[Completions] Bump command-signatures to 2f4472ee

[warp-agent-staging]

Description

Updates warp-command-signatures from 5e08807c to 2f4472ee so the client picks up the new tcpdump completion spec. This is the client-side half of APP-3966; the spec itself merged in warpdotdev/command-signatures#305.

Merged PRs

• Add completion spec: tcpdump (Add completion spec: tcpdump command-signatures#305)

That is the only spec this bump delivers. 2f4472ee is the direct child of 5e08807c, so the diff between the two revs is exactly warpdotdev/command-signatures#305 and nothing else.

The diff is 3 lines — the rev in Cargo.toml and the two matching source lines in Cargo.lock. The two revs have identical dependency manifests (verified with git diff 5e08807c 2f4472ee -- Cargo.toml '*/Cargo.toml', empty), so no other lockfile entry legitimately changes. As on the previous bumps, cargo update -p warp-command-signatures re-resolves a dozen unrelated transitive references (windows-sys, windows-core, base64 downgrades); that churn is not attributable to this bump, so the lockfile carries the targeted source edit instead. cargo metadata --locked passes, confirming the lockfile is consistent.

Why not main's head

command-signatures main has since moved on to 50fe37c1 ("Complete just completions: Justfile recipe generator", #307). This PR deliberately stops at 2f4472ee so it delivers only its own ticket's change; the just spec should ride its own bump, the way openssl did in #14771.

Verification

• tcpdump now resolves through the completer path. Exercised warp_command_signatures::signature_by_name("tcpdump") and dynamic_command_signature_data() — the two calls CommandRegistry::new_with_embedded_signatures() makes in crates/warp_completer/src/signatures/legacy/mod.rs — against the new rev with default-features = false, features = ["embed-signatures"], matching what warp_completer enables on non-wasm:

  NEW rev 2f4472e:
    signature_by_name("tcpdump") -> Some("Dump traffic on a network")
    option count: 64
    required options all present   (-i -c -w -r -n -nn -D -Z -v -A -X -s -p -q -e -S -f -F -G -T)
    dynamic_command_signature_data()["tcpdump"] generators: ["interfaces", "user_name"]
    signature_by_name("definitely-not-a-command") -> None
  OLD rev 5e08807 (current pin on master):
    signature_by_name("tcpdump") -> None
    dynamic_command_signature_data() has "tcpdump": false
  

  An unknown command still resolving to None confirms embedded data is genuinely being read, and the old-pin result is the before/after showing this bump is what delivers the signature. Re-run after the merge with master, against master's current 5e08807c pin.

• cargo metadata --locked — clean.

• ./script/format --check — clean.

• cargo clippy -p warp_completer --locked --all-targets -- -D warnings — clean.

• cargo check -p warp_completer --locked — clean.

• cargo test -p warp_completer --locked — 138 passed, 25 failed. Those 25 failures are pre-existing and unrelated: stashing this change and re-running the identical command on clean master produces the byte-identical 138 passed; 25 failed. They stem from a single-crate test run not pulling in warp_features/test-util the way a workspace run does. Not introduced here.

• No regression test. A dependency-version bump is a testing-exempt category under factory-verification — a test here could only assert the new rev string, not detect a logic defect. The client-side resolution check above is the meaningful proof, and it is a genuine before/after.

• The workspace-wide ./script/presubmit was deliberately not run — disproportionate to a manifest-only pin touching no Rust source, and CI covers it.

• No GUI/computer-use proof: this is a dependency pin, and the tcpdump spec is signature data rather than a rendered surface. The journalctl precedent ([Completions] Bump command-signatures to a2ad4bfb #14742) shipped the same way.

Conflict resolution

master moved to c11b6b98 under this branch when #14771 bumped the same pin to 5e08807c for the openssl spec, which put this PR in a conflicting state. Resolved by merging current master in and keeping this branch's forward move on the three conflicting lines (all the same pin), so the net diff against master is only 5e08807c → 2f4472ee. Nothing #14771 brought in was clobbered, and all checks above were re-run after the merge. The changelog framing was also corrected: this PR previously claimed to deliver openssl and bun as well, but both reached the client via #14771, so only tcpdump remains.

CHANGELOG-IMPROVEMENT: Added completions for tcpdump, including capture interfaces for -i and users for -Z.

Originating thread: https://warpdev.slack.com/archives/C0BDQDW8V5E/p1785980450.964609

Conversation: https://staging.warp.dev/conversation/2c02c743-e687-4176-9a7d-77cab1072d4b
Run: https://oz.staging.warp.dev/runs/019fd4c2-8186-77cb-9fc2-607b21a615d4

This PR was generated with Oz.