You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
World rebuild.lz4 declares replace_on_cycle in build.ncl (bedrock-root
prebuilt), so it takes part in the toolchain rebuild graph and its hash change
cascades through the tree, the same shape as #1351 (zlib 1.3.2 -> 1.4).
Direct dependents in this tree: file, postgres, rizin, rsync, zstd, and through zstd most of the world.
Recipe changes beyond version + sha256 (both found by a local clean-room build):
findutils added to build_deps. 1.11.0's new make layer (build/make/multiconf.make)
discovers sources with find. Without it, make fails with find: command not found / no source for lz4.o ....
build.sh builds through make install only. In 1.11.0, lib: still requires liblz4.a, but the liblz4.a rule is only defined when BUILD_STATIC=yes. So the default
target fails under our BUILD_STATIC=no (No rule to make target 'liblz4.a', needed by 'lib'). install builds exactly the shared lib and CLI. -DNDEBUG (previously from lz4-release)
is now set through CPPFLAGS, because multiconf's phony lz4 target is re-resolved with install's flags and would otherwise yield an assert-enabled CLI.
Verification
Downloaded with gh release download v1.11.0 --repo lz4/lz4. The sha256 matches
both GitHub's published asset digest (releases/tags/v1.11.0, .assets[].digest)
and upstream's lz4-1.11.0.tar.gz.sha256 asset.
Upstream publishes no signature for this release, so there was nothing to verify with sq.
Mirrored with gcloud storage cp --no-clobber. The public object at https://storage.googleapis.com/minimal-staging-archives/lz4-1.11.0.tar.gz was
re-downloaded and re-hashed: it matches.
Build files diffed from tarball to tarball (lib/Makefile, programs/Makefile, lib/liblz4.pc.in):
The soname stays liblz4.so.1 (LZ4_VERSION_MAJOR 1).
BUILD_STATIC=no and PREFIX/DESTDIR install are still honored.
The installed program set (lz4, lz4c, lz4cat, unlz4) and output globs are unchanged.
The new Makefile layer (build/make/multiconf.make) keeps user CFLAGS/LDFLAGS, so the reproducibility flags still apply.
nickel format run on build.ncl.
Existing standalone smoke test (/usr/bin/lz4 --version) kept.
Local clean-room build (min session exec <s> -- min package build --rebuild --verbose lz4, fresh session per run):
run 1, version bump only: FTBFS, find: command not found -> fix 1;
run 2, with findutils: the CLI compiled and linked ("built with multithreading support"),
then the library failed on the liblz4.a rule -> fix 2;
fix 2 has NOT been built locally (session budget used up). Build-bot is the first build of the final recipe.
Reviewer heads-up (build risk)
Header API removal: functions deprecated since v1.7/v1.8 (LZ4_compress, LZ4_uncompress, LZ4_compress_limitedOutput, ...) and the unprefixed lz4frame enums
are no longer declared in headers. liblz4 still exports them, so already-built
binaries are ABI-compatible, but a dependent whose source still calls them will now
fail to build. The direct dependents to watch in the world rebuild are file, postgres, rizin, rsync and zstd.
LZ4F_compressOptions_t must now be zero-initialized (positional initializers break).
The dynamic library is now multithreaded by default: it links -pthread, detected by a
configure-time probe that uses mktemp. With glibc >= 2.34, pthread is in libc.
Update: zstd fix, and the final recipe builds locally
A local clean-room build of the earlier head failed in zstd 1.5.7, which is in lz4's own build closure: programs/fileio.c:1424: 'contentChecksum_t' undeclared. lz4 1.11.0 removed the unprefixed lz4frame names together with LZ4F_ENABLE_OBSOLETE_ENUMS, and zstd dev (49bf1d1, line 1690) still uses the same cast.
Improved build compatibility for LZ4 and Zstandard, helping ensure both packages can be built and installed successfully.
Note: the cycle-break prebuilt stays on 1.10.0
lz4's replace_on_cycle bedrock prebuilt is not touched here and is still 1.10.0 until it is re-rolled. zstd's fix (LZ4F_contentChecksum_t) works against either version. The world build on the head, rsync included, is green.
Reviewing files that changed from the base of the PR and between 14fce56 and 8209693.
You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.
Use the checkbox below for a quick retry:
Trigger review
No actionable comments were generated in the recent review.
Walkthrough
The LZ4 package now uses version 1.11.0 and updated build settings. The Zstandard build script rewrites and checks checksum type casts in programs/fileio.c.
The build script replaces contentChecksum_t casts in programs/fileio.c with LZ4F_contentChecksum_t. It exits with an error if a whole-word contentChecksum_t occurrence remains.
LZ4 is updated to 1.11.0 with a Zstandard compatibility rewrite. No merge-blocking issue was established.
Pre-merge checks: 5
Check name
Status
Explanation
Description Check
Check skipped - CodeRabbit’s high-level summary is enabled.
Title check
The title "Update stacked: lz4 1.10.0 -> 1.11.0" names the main change: the lz4 version bump from 1.10.0 to 1.11.0. The changeset updates the lz4 build.ncl and build.sh to 1.11.0 and its checksum. The…
Docstring Coverage
No eligible functions found in the changed files. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documentat…
Linked Issues check
Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check
Check skipped because no linked issues were found for this pull request.
Full details: Title check
Explanation
The title "Update stacked: lz4 1.10.0 -> 1.11.0" names the main change: the lz4 version bump from 1.10.0 to 1.11.0. The changeset updates the lz4 build.ncl and build.sh to 1.11.0 and its checksum. The title also omits the supporting changes, which include adding findutils as a build dependency, the build script changes, and the zstd build script fix for the contentChecksum_t cast. Those are expected to be left out of a short title. The word "stacked" describes the PR's stacking workflow, not the code change, but it does not make the title misleading.
Full details: Docstring Coverage
Explanation
No eligible functions found in the changed files. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documentation-only edits and deletions within surviving functions are included; unchanged diff context and fully deleted functions are excluded. Analyzed 0 functions across 2 files. (1 skipped: 1 unsupported.)
Finishing Touches
Generate docstrings
Commit to this branch
Create a new PR
Generate unit tests (beta)
Commit to this branch
Create a new PR
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Reviewing files that changed from the base of the PR and between 22201ed and 381909f.
Files selected for processing: 3
packages/lz4/build.ncl
packages/lz4/build.sh
packages/zstd/build.sh
Included review availability: This review used your included allowance. 0 included reviews remain after this review. Your included PR review attempts over the past 7 days set your current allowance at 1 review per hour.
… old name)
The local clean-room build of this PR failed in zstd 1.5.7:
fileio.c:1424: contentChecksum_t undeclared. lz4 1.11.0 removed
LZ4F_ENABLE_OBSOLETE_ENUMS, and zstd dev (49bf1d1) has the same cast.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The reason will be displayed to describe this comment to others. Learn more.
Review on Karl's behalf. This is a comment, not an approval. Nothing was built for it. The findings come from reading the declarations and fetching upstream sources, and each one says how it was checked. The CI rollup on this head is now SUCCESS.
What "stacked" means: lz4 is a bedrock-root package. It declares replace_on_cycle with a hex0-ladder prebuilt, so it sits in the toolchain cycle graph. Bumping it triggers a world rebuild through zstd, the same shape as #1351 (zlib).
CI: Every check is green except minimal build, which has no conclusion yet (the world build is still running). CodeRabbit skipped the PR ("no reviewable changes"), so there is no bot review.
Should-fix: none found. The world build is the real gate for the dependents.
Nit
PR body, "Direct dependents". It lists file, postgres, rizin and zstd, but rsync also imports lz4 (packages/rsync/build.ncl).
If rsync's source used an API that 1.11 removed from the headers, its failure in the world build would surprise reviewers.
I expect it to build: as far as I know, rsync uses LZ4_compress_default and LZ4_decompress_safe, which are still declared. I did not check this against rsync's source.
packages/lz4/build.ncl (replace_on_cycle). The bedrock prebuilt is unchanged and is presumably still 1.10.0, because minimermetic owns it.
That is fine for zstd: LZ4F_contentChecksum_t exists in both versions, so the zstd fix works against either lz4.
The body should say that the cycle-break artifact stays on the old version until minimermetic re-rolls it.
Checked, clean
sha256: I downloaded storage.googleapis.com/minimal-staging-archives/lz4-1.11.0.tar.gz. Its sha256 is 247e5fa6…5e26, which matches build.ncl. build.sh's lz4-1.11.0 dir name matches the tarball.
Patches: lz4 has none. The zstd sed targets (contentChecksum_t) and then grep -w fails the build on any leftover unprefixed use. -w does not match LZ4F_contentChecksum_t, because _ is a word character, so the check cannot false-fire after the rewrite.
Makefiles (from the tarball):
lib/Makefile:154 lib: liblz4.a liblz4, with the liblz4.a rule only under BUILD_STATIC=yes (:174). This confirms the default target fails.
lib's install (:293) builds $(LIBLZ4) when it is absent.
Top-level install depends on lz4.
User CFLAGS survive through USERCFLAGS := -O3 $(CFLAGS), and CPPFLAGS/LDFLAGS survive through multiconf's MCM_* split. So the env CPPFLAGS=-DNDEBUG reaches both the CLI and the lib.
Without lz4-release, DEBUGFLAGS returns to its default. That default is only -W warnings, which do not change the output.
mktemp (pthread/JCC probes) comes from coreutils, and find from the added findutils.
Soname: it stays liblz4.so.1. The output glob usr/lib/liblz4.so* covers liblz4.so.1.11.0. No package in the tree pins liblz4.so.1.10.
Verdict: Recipe looks correct. Hold the merge until minimal build (the world rebuild) is green, since file, postgres, rizin and rsync are proved only there.
Update:minimal build has passed on this head since this review was written, so the hold is cleared. The PR is ready.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Update stacked: lz4 1.10.0 -> 1.11.0
Note
World rebuild.
lz4declaresreplace_on_cyclein build.ncl (bedrock-rootprebuilt), so it takes part in the toolchain rebuild graph and its hash change
cascades through the tree, the same shape as #1351 (zlib 1.3.2 -> 1.4).
Direct dependents in this tree:
file,postgres,rizin,rsync,zstd, and throughzstdmost of the world.Summary
1.10.01.11.0github:lz4/lz4release assetlz4-1.11.0.tar.gz537512904744b35e...->247e5fa64da5f7b75ee3dd3bed484076396c9ebe8254d9ccf693eb62800c5e26gs://minimal-staging-archives/lz4-1.10.0.tar.gz->gs://minimal-staging-archives/lz4-1.11.0.tar.gzfindutilsadded tobuild_deps. 1.11.0's new make layer (build/make/multiconf.make)discovers sources with
find. Without it, make fails withfind: command not found/no source for lz4.o ....make installonly. In 1.11.0,lib:still requiresliblz4.a, but theliblz4.arule is only defined whenBUILD_STATIC=yes. So the defaulttarget fails under our
BUILD_STATIC=no(No rule to make target 'liblz4.a', needed by 'lib').installbuilds exactly the shared lib and CLI.-DNDEBUG(previously fromlz4-release)is now set through
CPPFLAGS, because multiconf's phonylz4target is re-resolved withinstall's flags and would otherwise yield an assert-enabled CLI.Verification
gh release download v1.11.0 --repo lz4/lz4. The sha256 matchesboth GitHub's published asset digest (
releases/tags/v1.11.0,.assets[].digest)and upstream's
lz4-1.11.0.tar.gz.sha256asset.sq.gcloud storage cp --no-clobber. The public object athttps://storage.googleapis.com/minimal-staging-archives/lz4-1.11.0.tar.gzwasre-downloaded and re-hashed: it matches.
lib/Makefile,programs/Makefile,lib/liblz4.pc.in):liblz4.so.1(LZ4_VERSION_MAJOR1).BUILD_STATIC=noandPREFIX/DESTDIRinstall are still honored.lz4,lz4c,lz4cat,unlz4) and output globs are unchanged.build/make/multiconf.make) keeps userCFLAGS/LDFLAGS, so the reproducibility flags still apply.nickel formatrun on build.ncl./usr/bin/lz4 --version) kept.min session exec <s> -- min package build --rebuild --verbose lz4, fresh session per run):find: command not found-> fix 1;then the library failed on the
liblz4.arule -> fix 2;Reviewer heads-up (build risk)
LZ4_compress,LZ4_uncompress,LZ4_compress_limitedOutput, ...) and the unprefixed lz4frame enumsare no longer declared in headers. liblz4 still exports them, so already-built
binaries are ABI-compatible, but a dependent whose source still calls them will now
fail to build. The direct dependents to watch in the world rebuild are
file,postgres,rizin,rsyncandzstd.LZ4F_compressOptions_tmust now be zero-initialized (positional initializers break).-pthread, detected by aconfigure-time probe that uses
mktemp. With glibc >= 2.34, pthread is in libc.Update: zstd fix, and the final recipe builds locally
A local clean-room build of the earlier head failed in zstd 1.5.7, which is in lz4's own build closure:
programs/fileio.c:1424: 'contentChecksum_t' undeclared. lz4 1.11.0 removed the unprefixed lz4frame names together withLZ4F_ENABLE_OBSOLETE_ENUMS, and zstd dev (49bf1d1, line 1690) still uses the same cast.381909f fixes it in zstd's build.sh:
LZ4F_contentChecksum_t, a name that has existed since lz4 1.7.Local results at 381909f (fresh
pkgbuildsession,--rebuild):lz4buildlz4 installation completedzstdbuildrelease build completedmin checkfor lz4 and zstdlz4's other direct dependents (file, postgres, rizin) are only covered by the world build.
🤖 Generated with Claude Code
Summary by CodeRabbit
Note: the cycle-break prebuilt stays on 1.10.0
lz4's
replace_on_cyclebedrock prebuilt is not touched here and is still 1.10.0 until it is re-rolled. zstd's fix (LZ4F_contentChecksum_t) works against either version. The world build on the head, rsync included, is green.