Potential fix for 'exec' failure case

[anwilli5]

• Adding more robust code to handle the case where the exec call within fails. Now, spawn will raise an exception if this happens.
• Adding a test to ensure this exception is raised when an invalid binary is run

[takluyver]

I don't think you need to use the low-level os.open interface here - you should just be able to open a regular file object in binary mode - with open(fullpath, 'wb') as f:

[takluyver]

Thanks for this, and especially for writing a test to go with it.

[takluyver]

Would it make more sense to pull it out with a regex or to send it again as, say, the first few bytes sent over the pipe?

I'd prefer to send it separately as the first few bytes, that way we're not relying on the way Python formats error messages. We're already using the struct module, so doing struct.pack('i', e.errno) + errbytes should work.

[anwilli5]

Thanks for the feedback - my python dev experience isn't too extensive, so I'm still learning the best ways of doing things. :)

[takluyver]

Theoretically, some infintesimal fraction of the time, this could accidentally generate a valid executable and do something random on your computer. I expect that's massively unlikely, but I might hardcode some nonsense bytes so it's always the same.

[takluyver]

You're welcome. There's just a couple more little things I spotted with the test, then I think this is good to go.

[takluyver]

Thanks. I think b'\x00\x00' should work on both Python versions.

[takluyver]

(It wouldn't work on Python 2.5 or older, but I'm not interested in supporting that)

[takluyver]

Thanks! Merging this now.

[takluyver]

Hmmm, tests are failing on my machine, with AssertionError: OSError was not raised. Is it working for you?

[anwilli5]

Weird... Yes, the tests pass for me:
platform linux -- Python 3.4.0 -- py-1.4.25 -- pytest-2.6.3
platform linux2 -- Python 2.7.6 -- py-1.4.25 -- pytest-2.6.3

Just as a sanity check, is the latest working directory in your PYTHONPATH?

[takluyver]

Ah, you're right, it was testing against an older version. D'oh. Normally py.test automatically adds the working directory to sys.path - I don't know why it's not at the moment.

[takluyver]

Got it: if the tests are a package, i.e. with an __init__.py, py.test adds the parent directory to sys.path. If they're just a bunch of Python files in a directory, it doesn't.