[Danigamer05]

What

Builds on #16113: this branch contains its commit, so until it merges only the two commits after it are this PR's change.

Export the versioned std::__sce_v2 locale::_Locimp::_Locimp_ctor and _Locimp_dtor from libc, and make the versioned _Locimp family count facet references the way the console does.

Export	NID	Original address
_ZNSt8__sce_v26locale7_Locimp12_Locimp_ctorEPS1_RKS1_	XBQPgR5196I	0xF25E0
_ZNSt8__sce_v26locale7_Locimp12_Locimp_dtorEPS1_	L2pyahLSGFk	0xD1030

Changed exports: _ZNSt8__sce_v26locale7_Locimp11_New_LocimpERKS1_ (ynwqogpIbL0, 0xD0F60) and _ZNSt8__sce_v26locale7_Locimp14_Locimp_AddfacEPS1_PNS0_5facetEm (7KPYZ-jSAto, 0xF2680).

The behavior comes from the libc.prx bundled with PPSA03671 v01.001.005 (J03848509, SHA-256 483961e086b8a1fffa8138b3c0badf798987fb5f35ec908d828946408be2d8a6). As in the module:

• _Locimp_ctor(self, source) builds only the facet vector. The caller has already copied the count. A zero count leaves the vector alone. Otherwise it calls malloc(count * 8), throws bad_alloc on null, and copies the facets from the last to the first, retaining each non-null one.
• _Locimp_dtor(self) releases each non-null facet from the last to the first, calls deleteObject on the ones that reach zero, and frees the vector unless it is the classic one. It does not free the _Locimp.
• _New_Locimp(const _Locimp&) allocates 0x30 bytes and sets the _Locimp vtable, one reference and a null vector. It copies the count, category, transparency and name pointer, then calls _Locimp_ctor. If that throws, it frees the object.
• _Locimp_Addfac grows the vector with realloc to max(id + 1, 40) slots and clears the new ones. It retains the new facet, releases the one it replaces (deleting it at zero) and stores the new one.
• The _Locimp objects built by _New_Locimp(const&), _New_Locimp(bool) and empty() get the module's _Locimp vtable. The last release returns the object, and deleteObject runs _Locimp_dtor and frees it. Before this PR they used the copy vtable and were never freed.

Before this PR, the versioned _New_Locimp(const&) and _Locimp_Addfac called the unversioned copy and insertion, which neither retain nor release. The unversioned functions are unchanged.

guest_sce_v2_locale passed facets with a null vtable to _Locimp_Addfac and read the copy after its last release. It now uses a counting facet and checks that the last release deletes the copy and its facet.

A new debt entry lists what still differs: the lock, null arguments, size overflow, the one-slot classic vector, and copies of locales built by unversioned calls. The versioned locale entry drops its sentence about the reused unversioned copy and insertion, and #16113's entry drops the one about the copy vtable.

Tested

Windows x86-64, WinLibs GCC 15.2.0, Release, BUILD_TESTING=ON, on top of #16113 (d18e1c71325, main d893997f1a6); the branch merges cleanly with main c4469b60daf. Exact tested commit: b24626696c2. It adds one commit to 4e7016e6926 that renames two internal helpers. The build, the tests, the side-by-side run and the conventions check ran on it; the negative control and the overlap checks below ran on 4e7016e6926.

• Built libc, libkernel, their NID-patched production libraries and the related tests. The new guest_sce_v2_locimp_references passes, along with guest_sce_v2_locale, guest_sce_v2_locale_basics, guest_locale and guest_sce_v2_exceptions; public_headers passed on 4e7016e6926.

• The NID-patched libc exports both new NIDs.

• The new test uses facets that count retains, releases and deletes. It checks:

  • the fields _New_Locimp(const&) copies, and that the copy owns its vector;
  • growth to 40 and to id + 1, the cleared slots, replacing a facet and re-adding the same one;
  • the references a nested copy takes, and that the last release deletes the copy and releases its facets;
  • that _Locimp_ctor leaves the vector alone for a zero count and copies only count facets, and that _Locimp_dtor frees the vector and deletes facets at zero;
  • that a _Locimp sharing the classic vector keeps it on destruction and gets a copy of it on growth;
  • that a failed allocation in each call throws the guest bad_alloc, retains nothing, keeps the vector and frees the half-built _New_Locimp object, and that another thread can still take the lock afterwards;
  • that _New_Locimp(bool) and empty() objects are freed with their facets on the last release.

• The original functions were loaded from the bundled module with their relocations applied. Their imports were bound by NID to this branch's libc and libkernel. The module's heap is not set up outside a console process, so its allocation calls were redirected to the host's. Both sides' _Xbad_alloc were redirected to one recorder. The recorder exits by longjmp, so after a failed _New_Locimp(const&) the harness frees the object as the module's landing pad does.

• Side by side with the patched production library, over two seeds with 200,000 random steps each: about 64,000 _New_Locimp(const&), 16,000 _New_Locimp(bool), 88,000 _Locimp_Addfac, 100,000 _Locimp_ctor/_Locimp_dtor pairs on caller-owned objects, and 131,000 retains, releases and destroy calls through the vtable. One step in six made its allocation fail (about 12,000 throws per seed), and about 13,000 _Locimps per seed reached zero and were deleted. Each step compared:

  • every _Locimp's references, count, category, transparency, name and slots;
  • each of 24 facets' references, retains, releases, deletes and destroys;
  • the sequence and sizes of malloc, realloc and free calls.

  The only difference is that the module also calls free(NULL), which the harness ignores. No comparison differed. feat(libc): export the std::__sce_v2 locale::global, empty, classic and _Locinfo category calls #16113's side-by-side run also still passes with two seeds on 4e7016e6926.

• Negative control: five lines were altered one at a time:

  • _Locimp_ctor did not retain;
  • _Locimp_dtor did not free the vector;
  • _Locimp_Addfac kept the replaced facet;
  • _New_Locimp(const&) did not copy the count;
  • _New_Locimp(bool) kept the copy vtable.

  CTest failed on each, and the side-by-side run reported differences each time. The source was then restored byte for byte (SHA-256 checked) and rebuilt, and the tests and side-by-side run passed again.

• Code conventions and git diff --check pass.

• tools/pr_overlap.py read 1,871 open PRs at 4e7016e6926 and reports no overlap. It keeps exit 2 for 36 branches that already conflict with main. Merging each of the 36 with this branch and with main using git merge-tree gives the same conflict blocks, apart from main's own movement since feat(libc): export the std::__sce_v2 locale::global, empty, classic and _Locinfo category calls #16113's base, except in docs/dev/debt/libc.md: for feat(aot): consolidate native ARM64 compiler, runtime and consumers #5021, fix(libc): follow a Windows symbolic link before .. in a guest path #9371, feat(libc): add the std::__sce_v2 thread start, timed wait, clock and processor count calls #9447 and feat(libc): add std::__sce_v2::_Random_device, _Fiopen and setw #9526 the edited _Locimp line and the new entry land inside the conflict block they already have there, as feat(libc): export the std::__sce_v2 locale::global, empty, classic and _Locinfo category calls #16113's lines do, and docs(debt): separate entries with blank lines and drop derived export totals #14212 gets no conflict block beyond the ones it has with feat(libc): export the std::__sce_v2 locale::global, empty, classic and _Locinfo category calls #16113. The debt edit adds no conflict over feat(libc): export the std::__sce_v2 locale::global, empty, classic and _Locinfo category calls #16113 with any of the 142 open PRs that touch that file. No open PR adds these names or NIDs. The branch also merges cleanly with fix(libc): preserve Dinkumware zero signs when no conversion occurs #14595, feat(libc): export the std::__sce_v2 extended-precision helpers #15336, feat(libc): export the std::__sce_v2 multiprecision helpers #15471, feat(libc): export the std::__sce_v2 _Poly and _XLgamma helpers #15484, feat(libc): export the std::__sce_v2 _Xtime_diff_to_millis2 helper #15490, feat(libc): export the std::__sce_v2 _Exp and _FExp helpers #15501, feat(libc): export the std::__sce_v2 _Pmsw, _FPmsw and _LPmsw helpers #15519, feat(libc): throw std::bad_cast and the std::__sce_v2 bad_any_cast, format_error and future_error #15552, feat(libc): forward the std::atomic wait and notify helpers to libkernel #15561, feat(libc): count the std::__sce_v2 _Winit and ios_base::Init initializations #15602, feat(libc): export the std::__sce_v2 floating-point constants #15614, fix(libc): store the std::__sce_v2 _Throw_Cpp_error value in the first code word #15667, feat(libc): export the std::__sce_v2 _Stollx, _Stoullx, _Stolx and _Stoulx parsers #15683, feat(libc): export std::_Throw_bad_any_cast and std::_Throw_bad_node_access #15791, feat(libc): export the std::error_category complete and deleting destructors #15796, feat(libc): export the std::__sce_v2 resetiosflags, setiosflags, setbase and setprecision manipulators #15809, feat(libc): export the std::__sce_v2 _Timespec64_get_sys and _Timespec64_diff_to_ts clock helpers #15834, feat(libc): export the std::__sce_v2 at-thread-exit registry calls #15879, feat(libc): export the remaining std::__sce_v2 facet ids and ios_base _Sync and _Index #15923, feat(libc): export the std::__sce_v2 clog, wclog and standard stream pointers #16083 and feat(libc): export the std::__sce_v2 locale::global, empty, classic and _Locinfo category calls #16113.

Not tested: Linux, macOS, another SDK, a game, GPU or console. No game compatibility claim. No original module, extracted bytes, harness or log is committed or uploaded.

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: feat(libc): export the std::__sce_v2 locale::global, empty, classic and _Locinfo category calls #16113
• Fixes: none
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

🤖 Generated with Claude Code

[JJenkx]

Windows check of head 4e7016e692, merged onto main b3189fd745 (da2e1b4a6f, 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_sce_v2_locale passes in 3 of 3 runs, none skipped.

AI-assisted: yes.

[github-actions]

Progress report (151a2cb → 8c982c6)

📈 System libraries: 27.09% (+0.02%, +10 implemented functions)

✅ 10 implemented

Library	Name
libc	_ZNSt8__sce_v215_sceLibcLocinfoEPKc_nid_postfix
libc	_ZNSt8__sce_v26locale16_SetgloballocaleEPv_nid_postfix
libc	_ZNSt8__sce_v26locale5emptyEv_nid_postfix
libc	_ZNSt8__sce_v26locale6globalERKS0__nid_postfix
libc	_ZNSt8__sce_v26locale7_Locimp11_New_LocimpEb_nid_postfix
libc	_ZNSt8__sce_v26locale7_Locimp12_Locimp_ctorEPS1_RKS1__nid_postfix
libc	_ZNSt8__sce_v26locale7_Locimp12_Locimp_dtorEPS1__nid_postfix
libc	_ZNSt8__sce_v26locale7classicEv_nid_postfix
libc	_ZNSt8__sce_v28_Locinfo13_Locinfo_ctorEPS0_iPKc_nid_postfix
libc	_ZNSt8__sce_v28_Locinfo16_Locinfo_AddcatsEPS0_iPKc_nid_postfix

Changed libraries (implemented exports)

Library	Before	After	Change
libc	1629/1649 (98.79%)	1639/1659 (98.79%)	+0.01%

[TechLevelZero]

❌ AMD GPU test: failed

PR head 4e7016e merged onto main 813f46c · tested 2026-11-05 20:28 UTC · 46 min

GPU	AMD Radeon RX 6300 (RADV NAVI24), driver Mesa 25.2.8-0ubuntu0.24.04.4, Vulkan 1.4.318
Host	Ubuntu 26.04.1 LTS, kernel 7.0.0-42-generic, container ubuntu:24.04
Logs	build and test logs for this run
Build	✅ passed (Release, -DANYPS5_ENABLE_SPIRV_TOOLS=ON)
Tests	❌ 3,978 passed · 18 failed (1 new: 1 not compared, 14 also fail on main, 3 unstable on main) · 71 skipped (ANYPS5_REQUIRE_VULKAN=1)
Test setup	CPU-only tests without a GPU; GPU tests (and any CPU-only test that failed or skipped without one) once on the GPU with VK_LAYER_KHRONOS_validation and ANYPS5_NO_SHADER_CACHE=1; main ran the same way
Test selection	4,067 of 4,097 tests run: the ones this change can reach, chosen by tools/select_tests.py from main's build graph
Vulkan validation	✅ no new validation IDs vs main (1409 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ agc_driver_pipeline_journal 120.4 s, ⚠️ agc_driver_graphics_maps_fallback 113.9 s, ⚠️ agc_driver_graphics 112.0 s, ⚠️ public_headers 63.0 s, ⚠️ agc_graphics_profile_draw 60.2 s

Failing tests

This PR changes test files, so each failing test was also built and run on main with only the PR's test changes applied. ✅ there means the failure comes from the PR's other changes; ❌ means the test fails on main too, so this run does not show the PR caused it.

Test	With this PR	Same test on main 813f46c (main + this PR's test changes only)
guest_sce_v2_locimp_references	❌ failed	not compared: test not defined on main with the PR's test changes

guest_sce_v2_locimp_references: Failed (also failed when re-run alone), assertion and diagnostics

ctest: Failed

2 failures

Unstable on main (3, these also failed in recent main runs on this machine; not counted against this PR)

agc_driver_array_layer_target_cpu: failed in 2 of the last 10 main runs
agc_driver_job_manager_compute_capacity: failed in 9 of the last 10 main runs
agc_driver_two_sample_alpha_to_coverage_sample_mask_export: failed in 6 of the last 10 main runs

Failing on main too (14, not counted against this PR)

agc_driver_alpha_to_coverage_sample_mask_export
agc_driver_guest_memory_lifetime_direct
agc_driver_image_table
agc_driver_native_depth_bda_clear
agc_driver_native_depth_buffer_clear
agc_driver_native_depth_clears
agc_driver_native_multisample_depth
agc_driver_pipeline_journal
agc_driver_submit_progress
agc_graphics_profile_draw
agc_submit_progress
agc_submit_progress_no_timeline
guest_mmap_executable
guest_socket_options_multicast

Validated but not compared (1, not in the main baseline's validation pass)

guest_sce_v2_locimp_references

Validation IDs already present on main (4 IDs in 808 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (808 tests): agc_compute_pipeline_replay, agc_driver, agc_driver_address_draw_space_reuse, agc_driver_address_draw_space_reuse_held, agc_driver_address_draw_space_reuse_off, agc_driver_address_draw_space_reuse_transitions, agc_driver_address_draw_space_reuse_verify, agc_driver_address_passes, and 800 more
VUID-vkCmdCopyImage-srcImage-01548 (11 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_cpu_write_after_memo, agc_driver_depth_plane_cpu_write_after_page_unwatch, agc_driver_depth_plane_cpu_write_after_partial_unwatch, agc_driver_depth_plane_partial_store_image_copy, agc_driver_depth_plane_partial_store_unimported, agc_driver_depth_plane_partially_unmapped_plane, agc_driver_depth_plane_unmapped_plane, and 3 more
VUID-vkCmdCopyImage-srcImage-01551 (11 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_cpu_write_after_memo, agc_driver_depth_plane_cpu_write_after_page_unwatch, agc_driver_depth_plane_cpu_write_after_partial_unwatch, agc_driver_depth_plane_partial_store_image_copy, agc_driver_depth_plane_partial_store_unimported, agc_driver_depth_plane_partially_unmapped_plane, agc_driver_depth_plane_unmapped_plane, and 3 more
VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 (2 tests): agc_driver_prepared_exports, agc_driver_srgb8_color_target

Skipped (71)

agc_driver_bda_reads_shadowed_results: skipped, the device reports watched host memory written when it imports or reads it
agc_driver_bda_reads_shadowed_results_per_build: skipped, the device reports watched host memory written when it imports or reads it
agc_driver_copied_writer_label_order: skipped, the device reports a minStorageBufferOffsetAlignment of 4, so the dispatch binds its record at +0x1004 in place and nothing is written back; the test needs an alignment above 4
agc_driver_copied_writer_label_order_drain: skipped, the device reports a minStorageBufferOffsetAlignment of 4, so the dispatch binds its record at +0x1004 in place and nothing is written back; the test needs an alignment above 4
agc_driver_float16_unused_nan_payloads: skipped, the device has no native f16 conversions with RTE rounding and kept f16 denormals
agc_driver_gcn_display_present: skipped, no display: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
agc_driver_image_table_vouched_lookups: skipped, the device reports 4 of 4 host-imported pages written after a GPU read, so a watched imported table cannot be vouched for
agc_driver_linear_target_direct_write_completion: skipped, the device does not keep host imports write-watched, so no completion stamp over the target is tracked
agc_driver_ngg_vertex_wave_reduction: skipped, s3 describes a 32-lane wave only on 32-lane subgroups, and the device has 64 lanes
agc_driver_recorded_draw_staging: skipped, the device does not keep host imports write-watched, so the target is never resident
agc_driver_scattered_runs: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_sparse_dma_buf_import: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_unit_shadow: skipped, the device does not import watched host memory for unit shadows
agc_driver_unit_shadow_budget: skipped, the device does not import watched host memory for unit shadows
agc_driver_vertex_elected_atomic_max: skipped, the device runs vertex programs in subgroups, so they keep their cross-lane reductions
agc_host_import_sparse: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_idle_publish_wait: skipped, the device reports 4 of 4 host-imported pages written after a GPU read, so a forced watch of the import reads its submissions as CPU stores
agc_import_settle_initial_dcc: skipped, the device has no deferred settle of host imports: imports under 1 MiB wait for one only on Windows
agc_shadowed_key_stores: skipped, the device imports no watched host memory under a unit shadow
agc_small_tiled_shadow_label: skipped, host imports are unavailable: the test block is not imported and watched
agc_storage_clear_hand_over: skipped, the device imports no watched host memory for the clear-key hand-over
agc_storage_pending_key_refresh: skipped, refreshes over a pending key store are not testable here
agc_unit_shadow: skipped, the device does not import watched host memory for unit shadows
agc_unit_shadow_completion_store: skipped, the device reports watched host memory written when it imports or reads it
guest_http_uri_locale: 
guest_net_ctl: skipped, the host has no IPv4 address
guest_net_ctl_disconnected: skipped, the host has no IPv4 address
guest_video_out_color_space: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_device_capability: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_hdr: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_high_frame_rate: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_host_resolution_disabled: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROL

Reproduce

git fetch origin 4e7016e6926d6cb2d41de9956b004d9e4308ecf1 813f46cd685f49aafa9ccac7ca7d65b9179b4974
git checkout 813f46cd685f49aafa9ccac7ca7d65b9179b4974 && git merge --no-ff 4e7016e6926d6cb2d41de9956b004d9e4308ecf1
git submodule update --init --recursive
cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DBUILD_TESTING=ON -DANYPS5_ENABLE_SPIRV_TOOLS=ON
cmake --build build && cmake --build build --target libs
VK_DRIVER_FILES=/nonexistent/icd.json ctest --test-dir build -E '^(agc|shader|recompiler|videoout|video_out|guest_video_out|vulkan)|gpu' --timeout 120 --parallel 12 --output-on-failure
ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1 VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation \
  ctest --test-dir build -R '^(agc|shader|recompiler|videoout|video_out|guest_video_out|vulkan)|gpu' --timeout 120 --parallel 16 --output-on-failure
# failures are then re-run alone: --timeout 360 --parallel 2; new validation IDs: the test alone, twice, --parallel 1
# the failing tests, on this merge and on main:
ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1 VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation \
  ctest --test-dir build -R '^(guest_sce_v2_locimp_references)$' --parallel 1 -V

_{Automated build and test on a contributor's AMD machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs.}

[oneandonlydean]

Hi @Danigamer05, the new locimp-references test failed on Linux: the last two checks (_New_Locimp(bool) and empty() use the owning vtable) read copy->base.vtable after deleteObject had already freed copy, so they compared against freed memory. I pushed a small fix-up that saves the vtable pointer before the free; with it all three guest_sce_v2_loc* tests pass here. No change to the library code.

[Danigamer05]

Thanks for the fix-up, I also carried it into #16178.