ff log¶

Show commit history, with the open change @ above the recorded commits ●. By default, show the last 25 rows from HEAD. The letters column contains change IDs, which ff evolog uses to show a change's recorded evolution.

Usage¶

Usage: ff log [OPTIONS] [path]...

Examples¶

ff log                          # Last 25 rows, including open work
ff log -n 0                     # Unlimited rows
ff log --commits                # Commit history without change IDs
ff log --signatures             # Verify signatures and show verdicts
ff log --body                   # Show full commit messages
ff log -p                       # Each row's patch under it
ff log --stat -n 5              # File counts under the last five rows
ff log -r main                  # Only main's tip
ff log -r 'trunk..@'            # Work beyond trunk, including @
ff log -r '@~3..@'              # Two commits and the open change
ff log src/parser.rs            # Follow a file through renames
ff log -r 'trunk..@' src/       # Filter a revision set by path

Options¶

Arguments:
  [path]...
          Files or directories to limit the log to; all of them when omitted

Options:
  -n, --max-count <COUNT>
          Number of rows to show; 0 means unlimited

          [default: 25]

  -r, --revisions <revset>
          Revisions to show, as a revset; without it, the walk from HEAD

      --commits
          Commits only — the plain history view

      --json
          Emit machine-readable JSON

      --fetch
          Fetch now on commands that support fetching, regardless of cadence

      --signatures
          Verify signatures and show verdicts; invokes external verifiers

      --body
          Print each row's message body under its subject

      --no-fetch
          Skip the fetch: read the tracking refs as they stand

  -p, --patch
          Print each row's patch under it

      --stat
          Print the diffstat under each row instead of the patch

      --name-only
          Print one path per line with its change type under each row instead of the patch

      --session <name>
          Session name for this invocation

      --fields <list>
          Keep only these dotted paths of the JSON data, comma-separated

  -U, --unified <n>
          Context lines around each change; 3 when omitted

      --at-op <op>
          Read as of this operation (a hex id or prefix, `@`, `@^`, `@~3`)

  -C, --cwd <dir>
          Run as if fufu had been started in <dir>

      --at <time>
          Read as of the operation current at this time (30m/2h/3d, or a date)

  -h, --help
          Print help (see a summary with '-h')

Choosing rows¶

-r (--revisions) selects a revision set (revset) instead of walking from HEAD. It accepts commit SHAs, change IDs, branches, tags, suffixes, and set operators. @ means the open change; @^ is HEAD, and @~3 is HEAD~2. The open row appears only if selected. See Revisions and IDs for grammar and current set boundaries.

Positional arguments are paths, never revisions: ff log main filters the path main. Paths select files or directory prefixes, without globs. The open row appears only when it touches a selected path. A file is followed through renames by default; a directory is not. With -r, paths filter the selected commits without rename following.

--commits omits change IDs. --body shows full commit messages, including the open change's pending description. These two options cannot be combined. Without --body, rows show only the subject.

--at and --at-op are declared but currently refused for log. Use ff op log for recorded operations or ff history for undo steps.

Patches under rows¶

Use -p (--patch) to print a patch below each row, --stat for per-file change counts, or --name-only for paths and change types. --stat and --name-only are mutually exclusive; either replaces the patch when combined with -p. None of these options combines with --commits.

-U <n> sets the number of context lines in patches, 3 by default. It has no effect without a patch.

Each commit is compared with its parent, and the open change with HEAD. Merges use the parents' auto-merge, falling back to the first parent if that auto-merge conflicts or has no common ancestor. This is the same comparison as ff show.

JSON output¶

Commit rows always include body. With -p, --stat, or --name-only, rows and the open block also include patch data. --stat omits each file's hunks; --name-only keeps path, from, kind, and binary per file and omits change counts.

Rows with patch data include against: parent, auto-merge, or first-parent. See JSON output and scripting for the complete report shape.

Change IDs and commit objects¶

A fufu commit stores a stable change-id header, also used by jj. Surviving changes keep it through fufu rewrites. A commit without the header derives its ID from its SHA, so an outside rewrite can change its identity.

The bold change-ID prefix is unique only on the displayed page. Resolution needs at least four characters and a unique repository match. Use more letters for an ambiguous prefix, or a commit SHA for divergent copies of one change. The prefix reference describes lookup limits.

The @ row's SHA identifies an internal open object, not a commit already recorded in branch history. A different message, partial commit, signing, or hook can change the SHA when ff commit records it. The SHA is blank on a clean tree.

Signatures and paging¶

signed means a signature is present, without verifying it. --signatures verifies signed commits and displays words such as verified gpg 9B295D68, bad signature, untrusted key, expired signature, expired key, revoked key, or unverifiable. Unsigned commits have no verdict. Verification uses up to eight workers; SSH normally invokes ssh-keygen twice per signed row, other formats once.

Terminal output uses fufu.pager, then FF_PAGER, then PAGER, then less. Piped output and JSON do not page.