[cberner]

TableTypeMismatch reported only the types a table is stored with, as {table} is of type Table<K, V>. Two problems:

• The niche case reads as a contradiction. A table of Option<&str> written by redb 2.6, opened with redb 5's niche Option<&str>, failed with values is of type Table<u64, Option<&str>>, which is exactly the type the caller asked for. The stored spelling and the niche spelling print the same in that message.
• Multimap tables were mislabelled. A multimap table opened with the wrong types was also described as Table<..>.

Changes:

• Wording, in both modes: the message no longer names the table type: {table} has key type K and value type V.

• Under experimental-api-5: the TableTypeMismatch and TypeDefinitionChanged variants of TableError and Error hold a boxed struct, TableTypeMismatch or TypeDefinitionInfo, with private fields and accessors:

  • TableTypeMismatch: table(), stored_key(), stored_value(), requested_key(), requested_value().
  • TypeDefinitionInfo: name(), alignment(), width().

  Boxing makes TableError and Error 32 bytes, down from 88 on master. With the requested types inline they would have been 152, over clippy's result_large_err limit at 20 functions. Private fields mean later releases can add to the structs without a break.

• New message under the feature: the message adds the requested types: values has key type u64 and value type Option<&str>, but was opened with key type u64 and value type niche::Option<&str>.

• Without the feature, the variants keep their 4.x fields; only the wording above changes.

• Construction goes through TableError::type_mismatch() and type_definition_changed(), so table_tree_base.rs has no cfgs.

• Patterns: existing TableTypeMismatch { .. } patterns still compile against the tuple variants.

Tests:

• niche_tests: checks the exact message on a real table written by redb 2.6.
• multimap_tests: checks the wording in both modes.
• basic_tests: checks the accessors and the message for a u32/NonZeroU32 mismatch.

Checked:

• cargo fmt.
• clippy --all --all-targets, with and without --all-features.
• clippy for thumbv7em (no_std).
• cargo doc.
• cargo test --all-features and cargo test with default features: the full suites before review, and the lib tests plus niche, multimap and basic tests after the review changes.
• The no_std regression crate.

🤖 Generated with Claude Code

https://claude.ai/code/session_01ALhxet6NP69dCfAUv8TAhR

[chatgpt-codex-connector]

Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-09T03:19:17.046731Z	1f2b002	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.