RERUN_LOG.txt -- re-runs of 27 October 2026 (GAP 4.16.1, Python 3.13.5 with networkx 3.6.1, macOS, one core
per program, nice -n 10), made from a copy of this folder, with outputs in rerun/ of that copy.

"same": the summary lines (orders, vertex and edge counts, chordality, C4, least hole lengths, the
"... every maximal subgroup has C4-free E(M)" and "ALL ... true" lines) are identical to the recorded output;
GAP's line wrapping and the "Syntax warning" lines are ignored.

program                     time      result
check_note.g                13 s      CHECK_NOTE ALL OK: true (recorded/check_note.out is this run)
check_c4_direct.g           927 s     CHECK_C4_DIRECT ALL AS EXPECTED: true (recorded/check_c4_direct.out is this run;
                                      PSU3(23): 12654 triples, no C4, 883 s)
programs/verify_holes.g     1.4 s     byte-identical to recorded/verify_holes.out (ignoring the warning lines)
programs/cert_m11_python.py <1 s      byte-identical to recorded/cert_m11_python.out
programs/cert_families.g    8 s       same (ALL FAMILY CERTIFICATES OK: true)
programs/cert_finite.g      5 s       same; the printed holes differ for J1 (orders 2,5,3,2,3,5; recorded 3,5,2,5,3,2)
                                      and PSU3(3) (2,3,2,3,2,3; recorded 2,3,4,3,2,3), both re-checked by IsCyclic
programs/cert_c4free.g      47 s      same
programs/cert_c4free2.g     26 s      same
programs/cert_c4free3.g     248 s     with GAP's default seed: stopped with a GAP error, see below;
                            476 s     with another seed: same
programs/cert_j3_tits.g     12 s      same
programs/cert_ree27.g       89 s      same
programs/cert_sporadic.g    21 s      same groups, other (random) C4s, all re-checked: "induced cycle in E: true"
programs/cert_containments.g 1 s      same
programs/cert_sanity.g      243 s     same (ALL AGREE: true for all 56 groups)
programs/cert_stress.g      1 s       same

cert_c4free3.g (PSU3(23) through the maximal subgroups listed by ClassicalMaximals). The recorded output
(recorded/cert_c4free3.out) shows all 8 classes C4-free (all with chordal E(M)). The file of the program was last
changed after that run (13:27 against 13:09 for the output), so the recorded output may come from a slightly
different version. Two re-runs of the present version stopped after about 4 minutes, at the second maximal
subgroup (GU2(23)/3, order 97152), with the GAP error
  Error, no 1st choice method found for `FamilyRange' on 1 arguments
raised inside SmallerDegreePermutationRepresentation (FittingFreeLiftSetup).
A third re-run with a different random seed,
  gap -q -o 8g -c 'Reset(GlobalMersenneTwister, 2026); Reset(GlobalRandomSource, 2026); PYTHON := "...";' \
      programs/cert_c4free3.g
completed in 476 s with the same results as the recorded output (all 8 classes C4-free, all with chordal
E(M); the same vertex and edge counts). So the error depends on the random choices of that GAP method.
PSU3(23) does not depend on this program: check_c4_direct.g treats the whole group without maximal subgroups,
and verification run B treated the same 8 classes with its own program (verification_run_B/vb_n4_b.out).
