mirror of
https://github.com/Twigpine/node.git
synced 2026-11-09 13:31:25 -10:00
beardthelionandKevin Codex
ed76adfe29
* fix(node): close the advisory-lock probe's session if it is dropped mid-acquire A cancelled .await does not cancel an already-sent SQL statement, so a pg_try_advisory_lock whose future is dropped still takes the lock server-side while the caller abandons the result, leaving nothing to release it. The connection then returns to the pool holding the lock and wedges that repo until sqlx recycles the session. Introduce LockProbe, which owns the connection across the in-flight try-lock and closes it in its own Drop if it is still held. close_on_drop is a one-way setter, so the arming lives in Drop rather than being set up front and cleared on success; disarming is Option::take, which is what into_conn does once an acquire is actually observed. This is now the only place that issues pg_try_advisory_lock. The committed gate drops a probe without taking its connection, which is the state a cancellation leaves behind, and polls a standalone observer until the lock frees. Deterministic on purpose: the timing sweep that found this window leaks roughly 1 in 600, which is not something a CI gate can rest on. Observed RED before this change with the lock still held for the full 10s window. Refs #279 * fix(node): add a dedicated, lazily-connected advisory-lock pool Pinning a connection for the lock's lifetime is only safe if those connections come from somewhere other than the pool serving ordinary request handlers, otherwise a push burst starves every other query. Add GITLAWB_DB_LOCK_POOL_MAX_CONNECTIONS (default 32) and a Db::lock_pool builder, with the sizing tradeoff documented on the field and in .env.example: every in-flight write pins one connection here, so the value is a hard ceiling on simultaneous writes node-wide. The pool connects lazily on purpose. The main pool must connect eagerly because it runs migrations, which is why it needs connect_db_with_retry's backoff and degraded-server handoff; that function is not a generic retry helper and the lock pool is built well after the db-ready handoff has already resolved. A lazy pool has no startup work, so it adds no new way for the process to fail to boot and needs no second copy of that machinery. If Postgres is unreachable when the first write arrives, that write fails on the pool's own acquire timeout, like any other database-backed request. Pure configuration, so no proof-first cycle: the knob is covered by a parse/default/reject-zero test. Db::lock_pool has no caller until the guard wiring lands, hence the temporary dead_code attribute. Refs #279 * fix(node): hold the lock-owning connection in RepoWriteGuard (#279) Postgres advisory locks are session-scoped: only the backend that took one can release it. acquire_write took the lock through fetch_one(&pool) and release unlocked through execute(&pool), two independent checkouts, so the unlock usually landed on a session that held nothing and returned false. Measured on main: two writers on one node and the same repo BOTH acquired, 50 of 50 sequential cycles leaked, and 100 writes left 100 orphaned advisory locks on the server. The guard now owns the PoolConnection that took the lock, drawn from the dedicated lock pool, and releases on that same session. The retry loop probes through LockProbe so a cancellation mid-acquire cannot strand the lock, and hands the connection back before each backoff so a spinner on a contended repo does not pin a slot while idle. Pool exhaustion is deliberately not retried. It is a different condition from lock contention, and retrying it would spend all 60 attempts on a capacity problem unrelated to this repo while reporting it as someone else holding the lock. Both #279 acceptance tests were observed RED first: the exclusion test admitted the second writer, and the leak test reported 1 lock held where 0 was required. Both GREEN after. Full crate suite 516 passed. Db::pool() is removed because this change was its only caller. Refs #279 * fix(node): free the advisory lock when a write guard dies without releasing A guard can exit without reaching release(): an early ? on the pre-write download, a panic, or an axum handler future cancelled when the client disconnects. Its session still holds the lock, so returning that connection to the pool would block every future write to the repo until sqlx recycles it, which on the 0.8.6 defaults is ten minutes idle or thirty minutes lifetime. Close the session instead and let Postgres free the lock at session end. One hazard found by writing the teardown test rather than by reasoning: PoolConnection::drop spawns onto the runtime for both closing and returning, and panics outright when no runtime handle exists. That panic would fire inside a Drop and abort the process during unwind. It is not introduced by this commit, it comes with owning a PoolConnection at all, but this is where it becomes reachable. So Drop checks for a runtime first and, with none, leaks the handle deliberately rather than panicking: the process is already exiting and socket teardown ends the session, which is what frees the lock at exit anyway. Observed RED before the fix, with the lock still held for the full 10s poll window against a standalone observer on a no-reap pool, and the teardown case panicking in sqlx-core connection.rs:208. Proven load-bearing after: neutering the close_on_drop call turns the drop test RED again. The must-not case (a released guard reuses its backend pid across four writes on a pool sized 1) passes in both states, so the signal is specific to the abandoned-guard path. Full crate suite 519 passed. Refs #279 * fix(node): read the advisory unlock's result instead of discarding it pg_advisory_unlock reports "you did not hold this lock" as a false RETURN VALUE plus a server WARNING, never an error, so let _ = execute(...) could not distinguish a real release from a no-op. Three blindnesses stacked in those four lines: execute discards the row, the boolean lives in the row, and let _ discarded the Result too. Read it through fetch_one into (bool,). A false means this session's lock state is not what we believe it is, so the connection stays in the guard for Drop to close rather than being handed back to the pool as clean. Only a confirmed unlock returns it. A query error gets the same treatment, since the lock must not outlive a session we can no longer reason about. Note this is the only unlock site: the pre-write download's error path returns through the guard, so Drop covers it and there is no second place to keep in sync. RED before: the connection came back on the same backend pid after an unlock that returned false. GREEN after, and proven load-bearing by treating false as success, which turns it RED again. The must-not case (a normal release still reuses its backend) passes in both states, so this does not over-close the happy path. Full crate suite 520 passed. Refs #279 * fix(node): bound the object-storage transfers that run under the write lock Two transfers happen while the per-repo advisory lock is held: the archive download inside acquire_write, which runs after the lock is taken and before the guard exists, and the upload inside release. Both were free before the guard pinned a lock-pool connection, because the lock's connection went back to the pool immediately. Now an unbounded stall holds a lock-pool slot for its whole duration, so enough concurrent stalls deny every write on the node with no reaping path. Add GITLAWB_LOCK_HELD_TRANSFER_TIMEOUT_SECS (default 300) and route both through it. A timed-out upload is UNKNOWABLE rather than failed, so the timeout arm takes no compensating action: the PUT may well have landed. The lock is released either way, which trades a narrow last-writer-wins window for not wedging the repo behind a stalled transfer. Deliberate, and stated here rather than discovered later. Note what is NOT bounded by this knob: acquire_fresh's download. It runs before any lock is taken, so it holds nothing. An earlier revision of this change put the bound there by mistake, because both functions contain an identical download call and the first match won. Two disclosures. Coverage: the committed tests cover the bound mechanism, not the wiring. Driving a genuinely stalled transfer through acquire_write needs either the object-store abstraction (out of scope) or a process-global AWS_ENDPOINT_URL_S3 mutation, which would make the suite order-dependent under the concurrent runner. That a stalled transfer is bounded is therefore verified by reading, not by execution. Behavior: on timeout with a local copy present, the download falls into the pre-existing self-healing fallback and the write proceeds against that local copy. This widens the conditions reaching that path from corrupt-or-unreachable to include merely-slow, so a stale tree could now be written and re-uploaded on a slow link. Kept consistent with the existing failed-download behavior rather than inventing new semantics here; changing it is its own decision. Full crate suite 522 passed. Refs #279 * fix(node): authorize before taking the write lock in close_issue close_issue acquired the per-repo advisory lock, then ran the owner-or-author check, then returned 403 with the lock still held. That was harmless only because the lock excluded nothing. Making it work, as the preceding commits do, turns this ordering into a denial primitive: any caller with repo read access could take the write lock on demand and be refused the write, while a legitimate writer burned its 60-attempt retry budget against a lock held by someone with no write authorization. On a public repo that is every permissionless identity. This series creates the exposure, so it closes it in the same series. The owner check is cheap and moves above the lock outright. The author fallback needs the issue's git-JSON blob, since there is no author column, so it now reads through acquire() rather than acquire_write(): an issue's author is set at creation and never changes, so reading it outside the lock races nothing, and only the mutation needs exclusion. Two deliberate behavior choices. A non-owner whose authorship cannot be established gets 403 rather than 404, so this route does not tell an unauthorized caller whether an issue exists. And the owner's existing 404-for-a-missing-issue path is preserved by re-reading under the guard, which the mutation wants anyway. Swept the other three acquire_write sites: merge_pr owner-gates at :200 before acquiring at :214, the repo write path checks did_matches at :916 before :933, and create_issue's read gate is legitimate because that caller IS authorized for the action it performs. close_issue was the only one with the wrong order. RED before: with an independent session holding the repo's lock, a stranger's request sat in the retry loop until the 3s deadline fired (Elapsed), which is the wedge itself. GREEN after, refused immediately. Proven load-bearing: disabling the pre-lock refusal restores the Elapsed. Full crate suite 523 passed. Refs #279 * fix(node): address the review findings on the advisory-lock series Seven fixes from a seven-reviewer pass. Eight of the nine substantive findings were defects introduced by this branch, and two contradicted claims made in its own earlier commit messages. P0, the worst of them. The under-lock download's timeout was folded into the same Err as a corrupt archive, so it fell through to the use-the-local-copy fallback and the write proceeded. Two ways that destroys data: we do not know whether we hold the latest tree, so re-uploading can overwrite another node's newer archive; and the abandoned download's extraction runs in an uncancellable spawn_blocking that ends in remove_dir_all + rename over the same path, so git would have been running against a directory a background task was about to delete. Now the Option is branched rather than collapsed: a genuine fetch error keeps the local-copy fallback, a timeout refuses the acquire and lets the guard's Drop free the lock. LockProbe closed its connection on ordinary contention, so a 60-attempt spinner tore down 60 backends -- the exact behavior an earlier commit claimed it avoided. The test written to guarantee that asserted the HOLDER's lock count, which cannot observe the probe's own connection, so it passed throughout. The correct predicate turned out to be narrower than the first attempt at this fix: not "we saw an answer" but "we positively know nothing was acquired," because a true answer means the lock IS held and dropping without handoff leaks it exactly as a cancellation would. The first attempt reopened U1's leak and U1's gate test caught it. tigris.exists() ran unbounded inside the lock-held span, so the knob's promise was not kept and the docs were wrong about the shape. The HEAD and the download now share one budget, so worst-case occupancy is one budget per span rather than two. close_issue's pre-lock author read used acquire(), whose fast path returns on a cached directory and never contacts object storage, so on a multi-node deploy an issue author would be refused their own issue. Now acquire_fresh, which refreshes without taking the lock -- the property the comment already claimed. That comment also asserted authorship is immutable, which a reviewer disproved by force-pushing a forged blob at refs/gitlawb/**; reworded to the real justification (a pre-check whose mutation re-reads under the guard). The underlying pushability is pre-existing and larger than this branch. The retry loop had no wall-clock cap, reaching ~360s against a 120s proxy idle timeout that was itself lowered from 600s after held connection slots caused a production outage. Capped at 90s, with the bail message naming the real bound. All four acquire_write call sites stringified their error, destroying the anyhow downcast that error.rs documents explicitly ("without this, every database outage surfaces as a 500 instead of a 503"). PoolTimedOut is in that 503 set, so a saturated lock pool told clients not to retry something transient. Now propagated. Tests: six that the plan required and the first pass never wrote. The R4 pool-isolation test (named in the plan as proving the pool split, entirely absent), close_issue's owner and non-owner-author twins (INV-21c, a two-principal gate with only its deny arm covered), a waiter-does-not-block- an-unrelated-repo case, and a config test for the transfer knob. The runtime-teardown test no longer returns green when DATABASE_URL is unset. Drop's no-runtime branch now detaches via leak() instead of mem::forget: PgConnection has no Drop impl, so the socket closes synchronously and the lock frees immediately rather than at process exit. Evidence. The probe predicate is proven load-bearing in BOTH directions: forcing always-close reddens the churn regression, forcing never-close reddens the cancellation gate, and only the correct predicate satisfies both. The author twin reddens when the fallback is removed. One honest limit: the author twin does NOT redden when acquire_fresh is reverted to acquire. With tigris disabled in tests the two calls are identical, so that fix is correct by reading, not by execution -- the same seam that leaves the transfer bound's wiring untested. Full crate suite 529 passed, clippy -D warnings clean. Refs #279 * fix(node): log lock-pool saturation in the request path, not via readiness Completes the shed-in-path half of the availability story. F7 already made a pool timeout a retryable 503 by letting the sqlx downcast through; this adds the operator-facing half, with the pool's own size/idle counters so an incident can distinguish "the pool is full" from "the database is gone" without reproducing it. Deliberately NOT a readiness probe. /ready gates Fly routing with no fail-open, so failing it on a saturated pool would pull this node's READS out of service too and push its write load onto peers carrying the same load. That is the downward spiral AWS's health-check guidance names and the SRE Book documents independently. A lock-pool readiness probe would also add nothing on the reachability axis, because both pools are built from the same database_url, so the existing app-pool ping already answers it. The error is still returned via .context() rather than replaced, because anyhow preserves downcastability through context layers and the 503 mapping depends on it. Cost is bounded by construction: one line per failed acquire, and a failed acquire has already paid a multi-second pool timeout. Refs #279 * fix(node): refuse a write when object storage is unknowable, and shed contention as 503 Four defects a seven-reviewer pass found in the previous two commits, three of them cases where one arm of a match contradicted its neighbour. The under-lock refresh read a failed Tigris HEAD as "no archive" (`exists().await.unwrap_or(false)`), skipped the refresh, and wrote against a possibly-stale tree, then re-uploaded over it. That is the outcome the timeout arm twenty lines below explicitly refuses. A failed HEAD and a failed download leave us knowing different things, so they no longer share a branch: only a download failure after a successful HEAD establishes that the local copy is a sound thing to fall back to and re-upload. A HEAD failure now refuses the write. Lock contention that ran out the acquire deadline surfaced as a 500 carrying the owner slug and repo name in the client-visible body, while pool exhaustion fifteen lines up returned a deliberate 503. Contention is transient and ordinary, so it now maps through a typed `RepoBusy` to a retryable 503 with a fixed body; the detail stays in the log at the raise site. `lock_not_taken` was assigned after the try-lock answered, so an error left a previous `true`-derived value standing and could return a lock-holding session to the pool. It is now cleared before the statement is sent. `leak()`'s no-panic property in the guard's no-runtime Drop branch depended on `min_connections == 0` without saying so; a future tuning change would have silently re-armed a panic inside Drop. Made explicit with the reason. Two tests that did not bind: - `waiter_on_one_repo_does_not_block_another` proved only that two lock keys do not collide. It now samples the pool counters across more than two backoff cycles. RED at 0/50 samples with `drop(probe)` moved after the sleep, the exact defect it names, which the previous version survived. - the exhaustion test's `Ok(Err(_)) | Err(_)` was satisfied by its own outer timeout. It now requires `PoolTimedOut` and asserts the slots are accounted for. RED with the pool sized N+1. `contended_acquire_sheds_as_repo_busy_not_internal_error` is new and drives the deadline path for real; the deadline became a field so it does not wait 90s. RED at 500-vs-503 with the downcast removed. The acquire deadline's docstring claimed a total under the 120s proxy idle timeout, which the 300s under-lock transfer bound in the same function contradicts. It bounds the wait only, and now says so. The backoff is clamped to the remaining budget. * fix(node): re-authorize close_issue under the guard instead of only re-checking existence The guarded re-read matched `Ok(Some(_))` and discarded the blob, so the authorization decision rested entirely on the pre-lock read. A comment above claimed the opposite. That is not a narrow window: `acquire_write` re-downloads the archive after locking, so the tree that gets mutated is routinely not the one the author was read from. With owner-push enforcement defaulting to false and branch protection covering only `refs/heads/*`, `refs/gitlawb/issues/*` is pushable, so a forged author blob landing between the two reads was honored. The blob is already in hand under the guard, so re-asserting owner-or-author costs a deserialize. A non-owner whose issue has vanished gets 403 rather than 404, matching the pre-check's existing refusal to reveal existence. `owner_can_still_close_after_the_reorder` seeded the owner as their own issue's author, which made it unable to fail: with the owner check disabled the author fallback granted the close and the test stayed green. It now seeds a third party, so only the owner arm can grant. RED with `is_owner = false`. Also drops the claim that the author twin covers the acquire-vs-acquire_fresh distinction. `RepoStore::for_testing` hardcodes `tigris: None`, so the two calls are identical in every test here and reverting that line leaves the twin green. Separating them needs an object-storage seam. Stating the gap beats asserting coverage that does not exist. * style(node): rustfmt the new contention test Whitespace only; cargo fmt --check gates the push. * fix(node): shed an unrefreshable write as a retryable 503 with a fixed body The two refusal arms added for the under-lock transfer bound returned bare anyhow errors, so AppError's From impl (which downcasts only sqlx::Error and RepoBusy) landed them in Internal: a 500 internal_error whose body was the error string. Two defects in one. A transient object-storage failure told the client not to retry, and the timeout arm's message interpolated the owner slug and repo name straight into the response body, contradicting the fixed-body policy the RepoBusy arm sets six lines above it. RepoUnavailable follows RepoBusy exactly: a fieldless type raised with the operator detail in a context string, downcast to its own rung, mapped to a 503 whose body interpolates nothing. The detail stays in the log at the raise site. The timeout arm logs at error! where its sibling logs at warn!, deliberately: a 300s stall pinned a lock-pool slot for five minutes and is the condition the transfer bound exists to surface, so it must keep paging when the handler classifier demotes the ordinary blip case. * fix(node): refuse a fresh-copy write when the archive HEAD cannot be read acquire_fresh collapsed a failed HEAD into "no archive exists" via unwrap_or(false) and served the local copy, while the sibling path under the lock already refused on the same condition. A push that hit a storage blip got an advertisement built from a possibly-stale tree, uploaded a whole pack, and was then refused by acquire_write for the reason the advertisement had already swallowed. The authorship pre-check on close_issue read the same way, which is an infrastructure failure resolving toward a denial. A failed HEAD now raises RepoUnavailable, so both callers surface it as the retryable 503 rather than a stale success. The download-failure fallback is unchanged: a present-but-unreadable archive is still self-healed from local. info_refs keeps its map_err for every other failure so the read path's error vocabulary does not move; only the typed error takes the From chain. A test-only TigrisClient constructor pointed at a closed port makes both refusals executable, so this is no longer verified by reading. It also proves the under-lock arm end to end, which the earlier draft had recorded as an untestable gap. * fix(node): log expected transient acquire failures at warn, not error Both acquire call sites logged every failure at error, including RepoBusy, which the raise site already logs at warn. Ordinary write contention paged. The previous commit widened the problem: info_refs now raises RepoUnavailable on a storage blip, so that site would have started paging on the condition this series just classified as transient and retryable. A classifier over the two typed refusals picks the level at both sites, mirroring the startup path's permanent-vs-transient split. Anything it cannot classify still logs at error, so an unknown failure keeps paging. * fix(node): log the read failure the close_issue pre-check folds into a denial The authorship pre-check treated a get_issue I/O error and a genuinely absent issue as the same None, with no log line. The client answer is deliberately identical, since a caller who cannot write must not learn whether the issue exists, but the two are not the same event and an operator had no way to tell a real authorization denial from a filesystem or parse failure behind it. Splitting the arm leaves the 403 exactly where it was and makes the read failure visible in the log. * test(node): poll for the closed session instead of sleeping a fixed 300ms The release-invariant test slept 300ms for a Drop-spawned close before comparing backend pids, which is a coin flip on a loaded CI runner. It now polls pg_stat_activity for the captured pid on a standalone connection, the same discipline poll_until_free documents: a pooled observer would be handed the session under measurement and hide the effect. The conversion was checked against the failure it exists to catch rather than assumed. Pooling the session on an unlock that returned false makes the test fail after the poll deadline, not hang, and no_reap_pool disables idle timeout and max lifetime so nothing but the close under test can retire that backend. A generous deadline would have been the same defect as the sleep. * fix(node): raise a failed fresh download with no local copy as RepoUnavailable acquire_fresh already refused a failed Tigris HEAD as RepoUnavailable, so the handler layer mapped it to a retryable 503. A failed GET on the same path still returned a bare anyhow error, which the info/refs map_err closure stringified to AppError::Git and answered 500. One endpoint therefore told the client a transient object-storage blip was permanent or retryable depending on which call failed, and close_issue's pre-check inherited the same split through its bare ?. Raise it at the source instead of at each consumer: From<anyhow::Error> for AppError already downcasts RepoUnavailable out of the context chain, so both callers pick up the retryable mapping without touching either. The new test drives HEAD 200 with GET 500, which is the exact state the refusal is for: archive present per HEAD, GET failed, no local copy to fall back on. * test(node): assert the second writer sheds as RepoBusy instead of timing out two_writers_on_the_same_repo_are_not_both_admitted wrapped the second acquire_write in an 8-second outer timeout and only checked the future had not finished. That passes for any stall, including lock-pool saturation or a slow CI box, so it could not tell a working shed from an unrelated hang, and it cost 8 seconds on every suite run. Use with_lock_acquire_deadline and assert the typed RepoBusy downcast while the first guard is still held, matching what contended_acquire_sheds_as_repo_busy_ not_internal_error already does. It now also fails loudly if a second writer is admitted, which the timeout version could not distinguish. * fix(node): read the close_issue author pre-check from a non-mutating snapshot The non-owner author fallback refreshed through acquire_fresh, which publishes into the live repo directory: its extract step removes the existing directory and renames the new one into place. That runs with no write lock held, so any signed non-owner could trigger a directory swap underneath an in-flight guarded write on the same path. read_snapshot downloads to a throwaway temp dir and hands back a RepoSnapshot that removes it on drop, so the pre-check still sees fresh data and the live path is never touched. download_to grows a publish flag to serve both shapes from one path. The wedge invariant still holds: a stranger is refused without waiting on the write lock. * fix(node): bound every await in the lock-acquire loop by the deadline The remaining budget was checked before the pool checkout but bounded neither the checkout nor the pg_try_advisory_lock query that follows it. A checkout starting just under the deadline could wait out the pool's own acquire timeout, and a slow query could be accepted after the budget was spent, so the advertised wall-clock cap held only on the fast path. Both awaits now run under the remaining budget and shed as RepoBusy when it runs out. The probe's Drop closes its session, which cannot hold a lock it never confirmed taking, so the query-timeout arm is a plain shed. * test(node): add an S3 mock that really enforces conditional writes The fence work landing next is only as good as what the tests can observe, and a mock that answers 200 to every PUT would make the whole suite vacuous. This one holds the object and its ETag, refuses a mismatched If-Match and an If-None-Match "*" over an existing object with 412, and mints a fresh ETag per successful PUT so two byte-identical archives never share a token. Capture-then-replay is deliberate rather than parking a handler and hoping it resumes: when tokio drops an SDK future the client can tear the connection down and cancel the server task with it. Replaying what arrived models the arm that matters (body fully transmitted, commit decided later) with no timing in it. Six tests pin the semantics in both directions so a hollowed mock cannot hide. * fix(node): stop pretending a held lock fences a late upload The timeout arm claimed that keeping the advisory lock protected a successor from an abandoned PUT. It does not. release takes mut self, so the guard drops the moment it returns and Drop closes the session; measured on this branch, a successor took the same repo's lock 5ms after release returned while the PUT was still in flight. The comment and warn now say what is actually true: the outcome is unknowable, the PUT may still land, the lock releases normally, and a conditional upload is what keeps a late publish from overwriting a successor's archive. Two tests replace the claim. Session disposition is the observable that separates the two shapes, so the unlock is pinned to run and be confirmed on the guard's own session with the connection returned to the pool, checked by backend pid. Successor admission is pinned too, but noted as not what proves the point, since the lock frees within milliseconds either way. * feat(node): let the storage client express and detect a conditional write upload takes an UploadPrecondition (IfMatch, IfAbsent, Unconditional) rather than an Option<String>, so the absent case and the deliberate no-fence case are distinguishable at the type level and a caller cannot lose the fence by passing None. head_etag reads the current ETag alongside exists, which keeps exists and its other callers untouched. A failed precondition has to be classified off the raw HTTP status: PutObjectError models no PreconditionFailed variant, so a 412 arrives as Unhandled with nothing useful on it. 412 is always a lost precondition and 409 is one under IfAbsent. 404 deliberately is not: no archive delete exists on this line, so a 404 on a conditional PUT means a wrong bucket or endpoint, and reporting that as retryable would send clients into a loop against a permanent fault. The three background uploads outside the write guard now publish with IfAbsent. They fire only where the archive is expected absent, and leaving them unconditional would defeat the fence from the side: init uploads an empty bare repo, so a push landing just before it could have its archive replaced by that empty one. A refusal there means someone else already published the key, which is logged as the correct outcome rather than a failure. * fix(node): fence the release publish so an abandoned upload cannot win acquire_write now reads the archive's ETag under the advisory lock and the guard carries it, so release publishes conditionally on the generation it actually refreshed from. An upload abandoned by an earlier writer no longer overwrites a successor's archive: the store rejects it, because the ETag it was written against is gone. A refused precondition gets exactly one supersede-retry, never a loop. The distinction that makes this sound is that a 412 is a definite answer, unlike the timeout arm where nothing is knowable, and the retrying writer still holds the lock, so whatever landed underneath was written without one and its tree is not the authority. Two losses in a row refuse instead of escalating. That retry is what keeps ordinary pushes working now that init publishes create-only: a first push racing init's upload of the empty repo loses once and then wins, rather than surfacing a 503 on the most common operation there is. release returns a must-use outcome and the four publishing handlers propagate it before any trust bump, webhook, or success body, so a publish the store refused reads as a retryable 503 instead of a 201. The three release(false) sites deliberately do not map it: they publish nothing, and a 503 there would shadow the 403 or 404 the route means to return. The download-failure fallback also publishes fenced now. That arm knows the stored generation (its HEAD succeeded, only the GET failed), so publishing unconditionally from it would reintroduce the same overwrite. * test(node): drive the abandoned-upload race end to end, and probe the real backend The headline test is the one that had to exist. Writer A's release is parked past its transfer bound and returns with the outcome unknowable, B acquires and publishes, and A's captured PUT is then replayed: the store answers 412 and B's archive survives. The create-only variant covers the arm whose real-world failure is silent rather than loud. The control is what makes those attributable. With no interleaved B, an abandoned-then-replayed PUT whose generation still matches lands. Without it the headline would only show that replays get rejected, not that staleness is what rejects them. Header assertions come last in all three, so a lost fence reds on the outcome it is about rather than on a wire-format check. A mock cannot prove Tigris honors any of this, and the vendor requires a Single-region or Multi-region bucket for conditional operations, so against a Global or Dual-region bucket the fence is a silent no-op. The credentials-gated probe checks both arms against the real endpoint and cleans up unconditionally, including when an assertion fails, which is the case it exists to catch. It accepts 412 or 409 on the create-only arm because both mean the precondition was enforced and both are already classified as a loss. * fix(node): reconcile F-series tests and test seams with the rebased guard shape The rebase onto main carried main's #174 F-series guard tests forward. The branch's guard replaces main's locked/released/test_pre_unlock_gate Drop mechanics with close-on-drop (runtime) and leak (off-runtime), so two tests that asserted the replaced mechanics are reconciled: the off-runtime disposal test now asserts the leak observable (the slot never returns to idle) and the detached-unlock-returns-connection test is dropped, its invariant covered by the U-series drop-frees-the-lock gates. The pre-unlock gate seam is restored (test-only) so the mid-unlock cancellation tests keep their deterministic park. ipfs.rs tests are updated to the branch's sync test-client constructor and 4-arg RepoStore::new. Refs #279 * fix(node): address the 2026-08-10 review round on the advisory-lock series - P1: serve the receive-pack advertisement from a non-mutating snapshot instead of acquire_fresh, so the unlocked advertisement cannot delete or swap the live repo directory under a concurrent guarded write (the stale-ETag success ordering). acquire_fresh loses its only production caller and is removed. - P1: read-gate, rate-limit, bound, and cancellation-clean the close_issue pre-lock snapshot, so a signed non-owner cannot drive unbounded Tigris downloads and blocking extractions, and an abandoned extraction no longer leaks its temp dir. - P1: classify raw 409/412 responses as a lost conditional write via SdkError::raw_response() in both upload and the publish supersede-retry, so an unparsable error body cannot acknowledge a write that was never published. - P2: wrap the cold-cache under-lock download failure in RepoUnavailable so it sheds as a retryable 503 like the HEAD arm, not a permanent 500. - P2: surface a refused create-only fork upload as PreconditionLost and refuse the fork, so an orphan archive cannot shadow a fork's DB record. Refs #279 * fix(node): record why the close-issue read gate cannot carry a synced row The gate added for the pre-lock snapshot reads a repo row's own visibility rules and public flag. A row synced from a peer is stored public and carries none of the owner's rules, so for that class the gate can only return allow. Refusing instead would deny the repo's real owner and the issue's real author on any node whose only copy is a synced one, and it would not buy the protection it appears to, because a synced row's recorded owner comes from the peer that sent it. Every other read gate in the API reaches the same verdict for such a row, so this is left as is and stated at the call site rather than special-cased here. The test that covered the gate seeds a locally created repo and cannot observe this. Adds one that pins the property directly: the gate allows an arbitrary caller on a synced row, and the handler refuses that caller anyway, because the owner-or-author check is what decides this route. * fix(node): address jatmn review on merged advisory-lock head Align .env.example db pool with validate floor, move snapshot cleanup into the blocking extract task, remove fork clones on PreconditionLost, restore upload_site_reached test plumbing, and use owner DIDs in receive-pack tests after #330 owner-push default. * fix(node): cap close_issue snapshot work on the read pool Take a git_read_semaphore permit before the non-owner author snapshot so parallel close attempts cannot each drive a full archive extraction while still under the hourly rate bucket. Format owner-DID test helpers. * fix(node): mark close_issue mirror visibility gate for push hook Annotate list_visibility_rules with mirror-rows-handled so the pre-push surface detector accepts the synced-row read gate. * fix: unblock pre-push after main merge Read CappedBody.text in gl whoami (#381 shape), restore legacy CID sweep wiring dropped in the merge, route rate-limiter cleanup through AppState, and allow build_lock_pool for test-only callers. * fix(node): carry transient acquire failures through every write handler Extend acquire_write_app_error so RepoBusy and RepoUnavailable map to fixed-body 503s on issue and pull paths, matching receive-pack. Fix the flaky u2 drain exhaustion test via a drain_faults seam and add handler regressions for contention and storage unavailability on create_issue. * fix(node): drop unused logcap helper after drain_faults seam * fix(node): route fork disk paths through validated_repo_disk_path CodeQL flagged remove_dir_all on a user-derived path in fork_repo. Use the same three-layer barrier as RepoStore::local_path and reject names the validator refuses before clone. * fix(node): harden release-upload outcomes and replication tail gating Treat timed-out or failed archive uploads as non-durable in into_result, gate pinata/gossip on confirmed publish durability while keeping inv22 spawn order, split lock-pool vs main DB pool validation, map LockPoolBusy through From<anyhow::Error>, and attach Retry-After on repo write contention errors. * fix(node): close jatmn round on publish authority and durability gating Revoke under-lock refresh swaps when the advisory lock ends, fail closed on unrecorded release outcomes before Pinata/gossip, compensate fork archives when create_repo fails, recompute acquire backoff from the live deadline, and route publish swaps through validated_repo_disk_path for CodeQL. * fix(node): harden issue close/create and fork publish paths on #285 Move close_issue rate limiting after the read gate so denied readers see 404 not 429, roll back local issue refs when publish refuses, re-check swap authority under publish_lock, propagate fork upload failures before DB insert, and update lock-pool config help text. * fix(node): close write-attempt ownership gaps on advisory-lock release Make publish swaps claim-exclusive via CAS, invalidate the local read cache on definite publish refusals, harden fork create compensation and clone cleanup, and keep snapshot temp dirs owned through the async handoff. * fix(node): gate replication tail on publish durability and split close_issue limits Wait for a confirmed release outcome before post_receive_replication_tail runs any walks or coalescing work. Give close_issue its own per-IP rate bucket so it cannot drain receive-pack quota. Add regression tests for both paths. * fix(node): add ValidatedRepoDiskPath for CodeQL and reconcile tail durability Introduce a validated-path newtype with the same inline join and component walk main uses, so path-injection sinks only accept sanitised repo paths. Reconcile the publish-durability gate with the disconnect-safe tail spawn via PublishDurabilitySlot and explicit UploadUnknowable handling. * fix(node): close three jatmn findings on advisory-lock PR Retry fork create confirmation before compensating or returning, schedule background recovery when the lookup stays unavailable, install publish durability outcomes before the recorded flag, and size the lock pool for non-push mutations (default 40). * fix(node): verify adversarial findings on advisory-lock PR Fail closed when publish durability never records, bound pg_advisory_unlock with the same transfer budget as upload, and cap close_issue pre-lock snapshots at git_acquire_timeout_secs instead of the write-lock transfer bound. * fix(node): compensate under the write lock and harden tail durability handoff Run create_issue rollback through release_compensating so issue refs are deleted only while the advisory lock is still held and only on definite publish refusals, not UploadUnknowable. Replace the tail durability slot's async try_lock drop with a std mutex so cancellation during polling cannot skip installing UploadUnknowable. * fix(node): refuse stale under-lock refresh and gate tail durability on release Under-lock GET failures now always shed as RepoUnavailable, even when a local copy exists, so a transient download error cannot publish over a newer stored generation. PublishDurabilitySlot only synthesizes UploadUnknowable after release starts, and receive-pack marks that boundary before spawning the tail. Adds regression tests for both paths; inv22 F4 gate accepts the release wrapper. * style(node): rustfmt advisory-lock gap-closure commits * fix(node): satisfy clippy on stale-local download test path join * feat(node): add a publish attempt/stage boundary The four lifecycle findings on this branch share one root cause: a resource was tracked by a coarse state ("release started", "the path exists", "a row with this name exists") when the safety question was about the IDENTITY and the PUBLICATION STAGE of one specific write attempt. This is the vocabulary that makes those questions answerable: - `PublishAttemptId` — minted before the request is built, carried with the bytes as object user metadata, read back off the store to decide whether what is published is THIS attempt's work. That turns "did my request succeed", which a lost response makes undecidable, into "are the published bytes mine", which the store can answer. - `PublishStage` / `PublishStageCell` — how far one attempt got, observable from OUTSIDE the future doing the work. A cancelled handler never returns an outcome, so the stage it had reached is the only thing that can classify it. - `UploadError` split by what a failure ENTITLES A CALLER TO DO rather than by which HTTP status came back. Destructive compensation is licensed only by `proves_not_published()`. Deliberately free of any object-storage type. #79 deletes `git/tigris.rs` and replaces it with a BlobStore layer; this module is what survives that swap, with each backend supplying only its own error classifier. * feat(node): stamp uploads with an attempt id and preserve dispatch ambiguity `upload` now reports its progress into a `PublishStageCell` and stamps every PUT with an attempt id in object user metadata, so both of the questions a cancelled or unanswered publish raises become answerable: - `PreparingArchive` is marked before the blocking compression and `PutDispatched` immediately before `send()`. The conditional PUT is not constructed until compression returns, so a caller abandoned in that window definitely never attempted publication — a fact nothing could previously observe, because the only report was the return value of a future that no longer existed. - `attempt_landed()` HEADs the key and compares the stored attempt id, which is what lets a client whose response was lost recover its own committed write instead of guessing. `UploadError::Other` is replaced by `NotPublished` (proven never committed) and `Ambiguous` (may have been dispatched and may have committed). The single backend-aware classifier is `classify_put_failure`: a 4xx is an answer the server gave before storing anything and proves non-publication; a 5xx does not, and neither does a timeout, a dispatch failure, or a response the SDK could not read. `SdkError` is `#[non_exhaustive]`, so the fallback arm is the cautious one. `delete` is replaced by `delete_if_attempt_matches`, which reads the attempt off the object and fences the DELETE with `If-Match` on the generation it came from. A second name lookup before an unconditional delete would only narrow the window in which a successor's object can be erased; the conditional delete closes it. * fix(node): key durability, cache validity and fork cleanup on the attempt Closes the four lifecycle findings by consuming the attempt/stage boundary at the sites that were deciding on coarse state. P1 — cancellation before PUT dispatch is not publish durability. `PublishDurabilitySlot` is armed on the guard's publish STAGE rather than on a "release started" flag, and its `Drop` classifies an abandoned handler by how far the attempt actually got: a cancellation during compression records a definite refusal, a cancellation after dispatch records unknowable, and a cancellation after the store acknowledged records `Released` (which the flag could never say). `publish_durability_confirmed` now requires `Released`, so the detached tail no longer does IPFS/Pinata/P2P/GraphQL/Arweave/peer work for refs that exist only in a rejected local write. A guard with no storage backend seeds `NoBackend`, keeping Tigris-less deployments' tails running. P1 — an unknowable generation is quarantined, not served as ordinary cache. `release` writes a sidecar marker naming the unresolved attempt beside the live tree (never inside it: it would be tarred into the next archive), and `acquire`'s existing-path fast path reconciles it before serving. Only the store confirming it holds that attempt lifts the quarantine; anything else is a retryable `RepoUnavailable`. The tree is deliberately NOT deleted — the PUT may have landed and this can be the only local copy. A confirmed publish, an under-lock refresh, and cache invalidation each clear the marker, so the quarantine is a bounded refusal rather than a standing outage. The release-side timeout arm also stops treating every stall alike: a bound that expires at `PreparingArchive` is a definite non-publication, and a bound that expires after dispatch spends one short, separately bounded HEAD trying to reconcile the attempt before falling back to unknowable. P1 — fork confirmation and cleanup bind to the creating attempt. The DB row id is minted up front and used as the attempt id, so the object's metadata, the clone's sidecar stamp and the row all name one attempt. `confirm_fork_repo_row` looks the row up BY ID and reports Ours / Foreign / Absent, so a concurrent create, mirror registration or retry can no longer be returned as this request's own commit. Every destructive step — `compensate_fork_archive`, its background retries, and `ForkCloneGuard::drop` — is conditional on the resource still belonging to this attempt, so recovery that observed `None` before a successor committed can no longer erase that successor's archive or directory afterwards. P2 — response-loss ambiguity is preserved rather than compensated. Guarded writes map an ambiguous publish to `UploadUnknowable` + quarantine instead of `UploadFailed` + cache invalidation + the caller's undo, so `create_issue` no longer deletes an issue ref whose archive is durable. Fork creation reconciles by attempt id: a create-only PUT that committed and lost its response is RECOVERED and the row inserted, and an unresolved one keeps its only local clone and refuses retryably. The fork name is no longer fenced behind the attempt's own orphan. Tests, each RED-checked by reverting its guard: - slot_drop_during_compression_records_a_definite_non_publication - slot_drop_after_dispatch_records_unknowable - slot_drop_after_the_store_acknowledged_records_released - slot_drop_with_no_storage_backend_records_released - publish_durability_confirmed_accepts_only_released - publish_durability_confirmed_refuses_quickly_after_unrecorded_slot_drop - receive_pack_cancelled_during_compression_publishes_nothing_and_runs_no_tail (+ receive_pack_that_completes_its_publish_still_runs_the_tail as control) - a_bound_that_expires_before_dispatch_is_a_definite_failure - an_unresolved_publish_quarantines_the_tree_and_a_later_read_refuses - a_quarantined_tree_is_served_once_the_store_confirms_the_attempt - the_next_write_clears_an_inherited_quarantine - a_confirmed_publish_leaves_the_tree_readable (control) - a_guarded_write_whose_response_is_lost_is_not_compensated - a_put_that_commits_and_loses_its_response_is_ambiguous_and_reconcilable - a_closed_response_with_no_http_status_is_ambiguous - upload_classifies_404_as_a_definite_non_publication - upload_classifies_500_as_ambiguous_not_definite_failure - a_generation_that_moves_between_the_head_and_the_delete_is_not_deleted - an_attempts_own_object_is_deleted_and_a_foreign_one_is_not - fork_confirmation_never_claims_a_concurrent_attempts_row - fork_confirmation_recognizes_this_attempts_own_committed_row - fork_confirmation_reports_absent_when_nothing_owns_the_name - fork_clone_guard_leaves_a_successors_mirror_alone - fork_recovery_resuming_after_a_successor_deletes_neither_object_nor_path - fork_compensation_removes_what_this_attempt_still_owns - fork_publish_that_loses_its_response_stays_recoverable - a_fork_whose_publish_lost_its_response_is_recovered_not_fenced - a_fork_whose_publish_is_unresolved_keeps_its_clone_and_refuses_retryably - an_ordinary_fork_publishes_and_commits (control) - mock_round_trips_the_attempt_metadata_a_put_stamped The S3 mock gains attempt metadata, conditional DELETE, a commit-then-lose- the-response mode, a delivered-but-not-committed mode, and a per-key object store (fork creation touches the source and fork keys in one request, and a single slot would have let an assertion about one silently read the other). The compression seam is a condvar gate so no test holds a guard across an await. * fix(node): settle the cache contract on the abandoned path and the attempt The publication boundary this branch introduced was applied at call sites that each had to be remembered, and five were not. Close the eight gaps that survived, grouped by what actually failed rather than by symptom. The marker's lifecycle was answered by the wrong event in three places. The swap that replaces a live tree now owns the clear, so a cache-miss download no longer leaves a stale marker that fences the archive it just installed, and the under-lock refresh no longer lifts a quarantine on a HEAD that downloaded nothing. init clears a predecessor's marker. A cancelled handler never reached release, so it applied the tail contract and not the cache contract. RepoWriteGuard::drop now settles the tree it abandoned, keyed on how far the publish actually got, and a guard that handed out its path for writing marks the tree whatever stage it reached. A marker naming no attempt is a definite non-publication rather than a permanent refusal: it invalidates and the next read serves the stored generation, so an abandoned first push cannot wedge a repo at 503. Only an unreadable sidecar still fails shut. Both of the store's live-path hand-outs now go through one confirming helper, so read_snapshot can no longer return an unconfirmed tree as an ordinary snapshot. The gate binds ahead of the lazy-migration spawn, because a quarantined tree backfilled to object storage by an IfAbsent PUT would publish exactly the refs the marker exists to withhold. The marker no longer rests on a single fs::write; an unwritable sidecar falls back to an in-memory record that reads consult. Fork cleanup claims ownership by renaming the stamp before it touches the tree, so a successor that claims the same path cannot lose its directory to a predecessor's check-then-unlink, and the directory-absent path no longer unlinks a stamp it does not own. A create-only precondition loss whose reconciliation HEAD fails is refused retryably instead of reported as a permanent name conflict, and its clone is released so the retry can re-clone rather than finding the destination occupied. The pin sweep skips a tree whose publish is unresolved, so provider CIDs are not rewritten from refs that may never become durable. That is the sidecar marker, distinct from the operator-level quarantined row flag. 26 tests, each written and observed failing before its fix. Suite 1211 passed, 0 failed, 1 ignored; fmt, clippy -D warnings and --locked clean. * fix(node): settle the abandoned tree at the disconnect, not at reaper exit The settlement that marks an abandoned write ran from RepoWriteGuard's Drop, and on a client disconnect that guard is held by the Arc clone riding the admission guard into the detached reaper, so it ran only once the process group had been reaped. Reads take no advisory lock and consult only the sidecar, so for the width of that window a concurrent fetch was served the abandoned refs off the live tree with no marker beside it. Confirmed by execution before the fix: no marker on disk, acquire returning the live path, and the abandoned ref still on the tree it served. Only the lock release has to wait for the reaper; the settlement does not. TreeSettlement is a token carrying the path, the publish stage and the hand-out flag, holding the exhaustive stage match that used to live in Drop. The handler takes it out of the guard before the guard is shared with the reaper, so on a disconnect the handler future drops first and the marker lands at the disconnect instant while the reaper still holds the lock. A successful release disarms it. Drop still settles a guard nobody took the token from, so the arms have one home. The regression test reads inside the window and asserts the abandoned refs are not served. It samples pg_locks from an independent session first, so a run where the reaper had already finished reports itself inconclusive rather than passing for the wrong reason. Suite 1212 passed, 0 failed, 1 ignored. The five settlement mutations are load-bearing, including one that reverts this split and reddens the window test. * fix(node): let exactly one classifier settle an abandoned tree Moving settlement into a token the handler holds left two classifiers able to run for one write. release() classifies the tree and only then awaits the unlock, while the handler disarms the token after that await returns, so a cancel in the unlock gap dropped a still-armed token and settled a second time. After a definite refusal that is a quarantine marker written beside the tree release had just deleted, which sends every later read down the refuse-and-reconcile path for a repo whose state was already resolved. Observed before the fix: the tree gone, the marker there. The arm is now an AtomicBool shared between the guard and its token. release stores false at the classification point, above the unlock await, and settle takes the arm with compare_exchange, so whichever classifier gets there first is the only one that acts. tree_settled stays, and now has a test that says why: after the handler takes the settlement the arm is still true, so tree_settled is the only thing stopping the guard's own Drop, which on a disconnect runs in the detached reaper, from settling again at reaper exit. The reap-window test was passing for weaker reasons than it claimed. Its fake git took SIGTERM and let the group die, so the reaper could free the advisory lock before the probe read; it now traps TERM and re-parks so only SIGKILL ends it. It asserts the marker exists at the disconnect instant rather than only mentioning it, and the served-refs half no longer passes when acquire fails for an unrelated reason. Suite 1214 passed, 0 failed, 1 ignored. Nineteen mutations load-bearing, including one that reverts the release-side disarm and one that reverts the settlement split. * test(node): use the test's own pool now that Db::pool is gone --------- Co-authored-by: Kevin Codex <[email protected]>