Skip to content

gcc: arm64 prebuilt onto gen-17's gcc-15.2.0-glibc-seed (arm64 world rebuild) - #1535

Draft
bryan-minimal wants to merge 1 commit into
mainfrom
bryan/arm64-gcc-seed-gen17
Draft

bryan-minimal wants to merge 1 commit into
mainfrom
bryan/arm64-gcc-seed-gen17

Conversation

@bryan-minimal

Copy link
Copy Markdown
Member

What

Moves the arm64 gcc (prebuilt) cycle-breaker from gen-13's gcc-15.2.0-glibc (4777e5fe…) to gen-17's gcc-15.2.0-glibc-seed (ad09a27c…). One file, one pin. The amd64 pin is untouched.

Why

The gen-13 arm64 object carries an RPATH into the bootstrap ladder's store paths on 33 of its ELF files. The seed (gominimal/bootstrap#16) is the same installed tree with those RPATHs removed and nothing else changed, so the prebuilt resolves its libraries from the sandbox like any other package. This is the arm64 counterpart of what #1499 does for amd64 binutils.

Evidence

  • The seed built on two arm64 lab machines (axion-1, axion-3) with identical spec hash, tree hash and object sha256.
  • The recipe's own gates passed: only the 33 files differ from the rung tree, no RPATH/RUNPATH remains, every binary resolves through the plain loader, and gcc and g++ compile and run C and C++.
  • gen-17 is published; its manifest is gominimal/bootstrap#17.

Impact

  • arm64: gcc's spec hash moves, so this is an arm64 world rebuild.
  • amd64: no change.

Ordering

🤖 Generated with Claude Code

The seed is the same gcc-15.2.0-glibc tree with the RPATH removed from its
33 ELF files, so the cycle-breaker no longer names ladder store paths. The
amd64 pin does not move.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
@coderabbitai

coderabbitai Bot commented Nov 16, 2026

Copy link
Copy Markdown
Contributor

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

  • Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

@bryan-minimal

Copy link
Copy Markdown
Member Author

Plan: fold this into the gcc 16 rotation rather than merging it on its own. The rotation is pkgs#1340 now, and a PR queued alone later. Each of #1499 and #1535 is a world rebuild. The registry spec's ruling (gominimal/arch#183, decisions §4) is to make gcc, glibc and bootstrap changes in one rotation, so the catalog is rebuilt once. That rotation also takes the OQ-42 breaker output fixes (bash-bootstrap, bzip2, binutils arm64, attr).

#1535 specifically: once gcc-runtime lands, gcc_runtime_guard.py requires gcc-runtime's breaker to pin the same object as gcc's. The arm64 repin to gen-17 has to move both breakers together.

This waits for the current canary run (230721a) to finish.

🤖 Generated with Claude Code

bryan-minimal added a commit that referenced this pull request Nov 17, 2026
Both rebuild the world, so they ride the same rotation as gcc 16
instead of each costing one of their own.

- #1499: amd64 binutils breaker and ghc-9.6.7 onto gen-16's
  plain-loader objects (binutils-2.46-glibc-seed, ghc-9.6.7).
- #1535: arm64 gcc breaker onto gen-17's gcc-15.2.0-glibc-seed.
  gcc-runtime's breaker moves with it, since gcc_runtime_guard.py
  requires the two pins to match.

All three objects exist in gs://minimal-bootstrap. #1499's build was
green. #1535's failed only on tmd's zig fetch from
deps.files.ghostty.org (connection reset), which is unrelated.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
bryan-minimal added a commit that referenced this pull request Nov 17, 2026
…308f31)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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.

1 participant