rm: don't treat symlinks as write-protected

[GomesGoncalo]

Detect symlinks in prompt_file() using fs::symlink_metadata() and handle them differently:

• InteractiveMode::Always should still prompt with "remove symbolic link ...?".
• For all other interactive modes, do not follow the symlink target or use the target's permissions to decide a write-protected prompt.

Adds a unit test test_symlink_not_write_protected_prompt to verify that a symlink to a read-only file is not treated as a write-protected regular file.

Closes #10222

[codspeed-hq]

CodSpeed Performance Report

Merging this PR will improve performance by 4.84%

_{Comparing GomesGoncalo:goncalo.gomes/rm-symlink-behaviour (d479e9b) with main (c498595)}

Summary

⚡ 3 improved benchmarks
✅ 279 untouched benchmarks
⏩ 38 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	rm_multiple_files	2.2 ms	2.1 ms	+4.84%
⚡	Memory	du_wide_tree[(5000, 500)]	1.2 MB	1.2 MB	+3.54%
⚡	Memory	cp_large_file[16]	119.4 KB	115.2 KB	+3.64%

Footnotes

1. 38 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. ↩

[GomesGoncalo]

CodSpeed Performance Report

Merging this PR will degrade performance by 4.57%

Comparing GomesGoncalo:goncalo.gomes/rm-symlink-behaviour (aad33db) with main (fd9157a)

Summary

❌ 1 regressed benchmark ✅ 140 untouched benchmarks ⏩ 38 skipped benchmarks1

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

Mode Benchmark BASE HEAD Efficiency
❌ Simulation rm_multiple_files 2.2 ms 2.4 ms -4.57%

Footnotes

1. 38 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](https://codspeed.io/uutils/coreutils/branches/GomesGoncalo%3Agoncalo.gomes%2Frm-symlink-behaviour?sectionId=benchmark-comparison-section-baseline-result-skipped&utm_source=github&utm_medium=comment&utm_content=archive). [↩](#user-content-fnref-skipped-65dfc430c8cce12c4551dab4f7d2efc4)

looks like entirely related to querying symlink metadata, I tried reversing the logic but found the branch needs to be always taken to keep the functionality

[ChrisDryden]

I'm trying to see if theres a way to deduplicate the metadata call so that theres no performance regression, I think it should be possible only have one call

[ChrisDryden]

For non-symlinks, fs::symlink_metadata is the same as fs::metadata, you should be able to combine the in a way to only have to call fs::symlink_metadata once and reuse that for later. Maybe if you move this above 803 and change it to symlink_metadata you can use metadata.is_symlink() and it means theres only one stat each time this function is called.

[GomesGoncalo]

that's a good idea, I tested and it still worked, should behave better now

[GomesGoncalo]

I believe that even improves performance slightly

Re-evaluating after double checking the performance regressions

[github-actions]

GNU testsuite comparison:

Congrats! The gnu test tests/basenc/bounded-memory is now passing!