git-sync

Git Sync

Bring the local branch current with origin. Fast-forward when the branch is only behind; merge to integrate when local and remote have both moved. Never rebase, never force-push, never resolve conflicts automatically.

When to run

• The router (blake-os:git-ops) sends a "get the latest" request
• As the integrate step of a full backup, after git-commit — a clean working tree is required (Step 3), which is why the router commits first
• Directly, only on the exact request "git pull" or "git sync"

Workflow

Step 1: Verify repo and remote

git rev-parse --is-inside-work-tree >/dev/null 2>&1 && echo "in_repo: yes"
git remote get-url origin
git branch --show-current

If not a git repo, stop: "This directory is not a git repository." If no origin remote, stop: "No remote configured." Do not invent one.

Step 2: Fetch and measure state

git fetch origin
git rev-list --left-right --count @{u}...HEAD   # output: "<behind>	<ahead>"
git status --porcelain                          # any output = dirty working tree

If git fetch fails (network, auth), report the error verbatim and stop. Do not pull.

The rev-list line prints two numbers: behind (commits the remote has that the local branch doesn't) and ahead (local commits the remote doesn't have). These two numbers drive the decision in Step 3 — not error-message text. (If @{u} errors because no upstream is set, the branch has never been pushed; there is nothing to sync — tell the user to git-push to publish it.)

Step 3: Act on the measured state

behind = 0 — already current. If ahead > 0, note there are unpushed commits (run git-push). Done.

behind > 0 and ahead = 0 — simple fast-forward:

git pull --ff-only

On success, report how many commits were pulled. If it fails because the working tree is dirty (local changes would be overwritten), stop — list the blocking files and tell the user to commit them with git-commit or set them aside, then sync again. Do not stash.