Trails in the Sky 2nd Chapter (PPSA35169)

[JJenkx]

Existing thread

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

Title

Trails in the Sky 2nd Chapter (PPSA35169)

Content version tested

01.003.000

AnyPS5 commit

main 44a59edc3a.

System tested

Linux 7.2.9 (Arch), Ryzen 9 7950X, Radeon RX 6650 XT, RADV Mesa 26.2.4, GCC, Release.

Furthest point

Boots (stops 2 s in, before the first frame)

What is shown

It stops 2 s in, before any frame:

what():  NGS2: the custom PitchShift module is not implemented

I found no open PR or issue for an NGS2 PitchShift module.

Current blocker

libSceNgs2: the custom PitchShift module is not implemented.

How it was run

relinker --registry --rpath <build>/core/libs/libs eboot.elf app.elf

Modules in sce_module are converted with the title. The dump has its PlayGo files, so no PlayGo setting was used.

AI-assisted testing and report: yes, Claude Code.

[JJenkx]

Update: with open PR #11315 (exact head aea8616433), the custom PitchShift refusal is gone. About 2 s in, the title then aborts on NGS2: the custom Reverb module is not implemented, which open PR #10599 implements. The two PRs conflict, so they were not run together. Details: #11315 (comment)

AI-assisted: yes (Claude Code).

[JJenkx]

Update: #11315 has merged. On plain main 60bdb06135 the title now stops 2 s in on NGS2: the custom Reverb module is not implemented, which open PR #10599 implements.

With #10599 applied I tested main 602c0e417c plus #12673 (head c17c71e4d1, which includes #10599), merged in locally. The Reverb refusal is gone. The title then stops in the same second on:

NGS2: voice input 1 is not implemented for this rack

A local diagnostic print at that refusal, which is not part of any PR, shows the patch: a voice of a standard submixer (0x2000) patched to input 1 of another standard submixer (0x2000) that was created with max_inputs 2.

main routes inputs above 0 only for custom submixers (0x4002). fma-rc's reading in #8039 covers custom racks: input N mixes into buffer N, item 15. What input 1 of the standard submixer feeds is still unconfirmed. Earlier notes in #8039 suggest a compressor sidechain. An implementation needs a firmware reading of the standard submixer's second input.

Linux, Radeon RX 6650 XT, RADV Mesa 26.2.6, Release, APS5_PLAYGO_UNLISTED_CHUNKS=installed.

AI-assisted testing and report: yes.

[JJenkx]

Update: map of the NGS2 standard submixer (rack 0x2000) features this title needs. Each item was found by letting the previous refusal through in a private, local-only experiment build. None of these local changes is proposed as a fix.

Base: main 93dbd0f82a plus open PR #10599 (head 1e4443ce43), merged in locally. Each run stopped about 2 s in, exit 134.

1. Second submixer input. A submixer voice is patched into input 1 of another submixer whose rack sets max_inputs 2. main refuses: NGS2: voice input 1 is not implemented for this rack. The title makes 8 such patches.
2. Submixer voice param 0x20000006, 56 bytes. Its words match the PS5 sampler filter layout, read as index 0, location 0, channel mask 0, type 0, frequency 0, Q 1.0, level 1.0. Type 0 turns a sampler filter off, so this looks like the submixer filter being disabled.
3. Submixer voice param 0x20000002, 40 bytes, matching shadPS4's OrbisNgs2SubmixerVoiceCompressorParam (flags, threshold, ratio, knee, attack, release, level). It is sent first with flags 0, then with flags 1: threshold 1.0, ratio 0.01, knee 0, attack 0, release 0, level 1.0.

The run stops at item 3 with flags 1. That is an enabled compressor. Together with the second input, which looks like a side-chain feed, it needs compressor processing whose arithmetic I cannot check without the firmware module. Mixing input 1 into the main mix, as my experiment did to get past item 1, is probably wrong, so I am not proposing it.

What would make this implementable: measurements of the PS5 libSceNgs2.native module, like the ones the existing NGS2 debt entries cite, covering:

• a two-input submixer, with and without the compressor enabled, showing what input 1 contributes to the output;
• the compressor's response to the parameters above.

AI-assisted: yes (Claude Code).

[zaryabzip]

@JJenkx Thanks for the map. I've filed #15090 asking for 13.42 readings of the standard submixer's second input and compressor, based on your three items.