Non-ASCII filename characters in plain diff files appear garbled

Description

When downloading plain diff files from Code > Plain diff, filenames with non-ASCII characters (such as Chinese, Japanese, Korean) appear garbled or corrupted.

Environment

  • Merge requests to change files with non-ASCII characters

Impacted offerings:

  • GitLab.com
  • GitLab Dedicated
  • GitLab Self-Managed

Impacted versions:

  • All versions

Solution

GitLab Self-Managed Solution

  1. Configure gitlab.rb as follows.
gitaly['configuration'] = {
   git: {
     config: [
       { key: 'core.quotepath', value: 'false' }
     ]
   }
}
  1. Run sudo gitlab-ctl reconfigure

GitLab.com & GitLab Dedicated Solution

  1. Clone/fetch your project into your local repository
  2. Run git config core.quotepath false
  3. Run git diff origin/<source branch>..origin/<target branch>

Cause

By default, Git commands that output paths escapes control characters or bytes with values larger than 0x80.

Since almost all non-Latin alphabet characters have a code greater than 0x80, merge requests that include filenames with such characters will be escaped and can appear to be garbled in the plain diff file.

Related Links

On this page...