Repository navigation
Match 16 USA functions and repair reload/padding accounting - #423
Merged
Merged
Conversation
Replace eight renderer, effect, input and cache markers with complete C owners, and recover the separately entered abort/assertion bodies from the old 64-byte boot row. The ten USA v1.01 functions add 8,052 distinct code bytes and occupy 489 verified C placements across the three carrying releases. Correct three existing compiler mechanisms: snapshot the initial non-spill exclusion before a reload marks its own register; count nops inserted before an already-visited loop branch; and count the chosen known non-PIC expansion when measuring the division-label distance. Add authored controls, including negative checks against the old behavior, without opcode/register rewriting or function-specific compiler cases. Keep the existing museum brightness update on one debug source line so its unchanged operations retain retail's division boundary. Use full UV, quad, matrix and menu-table owners, preserve real return and callback contracts, and represent the cache's -1 slot inside the full table owner. Document the original untouched trail-packet word. Hold the otherwise exact Joba renderer in nonmatching because its nc-copy spelling violates the helper's memory-dependency contract. Preserve useful variants privately and retire only the superseded matching/functional drafts. Validation on the complete staged production inputs: - tools/build.py all --strict --jobs 6 --version us: 28 programs OK. - The same strict command for us2 and pal: 28 programs OK each. - All 84 loaded images equal retail, zero C fallbacks; all 489 new C placements and the three existing museum controls checked in placed.json. - unittest discovery: 405 tests, OK (one source-dependent harness skipped in discovery; it ran successfully against prepared GCC during the compiler build). - Real-Git check_repository.py and progress.py --check for all three versions pass. - README and progress report bytes are unchanged. No native gameplay test run.
ByteMatch progress reportSCES_516.07
No function changes. SCUS_972.68
No function changes. SCUS_972.68_v2
No function changes. Commit 3b8cf1d · Ratchet & Clank: Going Commando on ByteMatch |
Complete the current matching lot with the 1,496-byte floating formatter, the RPC-buffer initializer and record-reset loop, and three independently entered empty callbacks. These six bodies add 1,948 distinct USA bytes and 360 C placements. Together with the first commit, the PR contains sixteen USA v1.01 functions, 10,000 distinct bytes and 849 verified C placements. Count the two-word expansion of an unannounced external SI load/address only in the existing post-delay-slot padding pass. Preserve scheduler lengths, instructions, register choices, prior metadata, explicit GP views, PIC/buffering exclusions and delay-slot behavior. Track definitions at G0 without changing emitted assembly. Seven authored controls pass; the prior compiler fails the three positive cases and passes the four controls. Validate every callback entry through the actual dispatch table: cameras in 26 levels and objects in Boldan, in every release. Keep complete-body, overlap and replay guards. Allow only the union of independently verified boot and level splits when checking stale reports; unknown gaps still fail. Separate the formatter's real integer-digit budget and decimal exponent, and group the character encoding/store. Preserve and document retail's low-word dispatch and unbounded scratch behavior. Record the pre-existing floor/fmod word-construction issue without claiming it was repaired. Retain measured interface/owner corrections and bounded negative results. Preserve distinct native-oriented variants with targeted ABI corrections; remove the unresolved counter-display draft from the functional registry while keeping its C in nonmatching. No failed trial earns matching credit. Validation: - All 84 programs passed tools/build.py all --strict on the fifteen-body integration with the final compiler. All three boot programs were rebuilt strictly after adding the formatter, and its C placement was verified. - Rechecked all 84 complete loaded images and all 849 C placements. The 81 level inputs are unchanged; the modified core parent's complete object is byte-identical, and the new formatter has no level placement. - 415 unit tests: OK (one conditional source harness skipped in discovery; it passed during the compiler build). Real-Git repository check and all three progress checks pass. README/report bytes remain unchanged. - Public edited text is UTF-8/LF. No native gameplay session was run.
llesieur99
changed the title
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.
What
Match 16 functions present in USA v1.01: 10,000 distinct code bytes and 849 verified C placements across their carrying releases.
func_00115E28func_00115E38func_001185E8func_00294420func_002E7338func_002EEBC8func_002F0FA0func_00316E08func_00337210func_L00_00418230func_L00_00424F98func_L00_0042A4E8func_L00_0042A4F0func_L00_0042A4F8func_L03_002ECF08func_L05_00419CC8The maintained generator separates the abort/assertion entries and the three empty callbacks only after checking complete bodies, bounds and real entry witnesses. The callback family uses camera dispatch in 26 levels and object dispatch in Boldan. Stale-report validation accepts only the union of those independently verified splits; unknown gaps and old matching credit still fail.
Correct four bounded cases in existing compiler mechanisms:
Authored positive and negative controls cover these cases. No correction selects a game function/address or rewrites its operations/registers. The existing museum brightness routine keeps its calculation, clamp and store on one debug source line to preserve retail's division boundary; it adds no matching credit.
The C uses complete quad, matrix, UV, camera and menu-table owners, actual call interfaces, and the full cache owner for its real
-1sentinel. The formatter separates its integer-digit budget from the decimal exponent and groups the actual character encoding/store. Retire superseded drafts and preserve useful near misses and their reopening conditions. Distinct native-oriented variants retain their bodies with targeted interface repairs.Explicit holds and limits: the otherwise exact Joba renderer remains
nonmatchingbecause itsqcopy_ncform violates the helper's memory-dependency contract. The unresolved counter-display draft is removed from the functional registry and retained for analysis. The trail packet's untouched word and the formatter's low-word dispatch/unbounded scratch are retail behavior. A pre-existing uninitialized scalar in the floor/fmod word macros is documented and unchanged; no native-safety claim is made.Checked
python3 tools/build.py all --strict --jobs 6 --version us, thenus2andpal: 28 programs OK each on the fifteen-function integration with the final compiler.tools/build.py boot --strict --version ...passed for all three releases, with its C placement checked.python3 -m unittest discover -s tests: 415 tests, OK (one conditional GCC-source harness skipped in discovery). That harness ran successfully against prepared GCC during the compiler build.python3 tests/check_repository.py: passed with real Git metadata.python3 tools/progress.py --check --version us,us2, andpal: passed. README and report bytes are unchanged.No native gameplay or emulator session was run. The PR is ready for review and has not been merged.