Repository navigation
Drop the hex0 -> stage0-* -> glibc-bedrock-2.42 chain (23 packages) - #1465
Conversation
|
Warning Review limit reached
This review includes 161 billable files and costs up to $40.25.
Reviews can continue after your included limit without a manual trigger. An admin must approve usage-based billing. Or wait 21 minutes for your next included review. View limit details
|
89a131f to
5b547c1
Compare
…sroot package gen-15 of gominimal/bootstrap builds the amd64 openjdk-21/25/26 and ghc-9.6.7 against the system glibc with the gcc seed: PT_INTERP /usr/lib/ld-linux-x86-64.so.2, no RUNPATH beyond their own $ORIGIN entries, and ghc-cc a plain gcc wrapper that adds that loader. The four pins move to gen-15 on both arches (arm64's objects are the gen-13 bytes under their gen-15 names, ghc-9.6.7's rebuilt from the shared script and identical apart from BUILDINFO) and their amd64 runtime dep becomes glibc. ghc drops the sysroot from its amd64 build deps; its check that nothing shipped names the sysroot stays. Nothing imports glibc-bedrock-2.42-sysroot any more, so the package goes. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> (cherry picked from commit 1151b19)
The arm64 JDK objects are the gen-13 bytes re-published under gen-15 (bootstrap generations/gen-15/manifest.json lists each sha256 at its gen-15 url), so the url moves and the sha256 correctly stays. Comments only; no spec hash changes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Nothing in pkgs reads it: the sysroot was its last consumer, and the bootstrap roots pin gominimal/bootstrap generation objects instead. The chain lives on there, built and reproduced per generation. The amd64 switch hex0/amd64-ladder.ncl goes with it. The CI-scope test used hex0's ladder helper as its shared-helper fixture; llvm/versioned.ncl is the one with the most importers now. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
5b547c1 to
608076b
Compare
norrietaylor
left a comment
There was a problem hiding this comment.
Scope matches the title and the body. The diff also carries the two stacked commits (ghc-9.6.7 and openjdk-21/25/26 move to gen-15, glibc-bedrock-2.42-sysroot goes). The body says this. Clobber none. git log on test_ci_scope.py, ghc/build.sh and openjdk-21/build.ncl shows only this stack after #1242 and #1372. Correctness no file outside the chain names hex0, amd64-ladder, a stage0-* package or glibc-bedrock-2.42 by import. The remaining hits are comments ("hex0-rooted toolchain") and gen-13 bootstrap object URLs in glibc and linux_headers. Those URLs are bootstrap objects, not packages, so the deletion does not break them. The amd64 runtime deps of ghc-9.6.7 and openjdk-21/25/26 now use glibc. No build script in those packages or in ghc/jdk-* references a sysroot path, so nothing else depended on the dropped dep. The ghc/build.sh guard against /usr/lib/glibc-bedrock-2.42 stays, which is a reasonable regression check. One finding: the new CI-scope test comment states a fact that is false for the new fixture. Not checked: the gen-15 hashes against the bootstrap manifest, and the 15/15 test run the body reports. Both need network or execution.
| self.assertGreaterEqual(len(importers), 3) | ||
| # The edge itself, not just reachability: the importers also reach | ||
| # llvm through build.ncl, which would hide a regex that only matches | ||
| # build.ncl imports. |
There was a problem hiding this comment.
low · The comment says the importers also reach llvm through build.ncl. That is false. llvm-21, llvm-22 and llvm-23 reach llvm only through versioned.ncl, and none of them imports ../llvm/build.ncl. The assertion is now stronger than the comment says, because the helper edge is the only path. Rewrite the comment so that it says this. As written, it will mislead the next person who edits the test or the regex. The issue is local to this test.
There was a problem hiding this comment.
Fixed in a54b98f: the comment now says llvm-21/22/23 reach llvm only through versioned.ncl (none imports ../llvm/build.ncl), so a regex matching only build.ncl imports would fail this assertion. test_ci_scope: 15/15 pass.
….ncl The comment said they also reach llvm through build.ncl; none imports ../llvm/build.ncl, so the helper edge is the only path (review on #1465). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stacked on #1462 and #1442 (both merged in); merge those first, then this lands as the single deletion commit.
Nothing in pkgs reads the chain any more. Its last consumer was
glibc-bedrock-2.42-sysroot(gone in #1462); the three orphans above it go in #1442. Every bootstrap root pins a gominimal/bootstrap generation object by sha-keyed URL, and the chain lives on there, built and reproduced per generation.Removed
hex0(withamd64-ladder.ncl, the amd64 switch that has beenenabled = falsesince 2026-10-07)stage0-*rungs: hex1, kaem, mes, mescc, mescc-full, tcc-0.9.26, tcc-0.9.27 (+ musl-s1/s3/s4), musl-1.1.24 (+ cc2), musl-1.2.5, binutils-2.30/2.41, gcc-4.0.4/4.7.4/10.4.0/15.2.0, gmp-mpfr-mpc, linux-headers-6.12.43glibc-bedrock-2.42No package outside the chain imports any of them (checked every
../<pkg>/reference underpackages/).CI scope test
test_ci_scope.pyusedhex0/amd64-ladder.nclas its shared-helper fixture. It now usesllvm/versioned.ncl, the helper with the most importers left; the comments inci_scope.pyandblast_radius_report.pyname the same example. 15/15 tests pass.🤖 Generated with Claude Code