Check for Mach-O using byte strings to allow case of non unicode chars

[QuantumEntangledAndy]

The recent change of not encoding byte strings in bdist_mac is breaking the build. Methods using bytes should not be mixed with str. This patch unifies the relevant code to use all bytes.

NOTE: There is one step where I choose to decode

else:
                        files.append(name.decode())

This is because

filePath = os.path.join(self.binDir, fileName)

The fileName is generated from type str (from os.walk) but when we add extra files in the loop they were added as type bytes.

Currently I decode back to string then append to the list, but alternatively it may be better to encode the results of os.walk so that it is all bytes throughout.

[QuantumEntangledAndy]

I have a further patch that gets @loader_path working (see #634) BUT when I came to merge I had conflicts from your changes to not decode. When I pulled your changes I ran into multiple errors related to using str when bytes was expected.

If you want I can roll the addition of @loader_path into this pull request too. Or I can wait until this is merged and pull request of that.

[marcelotduarte]

That is my patch. It was untested and I forget to declare this. In the issue #529 I requested tests but no one answered. As you use mac, if you find it better to revert the patch and you fix the issue, I even prefer it.

[QuantumEntangledAndy]

I actually had #529 and had a different fix. Rather than not decoding I did this instead

try:
    line = p.stdout.readline().decode()
except UnicodeDecodeError:
    continue
if "Mach-O" in line:

Here I assume that if we can't decode it is not Mach-O. Which fixes #529 in my case I am not sure what the original case was.

We have two choice

1. resolve by changing all to bytes (many changes, but possibly more flexible)

2. resolve by skipping any files where it is not decodable (less changes, but possibly not as flexible).

Which would you prefer?

[marcelotduarte]

I reverted my patch. Please send your patch, your approach is ok to me, the option 2.

[QuantumEntangledAndy]

I have made this pull request work as indicated in option 2.

Fixes #529

[QuantumEntangledAndy]

Actually I am thinking of an option 3.

Use str everywhere except when testing for Mach-O this will let us test for Mach-O in files with non unicode bytes, while still not requiring many changes

[marcelotduarte]

ok, put it in another PR

[QuantumEntangledAndy]

A completely separate PR? Or ammend this one?

[marcelotduarte]

Can be in this PR, but change the title to not confuse anthony.

[QuantumEntangledAndy]

Latest commit is using option number 3. This allows searching for Mach-O even if there are non unicode chars but keeps the rest of the code as str