I managed to fix PY3 TCs with f801a3a, but unfortunately, PY2 encodings for filepaths is a "casino"; surprisingly, Windows had no problem! Cannot reproduce this under Windows, win-git refused to start with unicode TMPDIR vars.

@@yarikoptic can you live by that alone?

Both smmap and gitdb projects fails PY2 with a unicode tmp-dir :-(~~
[edit:] False alarm for smmap; PY2 was failing when pip-installing codecov; pip is not happy with symlinked unicode temp-folders...

Dear @ankostis - if you are still interested in my opinion/desires -- I just want ideally both PY2 and PY3 supported. ATM we do not actively support Windows (do Linux and OSX), so those platforms are of interest to me

Of course!
Why do you say so?
[edit:] I run TCs also on Linux. But I cannot fix it.
My comment about Windows, was for-the-record, nothing to do with my efforts to fix Linux-PY2.

Saying just because you asked and I didn't reply ? In time so thought that may be you fixed it all and no feedback was needed any longer ;-)

Anyhow, I've managed to fix PY3 on linux, please do try and fix the issue also in PY2; commence from f801a3a, as explained above.

I ahead stashed my laptop away for a transatlantic flight to commence in few minutes... Will try when across the pond and finish driving 300 miles north unless someone beats me to it ;-)

@yarikoptic I do hope you have landed safely :)! I would also be looking forward to your results whenever you have some time.

thanks for the buzz @Byron ! Yes, I have landed successfully ;) and today seems to be the day to respond to my own whinings on github (trying to fixup pandas as well atm).

Could you please clarify where that commit/branch you pointed to lives at since I can't find it on a quick look :-/

$> git log | grep surrogate
    fix(surrogateescape): enable on py2, fix tests
    fix(unicode): use surrogateescape in bytes.decode
changes on filesystem:                                                           
 git/ext/gitdb | 2 +-

$> git show f801a3aed7ab28b1c80fc1b573afb2b182eca995
fatal: bad object f801a3aed7ab28b1c80fc1b573afb2b182eca995

$> git show | head
commit b0c187229cea1eb3f395e7e71f636b97982205ed
Author: Sebastian Thiel <[email protected]>
Date:   Thu Dec 8 16:07:11 2016 +0100

meanwhile I will merge current master, resolve conflict and push...

Codecov Report

Merging #543 into master will decrease coverage by 8.3%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #543      +/-   ##
==========================================
- Coverage   94.66%   86.36%   -8.31%     
==========================================
  Files          60       60              
  Lines        9340     9338       -2     
==========================================
- Hits         8842     8065     -777     
- Misses        498     1273     +775

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f14a596...aace290. Read the comment docs.

@yarikoptic Actually I don't know where said commit is, maybe @ankostis can help finding it. So far this issue seemed like a conversation between you to, I just wanted to reignite it after the transatlantic flight.

I have f801a3a in my repo: https://github.com/ankostis/GitPython/commits/fix_py2_enc

Is there something we should do with this PR?

I've managed to address the problem described only for python-3, as explained in this comment.

@ankostis Thanks again for your contribution, and sorry for the great delay (once again). It appears it is still having trouble to deal with this :/.

pushed with current master merged in -- may be some things are already better ;-)

@yarikoptic This one will probably have to wait for another release. The current one should already make things better though :).

sure! I will merge current improved master again to see where we stand though
Cheers