fix(bazel): remove host-go dep from compact GC test; drop frozen shard blocks (#7478)

* fix(bazel): remove host-go dep from compact GC test; drop frozen shard blocks

Replace the ungated `go build` in compact_gc_target_test.go's dolt-stand-in
fixture with a Bazel-built go_binary (cmd/bd:compact_gc_fixture) wired via
rlocationpath env + data, mirroring the existing preflight_go_fixture
pattern. Falls back to `go build` only when not running under Bazel
(bazeltest.IsBazel()). Also gate test_repo_beads_guard_test.go's `go env
GOCACHE`/`GOMODCACHE` probes on !bazeltest.IsBazel(), since they previously
ran unconditionally in every cmd/bd TestMain.

Separately, delete the legacy 20-shard embedded-cmd and 15-shard
proxied-cmd "frozen" manifest blocks in .github/scripts/*-test-shards.txt:
their only consumers were pr-risk.yml fork/push jobs that have since
migrated to bazel.yml. Update gen_embedded_cmd_shard_manifest.py and
gen_proxied_shard_manifest.py to drop the frozen-block rendering/guard
logic and default to the live Bazel-only shard counts (100 and 34),
and refresh stale comments/docs (cmd/bd/BUILD.bazel,
scripts/shard_manifest_test.go, engdocs/TESTING.md,
engdocs/CI_REQUIRED_CHECK_TOPOLOGY.md) that still described the deleted
blocks. Verified no other BUILD/.github/scripts references read them.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* build(bazel): auto-sync BUILD files

Applied from the bazel-sync-patch artifact of https://github.com/gastownhall/beads/actions/runs/38069936930
(`make bazel-sync`: gazelle, tools/bazel/go_srcs.py, bazel mod tidy).
Only BUILD.bazel, MODULE.bazel and MODULE.bazel.lock are ever pushed; see
scripts/bazel-autofix-push.sh.

* fix(bazel): gate remaining host-go calls; fix shard-manifest --weights footgun

Address PR #7478 review findings (bee-ghosttrack):

- scripts/ci/lib/test-env.sh: skip the GOCACHE/GOMODCACHE `go env` probes
  when TEST_SRCDIR is set, so scripts/test.sh's beads_test_env_enter never
  shells out to a bare `go` when reached from a Bazel-run go_test (e.g.
  //scripts:shell_scripts_test's TestTestScriptPrebuiltTestBinaryContract),
  where a farm masks host go/dolt to exit 127. Nothing on that path needs a
  cache directory.
- scripts/ci/fmt-check.sh: describe_gofmt() no longer calls a PATH `go`
  under Bazel. It now resolves BEADS_TEST_GO (the hermetic SDK go already
  wired into scripts/prlintmake:prlintmake_test's env) via TEST_SRCDIR, and
  skips the version-reporting arm entirely when that isn't set (e.g.
  scripts/repochecks:fmt_test), rather than ever touching a possibly-masked
  PATH `go`.
- Verified with a PATH stub (go/dolt scripts that log and exit 127) via
  `bazel --nohome_rc test --config=prcore //scripts/... //cmd/bd:bd_test
  --test_env=PATH=<stubdir>:/usr/bin:/bin:/usr/local/bin`: 26/26 targets
  pass, zero stub invocations logged.
- gen_proxied_shard_manifest.py / gen_embedded_cmd_shard_manifest.py:
  default --weights to duration now that the manifests' only remaining
  blocks are duration-packed (the legacy inits-packed blocks were deleted
  alongside the PR Risk fork/push jobs that read them). Previously the bare
  `--write` default (inits) silently clobbered the live block's duration
  header and repacked with the wrong cost model; --check couldn't see it.
  Added TestProxiedShardManifestBareWriteIsIdempotent and
  TestEmbeddedCmdShardManifestBareWriteIsIdempotent to pin a bare --write
  as a no-op against the committed files.
- cmd/bd/compact_gc_target_test.go: resolve the non-Bazel fallback fixture
  path from this file's own directory (runtime.Caller) instead of the test
  binary's working directory.
- Refresh stale comments in .github/workflows/bazel.yml,
  scripts/shard_manifest_test.go, scripts/pr_risk_bazel_coverage_test.go,
  and scripts/embedded_shard_count_test.go that still described the deleted
  legacy 20/15-shard blocks and their retired fork/push jobs in the present
  tense.
- Confirmed cmd/bd/BUILD.bazel's compact_gc_fixture go_binary (implicit
  pure/race defaults) still passes TestRunCompactDoltTargetsOnlyAuthorizedActiveDatabase
  under `bazel test //cmd/bd:bd_test`.

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

---------

Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
This commit is contained in:
Julian Knutsen 2026-10-10 19:24:21 +00:00 • committed by GitHub
commit 3e7705a2cd
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
16 changed files with 382 additions and 634 deletions

View file

@ -1,161 +1,3 @@
# Embedded cmd test shard manifest.
#
# Format: <total_shards> <shard_number> <top_level_test_name>
#
# There is no committed recent duration artifact for these tests. This manifest
# pins a conservative even split for the 20-shard CI matrix while
# embedded-test-shard.sh continues to hash-distribute newly-added tests that are
# not listed here.
20 1 TestEmbeddedBackup
20 2 TestEmbeddedBackupConcurrent
20 3 TestEmbeddedBlocked
20 4 TestEmbeddedBlockedConcurrent
20 6 TestEmbeddedBootstrapConcurrent
20 7 TestEmbeddedBranch
20 8 TestEmbeddedBranchConcurrent
20 10 TestEmbeddedChildrenConcurrent
20 11 TestEmbeddedCloseGuardsAndEpics
20 12 TestEmbeddedCloseConcurrent
20 13 TestEmbeddedComments
20 14 TestEmbeddedCommentsCLI
20 15 TestEmbeddedCommentsCLIConcurrent
20 16 TestEmbeddedCompact
20 17 TestEmbeddedCompactConcurrent
20 18 TestEmbeddedConfig
20 19 TestEmbeddedConfigConcurrent
20 20 TestEmbeddedConfigNoWorkspaceStoreFreeCommands
20 1 TestEmbeddedConfigValidateJSONNoWorkspaceWritesStdout
20 2 TestEmbeddedContext
20 3 TestEmbeddedContextConcurrent
20 5 TestEmbeddedCount
20 6 TestEmbeddedCountConcurrent
20 7 TestEmbeddedCreateScalarAndRelationships
20 7 TestEmbeddedCreateGraphAndBulk
20 8 TestEmbeddedCreateCommitPending
20 9 TestEmbeddedCreateConcurrent
20 10 TestEmbeddedCreateCrossRepo
20 11 TestEmbeddedCreateCrossRepoUninit
20 12 TestEmbeddedCreateCrossRepoWithParent
20 13 TestEmbeddedCreateWithGitRemote
20 14 TestEmbeddedDefer
20 15 TestEmbeddedDeferConcurrent
20 18 TestEmbeddedDepConcurrent
20 19 TestEmbeddedDiff
20 20 TestEmbeddedDiffConcurrent
20 1 TestEmbeddedDoltBlockedAndCommit
20 1 TestEmbeddedDoltRemoteAndPush
20 2 TestEmbeddedDoltConcurrent
20 3 TestEmbeddedDuplicate
20 4 TestEmbeddedDuplicateConcurrent
20 6 TestEmbeddedDuplicatesConcurrent
20 7 TestEmbeddedEpicStatus
20 7 TestEmbeddedEpicCloseEligibleReasonFlag
20 7 TestEmbeddedEpicCloseEligibleReasonOutput
20 8 TestEmbeddedEpicConcurrent
20 9 TestEmbeddedExportFormats
20 9 TestEmbeddedExportFlags
20 10 TestEmbeddedExportConcurrent
20 11 TestEmbeddedFederationPeers
20 11 TestEmbeddedFederationStatus
20 12 TestEmbeddedFederationConcurrent
20 13 TestEmbeddedFindDuplicates
20 14 TestEmbeddedFindDuplicatesConcurrent
20 15 TestEmbeddedFlatten
20 16 TestEmbeddedFlattenConcurrent
20 18 TestEmbeddedGCConcurrent
20 20 TestEmbeddedGateConcurrent
20 2 TestEmbeddedGetDependencies
20 3 TestEmbeddedGraph
20 4 TestEmbeddedGraphConcurrent
20 5 TestEmbeddedHistory
20 6 TestEmbeddedHistoryConcurrent
20 7 TestEmbeddedHooks
20 8 TestEmbeddedHooksConcurrent
20 10 TestEmbeddedHumanConcurrent
20 11 TestEmbeddedImportFromSources
20 11 TestEmbeddedImportUpsertAndPrefixSync
20 12 TestEmbeddedImportConcurrent
20 13 TestEmbeddedInfo
20 14 TestEmbeddedInfoConcurrent
20 15 TestEmbeddedInitC
20 16 TestEmbeddedInitConcurrent
20 17 TestEmbeddedKV
20 18 TestEmbeddedKVConcurrent
20 19 TestEmbeddedLabelPropagateRename
20 20 TestEmbeddedLabelConcurrent
20 1 TestEmbeddedLint
20 2 TestEmbeddedLintConcurrent
20 4 TestEmbeddedListConcurrent
20 13 TestEmbeddedMaxRowsListBasic
20 13 TestEmbeddedMaxRowsListLimitInteraction
20 11 TestEmbeddedMaxRowsDepTreeAndDuplicates
20 12 TestEmbeddedMaxRowsDoctorAndConfig
20 14 TestEmbeddedMaxRowsGraph
20 15 TestEmbeddedMaxRowsReady
20 16 TestEmbeddedMaxRowsReadyMerge
20 5 TestEmbeddedMemory
20 6 TestEmbeddedMemoryConcurrent
20 7 TestEmbeddedMigrate
20 8 TestEmbeddedMigrateConcurrent
20 9 TestEmbeddedOnboard
20 10 TestEmbeddedOnboardConcurrent
20 11 TestEmbeddedOrphans
20 12 TestEmbeddedOrphansConcurrent
20 13 TestEmbeddedPrime
20 14 TestEmbeddedPrimeConcurrent
20 15 TestEmbeddedPromote
20 16 TestEmbeddedPromoteCLI
20 17 TestEmbeddedPromoteCLIConcurrent
20 18 TestEmbeddedPruneSafetyAndScope
20 18 TestEmbeddedPruneDryRunAndFilters
20 19 TestEmbeddedPurgeBasics
20 19 TestEmbeddedPurgeFiltered
20 20 TestEmbeddedPurgeConcurrent
20 1 TestEmbeddedQuery
20 2 TestEmbeddedQueryConcurrent
20 3 TestEmbeddedQuick
20 4 TestEmbeddedQuickConcurrent
20 5 TestEmbeddedQuickstart
20 6 TestEmbeddedQuickstartConcurrent
20 8 TestEmbeddedReadyConcurrent
20 9 TestEmbeddedRenamePrefixBasic
20 9 TestEmbeddedRenamePrefixEdgeCases
20 10 TestEmbeddedRenamePrefixConcurrent
20 11 TestEmbeddedReopen
20 12 TestEmbeddedReopenConcurrent
20 13 TestEmbeddedRestore
20 14 TestEmbeddedRestoreConcurrent
20 16 TestEmbeddedSearchConcurrent
20 17 TestEmbeddedSetup
20 18 TestEmbeddedSetupConcurrent
20 19 TestEmbeddedShip
20 20 TestEmbeddedShipConcurrent
20 1 TestEmbeddedShowBasicsAndJSON
20 2 TestEmbeddedShowConcurrent
20 3 TestEmbeddedStale
20 4 TestEmbeddedStaleConcurrent
20 5 TestEmbeddedState
20 6 TestEmbeddedStateConcurrent
20 7 TestEmbeddedStatus
20 8 TestEmbeddedStatusConcurrent
20 9 TestEmbeddedSupersede
20 10 TestEmbeddedSupersedeConcurrent
20 12 TestEmbeddedSwarmConcurrent
20 13 TestEmbeddedTodo
20 14 TestEmbeddedTodoConcurrent
20 15 TestEmbeddedTypes
20 16 TestEmbeddedTypesConcurrent
20 17 TestEmbeddedUndefer
20 18 TestEmbeddedUndeferConcurrent
20 20 TestEmbeddedUpdateConcurrent
20 1 TestEmbeddedVCStatusAndCommit
20 1 TestEmbeddedVCCommitVariantsAndMerge
20 2 TestEmbeddedVCConcurrent
20 3 TestEmbeddedVersion
20 4 TestEmbeddedVersionConcurrent
20 5 TestEmbeddedWhere
20 6 TestEmbeddedWhereConcurrent
# Embedded-Dolt cmd/bd test shard manifest.
#
# Format: <total_shards> <shard_number> <top_level_test_name>

View file

@ -1,235 +1,3 @@
# Proxied-server cmd test shard manifest.
#
# Format: <total_shards> <shard_number> <top_level_test_name>
#
# 15-shard split, bin-packed longest-processing-time-first by
# estimated cost (bd-init count; per-init migration chains dominate), as of
# the test set that existed when this block was last generated.
# Regenerating it (scripts/ci/gen_proxied_shard_manifest.py 15 --write) is
# deliberately NOT part of normal maintenance: this block is frozen for
# pr-risk.yml's/main.yml's legacy fork/push jobs (see engdocs/TESTING.md),
# so every TestProxiedServer*/TestServerMode* function added since it was
# last generated hash-distributes via proxied-test-shard.sh (cksum(name) %
# total) instead of being explicitly listed here. Only reweight it on
# purpose, and expect the shard assignments below to change substantially
# if you do (a fresh run packs every currently-discovered test, not just
# the ones already listed here).
#
# Deliberate exception (Oct 2026): four tests that had been hash-placed
# onto the two shards closest to the 15m -test.timeout are pinned here
# by hand, onto shards whose measured CI job time was lowest; nothing
# that was already listed moved. TestProxiedServerStorageClass and
# TestProxiedServerServeRefusesAStaleCheckpointWithGone (discovered
# from #7129 on) had hashed onto shard 5, which then hit the timeout on
# most Main runs; pinning them returns shard 5 to its earlier test set.
# TestProxiedServerListComments and
# TestProxiedServerServeRefusesAStreamOnADisabledJournal had hashed
# onto shard 3, which runs within about a minute of the timeout. An
# incremental --write keeps pinned lines where they are; a pin is a
# reviewed choice, not a regeneration of this block.
#
# TestProxiedServerOutageReconnectAcceptanceMatrix is pinned to shard 1
# (the lightest in recent Main runs, ~6.5 min) instead of hashing onto
# shard 14, which already runs ~13.5 min of the 15m timeout. These jobs
# have no socat, so today it skips here; the pin keeps shard 14 safe if
# socat is ever installed.
#
# TestProxiedServerClose3 (shard 13) and TestProxiedServerClose4 (shard 2)
# are the two parents gastownhall/beads#7173 split off TestProxiedServerClose
# to pull shard 5 (15m+) back under the timeout, pinned onto the two
# lightest shards of that PR's own run other than shard 5 itself (6m51s
# and 7m57s; shard 5 ran 5m50s only because their 12 subtests had just
# left it, and putting Close3 back would recreate most of the set that
# timed out) rather than hashed; the lines are listed under their shards
# like every other pin.
#
# This file is generated: notes added here are erased by the next
# regeneration. Add them to the generator instead. Known blind spots
# of the cost proxy (it counts inits, not the work done per init):
#
# TestProxiedServerListSortRetiresTheWalk is the most expensive
# entry and one of the cheapest-looking: off a single init it seeds
# 1400 rows and walks the listing three ways. It is the measurement
# behind the claim that sort=priority retires the client-side walk,
# so it has to run on PRs.
#
# TestProxiedServerServeReadParity is the CLI/HTTP read parity
# oracle, the behavioural half of the anti-drift claim. It being in
# this lane, and this lane being required by the PR Risk CI gate,
# is what makes that claim enforced on PRs rather than asserted in
# a doc comment.
15 1 TestProxiedServerBlocked
15 1 TestProxiedServerCompactDolt
15 1 TestProxiedServerCreate
15 1 TestProxiedServerDep2
15 1 TestProxiedServerDepConcurrent
15 1 TestProxiedServerListSortRetiresTheWalk
15 1 TestProxiedServerMolLastActivity
15 1 TestProxiedServerOutageReconnectAcceptanceMatrix
15 1 TestProxiedServerReopenConcurrent
15 1 TestProxiedServerServeReadParity
15 1 TestProxiedServerServeRefusesAStreamOnADisabledJournal
15 1 TestProxiedServerUpdateConcurrentClaim
15 2 TestProxiedServerClose4
15 2 TestProxiedServerCompactHistory
15 2 TestProxiedServerCreate2
15 2 TestProxiedServerDepListSingleIDWarnsOnDroppedExternalEdge
15 2 TestProxiedServerFindDuplicates
15 2 TestProxiedServerListWatchCountsTheOffsetIntoMaxRows
15 2 TestProxiedServerReady2
15 2 TestProxiedServerSQL
15 2 TestProxiedServerServeReadRevision
15 2 TestProxiedServerUpdateConcurrentMetadata
15 2 TestProxiedServerUpdateIfGuards
15 3 TestProxiedServerCascadeParity
15 3 TestProxiedServerDoltLifecycle
15 3 TestProxiedServerMolBurn
15 3 TestProxiedServerMolStale
15 3 TestProxiedServerMoleculeFrontDoorDirectServerProxyParity
15 3 TestProxiedServerSQLDatabaseFlag
15 3 TestProxiedServerServeRelatedIssues
15 3 TestProxiedServerTag
15 3 TestProxiedServerUpdate
15 3 TestProxiedServerWispCreateRejectsEnumViolatingVar
15 4 TestProxiedServerCreateInfraTypeRoutesToWisps
15 4 TestProxiedServerEdit
15 4 TestProxiedServerListComments
15 4 TestProxiedServerMoleculeFrontDoorWriteParity
15 4 TestProxiedServerPour
15 4 TestProxiedServerReassignFence
15 4 TestProxiedServerServeAddComment
15 4 TestProxiedServerServeRelease
15 4 TestProxiedServerUpdate2
15 4 TestProxiedServerUpdateHooks
15 5 TestProxiedServerClose
15 5 TestProxiedServerConfigConcurrent
15 5 TestProxiedServerExport
15 5 TestProxiedServerGateLifecycle
15 5 TestProxiedServerMolReadyGated
15 5 TestProxiedServerMultiStatementSQL
15 5 TestProxiedServerServeAddDependencies
15 5 TestProxiedServerServeRemoveDependency
15 5 TestProxiedServerWispCreate
15 6 TestProxiedServerClose2
15 6 TestProxiedServerContext
15 6 TestProxiedServerExternalCreate
15 6 TestProxiedServerLink
15 6 TestProxiedServerNote
15 6 TestProxiedServerRelate
15 6 TestProxiedServerServeBatchClose
15 6 TestProxiedServerServeSettingsWrites
15 6 TestProxiedServerWispList
15 7 TestProxiedServerAssign
15 7 TestProxiedServerBackupRefusedOnExternalTopology
15 7 TestProxiedServerCount
15 7 TestProxiedServerGC
15 7 TestProxiedServerList
15 7 TestProxiedServerOrphans
15 7 TestProxiedServerReady
15 7 TestProxiedServerServeClaim
15 7 TestProxiedServerServeRefusesAStaleCheckpointWithGone
15 7 TestProxiedServerServeUpdate
15 7 TestProxiedServerState
15 8 TestProxiedServerBatchParity
15 8 TestProxiedServerCountIncludeInfra
15 8 TestProxiedServerDoltRemoteRemove
15 8 TestProxiedServerGateCreateInheritsRepoMetadata
15 8 TestProxiedServerLabel
15 8 TestProxiedServerMolBond
15 8 TestProxiedServerPathHelpers
15 8 TestProxiedServerRecomputeBlocked
15 8 TestProxiedServerServeClaimNext
15 8 TestProxiedServerSetState
15 9 TestProxiedServerChildren
15 9 TestProxiedServerComment
15 9 TestProxiedServerDeferAutoWake
15 9 TestProxiedServerGateFrontDoorDirectServerProxyParity
15 9 TestProxiedServerMemory
15 9 TestProxiedServerPing
15 9 TestProxiedServerPrune
15 9 TestProxiedServerReopen
15 9 TestProxiedServerServeClose
15 9 TestProxiedServerStale
15 10 TestProxiedServerCleanDatabases
15 10 TestProxiedServerDeferUndefer
15 10 TestProxiedServerDuplicates
15 10 TestProxiedServerGraph
15 10 TestProxiedServerPourRejectsEnumViolatingVar
15 10 TestProxiedServerSearch
15 10 TestProxiedServerServeCount
15 10 TestProxiedServerShow
15 10 TestProxiedServerStatuses
15 10 TestProxiedServerUpdateWisp
15 11 TestProxiedServerDelete
15 11 TestProxiedServerDep
15 11 TestProxiedServerGateCheck
15 11 TestProxiedServerGateList
15 11 TestProxiedServerHeartbeat
15 11 TestProxiedServerMolProgress
15 11 TestProxiedServerPriority
15 11 TestProxiedServerPromote
15 11 TestProxiedServerServeCreate
15 11 TestProxiedServerSwarm
15 12 TestProxiedServerDeleteBlockedJSONSingleDoc
15 12 TestProxiedServerHistory
15 12 TestProxiedServerKV
15 12 TestProxiedServerMolShow
15 12 TestProxiedServerQuery
15 12 TestProxiedServerQuick
15 12 TestProxiedServerServeDelete
15 12 TestProxiedServerShow2
15 12 TestProxiedServerTypes
15 12 TestProxiedServerUpdateClosePolicy
15 13 TestProxiedServerClose3
15 13 TestProxiedServerDeleteConcurrent
15 13 TestProxiedServerInfo
15 13 TestProxiedServerMolSquash
15 13 TestProxiedServerReadyClaimConcurrent
15 13 TestProxiedServerReadyWisp
15 13 TestProxiedServerServeEdgeCounts
15 13 TestProxiedServerShow3
15 13 TestProxiedServerStatus
15 13 TestProxiedServerUnclaim
15 13 TestProxiedServerWispGC
15 14 TestProxiedServerCleanDatabasesPurgeDropped
15 14 TestProxiedServerConfig
15 14 TestProxiedServerConfigSetMany
15 14 TestProxiedServerDeleteForceOrphans
15 14 TestProxiedServerImport
15 14 TestProxiedServerInitIfMissing
15 14 TestProxiedServerMolDistill
15 14 TestProxiedServerPrime
15 14 TestProxiedServerReadyConcurrent
15 14 TestProxiedServerServeLifecycle
15 14 TestProxiedServerStorageClass
15 14 TestProxiedServerUnclaimIfAssigneeParity
15 15 TestProxiedServerCloseConcurrent
15 15 TestProxiedServerDeleteWisp
15 15 TestProxiedServerEpic
15 15 TestProxiedServerHuman
15 15 TestProxiedServerLint
15 15 TestProxiedServerMolCurrent
15 15 TestProxiedServerReclaim
15 15 TestProxiedServerServeModeGate
15 15 TestProxiedServerTodo
15 15 TestProxiedServerUpdate3
15 15 TestProxiedServerUpdateConcurrentAppendNotes
# Proxied-server cmd test shard manifest.
#
# Format: <total_shards> <shard_number> <top_level_test_name>

View file

@ -1337,9 +1337,11 @@ jobs:
JOB_STATUS: ${{ job.status }}
run: echo "result=$JOB_STATUS" >> "$GITHUB_OUTPUT"
# pr-risk.yml's embedded-Dolt tier ("Test (Embedded Dolt Cmd N/20)",
# pr-risk.yml's retired embedded-Dolt tier ("Test (Embedded Dolt Cmd N/20)",
# "Test (Embedded Dolt Storage N/5)", "Test (Embedded Dolt Conformance -
# core/audit)") as --config=embedded: the embedded-tagged variants of
# core/audit)") as --config=embedded: those jobs, and the frozen 20-shard
# cmd block they read, no longer exist (see engdocs/TESTING.md) — this runs
# the embedded-tagged variants of
# //cmd/bd:bd_test and //internal/storage/embeddeddolt:embeddeddolt_test.
# The shard variants run the jobs' own shard scripts and manifests, so Bazel
# shard k runs exactly the tests of job k. A job of its own rather than a
@ -1830,13 +1832,15 @@ jobs:
JOB_STATUS: ${{ job.status }}
run: echo "result=$JOB_STATUS" >> "$GITHUB_OUTPUT"
# Like pr-risk.yml's proxied-server tier, "Test (Proxied Dolt Cmd N/15)",
# as --config=doltserver-proxied: //cmd/bd:bd_proxied_test runs the shard
# script, so shard k runs exactly that manifest entry's tests. This lane
# runs 34 shards of its own (.github/scripts/proxied-cmd-test-shards.txt's
# duration-balanced 34-block; see scripts/ci/proxied_test_durations.json),
# not the legacy jobs' 15 — shard k here is not the same split as their
# shard k. Same build as --config=doltserver (race bd_test, non-race
# Like pr-risk.yml's retired proxied-server tier, "Test (Proxied Dolt Cmd
# N/15)" (that job, and the frozen 15-shard block it read, no longer exist;
# see engdocs/TESTING.md), as --config=doltserver-proxied:
# //cmd/bd:bd_proxied_test runs the shard script, so shard k runs exactly
# that manifest entry's tests. This lane runs 34 shards of its own
# (.github/scripts/proxied-cmd-test-shards.txt's duration-balanced 34-block;
# see scripts/ci/proxied_test_durations.json) — shard k here was never the
# same split as the retired job's shard k. Same build as --config=doltserver
# (race bd_test, non-race
# bd_for_tests), shared with bazel-doltserver through the remote cache.
# The test binary is race, as PR Risk's retired jobs built it. It also runs
# //cmd/bd:bd_managed_local_test, the managed-local proxied lifecycle

View file

@ -12,6 +12,16 @@ go_binary(
srcs = ["testdata/preflight-go.go"],
)
# The external `dolt` stand-in for TestRunCompactDoltTargetsOnlyAuthorizedActiveDatabase
# (compact_gc_target_test.go), entirely env-driven at run time, so one
# prebuilt copy covers every test case and mode: the test never needs a host
# Go toolchain under Bazel (bazeltest.RunfileEnv, BEADS_TEST_COMPACT_GC_FIXTURE).
go_binary(
name = "compact_gc_fixture",
testonly = True,
srcs = ["testdata/compact-gc-fixture.go"],
)
go_library(
name = "bd_lib",
srcs = [
@ -1269,6 +1279,7 @@ go_test(
],
data = glob(["testdata/**"]) + [
":bd_for_tests",
":compact_gc_fixture",
":go_srcs",
":preflight_go_fixture",
"//:bd_test_files",
@ -1281,6 +1292,7 @@ go_test(
env = {
"BEADS_TEST_BD_BINARY": "$(rlocationpath :bd_for_tests)",
"BEADS_TEST_GOFMT": "$(rlocationpath @go_sdk//:bin/gofmt)",
"BEADS_TEST_COMPACT_GC_FIXTURE": "$(rlocationpath :compact_gc_fixture)",
"BEADS_TEST_PREFLIGHT_GO": "$(rlocationpath :preflight_go_fixture)",
},
shard_count = 16,
@ -1413,13 +1425,12 @@ go_test(
# check_shard_coverage.py fails unless the two ranges tile the block exactly
# once). Both run this configuration's bd_test with
# BEADS_TEST_EMBEDDED_DOLT=1 and the race bd (//cmd/bd:bd under
# --config=embedded) as the subprocess binary. This is a *different* shard
# count from PR Risk's legacy fork/push jobs, "Test (Embedded Dolt Cmd
# N/20)" (pr-risk.yml): those keep running 20 shards of their own, frozen
# hand-assigned block in the same manifest file — shard k here is not the
# same split as their shard k. Each repeats bd_test's `env` (the wrapper
# cannot inherit it). The two shard_counts must sum to --shard-total, the
# manifest block's shard total.
# --config=embedded) as the subprocess binary. PR Risk's legacy fork/push
# jobs, "Test (Embedded Dolt Cmd N/20)" (pr-risk.yml), which used to run 20
# shards of their own hand-assigned block in the same manifest file, no
# longer exist; the manifest now holds only this 100-shard block. Each
# repeats bd_test's `env` (the wrapper cannot inherit it). The two
# shard_counts must sum to --shard-total, the manifest block's shard total.
#
# Tags:
# embedded: the embedded-Dolt tier; excluded from --config=prcore/ci,
@ -1511,15 +1522,14 @@ sh_test(
# functions, assigned by the manifest's 34-shard block (duration-balanced;
# see scripts/ci/proxied_test_durations.json), against this configuration's
# (race) bd_test with BEADS_TEST_PROXIED_SERVER=1 and the non-race
# bd_for_tests as the subprocess binary. This is a *different* shard count
# from PR Risk's legacy fork/push jobs, "Test (Proxied Dolt Cmd N/15)"
# (pr-risk.yml): those keep running 15 shards of their own,
# frozen bd-init-cost-proxy block in the same manifest file — shard k here
# is not the same split as their shard k. PR Risk builds its bd-cmd-test
# with -race, as here. The shared proxied harness gets its dolt sql-server from
# internal/testutil's local backend (the pinned dolt in the test
# wrapper's runfiles) instead of the jobs' container, and fails instead
# of skipping when it cannot start one. It repeats bd_test's `env`.
# bd_for_tests as the subprocess binary. PR Risk's legacy fork/push jobs,
# "Test (Proxied Dolt Cmd N/15)" (pr-risk.yml), which used to run 15 shards
# of their own frozen bd-init-cost-proxy block in the same manifest file, no
# longer exist; the manifest now holds only this 34-shard block. The shared
# proxied harness gets its dolt sql-server from internal/testutil's local
# backend (the pinned dolt in the test wrapper's runfiles) instead of a
# container, and fails instead of skipping when it cannot start one. It
# repeats bd_test's `env`.
# shard_count must stay equal to the manifest block's shard total.
#
# Tags:
@ -1604,6 +1614,7 @@ sh_test(
data = [
":bd_for_tests",
":bd_test",
":compact_gc_fixture",
":go_srcs",
":preflight_go_fixture",
"@go_sdk//:bin/gofmt",
@ -1612,6 +1623,7 @@ sh_test(
"BEADS_TEST_BD_BINARY": "$(rlocationpath :bd_for_tests)",
"BEADS_TEST_DOLT_SERVER": "local",
"BEADS_TEST_GOFMT": "$(rlocationpath @go_sdk//:bin/gofmt)",
"BEADS_TEST_COMPACT_GC_FIXTURE": "$(rlocationpath :compact_gc_fixture)",
"BEADS_TEST_PREFLIGHT_GO": "$(rlocationpath :preflight_go_fixture)",
"BEADS_TEST_PROXIED_LOCAL": "1",
"BEADS_TEST_REQUIRE_DOLT_CONTAINER": "1",
@ -1660,6 +1672,7 @@ sh_test(
data = [
":bd_for_tests",
":bd_test",
":compact_gc_fixture",
":dolt_server_pinned_shards.txt",
":preflight_go_fixture",
"@go_sdk//:bin/gofmt",
@ -1670,6 +1683,7 @@ sh_test(
# The integration jobs' global git identity (tools/bazel/test_env.sh).
"BEADS_TEST_GIT_IDENTITY": "1",
"BEADS_TEST_GOFMT": "$(rlocationpath @go_sdk//:bin/gofmt)",
"BEADS_TEST_COMPACT_GC_FIXTURE": "$(rlocationpath :compact_gc_fixture)",
"BEADS_TEST_PREFLIGHT_GO": "$(rlocationpath :preflight_go_fixture)",
"BEADS_TEST_REQUIRE_DOLT_CONTAINER": "1",
# TestSharedServerConcurrent (skipped everywhere else) as a bounded

View file

@ -14,6 +14,7 @@ import (
"github.com/steveyegge/beads/internal/beads"
"github.com/steveyegge/beads/internal/hooks"
"github.com/steveyegge/beads/internal/storage"
"github.com/steveyegge/beads/internal/testutil/bazeltest"
)
type compactGCStoreStub struct {
@ -292,31 +293,59 @@ func TestRunCompactDoltTargetsOnlyAuthorizedActiveDatabase(t *testing.T) {
}
}
// compactGCFixtureEnv names the environment variable a go_test sets from its
// BUILD file to the prebuilt compact-gc-fixture.go binary (built once by
// Bazel as a testonly go_binary), the same pattern preflight_test.go uses for
// BEADS_TEST_PREFLIGHT_GO:
//
// data = [":compact_gc_fixture"],
// env = {"BEADS_TEST_COMPACT_GC_FIXTURE": "$(rlocationpath :compact_gc_fixture)"},
//
// so this test never needs a host Go toolchain under Bazel.
const compactGCFixtureEnv = "BEADS_TEST_COMPACT_GC_FIXTURE"
// buildCompactGCFixture produces a directory holding a `dolt` executable that
// stands in for the external Dolt CLI. Under Bazel it copies the prebuilt
// fixture from runfiles (bazeltest.RunfileEnv); outside Bazel (plain
// `go test`) it falls back to compiling testdata/compact-gc-fixture.go with
// the host Go toolchain, as before.
func buildCompactGCFixture(t *testing.T) string {
t.Helper()
dir := t.TempDir()
const source = `package main
import ("encoding/json"; "fmt"; "os")
func main() {
dir, err := os.Getwd(); if err != nil { panic(err) }
log, err := os.OpenFile(os.Getenv("BEADS_COMPACT_GC_FIXTURE_LOG"), os.O_CREATE|os.O_APPEND|os.O_WRONLY, 0600); if err != nil { panic(err) }
if err := json.NewEncoder(log).Encode(struct { Dir string; Args []string }{dir, os.Args[1:]}); err != nil { panic(err) }; if err := log.Close(); err != nil { panic(err) }
mode := os.Getenv("BEADS_COMPACT_GC_FIXTURE_MODE")
if mode == "fallback" && len(os.Args) == 4 { fmt.Fprintln(os.Stderr, "unknown flag: --archive-level"); os.Exit(23) }
if mode == "failure" { fmt.Fprintln(os.Stderr, "genuine GC failure"); os.Exit(23) }
if err := os.WriteFile("gc-ran", []byte("collected"), 0600); err != nil { panic(err) }
}
`
path := filepath.Join(dir, "fixture.go")
if err := os.WriteFile(path, []byte(source), 0o600); err != nil {
t.Fatal(err)
}
name := "dolt"
if runtime.GOOS == "windows" {
name += ".exe"
}
cmd := exec.Command("go", "build", "-o", filepath.Join(dir, name), path)
cmd.Dir = dir
target := filepath.Join(dir, name)
if bazeltest.IsBazel() {
fixture, err := bazeltest.RunfileEnv(compactGCFixtureEnv)
if err != nil {
t.Fatal(err)
}
data, err := os.ReadFile(fixture)
if err != nil {
t.Fatalf("read prebuilt GC fixture: %v", err)
}
if err := os.WriteFile(target, data, 0o755); err != nil {
t.Fatalf("install prebuilt GC fixture: %v", err)
}
return dir
}
realGo, err := exec.LookPath("go")
if err != nil {
t.Fatalf("locate Go toolchain for subprocess fixture: %v", err)
}
// Absolute, resolved from this file's own directory rather than the test
// binary's working directory: `go test` happens to run from the package
// directory, but nothing here should depend on that (and t.Chdir, a
// -run subset launched from elsewhere, etc. would otherwise break this
// silently with a confusing "no such file" from `go build`).
_, thisFile, _, ok := runtime.Caller(0)
if !ok {
t.Fatal("runtime.Caller failed")
}
path := filepath.Join(filepath.Dir(thisFile), "testdata", "compact-gc-fixture.go")
cmd := exec.Command(realGo, "build", "-o", target, path)
cmd.Env = append(os.Environ(), "CGO_ENABLED=0", "GOWORK=off")
if output, err := cmd.CombinedOutput(); err != nil {
t.Fatalf("build owned GC fixture: %v\n%s", err, output)

View file

@ -16,6 +16,7 @@ import (
"github.com/steveyegge/beads/internal/metrics"
"github.com/steveyegge/beads/internal/migration"
"github.com/steveyegge/beads/internal/testutil"
"github.com/steveyegge/beads/internal/testutil/bazeltest"
"github.com/steveyegge/beads/internal/testutil/credentialcmd"
"github.com/steveyegge/beads/internal/workspacegate"
)
@ -176,23 +177,28 @@ func testMainInner(m *testing.M) int {
// /tmp and exhausted tmpfs over time (bd-3q2u).
testTempRoot = tmp
// Preserve Go build cache before changing HOME.
// On macOS, GOCACHE defaults to $HOME/Library/Caches/go-build.
// Changing HOME would cause tests that run `go build` (e.g., TestShow)
// to miss the cache and do a full CGO rebuild (~80s each).
if os.Getenv("GOCACHE") == "" {
if out, err := exec.Command("go", "env", "GOCACHE").Output(); err == nil {
_ = os.Setenv("GOCACHE", strings.TrimSpace(string(out)))
// Preserve Go build cache before changing HOME. On macOS, GOCACHE
// defaults to $HOME/Library/Caches/go-build. Changing HOME would cause
// tests that run `go build` (e.g., TestShow) to miss the cache and do a
// full CGO rebuild (~80s each). Under Bazel there is no host Go
// toolchain on the hermetic test PATH (tools/bazel/test_env.sh) and no
// in-test `go build` to preserve a cache for, so skip the host `go env`
// exec entirely rather than rely on its error being ignored.
if !bazeltest.IsBazel() {
if os.Getenv("GOCACHE") == "" {
if out, err := exec.Command("go", "env", "GOCACHE").Output(); err == nil {
_ = os.Setenv("GOCACHE", strings.TrimSpace(string(out)))
}
}
}
// Same for the module cache: GOMODCACHE defaults to $HOME/go/pkg/mod,
// so without this the in-test `go build` (buildEmbeddedBD) re-downloads
// every dependency into the temp HOME on each run — slow, and a hard
// failure when the network is unavailable.
if os.Getenv("GOMODCACHE") == "" {
if out, err := exec.Command("go", "env", "GOMODCACHE").Output(); err == nil {
_ = os.Setenv("GOMODCACHE", strings.TrimSpace(string(out)))
// Same for the module cache: GOMODCACHE defaults to $HOME/go/pkg/mod,
// so without this the in-test `go build` (buildEmbeddedBD) re-downloads
// every dependency into the temp HOME on each run — slow, and a hard
// failure when the network is unavailable.
if os.Getenv("GOMODCACHE") == "" {
if out, err := exec.Command("go", "env", "GOMODCACHE").Output(); err == nil {
_ = os.Setenv("GOMODCACHE", strings.TrimSpace(string(out)))
}
}
}

45
cmd/bd/testdata/compact-gc-fixture.go vendored Normal file
View file

@ -0,0 +1,45 @@
// Command compact-gc-fixture stands in for an external `dolt` binary in
// TestRunCompactDoltTargetsOnlyAuthorizedActiveDatabase
// (cmd/bd/compact_gc_target_test.go). It is entirely env-driven at run time
// (BEADS_COMPACT_GC_FIXTURE_LOG, BEADS_COMPACT_GC_FIXTURE_MODE), so one
// prebuilt copy serves every test case and every mode: the test never needs
// to invoke a Go toolchain to produce it.
package main
import (
"encoding/json"
"fmt"
"os"
)
func main() {
dir, err := os.Getwd()
if err != nil {
panic(err)
}
log, err := os.OpenFile(os.Getenv("BEADS_COMPACT_GC_FIXTURE_LOG"), os.O_CREATE|os.O_APPEND|os.O_WRONLY, 0600)
if err != nil {
panic(err)
}
if err := json.NewEncoder(log).Encode(struct {
Dir string
Args []string
}{dir, os.Args[1:]}); err != nil {
panic(err)
}
if err := log.Close(); err != nil {
panic(err)
}
mode := os.Getenv("BEADS_COMPACT_GC_FIXTURE_MODE")
if mode == "fallback" && len(os.Args) == 4 {
fmt.Fprintln(os.Stderr, "unknown flag: --archive-level")
os.Exit(23)
}
if mode == "failure" {
fmt.Fprintln(os.Stderr, "genuine GC failure")
os.Exit(23)
}
if err := os.WriteFile("gc-ran", []byte("collected"), 0600); err != nil {
panic(err)
}
}

View file

@ -1159,11 +1159,12 @@ Required` requires them to have run remotely and passed.
- every Bazel shard of `//cmd/bd:bd_embedded_test` and
`//cmd/bd:bd_embedded_part2_test` (50 each: shards 1-50 and 51-100 of
the manifest's 100-shard block, which their ranges must tile exactly
once; the manifest's frozen 20-shard block was the retired
`test-embedded-cmd` jobs' split, F1), `//internal/storage/embeddeddolt:embeddeddolt_embedded_test` (40;
once; the manifest's frozen 20-shard block, once the retired
`test-embedded-cmd` jobs' split, has been deleted, F1),
`//internal/storage/embeddeddolt:embeddeddolt_embedded_test` (40;
the frozen 5-shard block was `test-embedded-storage`'s, F1),
`//cmd/bd:bd_proxied_test` (34; the frozen 15-shard block was
`test-proxied-cmd`'s, F2) and
`//cmd/bd:bd_proxied_test` (34; the manifest's frozen 15-shard block,
once `test-proxied-cmd`'s split, has been deleted, F2) and
`//internal/storage/dolt:dolt_server_full_test` (16) to have run
exactly the tests its shard script lists (list-only mode, minus
`TestMain`, which `grep '^func Test'` lists but which is never a

View file

@ -184,9 +184,9 @@ builds with the integration tag like `--config=integration`. Each shard runs
its CI job's shard script, so for `--config=doltserver-integration` Bazel
shard k runs the tests of job k+1 (both split the manifest's 16-shard block
the same way). `--config=doltserver-proxied`'s `bd_proxied_test` instead
runs the manifest's own 34-shard block — bin-packed by measured duration,
not the legacy jobs' 15-shard, bd-init-cost-proxy block — so shard k there
is not job k+1's tests; it is a different split of the same tests.
runs the manifest's own 34-shard block, bin-packed by measured duration (the
manifest's legacy 15-shard, bd-init-cost-proxy block, once read by PR Risk's
retired fork/push jobs, has been deleted; those jobs no longer exist).
`--config=doltserver-cmd` (`//cmd/bd:bd_dolt_server_test`, bazel.yml's
advisory `bazel-cmd-dolt` job) runs the whole integration-tagged cmd/bd

View file

@ -15,9 +15,29 @@ cd "$REPO_ROOT"
GOFMT_BIN="$("$SCRIPT_DIR/gofmt-bin.sh")"
describe_gofmt() {
local bin="$1" version=""
if command -v go >/dev/null 2>&1; then
version="$(go version "$bin" 2>/dev/null | awk '{ print $NF }')"
local bin="$1" version="" go_bin=""
if [[ -n "${TEST_SRCDIR:-}" ]]; then
# Under `bazel test` a bare `go` on PATH can be the farm's exit-127
# mask for the host toolchain, so only call a go binary the test
# explicitly declared as data: BEADS_TEST_GO, set the same way
# scripts/prlintmake:prlintmake_test wires in the registered SDK's go
# ($(rlocationpath @go_sdk//:bin/go)). A caller that did not wire it
# up (e.g. scripts/repochecks:fmt_test) just gets the version-less
# fallback below, never a PATH `go`.
if [[ -n "${BEADS_TEST_GO:-}" ]]; then
local candidate="$BEADS_TEST_GO"
if [[ "$candidate" != /* ]]; then
candidate="$TEST_SRCDIR/$candidate"
fi
if [[ -x "$candidate" ]]; then
go_bin="$candidate"
fi
fi
elif command -v go >/dev/null 2>&1; then
go_bin="go"
fi
if [[ -n "$go_bin" ]]; then
version="$("$go_bin" version "$bin" 2>/dev/null | awk '{ print $NF }')"
fi
if [[ -n "$version" ]]; then
printf '%s (%s)' "$version" "$bin"

View file

@ -4,32 +4,35 @@
Discovery must stay in step with embedded-test-shard.sh: both select every
top-level TestEmbedded* function from cmd/bd/*_embedded_test.go.
Two cost models, selected by --weights (default: inits, matching
Two cost models, selected by --weights (default: duration, matching
gen_proxied_shard_manifest.py's convention — see that script's module
docstring for the full rationale this one shares):
--weights=inits (default): a cheap static proxy for wall-time — the
function's subtest count (t.Run( occurrences), floored at 1. Unlike the
proxied generator's bd-init count, embedded cmd tests do not share one
dominant per-call cost center, so subtest count is the simplest proxy
that is still monotonic in "more scenarios this function exercises."
--weights=duration: measured wall-time from
--weights=duration (default): measured wall-time from
scripts/ci/embedded_cmd_test_durations.json (see that file's header for
provenance). A function missing from that file — a test added since it
was last captured — falls back to its inits cost times the file's
"seconds_per_init_fallback" ratio. A from-scratch pack models the shard
process (pack_concurrent): functions without a top-level t.Parallel()
run one after another, the rest share -test.parallel=4 slots.
run one after another, the rest share -test.parallel=4 slots. The
manifest's only remaining block is duration-packed (see below), so this
is also the only --weights whose render() header matches what is
committed — a bare `--write` with no flags reproduces the committed file
instead of silently clobbering its header with the inits one.
This file's existing 20-shard block (hand-assigned, round-robin; see its own
header) is the FROZEN block PR Risk's and main.yml's legacy fork/push jobs
read — see engdocs/TESTING.md. Do NOT regenerate it. This generator's
--weights=duration path targets a *separate*, Bazel-only shard total (the
bazel-embedded job in .github/workflows/bazel.yml): a different total_shards
value, written as its own block in the same manifest file (see
split_blocks()/write_block() in _embedded_shard_manifest_lib.py for how one
file holds more than one independent block).
--weights=inits: a cheap static proxy for wall-time — the function's
subtest count (t.Run( occurrences), floored at 1. Unlike the proxied
generator's bd-init count, embedded cmd tests do not share one dominant
per-call cost center, so subtest count is the simplest proxy that is
still monotonic in "more scenarios this function exercises." Pass this
explicitly if you want it; --write still honors it, it is just no longer
the default.
This generator's --weights=duration path targets the Bazel-only shard total
(the bazel-embedded job in .github/workflows/bazel.yml). The legacy,
hand-assigned 20-shard block that PR Risk's and main.yml's retired fork/push
jobs used to read has been deleted (those jobs no longer exist; see
engdocs/TESTING.md): the manifest file now holds exactly the one live block.
--write is INCREMENTAL by default (see --repack to force a full rebalance);
--check verifies exact-once NAME coverage only, not shard assignments. Both
@ -54,7 +57,9 @@ from _embedded_shard_manifest_lib import ( # noqa: E402
default_manifest_path = '.github/scripts/embedded-cmd-test-shards.txt'
durations_filename = 'embedded_cmd_test_durations.json'
default_total_shards = 20
# The Bazel-only block's total (bd_embedded_test + bd_embedded_part2_test's
# shard_counts in cmd/bd/BUILD.bazel must sum to this).
default_total_shards = 100
# .bazelrc's test:embedded --test_arg=-test.parallel=4: how many t.Parallel
# top-level tests one shard's process runs at once (--weights=duration).
test_parallelism = 4
@ -131,35 +136,27 @@ def render(total, shards, weights):
out.append('#')
out.append('# Format: <total_shards> <shard_number> <top_level_test_name>')
out.append('#')
if total == default_total_shards and weights == 'inits':
out.append(f'# {total}-shard split. This block is hand-assigned (round-robin,')
out.append('# not cost-balanced) and FROZEN for pr-risk.yml/main.yml\'s legacy')
out.append('# fork/push jobs (see engdocs/TESTING.md): do not regenerate it. Newly-')
out.append('# added tests not listed here hash-distribute via embedded-test-shard.sh:')
out.append('# cksum(name) % total. --check does not cover this block; only the')
out.append('# Bazel-only --weights=duration block is checked in CI.')
else:
out.append(f'# {total}-shard split for the Bazel-only embedded-Dolt cmd tier')
out.append('# (bazel-embedded in .github/workflows/bazel.yml), bin-packed')
out.append('# longest-first by measured wall-time from')
out.append(f'# scripts/ci/{durations_filename} (see that file for provenance')
out.append('# and its "relative weight, not absolute SLA" caveat) onto the shard')
out.append('# whose estimated wall time grows least: serial tests (no top-level')
out.append(f'# t.Parallel) add up, parallel ones share -test.parallel={test_parallelism}.')
out.append(f'# Regenerate with scripts/ci/gen_embedded_cmd_shard_manifest.py {total}')
out.append('# --weights=duration --write after adding, splitting or removing')
out.append('# TestEmbedded* functions in cmd/bd/*_embedded_test.go. --write is')
out.append('# incremental: it keeps every already-assigned test on its current shard')
out.append('# and only places newly-discovered names, onto whichever shard is')
out.append('# currently lightest, so two unrelated PRs that each add one test do not')
out.append('# textually conflict or silently invalidate each other\'s packing. Pass')
out.append('# --repack to force a full from-scratch rebalance instead (a deliberate,')
out.append('# separate change — expect a large diff). --check only verifies every')
out.append('# discovered name is listed here exactly once (no stale or duplicate')
out.append('# entries); it does NOT verify the packing is still well-balanced.')
out.append('#')
out.append('# This file is generated: notes added here are erased by the next')
out.append('# regeneration. Add them to the generator or the durations file instead.')
out.append(f'# {total}-shard split for the Bazel-only embedded-Dolt cmd tier')
out.append('# (bazel-embedded in .github/workflows/bazel.yml), bin-packed')
out.append('# longest-first by measured wall-time from')
out.append(f'# scripts/ci/{durations_filename} (see that file for provenance')
out.append('# and its "relative weight, not absolute SLA" caveat) onto the shard')
out.append('# whose estimated wall time grows least: serial tests (no top-level')
out.append(f'# t.Parallel) add up, parallel ones share -test.parallel={test_parallelism}.')
out.append(f'# Regenerate with scripts/ci/gen_embedded_cmd_shard_manifest.py {total}')
out.append('# --weights=duration --write after adding, splitting or removing')
out.append('# TestEmbedded* functions in cmd/bd/*_embedded_test.go. --write is')
out.append('# incremental: it keeps every already-assigned test on its current shard')
out.append('# and only places newly-discovered names, onto whichever shard is')
out.append('# currently lightest, so two unrelated PRs that each add one test do not')
out.append('# textually conflict or silently invalidate each other\'s packing. Pass')
out.append('# --repack to force a full from-scratch rebalance instead (a deliberate,')
out.append('# separate change — expect a large diff). --check only verifies every')
out.append('# discovered name is listed here exactly once (no stale or duplicate')
out.append('# entries); it does NOT verify the packing is still well-balanced.')
out.append('#')
out.append('# This file is generated: notes added here are erased by the next')
out.append('# regeneration. Add them to the generator or the durations file instead.')
out.append('')
for i in range(total):
for name in sorted(shards[i]):
@ -171,7 +168,7 @@ def render(total, shards, weights):
def main():
ap = argparse.ArgumentParser(description=__doc__.split('\n\n')[0])
ap.add_argument('total_shards', nargs='?', type=int, default=None)
ap.add_argument('--weights', choices=['inits', 'duration'], default='inits')
ap.add_argument('--weights', choices=['inits', 'duration'], default='duration')
ap.add_argument('--manifest', default=default_manifest_path,
help=f'manifest file to read/write (default: {default_manifest_path})')
mode = ap.add_mutually_exclusive_group()
@ -186,20 +183,6 @@ def main():
'from-scratch LPT pack')
args = ap.parse_args()
if args.total_shards is None:
# N3 (F1 review): omitting total_shards used to default to 20, the
# FROZEN legacy block's own total, with --weights=inits also
# defaulting on -- so a bare `--write` silently rewrote the frozen
# legacy 20-shard block instead of erroring or targeting the
# Bazel-only block. Require the caller to say which total they mean
# whenever they're about to write; --check/dry-run keep the old
# default_total_shards=20 fallback since reading the legacy block is
# harmless and already how engdocs/TESTING.md documents inspecting it.
if args.write:
ap.error(f'total_shards is required with --write (the default, {default_total_shards}, '
'is the FROZEN legacy block -- see its header in embedded-cmd-test-shards.txt; '
'it must not be regenerated). Pass the Bazel-only total explicitly instead, e.g. '
'"100 --weights=duration" for bazel-embedded\'s block (see cmd/bd/BUILD.bazel\'s '
'bd_embedded_test/bd_embedded_part2_test --shard-total for the current value)')
args.total_shards = default_total_shards
inits = discover_inits_cost()

View file

@ -5,24 +5,24 @@ Discovery must stay in step with proxied-test-shard.sh: both select the
TestProxiedServer* and TestServerMode* entry points, which share the
requireSharedProxiedServer gate and the shared dolt sql-server.
Two cost models, selected by --weights (default: inits, unchanged from
before this flag existed — the committed 15-shard block for the legacy
pr-risk.yml/main.yml jobs was generated this way and is NOT regenerated by
this script's --weights=duration path; it stays frozen so those jobs' shard
assignment keeps its exact current behavior):
Two cost models, selected by --weights (default: duration):
--weights=inits (default): bd-init count (newSharedProxiedProject
occurrences) per top-level function — a cheap static proxy for
wall-time that undercounts a test whose cost is dominated by what it
does per init rather than how many inits it runs (see the two
known-expensive entries in the committed 15-shard manifest's header).
--weights=duration: measured wall-time from
--weights=duration (default): measured wall-time from
scripts/ci/proxied_test_durations.json (that file's header explains its
provenance and limits). A function missing from that file — a test
added since it was last captured — falls back to its inits cost times
the file's "seconds_per_init_fallback" ratio, so a brand-new test still
gets a plausible seconds-shaped weight instead of a unit-mismatched one.
The manifest's only remaining block is duration-packed (see below), so
this is also the only --weights that renders a header matching what is
committed — a bare `--write` with no flags reproduces the committed file
instead of silently clobbering its header with the inits one.
--weights=inits: bd-init count (newSharedProxiedProject occurrences) per
top-level function — a cheap static proxy for wall-time that undercounts
a test whose cost is dominated by what it does per init rather than how
many inits it runs. Pass this explicitly if you want it; --write still
honors it, it is just no longer the default.
Either way, a from-scratch pack puts TOTAL shards' worth of tests together
longest-processing-time-first (heaviest first onto the currently lightest
@ -54,16 +54,15 @@ and read the stderr shard-load report, or pass --check anyway: it still
reports the block's current loads to stderr even though it only gates on
name coverage.
The manifest file holds more than one block (one per total_shards value in
use): the legacy, frozen 15-shard block pr-risk.yml/main.yml's jobs read, and
the Bazel-only lane's own block (34 shards since 2026-10-09). By default this prints
only the requested total's block to stdout, which is NOT safe to redirect
straight into the manifest file (`gen... 30 > file` deletes every other
block). Use --write to update the file in place instead: it replaces only
the block for the requested total_shards and leaves every other block
untouched. Use --check to verify the committed block for a total_shards
has exact-once name coverage, without writing anything — this is the form a
CI policy check should call.
The manifest file holds the Bazel-only lane's one block (34 shards since
2026-10-09; the legacy, frozen 15-shard block pr-risk.yml/main.yml's retired
fork/push jobs used to read has been deleted — those jobs no longer exist,
see engdocs/TESTING.md). By default this prints the requested total's block
to stdout, which is NOT safe to redirect straight into the manifest file
(`gen... 30 > file` loses the header if total does not match what is
committed). Use --write to update the file in place instead. Use --check to
verify the committed block for a total_shards has exact-once name coverage,
without writing anything — this is the form a CI policy check should call.
Usage: gen_proxied_shard_manifest.py [total_shards] [--weights=inits|duration]
[--write [--repack] | --check] [--manifest PATH]
@ -187,44 +186,12 @@ def render(total, shards, weights):
if weights == 'inits':
out.append(f'# {total}-shard split, bin-packed longest-processing-time-first by')
out.append('# estimated cost (bd-init count; per-init migration chains dominate).')
out.append(f'# Regenerating it (scripts/ci/gen_proxied_shard_manifest.py {total} --write)')
out.append('# is deliberately NOT part of normal maintenance; see the frozen note')
out.append('# below. Newly-added tests not listed here hash-distribute via')
out.append(f'# Regenerate with scripts/ci/gen_proxied_shard_manifest.py {total} --write.')
out.append('# Newly-added tests not listed here hash-distribute via')
out.append('# proxied-test-shard.sh: cksum(name) % total.')
out.append('#')
out.append('# This file is generated: notes added here are erased by the next')
out.append('# regeneration. Add them to the generator instead. Known blind spots')
out.append('# of the cost proxy (it counts inits, not the work done per init):')
out.append('#')
out.append('# TestProxiedServerListSortRetiresTheWalk is the most expensive')
out.append('# entry and one of the cheapest-looking: off a single init it seeds')
out.append('# 1400 rows and walks the listing three ways. It is the measurement')
out.append('# behind the claim that sort=priority retires the client-side walk,')
out.append('# so it has to run on PRs.')
out.append('#')
out.append('# TestProxiedServerServeReadParity is the CLI/HTTP read parity')
out.append('# oracle, the behavioural half of the anti-drift claim. It being in')
out.append('# this lane, and this lane being required by the PR Risk CI gate,')
out.append('# is what makes that claim enforced on PRs rather than asserted in')
out.append('# a doc comment.')
out.append('#')
out.append(f'# This {total}-shard block is frozen for pr-risk.yml/main.yml\'s legacy')
out.append('# fork/push jobs (see engdocs/TESTING.md): do not regenerate it even')
out.append('# after adding tests (they hash-distribute until a maintainer chooses')
out.append('# to reweight on purpose). --check does not cover this block; only the')
out.append('# Bazel-only --weights=duration block is checked in CI.')
out.append('#')
out.append('# Hand-pinned lines are allowed: when a hash-placed test overloads a')
out.append('# shard toward the 15m -test.timeout, a reviewed change may list it')
out.append('# on a lighter shard without moving anything already listed (done in')
out.append('# Oct 2026 for TestProxiedServerStorageClass,')
out.append('# TestProxiedServerServeRefusesAStaleCheckpointWithGone,')
out.append('# TestProxiedServerListComments and')
out.append('# TestProxiedServerServeRefusesAStreamOnADisabledJournal; and for')
out.append('# TestProxiedServerOutageReconnectAcceptanceMatrix, pinned to shard 1')
out.append('# instead of hashing onto shard 14; and for TestProxiedServerClose3')
out.append('# and TestProxiedServerClose4, the parents #7173 split off')
out.append('# TestProxiedServerClose, pinned to shards 13 and 2).')
out.append('# regeneration. Add them to the generator instead.')
else:
out.append(f'# {total}-shard split for the Bazel-only proxied-server tier')
out.append('# (bazel-proxied in .github/workflows/bazel.yml), bin-packed')
@ -361,8 +328,8 @@ def check_coverage(path, total, universe):
def main():
ap = argparse.ArgumentParser(description=__doc__.split('\n\n')[0])
ap.add_argument('total_shards', nargs='?', type=int, default=15)
ap.add_argument('--weights', choices=['inits', 'duration'], default='inits')
ap.add_argument('total_shards', nargs='?', type=int, default=34)
ap.add_argument('--weights', choices=['inits', 'duration'], default='duration')
ap.add_argument('--manifest', default=default_manifest_path,
help=f'manifest file to read/write (default: {default_manifest_path})')
mode = ap.add_mutually_exclusive_group()

View file

@ -19,18 +19,25 @@ beads_test_env_enter() {
export BEADS_TEST_ENV_ROOT="$root"
export BEADS_TEST_ENV_ACTIVE=1
if [[ -z "${GOCACHE:-}" ]]; then
local go_cache
go_cache="$(go env GOCACHE 2>/dev/null || true)"
if [[ -n "$go_cache" ]]; then
export GOCACHE="$go_cache"
# Under `bazel test` (TEST_SRCDIR set) a bare `go` on PATH can be the
# farm's exit-127 mask for the host toolchain, and nothing downstream of
# this wrapper in that mode builds Go code or needs a cache directory
# (scripts/test.sh's prebuilt-test-binary path runs a binary Bazel already
# built), so skip the probe entirely rather than risk invoking it.
if [[ -z "${TEST_SRCDIR:-}" ]]; then
if [[ -z "${GOCACHE:-}" ]]; then
local go_cache
go_cache="$(go env GOCACHE 2>/dev/null || true)"
if [[ -n "$go_cache" ]]; then
export GOCACHE="$go_cache"
fi
fi
fi
if [[ -z "${GOMODCACHE:-}" ]]; then
local go_mod_cache
go_mod_cache="$(go env GOMODCACHE 2>/dev/null || true)"
if [[ -n "$go_mod_cache" ]]; then
export GOMODCACHE="$go_mod_cache"
if [[ -z "${GOMODCACHE:-}" ]]; then
local go_mod_cache
go_mod_cache="$(go env GOMODCACHE 2>/dev/null || true)"
if [[ -n "$go_mod_cache" ]]; then
export GOMODCACHE="$go_mod_cache"
fi
fi
fi

View file

@ -82,10 +82,12 @@ func bazelEmbeddedCmdParts(t *testing.T) []embeddedCmdPart {
// bazelEmbeddedCmdShardCount returns the shard total of the bazel-embedded
// lane's cmd/bd manifest block, which embeddedCmdTargets split between them
// (bazelEmbeddedCmdParts): the single source of truth for the Bazel-only
// lane's cmd/bd shard split (slice F1), which no longer has to equal PR
// lane's cmd/bd shard split (slice F1). It no longer has to equal PR
// Risk's/main.yml's legacy "Test (Embedded Dolt Cmd N/20)" fork/push jobs'
// matrix size — mirrors bazelProxiedShardCount (F2; see that function's doc
// comment below for the shared rationale, not repeated here).
// matrix size — those jobs, and the frozen 20-shard block they read, have
// since been deleted (see engdocs/TESTING.md) — mirrors bazelProxiedShardCount
// (F2; see that function's doc comment below for the shared rationale, not
// repeated here).
func bazelEmbeddedCmdShardCount(t *testing.T) int {
t.Helper()
return bazelEmbeddedCmdParts(t)[0].total

View file

@ -333,9 +333,12 @@ func TestBazelRetiredLanesCheckListedTestsRan(t *testing.T) {
{bazelEmbedJobName, "embedded", []suite{
// bazel-embedded runs its own duration-balanced manifest blocks
// (scripts/ci/embedded_{cmd,storage}_test_durations.json), not
// PR Risk's frozen 20- and 5-shard blocks (slice F1): neither is
// a drop-in retirement of its legacy job's shard count, just its
// tests. bazelEmbeddedCmdShardCount/bazelEmbeddedStorageShardCount
// PR Risk's 20-shard cmd block, which was frozen and has since
// been deleted (those jobs no longer exist; see
// engdocs/TESTING.md), or the still-frozen 5-shard storage block
// (slice F1): neither is a drop-in retirement of its legacy
// job's shard count, just its tests.
// bazelEmbeddedCmdShardCount/bazelEmbeddedStorageShardCount
// read the real counts from BUILD.bazel, making these suites'
// `want` (below) genuine cross-file pins against bazel.yml's own
// check_shard_coverage.py arguments, not independently
@ -347,9 +350,11 @@ func TestBazelRetiredLanesCheckListedTestsRan(t *testing.T) {
}, []string{"//internal/storage/embeddeddolt:embeddeddolt_conformance_core_test", "//internal/storage/embeddeddolt:embeddeddolt_conformance_core_slow_test", "//internal/storage/embeddeddolt:embeddeddolt_conformance_audit_test"}},
{bazelProxiedJobName, "doltserver-proxied", []suite{
// bazel-proxied runs its own duration-balanced manifest block
// (scripts/ci/proxied_test_durations.json), not PR Risk's frozen
// 15-shard bd-init-cost-proxy block: it is not a drop-in
// retirement of test-proxied-cmd's shard count, just its tests.
// (scripts/ci/proxied_test_durations.json), not PR Risk's
// 15-shard bd-init-cost-proxy block, which was frozen and has
// since been deleted (that job no longer exists; see
// engdocs/TESTING.md): it was never a drop-in retirement of
// test-proxied-cmd's shard count, just its tests.
// bazelProxiedShardCount reads the real count from
// cmd/bd/BUILD.bazel, making this suite's `want` (below) a
// genuine cross-file pin against bazel.yml's own
@ -913,8 +918,9 @@ func TestBazelRetiredLanesCannotBeNarrowed(t *testing.T) {
"//cmd/bd:bd_managed_local_test": {
[]string{"$(rootpath :bd_test)", "-test.run=^TestManagedLocalProxied", "-test.timeout=15m"},
map[string]string{
"BEADS_TEST_BD_BINARY": "$(rlocationpath :bd_for_tests)", "BEADS_TEST_DOLT_SERVER": "local",
"BEADS_TEST_GOFMT": "$(rlocationpath @go_sdk//:bin/gofmt)", "BEADS_TEST_PREFLIGHT_GO": "$(rlocationpath :preflight_go_fixture)",
"BEADS_TEST_BD_BINARY": "$(rlocationpath :bd_for_tests)", "BEADS_TEST_COMPACT_GC_FIXTURE": "$(rlocationpath :compact_gc_fixture)",
"BEADS_TEST_DOLT_SERVER": "local",
"BEADS_TEST_GOFMT": "$(rlocationpath @go_sdk//:bin/gofmt)", "BEADS_TEST_PREFLIGHT_GO": "$(rlocationpath :preflight_go_fixture)",
"BEADS_TEST_PROXIED_LOCAL": "1", "BEADS_TEST_REQUIRE_DOLT_CONTAINER": "1", "BEADS_TEST_SKIP": "dolt",
},
},

View file

@ -19,19 +19,20 @@ import (
// S3 (F1, mirroring F2's TestProxiedShardManifestGeneratorNotStale below —
// see that test's doc comment for the full --check rationale, not repeated
// here): the Bazel-only 100-shard cmd block and 40-shard storage block are
// not frozen like their files' legacy 20- and 5-shard blocks.
// here): the Bazel-only 100-shard cmd block and 40-shard storage block.
// gen_embedded_{cmd,storage}_shard_manifest.py --check verifies only that
// the committed block names every discovered test exactly once, failing with
// the exact command to fix it when a name is missing, stale, or duplicated —
// run here so a drifted block fails `go test ./scripts/...`
// (//scripts:go_test_sources_test under Bazel) instead of only surfacing as
// a test silently never running in any shard. The legacy blocks are
// deliberately excluded: both files document that their 20- and 5-shard
// blocks are frozen (see .github/scripts/embedded-{cmd,storage}-
// test-shards.txt and engdocs/TESTING.md), so a --check against them is
// expected to report "missing" entries by design (see those generators'
// module docstrings) and is not what this test runs.
// a test silently never running in any shard. The cmd file's legacy,
// hand-assigned 20-shard block (frozen for PR Risk's and main.yml's retired
// fork/push jobs) has been deleted; those jobs no longer exist (see
// engdocs/TESTING.md). The storage file's legacy 5-shard block is still
// frozen and deliberately excluded here: its header documents that (see
// .github/scripts/embedded-storage-test-shards.txt and engdocs/TESTING.md),
// so a --check against it is expected to report "missing" entries by design
// (see that generator's module docstring) and is not what this test runs.
func TestCmdEmbeddedShardManifestGeneratorNotStale(t *testing.T) {
python := requireHostTool(t, "python3")
root := sourceRepoRoot(t)
@ -158,9 +159,13 @@ func TestShardScriptsListOnlyRealTests(t *testing.T) {
case ".github/scripts/proxied-test-shard.sh":
totalsSet[bazelProxiedShardCount(t)] = true
case ".github/scripts/embedded-test-shard.sh":
// F1: the Bazel-only bazel-embedded lane reads a 50-shard cmd
// block that no PR-Risk-matrix-only (20-shard) check exercises;
// without this, "BUILD.bazel's shard_count and bazel.yml's
// F1: the Bazel-only bazel-embedded lane reads a 100-shard cmd
// block (split over embeddedCmdTargets); the legacy, frozen
// 20-shard block PR Risk's/main.yml's retired fork/push jobs
// used to read has been deleted (those jobs no longer exist; see
// engdocs/TESTING.md), so there is no longer a second check that
// would catch this total drifting on its own. Without this,
// "BUILD.bazel's shard_count and bazel.yml's
// check_shard_coverage.py arg both drift to a new total with no
// manifest block" passes every policy test (see review S3) --
// the lane then silently goes 100% hash fallback and loses its
@ -249,26 +254,24 @@ func TestShardScriptsListOnlyRealTests(t *testing.T) {
}
}
// S3: the Bazel-only 34-shard block is not frozen like the legacy 15-shard
// block (TestShardScriptsListOnlyRealTests's B1 fix catches outright
// corruption, but not a committed block that has drifted from the currently
// discovered TestProxiedServer*/TestServerMode* test set, e.g. a test added,
// renamed, or removed without anyone running --write). gen_proxied_shard_
// manifest.py --check verifies only that the committed block names every
// discovered test exactly once -- not that its shard *assignments* match a
// fresh LPT pack -- and fails with the exact command to fix it when a name
// is missing, stale, or duplicated. It deliberately does NOT fail merely
// because proxied_test_durations.json's weights changed and the existing
// packing is now suboptimal: two PRs each adding one proxied test would
// otherwise force a full repack and conflict on unrelated shard lines (see
// --repack below for the explicit opt-in to that). Run --check here so a
// block with missing/stale/duplicate names fails go test ./scripts/...
// (//scripts:go_test_sources_test under Bazel) instead of only
// surfacing as a test silently never running in any shard. The legacy
// 15-shard block is deliberately excluded: its header documents that it is
// frozen and must not be regenerated (see
// .github/scripts/proxied-cmd-test-shards.txt and engdocs/TESTING.md), so a
// --check against it would always fail by design.
// S3: the manifest's Bazel-only 34-shard block, the file's only remaining
// block now that its legacy, frozen 15-shard block (read by PR Risk's and
// main.yml's retired fork/push jobs) has been deleted — those jobs no
// longer exist (see engdocs/TESTING.md). TestShardScriptsListOnlyRealTests's
// B1 fix catches outright corruption, but not a committed block that has
// drifted from the currently discovered TestProxiedServer*/TestServerMode*
// test set, e.g. a test added, renamed, or removed without anyone running
// --write. gen_proxied_shard_manifest.py --check verifies only that the
// committed block names every discovered test exactly once -- not that its
// shard *assignments* match a fresh LPT pack -- and fails with the exact
// command to fix it when a name is missing, stale, or duplicated. It
// deliberately does NOT fail merely because proxied_test_durations.json's
// weights changed and the existing packing is now suboptimal: two PRs each
// adding one proxied test would otherwise force a full repack and conflict
// on unrelated shard lines (see --repack below for the explicit opt-in to
// that). Run --check here so a block with missing/stale/duplicate names
// fails go test ./scripts/... (//scripts:go_test_sources_test under Bazel)
// instead of only surfacing as a test silently never running in any shard.
func TestProxiedShardManifestGeneratorNotStale(t *testing.T) {
python := requireHostTool(t, "python3")
root := sourceRepoRoot(t)
@ -279,3 +282,54 @@ func TestProxiedShardManifestGeneratorNotStale(t *testing.T) {
t.Errorf("gen_proxied_shard_manifest.py %s --weights=duration --check: %v\n%s", shards, err, out)
}
}
// bareWriteReproducesManifest runs `generator --write --manifest COPY` with
// no other flags (in particular no --weights) against a scratch copy of
// manifest, and asserts the result is byte-identical to the committed file.
//
// This pins the --weights default against the footgun a reviewer caught on
// this PR: the manifest's only remaining block is duration-packed (the
// legacy, frozen inits-packed blocks were deleted alongside the PR Risk
// fork/push jobs that read them — see engdocs/TESTING.md), so a --weights
// default of "inits" made a bare `--write` silently rewrite the live
// block's header (and, since costs then came from the wrong model, its
// incremental placement of any newly-discovered test) to the inits form
// without --check ever catching it. Defaulting --weights to "duration"
// makes a bare `--write` idempotent on an up-to-date, duration-packed file.
func bareWriteReproducesManifest(t *testing.T, root, generator, manifest string) {
t.Helper()
python := requireHostTool(t, "python3")
want, err := os.ReadFile(filepath.Join(root, manifest))
if err != nil {
t.Fatalf("read %s: %v", manifest, err)
}
scratch := filepath.Join(t.TempDir(), filepath.Base(manifest))
if err := os.WriteFile(scratch, want, 0o644); err != nil {
t.Fatalf("seed scratch copy of %s: %v", manifest, err)
}
cmd := exec.Command(python, generator, "--write", "--manifest", scratch)
cmd.Dir = root
if out, err := cmd.CombinedOutput(); err != nil {
t.Fatalf("%s --write --manifest %s (no --weights): %v\n%s", generator, scratch, err, out)
}
got, err := os.ReadFile(scratch)
if err != nil {
t.Fatalf("read %s after --write: %v", scratch, err)
}
if string(got) != string(want) {
t.Errorf("%s --write with no --weights changed %s; a bare --write should reproduce the "+
"committed, duration-packed file exactly (--weights defaulting to anything but "+
"\"duration\" clobbers its header and repacks with the wrong cost model):\ngot:\n%s\nwant:\n%s",
generator, manifest, got, want)
}
}
func TestProxiedShardManifestBareWriteIsIdempotent(t *testing.T) {
bareWriteReproducesManifest(t, sourceRepoRoot(t),
"scripts/ci/gen_proxied_shard_manifest.py", ".github/scripts/proxied-cmd-test-shards.txt")
}
func TestEmbeddedCmdShardManifestBareWriteIsIdempotent(t *testing.T) {
bareWriteReproducesManifest(t, sourceRepoRoot(t),
"scripts/ci/gen_embedded_cmd_shard_manifest.py", ".github/scripts/embedded-cmd-test-shards.txt")
}