Tracking the two self-proof corpora that the weekly verification-proofs-cron does not
yet gate, because they have genuine (not config) verification errors. They need dedicated,
locally-tested repair — both toolchains' download hosts (isabelle.in.tum.de,
mizar.org) are 403-blocked from the dev sandbox, so they can only be iterated on CI,
which is too slow/uncertain for proof repair. Context: the proof-CI sweep (#234, #275, #276,
#277, #278, #279) gated Coq, Lean 4, Agda, Idris2, and the Isabelle base (Basic,
EchidnaList, EchidnaNat, Propositional). These two are the remainder.
1. proofs/isabelle/algebra/GroupTheory.thy (excluded from proofs/isabelle/ROOT)
"ML training corpus" material that was never checked. On Isabelle2025-2 it has genuine bugs:
mul_right_inv, mul_right_id: omit assumes "group G" but use assms → "Undefined fact: assms".
center_closed: proved with group.m_comm, which holds only for comm_group, not a bare group.
- Dubious notation / lemma names:
(⊕ₘ) for power, Prod.mult_def/Prod.inv_def/Prod.one_def, power_add/power_0/power_1.
- Uses the deprecated path import
~~/src/HOL/Algebra/Group (already changed to HOL-Algebra.Group).
Needs each lemma's statement + proof checked against HOL-Algebra. Once green, re-add an
Echidna_Isabelle_Algebra session (parent HOL-Algebra) to proofs/isabelle/ROOT and a
just proofs-isabelle target.
2. proofs/mizar/*.miz
Mizar is the project's mock-only Tier-4 tail. The cron's Mizar install fully works
(resolves + downloads mizar-8.1.15 from mizar.uwb.edu.pl, binaries on PATH via
MIZFILES), but the .miz files have genuine, heterogeneous errors:
propositional.miz: 24 errors — every thus A c= B; thus B c= A set-equality
double-inclusion step (code 51) and let x be object (code 55). Likely a missing
definitions environ item, but unconfirmed.
basic.miz: a makeenv error in the Notations phase.
numbers.miz, mvp_basic.miz: not yet diagnosed (mvp_basic is trivial, likely fine).
The just proofs-mizar recipe (makeenv + verifier, failing on a non-empty .err) remains
for local use. Once the files verify, re-add the mizar job + proofs/mizar/** trigger to
.github/workflows/verification-proofs-cron.yml.
Suggested approach
Repair in an environment where Isabelle2025-2 and the Mizar system are installable and
runnable locally (the sandbox here cannot reach either host), then re-gate via the weekly cron.
Tracking the two self-proof corpora that the weekly
verification-proofs-crondoes notyet gate, because they have genuine (not config) verification errors. They need dedicated,
locally-tested repair — both toolchains' download hosts (
isabelle.in.tum.de,mizar.org) are403-blocked from the dev sandbox, so they can only be iterated on CI,which is too slow/uncertain for proof repair. Context: the proof-CI sweep (#234, #275, #276,
#277, #278, #279) gated Coq, Lean 4, Agda, Idris2, and the Isabelle base (Basic,
EchidnaList, EchidnaNat, Propositional). These two are the remainder.
1.
proofs/isabelle/algebra/GroupTheory.thy(excluded fromproofs/isabelle/ROOT)"ML training corpus" material that was never checked. On Isabelle2025-2 it has genuine bugs:
mul_right_inv,mul_right_id: omitassumes "group G"but useassms→ "Undefined fact: assms".center_closed: proved withgroup.m_comm, which holds only forcomm_group, not a baregroup.(⊕ₘ)for power,Prod.mult_def/Prod.inv_def/Prod.one_def,power_add/power_0/power_1.~~/src/HOL/Algebra/Group(already changed toHOL-Algebra.Group).Needs each lemma's statement + proof checked against HOL-Algebra. Once green, re-add an
Echidna_Isabelle_Algebrasession (parentHOL-Algebra) toproofs/isabelle/ROOTand ajust proofs-isabelletarget.2.
proofs/mizar/*.mizMizar is the project's mock-only Tier-4 tail. The cron's Mizar install fully works
(resolves + downloads
mizar-8.1.15frommizar.uwb.edu.pl, binaries on PATH viaMIZFILES), but the.mizfiles have genuine, heterogeneous errors:propositional.miz: 24 errors — everythus A c= B; thus B c= Aset-equalitydouble-inclusion step (code 51) and
let x be object(code 55). Likely a missingdefinitionsenviron item, but unconfirmed.basic.miz: amakeenverror in the Notations phase.numbers.miz,mvp_basic.miz: not yet diagnosed (mvp_basic is trivial, likely fine).The
just proofs-mizarrecipe (makeenv + verifier, failing on a non-empty.err) remainsfor local use. Once the files verify, re-add the
mizarjob +proofs/mizar/**trigger to.github/workflows/verification-proofs-cron.yml.Suggested approach
Repair in an environment where Isabelle2025-2 and the Mizar system are installable and
runnable locally (the sandbox here cannot reach either host), then re-gate via the weekly cron.