Skip to content

fslock: build on freebsd, openbsd, netbsd and dragonfly again - #6

Merged
reltuk merged 1 commit into
dolthub:mainfrom
shrisha:unix-build-tag
Oct 18, 2026
Merged

reltuk merged 1 commit into
dolthub:mainfrom
shrisha:unix-build-tag

Conversation

@shrisha

@shrisha shrisha commented Oct 15, 2026

Copy link
Copy Markdown
Contributor

Fixes #5.

lock_unix.go is constrained to linux || darwin. On the other unixes, lock.go calls sysOpen, sysLock and sysUnlock, but nothing defines them, so the package doesn't compile there. That became a regression for Dolt users when dolthub/dolt#11935 swapped dolthub/fslock for this module. The old fslock_nix.go was constrained to darwin || dragonfly || freebsd || linux || netbsd || openbsd and used the same flock(2) calls, including LOCK_EX|LOCK_NB polling for waits that have a deadline. Since the Dolt v2.4.1 bump, GOOS=freebsd builds of Dolt-embedding programs fail. gastownhall/beads's freebsd/amd64 release build is one of them; it is the last red check on gastownhall/beads#7441.

Change

  • lock_unix.go: use fslock's constraint. The code is unchanged, because golang.org/x/sys/unix provides Flock, LOCK_*, EINTR and EWOULDBLOCK on all six systems.
  • README.md: list the supported systems.
  • ci.yml: add a cross job that runs go vet (which also type-checks the tests) for the four BSDs on amd64 and arm64 (dragonfly is amd64-only), so the constraint can't quietly narrow again. arm64 is included because x/sys reaches Flock differently per arch on openbsd.

Why not //go:build unix

#5 suggests unix, but it fails on aix (undefined: unix.Flock; x/sys has no Flock there). On solaris and illumos, Flock links to libc's flock, which illumos and recent Solaris have, but I haven't tested those. Matching fslock's list restores exactly what Dolt had before, and nothing more. I'm happy to widen it if you'd like.

Testing

  • go test -race -count=3 ./... passes on linux/amd64.

  • go vet ./... and go test -c are clean for linux, darwin, freebsd, openbsd, netbsd, dragonfly and windows (amd64), plus freebsd arm64, 386, arm and riscv64. aix still fails, which is expected.

  • Downstream: GOOS=freebsd GOARCH=amd64 CGO_ENABLED=0 go build ./cmd/bd on build(deps): bump the go-deps group across 1 directory with 18 updates gastownhall/beads#7441 fails with v0.1.1 (lock.go:272:12: undefined: sysOpen) and succeeds with this branch.

  • On real BSD kernels (run on my fork, via a fork-only vmactions workflow that is not part of this PR):

    • FreeBSD 15.1-RELEASE (go1.27.2): go vet, go test -v -count=3 and go test -race all pass.
    • OpenBSD 7.9 (go1.26.2): go vet and go test -v -count=3 pass.

    That covers all eight tests, including TestAnotherProcess (exclusion across processes), TestExclusion (two Locks in one process) and TestReplacedLockFile / TestRemoveWhileHeld, so the open-file-description semantics the package relies on hold there. If you want this in upstream CI, I'm happy to add the VM jobs; I left them out because they pull in a third-party action and take a few minutes per run.

After merge

Could you tag v0.1.2? Downstreams can then require it directly (MVS picks it over Dolt's v0.1.1) without waiting for a Dolt release. A follow-up Dolt bump would fix it for everyone else.

Thanks to @bee-ghosttrack for the diagnosis in #5.

This PR was prepared with the help of an AI coding agent (Claude Code). I reviewed the change, and the test results above are from real runs.

🤖 Generated with Claude Code

lock_unix.go was constrained to linux || darwin, so on freebsd, openbsd,
netbsd and dragonfly lock.go referenced sysOpen/sysLock/sysUnlock with
nothing defining them. github.com/dolthub/fslock, which this package
replaces in Dolt, built on all six with the same flock(2) calls, so Dolt
consumers lost those targets with the switch (dolthub#5).

Use the constraint fslock had rather than unix: x/sys/unix has no Flock
on aix, and on solaris it depends on libc's flock, which nobody has
tested here. Add a CI job that vets the four BSDs so the constraint
cannot quietly narrow again.

Fixes dolthub#5

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@shrisha

shrisha commented Oct 18, 2026

Copy link
Copy Markdown
Contributor Author

A quick ping on this: gastownhall/beads#7646 (the Dolt 2.4.1 bump for beads) is now approved, and its only remaining red check is the freebsd/amd64 release cross-compile, which needs this fix tagged as v0.1.2 (see "Known red" in its description). Dolt v2.5.0 still pins v0.1.1, so other projects that embed Dolt and release for the BSDs are in the same position.

@bheni, or @reltuk since you reviewed dolthub/dolt#11935: could one of you take a look? The CI run is also waiting for first-time-contributor approval. Happy to change anything.

@reltuk

reltuk commented Oct 18, 2026

Copy link
Copy Markdown
Contributor

Thank you for the contribution! I'll make sure this gets merged and the bump gets into dolt.

@reltuk
reltuk merged commit 5b3ec47 into dolthub:main Oct 18, 2026
10 checks passed
@reltuk

reltuk commented Oct 18, 2026

Copy link
Copy Markdown
Contributor

Dolt bump PR is at dolthub/dolt#12136

@reltuk

reltuk commented Oct 18, 2026

Copy link
Copy Markdown
Contributor

OK, the bump has landed in Dolt. It will be included in the next Dolt release, which should happen sometime this week.

gascityinc-randy Bot pushed a commit to gastownhall/beads that referenced this pull request Oct 18, 2026
file-locks v0.1.1 (pulled in transitively by Dolt 2.4.1) had no build
tags for freebsd/openbsd/netbsd/dragonfly, so the freebsd/amd64 release
cross-compile failed. dolthub/file-locks#6 (by @shrisha) adds them and
ships in v0.1.2.

Only file-locks moves in go.mod/go.sum; tools/nogo is unaffected. The
Nix vendorHash is recomputed for the new module set.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
gascity-eddie Bot pushed a commit to gastownhall/beads that referenced this pull request Oct 18, 2026
file-locks v0.1.1 (pulled in transitively by Dolt 2.4.1) had no build
tags for freebsd/openbsd/netbsd/dragonfly, so the freebsd/amd64 release
cross-compile failed. dolthub/file-locks#6 (by @shrisha) adds them and
ships in v0.1.2.

Only file-locks moves in go.mod/go.sum; tools/nogo is unaffected. The
Nix vendorHash is recomputed for the new module set.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
gascityinc-randy Bot pushed a commit to gastownhall/beads that referenced this pull request Oct 23, 2026
file-locks v0.1.1 (pulled in transitively by Dolt 2.4.1) had no build
tags for freebsd/openbsd/netbsd/dragonfly, so the freebsd/amd64 release
cross-compile failed. dolthub/file-locks#6 (by @shrisha) adds them and
ships in v0.1.2.

Only file-locks moves in go.mod/go.sum; tools/nogo is unaffected. The
Nix vendorHash is recomputed for the new module set.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
timothybrush pushed a commit to timothybrush/beads that referenced this pull request Oct 23, 2026
…bility fixes (gastownhall#7646)

* build(deps): bump the go-deps group across 1 directory with 20 updates

Bumps the go-deps group with 20 updates in the / directory:

| Package | From | To |
| --- | --- | --- |
| [github.com/anthropics/anthropic-sdk-go](https://github.com/anthropics/anthropic-sdk-go) | `1.71.0` | `1.80.0` |
| [github.com/dolthub/driver/v2](https://github.com/dolthub/driver) | `2.2.0` | `2.4.1` |
| [github.com/gowebpki/jcs](https://github.com/gowebpki/jcs) | `1.0.1` | `1.0.2` |
| [go.opentelemetry.io/otel](https://github.com/open-telemetry/opentelemetry-go) | `1.46.0` | `1.47.0` |
| [go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetrichttp](https://github.com/open-telemetry/opentelemetry-go) | `1.46.0` | `1.47.0` |
| [go.opentelemetry.io/otel/exporters/stdout/stdoutmetric](https://github.com/open-telemetry/opentelemetry-go) | `1.46.0` | `1.47.0` |
| [go.opentelemetry.io/otel/exporters/stdout/stdouttrace](https://github.com/open-telemetry/opentelemetry-go) | `1.46.0` | `1.47.0` |
| [go.opentelemetry.io/otel/metric](https://github.com/open-telemetry/opentelemetry-go) | `1.46.0` | `1.47.0` |
| [go.opentelemetry.io/otel/sdk](https://github.com/open-telemetry/opentelemetry-go) | `1.46.0` | `1.47.0` |
| [go.opentelemetry.io/otel/sdk/metric](https://github.com/open-telemetry/opentelemetry-go) | `1.46.0` | `1.47.0` |
| [go.opentelemetry.io/otel/trace](https://github.com/open-telemetry/opentelemetry-go) | `1.46.0` | `1.47.0` |
| [golang.org/x/sync](https://github.com/golang/sync) | `0.23.0` | `0.24.0` |
| [golang.org/x/sys](https://github.com/golang/sys) | `0.48.0` | `0.49.0` |
| [golang.org/x/term](https://github.com/golang/term) | `0.46.0` | `0.47.0` |
| [github.com/charmbracelet/x/ansi](https://github.com/charmbracelet/x) | `0.11.8` | `0.11.9` |
| [github.com/dolthub/dolt/go](https://github.com/dolthub/dolt) | `0.40.5-0.20260806213044-796d07497741` | `0.40.5-0.20261002233748-c6af56fead88` |
| [github.com/dolthub/go-mysql-server](https://github.com/dolthub/go-mysql-server) | `0.20.1-0.20260805191915-e5eafe0da809` | `0.20.1-0.20261001232740-19405eb6f203` |
| [github.com/dolthub/vitess](https://github.com/dolthub/vitess) | `0.0.0-20260728212736-0542037326d7` | `0.0.0-20260930232419-a6b7b0c65d17` |
| [golang.org/x/mod](https://github.com/golang/mod) | `0.40.0` | `0.41.0` |
| [golang.org/x/net](https://github.com/golang/net) | `0.58.0` | `0.59.0` |

Updates `github.com/anthropics/anthropic-sdk-go` from 1.71.0 to 1.80.0
- [Release notes](https://github.com/anthropics/anthropic-sdk-go/releases)
- [Changelog](https://github.com/anthropics/anthropic-sdk-go/blob/main/CHANGELOG.md)
- [Commits](anthropics/anthropic-sdk-go@v1.71.0...v1.80.0)

Updates `github.com/dolthub/driver/v2` from 2.2.0 to 2.4.1
- [Release notes](https://github.com/dolthub/driver/releases)
- [Commits](dolthub/driver@v2.2.0...v2.4.1)

Updates `github.com/gowebpki/jcs` from 1.0.1 to 1.0.2
- [Release notes](https://github.com/gowebpki/jcs/releases)
- [Commits](gowebpki/jcs@v1.0.1...v1.0.2)

Updates `go.opentelemetry.io/otel` from 1.46.0 to 1.47.0
- [Release notes](https://github.com/open-telemetry/opentelemetry-go/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-go/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-go@v1.46.0...v1.47.0)

Updates `go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetrichttp` from 1.46.0 to 1.47.0
- [Release notes](https://github.com/open-telemetry/opentelemetry-go/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-go/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-go@v1.46.0...v1.47.0)

Updates `go.opentelemetry.io/otel/exporters/stdout/stdoutmetric` from 1.46.0 to 1.47.0
- [Release notes](https://github.com/open-telemetry/opentelemetry-go/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-go/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-go@v1.46.0...v1.47.0)

Updates `go.opentelemetry.io/otel/exporters/stdout/stdouttrace` from 1.46.0 to 1.47.0
- [Release notes](https://github.com/open-telemetry/opentelemetry-go/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-go/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-go@v1.46.0...v1.47.0)

Updates `go.opentelemetry.io/otel/metric` from 1.46.0 to 1.47.0
- [Release notes](https://github.com/open-telemetry/opentelemetry-go/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-go/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-go@v1.46.0...v1.47.0)

Updates `go.opentelemetry.io/otel/sdk` from 1.46.0 to 1.47.0
- [Release notes](https://github.com/open-telemetry/opentelemetry-go/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-go/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-go@v1.46.0...v1.47.0)

Updates `go.opentelemetry.io/otel/sdk/metric` from 1.46.0 to 1.47.0
- [Release notes](https://github.com/open-telemetry/opentelemetry-go/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-go/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-go@v1.46.0...v1.47.0)

Updates `go.opentelemetry.io/otel/trace` from 1.46.0 to 1.47.0
- [Release notes](https://github.com/open-telemetry/opentelemetry-go/releases)
- [Changelog](https://github.com/open-telemetry/opentelemetry-go/blob/main/CHANGELOG.md)
- [Commits](open-telemetry/opentelemetry-go@v1.46.0...v1.47.0)

Updates `golang.org/x/sync` from 0.23.0 to 0.24.0
- [Commits](golang/sync@v0.23.0...v0.24.0)

Updates `golang.org/x/sys` from 0.48.0 to 0.49.0
- [Commits](golang/sys@v0.48.0...v0.49.0)

Updates `golang.org/x/term` from 0.46.0 to 0.47.0
- [Commits](golang/term@v0.46.0...v0.47.0)

Updates `github.com/charmbracelet/x/ansi` from 0.11.8 to 0.11.9
- [Commits](charmbracelet/x@ansi/v0.11.8...ansi/v0.11.9)

Updates `github.com/dolthub/dolt/go` from 0.40.5-0.20260806213044-796d07497741 to 0.40.5-0.20261002233748-c6af56fead88
- [Release notes](https://github.com/dolthub/dolt/releases)
- [Commits](https://github.com/dolthub/dolt/commits)

Updates `github.com/dolthub/go-mysql-server` from 0.20.1-0.20260805191915-e5eafe0da809 to 0.20.1-0.20261001232740-19405eb6f203
- [Release notes](https://github.com/dolthub/go-mysql-server/releases)
- [Commits](https://github.com/dolthub/go-mysql-server/commits)

Updates `github.com/dolthub/vitess` from 0.0.0-20260728212736-0542037326d7 to 0.0.0-20260930232419-a6b7b0c65d17
- [Commits](https://github.com/dolthub/vitess/commits)

Updates `golang.org/x/mod` from 0.40.0 to 0.41.0
- [Commits](golang/mod@v0.40.0...v0.41.0)

Updates `golang.org/x/net` from 0.58.0 to 0.59.0
- [Commits](golang/net@v0.58.0...v0.59.0)

---
updated-dependencies:
- dependency-name: github.com/anthropics/anthropic-sdk-go
  dependency-version: 1.80.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: github.com/dolthub/driver/v2
  dependency-version: 2.4.1
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: github.com/gowebpki/jcs
  dependency-version: 1.0.2
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: go-deps
- dependency-name: go.opentelemetry.io/otel
  dependency-version: 1.47.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetrichttp
  dependency-version: 1.47.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: go.opentelemetry.io/otel/exporters/stdout/stdoutmetric
  dependency-version: 1.47.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: go.opentelemetry.io/otel/exporters/stdout/stdouttrace
  dependency-version: 1.47.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: go.opentelemetry.io/otel/metric
  dependency-version: 1.47.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: go.opentelemetry.io/otel/sdk
  dependency-version: 1.47.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: go.opentelemetry.io/otel/sdk/metric
  dependency-version: 1.47.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: go.opentelemetry.io/otel/trace
  dependency-version: 1.47.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: golang.org/x/sync
  dependency-version: 0.24.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: golang.org/x/sys
  dependency-version: 0.49.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: golang.org/x/term
  dependency-version: 0.47.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: github.com/charmbracelet/x/ansi
  dependency-version: 0.11.9
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: go-deps
- dependency-name: github.com/dolthub/dolt/go
  dependency-version: 0.40.5-0.20261002233748-c6af56fead88
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: go-deps
- dependency-name: github.com/dolthub/go-mysql-server
  dependency-version: 0.20.1-0.20261001232740-19405eb6f203
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: go-deps
- dependency-name: github.com/dolthub/vitess
  dependency-version: 0.0.0-20260930232419-a6b7b0c65d17
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: go-deps
- dependency-name: golang.org/x/mod
  dependency-version: 0.41.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
- dependency-name: golang.org/x/net
  dependency-version: 0.59.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: go-deps
...

Signed-off-by: dependabot[bot] <support@github.com>
(cherry picked from commit 1a1fd25)

* build(nix): move the flake to nixos-26.05 for go 1.26.5+

dolthub/driver v2.4.1 declares go 1.26.5, but nixos-25.11 is frozen
(last commit 2026-06-30) at go 1.26.4. default.nix's postPatch can only
lower the root module's go directive, not a dependency's, so the Nix
build failed with "requires go >= 1.26.5 (running go 1.26.4;
GOTOOLCHAIN=local)". nixos-26.05 ships go 1.26.8 and keeps
buildGo126Module and go_1_26, and the weekly flake-lock update can
track it.

Recompute vendorHash for the bumped go.sum.

Verified with nixos/nix: nix flake check passes, nix run .#default --
--help works, and the devShell reports go1.26.8.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* build(nogo): align tools/nogo x/mod, x/sync, x/text with the root module

TestModuleRequirementsAgreeWithRoot (//tools/nogo:nogo_test) requires
the analyzer module's shared requirements to match go.mod. The go-deps
bump moved the root to x/mod v0.41.0, x/sync v0.24.0 and x/text v0.42.0.

After the rebase onto main, gastownhall#6355 had already raised tools/nogo to
x/mod v0.41.0 and x/text v0.42.0, so this commit now moves only x/sync.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* fix(storage): compute the next version revision with a scalar subquery

go-mysql-server's aggregate-over-index rewrite
(dolthub/go-mysql-server#3907, pulled in by the driver v2.4.1 bump)
returns zero rows, not one NULL row, for an ungrouped MAX/MIN over an
empty primary-key prefix:

  SELECT COALESCE(MAX(revision), 0) + 1 FROM issue_versions WHERE issue_id = ?

so every first mint on an issue failed with "versioned history: compute
next revision for X: sql: no rows in result set", breaking the
versioned-history contract tests on the embedded, served, server and UoW
tiers. Wrap the aggregate in a scalar subquery: the outer SELECT has no
FROM and yields exactly one row on every engine, including dolt 2.2.0.

This is the only production query with the affected shape; other
MAX/MIN uses are grouped, unfiltered, range-filtered, or on a secondary
index. Versioned history is only enabled in tests today, so no release
is affected.

Add TestEmbeddedFirstMintOnEmptyHistoryIsRevisionOne, which fails on the
old query and pins that the first mint is revision 1 and the second 2.

Main's gastownhall#6661 pins the old mint SQL in two sqlmock expectations
(version_history_preenable_scan_test.go); move both to the
scalar-subquery literal, and name recordVersionAtInTx, where the read now
lives, in the single-writer comment.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* build(dolt): raise the Dolt CLI, container and Bazel pins to 2.4.1

Carry gastownhall#7254's pin bump onto the driver v2.4.1 upgrade. The pinned
dolt 2.2.0 CLI cannot write storage the 2.4.1 engine created: after a
`bd init` through driver v2.4.1, `dolt sql -q "INSERT ..."` fails with
"table has unknown fields" on dolt 2.2.0 and 2.3.1 and succeeds on
2.4.1. That broke the protocol, proxied and TestEmbeddedInit* lanes
wherever a test shells out to `dolt sql`.

The pin policy in install-dolt.sh / dolt.md requires the DOLT_RESET
('--hard') measurement before raising the pin. Measured 2026-10-10:
2.3.1 8/150 broken (positive control), 2.4.1 0/250 over two runs, the
embedded engine via driver v2.4.1 0/60. The docs keep the 2.3.x rows and
the "raise only once measured" rule and add the 2.4.1 row.

- install-dolt.sh, pull-dolt-image.sh, testdoltcommon.go, dolt.bzl
  (DOLT_VERSION plus release digests; linux-amd64 re-hashed locally) and
  the script test constants move to 2.4.1.
- doltversion.RecommendedMin is 2.4.1, documented with the unknown-fields
  evidence; it stays a recommendation, not a hard floor.
- Three fixes gastownhall#7254 lacked: the doltversion cache tests derive their
  "recent version" stub from RecommendedMin instead of 2.1.0 (now below
  the minimum, which failed the Windows lane); the server fingerprint
  golden takes dolt 2.4.x's default sql_mode; and doltSQLServerImage in
  scripts/repo_helpers_test.go, where main moved it from
  pull_dolt_image_test.go.
- Refresh comments that called 2.2.0 the current CLI pin.

Release note: managed proxied-server mode now recommends dolt >= 2.4.1,
and server-mode users on dolt 2.4.x get the stricter default sql_mode
(adds NO_ZERO_IN_DATE, NO_ZERO_DATE, ERROR_FOR_DIVISION_BY_ZERO).

Co-authored-by: coffeegoddd <dustin@dolthub.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* docs(changelog): note the Dolt 2.4.1 engine's descending-index storage break

The driver v2.4.1 bump changes what `bd init` writes. Dolt 2.4 stores a
descending index column as descending, where earlier Dolt silently stored
it ascending, and migration 0010 declares compaction_snapshots'
idx_comp_snap_issue with created_at DESC. Dolt older than 2.4 refuses a
database holding such a table with "table has unknown fields" on every
write and on information_schema reads, so a database created by this
branch cannot be opened by a bd built against driver v2.2.0, which every
current release is. Measured with bd built from this branch and from main:
on a database the new binary created, the old one fails create, list,
update and close at the pre-open schema skew probe. A database the older
engine created keeps its ascending index and stays usable by both
binaries after the new engine writes to it, and an old bd running against
a dolt sql-server 2.4.1 already creates the descending index.

This is the root cause behind the "table has unknown fields" evidence the
RecommendedMin bump cites. Record it under [Unreleased], together with the
proxied-server recommendation and the stricter default sql_mode Dolt 2.4
reports, so the release notes say which databases are affected and which
binaries have to move together.

Correction on re-landing: the descending-index storage change is
dolthub/dolt#11746, which shipped in Dolt 2.3.3, not 2.4.0 (bracketed by
bee-ghosttrack on gastownhall#7441: dolt 2.3.2 fails, 2.3.3 and 2.3.5 open the
database). The entry now gives 2.3.3 as the compatibility floor and keeps
2.4.1 as the recommendation.

Agent-Signature: claude-agent-sdk-claude-opus-5-5-max on behalf of Julian Knutsen
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* fix(deps): move go-runewidth to v0.0.28 to drop a 35 ms init from every bd run

The go-deps group's charmbracelet/x/ansi v0.11.9 requires
mattn/go-runewidth v0.0.27. In v0.0.26 and v0.0.27 that package
fills a 2 MB width table eagerly in init (mattn/go-runewidth#104), and bd
links it through lipgloss, bubbles and dolt's tabular printer, so every bd
command paid for it before main ran:

  GODEBUG=inittrace=1 bd version   go-runewidth init 33-36 ms (v0.0.27),
                                   0.03-0.04 ms (v0.0.28),
                                   0.12-0.17 ms (main's v0.0.24)
  bd version wall time, median of 40: 147 ms (v0.0.27), 103 ms (v0.0.28),
                                      101 ms (main's v0.0.24)

v0.0.28 builds the table lazily (mattn/go-runewidth#105) and changes
nothing else that is observable: RuneWidth over all 1,114,112 code points
in all four EastAsianWidth/StrictEmojiNeutral combinations, and StringWidth
and Truncate over an emoji, flag, ZWJ, Indic and Hangul sample, are
byte-identical to v0.0.27. go-runewidth is an indirect dependency, so
Dependabot will not move it on its own.

TestTerminalStackPackageInitStaysUnderBudget re-executes the internal/ui
test binary with GODEBUG=inittrace=1 and fails when any package in the
terminal rendering stack spends more than 10 ms in init, taking the fastest
of three starts with GOGC=off so a garbage collection inside one init is not
billed to it. On v0.0.27 it fails naming go-runewidth (32 ms, 381 ms under
-race); on v0.0.28 every package is under 1.1 ms, and it passed 40 -race and 50
plain repetitions, then 30 of each with 24 busy-loop processes running.

Recompute default.nix's vendorHash with a fake-hash build in nixos/nix;
nix run .#default -- --help and nix flake check pass with it.

Agent-Signature: claude-agent-sdk-claude-opus-5-5-max on behalf of Julian Knutsen
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* docs(dolt): name 2.3.3 as the write boundary and drop the stale 2.3.x DOLT_RESET claim

The CHANGELOG entry gave a wrong reason to prefer 2.4.1 over 2.3.x: it said
the 2.3.x releases that can open a new database carry the
DOLT_RESET('--hard') regression. That regression is dolthub/dolt#11581,
fixed by dolthub/dolt#11652, which first shipped in 2.3.2, before the
2.3.3 write boundary. The reasons that do hold are that 2.4.1 is the
release CI pins and measured, and that releases before 2.4.0 mis-plan a
lookup join whose ON clause has an OR over indexed columns
(dolthub/dolt#11886, fixed by go-mysql-server#3899, the cause of gastownhall#7037):
dolt 2.3.5 pins a go-mysql-server without that fix, and 2.4.0 pins one
with it.

The surrounding text this branch added also implied a later boundary than
the real one:

- troubleshooting.md said storage the v2.4.1 engine writes "is not
  writable by older dolt CLIs" and that "2.3.x carries" the DOLT_RESET
  defect. Re-measured 2026-10-17 against a database bd init created with
  this branch: dolt 2.2.0 and 2.3.2 fail a plain INSERT and an
  information_schema read with "table has unknown fields"; 2.3.3, 2.4.0
  and 2.4.1 succeed, and SHOW CREATE TABLE reports created_at DESC on
  idx_comp_snap_issue. The release tarballs match GitHub's asset digests.
- dolt.md and RecommendedMin's comment named only the two failing
  versions measured on 2026-10-10. Both now name 2.3.3 and
  dolthub/dolt#11746 (merged into 2.3.3, absent from 2.3.2), and the
  comment says why the recommendation is still 2.4.1.
- dolt.md said "Versions after 2.4.1 have not been measured", which
  implied 2.3.2 through 2.4.0 had been. None of them is in that table.
- The CHANGELOG put the stricter default sql_mode on "Dolt 2.4.1
  servers". SELECT @@GLOBAL.sql_mode in a fresh dolt init: 2.2.0 and 2.3.2
  report NO_ENGINE_SUBSTITUTION,ONLY_FULL_GROUP_BY,STRICT_TRANS_TABLES;
  2.3.3, 2.4.0 and 2.4.1 add NO_ZERO_IN_DATE, NO_ZERO_DATE and
  ERROR_FOR_DIVISION_BY_ZERO. It now says 2.3.3 and later.

RecommendedMin stays 2.4.1, and TestPinnedDoltCLIAtLeastRecommendedMin
still holds against the 2.4.1 pins.

Agent-Signature: claude-agent-sdk-claude-opus-5-5-max on behalf of Julian Knutsen
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* build(deps): bump dolthub/file-locks to v0.1.2 for BSD builds

file-locks v0.1.1 (pulled in transitively by Dolt 2.4.1) had no build
tags for freebsd/openbsd/netbsd/dragonfly, so the freebsd/amd64 release
cross-compile failed. dolthub/file-locks#6 (by @shrisha) adds them and
ships in v0.1.2.

Only file-locks moves in go.mod/go.sum; tools/nogo is unaffected. The
Nix vendorHash is recomputed for the new module set.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* build(deps): run the embedded Dolt engine on 2.5.1

Raise the direct requirements on github.com/dolthub/dolt/go,
go-mysql-server and vitess to the exact Dolt v2.5.1 release set and keep
the tagged github.com/dolthub/driver/v2 v2.4.1:

  dolthub/dolt/go        v0.40.5-0.20261020191035-d1f8f8e76386 (v2.5.1 tag)
  dolthub/go-mysql-server v0.20.1-0.20261016230655-239e0ad6b2f9
  dolthub/vitess          v0.0.0-20261014211039-3e903c0f05d3
  dolthub/jsonpath        follows (indirect)

No driver release exists on 2.5.x, and driver main now declares go
1.27.3, which would move every toolchain pin. The driver changed no Go
source between v2.4.1 and its 2.5-era commit (89f53f8d3f72), only
go.mod/go.sum, so this pairing is source-identical to DoltHub's own
driver build, on the release commit and with no replace directive.
dolt/go 2.5.1 says go 1.26.2, so the go directive is unaffected, and no
new module enters the graph.

Dolt 2.5.0 fixes the engine bug the scalar-subquery mint read works
around (dolthub/dolt#12093: 2.4.0 through 2.4.2 return no row for an
ungrouped MAX or MIN over an unmatched primary-key prefix) and the
replaceAgg panic on an ungrouped MAX of a non-key column under a fully
bound primary key (dolthub/go-mysql-server#4037). The workaround stays
for server-mode users on dolt 2.4.0 through 2.4.2.

Measured on this engine: the embedded tier's per-test outcome is
identical to the 2.4.1 engine (1755 pass, 59 skip), the default
sql_mode is unchanged, and the storage floor stays Dolt 2.3.3.

default.nix: vendorHash for the new go.sum. The example modules replace
github.com/steveyegge/beads with the repo root, so they follow its
requirements: go mod tidy in examples/library-usage and
examples/bd-example-extension-go (driver v2.4.1, dolthub/file-locks
v0.1.2, the go-deps group and this engine). Without it
scripts/build-examples.sh fails with "updates to go.mod needed"; this
was a review finding from gastownhall#7646's adoption review, re-applied here.

Agent-Signature: claude-agent-sdk-claude-opus-5-5-unknown-reasoning on behalf of Julian Knutsen
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* build(dolt): move the CLI, container and Bazel pins to 2.5.1 and recommend 2.5.0

The Dolt CLI pin (scripts/ci/install-dolt.sh), the test container
(dolthub/dolt-sql-server:2.5.1 in testdoltcommon.go and
pull-dolt-image.sh), the hermetic Bazel dolt (tools/bazel/dolt.bzl
DOLT_VERSION plus the four release-asset digests) and the docs snippet
move to 2.5.1, the release the embedded engine now links, so the CLI,
the container and embedded mode run the same Dolt.
TestPinnedDoltCLIMatchesContainerImage, TestPinnedDoltCLIAtLeastRecommendedMin
and test/docsync's doltpin test hold those in step.

2.5.1 passes the pin criterion install-dolt.sh states: measured
2026-10-21 by creating fresh databases and immediately calling
CALL DOLT_RESET('--hard'), 2.5.0 and 2.5.1 broke 0/500 each, 2.4.2
0/250 and the embedded 2.5.1 engine 0/100, against a 2.3.1 positive
control of 4/300. TestDoltServerFingerprint matches the existing golden
on the 2.5.1 CLI and container: the default sql_mode did not change
from 2.4.1.

doltversion.RecommendedMin moves to 2.5.0, the first release that
answers an ungrouped MAX or MIN over an unmatched primary-key prefix
with one NULL row (dolthub/dolt#12093; 2.4.0 through 2.4.2 return no
row, and panic on dolthub/go-mysql-server#4037). The old-version
warning now names what goes wrong below it, release range by range:
2.4.0-2.4.2 and #12093, the join mis-plan before 2.4.0
(dolthub/dolt#11886), the DOLT_RESET('--hard') regression on 2.3.0
and 2.3.1, and "table has unknown fields" before 2.3.3, the storage
floor. TestWarningMessage pins each clause. It stays a warning, not a
refusal. The scoped warning was a gastownhall#7646 adoption-review finding,
re-worded here for 2.5.0 and without migration 0071.

Agent-Signature: claude-agent-sdk-claude-opus-5-5-unknown-reasoning on behalf of Julian Knutsen
Co-authored-by: Eddie the Engineer <338826946+gascity-eddie[bot]@users.noreply.github.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* docs(code): correct the parity-oracle and port-check comments for the 2.5.1 pin

Review fixes from gastownhall#7646's adoption review, iterations 1-3, re-applied
without migration 0071 (they were carried by 2068dbe, which this
branch drops):

- cli_migrations.go puts the parity oracle's green run against the
  then-pinned 2.2.0 sql-server in the past tense. A 2.3.0-or-newer CLI
  executes the prepared ALTER, so with the CLI pinned at 2.5.1
  TestCLIBundleMatchesRuntimeCommittedSchema cannot catch a dropped
  override (re-probed 2026-10-21: a prepared ADD COLUMN vanishes on the
  2.2.0 CLI and lands on 2.5.1). The comment names the two binary-free
  tests that still can.
- localdoltserver.go's port comment covers 2.4.1 and 2.5.1 as well as
  2.2.0: each prints "Port N already in use." on a held port (2.5.1
  measured 2026-10-23).

Agent-Signature: claude-agent-sdk-claude-opus-5-5-unknown-reasoning on behalf of Julian Knutsen
Co-authored-by: Eddie the Engineer <338826946+gascity-eddie[bot]@users.noreply.github.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* test(storage): pin per-issue first-mint revisions

TestEmbeddedFirstMintOnEmptyHistoryIsRevisionOne now also creates a
second issue and pins that its first mint is revision 1: revisions
count per issue. On its own this kills a mutant that drops the
issue_id filter from the scalar subquery (the second issue then
continues the first issue's sequence: "first mint revision for a
second issue = 3, want 1").

A gastownhall#7646 adoption-review finding (iteration 1, gap 2), carried by
2068dbe and re-applied here without migration 0071.

Agent-Signature: claude-agent-sdk-claude-opus-5-5-unknown-reasoning on behalf of Julian Knutsen
Co-authored-by: Eddie the Engineer <338826946+gascity-eddie[bot]@users.noreply.github.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* docs(dolt): Dolt 2.5 version table; the descending-index change is one-way (floor 2.3.3)

Maintainer decision on gastownhall#7646: the descending-index storage change is
accepted as one-way. Dolt 2.3.3 and later store migration 0010's
created_at DESC on compaction_snapshots' idx_comp_snap_issue as
descending (dolthub/dolt#11746), and Dolt older than 2.3.3 cannot read
such a table ("table has unknown fields"). bd does not convert the index
back, so every database bd creates from this release on needs Dolt 2.3.3
or newer, and a bd built on it, for good. Databases created earlier keep
an ascending index and stay readable and writable by older Dolt even
after this release migrates and writes to them (measured 2026-10-21 with
Dolt 2.2.0 and 2.3.2).

- CHANGELOG: the entry now says the embedded engine is Dolt 2.5.1
  (driver v2.4.1 with dolt/go at the v2.5.1 release), that the change is
  one-way, that the recommended dolt is 2.5.0 (CI pins 2.5.1), and lists
  the defects below it: dolthub/dolt#12093 on 2.4.0-2.4.2, the join bug
  before 2.4.0, the DOLT_RESET('--hard') regression on 2.3.0 and 2.3.1.
- docs/architecture/dolt.md: the pin paragraph says embedded mode links
  the 2.5.1 engine and the CLI pin tracks it; a one-way statement; and
  one version table merging the DOLT_RESET measurements with the
  2.3.3 storage floor and the MAX/MIN engine bug (2.4.0-2.4.2 failing,
  2.5.0 and later correct; 2026-10-21: 2.3.1 4/300, 2.5.0 and 2.5.1
  0/500 each, 2.4.2 0/250).
- troubleshooting.md: the old-version warning section recommends 2.5.0
  and lists the three defect ranges, pointing at dolt.md.
- scripts/test-git-remote.sh: dolt >= 2.3.3. Its 2.2.0 floor rested on
  migration 0071, which this branch does not carry.
- sync-setup.md scopes the DOLT_RESET regression to 2.3.0 and 2.3.1
  (2.3.2 carries the fix); upgrading.md scopes the gastownhall#7037 embedded-engine
  caveat to bd releases that link Dolt older than 2.4.0.

The MAX/MIN column is bee-ghosttrack's reproduction from gastownhall#7646.

Agent-Signature: claude-agent-sdk-claude-opus-5-5-unknown-reasoning on behalf of Julian Knutsen
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

---------

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Julian Knutsen <julianknutsen@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-authored-by: coffeegoddd <dustin@dolthub.com>
Co-authored-by: Eddie the Engineer <338826946+gascity-eddie[bot]@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

lock_unix.go builds only on linux || darwin; freebsd (and other unix) has no sysOpen/sysLock/sysUnlock

2 participants