Skip to content

Bump version from 0.13.0 to 0.13.1 - #252

Merged
helly25 merged 2 commits into
mainfrom
chore/bump_version_to_0.13.1
Jul 12, 2026
Merged

Bump version from 0.13.0 to 0.13.1#252
helly25 merged 2 commits into
mainfrom
chore/bump_version_to_0.13.1

Conversation

@helly25

@helly25 helly25 commented Jul 11, 2026

Copy link
Copy Markdown
Owner

Created by tools/trigger_release.sh.

@helly25
helly25 enabled auto-merge (squash) July 11, 2026 22:58
@helly25
helly25 requested a review from Fab-Cat July 11, 2026 22:59
The mangle constant is version-derived (internal/hash_mangle_seed.bzl
folds native.module_version()), so bumping MODULE.bazel to 0.13.1
rotated it. Regenerate the checked-in non-Bazel fallback so
//mbo/hash:hash_mangle_seed_default_test passes.
@helly25
helly25 merged commit 88342b4 into main Jul 12, 2026
39 checks passed
@helly25
helly25 deleted the chore/bump_version_to_0.13.1 branch July 12, 2026 15:32
helly25 added a commit that referenced this pull request Jul 12, 2026
… leaner archive (#253)

* mbo/hash: generate mangle-seed header per build; dumbo/fnv1a in Starlark

The checked-in mangle-seed fallback carried a version-derived constant, so every
version bump made it stale and broke hash_mangle_seed_default_test (the CI
failure on release PRs, e.g. #252). Generate the header per build and stop
committing it: remove internal/hash_mangle_seed.h.in and the diff test;
hash_mangle.h includes the generated header directly and #errors when it is
missing (clangd falls back to a stable constant under -DIS_CLANGD). A version
bump no longer needs a committed regeneration, and the --//mbo/hash:mangle_seed*
flags are unaffected.

Also switch the version/seed fold from FNV-1a to the in-house dumbo hash
(SMHasher3-proven) and add Starlark ports of dumbo and fnv1a in
//mbo/hash:hash.bzl (public hash struct), kept byte-for-byte identical to the
C++ prime implementation and verified against it by hash_bzl_vs_cpp_dumbo_test
and hash_bzl_vs_cpp_fnv1a_test via the new //mbo/hash:hash_tool CLI. README and
CHANGELOG updated.

Verified: bumping 0.13.1 -> 0.13.3 rotates the generated constant and all hash
tests pass; reverting restores it and all tests pass.

Also condensed the 0.13.0 CHANGELOG entries to the terse, verb-first house style
(presentation only; released content unchanged).

* release: drop dev-only files from the release archive

Add .trunk, .gitattributes, .gitignore (and mbo/hash/measurements) to the
release_prep.sh EXCLUDES so the released tarball carries only what a consumer
needs to build the library. Kept bazelmod (useful to copy for other libs),
.clang-format (mope formats generated output with it), the other lint/format
configs, and compile_commands-update.sh (a courtesy for consumers).

* ci: remove the informational hash benchmark job

It never gated (continue-on-error), ran only on main when mbo/hash changed, did
no baseline comparison, and produced noisy shared-runner artifacts nothing
consumed. Reliable, comparable numbers now come from the out-of-band
mbo/hash/measurements bundles. The hash_benchmark binary stays for those and for
manual runs; only the CI job and its done-gate wiring are removed.

* ci: test one Bazel version per supported major (drop 9.0.2)

Policy is the last three majors (7, 8, 9). 9.1.1 already covers major 9, so the
extra 9.0.x rung (9.0.2) was redundant; dropping it removes 4 matrix jobs. Kept
7.2.1 (the earliest working 7.x floor) and 8.7.0 (latest 8.x).

* ci: shrink the test matrix and make test-clang gating

- test-clang now gates (continue-on-error: false) via the done job; the
  hermetic-clang coverage was informational-only before.
- Reduced the clang LLVM ladder to oldest+newest (dropped 21.1.8) and dropped
  the fastbuild config.
- Consolidated asan onto the newest clang on both OSes (dropped the redundant
  clang-20 asan); gcc-14 asan stays and runs early in test-gcc.
- test-bcr runs the 7.x/8.x Bazel rungs on ubuntu only (platform-agnostic
  build-system compat); macOS keeps the default 9.1.1 for platform coverage.
- test-gcc drops gcc-13 (still built at opt later by test-bcr's ubuntu+gcc).

Net: 31 -> 21 test jobs, all gating.

* ci: fix stale LLVM default-pin reference (it is 22.1.8)

The clang-ladder comment claimed 20.1.8 was the default pin; the pin is 22.1.8
(bazelmod/llvm.MODULE.bazel). Also align test-bcr's clang combos to that pinned
default instead of 20.1.8.

* bazelmod: correct LLVM distribution comments

22.1.8 is the default pin (not just a forward-looking rung); 21.1.8 matches
macOS 26's native Apple clang 21 (a pairing that tracks Xcode). Also replaces the
em-dashes with hyphens.

* ci: exercise clang cpp23 on the newest LLVM only

C++23 is used on recent compilers, so testing the cpp23 mode on the oldest
clang (20.1.8) is low value; the oldest rung now just proves the minimum
supported LLVM still builds (opt). Matches the asan 'newest-only' rule.
test-clang 10 -> 8.
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.

2 participants