New Super Lucky's Tale (PPSA34823)

[JJenkx]

Existing thread

• I searched this category for PPSA34823 and "Lucky"; no thread exists.

Title

New Super Lucky's Tale (PPSA34823)

Content version tested

The extraction's eboot.bin (Unity, IL2CPP), with PlayGo files present.

AnyPS5 commit

main 474030e584, with no extra PRs. Four open PRs were each merged onto the same main for the comparison below.

System tested

Linux 7.2.11 (Arch), Ryzen 9 7950X, Radeon RX 6650 XT, RADV Mesa 26.2.6, GCC, Release.

Furthest point

None: it stops before the first frame, at 2 s, exit 134.

What is shown

compute shader 0x21ecb9200: AGC graphics: a shader buffer access of 0x21fac1000+0x28000 overlaps multisampled color target 0x21fac1000+0x28000 (8 samples), whose samples are only on the host: guest access to it is not implemented

With #7002 applied, the same stop names the access as a write: a shader buffer write of 0x21fac1000+0x28000 …. The compute shader's buffer covers exactly the 0x28000 bytes of the 8-sample color target.

Current blocker

A compute buffer write over a multisampled color target whose samples exist only on the host. These PRs, each merged alone onto the same main, give the same stop:

PR	Merge
#7002, publish native color for read-only shader buffers	01ec1d7a39
#13248, claim a multisampled color target that a single-sample storage image store overlaps	0fc12fa21a
#13963, store storage results pending under a multisampled color target before it takes the memory	2fbcc57b1f
#10049, retire a multisampled color target a write-only storage image overwrites	d538a6004c

Whether this write replaces the target's memory or reads it as well is not established from this run.

How it was run

relinker --registry --rpath <build>/core/libs/libs --module-dir Media/Modules --module-dir Media/Plugins eboot.elf app.elf
./app.elf

AI-assisted testing and report: yes.

[JJenkx]

Linux follow-up: PR #10893 (bffb98b42b) advances PPSA34823 01.000.001 from the early compute-buffer/native-8x-MSAA abort to the rendered title menu. Main 0836dfdf4fe exits134 at2.0s; current main 9fe758d7520 plus #10893 shows readable Start Game and Import PlayStation4 save data text. AMD RX6650XT / RADV26.2.6, isolated runtime, unchanged original asset hashes.

The matched GPU regression gate verifies register-allocation variants, bounded patterned writes and full clears of drawn 4x/8x native MSAA targets. Main fails the new recognition case, the change passes; actual Vulkan validation was inserted, no skips or VUIDs. Full qualification comment includes the reusable supplemental test patch and exact integration hashes.

Remaining blocker: after Cross activates Start Game, the native run exits139 at21.1s while loading the intro scene bundle. A separate debugger run reproduces a guest-code SIGSEGV; its cause is still under investigation. Gameplay/audio/performance and later scenes are not established.

Evidence archive (SHA-256 e897c6d8553a2a3c606b8aba0cc20f9c494130b143c71dd97f2c64f649b069bb)

Reliability follow-up (2026-10-31): a repeat using the same frozen parent SDK exits139 at6.1s before Start Game; a private copy-back candidate (#6185) also stops at6.1s. This does not establish a regression or fix in #6185. The earlier title-menu screenshot is a directly observed result, but reaching that menu is not yet reliable. A debugger stopped at the guest allocator entry with a zero size header before the free modified it; the first corrupting writer and any SDK cause remain unknown. All original input hashes and both SDK banks were reverified unchanged.