Timeline

Two concurrent libfossil.Repo handles on the same SQLite file fail their writes with SQLITE_BUSY immediately, well under the configured busy_timeout=5000 window. Reproduces ~50% of the time on contended manifest.Checkin paths.

Root cause: db.WithTx opens BEGIN DEFERRED. Under WAL, a deferred transaction acquires SHARED on first read, then upgrades SHARED→RESERVED on first write. If two connections race that upgrade, SQLite's deadlock-avoidance returns SQLITE_BUSY *immediately*, bypassing busy_timeout entirely (documented behavior, sqlite.org/c3ref/busy_timeout.html).

Fix: prepend _txlock=immediate to the DSN in both modernc and ncruces drivers. Every BEGIN now acquires RESERVED upfront, so the second writer blocks at BEGIN — where busy_timeout *does* apply — and the two transactions serialize cleanly.

Regression tests added to both driver packages: - TestBuildDSN_TxlockImmediate covers the DSN shape across empty/non-empty pragma maps. - TestConcurrentWritersDoNotRace reproduces the issue-#33 contention pattern (two conns, SELECT then INSERT under DEFERRED) over 25 iterations and asserts zero SQLITE_BUSY. Without the fix, this test fails 25/25; with the fix, 0/25.

Closes #33. check-in: 202576c58c user: [email protected] tags: trunk, v0.6.3, db/driver/ncruces/v0.2.0, db/driver/modernc/v0.2.0

Repo.Commit was a thin pass-through to manifest.Checkin: callers who supplied a subset of the parent's tracked files got a partial manifest, so files tracked at the parent silently disappeared from the new commit. A checkout at the child's rev saw only the supplied subset — bones, EdgeSync, and sesh all hit this as silent data loss across slot commits.

Default behavior now matches `fossil ci`: when ParentID > 0 the parent's F-cards are merged into the commit's file set (caller wins on name collision; untouched parent files are loaded by UUID and carried forward).

For the rare case where a caller intentionally wants omission to mean deletion, CommitOpts.PartialManifest opts out of the merge. The internal diff/deletion tests that previously relied on the partial behavior now set this flag explicitly. A first-class deletion API may replace the escape hatch later.

Adds TestCommitPreservesParentFiles as the regression: parent [a,b,c], child supplies only [b'], child's ListFiles must return [a,b',c] with a and c keeping the parent's blob UUIDs. check-in: 862dcb20c3 user: [email protected] tags: trunk, v0.6.2

* feat: CreateOpts.ProjectCode — accept explicit project-code on Create

Downstream EdgeSync coordinates a shared project-code across sibling Fossil repos created on different machines so its NATS sync subject ("<prefix>.<project-code>.sync") matches across leaves. Today every Create generates a fresh project-code, so commits never propagate.

Add CreateOpts.ProjectCode: empty preserves current behavior; non-empty must be 40-char lowercase hex (^[0-9a-f]{40}$) — the same format generated project-codes use. Validation runs at the top of libfossil.Create before any disk I/O, so invalid input cannot leave a half-written file behind. Server-code stays auto-generated.

Threaded through repo.Create / repo.CreateWithEnv / db.SeedConfig. SeedConfig validates as defense-in-depth (public package, multiple callers possible).

Clone accepting ProjectCode is out of scope; will follow separately if EdgeSync needs it.

* test: round-trip explicit ProjectCode through fossil rebuild

Asserts that vanilla `fossil` (2.28) accepts a Go-side repo created with an explicit ProjectCode, and that `fossil sql` reads back the same value we set. Closes the gap where the existing TestCreate_FossilValidation only covered the generated-project-code path. check-in: 6b66ea6b2d user: [email protected] tags: trunk, v0.6.1

Closes #27.

libfossil opens repos in WAL journal mode but the Close chain never ran wal_checkpoint, so closed repo files were a 4 KiB stub plus a fat -wal sidecar. Vanilla fossil's validity probe stats the main file before consulting the WAL → repo rejected as "not a valid repository".

- Adds a typed CheckpointMode (Passive / Full / Restart / Truncate) and a Checkpoint(mode) method on the public Repo (mirrored through internal repo and db.DB). CheckpointMode and its constants are re-exported from the package root so callers don't have to import db. - db.DB.Close now runs PRAGMA wal_checkpoint(TRUNCATE) before closing the underlying *sql.DB. Checkpoint errors are joined with the close error via errors.Join so the connection is always closed regardless of checkpoint outcome. - Skips the close-time checkpoint on the WASM/WASI build path where WAL isn't used. check-in: faf57ad9c7 user: [email protected] tags: trunk, v0.6.0

Repo.Diff has a multi-file return shape ([]DiffEntry) but previously rejected filePath == "" with "filePath is required". Consumers that wanted a whole- checkin diff (the equivalent of `git diff revA..revB` or `fossil diff --from revA --to revB`) had to enumerate changed files themselves and call Diff per file; EdgeSync's Agent.Diff is the in-tree example of that workaround.

This relaxes the guard: when filePath == "", Diff now returns one DiffEntry per file that changed between ridA and ridB, sorted by Name for deterministic ordering. The non-empty filePath path is unchanged. Identical checkins return ([]DiffEntry{}, nil).

Whole-checkin enumeration is name-keyed (union of files across both manifests where the content UUID differs or the file exists on only one side). Proper rename detection via mlink (so a pure rename produces a single entry rather than an add+delete pair) and permission-only change reporting are deferred to a follow-up; the limitation is documented on Diff's godoc and called out in the PR description. The single-file path is unaffected.

The previous TestDiff_EmptyFilePath (which asserted the rejection error) is replaced with positive coverage for identical checkins, single-file change, multi-file change, addition-only, and deletion-only.

closes #24 check-in: 764a82ecdb user: [email protected] tags: trunk, v0.5.0

Consumers (bones, EdgeSync) currently bypass the methods-only Repo API to convert RIDs to UUIDs by reaching into r.DB() and running raw SQL. The lookup is well-defined and recurs throughout the libfossil internals; expose it as a first-class method.

Add Repo.UUIDFromRID(rid int64) (string, error) backed by a small internal helper Repo.UUIDByRID on internal/repo so the SELECT lives in one place. Missing rids surface ErrArtifactNotFound (wraps sql.ErrNoRows behind a typed sentinel) with the rid in the message.

closes #23 check-in: 61aa4dd022 user: [email protected] tags: trunk

* docs: add cross-repo release automation design spec

Designs an automated release pipeline across libfossil → edgesync → bones: - repository_dispatch chain replaces manual go.mod bumps - linear-only graph (libfossil does not dispatch directly to bones) - per-repo `make ci` mirrors GitHub CI verbatim; `make ci-fast` for pre-push - `make release VERSION=...` helper with annotated tag template - GoReleaser auto-changelog (edgesync, bones); `gh release --generate-notes` for libfossil - 6-PR rollout: 3 hygiene PRs first, then 3 dispatch wiring PRs

See docs/superpowers/specs/2026-04-30-cross-repo-release-automation-design.md

* docs: add cross-repo release automation implementation plan

Breaks the design spec into 6 sequential PRs across libfossil, edgesync, and bones — 3 hygiene PRs (Makefile + pre-push hook + release helper) followed by 3 dispatch-wiring PRs.

Plan includes: - Phase 0 prerequisites (DISPATCH_PAT setup per upstream repo) - Per-PR branch creation, file-level steps, expected commands/outputs, commit messages, and PR-creation gh commands - Post-merge smoke tests for each receiver/sender combo - Final end-to-end cascade validation via v0.4.6-rc1 tag

See docs/superpowers/plans/2026-04-30-cross-repo-release-automation.md

* docs: correct bones make ci scope to include make check

bones' .github/workflows/ci.yml runs make check (fmt-check + vet + lint + race + todo-check) before the otel-tagged lanes. The earlier scope captured only the otel lines, missing the discipline suite that is the primary CI gate. Update both spec and plan to match the actual workflow.

The cross-repo leaf-binary integration step (sibling EdgeSync + make leaf) remains intentionally CI-only — it requires a sibling working tree that doesn't generalize to a single-repo make target.

* docs: capture libfossil deviation (no pre-push, no ci-fast)

PR #3 (#19, merged 2026-04-30) shipped a tighter scope than PR #1/#2 because libfossil's existing pre-commit is already comprehensive. Document the option (A) decision in both spec and plan, and add a multi-module note for the future case if pre-push is ever added.

* docs: add go get retry + actor field to receiver workflows

PR #4 (bones #90) added a follow-up commit per code review with these two improvements; capture them in spec/plan so PR #5's edgesync receiver picks them up automatically:

- 3-attempt retry with 15s backoff covers GOPROXY propagation race (dispatcher fires immediately on tag push, receiver may outrun the module proxy). workflow_dispatch fallback still exists for manual recovery. - Actor field plumbed payload→input→github.actor with triple fallback and rendered as 'Triggered by: @actor' in bump PR body for audit. check-in: 339643c721 user: [email protected] tags: trunk

On tag push (v*): 1. Create GitHub release with auto-generated notes from PRs/commits since previous tag. Source tarball auto-attached. 2. Dispatch repository_dispatch event to danmestas/EdgeSync to trigger a chore-bump PR there.

libfossil is a library — no binaries are published, so goreleaser is overkill. 'gh release create --generate-notes' provides equivalent auto-changelog and source archives in 5 lines.

Closes the cascade chain. Once merged, the next libfossil tag will fire PRs in EdgeSync (PR #5 receiver) and via that, in bones (PR #4 receiver). check-in: 38e4948972 user: [email protected] tags: trunk

* feat(make): add ci targets mirroring test.yml verbatim

ci aggregates 3 sub-targets matching the 3 GitHub Actions jobs: ci-default — modernc driver tests + DST + vet + cmd build ci-ncruces — ncruces driver tests with -tags test_ncruces ci-otel-target — observer/otel sub-module tests

All use GOWORK=off (matching CI). Existing test/test-drivers/test-otel targets are left untouched — they don't use GOWORK=off and have drifted from CI; ci-* are the single source of truth going forward.

No ci-fast / pre-push hook in this PR — libfossil's existing pre-commit hook (modernc + ncruces + vet + otel + build + SDK drift, ~45s) already exceeds what ci-fast would provide. Adding a pre-push would just double push cost without catching new bugs.

* feat(make): add release VERSION=... helper

Validates clean tree, runs make ci, opens \$EDITOR with auto-populated tag annotation template, creates annotated tag locally. Push is intentionally manual.

* feat: add release annotation template

Used by 'make release VERSION=...' as the body skeleton for annotated tags. check-in: 5fd8907820 user: [email protected] tags: trunk