Py3: Fix ToASCII method

[MartinBasti]

in Py2 to_text method returns Py2 non-unicode string, but in Py3 to_text method
returns Py3 default (unicode) string. So only in Py2 we have to decode
str to unicode.

https://fedorahosted.org/freeipa/ticket/5935

Edit: ticket number

[tiran]

On Python 3 the ToASCII method can theoretically return non-ASCII text.

if six.PY2:
    return res.decode('ascii') # must be unicode string in Python 2
else:
    res.encode('ascii')  # ensure it's only ASCII
    return res

[MartinBasti]

but to_text() returns punycoded values, escaped values as octal numbers, and ascii 32<characters<=127. So I don't think that string can contain something else than ASCII characters < 128

[MartinBasti]

but anyway I will add there the check you mentioned

[tiran]

I left a comment

[HonzaCholasta]

Don't bump BuildRequires unless it is actually required for the build. Doing so does not add any value, only the chore of having to maintain additional packages in our build roots for no reason.

[MartinBasti]

But this is conditional just for lint checking, IMO we may need to test with lint the same packages as are in Requires

[HonzaCholasta]

This does not guarantee that the package checked during build will be the same as the package used in runtime in any way. But given that an API we use was changed in 1.15, the bump is in fact required.

[MartinBasti]

I reworked commit:

• keep python-dns 1.15 in build requires at it changed return type of to_text function
• removed .encode() check

[tiran]

I have kicked Travis. Let's see if the test is passing now.

[MartinBasti]

bump for review

[MartinBasti]

Fixed upstream
master:
https://fedorahosted.org/freeipa/changeset/35ba724de90b270773d91596de310291df745df0