[MarcZX10]

Fixes #5095.

What

On Windows, CreateFileW rejects TRUNCATE_EXISTING with GENERIC_READ alone, so open(path, O_RDONLY | O_TRUNC) fails instead of truncating. For that flag combination, open with the access needed to truncate, then narrow the resulting handle to read-only. This matches the FreeBSD/Linux behavior while keeping the guest descriptor read-only. The implementation records the remaining Windows host sharing difference in technical debt.

The three commits carry David’s implementation, regression, and debt update with his original commit authorship preserved.

Tested

• Linux, GCC 12.2, Release: build succeeded; guest_filesystem and guest_kernel_fcntl passed (2/2).
• tools/check_conventions.py --base upstream/main passed.
• Windows-specific regression matrix is documented in libkernel: open with O_RDONLY|O_TRUNC fails with EINVAL on Windows, where FreeBSD truncates the file; fix branch ready #5095; it was not rerun locally.

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-path policy: not applicable; no new unimplemented path
• No notes, investigation files, or images added
• No third-party code added
• Generic behavior, not title-specific
• Depends on: none
• Unblocks: none identified in libkernel: open with O_RDONLY|O_TRUNC fails with EINVAL on Windows, where FreeBSD truncates the file; fix branch ready #5095
• GPUs tested: n/a
• AI-assisted: yes

[github-actions]

Other open pull requests touch the same code. Merging one will break the other, or both implement the same thing. If this one builds on them, list them in Depends on.

Pull request	Conflicting files (lines on main)	Same exports	Same new files
#5434	core/libs/tests/GuestFilesystem.cpp L232, L235

[MarcZX10]

Closing this duplicate: it carries the same Windows O_RDONLY|O_TRUNC fix already reviewed in #5112. The maintainer closed #5112 because no title-specific failure was identified and asked contributors to hold generic PRs while the queue is high. I should have found that history before opening this PR.