mirror of
https://github.com/Druthulu/BFM-decomp
synced 2026-09-27 22:45:39 -04:00
Drew T
3f51101ca9
Paused at Drew's request for a Windows restart. Nothing running, tree lock free, tree clean (R23 db churn aside). R22 run 19x this session, 140/140 every time. THE ONE THING OWED: `.run/s37/*/func_*.c` — 16 wave-5 drafts, all claiming MATCH, NONE gated. Force-added to git (.run/ is gitignored) because they cost ~2.7M agent tokens and Workflow's resumeFromRunId cache is SAME-SESSION-ONLY, so it does not survive the restart. Resume by GATING them, not by re-running the wave. Also preserved: the wave scripts (.run/s36w.js, .run/s37w.js), manifests, the hardened gate driver, and the four capture drivers. MEASURED THIS SESSION (both answer questions Drew asked): - The wave PROMPT is the lever. Bank rate 76% -> 77% -> 100% -> 100% on the same models and the same gate, prompt the only variable. The jump was STEP 0 (a magic-literal grep of src/, ahead of engine_core.h) — and that step came from a wave-2 agent's index_gap report, i.e. the agents write the next prompt. - The pipeline() fix, before/after: wave 4 parallel() 14 targets / 136 min / 2.5x parallelism; wave 5 pipeline() 16 targets / 82 min / 3.8x. 40% faster on 14% more targets. The two-batch design was a hard barrier with 46-min dead gaps at each boundary; the harness already caps at 16 so the batching bought nothing. Housekeeping: removed 3 stray cc1 intermediates (t.i, t.i.greg, t.s) that an agent left at the repo ROOT — scratch belongs under .run/ (R12), same class as the gccdump.lreg noted at the Phase-24 close.
38 lines
2.0 KiB
Python
38 lines
2.0 KiB
Python
#!/usr/bin/env python3
|
|
"""Capture the whole-binary blocker for a gate-refused draft: splice -> build -> read the
|
|
compiler's OWN error -> revert. Agents cannot run the gate, so a TU-level declaration conflict is
|
|
invisible to them (§136a: classify on the build's OUTPUT, never its exit status)."""
|
|
import sys, os, re, subprocess, shutil
|
|
sys.path.insert(0, 'tools'); import corpus
|
|
for spec in sys.argv[1:]:
|
|
ov, fn = spec.split(':')
|
|
addr = int(fn.split('_')[1], 16)
|
|
st = corpus.stubs(ov); rec = st.get(addr)
|
|
if not rec:
|
|
print(f"== {ov} {fn}: NOT A LIVE STUB (already banked?)"); continue
|
|
tu = rec.path
|
|
draft = f".run/s35/{ov}/{fn}.c"
|
|
stub = f'INCLUDE_ASM("{rec.asm_dir}", {fn});'
|
|
t = open(tu).read()
|
|
if t.count(stub) != 1:
|
|
print(f"== {ov} {fn}: stub line not found verbatim in {tu}"); continue
|
|
bak = t
|
|
open(tu, 'w').write(t.replace(stub, open(draft).read()))
|
|
r = subprocess.run(['make', 'build', f'BINARY={ov}'], capture_output=True, text=True)
|
|
open(tu, 'w').write(bak) # revert ALWAYS
|
|
out = r.stdout + r.stderr
|
|
# §136a, applied to MYSELF: a narrow keyword filter is exactly how a real error goes unseen.
|
|
# Keep every cc1/ld diagnostic line that names a source position, minus the known SHB noise.
|
|
HARD = re.compile(r'undeclared|redefinition|conflicting types|parse error|undefined reference'
|
|
r'|syntax error|too (many|few) arguments|invalid|incompatible')
|
|
lines_ = [l for l in out.splitlines()
|
|
if re.search(r'\.[ch]:\d+|undefined reference', l)
|
|
and 'built-in function' not in l and 'Makefile:' not in l]
|
|
hard = [l for l in lines_ if HARD.search(l) and 'warning' not in l.lower()]
|
|
errs = hard or [l for l in lines_ if 'warning' not in l.lower()] or lines_
|
|
print(f"== {ov} {fn} (build rc={r.returncode})")
|
|
if errs:
|
|
for l in errs[:6]: print(" ", l.strip())
|
|
else:
|
|
print(" NO COMPILE ERROR -> the draft builds clean; the miss is a byte DIFF at whole-binary")
|