[MarrasFederico]

What

Builds on #9896, which implements sceNpMatching2Initialize and sceNpMatching2Terminate. #9896 is merged and this branch is merged with current main, so the diff now holds only this PR's change.

The context calls of libSceNpMatching2 threw. They now follow shadPS4's np_matching2.cpp and its context manager, offline:

• Contexts: sceNpMatching2CreateContext returns NOT_INITIALIZED before any argument check, then refuses a null parameter, a parameter whose size is not 0x28 and a null output with INVALID_ARGUMENT. Ids 1 to 10 are handed out round-robin; an eleventh context is CONTEXT_MAX. sceNpMatching2DestroyContext and the room event, room message and signaling callback registrations answer INVALID_CONTEXT_ID for an unknown context; sceNpMatching2SetDefaultRequestOptParam checks the null option first. sceNpMatching2Terminate now also releases the contexts and restarts the ids.
• Starting: sceNpMatching2ContextStart returns 0x804101E2 (SCE_NET_ERROR_RESOLVER_ETIMEDOUT) after the initialization check. shadPS4 returns it when there is no network and notes that the "error [was] confirmed with a real console disconnected from the internet". No context starts, so sceNpMatching2ContextStop answers CONTEXT_NOT_STARTED for a known context.

Callbacks and the default request option are checked but never used, as no event is generated offline. The room, user, world, server-id and signaling-info calls still throw: sceNpMatching2GetServerId in shadPS4 writes server id 1 from its own server, which is not an answer to copy. Nothing here is measured on a PS5 beyond shadPS4's note; the debt entry says so.

#9874 declares other libSceNpMatching2 exports and creates the same debt file as #9896; pr_overlap.py finds no conflict in Export.cpp.

Tested

Ubuntu 26.04.1, kernel 7.0.0-38, GCC 15.2.0, Release, full build with BUILD_TESTING=ON. Windows was not compiled. No title was run.

• New guest_np_matching2_offline (core/libs/tests/GuestNpMatching2Offline.cpp): NOT_INITIALIZED from every context call before initialization (ahead of the argument checks), the create parameter checks, ids 1 to 10 with a freed id skipped until the allocation wraps, CONTEXT_MAX, the offline start for known and unknown contexts, CONTEXT_NOT_STARTED, the callback and option checks, and termination releasing the contexts and restarting the ids.

• On feat(np): answer and declare the PS4 NP and trophy imports of CUSA03312 #9896's branch the test aborts (sceNpMatching2CreateContext not implemented); with this change it passes, and feat(np): answer and declare the PS4 NP and trophy imports of CUSA03312 #9896's guest_np_ps4_imports still passes. 11 mutations of the new code (parameter size, check order, round-robin, the limit, the offline start before the context check, the stop and option checks, both parts of the termination reset, the context callback's initialization check, destroying) each fail it.

• tools/check_conventions.py --base upstream/main passes, and the branch merges cleanly into current main.

• Updated onto main 15e71dbe11 with merge commit d90537bf06, preserving hashes referenced in the review. Resolved the CMake conflict by keeping both guest_np_common and guest_np_ps4_imports, and retained both Signaling debt sections. Full Release build succeeds; guest_np_matching2_offline, guest_np_ps4_imports, all three guest_np_common cases and guest_np_signaling_offline pass (6/6). Conventions and whitespace checks pass. Windows was not compiled.

• After merging current main (feat(np): answer and declare the PS4 NP and trophy imports of CUSA03312 #9896 included), the full build succeeds, the 34 guest_np tests pass and conventions pass.

• Refreshed overlap check: feat(np): answer and declare the PS4 NP and trophy imports of CUSA03312 #9896 is the declared prerequisite; feat(prx): declare exports for 3 inventory-covered NP libraries #9874 adds different exports but shares Matching2/Trophy debt sections, which must retain both entries on merge. The checker could not compare CMake changes in feat(aot): consolidate native ARM64 compiler, runtime and consumers #5021, test: organize library tests and label CTest suites #6044, Macos port #7201, feat(agc): add opt-in render-state overrides #7275, fix(runtime): isolate host worker MXCSR #7622, fix(savedata): build libSceSaveData and libSceSaveDataDialog from the native sources #7780, feat(savedata): add sceSaveDataMount2 #8285, feat(libSceVrTracker): answer before initialization in sceVrTrackerGetTime, sceVrTrackerGpuSubmit and sceVrTrackerGpuWait #8650, feat(libSceGnmDriverForNeoMode): write the shader register packets and answer the workload and clock queries #9212, feat(libSceGnmDriverForNeoMode): set the embedded fullscreen VS and PS shaders #9221, feat(libkernel): implement the unnamed ksem semaphores #9284, feat(libkernel): implement the named ksem semaphores #9293, feat(libkernel_sys): implement POSIX message queues #9297, feat(libkernel_sys): implement kmq_notify registration #9302 and feat(libSceLibreSsl): implement CMAC with AES-CBC methods #9571. I inspected those CMake diffs; none edits the guest_np_ps4_imports registration.

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

[github-actions]

Progress report (aeeaefa → aff9389)

📈 System libraries: 48.79% (+0.08%, +9 implemented functions)

✅ 9 implemented

Library	Name
libSceNpMatching2	sceNpMatching2ContextStart
libSceNpMatching2	sceNpMatching2ContextStop
libSceNpMatching2	sceNpMatching2CreateContext
libSceNpMatching2	sceNpMatching2DestroyContext
libSceNpMatching2	sceNpMatching2RegisterContextCallback
libSceNpMatching2	sceNpMatching2RegisterRoomEventCallback
libSceNpMatching2	sceNpMatching2RegisterRoomMessageCallback
libSceNpMatching2	sceNpMatching2RegisterSignalingCallback
libSceNpMatching2	sceNpMatching2SetDefaultRequestOptParam

[MarcZX10]

I checked the context lifecycle and offline behavior against the added test: the context set is serialized under a mutex, ids wrap across 1–10 while skipping live ids, termination clears state and resets allocation, and ContextStart consistently returns the documented offline resolver timeout before validating the id. The regression covers initialization precedence, output preservation on failed creation, capacity/wraparound, callback/options checks and reinitialization. git diff-tree --check -r passes on head 4bb91851bbcd5fcda792dd0097697c3a936d162a; current conventions/intake/labels checks pass and product build jobs are skipped. Network-driven behavior and unsupported server calls remain explicitly unimplemented/debt.

[MarcZX10]

I also independently built and ran the exact GuestNpMatching2Offline.cpp with the PR’s libSceNpMatching2/Export.cpp under GCC 12.2 in an isolated Linux harness (NpMatching2 offline tests passed). This covers the context lifecycle/model, not the separate trophy or PS4 import integration tests.

A scope/checklist question before I can complete approval: the current three-commit diff includes offline Matching2 contexts, libSceNpTrophy game/group/trophy behavior, and import declarations for InvitationDialog and Signaling, while the PR title/body center on Matching2. CONTRIBUTING says one topic per PR. Are these all part of one documented CUSA03312 bring-up objective? If so, please identify the relevant Game bring-up discussion in the Unblocks checklist as the template requires; otherwise, please keep the distinct contributions in separate PRs and preserve the declared #9896 dependency/order. I have not run the trophy or PS4 import CTests locally.

[MarcZX10]

Correction to my scope question in #6354074656: I compared the complete commit ancestry with #9896 and confirmed that the PS4 NP imports and Trophy changes are the exact declared prerequisite history, not separate contributions newly bundled into this PR. The Depends on: #9896 declaration is appropriate, so the request to split them or add an Unblocks entry was mistaken; I withdraw it. My independent Linux run remains limited to the new guest_np_matching2_offline test and does not cover the prerequisite’s separate tests.

[MarrasFederico]

@MarcZX10 thanks for the independent context test and for clarifying the prerequisite scope. Updated in d90537b, merging main while preserving the reviewed hashes: both CMake test registrations and both Signaling debt sections are retained. The full Release build and all six focused NP tests pass, including guest_np_matching2_offline and the prerequisite guest_np_ps4_imports. Depends on #9896 remains; Windows was not compiled here.