[Noxy-HNR]

What

FT_Stroker_GetCounts and FT_Stroker_GetBorderCounts create output write guards but ignore refusal. They now check both non-null destinations before writing either count, so a read-only guest output produces a controlled error without partially changing the other output.

The native query still runs first. Writable destinations retain FreeType's zero-on-error writes, optional-null outputs and points-before-contours alias order. A refused output preserves an existing native error; otherwise it returns FT_Err_Invalid_Argument. This is host-boundary policy, not measured console behavior, and is recorded in debt. Ordinary write-watch CPU stores are not the issue addressed here. Native output order follows the bundled ftstroke.c.

Tested

Current-head hosted verification (8024e45a47): the Windows changed-test job passed; the existing stroker test and both new raw/NID output regressions passed. The job reports zero failures among 3,344 selected CTest entries, including skipped entries. Its log identifies this exact PR head. This is not a fresh full Linux run; earlier local and Linux results below retain their stated commits and scope.

Windows x86-64, WinLibs GCC 15.2.0, Ninja, Release, full project configuration and source-built dependencies.

• New raw and NID-patched runtime tests compare empty, completed and unfinished strokers with bundled FreeType, including both borders, invalid borders, null strokers, optional outputs and aliased outputs.
• Two guest direct-memory pages are protected independently and together. The tests check whole-buffer preservation on refusal and compare results after restoring writes. A host-writable/logically-read-only mode gives an assertion-based negative control; actual PAGE_READONLY is confirmed with VirtualQuery for the protected cases.
• With the unchanged implementation at c55222cfc4 (new tests on main f5844034d9), both regressions fail by assertion at the refused-output result check.

Final validation at 8024e45a47, including main 3cb1cd29ec:

• Built the complete libs target and both focused test executables. guest_freetype_stroker, guest_freetype_stroker_outputs and guest_freetype_stroker_outputs_nid passed 3/3 in each of three runs.
• All five deliberate-break controls made both output-test variants fail by assertion: removing guard refusal, writing points before checking the second output, skipping native error zeroing, reversing aliased writes, and replacing native-error precedence with unconditional Invalid_Argument. Restored sources were rebuilt and both variants passed again.
• Public-header audit: 406 libraries / 30,649 exports verified. Conventions, strict C++20 test syntax and whitespace checks pass.
• tools/pr_overlap.py --repo boykopovar/AnyPS5 --base main --branch fix/freetype-stroker-output-guards scanned 1,886 open PRs. It returned exit 2 for 14 branches that conflict with its base. I inspected their shared CMake/debt changes directly: feat(libSceFreeType): query glyph control boxes #9588, feat(libSceFreeType): implement FT_Atan2 and the FT_Vector polar calls #10600, feat(libSceFreeType): implement FT_Get_Sfnt_Table and FT_Outline_Get_Bitmap #10601, feat(libSceFreeType): implement FT_Open_Face for memory and path sources #10604, feat(libSceFreeType): implement FT_Done_Library and FT_Reference_Library #10606, feat(libSceFreeType): implement FT_Outline_Render #10609, feat(libSceFreeType): implement FT_Set_Renderer #10611, feat(libSceFreeType): implement the variation and multiple-master calls #10613, feat(libSceFreeType): implement FT_Get_PS_Font_Value and FT_Attach_File #10616, feat(freetype): implement font table validation #10897, feat(libSceFreeType): implement the legacy memory allocators #12004, feat(libSceFreeType): implement the legacy reallocation helpers #12373, fix(relinker): match AppContent utility import to provider #13486 and docs(debt): separate entries with blank lines and drop derived export totals #14212 concern other APIs or formatting. None changes these two Stroker count APIs or adds these new files. Combined integration with those branches is untested. The current-main merge preview is clean; there is no functional prerequisite.

No Linux, console or retail-title run.

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

[JJenkx]

Windows check of head 8024e45a47, merged onto main 7015a659ca (15a416b77b, clean merge).

Setup: Windows 11 Enterprise 22621, WinLibs GCC 15.2 UCRT, Release. The hosted Windows check does not run these tests on pull requests.

• guest_freetype_stroker_outputs and guest_freetype_stroker_outputs_nid pass in 3 of 3 runs, none skipped.

AI-assisted: yes.