Server object at 0x2eec890>>
Traceback (most recent call last):
  File "/home/wynand/.local/lib/python/pp.py", line 771, in __run
    job.finalize(sresult)
UnboundLocalError: local variable 'sresult' referenced before assignment


I've looked for everyone with similar problems and the problems seem to occur with Python 2.5.

On my machine with Python 2.6, everything works. But when running the code on a 32-bit machine with Python 2.5.1 or a 64-bit machine with Python 2.5.2, I get the above error..

I have verified that the client function never gets called (I tried logging to a file, but nothing appeared, even when the logging code is executed before anything else).

I checked if maybe I was passing parameters that pickle in a weird way, but I get the same issues even if I pass only integers and lists.

I am afraid that I cannot give you a simple example that triggers this, since my code is very convoluted and intertwined (academic code ) and I would have to start from scratch to get a minimal example.

It seems as if the subprocess dies, probably due to an exception. But I can't see what that exception is - is there a way to see it?

I am now getting this error in python 2.5 (and 2.4) on Debian testing (both 32- and 64-bit).  My code had worked for over a year, but then returning to it after a short break I now get these Broken Pipe issues.

In trying to make a short example (my actual code is long) I ran into the numpy bug that I posted separately.  I have yet to get a short example to break, but I'll add to this post if I do.

In the meantime, is there anything I could look at to narrow down why this Broken Pipe is occurring?

Thanks,
Per

I got this kind of error from using a simple subprocess call, if you use subprocess you should invoke it the following way:


and then either call wait() on p or communicate().

Hi Vitalii,

The first post by mako01 is closest to what I am getting reported at
http://www.parallelpython.com/component/option,com_smf/Itemid,29/topic,414.0

It appears to occur whenever a module that prints to stdout/stderr is passed to submit(). This printed string is later caught by worker.t.receive() rather than being sent to stdout.
Could you propose a workaround?
I have been able to reproduce it using a simple module that prints a string. And it goes away if i redirect sys.stdout. But somehow the trick of redirecting sys.stdout doesn't work for moose. Not even redirecting both sys.stdout and sys.stderr. See my original post at above link.
Hope you can elucidate.
Thanks,
Aditya.