[Ondra97]

What

Json2 array parsing constructs each value before linking it into its owning list. If list-node allocation fails, destination cleanup cannot reclaim that value. On the current main, parsing [1] with allocation budget 2 retains one allocation after cleanup.

Construct each array entry directly in its owning list so existing partial-tree cleanup owns it before further allocation. Current main already handles the object-entry failure paths from the original patch; retain its ordered object insertion, duplicate-key replacement and incremental parser behavior.

Extend allocation-failure coverage for valid and malformed documents, nested containers, long keys, callback reads and missing-key insertion. Preserve upstream iterator-copy and swap regressions. Check destination values, entry identity, container sizes and mismatch handlers after failures. Related to #8005, finding C9; no title, OOM or performance improvement is claimed.

Tested

Rebased on a6dcea9279b53b3a39c13695eac66f7c8b5b503c; tested head 1f0a363b503d5e689bae3a5667060b281a70fe9c.

Windows 11 Pro 10.0.26300, WinLibs MinGW-w64 GCC 15.2.0r7 UCRT, Release, full build configuration with BUILD_TESTING=ON. CPU: AMD Ryzen 9 7900X; RAM: 32 GB; GPU: NVIDIA GeForce RTX 4070 Ti SUPER.

The final allocation-failure regression fails with unchanged current-main production code at parse array budget=2: operation leaked allocations. With the fix it covers 37 operations and 549 forced allocation failures with no tracked leaks. The parser checks use current-main error codes, callback error propagation and acceptance of trailing input.

The production build (libs, relinker, patched-library fixture and both Json2 fixtures) passes. All three registered runtime tests pass without skips (3/3, 0.66 s). Conventions, main-sync and whitespace checks pass. The overlap checker flags conflicting #7215 and this PR itself; manual review confirms #7215 changes numeric locale conversion and does not implement allocation ownership.

cmake --build build --target libs relinker patched_libraries_tests guest_json_tests guest_json_copy_failure_tests --parallel 8
ctest --test-dir build -R "^(guest_json|guest_json_copy_failure|patched_libraries)$" -V --output-on-failure --no-tests=error
python -B tools/check_conventions.py --base origin/main
python -B tools/check_main_sync.py --base origin/main
python -B tools/pr_overlap.py --branch codex/fix-json2-parse-allocation --repo boykopovar/AnyPS5 --jobs 3
git diff --check origin/main...HEAD

The default all-target build, complete CTest suite and game/title runs were not performed on this rebased head.

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

[Ondra97]

Could someone help establish the PS5 Json2 allocator/initialization contract for the separate C2 follow-up in #8005? This is a contract question for that follow-up; the implementation in this PR covers allocation-failure ownership.

On tested revision c9f680fb, a synthetic parse/serialize probe performs 21 tracked host C++ allocations and makes zero calls to the supplied allocator's allocate/free/notify callbacks. Both the normal and null-returning callback settings still give successful initialize/parse results. Since no callback runs, this confirms ignored configuration but does not validate the candidate callback ABI.

I need to establish:

• The allocator virtual-method signatures and alignment requirements, allocation-failure reporting, and how the RTTI allocator variant differs.
• Whether initialization state is global or per Initializer, and the required allocator/object lifetime across repeated initialize/terminate calls and reinitialization with live Values.
• Which parser overloads consume fileBufferSize, including what a zero size means.
• The allowed SpecialFloatFormatType values and their behavior, especially value 2 in the supplied initialization configuration.

An API reference or results from a small PS5 test covering these cases would help me implement the general behavior and add a regression that fails before the fix and passes afterward. The synthetic probe uses no title assets.

[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 9826	core/libs/prx/libSceJson2/Export.cpp