Mailing List Archive: bug report: profiler

Hi,

when trying to run the profiler calibration explained in the python
documentation, I got the following error:

Python 1.5.1 (#8, Dec 18 1998, 09:42:46) [GCC 2.7.2.1] on linux2
Copyright 1991-1995 Stichting Mathematisch Centrum, Amsterdam
>>> import profile
>>> pr = profile.Profile()
>>> print pr.calibrate(100)
Traceback (innermost last):
File "<stdin>", line 1, in ?
File "/usr/local/lib/python1.5/profile.py", line 498, in calibrate
self.instrumented()
File "/usr/local/lib/python1.5/profile.py", line 515, in instrumented
self.profiler_simulation(a, a, a)
File "/usr/local/lib/python1.5/profile.py", line 520, in profiler_simulation
t = t[0] + t[1]
AttributeError: __getitem__

Ralph
--
------------------------------------------------------
Ralph Heinkel
European Molecular Biology Laboratory (EMBL)
Meyerhofstr. 1
69012 Heidelberg
Tel. +49 6221/387 529
eMail: heinkel@dummy.embl-heidelberg.de

Please remove the 'dummy' from my email address.

[Ralph Heinkel]
> when trying to run the profiler calibration explained in the python
> documentation, I got the following error:
>
> Python 1.5.1 (#8, Dec 18 1998, 09:42:46) [GCC 2.7.2.1] on linux2
> Copyright 1991-1995 Stichting Mathematisch Centrum, Amsterdam
> >>> import profile
> >>> pr = profile.Profile()
> >>> print pr.calibrate(100)
> Traceback (innermost last):
> File "<stdin>", line 1, in ?
> File "/usr/local/lib/python1.5/profile.py", line 498, in calibrate
> self.instrumented()
> File "/usr/local/lib/python1.5/profile.py", line 515, in instrumented
> self.profiler_simulation(a, a, a)
> File "/usr/local/lib/python1.5/profile.py", line 520, in
> profiler_simulation
> t = t[0] + t[1]
> AttributeError: __getitem__

Suggest you wait a day or two for 1.5.2 final to be released, or at least
extract a 1.5.2 profile.py; this was fixed quite some time ago (the profiler
uses one of several timing methods depending on the platform, but
profiler_simulation *assumed* a particular one was in use -- presumably the
one in use on the platform on which it was written <wink>).

BTW, calibration is a touchy process, and in my experience unless you always
run on a dedicated dead-quiet machine, the result of fiddling with it is
anomalies like negative reported runtimes. Profiling to get a feel for
where the *bulk* of the time is being spent works fine without bothering.

a-percent-or-two-here-or-there-may-vary-with-the-humdity-ly y'rs - tim