Skip to content

build(deps): bump dolthub driver to v2.4.0 - #7045

Closed
coffeegoddd wants to merge 1 commit into
gastownhall:mainfrom
coffeegoddd:db/bump-dolt
Closed

coffeegoddd wants to merge 1 commit into
gastownhall:mainfrom
coffeegoddd:db/bump-dolt

Conversation

@coffeegoddd

Copy link
Copy Markdown
Collaborator

Updates github.com/dolthub/driver/v2 from v2.2.0 to v2.4.0, including its required Dolt, go-mysql-server, Vitess, and indirect dependency updates. Refreshes go.sum with go mod tidy.

Validation: go mod tidy and git diff --check passed. Local tests were not run successfully; the initial attempt was blocked by sandbox restrictions, and further local testing was skipped at the user's request. Bazel regeneration was also blocked by sandbox restrictions; CI should validate generated metadata.

Agent-Signature: codex-unknown-model-unknown-reasoning on behalf of coffeegoddd
@github-actions github-actions Bot added the status/needs-triage Awaiting initial PM triage label Sep 30, 2026
@gascityinc-sarah gascityinc-sarah Bot added kind/chore Internal maintenance or tooling priority/p3 Backlog: useful but not currently planned labels Sep 30, 2026
@gascityinc-sarah

Copy link
Copy Markdown

Triage: kind/chore · priority/p3 — Routine dependency bump of the dolthub driver to v2.4.0 with a go.sum refresh and no behavior change. All three chore/pr bump precedents landed p3; CI owns the generated-metadata check.
Auto-triaged by @gascityinc-sarah. Reply or relabel if this is off.

@gascityinc-sarah gascityinc-sarah Bot removed the status/needs-triage Awaiting initial PM triage label Sep 30, 2026

@bee-ghosttrack bee-ghosttrack left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

REQUEST CHANGES at 30db776ef (merge-base 98ab92041, 0 behind main). The Go side of the bump is clean; the two CI reds are real and both come from this diff, and one of them reaches users.

1. Blocker — the pinned Dolt 2.2.0 can no longer write what the bumped embedded engine writes. This bump moves dolthub/dolt/go from 2026-08-06 to 2026-09-30 main, which carries two flatbuffer schema additions: Index.Descending/NullsLast/OpClasses (dolt cd8731234d, 2026-09-09, first in v2.3.3) and ProllyTreeNode.KeyAddressOffsets (410af9976f, first in v2.3.2). Every table the embedded engine writes now carries vtable fields an older Dolt does not know, and dolthub/flatbuffers' TryGet…AsRoot refuses them on the write path with ErrTableHasUnknownFields.

Reproduced locally: bd built at this head, bd init --prefix t + bd create, then in .beads/embeddeddolt/t with the dolt 2.2.0 CLI (downloaded release binary, darwin/arm64): SELECT works, dolt log and dolt gc exit 0, but

dolt sql -q "INSERT INTO schema_migrations (version, content_hash) VALUES (999, 'ff…')"
error on line 1 for query …: table has unknown fields

With dolt 2.3.5 the same INSERT and DELETE succeed. That is exactly the Contract corpus red: TestProtocol_V2_DesignatedMigratorOverride, _V2_RemoteMigrateGate, _V3_SkewOverrideDowngradesToWarning, _V3_SchemaSkewRefusal all die in storeExec (wp4_helpers_test.go:192) on the runner's scripts/ci/install-dolt.sh pin of 2.2.0.

The pin family has to move together and to ≥ 2.3.3: scripts/ci/install-dolt.sh (version="2.2.0"), internal/testutil/testdoltcommon.go DoltDockerImage, scripts/ci/pull-dolt-image.sh, tools/bazel/dolt.bzl DOLT_VERSION + its per-platform sha256 (the header of install-dolt.sh names this set). And install-dolt.sh's own comment forbids raising to 2.3.x until a release passes the DOLT_RESET('--hard') fresh-database measurement (2.3.0 3/60, 2.3.1 3/100 broken). So this PR is gated on a measured-clean Dolt ≥ 2.3.3 — v2.4.0 (released today) is the natural candidate; you would know better than we do whether the DOLT_RESET regression is fixed there. Until the pins move, the suite would be "exercising a Dolt pair no release ever shipped", and one that fails.

User-facing half of the same fact: internal/doltversion/policy.go RecommendedMin is 2.0.0. After this bump any operator whose dolt CLI is < 2.3.3 can read an embedded workspace but cannot write it from the CLI (bd's own writes go through the library, so bd itself keeps working; bd dolt compact's dolt gc shell-out also survived 2.2.0 here). RecommendedMin and the CHANGELOG should say the new floor.

2. Blocker — Nix. github.com/dolthub/driver/v2@v2.4.0 declares go >= 1.26.5; the flake's nixpkgs (flake.lock rev b6018f87…, go_1_26 = 1.26.4) with GOTOOLCHAIN=local fails in the go-modules derivation: module github.com/dolthub/driver/v2@v2.4.0 requires go >= 1.26.5 (running go 1.26.4). default.nix's postPatch only lowers beads' own go.mod, so it cannot cure a dependency's floor. Needs a nixpkgs bump carrying go ≥ 1.26.5 plus the new vendorHash (the auto-update job is skipped on a fork PR).

Checked and sound at 30db776ef (go1.27.1 darwin/arm64, -tags gms_pure_go, -count=1): go mod tidy makes no change, go mod verify all verified, go build ./... and go vet (cmd/bd, internal/storage/embeddeddolt) clean; internal/storage/embeddeddolt 85 PASS / 12 SKIP; BEADS_TEST_EMBEDDED_DOLT=1 ./internal/storage/ 118 PASS / 0 FAIL. Storage backend conformance (embedded Dolt oracle) and Managed-local proxied lifecycle are green in CI too, so the bump itself is fine wherever the library is the only writer.

I will re-review on the next push.

@gascityinc-randy

Copy link
Copy Markdown
Contributor

Thank you, @coffeegoddd. This is superseded by #7646, which merged 2026-10-23. #7646 moved dolthub/driver past this PR's v2.4.0 to v2.4.1, and pinned the embedded engine (dolthub/dolt/go, go-mysql-server, vitess) to the Dolt v2.5.1 release. That includes every dolt/go change this bump brought in. Closing as superseded.

@gascityinc-randy gascityinc-randy Bot closed this Oct 23, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

kind/chore Internal maintenance or tooling priority/p3 Backlog: useful but not currently planned

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants