)]}'
{
  "commit": "af045c12cbbafbfd906ce5e66113fc89a1aa2b0d",
  "tree": "aa011ee2f7bb189066fb4a836ba35814cc988425",
  "parents": [
    "0eacdb809995ae07a9a4a97860f3ad85da105880"
  ],
  "author": {
    "name": "Andrew Musselman",
    "email": "akm@apache.org",
    "time": "Fri Oct 23 17:12:00 2026 -0700"
  },
  "committer": {
    "name": "Andrew Musselman",
    "email": "andrew.musselman@gmail.com",
    "time": "Fri Oct 23 17:09:07 2026 -0700"
  },
  "message": "ledger: an embedded server that will not create a database is a skip\n\n`check (linux)` and `coverage` both fail on this, and have been failing on main:\n\n    called `Result::unwrap()` on an `Err` value: \"embedded PostgreSQL will not\n    create a database: error returned from database: invalid value for parameter\n    \\\"TimeZone\\\": \\\"UTC\\\"\"\n\n19 of the 38 `dm-ledger` tests, plus\n`the_seven_steps_pass_over_stdio_with_a_postgres_ledger` in the conformance\nsuite. It is not the extraction: the PostgreSQL 18.6.1 archive\npostgresql_embedded downloads ships a trimmed share/timezone -- 30 top-level\nentries with no `UTC`, and an `Etc` holding only `GMT+-N` -- and `UTC` is the one\nzone name create_database asks for. Reproducible after `rm -rf ~/.theseus`, and\non GitHub\u0027s runners as well as a local Ubuntu 25.10 box.\n\nThis module already says what to do about a server that cannot be had: \"where\nthat cannot be done, the test is skipped with the reason on stderr\". setup() and\nstart() failures are skips; create_database() was the one step that hard-errored,\nso an embedded server that starts and then refuses a database took the whole\nsuite down instead of stepping aside. A server we started ourselves and cannot\nuse is the same situation whichever call it refused at.\n\nMEMORY_PG_URL keeps its behaviour: a server someone named and we could not use\nis still an error, because somebody asked for that one by name. Running the\nPostgreSQL half against a real server is unaffected and is how it should be\nexercised in earnest:\n\n    docker run -d -e POSTGRES_PASSWORD\u003dpg -p 55432:5432 postgres:17\n    MEMORY_PG_URL\u003d\u0027postgres://postgres:pg@127.0.0.1:55432/postgres\u0027 make test\n\nThis does mean the PostgreSQL half goes unrun in CI until the archive is fixed or\na service container is added to the workflow, which is worth doing -- but a suite\nthat cannot pass anywhere hides every other failure behind it, and it has been\nhiding them on main for days.\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "c012bc77a37faa7a666a91402bfce1256ae089b4",
      "old_mode": 33188,
      "old_path": "crates/ledger/src/testing.rs",
      "new_id": "855340675dffc720c8351a2087c1df6257e477f2",
      "new_mode": 33188,
      "new_path": "crates/ledger/src/testing.rs"
    }
  ]
}