CI: vendor silent-error tests join vendor suite (#1089)

[Mikola Lysenko (mikolalysenko)]

LLM Description written by Claude Code:claude-opus-5-5

Refs #1089. This is the last open piece of child 3: the vendor twin of the apply slice in #1546.

What changes

• tests/coverage_fix_vendor_silent_mute_exit.rs was a standalone test binary. It is now tests/vendor/vendor_silent_errors.rs inside the vendor suite binary, so CI links one fewer integration test binary.
• The --json test reads the envelope through common::envelope. It checks for a failed event with errorCode = apply_failed whose purl is the patch, instead of a raw events[0]["errorCode"] read.
• The two --silent tests now check that stderr has an Error: line naming the patch purl. They used to match free-form words: offline, and request, which comes as much from reqwest's error text as from ours.
• The module docs now say what the tests actually pin: vendor's per-patch Error: line, offline and against a dead patch service. The old docs named a stage_vendor_sources_in_memory staging layer and a no_local_source envelope code, and vendor uses neither. The apply twin's cross-reference is corrected to match.
• dead_endpoint() moves into vendor/npm_fixture.rs. It had three identical copies in the vendor binary, and two of them now import it from there.
• scripts/ci-test-durations.json (Windows shard timings only): the old binary's entry is dropped and vendor goes from 11.52s to 12.8s. That adds only the moved tests' runtime (~1.3s); the rest of the old 7.4s was the removed binary's link time.

CI cost

Workflows are unchanged. A lean PR run has the same ~28 Linux jobs before and after, at roughly the same job-minutes. The Linux test shards build one fewer test binary, which saves a few seconds of link time per shard. That is too small to justify a ci-perf label.

Test evidence

• cargo test -p socket-patch-cli --all-features --test vendor: 151 passed, including the 3 moved tests.
• cargo fmt --all -- --check: clean.
• cargo clippy --locked --workspace --all-features -- -D warnings (CI's invocation): clean.
• python3 -m unittest discover -s scripts/tests: 347 tests OK, 6 skipped. This includes the shard planner test that reads ci-test-durations.json.
• CI on 69c08f0: ci-ok and clippy green, and every lean job passed. Cursor Bugbot found no issues.

/code-review high

I fixed these findings: the stale docs, the sentence-fragment matches, the offline run spawning an unneeded fixture patch service (it now uses hermetic::binary_command directly, as before), the staging misnomer in the file name, the questionable partialFailure assertion (dropped), the duplicated dead_endpoint, the apply twin's stale cross-reference, and the inflated shard timing.

I did not take these:

• Use the siblings' sktsec_placeholder_value_for_tests_api token. That token triggers the token-shape warning on stderr, and the --json test requires stderr to stay clean. I tried it and the test failed, so the well-formed token stays, with a comment explaining why.
• Share stderr_chatter through common. It has five copies across five different test binaries. Merging them is a cross-suite change outside this slice.

🤖 Generated with Claude Code

https://claude.ai/code/session_017QqpxAuxWR1QbGKykwjMUV

---

Generated by Claude Code

[Mikola Lysenko (mikolalysenko)]

BugBot review

---

Generated by Claude Code

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 69c08f0. Configure here.}

[Mikola Lysenko (mikolalysenko)]

Ready for review (burn-down agent)

• Head: 69c08f00427a880020f375c324f4f2723048927d
• CI: 86/86 check runs green on this head (36 success, 50 path-skipped); no main-wide failures.
• Bugbot: reviewed 69c08f0, no findings; no open review threads.
• Already approved by Tanmay Singla (@Tanmay182003) on this head. Test-only change (vendor silent-error tests moved into the vendor suite, asserting on error codes); no CHANGELOG.md change.

---

Generated by Claude Code