fix(printf): compute asterisk width magnitude in unsigned arithmetic

[bolverk]

Fixes #13766

Summary

A negative * field-width argument of i64::MIN (printf '%*d' -9223372036854775808 1) made resolve_asterisk_width panic with attempt to negate with overflow: the old code computed the magnitude as -(nb as isize), and |i64::MIN| = 2^63 is not representable in i64/isize.

The fix computes the magnitude with nb.unsigned_abs(), which is designed for exactly this case and returns 2^63 in unsigned (u64) arithmetic.

Behavior

• Before: panic / exit 134 (attempt to negate with overflow)
• After: the width 2^63 exceeds MAX_FORMAT_WIDTH, so printf fails gracefully with printf: write error: formatting width too large (exit 1), matching the handling of other oversized widths.

Verification

Formal verification of the width-resolution arithmetic was done in Dafny (4.11): it specifies the intended behavior (negative width ⇒ left-align with magnitude), proves |i64::MIN| overflows i64 but fits u64/usize, and proves the unsigned_abs()-based implementation is overflow-free and matches the spec. Proof: 15 verified, 0 errors.

Tests

• New unit test resolve_asterisk_width::asterisk_i64_min_width (both SignedInt and Unparsed argument forms)
• New integration test test_printf::test_asterisk_width_i64_min_no_panic
• Full printf suite passes: 131 passed, 0 failed, 1 ignored; cargo fmt and clippy clean.

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/date/date-locale-hour (fails in this run but passes in the 'main' branch)
Skipping an intermittent issue tests/tail/symlink (passes in this run but fails in the 'main' branch)
Note: The gnu test tests/misc/write-errors was skipped on 'main' but is now failing.

[bolverk]

Updated the integration test to be portable: on 32-bit targets (usize = 32 bits) |i64::MIN| = 2^63 does not fit, so the width is clamped to 0 (no padding) and printf succeeds with |1|; on 64-bit targets it exceeds MAX_FORMAT_WIDTH and fails with a write error. In both cases the original panic ("attempt to negate with overflow") is gone. Verified locally on both x86_64-musl and i686-musl (131/131 printf tests pass on each).

[sylvestre]

Please make the comment shorter

[codspeed-hq]

Merging this PR will improve performance by 5.1%

⚡ 3 improved benchmarks
✅ 344 untouched benchmarks
⏩ 46 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	du_max_depth_balanced_tree[(6, 4, 10)]	63.2 ms	59.9 ms	+5.51%
⚡	Simulation	ls_recursive_deep_tree[(200, 2)]	3.6 ms	3.4 ms	+5.04%
⚡	Simulation	du_summarize_balanced_tree[(5, 4, 10)]	16.2 ms	15.5 ms	+4.76%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing bolverk:fix/printf-asterisk-width-i64-min-overflow (afbc154) with main (9f29502)}

Footnotes

1. 46 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. ↩