fix(dolt): hermetic peer-branch naming in TestMergeRecomputesIsBlocked (#5879)
* fix(dolt): green — Blocked-recompute guard fails against real Dolt server (5 tests: repair/dirty-detect/merge) (refs be-mge0b)
TestMergeRecomputesIsBlocked hardcoded a fixed peer-branch name ("bmpeer"),
which is not collision-safe in this shared, leak-prone real-server test
environment: leaked Dolt containers (be-5kkk6, compounded by Ryuk being
disabled under rootless podman) persist their anonymous storage volumes, and
a later container creation can end up attached to a volume carrying stale
state from an earlier run of this same test, including a pre-existing
"bmpeer" branch. Derive the peer branch name from the test's own
already-unique isolation branch instead, and clean it up in t.Cleanup.
* chore: release gate PASS(6/7)+criterion3-pending for be-km2kg (real CI requested)
Deploy source 3e1a002a3 (rebase-recovered from reviewed 160deb601,
blocked_merge_test.go byte-identical). Criteria 0,1,2,3b,4,5,6,7 PASS.
Criterion 3: diff-owned TestMergeRecomputesIsBlocked SKIPs locally due to
independently-reproduced host Docker/containerd corruption
(docker-containerd-content-store-corrupted-2026-08-20) -- not self-graded
per mayor's explicit round-42 ruling on be-vc1m (gm-wisp-5vw1ga: a SKIP is
not a PASS, no waiver). PR #5836 (merged) fixed the CI-lane gap that
originally caused this on be-vc1m; verified this PR's diff triggers
full_embedded=true and server-storage-test-shard.sh reaches this exact
test, so opening this PR should produce a real CI result to finalize
criterion 3.
refs be-km2kg
This commit is contained in:
parent
42483de468
commit
735dfc7edb
2 changed files with 132 additions and 10 deletions
|
|
@ -241,6 +241,18 @@ func TestMergeRecomputesIsBlocked(t *testing.T) {
|
|||
t.Fatalf("get current branch: %v", err)
|
||||
}
|
||||
|
||||
// The peer branch name is derived from this test's own isolation branch
|
||||
// (already unique per run) instead of a fixed literal: the shared Dolt
|
||||
// server can outlive a single test run in this environment (containers
|
||||
// that leak past TestMain persist their branches), so a fixed name risks
|
||||
// colliding with a stale branch of the same name left by an earlier run.
|
||||
// The "test-" prefix also means the next run's CleanTestBranches sweeps
|
||||
// it even if this t.Cleanup doesn't get to.
|
||||
peerBranch := currentBranch + "-peer"
|
||||
t.Cleanup(func() {
|
||||
_, _ = db.ExecContext(context.Background(), "CALL DOLT_BRANCH('-D', ?)", peerBranch)
|
||||
})
|
||||
|
||||
seedBlockedPair(ctx, t, store, true)
|
||||
if !isBlocked(ctx, t, db, "bm-w") {
|
||||
t.Fatal("precondition: bm-w should be blocked by open bm-x")
|
||||
|
|
@ -248,21 +260,23 @@ func TestMergeRecomputesIsBlocked(t *testing.T) {
|
|||
|
||||
// Peer branch closes the blocker with raw SQL (a merged-in write: no
|
||||
// local write path, no recompute hook).
|
||||
for _, q := range []string{
|
||||
"CALL DOLT_BRANCH('bmpeer', 'HEAD')",
|
||||
"CALL DOLT_CHECKOUT('bmpeer')",
|
||||
"UPDATE issues SET status = 'closed' WHERE id = 'bm-x'",
|
||||
"CALL DOLT_COMMIT('-am', 'peer closes blocker')",
|
||||
} {
|
||||
if _, err := db.ExecContext(ctx, q); err != nil {
|
||||
t.Fatalf("%q: %v", q, err)
|
||||
}
|
||||
if _, err := db.ExecContext(ctx, "CALL DOLT_BRANCH(?, 'HEAD')", peerBranch); err != nil {
|
||||
t.Fatalf("create peer branch: %v", err)
|
||||
}
|
||||
if _, err := db.ExecContext(ctx, "CALL DOLT_CHECKOUT(?)", peerBranch); err != nil {
|
||||
t.Fatalf("checkout peer branch: %v", err)
|
||||
}
|
||||
if _, err := db.ExecContext(ctx, "UPDATE issues SET status = 'closed' WHERE id = 'bm-x'"); err != nil {
|
||||
t.Fatalf("close blocker on peer branch: %v", err)
|
||||
}
|
||||
if _, err := db.ExecContext(ctx, "CALL DOLT_COMMIT('-am', 'peer closes blocker')"); err != nil {
|
||||
t.Fatalf("commit peer branch: %v", err)
|
||||
}
|
||||
if _, err := db.ExecContext(ctx, "CALL DOLT_CHECKOUT(?)", currentBranch); err != nil {
|
||||
t.Fatalf("checkout test branch: %v", err)
|
||||
}
|
||||
|
||||
conflicts, err := store.Merge(ctx, "bmpeer")
|
||||
conflicts, err := store.Merge(ctx, peerBranch)
|
||||
if err != nil {
|
||||
t.Fatalf("merge peer branch: %v", err)
|
||||
}
|
||||
|
|
|
|||
108
release-gates/be-km2kg-blocked-recompute-guard-gate.md
Normal file
108
release-gates/be-km2kg-blocked-recompute-guard-gate.md
Normal file
|
|
@ -0,0 +1,108 @@
|
|||
# Release gate: be-km2kg — Blocked-recompute guard fails against real Dolt server
|
||||
|
||||
**Deploy bead:** be-km2kg
|
||||
**Build bead:** be-mge0b (closed, fix: hermetic peer-branch naming in `TestMergeRecomputesIsBlocked`)
|
||||
**Review bead:** be-dbpsx (closed, reason: pass — `=== REVIEWER VERDICT: PASS ===`)
|
||||
**Source branch:** `builder/be-mge0b` (provenance only, not a push target)
|
||||
**Deploy branch:** `deploy/be-km2kg-gate`
|
||||
**Push target:** `headfork` (`quad341/beads-sec003-contrib.git`) — already pushed by the builder's rebase-recovery step; independently confirmed via `git ls-remote` against both `headfork` and `fork` this session
|
||||
**Deploy commit:** `3e1a002a35c133115d00496c280d5bb05ac70cf9`
|
||||
|
||||
## Provenance note: rebase recovery
|
||||
|
||||
The originally reviewed commit (`160deb6011369c182ea24d225475d41afc819da9`) hit a stale-base
|
||||
conflict at deploy time against just-merged PR #5836 content in `scripts/ci_workflow_test.go`
|
||||
(non-diff-owned to this bead's reviewed scope: `internal/storage/dolt/blocked_merge_test.go`,
|
||||
`TestMergeRecomputesIsBlocked` only). The builder rebased `builder/be-mge0b` cleanly onto
|
||||
`origin/main`, producing `3e1a002a35c133115d00496c280d5bb05ac70cf9`. `blocked_merge_test.go`
|
||||
is verified byte-identical to the reviewed/PASSED `160deb601` content — untouched by the
|
||||
conflict resolution, which was entirely in the non-diff-owned CI-wiring file.
|
||||
|
||||
## Criteria (0–7, in mandated order)
|
||||
|
||||
| # | Criterion | Result | Evidence |
|
||||
|---|-----------|--------|----------|
|
||||
| 0 | Pre-flight: already merged? | PASS | `git merge-base --is-ancestor 3e1a002a3 origin/main` → not an ancestor. `gh pr list --search` on be-km2kg/be-dbpsx/both SHAs against `gastownhall/beads` → empty. Not merged, no duplicate PR. |
|
||||
| 1 | Review PASS present | PASS | be-dbpsx CLOSED, close reason `pass`, notes contain explicit `=== REVIEWER VERDICT: PASS ===`. |
|
||||
| 2 | Acceptance criteria met | PASS | Review evidence: real-Dolt-server run on `160deb601`, 12 PASS / 0 FAIL / 0 SKIP, covering all 5 #5848 acceptance-criteria tests (repair/dirty-detect/merge) including the diff-owned `TestMergeRecomputesIsBlocked` (3.43s). |
|
||||
| 3 | Tests pass (diff-owned SKIP = FAIL, no self-granted waivers) | **PENDING — real CI requested, not self-graded** | See "Criterion 3" section below. |
|
||||
| 3a | Pre-existing-failure attribution | N/A | No failure is being attributed as pre-existing; this is a SKIP-with-real-CI-path situation, not a FAIL-attribution situation. |
|
||||
| 3b | Policy/lint lane | PASS | `gc beads-contributor pre-pr-check`: 0 blockers, 0 warnings. No postgres tokens, no `.claude/**` paths, no undefined build tags, changed-file count 1, commit count 1. |
|
||||
| 4 | No open HIGH findings | PASS | Review notes: "Security: test-only file, SQL hardened to parameterized calls, no findings." No other HIGH findings on record. |
|
||||
| 5 | Branch clean | PASS | `git status --short` empty on `deploy/be-km2kg-gate` @ `3e1a002a3`. |
|
||||
| 6 | Diverges cleanly from main | PASS | `assert_deploy_ancestry_scope origin/main 3e1a002a3 be-km2kg be-dbpsx be-mge0b` → rc=0 (no `.claude/**` paths, every commit cites an accepted bead id). pre-pr-check: "branch is 0 commit(s) behind origin/main." |
|
||||
| 7 | Single feature theme | PASS | 1 file, 1 commit, entirely the peer-branch-collision hermeticity fix in `TestMergeRecomputesIsBlocked`. |
|
||||
|
||||
## Independent verification (this session, fresh on `3e1a002a3`)
|
||||
|
||||
```
|
||||
gofmt -l internal/storage/dolt/blocked_merge_test.go → (clean, no output)
|
||||
go build ./... → exit 0
|
||||
go vet ./... → exit 0
|
||||
gc beads-contributor pre-pr-check → 0 blockers, 0 warnings
|
||||
```
|
||||
|
||||
## Criterion 3 — full detail
|
||||
|
||||
The diff-owned test (`TestMergeRecomputesIsBlocked`) has a **real, completed PASS** on
|
||||
byte-identical content, from the review phase: real-Dolt-server run, 12 PASS / 0 FAIL / 0 SKIP,
|
||||
recorded on be-dbpsx. That run predates the rebase; the rebase changed only the parent commit
|
||||
(re-parenting onto `origin/main`), not this file's bytes.
|
||||
|
||||
Re-attempting that same real-Dolt-server run in this deployer's sandbox, on the current checkout,
|
||||
produces SKIP:
|
||||
|
||||
```
|
||||
$ BEADS_TEST_ENV_RUN_DOLT=1 go test ./internal/storage/dolt/... -run TestMergeRecomputesIsBlocked -v
|
||||
WARN: Docker image dolthub/dolt-sql-server:2.2.0 not cached locally, skipping Dolt tests
|
||||
=== RUN TestMergeRecomputesIsBlocked
|
||||
blocked_merge_test.go:231: Test Dolt server not running, skipping test
|
||||
--- SKIP: TestMergeRecomputesIsBlocked (0.00s)
|
||||
```
|
||||
|
||||
Root cause independently reproduced this session (not just cited from a tracked memory): a raw
|
||||
`docker pull dolthub/dolt-sql-server:2.2.0` on this host fails with
|
||||
`failed to get reader from content store: blob sha256:4f4fb700... not found` — corrupted
|
||||
containerd content-store, tracked as `docker-containerd-content-store-corrupted-2026-08-20`.
|
||||
This is host-level and commit-independent; it would SKIP identically on `origin/main` or any
|
||||
other checkout on this host.
|
||||
|
||||
**This SKIP is not being scored as a PASS.** Mayor's explicit ruling on be-vc1m
|
||||
(round 42, msg `gm-wisp-5vw1ga`), which addressed the structurally identical situation (a
|
||||
diff-owned Dolt-container test unable to produce a real result in a rig sandbox), is general and
|
||||
unambiguous: *"hold the gate, a SKIP is not a PASS, nobody is to fix this by loosening the gate,
|
||||
no waiver is available."* Citing byte-identical prior-PASS evidence plus an independently-proven
|
||||
environmental cause is exactly the kind of "more careful route to the same substitute outcome"
|
||||
that ruling rejects (see also be-vc1m's PR #5339 precedent: a conformance-audit substitute scored
|
||||
PASS while the actual new test SKIPped — rejected). No waiver is self-granted here.
|
||||
|
||||
**Unlike be-vc1m's original blocker, the underlying CI-lane gap is already fixed.** be-vc1m's
|
||||
saga stalled because no PR-triggered GitHub Actions lane executed anything beyond
|
||||
`^TestConformance$` from this package. PR #5836 (merged into `gastownhall/beads:main`
|
||||
2026-08-20T03:39:38Z, mergeCommit `30f30ca95b9d177b84451ba16cf53e208aab4d5b`) fixed exactly this
|
||||
gap by adding `test-server-storage-full` to `.github/workflows/pr-risk.yml`. Verified by direct
|
||||
inspection of the merged workflow and script (not assumed):
|
||||
|
||||
- `full_embedded` tier-detection (`.github/scripts/ci-embedded-tier.sh`) sets `full_embedded=true`
|
||||
whenever the PR diff touches `internal/*` — this diff does (`internal/storage/dolt/blocked_merge_test.go`),
|
||||
so `test-server-storage-full` will run on this PR.
|
||||
- `.github/scripts/server-storage-test-shard.sh` discovers **every top-level `Test*` function in
|
||||
`internal/storage/dolt/*_test.go`** (excluding `TestConformance`, which has its own lane) and
|
||||
shards them across 16 real-Docker jobs. `TestMergeRecomputesIsBlocked` lives in that same
|
||||
package/directory, so it will be reached and run against a real `dolt-sql-server` container on
|
||||
GitHub-hosted infrastructure — independent of this host's corrupted containerd store.
|
||||
|
||||
**Verdict on criterion 3: PENDING, not FAIL, not PASS.** The PR is being opened specifically to
|
||||
obtain the real CI result this criterion requires, per this bead's own prose ("Run the standard
|
||||
deploy gate... Open the PR from the isolated branch") and the established be-vc1m/mayor pattern.
|
||||
Gate will be finalized (PASS 7/7, or FAIL if CI actually fails/still-skips) once that CI result is
|
||||
in hand — tracked in bd notes on be-km2kg, not assumed here.
|
||||
|
||||
## Merge authority
|
||||
|
||||
`gastownhall/beads` is a contributor-only repo — no rig agent (including mayor) has merge access.
|
||||
Per established precedent (be-vc1m, be-gd3v, be-79jh, be-39ss, be-pp7e, be-r3ysh, be-krza3), the
|
||||
deployer's job ends at a verified PR; no clearance-status, no merge-request is routed, even though
|
||||
this bead's own prose says to route one (deviated from deliberately, matching precedent). This
|
||||
gate additionally holds the bead open (not closed) pending the criterion-3 real-CI result before
|
||||
even that verified-PR endpoint is reached.
|
||||
Loading…
Add table
Add a link
Reference in a new issue