libc: Add missing size check to qsort_s(3)

[• hselasky]

Summary

I find it very strange both the C11 standard
(ISO/IEC 9899:2011, K.3.6.3.2) and the ISO/IEC JTC1 SC22 WG14 N1172 standard,
does not define sorting an array having objects of zero size,
as undefined behaviour.

Add proper checks for this. Found while working on bsort(3).

MFC after: 1 week
Sponsored by: NVIDIA Networking

[• hselasky]

Can I have a quick go on this one?

[• hselasky]

The qsort code actually use the object size for advancing for loops, so clearly we should catch this regardless of ISO definitions.

[emaste]

Interesting, https://en.cppreference.com/w/c/algorithm/qsort

Unlike other bounds-checked functions, qsort_s does not treat arrays of zero size as a runtime constraint violation and instead returns successfully without altering the array

I'd take this to mean n == 0 but I think it's ambiguous; their previous use of size is in reference to element size.

But Microsoft's docs https://github.com/MicrosoftDocs/cpp-docs/blob/main/docs/c-runtime-library/reference/qsort-s.md explicitly show width <= 0 as EINVAL.

I think your change is good.

[kib]

lib/libc/stdlib/qsort.3

273–277

[• hselasky]

@emaste : "<= 0" is for the future, in case the type changes to be signed.

[emaste]

[emaste]

In D39687#903226, @hselasky wrote:

@emaste : "<= 0" is for the future, in case the type changes to be signed.

Yep, my point is just that the MS docs already have exactly this case.

[• hselasky]

Thank you for your reviews!