Skip to content

Re-apply fork work as one squash on top of sonos/tract main - #10

Merged
AdaWorldAPI merged 2 commits into
mainfrom
squash/our-work-on-upstream-2026-10-29
Oct 29, 2026
Merged

AdaWorldAPI merged 2 commits into
mainfrom
squash/our-work-on-upstream-2026-10-29

Conversation

@AdaWorldAPI

Copy link
Copy Markdown
Owner

This branch carries the fork's own work (105 commits from PRs #1–#9) as one commit on top of the current sonos/tract main (7e8be2f), plus a small follow-up so it builds there.

Intended order:

  1. Use GitHub's "Sync fork → Discard commits", so main becomes exactly sonos/tract main.
  2. This PR is then exactly 2 commits ahead and merges with no conflicts.

The old history stays on backup/main-pre-upstream-sync-2026-10-29 and backup/main-pre-upstream-sync-2026-10-29-b (both at 6ce7c91).

Commits:

  • e512dfe: a squash of all fork changes since the last shared commit (7ef965c). It covers the same 48 files as the fork's diff. Conflicts were resolved like this:
    • The Cargo.lock files stay deleted (fork convention).
    • Both deny.toml files keep upstream's new skip entries plus our [sources] allow-git list.
  • aeafcd3:
    • try_as_plain_mut → try_as_plain_ram_mut, because upstream renamed the method.
    • The toolchain moves 1.98.1 → 1.99.0, because the AdaWorldAPI ndarray fork now requires rustc 1.99 and 1.98.1 no longer resolves (this already affects the current fork main).

Verified locally:

  • cargo fmt --all --check passes.
  • cargo check --all-targets on the default members passes.
  • cargo test --release -p tract-linalg --lib: 4804 passed, 0 failed.

Known red: the fork's lint gate (cargo clippy --all-targets -- -D warnings) fails on Rust 1.99 because of upstream code, not ours:

  • About 100+ redundant_field_names findings in tract-core and other crates, coming from derive-new expansions.
  • A few chunks_exact / useless_conversion / too_many_arguments findings.

Upstream's CI does not deny clippy warnings, so these are new to us. Fixing them is a separate decision.


Generated by Claude Code

…ract main

Our fork-only changes since the last shared commit 7ef965c, replayed as one
commit on upstream 7e8be2f: ndarray SIMD/GEMM/AMX-BF16 pilots in linalg, the
CI lint gate, and the .claude knowledge files. Lock files stay deleted, and
both deny.toml files keep upstream's new skip entries plus our allowed git
sources. Full history is on backup/main-pre-upstream-sync-2026-10-29.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XAqKMMjKnig6MXTd41Lf2v
Upstream renamed Tensor::try_as_plain_mut to try_as_plain_ram_mut, so the AMX
native packer uses the new name. The toolchain moves to 1.99.0 because the
AdaWorldAPI ndarray fork now requires it and 1.98.1 no longer resolves.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XAqKMMjKnig6MXTd41Lf2v
@AdaWorldAPI
AdaWorldAPI marked this pull request as ready for review October 29, 2026 07:21
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Oct 29, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-10-29T07:22:26.584860Z aeafcd3 Draft marked ready
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@AdaWorldAPI
AdaWorldAPI merged commit 1136e86 into main Oct 29, 2026
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