[SleglJan]

What

Dispatch in System/src/Signals.cpp tested the blocked mask with 1u << guest, bit n. sigprocmask keeps the first word of the title's set as it is, and FreeBSD keeps signal n at bit n − 1 (_SIG_BIT), which is what FreeBSD's sigaddset (and #1641's) produces. So a title that blocked SIGTERM (bit 14) was not protected, and a set with bit 15 blocked SIGTERM instead of SIGURG.

Dispatch now tests 1u << (guest - 1). The stored mask and the set sigprocmask returns are unchanged, so they keep the title's layout. Nothing else interprets the mask's bits; sigprocmask only combines and returns it.

Found while adding the sigset functions in #1641: with them a title can build such a set without importing anything else, so the mismatch stops being latent. The two are independent changes; #1641 lists this one under Depends on so they land in order.

Tested

• Linux (CachyOS, kernel 7.2.9, GCC 16.2.1), full build with BUILD_TESTING=ON and ANYPS5_ENABLE_SPIRV_TOOLS=ON on this branch (based on main 4b005b2): ctest --repeat until-fail:3 466/466; tools/check_conventions.py --base upstream/main passes.
• guest_signals gains two cases: block {1u << 14} with SIG_BLOCK, raise(15) is not delivered, SIG_UNBLOCK, raise(15) is delivered; and block {1u << 15}, raise(15) is still delivered. The first was written before the change: on main it aborts (the raise is delivered through the block); it passes with the change. The existing round-trip assertions are untouched.
• Windows: run by @hikazey on Windows 11 with the CI toolchain (comment below): guest_signals 5/5 with this change, 5/5 failures with main's Signals.cpp and the new cases, full suite 457/457. Pending delivery of a blocked signal after unblocking is not implemented here or on main; this only corrects which bit is tested.
• Also the same Build workflow run on this commit in the fork's Actions (merged onto main 6b56175), run 37822080210: guest_signals passes its three repeats on Windows, full suite 473/473. Linux passed on a rerun after agc_driver_buffer_format hit its 120 s limit on the software Vulkan runner (the test takes 92 to 108 s on main's own runs; ci: warn about tests that take more than 30 s #1744).

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 in TechnicalDebt
• 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
• AI-assisted: yes

🤖 Generated with Claude Code

[hikazey]

Windows run, since the description says it wasn't tested there: Windows 11 Pro (26200), MinGW-w64 GCC 15.2.0 (WinLibs r7, the CI build), Release, configured like CI, Ryzen 9 5900X. This commit merged onto main 7995d657, which merges cleanly.

build	guest_signals
this PR	5/5 passed
Signals.cpp from main, new cases kept	5/5 failed
full suite with this PR	457/457 passed

So the bit n − 1 check also holds on the CRT signal/raise path that Dispatch uses on Windows: SIGTERM blocked through bit 14 isn't delivered, and bit 15 no longer blocks it.

AI-assisted: yes (Claude Code).

[SleglJan]

Thanks for the Windows run; the description now points to it, so the CRT path is covered on both hosts.

[SleglJan]

Rebased onto main ff8961d (head c26c1ca). #1916 (sigaction) and this PR both append cases to GuestSignals.cpp; both are kept, this PR's 12 lines after #1916's, with its two set variables renamed because term was taken. The Signals.cpp change is the same one line, and main still tests 1u << guest.

Linux (CachyOS, kernel 7.2.9, GCC 16.2.1), Release, BUILD_TESTING=ON, ANYPS5_ENABLE_SPIRV_TOOLS=ON, on this head:

• guest_signals and guest_signal_sets pass three repeats; check_conventions.py --base upstream/main passes.
• Full ctest on the RTX 3080 (615.71.09): 563 passed, 2 skipped, 1 failed of 566; the failure is agc_driver_dcc_keys_first_write ([bug] agc_driver_dcc_keys_first_write fails on main on a real NVIDIA GPU #2188, NVIDIA proprietary driver), which passes three repeats on RADV.
• Fork Build workflow on this head merged into main ff8961d, run 38005245415: relinker green on all three platforms; linux 536 passed, 29 skipped, 1 failed of 566 with guest_signals and guest_signal_sets among the passed, the failure guest_video_out_resolution_status (feat(videoout): implement sceVideoOutGetResolutionStatus #1714) aborting on the runner's offscreen SDL driver without Vulkan; windows stops at libSceVideoOut.prx with Windows build fails on main: duplicate patched export name: LibwuIonIBw in libSceVideoOut (since #1339) #2462's duplicate export. The fork's build of main ff8961d alone, run 38004711062, fails the same two ways with the same 29 skips.

AI-assisted: yes (Claude Code)