Demonstrating py2.6 shlex madness

[lrq3000]

On python 2.6, shlex fails miserably the unit test because it inserts null characters after each character of any command read from the Makefile:

>>  assert ("hello world" in u'[\'k\\x00e\\x00y\\x00b\\x00o\\x00a\\x00r\\x00d\\x
00m\\x00a\\x00s\\x00h\\x00i\\x00t\\x00a\\x00l\\x00l\\x00t\\x00o\\x00g\\x00e\\x00
t\\x00h\\x00e\\x00r\\x00t\\x00h\\x00i\\x00s\\x00s\\x00h\\x00o\\x00u\\x00l\\x00d\
\x00n\\x00o\\x00t\\x00r\\x00u\\x00n\\x00o\\x00t\\x00h\\x00e\\x00r\\x00w\\x00i\\x
00s\\x00e\\x00u\\x00n\\x00i\\x00t\\x00t\\x00e\\x00s\\x00t\\x00s\\x00w\\x00i\\x00
l\\x00l\\x00f\\x00a\\x00i\\x00l\']\r\nkeyboardmashitalltogetherthisshouldnotruno
therwiseunittestswillfail\r\nTraceback (most recent call last):\r\n  File "<stri
ng>", line 1, in <module>\r\n  File "pymake\\_main.py", line 56, in main\r\n
ignore_errors=opts[\'--ignore-errors\'])\r\n  File "pymake\\_pymake.py", line 16
7, in execute_makefile_commands\r\n    check_call(parsed_cmd)\r\n  File "c:\\pyt
hon26\\lib\\subprocess.py", line 483, in check_call\r\n    retcode = call(*popen
args, **kwargs)\r\n  File "c:\\python26\\lib\\subprocess.py", line 470, in call\
r\n    return Popen(*popenargs, **kwargs).wait()\r\n  File "c:\\python26\\lib\\s
ubprocess.py", line 623, in __init__\r\n    errread, errwrite)\r\n  File "c:\\py
thon26\\lib\\subprocess.py", line 833, in _execute_child\r\n    startupinfo)\r\n
TypeError: CreateProcess() argument 2 must be string without null bytes or None,
 not str\r\n')

If you look carefully, this:

'k\\x00e\\x00y\\x00b\\x00o\\x00a\\x00r\\x00d\\x
00m\\x00a\\x00s\\x00h\\x00i\\x00t\\x00a\\x00l\\x00l\\x00t\\x00o\\x00g\\x00e\\x00
t\\x00h\\x00e\\x00r\\x00t\\x00h\\x00i\\x00s\\x00s\\x00h\\x00o\\x00u\\x00l\\x00d\
\x00n\\x00o\\x00t\\x00r\\x00u\\x00n\\x00o\\x00t\\x00h\\x00e\\x00r\\x00w\\x00i\\x
00s\\x00e\\x00u\\x00n\\x00i\\x00t\\x00t\\x00e\\x00s\\x00t\\x00s\\x00w\\x00i\\x00
l\\x00l\\x00f\\x00a\\x00i\\x00l\'

Is in fact the same as:

keyboardmashitalltogetherthisshouldnotruno therwiseunittestswillfail

But with null characters between each character.

I tried everything I could, but I could not overcome this issue. It does not happen on newer versions of Python.

See this issue for more details. A possible alternative here (but it does not support comments yet) and also interesting reads here and here.

[coveralls]

Coverage increased (+0.2%) to 95.935% when pulling f940f35 on py2-shlex-madness into e7cbe64 on master.

[codecov-io]

Current coverage is 93.49% (diff: 78.94%)

Merging #3 into master will decrease coverage by 0.57%

@@             master         #3   diff @@
==========================================
  Files             6          6          
  Lines           118        123     +5   
  Methods           0          0          
  Messages          0          0          
  Branches         25         24     -1   
==========================================
+ Hits            111        115     +4   
- Misses            4          5     +1   
  Partials          3          3          

Powered by Codecov. Last update b882682...f940f35