W5-8c 1RB0RD_0RC0LB_0RD1RD_1LE1RA_1RC1LF_---1LA: Rocq log
Round 1 (2026-10-13): check_kill PASS
Kill file: ../1RB0RD_0RC0LB_0RD1RD_1LE1RA_1RC1LF_---1LA.v (458 lines), sha256
8c597d875a1125131e6a9f88fe63fc994b55df1ccecaba84b2c1d29909d00584.
Built from the parent's kill file (W5-8, sha256 beef54a5…cf48) by assemble.py (asserted single-match edits,
adapted from W5-8b's assemble.py; no hand edits). Route of proof.md 6: Perm bridge, no flip.
What is proved
- The parent's whole development unchanged, on
tmP(\btm\b->tmP):S_step,TE,TO,T6,T8(es'), primed versions,S_iter,DC,cc_steps(_spec),cc_steps'(_spec),rl,LR,Kr_list,Kr_cc. - New opening for
tmPfromE;; tape0:init : E;; tape0 -->* Kr 1291 35 38 21 9(6,950,830 steps), ten pieces cut at the left records 694,787 / 1,378,765 / 2,077,151 / 2,768,045 / 3,469,055 / 4,155,515 / 4,854,845 / 5,552,637 / 6,254,851 (opening_chunks.py->chunks.txt; the script asserts that step 6,950,830 is a left record and that its configuration isKr_list 1291 35 38 21 9= K(2583,70,76,43,18)). Piece lengths 683,978 .. 702,214 (parent's max ~702k). First piece starts bychange (E;; tape0) with (cconfig_to_config ([], [], 0, E)). opening : Kr 1291 35 38 21 9 -->* Ev 6649 21: the 18 follows of proof.md 6 (opening_args.py->opening_args.txt; its K values equal proof.md 2.2's table row by row, and the arguments equal proof.md 6's list).nonhaltP : ~halts tmP (E;; tape0):multistep_nonhalt+progress_nonhalt_simpleon (N, X) from (6649, 21).- Bridge:
phi = sigma^-1(A->E, B->F, C->B, D->C, E->A, F->D),perm : Perm tm tmP phi(split; intros [] [] ...; cbn in *; try discriminate; reflexivity/inversion H; subst; reflexivity),nonhaltbyintros H; change (halts tm (A;; tape0)) in H; exact (nonhaltP (perm_halts _ _ _ _ _ perm H)). - Scratch compile (
cc.sh, native on, 8 GB cap): 7.6 s, max RSS 0.52 GB.
check_kill: PASS (round 1, first run)
tools/check_kill.sh <kill file> 1RB0RD_0RC0LB_0RD1RD_1LE1RA_1RC1LF_---1LA under ulimit -s unlimited; ulimit -v 8000000:
RESULT: PASS, exit 0 (2026-10-13T13:28:27Z). Compile 7.85 s native; coqchk -norec pass 64.8 s (inside the
600 s cap); check_kill max RSS 3.8 GB; 0 assumptions (closed under the global context). Only warning: the second
TM_from_str literal (the parent's string, tmP). Log: check_kill_r1.log.
What failed
- Nothing. 6.95M direct steps in ten ~700k pieces: coqchk 64.8 s (vs 44.5 s for 4.1M in six pieces), memory unchanged at ~3.8 GB peak, so pieces of ~700k stay safe with a tape of ~5,500 cells.
Files
assemble.py(parent .v +chunks.txt+opening_args.txt+table_comment.txt-> kill file; refuses repo-local paths),opening_chunks.py,chunks.txt,opening_args.py,opening_args.txt,table_comment.txt,cc.sh(one compile frombuild/cc, 8 GB cap),check_kill_r1.log.