Please describe the problem.

In git-annex version 10.20250721, certain non-Latin filenames, specifically those with Cyrillic characters, fail to be added, unlocked, or adjusted in repositories. The issue affects a range of filename patterns, including simple Cyrillic names, names with numbers, dashes, spaces, or special characters, and files with various extensions. This problem appears to be a regression in this version, as the same repository works perfectly with git-annex version 10.20220121.

What steps will reproduce the problem?

1. Create a new git repository and initialize git-annex:

   git init
   git annex init

1. Create test files with different Cyrillic filename patterns (both working and failing examples):

   echo "test" > "ИА_2222.07.xlsx"    # 2-char Cyrillic prefix - WORKS
   echo "test" > "ЦППП_202206.xlsx"    # no dot in date - WORKS
   echo "test" > "ААА_55.22.xlsx"      # different date format - WORKS
   echo "test" > "ЦППП_2022.06.xlsx"   # 4-char prefix + YYYY.MM - FAILS
   echo "test" > "ИАИА_2222.07.xlsx"   # 4-char prefix + YYYY.MM - FAILS

1. Add the files:

   git annex add *

1. You will see that some files are successfully added, while others fail with the error:

   git-annex: .git/annex/othertmp/.0: createSymbolicLink: already exists (File exists) failed

1. Additionally, in existing repos, attempts to unlock or adjust in failed files will show errors like:

   git-annex: ../.git/annex/othertmp/.22/SHA256E-s...: removeDirectoryRecursive: permission denied (Permission denied) failed

What version of git-annex are you using? On what operating system?

• git-annex version: 10.20250721 (broken)
• OS: Manjaro Linux (ext4 filesystem)
• git config: core.quotepath=false
• Note: The issue does not occur in git-annex version 10.20220121 (tested on WSL Ubuntu).

Please provide any additional information below.

• Problematic Filename Examples:

  • "ЦППП_2022.06.xlsx" (4-char Cyrillic prefix with YYYY.MM date format) — fails
  • "ИАИА_2222.07.xlsx" (4-char Cyrillic prefix with YYYY.MM date format) — fails
  • "ДПК_2021.06-2.xlsx" (Cyrillic prefix with number and dash) — fails
  • "ВУП Авто .pptx" (Cyrillic with spaces) — fails
  • "Ачох_кейс.dat" (Cyrillic with underscore and special characters) — fails

• Even Simple Non-Latin names:

  • пожелания.md — fails
  • обучение.xlsx — fails
  • Протокол.xlsx — fails
  • Согласие.docx — fails
  • Грейдинг.pptx — fails

• Working Examples:

  • "ИА_2222.07.xlsx" (2-char Cyrillic prefix)
  • "ЦППП_202206.xlsx" (no dot in date)
  • "ААА_55.22.xlsx" (different date format)
  • Latin-only filenames such as "IOIO_2222.07.xlsx" also work fine.

• Debug Output shows escaped Cyrillic sequences:

  git annex --debug whereis "ЦППП_2022.06.xlsx" 2>&1 | grep ls-files
  git [...] ls-files [...] "\1062\1055\1055\1055_2022.06.xlsx"

• Workaround: Renaming the problematic file by adding a special character or changing the filename slightly (e.g., using an em-dash or a different date separator) resolves the issue:

  mv "ЦППП_2022.06.xlsx" "ЦППП_2022.06—.xlsx"  # Add em-dash
  git annex add "ЦППП_2022.06—.xlsx"  # This works

• Possible Root Cause: May be the temp filename generation algorithm in git-annex appears to have conflicts when processing escaped Cyrillic sequences (e.g., \1062\1055\1055\1055) in filenames that have 4+ character Cyrillic prefixes and a YYYY.MM date format. This causes temp filenames like "ЦП{PID}-{counter}" to conflict with existing operations.

Have you had any luck using git-annex before?

Yes, git-annex has been fantastic for managing large datasets across multiple machines, and the same repository works perfectly with an older version (10.20220121) on Ubuntu WSL. However, this issue with non-Latin filenames is a regression in the newer version. Despite this, git-annex remains an invaluable tool for distributed file management.

This issue appears to affect all Cyrillic filenames, not just the initially identified patterns, making the current version of git-annex barely usable for repositories containing non-Latin filenames.

fixed --Joey