next.js

The toolbar was blocked on sourcemapping the stackframes for the "copy details button". This suspended the whole devtools which meant you couldn't even see the error Now we only block on sourcemapping when we actually copy. We can leverage now that devtools us using modern React. We're only blocking for up to 2s and then fallback to including the unsourcemapped stack. Before: https://github.com/user-attachments/assets/601c0aa5-fb3a-42ce-b41e-afff06c5aba2 After: https://github.com/user-attachments/assets/9d80b0c1-2346-421f-a4bc-d2e87753c46c

What Add support for multi-valued tables in . A multi-valued table allows multiple distinct values to be associated with a single key. Each family is independently configured as (existing behavior) or via the new enum. Why This will support the table (implemented in #88904), where keys will change to be _hashes_ instead of full values. This greatly decreases DB size and speeds up queries due to smaller key sizes, at the cost of hash collisions requiring multiple values per key. How API New enum ( / ) and per-family in for single-valued families (panics if called on multi-valued) for multi-valued families, returns — stack-allocated for the common 0–1 result case, heap-scales when needed and are unchanged — the family kind controls dedup/compaction behavior Write path & compaction Single-valued (unchanged): last-write-wins per key Multi-valued: all distinct pairs are preserved; true duplicates (same key AND same value) are dropped during compaction Deletion inserts a tombstone that shadows all older values for that key across SST layers. Values written after the tombstone in the same batch are retained. in implements the three-step multi-value dedup: (1) stable sort by key, (2) prune pre-tombstone entries, (3) sort by and remove true duplicates this logic is shared by writes and compacion Read path Controlled by a const generic on the internal lookup methods Single-valued (): binary search, return _last_ match, stop This fixes a bug where we might return Deleted when there is a value in the SST depending on what the search algorithm found first Multi-valued (): scan all matching entries in the SST block, then continue to older SSTs. If a tombstone is found, stop searching older layers. Final read-path dedup ensures consistency across SSTs that haven't been compacted yet this is implemented as a simple sort/dedup, which is reasonable since we don't expect many duplicates, but this could be optimized if we expect large sets SST file changes New abstraction () for uniform entry handling for writing and compaction extended with path: scans backwards and forwards to find all matches Canonical value ordering () ensures deterministic sort and dedup when writing and compacting. Tests ~1500 lines of new tests covering: Basic multi-value get/put across batches and within batches Compaction preserving distinct values while dropping duplicates Deletion semantics: tombstone clears all prior values, no resurrection after GC Deduplication across SSTs and within single SSTs Edge cases: trailing deletes, interleaved batches, repeated compactions, 20+ values per key