Slice 010: core domain and persistence - #17
Merged
Merged
Conversation
ROADMAP row 010 to in_progress (plan_check rule 11 on a slice branch). NOTES.md carries the twelve-line G1 plan with the test or command each line names, the empty manual queue, and two deviations stated before any code: postgrex optional rather than default, and an in-tree UUIDv7 generator rather than a new dependency. plan_check: PASS. Signed-off-by: Ayla Croft <aylacroft@proton.me>
Red demonstrated on the exact defect PR #7 fixed, planted in the working tree and reverted; green on the clean tree; both outputs in NOTES.md. The population derives from git ls-files over .github; a missing YAML parser is a FAIL, never a skip. G1 line 1. Signed-off-by: Ayla Croft <aylacroft@proton.me>
…ts repo slot The adapter is chosen at compile time from TRINITY_DB (sqlite by default, postgres for the CI matrix), stated in config because use Ecto.Repo takes it as a literal. On SQLite the pool holds one connection in every environment and each pragma is named rather than inherited. Trinity.Repo.Receipts is declared, unstarted, for slice 024's own file. A test reads the pragmas back from a live connection; busy_timeout cannot be read through the pragma because exqlite installs its own busy handler, and the test says so with the driver's reason. 79 tests pass; compile with warnings-as-errors and credo --strict clean. G1 line 2. Signed-off-by: Ayla Croft <aylacroft@proton.me>
… job with a Postgres service TRINITY_DB=postgres selects the adapter at compile time (config/config.exs from line 2); config/test.exs takes DATABASE_URL under it; postgrex is an optional dependency so the desktop build carries no Postgres driver. A second job in gate.yml, named postgres so the required check keeps its context name, runs a postgres:17 service, asserts the compiled adapter is Ecto.Adapters.Postgres, then mix ecto.reset and the suite with :sqlite tagged tests excluded by tag. The version rows split: postgrex (in mix.lock) and pgvector (not yet a dependency). G1 line 3; the job's first green run on this branch is AC1's Postgres half. Signed-off-by: Ayla Croft <aylacroft@proton.me>
…tory by pid and mode Trinity.DataDir.Lock takes <data_dir>/LOCK with an exclusive create, records OS pid, mode, a per-boot token and the time, and is a supervised child placed before the Repo so a refused boot has opened no database file. A second acquire is refused naming the live holder's pid and mode; a file from a dead pid is taken over on Linux through /proc and treated as held elsewhere; a malformed file is never taken over. Ten tests, including the boot half against the running application and the child's position before the Repo. NOTES.md also records line 3's CI evidence: the postgres job on run 35509095238 dropped and created the database and ran the suite with the :sqlite tag excluded, 72 passed. G1 line 4. Signed-off-by: Ayla Croft <aylacroft@proton.me>
…do --strict was red on nesting depth Signed-off-by: Ayla Croft <aylacroft@proton.me>
…n's data-dir call mix compile --warnings-as-errors --force was red: forbidden reference to Trinity.Paths from lib/trinity/application.ex:74, which the previous commit's grep swallowed. Exported with the reason in lib/trinity.ex; the compile exits 0 and the suite passes. Signed-off-by: Ayla Croft <aylacroft@proton.me>
… seq, the stress test Trinity.UUID is an Ecto.Type minting RFC 9562 v7 ids with a monotonic rand_a, so rows sort by creation without a second column and both adapters keep their storage rule. Three migrations: personas, sessions, messages with the unique index on (session_id, seq). Trinity.Sessions is the only public surface over the tables; Store and the schemas stay inside a sub-boundary that Trinity exports, so TrinityWeb reaches the context and nothing else, proven by a compile that fails on a planted Store call and passes on the lawful one. append_message/2 assigns seq inside one transaction, locking the session row FOR UPDATE on a Postgres build (the lock is compiled in or out from the adapter, because the type checker refuses a runtime branch on a constant). Unknown roles, blank content and a missing session are refused by name. history/2 orders by seq with limit and offset. Factories are three plain functions. Stress: 20 writers, 200 appends each, 5 sessions, gapless, integrity ok, in the sandbox and again outside it on the dev database through scripts/stress_010.exs: 4000 appends in 1.2 to 1.3 s, about 3,000 per second, -wal at 4,152,992 bytes (1,014 pages, at the 1,000-page autocheckpoint). 105 tests pass; compile with warnings-as-errors, credo, plan_check and the name check are clean. G1 lines 5 to 9. Signed-off-by: Ayla Croft <aylacroft@proton.me>
…reasons, the boundary probe compiles forced coverage.tsv gains 010 at 44.88% (up from 30.37%); trinity.coverage OK. sobelow flagged four File calls in Trinity.DataDir.Lock as low-confidence traversal; every path is composed from the application's own data directory and a constant name, and each fingerprint carries that reason in .sobelow-skips.reasons. The boundary compile test now forces every compile: mix decides by mtime at one-second resolution, and the gate caught both a lawful half compiled against a stale violation and a red half that never compiled the violation. Four consecutive runs pass; the full gate exits 0. G1 line 11. Signed-off-by: Ayla Croft <aylacroft@proton.me>
PROOF.md carries every criterion's evidence: the gate at exit 0, 105 tests, coverage 44.88% (up from 30.37%), ecto.reset on SQLite here and on Postgres in CI, the stress run inside the sandbox and again outside it with the real WAL size, the boundary refusal text, the data-dir lock's refusal text. ROADMAP row 010 reads done; docs/05 names the id minter. The three deviations are stated in NOTES.md and repeated in PROOF.md. Signed-off-by: Ayla Croft <aylacroft@proton.me>
Signed-off-by: Ayla Croft <aylacroft@proton.me>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Slice 010 at G3. PROOF.md carries the evidence per criterion; NOTES.md carries the G1 plan, every line's record and three deviations. The postgres job on this pull request is AC1's and AC2's Postgres half on the closing tree. Merge-commit method, then tag slice/010 on main.