VoIP Mailing List Archives

[wiltingtree at gmail.com]

Thanks for the response Traun.  The version of Python is 2.4.3, and I didn't build it myself, I installed it with yum.
The version of Red Hat is  4.1.2-41.
"import threading" works fine, so I don't think it's a Python threading issue.
The FreeSWITCH version I installed is the freeswitch-1.0.3.tar.gz located at files.freeswitch.org.
I didn't make any major changes to the configuration; I enabled Python and set-up the SIP profile, directory and dialplan. No other changes.
Any other help would be appreciated, since I really don't know where to look.


Thanks,
Adam

[tleyden at branchcut.com]

Hi Adam,

I'm stumped .. I guess you could try the following:

* Try with the trunk version of freeswitch.  I don't think it will matter, but just in case

* Try to simulate the same test with a Lua script.  Do you see the same problem?

If those don't turn up anything, then the next logical step would be
to start adding printf() statements in the mod_python code and
find out where it is getting stuck.  In particular around the parts where
it swaps the threadstate in and out.  I might be able to create a patch
for you, but try those other tests first.

HTH,
Traun

 

[wiltingtree at gmail.com]

Traun, thanks again for your help.
I followed your advice and I made some progress!
 
I tested with the latest trunk version and also with 1.0.2, and both exhibited the same behavior.
I then tried writing a test script in Lua, and it worked fine.
So this meant the problem was in the Python module (I was sure it was some FS config issue).
So I started playing with a small test Python script, and I narrowed the problem down to when I'm using the "read" function.
Here is my test script:
 
from freeswitch import *
def handler(session, args):
  #answer the call
  session.answer();
  #play a file
  session.streamFile("long_prompt.mp3")
  # Test 1 - FAILED!
  digits = session.read(5, 10, "long_prompt.mp3", 3000, "#")
  # Test 2 - WORKED OK!
  #session.getDigits(1,"#",7000)
  #session.streamFile("long_prompt.mp3")
  # TEST 3 - WORKED OK!
  #digits = session.playAndGetDigits(5, 10, 1, 60, "#","long_prompt.mp3", "", "")
  #hangup
  session.hangup();

When I uncomment the code under test 1 and I make two simultaneous calls, the initial prompt plays for both calls just fine. But then the second prompt only plays on one of the channels and the other one just has dead air. When the first channel finishes playing the prompt, then the second channel starts playing it.
 
Then I re-comment test 1 and uncomment either test 2 or test 3, Both prompts play just fine for both channels.
 
So I think there may be a bug in the read() function somewhere.  I took a look at it, but it's way over my head.
 
Thanks again,
Adam
 
 
 

[anthony.minessale at g...]

easier than you think,

Just like Traun had suspected, the methods in the C++ wrapper for the read app was missing the begin and end allow threads calls
that are only important to python to tell it to suspend the thread state while FS specific code is being executed.

2 line fix in trunk rev 12958

On Wed, Apr 8, 2009 at 8:59 PM, Adam Wilt <wiltingtree@gmail.com (wiltingtree@gmail.com)> wrote:

[tleyden at branchcut.com]

Hey you beat me to it.  I was going to have a look this morning but
had no internet because some asswipe cut a bunch of fiber optic cables
and took out phone/internet for a big part of the bay area.

I haven't tried your patch yet, but I see something that looks suspect:

http://fisheye.freeswitch.org/browse/FreeSWITCH/src/switch_cpp.cpp?r=12958

In playAndGetDigits() there are now two calls to begin_allow_threads()
(line 778 and 780) followed by only one call to end_allow_threads() (line 793)

Also I guess it would have better to test against JS, since it should have
had the same bug right?  Lua just ignores the threadswapping stuff but
IIRC javascript uses it in much the same way as python.  Or did I miss something?

[mike at jerris.com]

On Apr 10, 2009, at 1:35 AM, Traun Leyden wrote:

[mike at jerris.com]

On Apr 10, 2009, at 1:35 AM, Traun Leyden wrote: