Skip to content

docs(handoff): rank 15 merged upstream as ZacxDev/cairn#7 β€” and NOT marked done, because merged is not live - #1419

Merged
ZacxDev merged 1 commit into
mainfrom
docs/handoff-rank15-merged-upstream
Sep 8, 2026
Merged

docs(handoff): rank 15 merged upstream as ZacxDev/cairn#7 β€” and NOT marked done, because merged is not live#1419
ZacxDev merged 1 commit into
mainfrom
docs/handoff-rank15-merged-upstream

Conversation

@ZacxDev

@ZacxDev ZacxDev commented Sep 8, 2026

Copy link
Copy Markdown
Member

docs(handoff): rank 15 fixed and MERGED upstream as ZacxDev/cairn#7 β€” and deliberately not marked done, because merged is not live

cairn recall's digest told readers to run --ref <name> / --limit N and the
client exited 2 with unrecognized arguments, so a reader following the output
it had just been shown fell back to the raw module.

The fix is not the flags. main() derived mode/limit/page AND enforced the
flag-conflict rules inline, while the client reaches the module as a LIBRARY and
never through main() β€” so offering the flags meant open-coding both at a second
site. recall_selection() and reject_recall_flags() are extracted and called
by both, so main() got shorter instead of the wrapper growing a copy. All four
flags wired, not the two the item named: fixing only --ref leaves the class
open, the same enumeration-vs-derivation shape rank 12 closed.

Evidence is recorded split rather than totalled. Two tests are RED at base on
their own assertions; a third fails at base with AttributeError and is NOT
counted as regression evidence. Four flags measured live against the real store,
because a passing suite and a working client are different claims.

A defect the change introduced is recorded rather than quietly fixed: exposing
--limit made recall()'s ValueError reachable from the command line for the
first time, and --limit 0 printed a traceback at rc 1. Guarded, watched before
and after, reachable by measurement.

Also recorded: the two failures the FULL suite found that a four-test subset did
not β€” the mutation battery refusing orphaned anchors after the guards moved, and
a TEXT ledger tripping on a comment quoting the callee's name. The second is a
false red, so the comment moved and the guard did not; that ledger cannot tell a
call from a comment, which is noted and not fixed.

πŸ”΄ The rank is NOT marked done. Its closing condition named a merged devrc PR and
the fix landed in the OSS client, so readlink -f ~/.local/bin/cairn still
resolves to devrc/scripts/cairn and --ref is still broken on this host. The
condition is re-pointed at "#7 merged AND #1406 landed AND the pin bumped past
059ec17", so it is checkable again rather than quietly unsatisfiable.

Rank 3 is untouched here: the session that owns it had already recorded #1406
across twelve mentions, and duplicating that would be the drift this doc keeps
paying for.


πŸ€– Generated with Claude Code

https://claude.ai/code/session_01MCjWicwrMAwjGTh3uXsSNT

… and deliberately not marked done, because merged is not live

`cairn recall`'s digest told readers to run `--ref <name>` / `--limit N` and the
client exited 2 with `unrecognized arguments`, so a reader following the output
it had just been shown fell back to the raw module.

The fix is not the flags. `main()` derived mode/limit/page AND enforced the
flag-conflict rules inline, while the client reaches the module as a LIBRARY and
never through `main()` β€” so offering the flags meant open-coding both at a second
site. `recall_selection()` and `reject_recall_flags()` are extracted and called
by both, so `main()` got shorter instead of the wrapper growing a copy. All four
flags wired, not the two the item named: fixing only `--ref` leaves the class
open, the same enumeration-vs-derivation shape rank 12 closed.

Evidence is recorded split rather than totalled. Two tests are RED at base on
their own assertions; a third fails at base with AttributeError and is NOT
counted as regression evidence. Four flags measured live against the real store,
because a passing suite and a working client are different claims.

A defect the change introduced is recorded rather than quietly fixed: exposing
`--limit` made `recall()`'s ValueError reachable from the command line for the
first time, and `--limit 0` printed a traceback at rc 1. Guarded, watched before
and after, reachable by measurement.

Also recorded: the two failures the FULL suite found that a four-test subset did
not β€” the mutation battery refusing orphaned anchors after the guards moved, and
a TEXT ledger tripping on a comment quoting the callee's name. The second is a
false red, so the comment moved and the guard did not; that ledger cannot tell a
call from a comment, which is noted and not fixed.

πŸ”΄ The rank is NOT marked done. Its closing condition named a merged devrc PR and
the fix landed in the OSS client, so `readlink -f ~/.local/bin/cairn` still
resolves to devrc/scripts/cairn and `--ref` is still broken on this host. The
condition is re-pointed at "#7 merged AND #1406 landed AND the pin bumped past
059ec17", so it is checkable again rather than quietly unsatisfiable.

Rank 3 is untouched here: the session that owns it had already recorded #1406
across twelve mentions, and duplicating that would be the drift this doc keeps
paying for.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MCjWicwrMAwjGTh3uXsSNT
Claude-Session-Id: 054db069-07b1-4abb-962a-f677e633829f
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