feat(exclusion): re-land --globus robustness layer dropped from #13 merge #14
+417
−2
Conversation
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
…-globus
Exclusion must work under --globus (gate runs on a compute node), so any failure
there must NEVER take down the archive, and the *reason* must travel back to the
launcher. This adds the robust layer (still behavior-neutral — nothing calls it
yet):
- evaluate_exclusions(matched_members, source_folder, spec, ...) — decides the
excluded set and returns a fully serializable diagnostic report; NEVER raises.
status OFF (no spec) / DISABLED (could not run safely -> back up everything,
with disabled_reason) / ON. Reports excluded[] (+blob_oid/remote_ref),
demoted[] (candidate kept, with verdict+reason), contested_data[] (DATA that
also matched a sub-rule — proof precedence, not luck, kept it), counts{}, and
excluded_bytes. Targeted handlers for malformed spec, missing repo_root,
precondition failure, and per-file gate errors, plus an absolute catch-all
backstop -> DISABLED + error recorded, drop nothing.
- companion_guard (INV-C): a directory that had matched files must ship >=1
member and never have a DATA member excluded; on violation, exclusion is
suppressed run-wide (drop nothing) and reported for the wiring to raise
[EXCLUSION-ABORT].
- Distinguish git-unavailable (rc 127, a likely --globus compute-node gap) from
repo-missing in preconditions, and from a false UNTRACKED in git_file_verdict
-> COV_ERROR "git unavailable", so a disabled run is diagnosable.
Tests (+11, suite 99 -> 110): OFF/DISABLED matrix (no spec, malformed spec, no
repo_root, repo missing, git-missing reported distinctly, never-raises on an
injected internal error), and live end-to-end (covered->excluded,
committed-not-pushed->STALE-demoted, DATA kept, dirty-subtree disables the whole
run, contested DATA logged, companion guard suppresses, report JSON-serializes).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Sign in
to join this conversation on GitHub.
1 participant
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.
Why this exists
The merge of #13 captured only its first commit (
1abc8e0, the classifier + git gate). The follow-up commit I pushed afterward — the--globusrobustness layer you asked for (evaluate_exclusions/companion_guard+ hardening + 11 tests) — was left behind (the merge's second parent is1abc8e0, not the later0095a27). Somaincurrently has the primitives but not the orchestrator.This PR re-lands exactly that dropped commit, unchanged, on top of current
main. Still behavior-neutral (nothing calls it yet).What it adds (the same content reviewed on #13)
The gate runs on a compute node under
--globus, so the contract is: exclusion can never fail an archive, and the reason must travel back to the launcher.evaluate_exclusions(...)— never raises. Returns one fully JSON-serializable diagnostic report (crosses the Compute boundary back to the launcher):statusOFF/DISABLED(→ back up everything, withdisabled_reason)/ON; plusexcluded[],demoted[],contested_data[],counts{},excluded_bytes. Targeted handlers for malformed spec / missingrepo_root/ precondition failure / per-file gate error, plus an absolute catch-all backstop.companion_guard(INV-C): on violation, suppresses exclusion run-wide (drops nothing) and reports it for a loud[EXCLUSION-ABORT].gitunavailable (rc 127 — likely compute-node PATH gap) reported distinctly from "repo missing," never a falseUNTRACKED.Tests
+11 (suite 99 → 110 green): OFF/DISABLED matrix incl. git-missing-reported-distinctly and never-raises-on-injected-error; live end-to-end (covered→excluded, committed-not-pushed→STALE-demoted, DATA kept, dirty-subtree disables the whole run, contested DATA logged, companion guard suppresses, report serializes).
Co-Authored-By: Claude Opus 4.8 noreply@anthropic.com
🤖 Generated with Claude Code