Dying Light: The Beast (PPSA24003)

[DProkos]

Existing thread

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

Title

Dying Light: The Beast (PPSA24003)

Content version tested

01.007.001

AnyPS5 commit

relinker from main a98590dec; system libraries from main 587bb343e + the libSceNpCppWebApi pull request below

System tested

Windows 11 Pro 26200, AMD Ryzen 9 5950X, AMD Radeon RX 7900 XTX (driver 32.0.31041.1004), WinLibs GCC 15.2.0 r7 UCRT

Furthest point

Menu

What is shown

The title screen ("Press ✕ to start") at 3840x2160 presented in a 1484x835 window, steady 59.9 FPS. Intro and title screen render correctly. The title logs Stage = 'InitialScreen' (prev value 'Intro'). Input, audio and anything past the title screen were not checked yet. The first run compiles about 18,000 shaders (about 8.8 GiB of shader_cache/) and takes about 80 s to reach the title screen; with the cache it takes about 20 s.

Current blocker

On main the title stops at start-up before the first frame: the Epic Online Services module (EOSSDK-PS5-Shipping.prx) imports 106 libSceNpCppWebApi exports (UserProfile::V1, ProfanityFilter::V2, CommunicationRestrictionStatus::V3 and their Common template members) that no library provides (FAIL: unresolved ELF import ... from libSceNpCppWebApi.prx). The dump ships no libSceNpCppWebApi.prx. #8561 adds them. Nothing past the title screen has been tried.

How it was run

eboot.bin, sce_module/libc.prx and the nine modules beside the executable (EOSSDK-PS5-Shipping.prx, ffmpeg.prx, runtime_prospero.prx, libNRD.prx, five Ak*.prx) unwrapped with tools/unwrap_self.py, then relinker --windows --registry --module-dir . eboot.elf PPSA24003.exe. app0/ holds the title's files; libs/ the built libraries and the three MinGW DLLs. No environment variables.

AI-assisted: yes

[DProkos]

Update: past the title screen. Same system and build as the first report (Windows 11 Pro 26200, RX 7900 XTX 32.0.31041.1004, main 587bb343e + #8561). Input injected with tools/pad_input.py (APS5_PAD_INPUT_FILE), frames from APS5_DUMP_FRAMES, calls from a --trace-imports build with APS5_TRACE_CALLS and APS5_HANG_DUMP_SECS.

Sign-in loop (needs #6990). Cross on the title screen moves to InitialScreen_ChoosenController, then InitialScreen_SigningIn, and the screen turns black with a spinner. One thread repeats sceSigninDialogUpdateStatus, Close, Terminate, Initialize, Open and sceNpGetAccountIdA (SCE_NP_ERROR_SIGNED_OUT), about 813 times in 50 s, as Dying Light 2 does in #1794. With #6990's library changes merged locally, sign-in finishes within 0.1 s:

Stage = 'InitialScreen_SigningIn' (prev value 'InitialScreen_ChoosenController')
Stage = 'InitialScreen_SigningInFinished' (prev value 'InitialScreen_SigningIn')
Stage = 'InitialScreen_SignedIn' (prev value 'InitialScreen_SigningInFinished')
[Savegame] > ScanForSavegames: saves 0 (prev value -1)

Current stop. After ScanForSavegames the title keeps the black screen with the spinner at 59.9 FPS and logs nothing else; the next stage in the executable is InitialScreen_Complete. Waited 6 min; Cross, Circle and Options presses change nothing. No crash, no FAIL, no unimplemented call.

Ruled out so far:

• The main thread's sceKernelUsleep(100) loop (exe+0x296ba0a) polls the frame's GPU completion label, once per frame; the frame loop keeps running, the stage just never advances.
• Save data, with APS5_SAVEDATA_TRACE=1: mount3 DATA mode=0x1 -> 0x809f0008 (not found), dirNameSearch hit 0, dirNameSearchPs4 hit 0, then no further save-data call.
• NP state: a local experiment reporting SIGNED_IN from sceNpGetState and the state callbacks (reachability still unavailable) leaves the title at the same point. Not proposed as a change.
• HTTP, with temporary logging in libSceHttp: only Epic Online Services makes requests, three non-blocking GET https://api.epicgames.dev/sdk/v1/default?platformId=PSN that fail in sceHttpSendRequest and are deleted; no request is pending afterwards.
• PlayGo reports every chunk local; no libSceMsgDialog or libSceSaveDataDialog call is made.

During the stop the only recurring calls outside libc/libkernel/AGC are once a second sceNpCheckCallback, sceNetCtlCheckCallback, sceSystemServiceGetStatus and sceNpWebApi2CheckTimeout, and the EOS thread's sceRtcGetCurrentTick/sceRtcSetTick. What the title waits for is still unknown.

AI-assisted: yes