bench: remove 'sort_random_strings'

[sylvestre]

too random !

[codspeed-hq]

CodSpeed Performance Report

Merging #9030 will improve performances by 4.23%

_{Comparing sylvestre:remove-bench (d26ff97) with main (a6f580c)}

Summary

⚡ 4 improvements
✅ 101 untouched
⏩ 74 skipped¹

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
⚡	du_human_balanced_tree[(5, 4, 10)]	10.4 ms	10.1 ms	+2.79%
⚡	du_max_depth_balanced_tree[(6, 4, 10)]	33.1 ms	32.3 ms	+2.49%
⚡	sort_ascii_c_locale	22.9 ms	22.5 ms	+2.07%
⚡	sort_ascii_utf8_locale	46.2 ms	44.3 ms	+4.23%

Footnotes

1. 74 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[github-actions]

GNU testsuite comparison:

Skipping an intermittent issue tests/tail/overlay-headers (passes in this run but fails in the 'main' branch)

[naoNao89]

@cakebaker this PR removed the generate_random_strings() function from src/uucore/src/lib/features/benchmark.rs on Oct 26.

the problem: this function used a deterministic seed ((i * length + j) * 17 + 42) to generate reproducible test data. Removing it eliminated benchmark stability.

cascading effect: since Oct 30, unrelated PRs (#9097, #9088, #9094, ..) show false CodSpeed regressions:

#9097 (doc-only): -31.87% degradation
#9088 (cksum debug): -4.31% degradation
#9094 (stty tests): -3% degradation
#9112 (.github/workflows/CICD.yml): -3.91% degradation
All regress the same benchmark (du_human_baland_tree) despite unrelated code changes—indicating measurement noise, not real performance issues.

[cakebaker]

@naoNao89 hm, I don't see the connection. du_human_balanced_tree didn't use generate_random_strings as far as I can see.

[naoNao89]

@cakebaker you're right, and I apologize for the confusion.

The real issue is that CodSpeed needs to establish a new baseline:

• Before PR bench: add cksum performance benchmarks #9075: Baseline = 19 benchmarks (no cksum)
• After PR bench: add cksum performance benchmarks #9075: Matrix = 20 benchmarks (with cksum)

Sorry for the initial misdiagnosis