* ci(bazel): split //scripts:scripts_test by what it reads; partition //:repo_files //scripts:scripts_test declared //:repo_files (the whole checkout), so every PR re-ran it, and it took ~181s on rbe-west. Most of that (60-79s) was TestBazelGatedLanesNeverRetryFlakyTests line-regexping the --run_under wrapper's pinned dolt binary (tools/bazel/hermetic_bin/dolt, in every test's runfiles); it now skips binary files, as git grep -I does. tools/bazel/go_srcs.py's per-package repo_files block now partitions each package's files into repo_go_srcs, repo_go_test_srcs, repo_doc_files and repo_other_files, aggregated at the root; //:repo_files is their union (same 4679 files). The //scripts tests are split into go_test targets by what they read (traced per test with strace over a runfiles-shaped tree): scripts_test CI/Bazel policy: //:repo_other_files (+1 Go file) shell_scripts_test script fixtures: //scripts files, a few root files go_test_sources_test _test.go scans: //:repo_go_test_srcs + shard files go_sources_test check-testing-short.sh: every .go file tracked_files_test every tracked file: //:repo_files Shared helpers move to helpers-only files; TestScriptsTestTargetsPartition- TheTests fails on a test file in no target or in two. //test/docsync and the //scripts/repochecks slices declare partitions or single files instead of //:repo_files where they can. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> * ci(bazel): skip only the non-checkout hermetic_bin dir in the flaky-retry walk Skipping every binary file also stopped scanning tracked binaries (images, .binpb fixtures). Skip just tools/bazel/hermetic_bin, the --run_under wrapper's runfiles directory, which is not in the checkout, so the walk's coverage of tracked files is unchanged. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| BUILD.bazel | ||
| main.go | ||
| main_test.go | ||
| README.md | ||
Historical release catalog
../release-catalog.json pins the exact historical release universe (v0.9.1
through v1.3.1; the v1.2 upgrade census plus the releases since): 127 stable
module versions authenticated through the Go
proxy and SumDB, each with module / go.mod / source-zip / proxy-origin
provenance, explicit proxy prerelease and repository-only tag exclusions, and
the current repository tag drift.
The catalog command in this directory both regenerates and validates that
artifact (usage: catalog generate|validate <manifest.json>):
go run ./scripts/migration-test/catalog validate scripts/migration-test/release-catalog.jsonruns the strict offline validator: schema checks, semantic validation, canonical byte-form, and the pinned SHA-256 identity digest. Even a well-formed substitution is rejected until that digest is deliberately updated.go run ./scripts/migration-test/catalog generate scripts/migration-test/release-catalog.jsondeterministically rebuilds the catalog from the live Go proxy and GitHub release APIs (network plusgo/ghrequired). This is a manual maintenance path, not a CI step.
Adding a release
Raise maximumVersion in main.go, run generate, and update the count
constants and expectedCatalogSHA256 the validator then reports. Every
release with a linux/amd64 asset digest automatically becomes a pinned Bazel
repository (@bd_releases//:<version>, tools/bazel/bd_releases.bzl).
Status: self-validated reference anchor
Today the catalog is a reviewed-identity anchor. Its only automated enforcement
is TestCheckedCatalogIsCanonicalAndComplete in main_test.go, which runs
under go test ./... and holds the checked JSON to the pinned digest and its
invariants. Bazel consumes it: tools/bazel/bd_releases.bzl declares one
sha256-verified repository per release asset it records, which the
regression, migration (tests/migration) and cross-version targets run
instead of downloading release binaries.
The intended consumer is the upgrade-coverage work stacked on top of this
change: runtime schema observations and family/path qualification will drive
upgrade tests against this finite, reproducible version universe. Until that
lands, regenerate deliberately with generate when upstream history changes —
drift from live upstream is not detected automatically.