[devin-ai-integration]

Summary

Fixes three generator bugs found in a pre-release sweep. Each one has a test that fails on main. The PR also fixes or documents the suspicions from that sweep.

1. PostgreSQL prepared-statement names could collide (High). Names were snake_cased and joined with _, so different class paths gave the same name. Options::PgV2::FirstSql and Options::Pg::V2::FirstSql both got options_pg_v2_first_sql_…, and names that PostgreSQL truncated at 63 bytes could also match. Two classes sharing a pool then ran each other's SQL or raised PG::DuplicatePstatement.

• Fix: Code.statement_name names each statement Module::Class#method. That name is unambiguous because :: and # never appear inside a constant or method name. Names over 63 bytes become a prefix, ~, and a 64-bit FNV-1a hash of the full name.
• Tests: tests/end2end_options.rb adds module pairs Options::PgV2 / Options::Pg::V2 and two long module names that share their first 63 bytes. Each pair runs on one ConnectionPool. Unit test: test_statement_names_of_different_classes_never_collide.

2. Columns named like private row hooks broke every row (Low-Med). A column named initialize (also initialize_copy, initialize_clone, initialize_dup, method_missing) was not renamed, so Data / Struct construction raised ArgumentError.

• Fix: these names are added to the reserved row-method names (Code::CONSTRUCTION_METHODS), so they get a trailing _ (initialize_).
• Tests: the example schemas of all three engines get an initialize column in name_cases. test_keyword_and_method_names_keep_distinct_row_and_argument_names reads it back as initialize_. Unit test test_columns_named_like_row_methods_still_build_and_read_rows goes through every public and private Struct / Data method on the running Ruby, then builds, reads, dups, clones and withs a row for each one.

3. Query file names were not CamelCased (Low). admin_users.sql generated Admin_usersSql, which is a RuboCop Naming/ClassAndModuleCamelCase offense.

• Fix: Code.constant_part CamelCases file names, so admin_users.sql becomes AdminUsersSql in admin_users_sql.rb. Two files that map to the same class are still rejected, and the error now names the files.
• Tests: the options E2E adds sqlite_queries/admin_users.sql and runs AdminUsersSql. Unit tests: test_query_file_names_become_camel_case_classes and the updated test_query_files_writing_the_same_file_are_rejected.

Breaking change: a query file with _ in its name now generates a different class name. For example, my_queries.sql used to give My_queriesSql and now gives MyQueriesSql. Callers have to rename the class. File names stay the same.

Suspicions from the sweep

• rubyVersion: "3" generated Data, which breaks on Ruby 3.1. A missing minor version now counts as .0 for this check, so 3 generates Structs (test_immutable_data_support).
• MySQL :execrows counts changed rows, not matched rows (mysql2's default). This is now documented with the flags: Mysql2::Client::FOUND_ROWS alternative (test_execrows_counts_changed_rows_unless_found_rows_is_set).
• A String passed to a PostgreSQL json / jsonb parameter is JSON text. This is now documented; the existing JSON round-trip test already covers it.
• SQLite busy_timeout= waits while holding the GVL, so other threads stop. This is now documented, recommending busy_handler_timeout= (test_busy_handler_timeout_waits_for_another_connection_without_blocking_threads).
• SQLite booleans: no change. The generated RBS already uses Integer, and passing true is a sqlite3 gem limitation.

examples/, test/golden/, test/fixtures/*.bin and steep_expectations.yml are regenerated (only line numbers moved in the Steep expectations). README, docs/Coverage.md and docs/Testing.md are updated.

Checklist

• examples/ is regenerated (rake sqlc_generate_process) and committed
• Tests added or updated (rake test, and rake end2end_* when the generated code changes)
• The generated code stays RuboCop-clean (rake rubocop) and Steep-clean (rake steep)
• Labelled (bug, enhancement, documentation, or breaking) so the release notes are grouped correctly

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring