test: restore removed test file

[Planeshifter]

Description

What is the purpose of this pull request?

This pull request:

Follow-up fixes for commits merged to develop between 2026-09-14T16:19:57-07:00 (adb952a) and 2026-09-15T02:37:51-07:00 (c24d71e).

• math/base/special: Fix: 5258215e0 clobbered lib/node_modules/@stdlib/math/base/special/test/test.js when adding tanhf, git recording the new package's lib/index.js as a rename of it and leaving directories.test in package.json pointing at nothing. Restored the test file verbatim from the parent commit.
• fft/base/fftpack/float32/decompose: Fixes a doc bug from a697555: the refactor of fft/base/fftpack/float32/decompose rewrote two description lines in lib/node_modules/@stdlib/fft/base/fftpack/float32/decompose/docs/repl.txt to backtick N but left the original two in place, duplicating the paragraph in the REPL help. Removes the stale pair.

Related Issues

Does this pull request have any related issues?

No.

Questions

Any questions for reviewers of this pull request?

No.

Other

Any other information relevant to this pull request? This may include screenshots, references, and/or implementation notes.

Validation

All 43 commits merged to develop in the review window were audited for style guide compliance (against docs/style-guides and established reference packages) and scanned for bugs (diff-scoped logic, values, and documentation checks; the new tanhf implementation was executed against its Python-generated fixtures in both JS and C, and the new blas/ext/base/glast-index-greater-than-equal examples were hand-evaluated). Only objective, independently re-verified defects were fixed; anything requiring interpretation, judgment calls, or changes outside the window's diff was deliberately excluded. Notably excluded: fft/base/fftpack/float32/rffti still passing Float32Array workspaces to the refactored decompose — behavior is unchanged and a change would fall outside the reviewed diff.

Checklist

Please ensure the following tasks are completed before submitting this pull request.

• Read, understood, and followed the contributing guidelines.

AI Assistance

When authoring the changes proposed in this PR, did you use any kind of AI assistance?

• Yes
• No

If you answered "yes" above, how did you use AI assistance?

• Code generation (e.g., when writing an implementation or fixing a bug)
• Test/benchmark generation
• Documentation (including examples)
• Research and understanding

Disclosure

If you answered "yes" to using AI assistance, please provide a short disclosure indicating how you used AI assistance. This helps reviewers determine how much scrutiny to apply when reviewing your contribution. Example disclosures: "This PR was written primarily by Claude Code." or "I consulted ChatGPT to understand the codebase, but the proposed changes were fully authored manually by myself.".

This PR was produced by a scheduled Claude Code review of the last 24 hours of commits merged to develop; the fixes were authored and validated by Claude Code.

🤖 Generated with Claude Code

https://claude.ai/code/session_01ApgqTYLbBALZcimF9KJqM5

---

Generated by Claude Code

[stdlib-bot]

Coverage Report

No coverage information available.

[kgryte]

As the CI failure is not due to changes introduced in this PR, I'll go ahead and merge.

[Planeshifter]

The Run affected tests failure on c42d57c is not this PR's: it comes in with the develop merge and reproduces verbatim on the develop tip (abd365a), which this PR's one-file diff (math/base/special/test/test.js) cannot affect.

Root cause: 7333861 updated fft/base/fftpack/float32/rffti to write factorization results through a Uint32Array view over the workspace (raw integer bits), but cosqi and sinqi — which delegate to rffti — still have tests (and doc examples) that read the factors back as Float32Array values. Reading integer bit patterns as float32 yields denormals, which is exactly what CI shows: expected 8, 2, 2, 4, got 1.12e-44, 2.8e-45, 2.8e-45, 5.6e-45 (the bit patterns of 8, 2, 2, 4). Deterministic, so a re-run won't help.

Proposed patch (not pushed here to avoid widening this PR into the in-flight integer-view migration): in cosqi/test/test.js and sinqi/test/test.js, read the factorization region through a Uint32Array view over the workspace buffer, following the pattern already used in rffti/test/test.js (fview = new Uint32Array( w.buffer, w.byteOffset, w.length )), and update the // returns <Float32Array>[ 8, 2, 2, 4 ] doc examples in both packages accordingly — i.e., the same refactor! treatment 7333861 gave rffti.

---

Generated by Claude Code