[squid-protocol]

Summary

Level 2 call graph, Rust (#3785): rust-analyzer SCIP is the reference. Rust impl methods carry no parent_class_name (class_start covers only struct/enum/union/trait), so self.m() fell through the resolver's methods-only bucket and resolved to none. For rust, a self receiver now searches bucket.all, so it reaches impl methods through the normal file/import/unique ladder.

serde-rs/json vs rust-analyzer	before	after
reference edges linked	432	651
recall	39.9%	60.1%
confident precision	92.1%	93.8%

No gated metric dropped. The widening is gated by _IMPL_SELF_LANGS = {"rust"}; a parametrized test pins that python, java and javascript are unchanged.

Rejected: widening every receiver for rust took external from 40 to 138 and precision to about 71%, so only self is widened.

Class / Sibling gaps

Class: a receiver whose owner the extractor does not record (impl blocks), so the method bucket excludes the target. Fixed here for self receivers in rust. Remaining siblings: (1) typed receivers (let x: Foo then x.m()) and Foo::m() path calls to impl methods need an impl-owner map in by_owner; (2) trait-method dispatch through generics; (3) CI gate and toolchain hook for the rust reference. Burned-estate programs needing these: none (call-graph work, no mainframe estate involved).

Golden

Re-blessed with golden-lock.sh crucible_check.py --update --yes. Only graph-derived files move (fan_in, fan_out, pagerank, llm_brief): the new rust edges change fan counts and global PageRank normalisation, which also shifts rounding in other languages' PageRank at the 1e-5 level. No extraction field moves.

Gates

pr_gates.py --fast 7/7, --ratchets 9/9, call_resolver tests (95) and the pre-PR guard tests (301 total) pass, rust resolution baseline regenerated.

Facts checked: the box had no rust toolchain; installed rustup 1.90.0 and rust-analyzer into scratch to reproduce.

Checklist: partial [c:b43cf8a0] -- CI gate and toolchain hook, typed receivers/owners left
Checklist: partial [c:a15ba693] -- known-pattern row added for rust self.m(); other languages' rows not touched

Lessons for agents

• The box has no rust toolchain: rustup into a scratch RUSTUP_HOME/CARGO_HOME, add rust-analyzer, then import_graph_accuracy.py --fetch-only rust before triage.
• Widening all receivers for rust regresses precision badly; widen self only.
• call_graph_resolution.py takes --regenerate, not --write-baseline.

🤖 Generated with Claude Code

[squid-protocol]

No description provided.

[squid-protocol]

PR #6448 @ 350a8a9: 1 failed check(s)

== squid-protocol.gitgalaxy (action_required) -- dirty
https://github.com/squid-protocol/gitgalaxy/runs/120859612883
repro: git merge origin/main (then regenerate generated files with their tools; see the PR worker rules)
log tail:
This PR conflicts with main, so no checks ran. Merge origin/main into it and regenerate generated files with their tools (never hand-merge them), then push.

[squid-protocol]

CI shepherd: the head of this PR (c24be65) is not the one a maintainer labelled shepherd:merge on (new commits, or never approved), so the label was removed and nothing merges unreviewed code. Re-add the label when the new head is ready (shepherd:reapprove marks this PR until then).

[github-actions]

🐦‍⬛ Muninn Security Scan

✅ No security issues found.

🐦‍⬛ Powered by Muninn · Skald Lab

[squid-protocol]

No description provided.

[squid-protocol]

CI shepherd: the head of this PR (2031c35) is not the one a maintainer labelled shepherd:merge on (new commits, or never approved), so the label was removed and nothing merges unreviewed code. Re-add the label when the new head is ready (shepherd:reapprove marks this PR until then).

[squid-protocol]

No description provided.

[squid-protocol]

No description provided.

[squid-protocol]

No description provided.

[squid-protocol]

No description provided.

[squid-protocol]

No description provided.

[squid-protocol]

No description provided.

[squid-protocol]

No description provided.

[squid-protocol]

No description provided.