ci(hypatia-scan): fix unresolvable setup-beam version pins - #63
Merged
Merged
Conversation
elixir 1.19.4 / otp 28.3 are not in the setup-beam version index, so the "Setup Elixir" step failed fast on every run, making the Hypatia dogfooding job red estate-wide regardless of PR content. Pin to currently-published loose major/minor versions instead. https://claude.ai/code/session_01GTo7dz32ZgxuHXefv8BGqn
hyperpolymath
marked this pull request as ready for review
May 15, 2026 20:57
hyperpolymath
added a commit
that referenced
this pull request
May 15, 2026
**Throwaway verification PR — do not merge.** One-line no-op comment in `hypatia-scan.yml` to run the `Hypatia Neurosymbolic Analysis (Dogfooding)` job in a PR context (which can be read from the agent environment) and confirm the corrected `elixir 1.18` / `otp 27` pins (merged via #63) actually make the job green. Will be closed once the result is observed. Permanent upstream hardening tracked in #64. https://claude.ai/code/session_01GTo7dz32ZgxuHXefv8BGqn --- _Generated by [Claude Code](https://claude.ai/code/session_01GTo7dz32ZgxuHXefv8BGqn)_ --------- Co-authored-by: Claude <noreply@anthropic.com>
hyperpolymath
added a commit
that referenced
this pull request
Sep 20, 2026
…not see GitHub-generated checks The five contexts this census flagged as unsatisfiable are all published: every one reports in the check-run history of the repository that requires it, as a job of a GitHub-generated workflow (code scanning default setup and friends) that no file in the repository mentions. Evidence, gathered 2026-09-20 from check-runs on the default branch: awesome-gleam Analyze (actions) 8 of the last 8 commits, and PR #12's head awesome-gleam Adjust Configuration reports on main (intermittent) casket-ssg Analyze Code (actions) 15 of the last 30 commits coord-tui CodeQL Analysis (actions, none) 5 of the last 8 commits, and PR #63's head vext CodeQL Analysis (actions, none) 12 reports over 8 commits, and PR #51's head The previous commit stated that all five were re-checked and are "live defects". That re-check answered the wrong question: it confirmed the *rules* were live (active enforcement, ~DEFAULT_BRANCH, integration_id 15368) and never asked whether a *publisher* exists. The publisher is the platform. This matters beyond the ledger: merged guidance to go and patch four repositories' required-status-check lists would have deleted working gates on false evidence, and the deleted names would have been the ones nothing could see — the checks GitHub generates itself. Changes: * the audit's results now read 0 unsatisfiable, with the five candidates and their evidence; the residues table becomes the refuted-candidates table; * the method note becomes the operational rule: a file-based pass may only produce candidates; an unsatisfiable verdict becomes a finding only after check-run history confirms the name never reports; a required context is never deleted on a file-based verdict alone; * the gate carries the same limit in its header, in its warning and error text, and in --print-commands, so the person reading a finding is told to confirm it before acting; * the census caveats section cites this as the worked example rather than a hypothetical. The class remains real — tropical-types#17 was a genuine mis-named requirement. What changes is the confidence level a file-based census may claim.
hyperpolymath
added a commit
that referenced
this pull request
Sep 20, 2026
…not see GitHub-generated checks (#876) ## What Withdraws the five "unsatisfiable required contexts" this census reported, because all five are published. Each one reports in the **check-run history** of the repository that requires it — as a job of a GitHub-generated workflow (code scanning default setup and friends) that no file in the repository mentions: | Repository | Context | Evidence (check-runs, default branch, 2026-09-20) | | --- | --- | --- | | `awesome-gleam` | `Analyze (actions)` | 8 of the last 8 commits, and PR #12's head | | `awesome-gleam` | `Adjust Configuration` | reports on `main` (intermittent) | | `casket-ssg` | `Analyze Code (actions)` | 15 of the last 30 commits | | `coord-tui` | `CodeQL Analysis (actions, none)` | 5 of the last 8 commits, and PR #63's head | | `vext` | `CodeQL Analysis (actions, none)` | 12 reports over 8 commits, and PR #51's head | ## Why the previous pass got it wrong The verification in the previous commit asked whether the *rules* were live — active enforcement, `~DEFAULT_BRANCH`, `integration_id: 15368`. It never asked whether a *publisher* exists. For these five the publisher is the platform itself, not the repository, so a census over repository files cannot see it. That is not a footnote. Acting on the previous version would have meant patching four repositories' required-status-check lists to delete working gates, on false evidence, and the names deleted would have been the ones nothing in the repository can see. ## What changes * results read **0 unsatisfiable**; the five are recorded as refuted candidates with their evidence; * the operational rule is stated where a reader will meet it: a file-based pass may only produce *candidates*; an `unsatisfiable` verdict becomes a *finding* only after check-run history confirms the name never reports; a required context is never deleted on a file-based verdict alone; * the gate says this in its header, in its warning and error text, and in `--print-commands`, next to the remediation it prints; * the census caveats section cites this as the worked example instead of a hypothetical. ## Verification * `bash scripts/tests/check-required-contexts-test.sh` — 5/5 pass. * `bash scripts/check-canonical-names.sh origin/main` — clean. * The refutation is reproducible: for each context, compare `GET /repos/{r}/commits/{default}/check-runs` and the check-runs on a recent PR head.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
The Hypatia Neurosymbolic Analysis (Dogfooding) job (
hypatia-scan.yml, jobscan) has been failing on 100% of runs across the estate, independent of PR content — red on already-merged PRs (#58, #60) and every new PR (#62).Root cause (inferred from a consistent ~12s fast-fail + workflow logic; the scan step uses
--exit-zeroand the only gate is "warn but don't fail"): theSetup Elixirstep pinnedelixir-version: '1.19.4'/otp-version: '28.3', which are not present in thesetup-beamversion index, so the step errored immediately and failed the whole job for everyone.This pins to currently-published, resolvable loose major/minor versions (
elixir 1.18/otp 27) so patch availability cannot re-break it.Changes
hypatia-scan.yml: correctedsetup-beamElixir/OTP pins + rationale comment.Test plan
Hypatia Neurosymbolic Analysis (Dogfooding)goes green on this PR (unrelated no-op change → proves the failure was environmental, not finding-driven).1.18/27actually build Hypatia's scanner (mix escript.build); bump deliberately if the scanner needs newer.hyperpolymath/hypatia+ a smoke job there so a bad pin fails at the source, not estate-wide.https://claude.ai/code/session_01GTo7dz32ZgxuHXefv8BGqn
Generated by Claude Code