Python内存泄漏？

Question

I am writing a python extension that seems to be leaking memory. I am trying to figure out the soure of the problem using valgrind.

However, it seems that python itself is leaking memory according to valgrind. Using the following simple script:

hello.py

  print "Hello World!"

and doing

> valgrind --tool=memcheck python ./hello.py

(...)
==7937== ERROR SUMMARY: 580 errors from 34 contexts (suppressed: 21 from 1)
==7937== malloc/free: in use at exit: 721,878 bytes in 190 blocks.
==7937== malloc/free: 2,436 allocs, 2,246 frees, 1,863,631 bytes allocated.
==7937== For counts of detected errors, rerun with: -v
==7937== Use --track-origins=yes to see where uninitialised values come from
==7937== searching for pointers to 190 not-freed blocks.
==7937== checked 965,952 bytes.
==7937== 
==7937== LEAK SUMMARY:
==7937==    definitely lost: 0 bytes in 0 blocks.
==7937==      possibly lost: 4,612 bytes in 13 blocks.
==7937==    still reachable: 717,266 bytes in 177 blocks.
==7937==         suppressed: 0 bytes in 0 blocks.
==7937== Rerun with --leak-check=full to see details of leaked memory.

Does anybody have an explanation for this strage behavior? Is the python interpreter really leaking memory?

What tool do python developers use to debug their memory leaks?

Answer 1

There's a whole README.valgrind in the Python sources that explains the various caveats trying to use Valgrind with Python:

http://svn.python.org/projects/python/trunk/Misc/README.valgrind

Python uses its own small-object allocation scheme on top of malloc,
called PyMalloc.

Valgrind may show some unexpected results when PyMalloc is used.
Starting with Python 2.3, PyMalloc is used by default.  You can disable
PyMalloc when configuring python by adding the --without-pymalloc option.
If you disable PyMalloc, most of the information in this document and
the supplied suppressions file will not be useful.  As discussed above,
disabling PyMalloc can catch more problems.

If you use valgrind on a default build of Python,  you will see
many errors like:

        ==6399== Use of uninitialised value of size 4
        ==6399== at 0x4A9BDE7E: PyObject_Free (obmalloc.c:711)
        ==6399== by 0x4A9B8198: dictresize (dictobject.c:477)

These are expected and not a problem. 

Answer 2

The leak is most likely coming from your own extension, not from Python. Large systems often exit with memory still allocated, simply because it isn't worth it to explicitly free it if the process is about to end anyway.