[colbymchenry]

Summary

Regression items 3, 4 and 6 of REG (Ruby): a call now follows a class's real ancestors, in Ruby's order. company.has_attribute? reaches the instance method, mapper.get and cookies.signed come back, and sinatra's bad_cookies.empty? no longer links Cookies::Jar.

Root cause

• e043efb (fix(resolution): bind Class.method() to the class itself, not a class whose name contains it #1935) walked a class's supertypes by bare names, and read no mixin. class Human < ActiveRecord::Base asked every class called Base, so 442 calls reached Journey's Node#name-style methods (human.name, column.type), and Topic.create reached Digestor::Node.create. Mapper lost include HttpHelpers, which 1.6.2 reached through a substring.
• A value of a known class walked extends edges by simple name, so company.has_attribute? took a ClassMethods module's.
• fix(ruby): calls to !/? methods and calls on what .new made link their method #3049's x.y? kept the lone-method guess, where a word of any module around the owner counted: bad_cookies shares Cookies with Sinatra::Cookies::Jar.

Fix

• Ancestors as Ruby reads them: superclass and mixins by constant lookup, a class's own body only, the module mixed in last asked first, and for an instance no module it extends.
• A receiver named after a class takes the class constant lookup finds first, then a namesake.
• A call on a class reaches a superclass's def self., what it extends and its concerns' ClassMethods, never a mixin's instance method or a mixin's own def self..
• A bare name that is a method of the class around the call, taking no arguments and handing back what one class makes (@cookie_jar ||= Cookies::CookieJar.build(…), the factory read too), types the call on it, unless the calling method binds that name.
• RUBY_STD_METHODS: a ? or ! method of Ruby's core classes needs a receiver named after the owner itself. record.new_record? and the rest of fix(ruby): calls to !/? methods and calls on what .new made link their method #3049's gain stay.

Validation

Red to green: 7 of 11 tests in ruby-receiver-class-ancestry.test.ts fail on main; the rest guard the rules beside them.

A/B against 4499d00, all CLI steps rc 0:

Repo	Edges	Moved	Added	Lost
rails-src 793d744c	171,738 to 179,247	2,159	8,179	670
sinatra cb22afd7	2,678 to 2,676	0	0	2
mastodon b0555670, Ruby files	23,466 to 23,499	1	34	1
jekyll 97c3bdc8	same edges, 9 guesses now typed
controls gin, flask	byte-identical

Classes, 86 sampled entries read against source, 0 false added:

• Added: 4,315 class calls (Topic.find, Topic.create!, HelloJob.set), 2,034 inherited instance calls (topic.id, topic.reload), 1,715 bare helper calls in view tests (concat, content_tag), now in scope.
• Moved: to the first method in Ruby's order (save to Suppressor#save, which calls super), off FakeRecord::Base, off Digestor::Node.create.
• Lost: 442 Node-style false edges, the false ClassMethods and Timestamp#type picks, Core::set on a Concurrent::Event. About 30 were true by luck, like 20 xml.first, whose String#first came through ActiveModel::Type::String read as core String.

REG's PUB entries: lost/20, lost/24, lost/25 and sinatra/added/3 hold. company.has_attribute? 14 of 14 on the instance method. rails-src/moved/24 can't be fixed without harm: table in add_to(table) is a TableDefinition only because its one caller passes self. 1.6.2 got it from a substring of the qualified name, the rule that also sent 429 error.message to WrongEncodingError and 144 response.body to Buffer. Reaching it needs call-site flow, which nothing here has.

Index CPU on rails-src, median of 5 interleaved: 74.9 to 73.2 s. No extraction change, so no kernel parity.

Suite on the top of this stack, 6e22039, on main 3bd9949: 857 files passed, 6 skipped; 10,662 tests passed, 0 failed, 87 skipped (kernel built, 4 workers). A rebuild after the rebase gave the same graphs as the A/B build on sinatra, serilog, eShopOnWeb, drupal-core and bookstack.

Stacked on the PHP PR.

🤖 Generated with Claude Code