* Scope Kamek bl-patch LR-continuation detection to genuine skip-return targets
Fix crash from Kamek skip-return hooks (Item Rain crash) (#182) added every
Kamek BranchLink patch target to lrContinuationCallTargets unconditionally,
with no filter analogous to the RetroWfcHookSetsLinkRegister check already
used for RetroWFC hooks. Since bl is the ordinary PowerPC call instruction,
this made the codegen treat effectively every patched call in the mod as a
potential skip-return hook, forcing conservative handling (full register
reload, disabled resident-call fast paths, local LR-continuation dispatch
tables) onto thousands of calls that just return normally.
For Retro Rewind this inflated total translated mod size by +42%
(1,414,327 -> 2,005,284 lines), concentrated in ~10 unrelated overlay
functions that happened to call a patched target, and was enough to make
one aggregate build shard pathologically slow to compile (hangs Linux CI).
Instead, only mark a bl target as LR-continuation-aware if a lightweight
discovery-only decode of its own body actually finds evidence of
skip-return behavior via DiscoverLrRelativeIndirectJumpOffsets. Falls back
to the conservative (old) behavior if a target can't be statically
analyzed, so no skip-return case is silently missed.
Verified against the real Retro Rewind mod: total mod size returns to
1,416,350 lines (+0.14% vs. pre-fix, down from +42%), all 6 genuinely new
continuation functions from the original fix are preserved, zero
functions lost, and all 609 existing translator tests still pass.
* Distinguish exhausted from truncated LR-relative offset search
CodeRabbit flagged that TargetExhibitsLrSkipReturn (added in ad2d4e7) treated
an empty DiscoverLrRelativeIndirectJumpOffsets result as a verified "this
target never skip-returns," but the analysis silently drops any path state
once more than MaxStatesPerInstruction (16) distinct states reach one
instruction - a bctr/return on a dropped state can never contribute its
offset, so an empty result could be an incomplete search rather than a real
negative. Treating every capped case as "skip-return possible" outright was
rejected as too broad a fallback given how conservative/expensive that path
already is.
Instead: raise MaxStatesPerInstruction 16 -> 512 (an arbitrary conservative
bound to begin with, not something correctness depended on) so genuinely
branchy functions have far more headroom to reach an exhaustive answer, and
give DiscoverLrRelativeIndirectJumpOffsets an optional onStateCapExceeded
callback that fires exactly when a state is dropped. TargetExhibitsLrSkipReturn
now only falls back to the conservative "treat as skip-return" answer when
the search both found nothing and the cap was actually hit during that run -
not whenever the cap merely exists - so a target is trusted as clean once the
search genuinely exhausts it.
Verified: all 609 translator tests pass, and a full translate-mod run against
the real Retro Rewind mod produces byte-for-byte identical output to the
prior fix (same 4,065 functions, 1,416,350 total lines) - confirming the
16-state cap was never actually the limiting factor in practice and this
change is a pure safety-net closure, not a behavior change for this mod.
* Add LR continuation regression tests
* Refine LR continuation hook analysis
---------
Co-authored-by: patchzyy <64382339+patchzyy@users.noreply.github.com>
fixes https://github.com/patchzyy/Wiicompiled/issues/159
previous requirement for ubuntu 24.04+ libstdc++ inherited from dawn prebuilds now dropped to ubuntu 22.04+ libstdc++ like the rest of the prebuilds
also add all architectures to the URL_HASH check since the dawn tag doesn't change but the binaries have
translate-recursive now starts at 0x800060A4 (__start) to match recomp.yml and system_bridge.h. The payload curl uses rwfc.net/api/wfc/payload like the rest of the guide, since nas.play.rwfc.net does not answer over https.
Co-authored-by: devangpratap <devangpratap@proton.me>
* fix: Kamek LR-continuation hook discovery and dispatch
* test: cover branching Kamek LR continuations
* review fix
* another review fix
fix: get the new tests to pass
test: expose LR restore and loop continuation regressions
* Update translator/src/Translator.Core/Mods/ContinuationPlanner.cs
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
* test: cover continuation regressions from the new path-sensitive planner
* Update translator/src/Translator.Core/Mods/ContinuationPlanner.cs
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
* test: cover continuation regressions from the new path-sensitive planner
* fix: preserve LR continuation analysis across large handlers and clobbers
* fix: track LR-relative r1 across update-form stack stores
* Harden LR-relative continuation test coverage
* Fix LR/SP continuation state tracking
---------
Co-authored-by: patchzyy <64382339+patchzyy@users.noreply.github.com>
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
Brings in patchzyy/Wiicompiled main: os_sleep parked-thread fix (#195),
HTTPS Retro WFC payload (#198), macOS build guide (#177), and the
reverse-Z depth fix (#134).
Conflicts were in aurora-main/lib/gfx/common.cpp and lib/gx/shader.cpp,
both from #134, which lands squarely on the VR stereo replay path.
#134 makes UseReversedZ genuinely reversed: the near/far correction now
applies exactly once, inside effective_projection(), instead of being
applied there AND per-vertex in the shader (the double application had
been cancelling out, so "reversed" Z silently behaved like forward Z).
Three pieces of the VR path were built against that old behaviour and
would have broken silently, so they are adapted here:
- shader.cpp exact-screen-depth parked the virtual screen at -0.5*w
specifically so the shader's following negation would land it at
+0.5*w. With that negation gone it now writes +0.5*w directly; keeping
the minus sign would park the screen at NDC -0.5, outside the clip
volume, discarding every 2D/HUD draw.
- stereo_replay.hpp backend_ndc_depth_row re-applied the correction to
the projection it was handed. That projection is effective_projection()
output, which now already carries it, so the function is a pass-through
of the Z row and no longer depends on the reversed-Z setting; the dead
bool parameter is dropped. Re-applying it would invert the virtual
screen's depth ordering, so 2D layers meant to sit on top would lose
the depth test to the ones behind them.
- shader_info.cpp stages the host depth window for that exact-depth path.
It now uses the same reversed-Z remap as upstream's new SetViewport
code, since frag_depth is written directly and has to reproduce the
window the fixed viewport transform would have applied. Restricted
depth windows (how the game forces an element in front of everything)
are exactly the 2D draws the virtual screen carries.
The SetViewport resolution keeps upstream's remap but retains the
ordering/clamp guard our version had: for any ordered guest range the
result is identical to upstream, and it avoids handing WebGPU
minDepth > maxDepth for the swapped pair MKW is known to emit. The VR eye
replay reuses these recorded values, so the guard covers that path too.
Test updates:
- stereo_replay_test now asserts the composed Z row against the staged
projection's own Z row rather than against the helper's output, so it
actually catches a re-introduced double correction (verified: it fails
when the old negation is put back; the previous self-consistent form
passed).
- gx_fifo_test's clearDepthValue expectation followed #134's deliberate
clear_depth_value() inversion, expressed through UseReversedZ rather
than hardcoded. Upstream changed the behaviour without updating this
test, so it fails on upstream/main as-is.
Verified: aurora suite 247 passed with the same 2 failures that already
fail on the pre-merge branch (IndexedPaletteHistoryKeepsAbsoluteVertexSlots,
PacksOneUniformWhenBothHalvesNeedInitialValue - both pre-existing, unrelated
to depth); shader.cpp and common.cpp compile clean; translator suite 577
passed. Not yet validated on-device in VR.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* Fix already downloaded toolchain re-use
the following mv command would move $work into $toolchain_dir if the $toolchain_dir folder already existed.
* resolve z-fighting
ProcessSleepTimers popped every due timer into a private vector and then
resumed the sleepers in a loop. OSResumeThread re-enters SelectThread, which
can switch fibers away mid-loop, so the timers still in that vector were
gone from gSleepTimers while their threads stayed parked (Ready, suspended,
no timer). The reconciler healed them 100ms later and the stale-timer drop
fired when the original fiber eventually resumed.
Pop one due timer at a time straight from the shared table instead, so any
timer not yet processed stays visible to every other pump while this call
is switched away.
Co-authored-by: jordanblakepp <slamuelrose2002@gmail.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Brings in 25 upstream commits: Dolphin-compatible input expressions and
GCPadNew.ini import, NAND setting.txt console identity, empty-MKW-save
handling, rumble toggle, LLVM 22, and CI caching.
The only conflict was runtime/CMakeLists.txt, where both sides appended
test targets after mkw_platform_paths_tests. Both blocks are kept: the VR
first-person test alongside upstream's NAND save/settings, SC serial, and
input expression tests.
runtime_config.h and settings_overlay.cpp auto-merged; the VR settings
menu, recenter hotkey, and stereo/first-person init calls are intact
alongside upstream's InputBindings wiring.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* caching a little bit
* provide CMake with the explicit path to sccache.exe
* map ACTIONS_RESULTS_URL to ACTIONS_CACHE_URL so sccache can upload the
files...
* i removed the parallel oops
* small change
* doing a little bit of flag editing
* update sccache and cache nuget stuff
* treat empty mkw save as missing
first-run format zero-fills rksys.dat before any real save; a quit before
the first save left an all-zero file that read back as corrupt and trapped
the user in a delete/recreate loop. read opens now treat an all-zero
rksys.dat as absent (a real save always begins with the RKSD0006 header),
so the game recreates it from scratch. also ignore native build output.
* shorten
* I dont really want to change this to be honest.
* extra safety
---------
Co-authored-by: patchzyy <64382339+patchzyy@users.noreply.github.com>
patchzyy