Hi, I am trying to import random on a power pc based machine and I see this exception. Could you please help me?

root@host# python
Python 2.7.3 (default, Apr  3 2016, 22:31:30)
[GCC 4.8.1] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import random
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python2.7/random.py", line 58, in <module>
    NV_MAGICCONST = 4 * _exp(-0.5)/_sqrt(2.0)
ValueError: math domain error
>>>

That's odd.

Try:
   import math
   print math.sqrt(2.0)
   print math

Post what you see.

The output of `struct.pack('<d', 2.0)` would also be worth knowing about. I'm wondering whether there's some kind of endianness issue going on in the float-to-string conversions.

Hi, I apologize. I didn't expect a quick reply.
Here are the outputs you requested.

root@host:~# python
Python 2.7.3 (default, Apr  3 2016, 22:31:30)
[GCC 4.8.1] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import math
>>> print math.sqrt(2.0)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: math domain error
>>> print math
<module 'math' from '/usr/lib/python2.7/lib-dynload/math.so'>
>>> import struct
>>> struct.pack('<d', 2.0)
'\x00\x00\x00\x00\x00\x00\x00@'
>>>

>>> print math.sqrt(2.0)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: math domain error

I'm surprised that such basic math function fails.

Can you please try to compile and run attached C program?

$ gcc sqrt.c -o sqrt -lm && ./sqrt
sqrt(2) = 1.41421

stinner victor, my host doesn't have a gcc compiler. is there a way you can give me the binary?

my host details:

Linux fpc0 3.10.62-ltsi-WR6.0.0.18_standard #1 SMP PREEMPT Sun Apr 3 23:17:02 PDT 2016 ppc64 ppc64 ppc64 GNU/Linux

What is your Linux distribution and what is your CPU? (try to read /proc/cpuinfo)

You cannot install a C compiler?

It's windriver linux. Processor is PPC64-E5500

What is the output of `struct.pack('d', 2.0)` and `struct.pack('d', float(2))`?

Serhiy,

root@host:~# python
Python 2.7.3 (default, Apr  3 2016, 22:31:30)
[GCC 4.8.1] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import struct
>>> struct.pack('d', 2.0)
'@\x00\x00\x00\x00\x00\x00\x00'
>>> struct.pack('d', float(2))
'@\x00\x00\x00\x00\x00\x00\x00'
>>>

What is the output of `python -m sysconfig`?

Python 2.7.3 is too old. May be this issue was already fixed in newer versions. Can you install 2.7.11?

There is no output for `python -m sysconfig`

What is the output of `import sysconfig; print sysconfig.get_config_vars()`?

Hi, This is my work system and I don't know if I am allowed to send out the complete output of "print sysconfig.get_config_vars()". could you please let me know which fields you are looking for?

Fields with ENDIAN or IEEE in key.

The keys ENDIAN and IEEE are not available.

I guess that Serhiy is looking for these variables defined in pyconfig.h:
---
/* Define if C doubles are 64-bit IEEE 754 binary format, stored in ARM
   mixed-endian order (byte order 45670123) */
#undef DOUBLE_IS_ARM_MIXED_ENDIAN_IEEE754

/* Define if C doubles are 64-bit IEEE 754 binary format, stored with the most
   significant byte first */
#undef DOUBLE_IS_BIG_ENDIAN_IEEE754

/* Define if C doubles are 64-bit IEEE 754 binary format, stored with the
   least significant byte first */
#undef DOUBLE_IS_LITTLE_ENDIAN_IEEE754
---

On Fedora, pyconfig.h is installed into /usr/include/python2.7/pyconfig.h. This file doesn't contain any sensitive information, it is the configuration used to build Python.

On Fedora, this file only dispatches between pyconfig-32.h and pyconfig-64.h:
---
$ cat /usr/include/python2.7/pyconfig.h 
#include <bits/wordsize.h>

#if __WORDSIZE == 32
#include "pyconfig-32.h"
#elif __WORDSIZE == 64
#include "pyconfig-64.h"
#else
#error "Unknown word size"
#endif
---

My system is 64-bit, so I can use pyconfig-64.h:
---
$ grep IEEE /usr/include/python2.7/pyconfig-64.h 
/* Define if C doubles are 64-bit IEEE 754 binary format, stored in ARM
/* #undef DOUBLE_IS_ARM_MIXED_ENDIAN_IEEE754 */
/* Define if C doubles are 64-bit IEEE 754 binary format, stored with the most
/* #undef DOUBLE_IS_BIG_ENDIAN_IEEE754 */
/* Define if C doubles are 64-bit IEEE 754 binary format, stored with the
#define DOUBLE_IS_LITTLE_ENDIAN_IEEE754 1
/* #undef HAVE_IEEEFP_H */
/* #undef HAVE_LIBIEEE */
/* Define to activate features from IEEE Stds 1003.1-2001 */
---

Here is one output:

root@fpc0:~# grep IEEE /usr/include/python2.7/pyconfig-32.h
/* Define if C doubles are 64-bit IEEE 754 binary format, stored in ARM
/* #undef DOUBLE_IS_ARM_MIXED_ENDIAN_IEEE754 */
/* Define if C doubles are 64-bit IEEE 754 binary format, stored with the most
/* #undef DOUBLE_IS_BIG_ENDIAN_IEEE754 */
/* Define if C doubles are 64-bit IEEE 754 binary format, stored with the
/* #undef DOUBLE_IS_LITTLE_ENDIAN_IEEE754 */
/* #undef HAVE_IEEEFP_H */
/* #undef HAVE_LIBIEEE */
/* Define to activate features from IEEE Stds 1003.1-2001 */
root@fpc0:~#

Couple more outputs:

----------------
/* Define WORDS_BIGENDIAN to 1 if your processor stores words with the most
   significant byte first (like Motorola and SPARC, unlike Intel). */
#if defined AC_APPLE_UNIVERSAL_BUILD
# if defined __BIG_ENDIAN__
#  define WORDS_BIGENDIAN 1
# endif
#else
# ifndef WORDS_BIGENDIAN
#  define WORDS_BIGENDIAN 1
# endif
#endif
----------------

Math works in some cases.
----------------
>>> import math
>>> math.sqrt(2)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: math domain error
>>> math.floor(2.5)
2.0
>>>
----------------


Also, could you please let me know the train of thought and ask for all the outputs as much as possible in one go? because I feel like I am delaying your debugging.

@stinner, my host doesn't have internet connection and it doesn't have gcc installed.

I don't know how to install gcc in windriverlinux.
In famous linux distributions like ubuntu and centos, i install gcc using 'yum install gcc' or 'apt-get install gcc' commands.

Hey Guys/Gals, are you still debugging the issue? Do you need more info?

An option would be to use the support of the probably commercial

  http://www.windriver.com/products/linux/

and ask them why they're still on Python 2.7.3.  They'll also
have gcc installed and can run the test case proposed in this
issue.

Also, I don't understand how the test suite could pass with
these kinds of errors. Are you sure this is the Wind-River
system Python?  Did they run the test suite?

Yes, it is a WindRiver linux. I will check if Commercial WindRiver team can help me.

Host details:
Linux fpc0 3.10.62-ltsi-WR6.0.0.18_standard #1 SMP PREEMPT Sun Apr 3 23:17:02 PDT 2016 ppc64 ppc64 ppc64 GNU/Linux



Also is there a protocol followed in this website? If I have done something wrong it's because I didn't know I was doing it incorrectly. I ask because, a new person replies everyday and they don't even completely tell me the solution or what conclusions they have made. I am lost.

My conclusion is:  The error in msg263292 should have been
caught by the test suite.

I'd ask the WindRiver support if they actually run the test suite
before shipping the binaries.

Thank you skrah. I raised a support request from WR. I will keep my fingers crossed.

Since there hasn't been activity on this for nearly 2 years, should it be closed as third party?  Stefan seemed to conclude that it was a WindRiver issue.  Thanks!

@csabella: Agreed. Let's close. Thanks!

stage: resolved