2026-09-04 - (/loop) strftime, and the question that found it

differential   326  ->  345 cases
suites 125   tests 2013   clippy/fmt/identity clean

Eighteenth cron tick.

The question, which is the reusable part

Two ticks running had paid off in the same place - a function orbistoun implements and nothing has ever compared - so this asked it mechanically instead of picking a third by hand: which implemented libc/libScePosix functions have no differential case?

Twenty. Most are correctly out of scope, and why is worth as much as the answer:

• Annex K _s forms and strnstr - glibc does not provide them. D512 ruled these out and was right.
• rand - would pin orbistoun to glibc's generator, which is not the platform's.
• strerror/strerror_r - message text is implementation-specific; the cases would enforce glibc's wording as though it were the contract.
• memalign, strdup, strndup - answer addresses. strdup's contents are comparable and worth doing later; the pointer is not.
• strtoll, strtoimax, strtoumax - strtoimax delegates to strtoll, and on LP64 strtol and strtoll clamp identically. The overflow boundary is already covered.

That leaves strftime.

Why it was the right one

A specifier parser - the shape D511 established as where bugs hide, since both sprintf bugs were in the parser rather than in any conversion. So the format is the case input and the date is held still, except where a boundary is the point: %F/%T expansions, %e against %d, %y at 1999/2000, %p at midnight and noon, %j being one-based, and the room boundaries.

All nineteen agree. No bug, and a previously uncompared function is now compared.

What crosses, and what does not

Only the nine int fields of struct tm, as a b: blob - ISO C fixes their names and order, and orbistoun cites FreeBSD's LP64 form where it reads them. Fields past the ninth are not recorded because orbistoun does not read them, and %Z/%z are among the conversions it refuses: a timezone case would test this emulator's lack of a timezone, not its formatting.

The buffer is compared only when the call succeeded - ISO C leaves the contents unspecified otherwise, so comparing them would compare something neither implementation promises.

The break, on exactly one case

strftime/e-pads-with-space: buffer is 5b30345d..., expected 5b20345d...

[04][04] where it should be [ 4][04] - 0x20 against 0x30. One byte, one case, no other case noticed.

Decision: D532.