Read Cargo.toml [package] identity through formats::cargo::manifest (#693)

[Mikola Lysenko (mikolalysenko)]

LLM Description written by Claude Code:claude-opus-5-5

Fixes #693 (slice 1 of 2)

Summary

Three parsers read a Cargo.toml's [package] name and version: the cargo crawler's hand-rolled line scanner, VEX product detection's scan_toml_section, and cargo_tag's own toml_edit lookup. This PR adds one toml_edit reader, formats::cargo::manifest, moves all three onto it, and deletes the crawler's scanner.

Why

• Issue #693, register row E15 (living document §5.4, Cargo).
• The three readers had drifted. On a BOM, the legacy [project] table, dotted keys or invalid TOML, the crawler, VEX and cargo_tag gave different answers (see the table in Read Cargo.toml package name and version through one shared toml_edit reader #693).
• In a cargo vendor directory, where the dir name carries no version, the crawler couldn't find a crate with such a manifest at all.
• Leverage: B 0; U 1 (Vendored cargo warns cargo_multi_version_old_cargo ("declares no rust-version") when the root package inherits rust-version from [workspace.package] #651 and the vendor/cargo.rs slice build on this reader); D 3 parsers into 1, about −110 production lines; R L.

What changed

• New formats/cargo/manifest.rs:
  • package_table(&Table) returns [package], else legacy [project];
  • package_name_version(&str) returns the literal name and version, or None for invalid TOML, a missing or empty field, or a non-literal field such as version.workspace = true. It is pure, so it passes the formats architecture guard.
• crawlers/cargo_crawler.rs (and its test oracle) calls package_name_version. Deleted: parse_cargo_toml_name_version, parse_table_header, extract_string_value and the scanner-specific unit tests. The tests that pin crawler behavior (comments, quotes, subtables, workspace inheritance) stay and now run through the shared reader.
• vex/product.rs parse_cargo_toml calls package_name_version. scan_toml_section stays for pyproject.toml (E38's).
• vendor/cargo_tag.rs version_literal finds its table through package_table and keeps only the span and quote logic.
• tests/crawler_cargo_e2e.rs imports the reader from formats::cargo::manifest.

Left for slice 2: vendor/cargo.rs (path_crate_version, declared_cargo_minor, and #651's rust-version inheritance). Open PRs #1026, #1039, #1041, #1043 and #1050 change that file.

Diff

7 files, +250 / −247: production about +45 / −115, the rest tests.

Behavior

Lockfiles, JSON output, exit codes and the CLI contract don't change. Cargo-style parsing changes the reader in these cases:

• Crawler: now reads a manifest with a BOM, [project], dotted keys or an inline package = { … }. It no longer reads invalid TOML ([package] junk, a stray [oops line); those fall back to the directory name, as version.workspace = true already did.
• VEX product: now reads [project], dotted keys and the inline table.
• cargo_tag: unchanged.

Normalized crates.io manifests (plain [package]) take none of these paths. On all 208 real registry manifests in ~/.cargo/registry/src, the old scanner and the new reader agree on every file.

Cost on the crawl path

A full toml_edit parse is slower than the line scanner, which stopped early. Release build, 208 real registry manifests × 20 iterations:

• old scanner: 0.5 µs per manifest;
• new reader: 81 µs per manifest.

crawl_all reads each crate's Cargo.toml on the walk pool, so 1,000 cached crates cost about 80 ms of extra CPU spread across the pool's threads. This is the price of agreeing with cargo, and #693 asks for toml_edit explicitly. S = 0 (no speed-up claimed).

Tests

• New: vex::product::tests::cargo_manifest_readers_agree. It runs the issue's drift rows (plain, BOM, [project], dotted, inline, workspace, invalid, no package) through the crawler's reader, VEX parse_cargo_toml and cargo_tag::tag_manifest_text, and asserts they agree.
• New: formats::cargo::manifest::tests, covering every spelling, [package] over [project], and the inputs it can't read.
• New: find_by_purls_vendor_layout_reads_every_manifest_spelling (crawler e2e): vendored crates with a BOM, [project] or dotted-key manifest are found. It fails on main, where the scanner returns None and the version-less dir name can't stand in.
• cargo clippy --workspace --all-features -- -D warnings: clean.
• cargo test -p socket-patch-core --lib: 5711 passed, 4 failed. The 4 are the known root-only sandbox failures that also fail on main (copy_tree::relax_loop_must_not_traverse_symlinked_root, vlt_heal::an_unremovable_hidden_lock_keeps_every_store_entry, pypi_poetry::wire_write_failure_maps_error_and_leaves_lock_untouched, pypi_requirements::wire_failure_rolls_back_already_written_files).
• cargo test -p socket-patch-core --test crawler_cargo_e2e: 34 passed.
• cargo test -p socket-patch-cli --all-features with --test in_process_cargo_apply, mode_migration_cargo, e2e_cargo, covgap_commands_vex, e2e_vex_redirect and e2e_vex_vendor: all green.

Risk

Low. The changes are the documented reader cases above, and the extra parse CPU on the cargo crawl path.

🤖 Generated with Claude Code

---

Note

Medium Risk
Changes crate discovery and VEX product PURLs for non-standard but cargo-valid manifests; crawl path is slower per manifest though normal registry layouts are unchanged.

Overview
Unifies Cargo.toml package identity behind a single toml_edit reader in formats::cargo::manifest (package_name_version / package_table), replacing three divergent parsers (cargo crawler line scanner, VEX scan_toml_section, and duplicated table lookup in cargo_tag).

The cargo crawler and its test oracle now call package_name_version; the hand-rolled parser and most of its unit tests are deleted. VEX parse_cargo_toml and cargo_tag::version_literal route through the same module. Behavior shifts for edge manifests cargo accepts: UTF-8 BOM, legacy [project], dotted keys, and inline package = { … } are read consistently (fixing vendored crates whose dir name has no version). Invalid TOML is no longer partially parsed by the crawler scanner—it returns None and may fall back to directory-name parsing. Crawl path pays a full toml_edit parse per manifest.

Tests add manifest unit coverage, cargo_manifest_readers_agree, and vendor-layout e2e for alternate spellings.

^{Reviewed by Cursor Bugbot for commit 3f4b741. Configure here.}

---

Generated by Claude Code

[Mikola Lysenko (mikolalysenko)]

BugBot review

---

Generated by Claude Code

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 3f4b741. Configure here.}

[Mikola Lysenko (mikolalysenko)]

Ready for review (burn-down agent).

• Head: 3f4b7417ccabb9577718e53ecaa97893cffc4a7e
• CI: 503/503 check runs green (success/skipped/neutral) on this head, mergeable, no conflicts.
• Bugbot: reviewed this head (Cursor Bugbot check: success), no unresolved review threads.
• Changelog: untouched.

Nothing specific flagged for the reviewer beyond the PR description.

---

Generated by Claude Code