add-row's archive scan treats an unparseable archive row as a non-row, so a retired position can be reissued — and the obvious refusal would fire on the archive's own header (#93 round 6, filed rather than taken) #122

Copilot's round 6 on #93 (review of 4bc90d2, 2026-09-16T13:22Z), filed rather
than taken under Brett Heap's ruling of 2026-09-16: Copilot is capped at two
rounds per pull request, and #93 is at six. The thread is replied to with this
number and resolved on the pull request.

The finding, in Copilot's words:

This parser silently skips any archive table line that has no backticked lane
token, so a readable but malformed LANES-retired.md can make a retired
lane/position disappear from ar_arch; the later add-row then treats the
position as free and reissues it. Because this check is the guard for the
permanent-reservation rule, malformed archive rows should cause a refusal
(and the scan status must not be collapsed to success) rather than being
treated as non-rows.

lanes-edit.sh:8852-8859, inside add-row's archive scan:

substr($0,1,1) == "|" {
  p1 = index($0, "`"); if (p1 == 0) next
  r = substr($0, p1 + 1); p2 = index(r, "`"); if (p2 == 0) next

What is right about it

The consistency argument is the strongest part and it is this repository's
own: archive_rows already refuses — for the whole repository, before a byte is
written — a row of that repository whose | count hides which text is the
state cell, because "this act moves EVERY retired row of a repository or none
of them" (Copilot round 1 on #93, taken in 5c49740). A reader whose answer
decides whether a position is reissued is at least as load-bearing as the writer
that moved the row there, and it currently treats an unparseable row as a
non-row. Amendment 19(b) — a retired position is NEVER reissued — is the
guarantee on the other side of that next.

The || : half is real and separate. The trailing || : on the pipeline
collapses an awk that FAILED into "no archived row matched". The read of the
archive TEXT is already fail-closed one screen up (ar_arch_rc, the die at
lanes-edit.sh:8838), so this is the one remaining status on that path that is
thrown away, and it is Amendment 7(d) either way: a scan that failed is not
"it is not retired".

What the fix must not do, and Copilot's wording would

The archive's own header lines are |-leading and carry no backtick.
archive_rows writes them itself (lanes-edit.sh:8508-8509):

| lane | session id | workstation / env / user | started (UTC) | objects owned | handoff path | state |
|---|---|---|---|---|---|---|

So "any archive table line that has no backticked lane token" is, on every
archive this act has ever written, two lines out of every file — and a
refusal written to that description would make add-row refuse against a
perfectly good archive, on the path that creates every lane. That is worse than
the defect: it is the same guarantee failing closed on the whole estate instead
of open on a hand-mangled row.

The refusal has to be able to tell a header from a mangled row. The shape that
does is the one archive_rows already uses — row_split_state_cell /
row_sep_count, which count | separators and know what a seven-column row
is — so a line with a data row's separator count and no backticked token is the
refusal, and the header and the |---| rule are not. Reusing that helper rather
than writing a second rule in awk is also what keeps the reader and the writer
from disagreeing about what a row is, which is the defect underneath every one
of this pull request's rounds.

Reachability

Nothing this toolset ships writes such a line: archive_rows copies rows
verbatim out of the register with sed -n "${ar_num}p", and it refuses the
whole move on a row of that repository it cannot take apart. The archive is a
hand-editable file, though, and Rule 9 invites people to read it — so the state
is a hand edit or a bad merge, which is exactly the state the sibling refusals
in this amendment are written for.

Filed by lane openRepoTools-3. Round 5's three findings are #120.