git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH v11] setup: improve error diagnosis for invalid .git files

From
Junio C Hamano <gitster@pobox.com>
Date
Feb 26, 2026, 23:03 UTC
Message-ID
<xmqqpl5rumy0.fsf@gitster.g>
In-Reply-To
<20260223074410.917523-1-a3205153416@gmail.com>
Tian Yuchen <a3205153416@gmail.com> writes:
Show 32 quoted lines
> 'read_gitfile_gently()' treats any non-regular file as
> 'READ_GITFILE_ERR_NOT_A_FILE' and fails to discern between 'ENOENT'
> and other stat failures. This flawed error reporting is noted by two
> 'NEEDSWORK' comments.
>
> Address these comments by introducing two new error codes:
> 'READ_GITFILE_ERR_MISSING'(which groups the "file missing" scenarios
> together) and 'READ_GITFILE_ERR_IS_A_DIR'.
>
> To preserve the original intent of the setup process:
> 1. Update 'read_gitfile_error_die()' to treat both 'IS_A_DIR' and
>    'MISSING' as no-ops, while continuing to call 'die()' on true
>    'NOT_A_FILE' errors to prevent security hazards (like FIFOs).
> 2. Unconditionally pass '&error_code' to 'read_gitfile_gently()'.
> 3. Only invoke 'is_git_directory()' when we explicitly receive
>    'READ_GITFILE_ERR_IS_A_DIR', avoiding redundant filesystem checks.
> 4. Correctly return 'GIT_DIR_INVALID_GITFILE' on unrecognized errors
>    when 'die_on_error' is false.
>
> Additionally, audit external callers of 'read_gitfile_gently()' in
> 'submodule.c' and 'worktree.c' to accommodate the refined error codes.
>
> Signed-off-by: Tian Yuchen <a3205153416@gmail.com>
> ---
>  setup.c                       | 45 ++++++++++++++------
>  setup.h                       |  2 +
>  submodule.c                   |  2 +-
>  t/meson.build                 |  1 +
>  t/t0009-git-dir-validation.sh | 77 +++++++++++++++++++++++++++++++++++
>  worktree.c                    |  6 ++-
>  6 files changed, 118 insertions(+), 15 deletions(-)
>  create mode 100755 t/t0009-git-dir-validation.sh

Unfortunately this seems to break almost all the tests, not just the test the patch adds, on Windows (which I almost know nothing about, but I can observe that CI jobs die).

https://github.com/git/git/actions/runs/22464017037 is a CI run that merged this patch on top of the commit that corresponds to the tip of 'next' as of today. We can see "win test (N)" jobs dying all over. I cancelled the workflow before seeing everything die, though.

https://github.com/git/git/actions/runs/22464479533 is a CI run that tests this patch applied directly on v2.53.0 in isolation.

As I said, I do not know Windows well, so this may be a red-herring, but in this CI run, we see "GIT_DIR=/dev/null git diff --no-index ..." results in "fatal: error reading 'nul'":

  https://github.com/git/git/actions/runs/22464479533/job/65067515458#step:5:95419

which is an expected thing to happen, but we probably used to ignore it as a non-error?

For now, I'll kick this topic out of my tree to give other topics a bit more test exposure so that we can notice new bugs in them (not in this topic) that causes the tests fail. With this topic in 'seen', such bugs in other topics are all masked.

Can folks equipped with knowledge and environment to debug Windows only breakage lend a hand to figure out what this patch gets wrong to help it move forward?

Thanks.
Previous: Tian YuchenNext: Tian Yuchen
Message 30 of 45 in “setup: allow cwd/.git to be a symlink to a directory”
  1. 0/2 setup: allow cwd/.git to be a symlink to a directoryTian Yuchen, Feb 18, 2026
  2. 1/2 setup: distinguish ENOENT from other stat errorsTian Yuchen, Feb 18, 2026
  3. 2/2 setup: allow cwd/.git to be a symlink to a directoryTian Yuchen, Feb 18, 2026
  4. setup: allow cwd/.git to be a symlink to a directoryTian Yuchen, Feb 19, 2026
  5. Junio C HamanoFeb 20, 2026
  6. Tian YuchenFeb 20, 2026
  7. setup: allow cwd/.git to be a symlink to a directoryTian Yuchen, Feb 20, 2026
  8. Junio C HamanoFeb 20, 2026
  9. Tian YuchenFeb 21, 2026
  10. Junio C HamanoFeb 21, 2026
  11. Tian YuchenFeb 22, 2026
  12. setup: improve error diagnosis for invalid .git filesTian Yuchen, Feb 21, 2026
  13. Junio C HamanoFeb 22, 2026
  14. Tian YuchenFeb 22, 2026
  15. setup: improve error diagnosis for invalid .git filesTian Yuchen, Feb 22, 2026
  16. Karthik NayakFeb 22, 2026
  17. Tian YuchenFeb 23, 2026
  18. Junio C HamanoFeb 22, 2026
  19. Junio C HamanoFeb 23, 2026
  20. Tian YuchenFeb 23, 2026
  21. Junio C HamanoFeb 23, 2026
  22. Junio C HamanoFeb 23, 2026
  23. Tian YuchenFeb 23, 2026
  24. Junio C HamanoFeb 23, 2026
  25. Tian YuchenFeb 24, 2026
  26. Tian YuchenFeb 24, 2026
  27. Junio C HamanoFeb 25, 2026
  28. Tian YuchenFeb 25, 2026
  29. setup: improve error diagnosis for invalid .git filesTian Yuchen, Feb 23, 2026
  30. Junio C HamanoFeb 26, 2026
  31. Tian YuchenFeb 27, 2026
  32. Junio C HamanoFeb 27, 2026
  33. Tian YuchenFeb 28, 2026
  34. Junio C HamanoMar 2, 2026
  35. Phillip WoodMar 3, 2026
  36. Junio C HamanoMar 4, 2026
  37. Tian YuchenMar 4, 2026
  38. Junio C HamanoMar 4, 2026
  39. Tian YuchenMar 4, 2026
  40. Junio C HamanoMar 4, 2026
  41. Tian YuchenMar 4, 2026
  42. Junio C HamanoMar 4, 2026
  43. Tian YuchenMar 5, 2026
  44. Junio C HamanoMar 9, 2026
  45. setup: improve error diagnosis for invalid .git filesTian Yuchen, Mar 4, 2026

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.