[reyani-dev]

What

On FreeBSD, stat("file/") fails with ENOTDIR: a trailing slash asks for a directory. On main:

• Windows: std::filesystem::exists("file/") is true (Win32 ignores the slash), then _wstat64("file/") fails, and File::FillFileStat throws FillFileStat: stat failed for ..., which aborts the title.
• Linux: stat reports ENOTDIR, which std::filesystem::exists reads as "not found", so the guest got ENOENT.

sceKernelStat now takes the std::filesystem::status once and returns ENOTDIR if the host reports not_a_directory, or if the guest path ends in / and the target exists but isn't a directory. A missing path is still ENOENT, and stat("dir/") still succeeds. stat (the POSIX wrapper) goes through sceKernelStat, so both are covered.

The same trailing slash on other calls returns the wrong error without aborting: open("file/") and unlink("file/") give EINVAL instead of ENOTDIR on Windows. open is being reworked by several open PRs (#1628, #1777, #2144, #2170), and unlink is in my #2289, so I kept this to stat.

Tested

Windows 11, WinLibs GCC 15.2.0 r7, Release, -DBUILD_TESTING=ON:

• New checks in guest_filesystem_tests, next to the other stat checks:

  • stat("present.txt/") returns -1 with errno 20;
  • sceKernelStat("present.txt/") returns 0x80020014;
  • stat("<dir>/") succeeds and reports a directory.

  On main the test process aborts with FillFileStat: stat failed.

• ctest -E "^agc_|video_out": 249/249 passed, including guest_path_case, which also uses sceKernelStat.

• tools/check_conventions.py passes, and tools/pr_overlap.py --branch reports no overlap. It merges cleanly with the open PRs that touch Open.cpp and GuestFilesystem.cpp (fix(kernel): return guest errors from scalar file operations #2144, fix(files): isolate guest descriptors from host resources #2170, fix(libkernel): honour open flags when Windows falls back to a directory descriptor #1628, feat(libkernel): open /dev/random and /dev/urandom for reading #1777, fix(libkernel): give flock BSD semantics on Windows #1755, fix(libkernel): return EWOULDBLOCK from a contended LOCK_NB flock #1764, feat(libs): implement font, libc, libkernel and misc prx stubs #2193, fix(libkernel): report Linux file birth times accurately #2115, feat(libkernel): implement lstat #1624, fix(libkernel): report EISDIR for path truncation #2086, feat(libkernel): implement fdatasync #1620, feat(libkernel): implement sceKernelFdatasync #2129) and with my fix(libkernel): report POSIX rename errors from sceKernelRename #2260 and fix(libkernel): return EPERM when unlinking a directory #2289. fix(linux): resolve loader and APR failures exposed by HFW #1616 already conflicts with main on its own.

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
• AI-assisted: yes

🤖 Generated with Claude Code

[reyani-dev]

Full Linux and Windows builds and tests for this head (d99a6ceb), using the project's Build workflow. I can't start the upstream manual run, so I ran the same build.yml in my fork through a mirror PR. Like the upstream run, the jobs check out the PR head merged into current main (d70b8998):

Job	Result
linux	success; full build, ctest: 516/516 passed
windows	success; full build, ctest: 507/507 passed
relinker (Linux, Windows, macOS)	success

Run: https://github.com/AchrafReyani/AnyPS5/actions/runs/37948551554

[reyani-dev]

Full Linux and Windows builds and tests for this head (d99a6ceb), using the project's Build workflow. I can't start the upstream manual run, so I ran the same build.yml in my fork through a mirror PR. Like the upstream run, the jobs check out the PR head merged into current main (d70b8998):

Job	Result
linux	success; full build, ctest: 516/516 passed
windows	success; full build, ctest: 507/507 passed
relinker (Linux, Windows, macOS)	success

Run: https://github.com/AchrafReyani/AnyPS5/actions/runs/37948551554

[adriadam10]

Checked on Linux merged onto current main: guest_filesystem aborted at the lstat checks, because the new directory check overwrote status, which the lstat checks (added by #1624) reuse. I pushed one commit that uses a separate FileStat dirStatus for it. After that, build and the scoped ctest (73 tests) are green, and the new stat checks fail when the Open.cpp change is reverted.

• Windows path (_wstat64) not run here; relying on your fork run, which predates feat(libkernel): implement lstat #1624.
  Merging once CI is green.

[reyani-dev]

Thanks @adriadam10, the separate dirStatus is right.

I checked the Windows (_wstat64) side with your commit (c1b714f) merged onto current main 03bf8ff (Windows 11, WinLibs GCC 15.2.0, Release):

• guest_filesystem passes 3/3 (--repeat until-fail:3), including the new trailing-slash checks and the lstat ones;
• ctest -E "^agc_|video_out" passes 335/335.

A fresh full Linux and Windows build of this head is running and will be posted here.

[reyani-dev]

Full Linux and Windows builds and tests for this head (c1b714f2), using the project's Build workflow. I can't start the upstream manual run, so I ran the same build.yml in my fork through a mirror PR. Like the upstream run, the jobs check out the PR head merged into current main (03bf8ff3):

Job	Result
linux	success; full build, ctest: 628/628 passed
windows	success; full build, ctest: 636/636 passed
relinker (Linux, Windows, macOS)	success

Run: https://github.com/AchrafReyani/AnyPS5/actions/runs/38031421856