Repository navigation
Fix the two CI test failures left after the v0.9.1 sync - #120
Merged
Merged
Conversation
test_sh2_discovery: the fork's cycle model opens every block, including a branch target's label, with a prologue (SH2_STATE_POINT/SH2_CYC/SH2_CYC2/ SH2_BLK) before its first instruction. The hook is still emitted after the hooked instruction on the branch path, which is the intended placement, so the regex now allows the block prologue and pins the hooked statement. gpu-renderer: upstream's new high-resolution floor case (a366075) expected the sprite red as 0xFF, from replicated low bits. The fork's geometry shader expands RGB dots as VDP2 does (5 bits shifted left by 3, no replication), as every other case in this file already expects, so the sprite channel is 0xF8. The trailing "Render target texture is NULL" was a stale SDL_GetError string, not the cause.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Follow-up to #119. Test-only changes; no runtime, shader or emitter code changes.
test_sh2_discovery
test_instruction_hook_survives_a_direct_internal_branchThe fork's cycle model (runtime-diagnostics.patch,
emit.py) opens every block, including a branch target's label, with a prologue ofSH2_STATE_POINT/SH2_CYC/SH2_CYC2/SH2_BLKbefore the first instruction. The hook is still appended to the hooked instruction (c.r[5] = 0x1u; sh2_hook(c, 0x06000006u);), which is the documented after-instruction placement and is reached on thegotopath. The emitter is correct; the regex now allows the block prologue, pins the hooked statement, and checks that there is exactly one hook. It still passes against an emitter with no cycle model.gpu_renderer_selftest "High-resolution floor must retain sprite colour-calculation order"
Upstream's new case (a366075) expects the sprite red channel as 0xFF (5-bit values with replicated low bits). The fork's
geometry.hlslexpands RGB dots the way VDP2 and the CPU compositor do (shifted left by 3, no replication), and every other case in this file already expects 0xF8. The expectation is updated to 0xF80000. The trailingSDL: Render target texture is NULLin the failure was a staleSDL_GetError()string, not the cause.Local (Linux, clang, lavapipe llvmpipe 19.1.7): gpu_renderer_selftest PASS (213 cases), image_scaler_selftest --gpu PASS, gpu_parity PASS, Python 338 run OK (4 skipped), source audit PASS (293 files).