`gwz fetch`¶

Contact every selected repository's remote and report what moved, integrating nothing.

gwz fetch

gwz fetch fetches the configured remote of each selected workspace target, updates that repository's remote-tracking refs, and reports one line per repository. Plain gwz fetch includes @root plus configured member repositories, and the selectors are gwz push's.

It is the answer to "what moved upstream while I was working?" across the whole workspace, in one pass and one voice, without changing a single file.

What it does not do¶

It never integrates. No merge, no rebase, no fast-forward, no reset: no branch, no HEAD, no index, no working-tree file. Use gwz pull to integrate what a fetch showed you.
It never writes workspace artifacts — no lock, no manifest, no boundary sync. Because of that it still runs while a merge is open, unlike pull and push.
It never prunes.
It never skips the network, with one exception: --dry-run. There is no --check-remotes and no "unchanged since the last fetch" short-circuit as there is on gwz push: a fetch that does not connect has answered nothing. The exception is gwz --dry-run fetch, which resolves the selection, prints one planned row per repository it would have contacted, and stops before the network.

Output¶

$ gwz fetch
status: Partial
@root      .         no change         (origin/main, +2 -0)
mem_core   gwz-core  90ef552..330a174  (origin/main, +0 -3)
mem_cli    gwz-cli   no change         (origin/main, +0 -0)
mem_local  local     no upstream
mem_priv   private   failed            RemoteRejected: member 'mem_priv' at 'private': failed to connect to 127.0.0.1: Connection refused

<before>..<after> — the remote-tracking ref moved, abbreviated as git abbreviates one.
new <after> — the remote-tracking ref did not exist before this fetch, so there is no left-hand side to show. This is the row a first fetch prints.
updated — the ref moved but neither object id is known, so there is nothing to print on either side.
no change — the repository was contacted and its tracking ref did not move.
no upstream — the repository has no fetch remote, or no attached branch whose tracking ref could move. A reported row, not a failure: one local-only member does not break a whole-workspace observation.
failed — the remote refused, or the fetch errored, with the reason.
would contact <remote> — --dry-run only. The repository was not contacted; see below.
(<remote>/<branch>, +A -B) — how far the current branch is ahead of and behind the tracking ref, counted after the fetch.

A dry run contacts nothing¶

gwz --dry-run fetch resolves the selection and reports the repositories it would contact, contacting none of them. Its rows say so in their own words:

$ gwz --dry-run fetch
status: Noop
@root      .         would contact origin
mem_core   gwz-core  would contact origin
mem_cli    gwz-cli   would contact origin
mem_local  local     no upstream
mem_priv   private   would contact origin

would contact <remote> is a plan, not a result, and no live fetch prints it. Its machine value is "result": "Planned". A repository with no fetch remote is still no upstream here, because that answer needs no network. A --remote <name> that a repository does not have is refused here too, as failed with MissingRemote, because that answer is also local: a dry run never promises to contact a remote the live run cannot. Nothing in a dry-run row comes from a remote, so a dry run cannot tell you what moved: run gwz fetch for that. A dry run exits with the same codes as the live run of the same selection.

Examples¶

Fetch every repository, root included:

gwz fetch

Members only:

gwz --no-target @root fetch

One member:

gwz --member mem_app fetch

Another remote, for every selected repository:

gwz --remote upstream fetch

Machine output, one object per repository under fetch_repos:

gwz --json fetch

Resolve the selection and report what would be contacted, contacting nothing:

gwz --dry-run fetch

Every row reads would contact <remote>, or no upstream where there is no fetch remote to contact.

Exit codes¶

Follow gwz push's, with one difference worth knowing: no change means contacted-and-answered rather than skipped, so a run in which one remote failed and every other repository read cleanly exits 1 — the report is incomplete — even though nothing moved.

Exit	Meaning
`0`	every selected repository answered
`1`	some answered and some failed; the report is incomplete
`2`	every selected repository was refused before the network, so nothing was contacted

Exit 2 is the whole-batch refusal: for example --remote <name> naming a remote that no selected repository has. A refusal that stops the request before it has a selection at all, such as no workspace under the current directory or an unknown member id, is a typed error and exits 1, as it does on every verb. A dry run shares these codes with the live run it rehearses.

gwz fetch¶