GeezerVison

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

Fixes #386

Summary

scan --mode hosted now pins a crates.io dependency declared with an explicit registry = "crates-io", in both the inline-table and the [dependencies.x] table spellings, instead of skipping it as "pinned to another registry".

Root cause

The hosted Cargo.toml planner (plan_cargo_toml in patch/redirect/mod.rs) let a registry = … value through only when it was our current pin or a stale socket-patch-<uuid> pin. Every other value was refused, including crates-io, Cargo's built-in name for crates.io. Cargo resolves that declaration exactly like one with no registry key, so hosted mode skipped a dependency it is built to redirect. It reported redirected: 0 and exited 0 with status: success while the project kept building the unpatched crate.

Changes

• is_repinnable_registry_name: crates-io is now overwritten in place, the same way a superseded socket-patch pin is, in both the table and inline arms. Any other registry name is still refused.
• Upstream restore (remove): as for every pin, the original spelling isn't recorded, so the declaration comes back without the redundant registry = "crates-io" key (the inline shorthand collapses to name = "1.0.4"). Cargo resolves it the same way and the lock is unchanged. This is documented in the module doc. unpin_line now keeps a trailing comment from a table-form pin line instead of dropping it with the line.
• mode_migration_cargo::takeover_over_unrewritable_spelling_keeps_vendored (Cargo vendored → hosted takeover un-vendors a crate before hosted mode refuses its Cargo.toml spelling (dotted keys, single quotes, registry = "crates-io"), leaving it unpatched, while --dry-run previews redirected: 1 #1020) used registry = "crates-io" as its example of a refused spelling. It now uses a multi-line inline table (Hosted cargo redirect refuses a dependency whose inline table has a multi-line features array, because it plans Cargo.toml pins with a line scanner #757), which hosted mode still refuses.

Tests (red → green)

Local runs on 46af5c2 (c2b3f89 adds the Bugbot CRLF fix; upstream::cargo tests, core clippy and e2e_redirect_cargo_shapes re-run green):

• cargo fmt --all -- --check: clean
• cargo clippy --workspace --all-features -- -D warnings: clean
• cargo test -p socket-patch-core --all-features: 6321 passed, 4 failed. The failures are copy_tree::relax_loop_must_not_traverse_symlinked_root, vlt_heal::an_unremovable_hidden_lock_keeps_every_store_entry, pypi_poetry::wire_write_failure… and pypi_requirements::wire_failure_rolls_back…. All four are chmod-based tests that fail because the agent sandbox runs as root, and they're unrelated (see Go go.work audit: false NotAMember for a use path through a missing dir on macOS (breaks macOS test job) #1409 for the same note).
• SOCKET_PATCH_CARGO_E2E_REQUIRED=1 cargo test -p socket-patch-cli --all-features --test e2e_redirect_cargo_shapes --test mode_migration_cargo --test e2e_redirect_cargo_build: 31 / 29 / 22 passed
• cargo test -p socket-patch-cli --lib --test in_process_cargo_apply: ok
• The full cargo test --workspace couldn't finish locally because the sandbox ran out of disk linking ~100 CLI test binaries. CI covers it.
• No wrapper changes: the fix is in core only, and npm/, pypi/ and gem/ just dispatch to the binary.

/code-review high: findings not fixed

• A triple-quoted registry = '''crates-io''' or a registry = '…' string inside a feature name defeats the line regex. The planned manifest then fails validation and the dep is refused, so it fails closed and is never mis-pinned. Both are contrived spellings; the general fix is Hosted cargo redirect refuses a dependency whose inline table has a multi-line features array, because it plans Cargo.toml pins with a line scanner #757's move onto toml_edit.
• A restored override isn't CRLF-converted in the e2e harness. No CRLF shape uses it.
• The explicit restored: vec[] on every Shape literal matches the harness's existing explicit-field style.

Not in scope

#757 (multi-line inline tables) touches the same planner but has a different cause: the planner reads one line per inline table. It needs the planner moved onto toml_edit and the upstream restore taught multi-line edits.

🤖 Generated with Claude Code

Generated by Claude Code