Process.get_connections() is broken on big-endian architectures

[giampaolo]

From matrixhasu on August 28, 2011 16:57:01

Hi,
followuing up issue 200 , I gave a look at what's happening on sparc about:

FAIL: test_get_connections (__main__.TestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "test/test_psutil.py", line 147, in inner
    return fun(self, *args, **kwargs)
  File "test/test_psutil.py", line 974, in test_get_connections
    self.assertEqual(ip, '127.0.0.1')
AssertionError: '1.0.0.127' != '127.0.0.1'

The problem (that I think it's related to the endianess of the architecture) is 
that _decode_address() expect 127.0.0.1 in a line like this

   2: 0100007F:1DD2 00000000:0000 0A 00000000:00000000 00:00000000 00000000     
0        0 6797 1 ffff880115776040 300 0 0 2 -1                      

while on sparc you'd get something like:

   2: 7F000001:A13F 00000000:0000 0A 00000000:00000000 00:00000000 00000000  
2940        0 41019895 1 fffff8103c59be80 300 0 0 2 -1                  

so the second field (the local address) is in reverse.

The different ordering might come up even on other part of the code.

If you need any help in debugging further, don't hesitate to contact me.

Cheers,
Sandro

Original issue: http://code.google.com/p/psutil/issues/detail?id=201

[giampaolo]

From matrixhasu on August 28, 2011 15:55:00

I can see an error like this even on: mips (but not on mipsel, which is little 
endian, so it's definitely alignment-related), powerpc and s390 .

[giampaolo]

From g.rodola on August 29, 2011 03:49:25

Can you please check whether this patch works (note: I still haven't took care of IPv6)?

Attachment: endianess.patch

[giampaolo]

From matrixhasu on August 29, 2011 10:54:59

Yep, it works fine (I tested both on a little and big endian box)

[giampaolo]

From g.rodola on August 29, 2011 13:06:03

This should now be fixed in r1113 and r1114 .
I've also added a brand new IPv6-related test so please re-run tests to make 
sure everything is fine.

Summary: Process.get_connections() is broken on big-endian architectures
Status: Started
Labels: OpSys-Linux Milestone-0.3.1

[giampaolo]

From matrixhasu on August 29, 2011 13:27:59

It works fine either on sparc and on amd64 (so 2 different endian machines). 
Ah, since you're giving me credit (thanks about that!) you might want to refer 
to my real name "Sandro Tosi" (for the email you can use sandro.tosi@gmail.com 
or morph@debian.org).

[giampaolo]

From g.rodola on August 29, 2011 13:31:43

Oh yeah, we met at Euro Python in Florence this very year. =)
Fixed in r1115 .

Status: FixedInSVN

[giampaolo]

From g.rodola on October 21, 2011 16:44:17

Labels: -Milestone-0.3.1

[giampaolo]

From g.rodola on October 21, 2011 16:45:27

Labels: Milestone-0.4.0

[giampaolo]

From g.rodola on October 28, 2011 20:44:14

Status: Fixed

[giampaolo]

From g.rodola on March 02, 2013 04:03:06

Updated csets after the SVN -> Mercurial migration: r1113 == revision 
e00994c657c0 r1115 == revision 3671d370e764