{} cannot be recreated as a directory: something else is the

Error message

{} cannot be recreated as a directory: something else is there now; nothing more was changed

What it means

Replay records empty directories that must be recreated (restore_dirs). Before mkdir, it verifies the path is not occupied by something else; if a non-directory file or symlink now sits at that path, it aborts to avoid destroying data.

Source

Thrown at src/system/history/replay.rs:690

                }
            }
        });
        let pending =
            journal::begin_changes("history", &display_path(&step.path), [step.path.clone()])?;
        match &step.action {
            Action::Write { mode, oid } => write_object(repo, step, mode, oid)?,
            Action::Delete => remove(&step.path)?,
            _ => unreachable!("filtered to mutating steps"),
        }
        journal::commit_changes(pending);
        touched.push(step.path.clone());
    }
    for dir in &exec.restore_dirs {
        if dir.is_dir() && !dir.is_symlink() {
            continue;
        }
        if dir.exists() || dir.is_symlink() {
            bail!(
                "{} cannot be recreated as a directory: something else is there now; nothing more was changed",
                display_path(dir)
            );
        }
        if let Some(parent) = dir.parent() {
            file::create_dir_all(parent)?;
        }
        let bits = exec.restore_modes.get(&display_path(dir));
        let builder = std::fs::DirBuilder::new();
        #[cfg(unix)]
        let builder = {
            let mut builder = builder;
            if let Some(bits) = bits {
                use std::os::unix::fs::DirBuilderExt;
                builder.mode(*bits);
            }
            builder
        };

View on GitHub (pinned to afd2eddd3a)

Solutions

Inspect the path named in the message; move/delete the occupying file or symlink, then re-run
Re-plan against current state so the restore_dirs list reflects reality
If the symlink is intentional, exclude that directory from the replay

Example fix

// before
$ rm -rf build && ln -s /shared/build build   # then replay fails
// after
$ unlink build && mkdir build                 # then re-run replay

Defensive patterns

Strategy: validation

Validate before calling

// preflight each restore dir
for dir in &restore_dirs {
    if dir.exists() && !(dir.is_dir() && !dir.is_symlink()) {
        eprintln!("clear the path first: {}", dir.display());
    }
}

Prevention

Don't replace directories with files or symlinks after checkpoints
Remove placeholders (files where empty dirs were) before replaying
Verify symlink placement in synced/mounted directories

When it happens

Trigger: For dir in exec.restore_dirs: dir exists or is a symlink but is not a real directory at restore time.

Common situations: A file was created where an empty directory used to be after the checkpoint; a symlink now occupies the recorded directory path; sync tools replaced directories with symlinks.

Understand the failure class

Background: "already exists" / EEXIST / FileAlreadyExistsException: what the 'file already exists' error means and how to fix it — this error's family across 37 libraries.

Related errors

AI-assisted analysis of jdx/mise@afd2eddd3a (2026-09-09). Data as JSON: /api/errors/95c0f19dc46d17a1. Report an issue: GitHub.