[milescoviello]

What

Guest path resolution trimmed the trailing slash of a path under a mount alias (/savedata0, /download0, /app0 and the others), so the host saw file for file/. On main, with a regular file f under an alias:

call on <alias>/f/	main	FreeBSD 11.4, and the same path outside an alias
unlink, remove, sceKernelUnlink	0, the file is deleted	ENOTDIR
rename("f/", "g")	0, the file is moved	ENOTDIR
open(O_RDONLY)	a descriptor for the file	ENOTDIR
chmod, access	0	ENOTDIR

FreeBSD 11.4 rejects it in lookup() (sys/kern/vfs_lookup.c):

	/*
	 * If we're processing a path with a trailing slash,
	 * check that the end result is a directory.
	 */
	if ((cnp->cn_flags & TRAILINGSLASH) && dp->v_type != VDIR) {
		error = ENOTDIR;
		goto bad2;
	}

#2300 made stat follow this; the other calls still acted on the file because ResolveAlias drops the slash before the host sees it, while a path outside an alias keeps it.

On Linux and macOS a path under an alias now keeps its trailing slash on the host path, exactly as paths outside an alias already do, so the host fails with ENOTDIR for a file while directories with a trailing slash (stat, mkdir, rmdir) keep working. Windows keeps the previous behaviour, since its trailing-separator handling differs (#4513) and I could not test it; an entry in docs/dev/debt/libc.md records that.

Tested

Ubuntu 24.04, GCC 13.3 (CI's compiler), Release with CI's configure flags, on main 1acfa7f. guest_filesystem gained checks under its parent-test-mount alias: unlink, remove, rename, open and sceKernelUnlink on file/ fail with ENOTDIR and every file is still there, stat of Existing/ reports a directory, and mkdir/rmdir of slash-directory/ work.

• Red: with main's General.cpp the test fails at the first new check (unlink returns 0 and deletes the file).
• Green: guest_filesystem passes, also in 30 repeated runs; all guest_* tests except guest_video_out_* pass (310/310, two runs).

Windows: @johnmaia built and ran it on Windows 11 with WinLibs GCC 15.2.0, merged into main 3faa9cb. The full suite passes 1074 of 1074, guest_filesystem included. The new checks are skipped there, so this shows only that nothing else changes on Windows.

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 in TechnicalDebt
• 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
• Unblocks: none
• AI-assisted: yes

🤖 Generated with Claude Code

[johnmaia]

Windows build check. Setup: Windows 11 (26300), WinLibs GCC 15.2.0 r7, Release, BUILD_TESTING=ON, ctest -j4 --timeout 300, RTX 4080. Head 516e90d merged into main 3faa9cb.

It builds, and the full suite passes 1074 of 1074, guest_filesystem included. The new trailing-slash checks are #ifndef _WIN32, so this only shows that nothing else changes on Windows.

AI-assisted: yes (Claude Code)

[adriadam10]

Checked the build and scoped ctest on the PR merged onto main, guest_filesystem repeated 5 times, and that the new checks fail with ResolveAlias reverted to main. The Windows runs by @johnmaia and @xXPBSXx confirm nothing else changes there. Merging after the full Build is green.

[milescoviello]

Rebased onto main 34355ee. The only conflict was docs/dev/debt/libc.md, next to the new tzname entry; both entries are kept. After the rebase, the build is clean, guest_filesystem passes 5/5 and all 354 guest_* tests pass (Ubuntu 24.04, GCC 13.3, Release).

AI-assisted: yes