[Python-bugs-list] [ python-Bugs-820605 ] tkinter's 'after' and 'threads' on multiprocessor

[SourceForge.net]

Bugs item #820605, was opened at 2003-10-09 15:15
Message generated for change (Settings changed) made by zoranbosnjak
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=105470&aid=820605&group_id=5470

Category: Tkinter
Group: Platform-specific
Status: Open
>Resolution: Fixed
Priority: 5
Submitted By: Zoran Bosnjak (zoranbosnjak)
Assigned to: Martin v. Löwis (loewis)
Summary: tkinter's 'after' and 'threads' on multiprocessor

Initial Comment:
I have attached simple testfile which works fine on
single processor machine, but it does not work on
multiprocessor machine (dual XEON - HT 2x 2.4GHz,
python 2.1.1 (or python 2.2), RedHat 9.0 distr.,
linux-kernel 2.4.20-20-9smp).
At the moment I can not tell where the problem is. It
could be 
- in python core,
- in threading module
- in tkinter, TCL, Tk
- in linux kernel
Problem description:
- I use tkinter's (Tk's) after method for periodic calls
- I also start a few dummy threads which have nothong
to do with tkinter.
- (see the attached file)
- After a while, one or more threads stop running and
'after' method stops working (they simply block). The
situation is different each time I run the script. The
problem is random and probability raises with number of
threads. Normaly the problem occours after a few
seconds with 8 threads.
I am willing to run additional tests if someone has any
idea where to start debugging.
regards,
Zoran Bosnjak


----------------------------------------------------------------------

Comment By: Zoran Bosnjak (zoranbosnjak)
Date: 2003-10-14 15:29

Message:
Logged In: YES 
user_id=583469

I couldn't reproduce this bug under python2.3.2 any more
(with the same testprogram). 


----------------------------------------------------------------------

Comment By: Drew Perttula (drewp)
Date: 2003-10-13 21:28

Message:
Logged In: YES 
user_id=127598

I forgot to show the after loop's output, which appears in
random
positions among the other threads' output. The more complete
version of the output I see looks like this:

3 1066073253.53
1066073253.53
1066073253.53
7 4 1066073253.53
5 1066073253.53
1 1066073253.53
6 1066073253.53
2 1066073253.58
8 1066073253.85
1066073254.03
1 1066073254.03
2 1066073254.03
5 1066073254.03

So, no bug confirmation yet.

----------------------------------------------------------------------

Comment By: Drew Perttula (drewp)
Date: 2003-10-13 09:59

Message:
Logged In: YES 
user_id=127598

The given test program works for me on a dual AMD 1400MHz with
py2.2.2 and linux 2.4.19. By 'works', I mean that after
60sec, the
program is still outputting spurts like this:

8 1066031758.16
5 1066031758.16
1 1066031758.16
7 1066031758.16
3 1066031758.16
6 1066031758.16
4 1066031758.16
2 1066031758.21

I am also available for running additional tests. My email is
drewp at big asterisk .com less the spaces.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=105470&aid=820605&group_id=5470