Monster Hunter Wilds (PPSA24191)

[Lindamea1]

Existing thread

• I searched this category for the title ID and no thread exists for this title on any platform.

Title

Monster Hunter Wilds (PPSA24191)

Content version tested

01.041.040

AnyPS5 commit

main 83bf557 with #7852 merged on top

System tested

Gentoo Linux (kernel 7.2.8), GCC 16.2.1, NVIDIA GeForce RTX 5090 (driver 595.99.02)

Furthest point

Stops before the first frame

What is shown

The title compiles its shaders, presents frames at about 5 to 10 fps and runs its save-data checks within a few minutes, but every presented frame is one flat dark grey (0x0c0c0c). Audio and input were not checked (audio driver set to dummy, no pad input).

Current blocker

• On main it stops at load, on three imports no library exports. feat(libs): add the Voice and IME exports PPSA24191 needs to load #7852 adds them; its description has the log lines and the runs.
• At Linux's default load address the title's own libc.prx rejects its first mspace buffer, as PRAGMATA (PPSA02530 v01.200.000) on Windows: the title's libc rejects its allocator's mspace buffer in the ASLR-relocated image above 1 TiB, so the first malloc returns NULL and start-up faults #3043 describes for PRAGMATA on Windows, and start-up fails. Relinking with --linux-fixed-base 0x400000000 (or 0x400000) gets past it (details in feat(libs): add the Voice and IME exports PPSA24191 needs to load #7852).
• With that, every frame is flat dark grey. The log has three queue 0x20 WAIT_REG_MEM_64 ... timed out after 1000ms (function 3 ref 0x63 ...) lines, with the awaited 64-bit word holding 1. Whether they are related to the grey picture is not established.

Open pull requests for this title:

• feat(libs): add the Voice and IME exports PPSA24191 needs to load #7852: the three loader imports (needed to start).
• fix(libkernel): return ESRCH from scePthreadAttrGet for a thread that has ended #7957: scePthreadAttrGet answers ESRCH for an ended thread. The title reuses an entry of its 64-entry thread table only on that answer. A 6-minute run filled 44 entries.
• fix(libSceNet): fill the address and port fields in sceNetGetSockInfo #7959: sceNetGetSockInfo fills the endpoint fields the title reads after bind, connect and accept. This is from the executable's code; an offline run does not reach it.
• fix(libSceSaveData.native): count PS5 save data in 64 KiB blocks, at most 16384 #7960: save data is counted in 64 KiB blocks, as the title's own size arithmetic expects.

How it was run

relinker --registry --module-dir PS5 --linux-fixed-base 0x400000000 eboot.elf MHWilds.elf

Run headless in gamescope (--backend headless), offline and not signed in to PSN, with SDL_AUDIODRIVER=dummy.

AI-assisted: yes (Claude Code).

[Lindamea1]

Update on main acda0b06f (same machine: Linux, RTX 5090 with NVIDIA 595.99.02; relinked with --linux-fixed-base 0x400000000; #7852 is merged).

• Rendering works now. With a warm shader cache the title shows the RE ENGINE logo, the Health Warning, creates its save data ("No saved data exists. Creating new save data."), and shows the Before You Play / EULA screens. The first run after a recompiler change spends about five minutes compiling shaders before anything appears.
• On NVIDIA it then stops on stencil reference export emulation requires a replay-safe pixel shader with late tests (this GPU has no VK_EXT_shader_stencil_export). What that shader does and why it is refused is in my comment on fix(recompiler): replay stencil exports of pixel shaders that only read through BDA #8145; fix(recompiler): replay stencil exports of pixel shaders that only read through BDA #8145 removes one of the reasons, not all of them.
• Past that, in local experiments only: the replayed draw hits a push-constant bug that is on main itself (fix(agc): put a stencil replay shader's guest buffer offsets where its shader data is #8615), then the replay's depth feedback check (also described on fix(recompiler): replay stencil exports of pixel shaders that only read through BDA #8145), then, at the language selection of a later boot, sample-preserving guest publication of native multisample depth is not implemented.
• Not checked: an AMD GPU (which has the extension, so the stencil replay path does not run), audio, and input beyond pressing Cross.

AI-assisted: yes (Claude Code). I ran the title on my machine.