Skip to content

Re-file the issues no bee can take again - #4441

Merged
gHashTag merged 1 commit into
masterfrom
feat/refile
Sep 20, 2026
Merged

gHashTag merged 1 commit into
masterfrom
feat/refile

Conversation

@gHashTag

Copy link
Copy Markdown
Owner

Closes #4440

A dispatch row that spends its retry ceiling is released once, then stays
rejected on purpose — "a third identical failure is evidence about the issue,
not about the attempt"
. Right rule. Its consequence was never acted on: the
issue stays claimed
, so no bee can take it again however much the brief
improves.

101 open issues carried a queen-* branch, no pull request, no running bee
 92 of them sat behind a branch that was EMPTY or had NO MERGE BASE with master
    — orphan debris from before the container stopped cloning shallow
    — held since 2026-09-17, through two corrections to their own brief

Measured after running it over all 92

before after
board review 59 15
board backlog 553 590
next dispatch #4429, a re-filed issue

It found its own blind spot first

git diff A...B on histories with no merge base prints fatal: ... no merge base to stderr and exits non-zero. The helper returned stdout and stderr
together, so that sentence read as "this branch changed one file" — and 93
branches were skipped as work to publish. Same defect class as a feeder that
exits green when it cannot see the swarm.

What it refuses

A branch with real commits (that is work to publish), a branch newer than
the cutoff, an issue that already has a pull request, an issue a bee is running
right now — and everything, if the board does not answer.

Daily at 04:41 with a two-day cutoff, so nothing in flight is ever touched.
Both ledgers a new workflow moves are updated in this same commit.

🤖 Generated with Claude Code

A dispatch row that spends its retry ceiling is released once, after an hour,
and then stays rejected on purpose - "a third identical failure is evidence
about the issue, not about the attempt". The rule is right; its consequence was
never acted on. The ISSUE stays claimed, so no bee can take it again however
much the brief improves.

Measured: 101 open issues carried a queen-* branch, no pull request and no
running bee. 92 sat behind a branch that was empty or had NO MERGE BASE with
master - orphan debris from before the container stopped cloning shallow - held
since 2026-09-17, through two corrections to the brief they were written
against (#4296, #4302).

There is no external way to release a row, so this copies the work into a new
issue, which has no row, and closes the old with a pointer. Title, body and
`## Boundary` carry over byte for byte; the new issue also carries today's
toolbelt.

Measured after running it over all 92: the review column fell from 59 to 15,
the backlog rose from 553 to 590, and the swarm's next dispatch was #4429 - one
of the re-filed issues.

The tool found its own blind spot first: `git diff A...B` on histories with no
merge base prints `fatal: ... no merge base` to STDERR and exits non-zero, and
the helper returned stdout and stderr together - so that sentence read as "this
branch changed one file" and 93 branches were skipped as work to publish.

CENSUS: quiet 59 -> 60 workflow files; shell 59 -> 60 files, 80 -> 81 jobs,
272 -> 273 run-steps, 251 -> 252 run with `bash -eo pipefail`. Classified in
check_pr_branch_filters.py in the same commit.

Closes #4440

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@github-actions

Copy link
Copy Markdown
Contributor

📓 NotebookLM Notebook linked to this PR

This notebook contains session context, decisions, and artifacts for this work.

@github-actions

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-09-20 13:05:55 UTC

Summary

Status Count
Total Open PRs 30
PRs with Failing Checks 27
PRs with All Checks Green 3
READY 2
FAILING 27
PENDING 0
NO CHECKS YET 0

These columns do not partition: 2 + 27 + 0 + 0 = 29, and there are 30 open PRs. A PR is being counted twice or not at all.

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=403499176a5d != manifest seal=87e5cbd3ad94.
    The committed NMSE numbers were certified against an older compiler.rs.
    Run scripts/reseal-check.sh locally for the two-step reseal command (advisory; not a merge gate).

@gHashTag
gHashTag merged commit 1302e05 into master Sep 20, 2026
28 of 30 checks passed
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.

A spent dispatch row keeps its issue for good, and 92 issues were held behind one

2 participants