[Noxy-HNR]

What

strcasecmp and strncasecmp used the host's tolower, and WebKit's strcasestr did the same. Changing host LC_CTYPE could make bytes compare equal even though the guest still exposes the C locale. On Windows code page 1252, for example, C4 and E4 became equal.

Use the existing guest case conversion for the libc comparisons and the existing ASCII folding helper for the substring search. The result difference and first-match behavior follow FreeBSD 11.4 strcasecmp.c and strcasestr.c. The debt entries retain the unmeasured console locale behavior.

Tested

Windows 11 26H2 (build 10.0.26300), WinLibs GCC 15.2.0 r7, full Release build, head 07bec52 on main d6e7ab4.

• guest_strings, guest_fnmatch, guest_locale and guest_rune_locale pass, 4/4.
• All 65,536 byte pairs are checked for comparisons, bounded comparisons, substring search and folded literal fnmatch, under C and an installed Windows code page 1252 locale. Empty strings, bounded nonterminated inputs, repeated prefixes, first match, no match and unchanged inputs are covered. The new regression fails against main on Windows.
• Linux: stock installs carry only UTF-8 locales, where glibc's tolower leaves bytes above 0x7F unchanged, so the host-locale check reports a skip and passes on main too. jotitadev's run with a Latin-1 locale (localedef -i en_US -f ISO-8859-1) fails on main and passes with this PR, so Linux only catches a regression when a single-byte locale is installed.
• tools/check_conventions.py --base origin/main and tools/check_test_registration.py pass. tools/pr_overlap.py and merge tests against the 103 open pull requests that touch these files find no new conflict.
• PS5 hardware and title behavior are untested.

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: none
• Fixes: none
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

[github-actions]

Some comparisons could not be completed because a pull request conflicts with the base branch. Review these shared files manually before concluding that there is no overlap.

Pull request	Shared files not compared
PR 3965	docs/dev/debt/libc.md
PR 5021	core/libs/prx/libc/src/Strings.cpp docs/dev/debt/libc.md
PR 5303	docs/dev/debt/libc.md
PR 6044	core/libs/tests/GuestFnmatch.cpp
PR 6469	docs/dev/debt/libc.md
PR 7081	docs/dev/debt/libc.md
PR 7201	core/libs/prx/libc/src/Strings.cpp
PR 7648	docs/dev/debt/libc.md
PR 7715	docs/dev/debt/libc.md
PR 8150	docs/dev/debt/libc.md
PR 8306	docs/dev/debt/libc.md
PR 8658	core/libs/prx/libc/src/Strings.cpp
PR 8965	core/libs/prx/libc/src/Strings.cpp docs/dev/debt/libc.md
PR 9133	docs/dev/debt/libc.md
PR 9390	core/libs/prx/libScePosixForWebKit/Export.cpp docs/dev/debt/libScePosixForWebKit.md
PR 9408	docs/dev/debt/libc.md
PR 9447	docs/dev/debt/libc.md
PR 9526	docs/dev/debt/libc.md
PR 9527	docs/dev/debt/libc.md
PR 9552	docs/dev/debt/libc.md
PR 9669	docs/dev/debt/libc.md
PR 9929	docs/dev/debt/libc.md

[jotitadev]

Linux run. The fix holds on glibc, but on a stock Linux install the new host-locale check never runs, so guest_fnmatch passes on main too.

Machine: Linux, Ubuntu 26.04, GCC 15.2.0 (glibc), x86-64. Head e52eac7ff merged with git merge --no-commit onto current main d6eafa54d. Full build, BUILD_TESTING=ON, ANYPS5_RELINKER_ONLY=OFF, Ninja Release.

Default locales (C, C.utf8, en_US.utf8): none of the candidate names resolve. CheckCaseInsensitiveHostLocale prints No installed locale changes host byte case folding; host-locale regression skipped. The PR test with main's Strings.cpp and Export.cpp passes. Under a UTF-8 LC_CTYPE, glibc's tolower leaves bytes above 0x7F unchanged, so stock Linux does not hit the bug. On an image with only UTF-8 locales, the test cannot catch a regression on Linux.

With a Latin-1 locale (localedef -i en_US -f ISO-8859-1 <dir>/en_US.ISO8859-1, run with LOCPATH=<dir>, so the test's en_US.ISO8859-1 candidate resolves):

	guest_fnmatch
PR test, main sources	Checking guest byte comparisons with host LC_CTYPE=en_US.ISO8859-1, then Guest byte case check failed (abort)
PR	Passes with the same locale; guest_posix_filesystem, guest_resolver and guest_service_database also pass

So the change also fixes the issue on Linux when the host has a single-byte locale. Today, only a Windows run, or a Linux host with a single-byte locale installed, catches a regression.

Written with AI assistance (Claude Code). I ran the builds and tests above on my machine.

[MarcZX10]

The core change has useful breadth: all 65,536 byte pairs are exercised for the locale-sensitive operations, with unchanged-input checks and Windows code-page coverage. I noted the report of an earlier full CTest run at a prior base with four failures that reproduce on unchanged main and this branch; please rerun the full suite on the newer base before treating overall suite status as current. The focused tests and current-main conventions are author-reported; the latest hosted conventions, eligibility, intake, labels, report, relinker, registration and write-watch jobs pass, while Linux/Windows full jobs are skipped.

[MarcZX10]

I checked the current implementation and the existing Linux report. The strcasecmp paths now use the guest C-locale conversion, while strcasestr uses the existing ASCII-only fold; I found no new code issue. The Windows report covers the locale-sensitive regression, and hosted checks pass, but product Linux/Windows jobs are skipped by path filters. One separate checklist item is stale: the PR targets b27924d, while current main is 47 commits ahead at 8b3a6ba; CONTRIBUTING.md requires a current-main base. The full CTest report is also from a prior base and recorded four failures, so please rebase and rerun the focused locale/string tests (and refresh the suite report) if you decide to move the draft forward. I did not run a project build in this session.

[Noxy-HNR]

Rebased onto main b7d1d47 and removed the blank lines that split the bullet lists in both debt files. guest_strings, guest_fnmatch, guest_locale and guest_rune_locale pass, 4/4, on Windows 11 with GCC 15.2.0; conventions and overlap are clean. Tested is updated, including @jotitadev's finding that stock Linux only catches a regression with a single-byte locale installed. Thanks for the Latin-1 run.

[MarcZX10]

I’ve noticed that asking you to hold this PR was unnecessary. It’s fine to keep submitting PRs; maintainers will handle them in due course.

[MaxLikesCode]

Data point for the expected values here: on #9163, @JJenkx ran the strcasecmp and strncasecmp of the libc.prx bundled with PPSA07923, PPSA05916 and PPSA13934 (comment). Over 200,000 random pairs they returned only -1, 0 or 1, and they fold ASCII only, so strcasecmp("\xc0", "\xe0") is -1. This PR's tests expect FreeBSD 11.4's folded byte difference (for example -'d'). Which module AnyPS5 follows when a title's libc.prx and libSceLibcInternal differ is the open question in #9691.

AI-assisted: yes (Claude Code)