#7900 Issue with encoding of filenames in git snapshot zip
If a git repo has a filename with unicode characters, and you download a snapshot zip file, then unzip it the filename will probably be wrong.
Example: "Download Snapshot" link on https://sourceforge.net/p/compilibre/code/ci/master/tree/ and then "COMPILIBRE-source/COMPILIBRE/installer/Cliquez ici pour démarrer COMPILIBRE.bat" can end up as "Cliquez ici pour d+?marrer COMPILIBRE.bat" or "Cliquez ici pour démarrer COMPILIBRE.bat"
I wonder how easy this is to fix? Does it matter what your git settings are, or how the file was added to the repo? Or your local filesystem?
Original report: https://sourceforge.net/p/forge/feature-requests/353/
So you're saying this example will work some of the time?
This example works all the time. (Although extracting on OSX ends up ok, whereas Linux has wrong characters)
I used the word "probably" because I don't know what triggers the issue and it might be dependent on how the file and repo are set up.
Ok, that is strange. I asked because I extracted it on OSX and worked correctly.
I'm getting "Cliquez ici pour d├йmarrer COMPILIBRE.bat" on Linux
I had the same issue in linux using unizp -- but 7zip properly detected the unicode names.
We might be able to address it server side to some degree with something like:
zip archive dir -r -UN=UTF8Or possibly upgrade the version of zip on the server to 3.0+.
...zip 3.0, in addition to the standard file path, now includes the UTF-8 translation of the path if the entry path is not entirely 7-bit ASCII. When an entry is missing the Unicode path, zip reverts back to the standard file path.I think we create the zip files for git with
git archivenot thezipcommand. Quick googling shows a long thread about the topic: http://git.661346.n2.nabble.com/git-archive-format-zip-utf-8-issues-td7564798.html I didn't read it all, I'll let whoever takes this ticket figure out what the results are. It sounds like they made some improvements during the course of that thread so versions of git after that date might be better than others.And the example referenced (on sourceforge.net) is using git 1.7.4.1 which is fairly old.
Work on this ticket should try to reproduce in that version of git, and then see if it's better with a new version of git.
Looks like it does depend on both the version of git on the server, and the version of unzip on the client.
d├йmarrer.txtd+?marrer.txtdémarrer.txtLast edit: Dave Brondsema 2015-06-19
To use different versions of git, it can be useful to set the
GIT_PYTHON_GIT_EXECUTABLEenv var for the taskd process.d+¬marrer.txtd+¬marrer.txtd├⌐marrer.txtdémarrer.txtSo Git 2.1.3 still isn't sufficient with windows vista, but is better in many other cases.
Seems like this isn't an Allura issue, but rather one between the version of Git and the unzipping software used. Upgrading to current version of
giton the server will help.