[XairRrR]

What

Agc::Command::WriteRegisters, behind sceAgcCbSetShRegistersDirect, sceAgcCbSetUcRegistersDirect and the sceAgcDcbSet{Cx,Sh,Uc}RegisterDirect writers, threw empty register list on a count of 0, although the matching ...GetSize functions already report 0 bytes for it (agc_register_command_tests checks that). PPSA05684 (Uncharted: Legacy of Thieves Collection) calls sceAgcCbSetShRegistersDirect with an empty list at its main menu and terminated with what(): sceAgcCbSetShRegistersDirect: empty register list.

An empty list now writes no packet, leaves the buffer untouched and returns its cursor, as its size says; a null buffer still throws. No packet is the behaviour the size function implies; the console's own return value for an empty list is not measured.

Tested

• Linux 7.2.9-zen, GCC, Ninja, Release, main 2b7e3ce: full ctest 1182 of 1182, tools/check_conventions.py --base origin/main passes.
• agc_register_command_tests now writes an empty list through each writer into a sentinel-filled buffer and requires the returned pointer to be the cursor, the cursor unchanged and the sentinel intact. With Packet.cpp from main it fails with sceAgcCbSetShRegistersDirect: empty register list.
• PPSA05684 on a local stack (main plus feat(agc): implement image and sampler tables indexed at runtime offsets #757/feat(agc): implement image tables loaded through 64-bit pointers #758): the main menu no longer terminates on this call.

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: none
• Unblocks: PPSA05684 Uncharted: Legacy of Thieves Collection (PPSA05684) on Linux / NVIDIA #4001 (main menu termination)
• GPUs tested: n/a (command buffer encoding only)
• AI-assisted: yes

[Dextroid17]

Measured on this branch (cf914b5, 238 commits behind main), default CMake build (CMAKE_BUILD_TYPE empty, so unoptimized), lavapipe not involved (command-buffer encoding only), CachyOS Linux 7.2.9, agc_register_command_tests, rebuilt:

• 3 of 3 runs pass.
• With Packet.cpp taken from main and the new test kept: the added call throws main's sceAgcCbSetShRegistersDirect: empty register list and the test exits 1. That matches the PR body.

I read the change: a count of 0 now returns buffer->cursor_up without writing, and a null buffer still throws (by reading; no test covers a null buffer with count 0). That agrees with ...GetSize(0) == 0, which the existing test already requires.

Not checked: what the console returns for an empty list (the PR body says the same), the title run on PPSA05684, and other targets.

AI-assisted: yes

[XairRrR]

Thanks for measuring it. The null-buffer case is now covered: f991052 adds expectFailure([&] { writers[i](nullptr, nullptr, 0); }) next to the empty-list check, for all three writers. With the Require(buffer != nullptr, ...) line removed from WriteRegisters the test crashes (segfault on buffer->cursor_up); with it, it passes.

AI-assisted: yes

[Dextroid17]

Thanks, that closes the gap I noted. Re-run at f991052 as-is (merge base 2b7e3ce, not rebased onto current main), Release build, Linux (CachyOS, kernel 7.2.9), agc_register_command_tests:

• 3 of 3 pass.
• With the Require(buffer != nullptr, ...) line in the count == 0 branch of WriteRegisters removed, the test segfaults (exit 139), 3 of 3, as you describe.

AI-assisted: yes

[adriadam10]

Checked on the PR merged onto current main: agc_register_command_tests passes, and with Packet.cpp reverted the new empty-list check fails. Empty list returns the cursor without writing, null buffer still throws, which matches what the ...GetSize functions report; merging.