Skip to content

Failure register: file the misread rows and record the causes of the four uninvestigated classes - #859

Merged
dmmdea merged 2 commits into
mainfrom
fix/register-rules-and-causes
Nov 12, 2026
Merged

dmmdea merged 2 commits into
mainfrom
fix/register-rules-and-causes

Conversation

@dmmdea

@dmmdea dmmdea commented Nov 12, 2026

Copy link
Copy Markdown
Owner

Intent

After the last release the failure register still had four classes marked "not investigated" (failed_verification, other, poll_deadline, stall_cold_load) and 30 rows in the residue class. This change records the cause of each class and files the misread rows under the class that names them.

Scope

In: internal/ledger (rules and the cause catalogue), one name added to a pin list in internal/pipeline/transcribe_noaudio_test.go, docs/systems/failure-register.md, the register's golden file, a changelog fragment.
Out: the writers of the rows, the placement engine, the research acceptance (a companion pull request).

What changed

  • Rules, each on the writer's own fixed words. no node that answered is called ... is node_busy. A dispatch that could not connect, N consecutive poll failures, rendered the job but fetching it failed, and a poll deadline that ends on a last poll error are node_unreachable. Error class no_node_configured and the audio check's missing-tool row are config. media convert: convert: in and out are required and error class lint_errors are bad_input. reasoning tier: low classify confidence ... and the browse door's browse: blocked: are abstention. Compose's check_failed is failed_verification.
  • Causes. failed_verification is mostly the research lane's own checks failing digests that were right (fixed in the same release by the companion change; the entry states what that fixed and what still fails). poll_deadline was a node that stopped answering. stall_cold_load is one placement event: a long-context subtask placed on a seat that shares its cards with the pool seat, with the wait behind the swap filed as a cold-load stall. other is a bundle of misread sentences, not one cause.
  • A comment that listed the known misfilings now points at the list in RefileStamped.

The real run and its output

The built binary and the installed one, both read-only on the real ledger, ledger --failures --since 30:

class                 before   after
other                     30       3
abstention               221     228
node_busy                728     736
node_unreachable           3       9
failed_verification      357     361
config                    32      35
bad_input                  2       4
poll_deadline              3       0
failure rows           2,592   2,592   (of 8,814 job rows in both)
causes not investigated: failed_verification, other, poll_deadline, stall_cold_load  ->  none

The 3 rows left in other are generate_svg rows with no reason, no reason code and no door.

Tests, each run once: go build ./... clean, go vet ./internal/ledger/... clean, go test ./internal/ledger/... ok, go test ./internal/pipeline/ ok (698 s), the root Failure|Ledger|Changelog|Fragment|Leak|Docs subset ok after the golden file was regenerated (catalogue text only; no class count in the fixture moves).

Blast radius

How ledger --failures reads rows, old and new. No row on disk is rewritten. New rows of these shapes are stamped with the class at write time.

Risk

Low: read rules and catalogue text. Two rules are a choice, named in the fragment: the fetch-failed row could read as infrastructure, and compose check_failed as bad_input.

🤖 Generated with Claude Code

dmmdea and others added 2 commits November 12, 2026 12:29
… record the causes of the four uninvestigated ones

The failure register still held 30 rows in other and 3 in poll_deadline after the last release. Each is a
sentence or error class one harness writer owns, and each now goes to the class that names it: node_busy (a media
call naming a node that did not answer), node_unreachable (a dispatch that could not connect, repeated poll
failures, a fetch that dropped, a poll deadline that ends on a last poll error), config (no_node_configured, audio
QA without ffmpeg), bad_input (media convert without in or out, lint_errors), abstention (low classify
confidence) and failed_verification (compose check_failed). Each rule needs the writer's own words, and the two
that wrap other text (media convert, reasoning tier) are matched as whole sentences. The browse "blocked" rows stay
in other: their cause is not settled.

rootcause.go now holds the findings for failed_verification, poll_deadline, stall_cold_load and other, with their
counts and what each rests on; the part not settled is stated in the text. The pin test over core.ErrClass*
omitted no_node_configured, so it could not have caught the missing table entry; it lists it now. The page and the
fixture golden follow the catalogue.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…the failed_verification entry, and keep the misfiling comment from going stale

The browse door's "browse: blocked:" (the model or the loop reported no progress) is abstention: the loop ended with
no usable answer. It is a prefix rule on the door's own sentence and not an errClassCodes entry, because the word
means other things elsewhere; the prefix is reached since the error class is not in the table. other falls to the 3
rows with no reason, and abstention takes the 4 browse rows.

The failed_verification entry now says the research acceptance fix ships in the same release, with the replay
numbers, what still fails and that rows already written keep their class. The comment in failureClassOf points at
RefileStamped's own list instead of naming the pairs, so it cannot go stale. The page, the fragment and the fixture
golden follow the catalogue.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
@dmmdea
dmmdea merged commit c107505 into main Nov 12, 2026
6 checks passed
@dmmdea
dmmdea deleted the fix/register-rules-and-causes branch November 12, 2026 18:03
@dmmdea dmmdea mentioned this pull request Nov 12, 2026
dmmdea added a commit that referenced this pull request Nov 12, 2026
Two changes, each merged to main through its own pull request (#858, #859)
on green CI:

- The research lane stops failing digests that are right. A digest that
  quotes its page is excused from the page-word tripwire; the check derived
  from a caller's schema is "a value in any required field", not an item
  count on the first required array; an HTML page under 150 characters of
  text is a skipped source and is not dispatched. An offline replay of 30
  days of saved pages and answers: of 356 failed rows 167 now pass, 16 are
  refused as pages with no content and 173 still fail; of 4,622 passes
  4,513 still pass, 109 are refused as pages with no content, none fails.
- The failure register files the misread rows under the class that names
  them and records the cause of the four classes it had not investigated
  (failed_verification, other, poll_deadline, stall_cold_load). On the real
  ledger, 30 days: other 30 to 3, poll_deadline 3 to 0, with the totals
  unchanged; no class is left marked not investigated.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

Deletes: changes/unreleased/fix-register-rules-and-causes.md, changes/unreleased/fix-research-acceptance-false-rejections.md

Co-authored-by: dmmdea <dmmdea@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant