Description
H4Lagrangian.lean has two unsolved goals, and the workflow that catches it ran once in sixty
lake build fails:
error: Trinity/H4Lagrangian.lean:73:59: unsolved goals
error: Trinity/H4Lagrangian.lean:108:46: unsolved goals
error: Lean exited with code 1
Those are the only two error lines in the whole run. Everything else the build emits is a
warning (This simp argument is unused, Variable name \x` is not explicitly referenced`).
Why nobody has seen it
Of the last 60 workflow runs in this repository, exactly one is Lean proofs — and that
one only exists because an unrelated pull request happened to touch three .lean files. The
workflow is paths-filtered, so it does not run on push to master; it has no baseline any green
state has ever been observed against, which is what tri pr ready reports as NO BASELINE and
VERDICT: CANNOT TELL:
a check that never runs on master has no green state anyone has ever seen.
This is the same shape as a paths:-filtered workflow with no push: trigger: the gate exists,
looks configured, and has effectively never executed. A proof with unsolved goals has been sitting
in the import closure of Trinity.lean — the root — and the build has been red for however long
without anything reporting it.
Two questions, and the second is the owner's
- Repair the proof. Two
unsolved goalsat named lines; that is ordinary Lean work and needs
somebody who can runlake(I cannot here — Lean is not installed in this environment, which
is why the header change in gate(commands): a hyphenated sibling is a command the space-separated matchers step over #3138 was deliberately shipped as "CI is the check and I am not
claiming it locally"). - Should
Lean proofsrun onpushto master? Right now it cannot fail visibly, so it
cannot protect anything. Adding apush:trigger would make it a real gate and would land red
until (1) is done — which is the honest sequencing, and is a CI-policy decision, not mine.
Related: the same run's build graph is the subject of the line-count correction in #3137, and
lake check-test / lake check-lint both report failed -> will not run in this log too, which
is a third thing nobody is seeing.
Found while landing #3138; anchored to that run, gh run view 33843156189.
User Scenarios
- Given
proofs/lean4/Trinity/H4Lagrangian.leancontains two unsolved goals at lines 73 and 108, whenlake buildruns in the Lean project, then it fails with exactly those two error lines and exits non-zero. - Given the "Lean proofs" workflow is paths-filtered and lacks a
push:trigger for master, when commits land on master, then the workflow does not execute and the red build remains undetected. - Given the unsolved goals are resolved and the workflow gains a
push:trigger for master, whenlake buildruns on master via the workflow, then it exits 0 and the gate becomes visible.
Requirements
- FR-001: The two unsolved goals in
proofs/lean4/Trinity/H4Lagrangian.leanat lines 73 and 108 MUST be resolved so thatlake buildproduces no errors. - FR-002: The "Lean proofs" workflow specification MUST include a
push:trigger for the master branch so it runs on every push and establishes a visible baseline. - FR-003: After both changes,
lake buildin the Lean project MUST exit 0.
Success Criteria
lake buildexecuted inproofs/lean4/exits 0- The "Lean proofs" workflow runs on push to master and reports success (exit 0)
- 2 unsolved goals resolved in
proofs/lean4/Trinity/H4Lagrangian.lean
Boundary
- proofs/lean4/Trinity/H4Lagrangian.lean
- specs/tri/pipeline/workflow.t27
Activity
fix(proofs): prove existing H4Lagrangian bounds
gHashTag commentedon Oct 7, 2026
Reopened 2026-10-07: the bulk close of issues without a
## Boundarysection was reverted by the owner. Nothing to do here; this comment replaces the earlier close notice.Fix unsolved goals in H4Lagrangian.lean and add push trigger to Lean …