test: migrate math/base/special/expm1rel to ULP-based assertions

[kgryte]

Resolves a part of #11352.

Description

What is the purpose of this pull request?

This pull request:

• migrates the tests for math/base/special/expm1rel from relative tolerance (EPS-scaled) assertions to ULP-based assertions using @stdlib/assert/is-almost-same-value, matching the idiom used by previously converted double-precision packages (e.g., hypot, ellipk, expit, cphase).
• updates both test/test.js and test/test.native.js, removing the abs import and the delta/tol computations in favor of t.strictEqual( isAlmostSameValue( v, expected[ i ], N ), true, 'returns expected value' );. The EPS import is retained in both files, as it is still used by the "returns 1.0 near 0.0" test.

Final ULP constants (per fixture set, identical in test/test.js and test/test.native.js):

Fixture set	ULP bound	Worst case
medium_negative.json	0	exact for all 1000 points
medium_positive.json	2	x = 87.55771771771771, y = 1.2121021459952258e+36, expected 1.2121021459952261e+36
small_negative.json	3	x = -0.22922922922922928, y = 0.8936633379449439, expected 0.8936633379449436
small_positive.json	3	x = 0.4644644644644645, y = 1.2727816274388977, expected 1.272781627438897
tiny.json	0	exact for all 1000 points

The previous tolerance was a uniform 3.0 * EPS * abs( expected[ i ] ) for every fixture set.

The bounds were tightened empirically rather than guessed. Starting from a high bound and lowering it, the measured maximum ULP distance was computed for every one of the 5000 fixture points, and the values above are the minima: lowering the 2 to 1 and the 3s to 2 produces 21 failing assertions, while at the stated bounds all 5008 assertions pass.

Related Issues

Does this pull request have any related issues?

This pull request has the following related issues:

• [RFC]: Migrate math/base/special packages from relative tolerance testing to ULP difference testing (tracking issue) #11352

Questions

Any questions for reviewers of this pull request?

The medium_negative and tiny fixture sets are reproduced exactly (0 ULP). Reviewers may prefer a floor of 1 there for robustness against future implementation changes; 0 is used here as it is the measured minimum, consistent with the guidance in #11352 and with the ~300 existing isAlmostSameValue( ..., 0 ) assertions already in the tree.

Other

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

• Only the two test files are modified; no implementation, fixture, or documentation changes.
• The suite was run twice at the final bounds to confirm the result is deterministic (no FMA/architecture-dependent flakiness).
• The native add-on could not be built in the authoring environment, so test/test.native.js skipped locally. To avoid guessing its bounds, the C implementation (src/main.c, plus its expm1/abs/is_nan dependencies) was compiled standalone with gcc -O2 and evaluated over the same fixture points: the C implementation reproduces the JavaScript results bit-for-bit, giving identical per-set bounds of 0, 2, 3, 3, 0. The two test files therefore use the same constants. Confirmation from CI (where the add-on is built) would be welcome.
• ESLint (etc/eslint/.eslintrc.tests.js) reports no errors or warnings for either file; the pre-commit JavaScript test lint and filename lint hooks both passed.

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 authored by Claude Code as part of an automated, unattended run: it selected the package, studied the idiom used in previously merged conversions, applied the test edits, and measured the minimum passing ULP bounds by running the suite.

---

@stdlib-js/reviewers

---

Generated by Claude Code

[stdlib-bot]

Coverage Report

Package	Statements	Branches	Functions	Lines
math/base/special/expm1rel	$\\color{green}198/198$ $\\color{green}+100.00\\%$	$\\color{green}10/10$ $\\color{green}+100.00\\%$	$\\color{green}2/2$ $\\color{green}+100.00\\%$	$\\color{green}198/198$ $\\color{green}+100.00\\%$

The above coverage report was generated for the changes in this PR.