fix(stdlib): return error instead of panicking on out-of-range combinations size 🧮

[timfennis]

Context

While fuzzing the pipeline for panics, I found that combinations(seq, k) aborts the process with capacity overflow for out-of-range k:

• combinations([1, 2, 3], -1) — k as usize turns -1 into usize::MAX
• combinations([1, 2, 3], 9223372036854775807) — a genuinely huge positive k

Both reach CombinationsIter::new, which eagerly built its index vector with (0..k).collect() and asked the allocator for a usize::MAX-element Vec.

Changes

• ndc_vm: CombinationsIter now stores k and defers the indices allocation until ensure_index(k - 1) confirms the buffer holds at least k elements. The allocation is then bounded by data actually materialised, so an over-large k over a short source yields no combinations rather than overflowing. This also covers a large positive k, which the signature change alone does not.
• ndc_stdlib: combinations, permutations, and subsequences took k/length as i64 and cast with as usize. They now take usize, so argument conversion rejects a negative value with the same error as the sibling windows/chunks/take functions. Previously permutations and subsequences silently returned [] for negative input.

Tests

Four regression tests under tests/functional/programs/900_bugs/: the huge-positive and negative cases for combinations, plus the negative cases for permutations and subsequences.

🤖

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f957eb7ec9

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".