> So, an initial test suggests that a recent version of git-p4 at least doesn't fail in > the same way under python3, in the face of at least some of these encoding > issues. I don't know yet whether failures will occur in other places, nor > whether the not-failing behavior is better, worse or the same as I had under > python2, but it seems plausible that I won't be filing any test_expect_failure > tests after all, and will instead say "yay, python3 ftw!" That would be fabulous. I myself have a repository that has a variety of such issues. A common case is CP-1252 Smart Quote characters produced on Windows which are incompatible with UTF-8, without explicit conversion. However, a lot of these problems can be avoided by simply avoiding conversion to text in the first place. In many cases the incoming data doesn't need to be converted and can be passed around as binary. I'm slowly working toward this goal, and once this patch-set it merged I have a couple of other decoding patches in the pipeline. If you have any other failure cases, please do submit them as test cases, or bug reports at least. I would prefer the script to discard Python 2 support, but even if the consensus is to retain it, Python 3 forces us to address these issues which is a great step in the right direction. Joel
