Mirror of github.com/cira0100/Eglantine-public nx2, a fork of Djihads80/Eglantine. https://git.cirakg.xyz/eglantine-cirakg/atom?h=nx2 2026-11-02T19:31:11+00:00 2026-11-02T19:31:11+00:00 snowstormxd 214507868+snowstormxd@users.noreply.github.com 2026-11-02T19:31:11+00:00 urn:sha1:43490c0417f6ef07a5999c2f8cf1c18e4e4a8520 Give the implementation-local TLS object one definition so ThinLTO retains its dynamic initializer. The state and its lifetime are unchanged. Assisted by Codex. 2026-11-02T19:07:54+00:00 snowstormxd 214507868+snowstormxd@users.noreply.github.com 2026-11-02T19:07:54+00:00 urn:sha1:00285925faebcd9fce6f12b4ebfaccdab491203c Adapted bitmap scans from sylvieistaken/selevet-eglantine performance build 4. Keep the wide-priority fallback and fix the first live slot after empty bitmap words. Include randomized lifecycle and ordering checks. Assisted by Codex. 2026-11-01T20:13:56+00:00 Takoroka 137927095+Takoroka@users.noreply.github.com 2026-11-01T20:13:56+00:00 urn:sha1:76812d0bd86c3dfbef473ad6236a68045ad3d4b5 # Conflicts: # src/common/host_memory.h 2026-11-01T18:26:12+00:00 Takoroka 137927095+Takoroka@users.noreply.github.com 2026-11-01T18:26:12+00:00 urn:sha1:f50b3e2b9d27fd6612dbc72893f6a37c06d77ed1 2026-11-01T18:26:03+00:00 Takoroka 137927095+Takoroka@users.noreply.github.com 2026-11-01T18:26:03+00:00 urn:sha1:7c9870b0e17b36240df0a8015421fc4c8067dd3b 2026-10-31T14:54:52+00:00 Alex 78863999+Sssnowyyy@users.noreply.github.com 2026-10-31T14:54:52+00:00 urn:sha1:ee2d55ee497c9467c8df82515165a21eac13fff5 * game_list: don't free the worker while it is processing events OnPopulatingCompleted runs a nested event loop from inside GameListWorker::ProcessEvents. If the directory watcher fires there, the repopulate destroys the worker while ProcessEvents is still on the stack and the next lock hits a freed mutex. Stop the worker synchronously but free it with deleteLater, and stop ProcessEvents once it has been replaced. Assisted-by: Claude Opus 5.5 (code changes, debugging and commit message) * Fix Switch 2 heap limits and game list reloads during emulation - SetHeapSize rejected heaps of 8 GiB or more (Switch 1's MainMemorySizeMax), so Switch 2 titles that size their heap from the 12 GiB layout (Pokopia asks for 9.5 GiB) got ResultInvalidSize and aborted with 2001-0101. Allow up to the memory the active layout provides. - Grow the 39-bit heap region from 8 to 16 GiB so such a heap fits. - Skip watcher-driven game list reloads while a game runs: Repopulate calls CreateFactories, which replaces content providers the running game reads (segfault in ContentProviderUnion::GetEntry). Stopping emulation already repopulates the list. Assisted-by: Claude Opus 5.5 (code changes, debugging and commit message) * nx2: answer aud:u 1 with an IAudioOutManager nn::audio::OpenAudioOut gets its manager from aud:u command 1 and calls OpenAudioOut through slot 1 of the returned object. The auto stub returned no object, so Pokopia jumped through a null vtable 5 s into boot and never got past its first frame. Assisted-by: Claude Opus 5.5 (code changes, debugging and commit message) * nx2: add a stpl:u (stream play) stub with a user service nn::streamplay::detail::GetUserService opens a session with stpl:u 0 and attaches a system event to the handle command 300 returns. The generic missing-service stub returned no session, so Pokopia's first nn::streamplay call jumped through null right after audio setup. Assisted-by: Claude Opus 5.5 (code changes, debugging and commit message) 2026-10-31T07:57:31+00:00 Takoroka 137927095+Takoroka@users.noreply.github.com 2026-10-31T07:57:31+00:00 urn:sha1:7d82146663c0960cc8d7acece8642c7c4c944e3f The 12 GB layout added 8 GiB of DRAM for memory the Switch 2 graphics server keeps, and the whole emulated DRAM is committed at boot, so Switch 2 games needed 20 GB of free commit ("Failed to allocate 20480 MiB of backing memory" on PCs with 16 GB). With GPU-only regions backed as they are written and sparse regions taking tiles from the game's memory, Mario Kart World keeps ~1.1 GiB of the pool (texture pools it fills at boot, flat afterwards) and Splatoon Raiders none. 2 GiB leaves room and brings the commit to 14 GB. A title that runs the pool dry loses GPU writes into its unbacked blocks and logs it once. 2026-10-30T16:29:40+00:00 Takoroka 137927095+Takoroka@users.noreply.github.com 2026-10-30T16:29:33+00:00 urn:sha1:36d9276b1ae582b2a941659b35ad684d0907a189 With EDEN_NX2_SVC_TRACE, WaitSynchronization now logs its handles and, for events services created, the names they gave them (a title that never draws usually polls one: Minecraft waits on AM's message event). 2026-10-30T08:40:24+00:00 Takoroka 137927095+Takoroka@users.noreply.github.com 2026-10-30T08:39:40+00:00 urn:sha1:f260bcb5a26a6b6b9675b41969a273858f024c55 The kernel clears every page it hands out, and the backing memory is a page-file section the host only gives RAM to once a page is touched. So the clears themselves made every page of a game's heap resident: Mario Kart World (12 GB layout) had a 12.5 GB working set 20 s after boot. HostMemory keeps two bits per 64 KiB of backing: used (handed out by the kernel at least once) and dirty (may hold data). Pages returned to the kernel mark their granules dirty, and a zero clear only touches dirty granules; a granule cleared in full is clean again. A new kernel on the same memory marks everything a previous one used dirty, so leaked processes can't leak data into the next game. KMemoryManager's optimized-process fill and KSharedMemory's clear go through the same path. Mario Kart World, title screen and attract demo, 180 s: working set 20 s in 4.7-6.4 GB instead of 12.5 GB, peak 14-17.6 GB instead of 20.1 GB. 2026-10-30T04:54:37+00:00 Takoroka 137927095+Takoroka@users.noreply.github.com 2026-10-30T04:54:09+00:00 urn:sha1:d22a03efcd0aeddaa7d0d2ca47bd345f6ecf7752 Loading a Switch 2 game after another one in the same window hung or crashed it at boot. The first game's process still had 55-60 references at kernel shutdown, so it was never finalized and its memory stayed mapped in the fastmem arena. The next game's mappings at the same addresses then failed ("Failed to split placeholder" / "Failed to map placeholder"), and it ran on the first game's memory: Mario Kart World loaded after another Switch 2 game jumped to a garbage address from rtld, or hung. Detach the closed process from the arena (a late finalize must not unmap the next game's memory), unmap the whole arena, and log when the process was still referenced.