Files
rustfs/crates
Zhengchao An f12f3c47e1 test(e2e): finish the serial sweep, leaving only markers that mean something (#6240)
Removes the remaining 230 no-op `#[serial]` markers across 77 files, plus the 77 imports that went with them. Stacked on the six-suite batch; together they take `crates/e2e_test` from 345 markers to 36.

The reasoning is unchanged from #6209: `#[serial]` is an in-process mutex and cargo-nextest gives every test its own process, so only a `[test-groups]` binding with `max-threads = 1` serializes anything.

Everything still carrying a marker now has a reason to:

  inline_fast_path_cluster_test     14   test-group e2e-inline-boundaries
  policy/policy_variables_test       6   binds a fixed port
  kms/kms_vault_test                 5   test-group e2e-vault
  reliability_disk_fault_test        4   test-group e2e-reliability
  degraded_read_eof_regression_test  3   test-group e2e-reliability
  replacement_privileged_e2e_test    2   test-group e2e-reliability
  protocols/webdav_core              1   binds fixed ports
  policy/test_runner                 1   binds a fixed port

The three fixed-port files are held back rather than swept, because their markers are not merely useless: `PolicyTestEnvironment::with_address("127.0.0.1:9000")` and webdav_core's `127.0.0.1:9080`/`:9010` bind fixed ports, which an in-process mutex cannot protect against a second test process. They need a test-group, which is a config change rather than a deletion, so it is filed on the issue instead.

`security_boundary_test` was checked and swept: its `127.0.0.1:8080` strings are SSRF targets fed to AddTier in a negative test, not ports it binds.

Refs backlog#1846 (T1)
2026-08-19 10:29:36 +08:00
..